text
stringlengths 10
379k
|
|---|
The dissociation of ribosomes into subunits is an indispensable event in translation. During the resting phase or under stress conditions, almost all of the ribosomes exist in the 80S form (). However, when conditions favor growth, and the protein synthesis must ramp up again, the 80S ribosomes rapidly dissociate into subunits to enter the cycle of translation (,).
In bacteria, dissociation of free 70S ribosomes is catalyzed by elongation factor G (EF-G) and ribosome recycling factor (RRF), powered by GTP hydrolysis. This dissociation of 70S ribosomes by EF-G·GTP/RRF is transient and requires stabilization by initiation factor 3 (IF3) to keep the subunits apart (). IF3 binds to the 30S subunit and prevents its re-association with the 50S subunit (). IF3 by itself can split 70S ribosomes. However, this dissociation is extremely slow and less extensive in comparison with EF-G·GTP/RRF-induced ribosome splitting ().
In eukaryotes, several translation initiation factors, such as eIF1A, eIF1 and eIF3, have been shown to possess ribosome anti-association/dissociation activity similar to that of bacterial IF3 (). Furthermore, translation factor eIF6 has a strong ribosome anti-association activity (,). This factor participates in the biogenesis of the 60S subunit (,) and remains associated with it until the late event of the initiation step (). However, the process of active dissociation of the 80S ribosome into its subunits, analogous to the EF-G·GTP/RRF-dependent splitting of 70S ribosomes, has not been described. Importantly, the homolog of RRF in eukaryotes appears to be localized in organelles, and is not found in the cytoplasm (,).
The eukaryotic elongation factor 2 (eEF2), like its bacterial homolog EF-G, catalyzes translocation of tRNA and mRNA in the presence of GTP (). eEF2 is larger than EF-G () and, in contrast to EF-G, has a binding site specific for adenine nucleotides in addition to the binding site for GTP/GDP. Thus, it has been shown that eEF2 of yeast () and mammalian () cells can bind ATP or ADP in the absence of ribosomes and in the presence of GTP. The functional importance of the ATP/ADP-binding site of eEF2 remained unknown so far.
In this report, we describe a novel activity of eEF2, the dissociation of 80S ribosomes into subunits in the presence of ATP. This splitting is transient and dependent on the energy released by hydrolysis of ATP. The resulting subunits can then be stabilized either by adding the ribosome anti-association factor or by treating the ribosomes/eEF2/ATP reaction mixture with glutaraldehyde. GTP inhibits the eEF2/ATP-dependent dissociation of 80S ribosomes. We propose that eEF2/ATP-dependent ribosome splitting is involved in catalysis of 80S ribosome dissociation during the shift up at the onset of favorable physiological conditions.
Buffers were prepared from reagent grade chemicals: buffer 5/100 (20 mM HEPES-KOH, pH 7.6, 5 mM MgCl, 100 mM KCl, 2 mM DTT); buffer 5/500 (buffer 5/100 with 500 mM KCl); buffer 2/150 (buffer 5/100 with 2 mM MgCl and 150 mM KCl); buffer 10/150 (buffer 5/100 with 10 mM MgCl and 150 mM KCl).
Yeast ribosomes, eEF2 and eIF6 were prepared as described in the Supplementary Data.
Prior to use, frozen ribosomes were incubated in buffer 2/150 for 5 min at 30°C, then sedimented at 4500 × for 4 min at room temperature. Ribosomes were treated with or without eEF2, eIF6 and nucleotides in 60 µl of buffer 2/150, 5/100 or 5/500 as specified in the figure legends (or tables). Then the reaction mixtures were loaded on a 5–30% (w/v) sucrose density gradient prepared in the same buffers, and were sedimented for 2 h at 4°C (Beckman rotor SW50.1, 46500 r.p.m.). In C and D, after incubation, the reaction mixtures were treated for 2 min on ice with a cold glutaraldehyde solution (final concentration 0.45%, v/v) prepared in 100 mM Tris-HCl (pH 7.6), 2 mM MgCl and 150 mM KCl. In F, ribosomes in buffer 2/150 were treated with glutaraldehyde under identical conditions except that the final concentration of glutaraldehyde was 1.5% (v/v) and fixation was performed in the presence of 0.3 mg/ml bovine serum albumin. The sedimentation behavior of ribosomes was monitored using an ISCO UA-6 spectrophotometer at 254 nm. The percentage of 80S ribosomes was calculated from the areas corresponding to 40S, 60S and 80S ribosomes measured by ImageJ 1.31j software (Bethesda, USA).
The light scattering experiments were performed at room temperature with a spectrofluorometer (Photon Technology International, incoming slits: 1 mm × 0.1 mm, outgoing slits: 0.3 mm × 1 mm; wavelength: 436 nm, angle: 90°). ‘Factor mixture’ (180 μl) containing eEF2, eIF6, ATP and/or ADP was mixed manually with 20 μl of 0.5 μM 80S ribosomes or 0.25 μM of 40S or 60S subunits. In each case, the final concentrations of MgCl and KCl were adjusted to 2 and 150 mM, respectively. The resultant mixture (200 μl) was placed in a cuvette immediately and the intensity of the light scattering (counts per second, CPS) was continuously recorded without stirring, beginning at 20 s after the mixing. The light scattering was calculated as the percentage of the initial value (×10) at 20 s is expressed as %CPS. Apparent rate constants of ribosome splitting were obtained using Kyplot software (Tokyo, Japan) by fitting data to a double exponential equation.
Ribosomes (0.05 µM) were incubated at 30°C in 80 µl of buffer 2/150 with 1 µM eEF2 alone or eEF2 with 0.5 mM ATP, GTP or a nonhydrolyzable analog of ATP, 5′-adenylyl-[β, γ-imido]diphosphate (ADPNP). After 15 min of incubation, the reaction mixtures were treated with 0.45% glutaraldehyde (see ‘Sedimentation profiles’ section). Gradients were fractionated from the bottom (10 drops per fraction) and the protein content in each fraction was precipitated by the addition of 100% TCA to a final concentration of 10%. Protein pellets were washed with a mixture of ether and ethanol (1:1) and resolved on 10% SDS-PAGE. eEF2 was detected by western blotting using rabbit antibodies against yeast eEF2 (dilution 1:20 000). The intensities of the bands corresponding to the antibody bound to eEF2 were determined using ImageJ 1.31j software and amounts of eEF2 were estimated using standard eEF2.
The release of inorganic phosphate from [γ-P]ATP upon hydrolysis was analyzed as described (). A standard reaction mixture (15 µl) contained 0.05 µM ribosomes, 2.5 µM eEF2 and 0.5 mM [γ-P]ATP (specific activity 0.12 µCi/nmol). Incubation was at 30°C for 1, 2, 4, 6 and 10 min.
Light scattering of ribosomes decreases upon the dissociation into subunits because the 80S ribosome scatters more light than its subunits due to the larger size (,). As shown in A, incubation of ribosomes with eEF2 and ATP caused a decrease in the scattered light (lower curve), suggesting that eEF2/ATP promoted splitting of the 80S ribosomes into subunits. In contrast, when ribosomes were incubated with eEF2 or ATP alone or eEF2 and ADP, the change in the light scattering was negligible during the incubation.
Since the light scattering change of ribosomes is an indirect method of assessing ribosome splitting and could occur for a number of reasons, further experiments were required before a definite conclusion could be reached. As shown in Figure S1 (Supplementary Data), addition of eEF2 to ribosomes alone, or together with ATP, caused an instantaneous increase in the light scattering. A similar increase in light scattering of and mammalian ribosomes upon addition of eIF3 was reported () and used for studying the binding of the factor to ribosomes (). B shows the binding of eEF2 to the ribosomes measured by the light scattering increase. As can be seen, ATP stimulated the binding of eEF2 to 80S ribosomes or 60S subunits. The binding of eEF2 to 40S subunits with or without ATP was not detected (data not shown). The difference of the affinity of eEF2 to the ribosomes due to ATP, however, does not influence the light scattering results shown in A, because 99% of ATP still remained at 10 min after the onset of the reaction (see B). Therefore, it was concluded that the decrease in light scattering observed in A indicates dissociation of 80S ribosomes into subunits.
For determination of the extent of the ribosome splitting by eEF2/ATP, the light scattered by completely dissociated 80S ribosomes must be compared with that by ribosomes in the buffer containing 2 mM MgCl and 150 mM KCl (2/150), in which the splitting reaction took place (A). Such comparison was performed as described in Figure S1 (Supplementary Data). These results indicated that 92% of the 80S ribosomes were converted into subunits by 2.5 µM eEF2 and ATP. It should be noted that the 80S preparation used had extra 60S subunits (see E) because of the procedure of the 80S ribosomes purification. The presence of the excess of 60S subunits inhibits dissociation of 80S ribosomes (see Supplementary Data, Figure S3A). However, even under these conditions, 2.5 µM eEF2 almost completely converted the existing 80S ribosomes into subunits.
In the experiment described in C, various amounts of eEF2 were added and the rate and extent of the ribosome splitting were determined. From these data, the Michaelis-Menten constant () value was 0.24 ± 0.10 µM. This matches the estimated dissociation constant () value of eEF2 to 80S ribosomes in the presence of ATP (B), which was 0.26 ± 0.04 µM.
To confirm the eEF2/ATP-dependent splitting of 80S ribosomes deduced from the measurements of the light scattering change, we compared the sedimentation patterns of ribosomes incubated alone and with eEF2/ATP (A and B). In the presence of eEF2/ATP, no significant dissociation of 80S ribosomes could be detected. These results are reminiscent of the finding that the transient splitting of bacterial 70S ribosomes by EF-G·GTP/RRF cannot be observed by the SDGC analysis (). Therefore, these results suggest that the dissociation of 80S ribosomes by eEF2/ATP detected by the light scattering decreases (A) is transient and that the separated subunits appear to re-associate during the SDGC analysis.
Even if the splitting of 80S ribosomes by eEF2/ATP is transient, it should still be detectable by SDGC upon stabilization of the dissociated subunits. In the experiment shown in D, the reaction mixture containing 80S ribosomes, 1 µM eEF2 and ATP was treated with 0.45% glutaraldehyde. This treatment is known to fix eEF2 on the 80S ribosome () without influencing the ribosomal sedimentation pattern. C is a control, where the ribosomes were treated with glutaraldehyde without eEF2/ATP. It is clear that upon incubation of ribosomes with eEF2/ATP, the apparent amount of 80S ribosomes decreased from 36 to 18% of the total ribosomes. This is equal to a 50% conversion of 80S ribosomes into subunits. The amounts of 80S ribosomes, 36 and 18% of the total ribosomes, actually correspond to 60 and 30% because of the induced dissociation of the 80S ribosomes during the ultracentrifugation analysis under the buffer conditions used (see the next section ‘Determination of …’). Therefore, the results shown in D indicate that eEF2/ATP-dependent splitting of 80S ribosomes is indeed transient and can be observed by SDGC only following stabilization of eEF2 on the subunits.
It should be noted that the results obtained by the SDGC analysis correlate well with the light scattering data. Thus, when ribosomes were incubated with 1 µM eEF2 and ATP, the decrease of the light scattering was ∼50% of the signal recorded at 2.5 µM eEF2 (C), which corresponds to almost complete dissociation of the 80S ribosomes (see Supplementary Data, Figure S1).
The experiments described in the preceding sections were carried out in buffer 2/150. These ionic conditions are close to the optimum for protein synthesis (). To determine the amount of 80S ribosomes split under conditions 2/150 (A), it was necessary to estimate the exact quantity of 80S ribosomes present in this buffer. As noted in A, under these conditions only 37% of the total ribosomes were detected as the 80S form by SDGC. This value does not represent the actual amount of 80S ribosomes present in buffer 2/150 because of the well-known effect of ultracentrifugation—dissociating ribosomes into subunits ().
In E–G, the ribosomes were analyzed with SDGC under three ionic conditions 5/100, 2/150 and 5/500. In buffers 5/100 (tightly associated 80S ribosomes; E) and 5/500 (complete dissociation into subunits; G), analyses were performed without glutaraldehyde fixation. To avoid induced ribosome dissociation during centrifugation in buffer 2/150, the ribosomes in this buffer were fixed with 1.5% glutaraldehyde as described previously (). The result showed that 57% of the ribosomes existed in the 80S form in buffer 2/150 (F). It is noted that the fixation with this high concentration of glutaraldehyde cannot be used for detection of the eEF2/ATP activity because treatment of the ribosomes/eEF2 mixture with 1.5% glutaraldehyde produces artifacts presumably due to fixation of eEF2 to 80S ribosomes (data not shown).
The determination of the amount of 80S ribosomes in buffer 2/150 by two other methods, light scattering analysis and gel-filtration, revealed similar values (see Supplementary Data, Figure S2). It is worth mentioning here, that the obtained value of 60% of 80S ribosomes correlates well with the data of light scattering by ribosomes (). In those experiments, it was shown that at 2 mM Mg, 65% of 80S ribosomes exist in equilibrium with ribosomal subunits. Therefore, under the experimental conditions used in , the starting concentration of 80S ribosomes was estimated at 0.03 µM. Using this value and the calculated maximum velocity of the splitting reaction (C), the turnover number () of the eEF2/ATP-dependent dissociation was estimated as 0.72 ± 0.10 min.
In the experiment described in A, after the eEF2/ATP-dependent splitting of 80S ribosomes, increasing amounts of anti-association factor eIF6, which binds to 60S subunits (,), were added followed by the SDGC analysis. As can be seen, stable subunits were formed in an eIF6-dose dependent manner to the maximum of 80% conversion of the existing 80S ribosomes into subunits. This is somewhat lower than what we observed with the light scattering analysis, possibly, due to the partial release of the bound eIF6 during the ultracentrifugation. The control experiments where eEF2 or eEF2/ATP was omitted revealed noticeably less dissociation, indicating that eEF2 increases the extent of the 80S ribosome dissociation into subunits in the presence of anti-association factor eIF6.
The effect of eIF6 on the eEF2/ATP-dependent dissociation of 80S ribosomes was confirmed by light scattering measurements of ribosomes as described in B. This figure shows that when eIF6 was added, the initial rate and the final value of the eEF2/ATP-dependent dissociation of 80S ribosomes were increased. We should emphasize that the rate of dissociation by eIF6/ATP alone is much slower than that by eEF2/ATP. These results indicate that the dissociation is performed by eEF2/ATP, which plays the major role in facilitating the splitting of 80S ribosomes, but not by eIF6, which stabilizes dissociated subunits.
In A, ATP was substituted by its non-hydrolyzable analog ADPNP or ADP and the reaction mixtures were analyzed by SDGC as described for A. The results showed that incubation of ribosomes with eEF2/ADPNP or eEF2/ADP did not lead to the splitting of 80S ribosomes. In the presence of other nucleotides, the splitting did not occur either (A, see results for GTP or GDP) or was much less than with ATP (A, see results for UTP or CTP). This indicates that the dissociation of 80S ribosomes by eEF2 is an energy-dependent process specifically involving ATP.
The fact that the non-hydrolyzable ATP analog was inactive in the splitting of 80S ribosomes strongly suggested that ribosome dissociation is accompanied by the hydrolysis of ATP. This is shown in B, where [γ-P]ATP was converted to inorganic phosphate (P) during the dissociation reaction (open circles). When incubated separately, eEF2 or ribosomes alone hydrolyzed much less ATP (open triangles and closed squares, respectively). As can be noted, the ribosome/eEF2-dependent hydrolysis of ATP in buffer 2/150 almost leveled off at 2 min (B, dashed line). Meanwhile, the splitting reaction went on up to 10 min (see A). The mechanism of such slow down of the ATPase reaction is not clear. Since the molar concentration of ATP is far higher than that of ribosomes in the ATPase reaction mixture, the background level of ATP hydrolysis must be sufficient for the ribosome splitting. It is possible that the major activity of ATP hydrolysis takes place upon binding of eEF2 to 80S ribosomes. However, after most ribosomes are split, the hydrolysis subsides.
When 80S ribosomes and eEF2 were incubated in the buffer with 10 mM MgCl (i.e. under the conditions where the splitting of 80S ribosomes by eEF2/ATP does not occur; data not shown and see Supplementary Data, Figure S3C), hydrolysis of ATP was considerably higher than in buffer 2/150 (B, closed circles). This indicates that ATP hydrolysis is not always coupled with the splitting reaction. Importantly, sordarin, which is inhibitory to the dissociation of 80S ribosomes (), stimulated the ATPase reaction significantly (open squares) indicating again that the ATPase activity is not entirely coupled with the ribosome splitting.
When various concentrations of ATP were added to ribosomes and eEF2 under the dissociation conditions, the rate of the P release increased at higher concentrations of ATP (C). From these data, the turnover number of ribosome/eEF2 ATPase was estimated as 40.2 ± 4 min, which is significantly higher than the value described above of the splitting reaction catalyzed by eEF2/ATP (0.72 ± 0.1 min).
To better understand the mechanism of the 80S ribosome dissociation by eEF2/ATP, the binding of eEF2 to subunits during the splitting reaction was analyzed. The ribosomes were incubated under the same conditions as shown in D (with 0.45% glutaraldehyde fixation), where 50% of the existing 80S ribosomes were converted to subunits (D). This condition was specifically chosen to examine the presence of eEF2 on the remaining 80S ribosomes in the presence of ATP. In parallel, the ribosomes were incubated with eEF2 alone or with GTP or ADPNP. The results showed (D) that in the presence of ATP, the binding of eEF2 to the 80S ribosomes is significantly less than that to the subunits. In contrast, GTP strongly stimulated the binding of eEF2 to the 80S ribosome. In the presence of ADPNP, a small amount of eEF2 was localized only on the 80S ribosomes suggesting that the binding of eEF2 to subunits observed in the presence of ATP is related to the splitting reaction. It should be noted that with ATP, eEF2 was found not only on 60S subunits but also on 40S subunits (D, diamonds). On the other hand, we did not observe the binding of eEF2 to isolated 40S subunits by the light scattering technique (data not shown). Hence, further studies will be necessary to assess the significance of the binding of eEF2 to 40S subunits under the splitting conditions.
As pointed out in the preceding section, GTP stimulated binding of eEF2 more to the 80S ribosomes than to the subunits (D, squares). On the contrary, ATP stimulated the eEF2 binding mostly to the subunits. These results suggest that the effect of eEF2/GTP on the ribosomes is different from that of eEF2/ATP. Therefore, it was important to examine the effect of GTP on the splitting and the ATPase activity. The results are shown in . With increasing concentrations of GTP, the eEF2/ATP-dependent ribosome splitting as well as the ATPase activity was progressively inhibited. The inhibition of the ATPase activity was more efficient than that of the dissociation reaction.
Various agents known to inhibit the conventional eEF2 activity were tested on the dissociation reaction catalyzed by eEF2/ATP (). Cycloheximide and sordarin at 0.2 mM inhibited more than 60%, while fusidic acid at the same concentration inhibited only 11%. In the presence of the aminoglycoside paromomycin, which prevents EF-G·GTP/RRF-dependent splitting of 70S ribosomes (), the dissociation of 80S ribosomes by eEF2/ATP was completely prevented and the amount of 80S ribosomes increased 40% over the initial amount of 80S ribosomes.
Since the 80S ribosome is in equilibrium with its subunits, conditions and factors that influence this equilibrium should have an effect on the eEF2/ATP-dependent splitting (see Supplementary Data). Polyamines, which stimulate association of the subunits, are expected to inhibit the dissociation of the 80S ribosomes by eEF2/ATP. However, as it is shown in , in the presence of spermidine at a concentration higher than the concentration of this polyamine (), the eEF2/ATP-dependent splitting still occurs. Furthermore, putrescine at a concentration stimulatory for protein synthesis (), had very little effect on dissociation.
Because magnesium ions have a similar effect as polyamines, shifting the equilibrium towards the 80S ribosomes, the increase in Mg led to the gradual inhibition of ribosome splitting by eEF2/ATP (Supplementary Data, Figure S3C). Thus, at 4 mM Mg only 10% of the existing 80S ribosomes were split. However, it should be pointed out that at this Mg concentration, protein synthesis is very inefficient (). As can be noted from Figure S3B, in the presence of higher concentrations of ribosomes, less dissociation was observed. At the same time, when the eEF2 concentration was increased, the splitting of ribosomes by eEF2/ATP was significantly restored even at a high ribosome concentration (Figure S3B).
There are physiological and genetic data consistent with the present finding that eEF2 may play an essential role in the dissociation of free 80S ribosomes in cells. Under stress conditions (,) or during resting phase (), the ribosome pool is primarily preserved in the 80S form. Nutrient depletion stress dramatically increases free 80S ribosomes, and decreases the ATP concentration (). In a separate experiment, the depletion of ATP elevated phosphorylation of eEF2 (), which inactivates the factor (). Similarly, the expression of a non-functional form of eEF2 (mutations H699K or V488A) led to an increase in 80S monomers (). All of these data point to the fact that inactivation of eEF2 with simultaneous decrease of ATP results in the accumulation of the ribosomes in the 80S form. Conversely, when the conditions are favorable for cell growth and there is enough ATP, eEF2 is activated, resulting in the rapid conversion of the accumulated 80S ribosomes to polysomes (,).
The translation factors eIF3, eIF1A and eIF6 have been reported to dissociate vacant 80S ribosomes into their subunits (,,). We propose that eEF2, but not the these factors, plays the active catalytic role in the splitting of 80S ribosomes for the following six reasons. First, eIF3 is an anti-association factor that exerts its activity from the solvent side of the 40S subunit, preventing subunits joining (,). In contrast, eEF2 translocates tRNA and mRNA through the inter-subunit space of the 80S ribosome (). The binding site of eEF2 is located mostly in the inter-subunit space of the 80S ribosome with domain IV contacting the broadest inter-subunit bridge B2a (). This makes eEF2 more suitable than eIF3 to get into the inter-subunits space to split the ribosome. Second, the cellular content of a core subunit of eIF3 (eIF3g) is less than the ribosome concentration (). There are only 2400 molecules of eIF3i, which is also a core subunit of eIF3, per yeast cell, but 78 100 molecules of eEF2 per cell (). Third, the rate of the binding of eIF3 to the 40S subunit is 12-times faster than that to the 80S ribosome () suggesting that the main target of eIF3 is the 40S subunit, not the 80S ribosome. Fourth, recent evidence indicates that eIF3 splits 80S ribosomes into subunits only in the presence of polyuridylic acid (). Fifth, the dissociation activity of eIF1A is weaker than that of eIF3 (). Finally, the speed of the 80S ribosome dissociation by eIF6 is incomparably slower than that by eEF2/ATP (B).
The kinetic parameters of the eEF2/ATP-dependent dissociation are consistent with the proposed physiological function. ATP stimulates binding of eEF2 to 80S ribosomes (). We estimated that of eEF2/ATP for the 80S ribosome is 0.2 µM, which is comparable with of eEF2 for the 80S ribosome in the post-translocation state in the presence of GDP (0.4 µM) or a non-hydrolyzable analog of GTP (0.3 µM) (). The turnover number () of eEF2/ATP for the 80S ribosome splitting is ∼0.7 min. This is comparable with the turnover number of the peptide bond formation in the yeast system (2 min) (). Due to the re-association of the split subunits into 80S ribosomes under our dissociation condition, the calculated value is much smaller than the actual turnover number of the splitting where the formed subunits are rapidly ‘removed’ by the initiation step. In addition, our value of the ribosome/eEF2 ATPase activity (40 min) is 4 times faster than of the GTPase activity of the ribosome/eEF2 complex (9.6 min) (). Furthermore, in yeast, all ribosomes accumulate in the 80S form under conditions of glucose depletion. However, they are completely converted to polysomes within 10 min after a shift up of the culture conditions to normal glucose levels (). This can be explained on the basis of the known cellular content of ribosomes () and eEF2, with the value we estimated for the eEF2/ATP-dependent 80S ribosome splitting.
One may argue that the spontaneous dissociation of 80S ribosomes into subunits under the optimum ionic conditions can be sufficient for the splitting, especially under the conditions where the formed subunits are efficiently ‘utilized’ for the initiation step. However, the experiment shown in B demonstrated that combining the spontaneously produced 60S subunits with eIF6, which prevents ribosome re-association, is not sufficient for rapid completion of 80S ribosome dissociation. The rate of the splitting by eIF6 is extremely slow compared with that of the splitting by eEF2/ATP.
The possibility that eEF2/ATP-dependent splitting produces inactive subunits is highly unlikely. First, eEF2 and ATP are present ; it is difficult to imagine that these natural components might harm the ribosomes. Second, we have shown that the 60S subunits formed by eEF2/ATP are capable of binding eIF6 (A). Third, the split subunits can re-associate to form 80S ribosomes (A and B). However, despite these facts and the evidences cited above, the idea that the eEF2/ATP-dependent splitting of the 80S ribosomes is a part of the utilization process of the dormant ribosomes should be tested by further experiments indicating that the formed subunits can engage in the initiation and elongation phases. Until this is accomplished, our proposal will remain a hypothesis.
The dissociation of 70S ribosomes of prokaryotes is catalyzed by EF-G, GTP and RRF (,,). It is therefore reasonable to expect that eEF2, the eukaryotic homolog of EF-G, can dissociate 80S ribosomes. In a similar manner to prokaryotic IF3, eukaryotic factors (eIF3, eIF1 (), eIF1A or eIF6) can stabilize the ribosome subunits once they are separated by eEF2/ATP. In the experiment described in , we used eIF6 (a strong ribosome anti-association factor ()) as a convenient tool to keep apart the subunits formed by the eEF2/ATP-dependent reaction. Which factors play this role is a subject of a separate study.
Since the EF-G·GTP/RRF-dependent splitting of 70S ribosomes is a part of the recycling step of the protein synthesis in bacteria, it is quite possible that the eEF2/ATP-dependent dissociation of 80S ribosomes is a part of the eukaryotic ribosome-recycling step. Our preliminary data obtained on yeast model post-termination complexes [puromycin-treated polysomes ()] strongly support this possibility. However, we should emphasize that in this article we do not intend to implicate the eEF2/ATP-dependent splitting of 80S ribosomes as a part of the eukaryotic recycling system until further solid evidence for this idea becomes available.
In our earlier studies of the dissociation of bacterial ribosomes, we showed that the splitting of vacant 70S ribosomes is dependent on GTP hydrolysis (). In analogy with this process, in the present article, it is demonstrated that the dissociation of the eukaryotic ribosome is dependent on the energy released upon hydrolysis of ATP. The specific involvement of ATP was shown by the inability of GTP, CTP and UTP to split 80S ribosomes (A) and the ATPase activity of the ribosome/eEF2 complex. Sordarin stimulated the ATPase activity (B), while it inhibited the dissociation (). The stimulatory effect of sordarin on the ATPase is of interest in view of the finding that this antibiotic apparently competes with ATP (or ADP) for binding to eEF2 in the solution (). Our results indicate that sordarin does not inhibit ATP binding to the complex of the 80S ribosome and eEF2. Cryo-EM studies of the stalled complex of yeast 80S ribosome/eEF2/sordarin suggested that sordarin restricts conformational changes of eEF2 by blocking rotation of domains III, IV and V of the factor (). It is possible that similar conformational changes are the key movements for the splitting activity of eEF2 in the presence of ATP, which would explain the inhibitory effect of sordarin on ribosome dissociation. The seemingly contradictory effects of sordarin (i.e. inhibition of the splitting and stimulation of the ATPase activity) are reminiscent of the similar effect of sordarin on the GTPase of the post-translocation form of 80S ribosome/eEF2 complex and the translocation ().
We showed that GTP inhibited the ATP-dependent splitting and the ATPase activity (). How then can eEF2/ATP mobilize 80S ribosomes in the presence of GTP? We assume that most of the free ribosomes are in the post-translocation state. The affinity of eEF2/ATP and that of eEF2/GTP to these ribosomes are approximately the same [B and ()]. During shift up conditions, the concentration of ATP is ∼6-fold higher than that of GTP (). The subunits produced by eEF2/ATP-dependent dissociation are rapidly utilized for the initiation step, driving the splitting reaction to the right by depleting the products. These considerations led us to suggest that the mobilization of the vacant 80S ribosomes by eEF2/ATP can occur .
There are a number of differences in the effect of ATP and GTP on eEF2 action on the ribosome. First, fusidic acid facilitates ribosome·eEF2/GTP complex formation and keeps eEF2 bound to the ribosome (), thereby stimulating subunits association activity by eEF2 (unpublished data). In contrast, the effect of fusidic acid on the eEF2/ATP-dependent splitting is nominal (). Second, the binding site for ATP on eEF2 is different from that for GTP (,). Depending on the binding site, nucleoside triphosphates exert different structural conformational changes on eEF2. In support of this hypothesis, GTP inhibited eEF2/ATP-dependent dissociation, as well as ATPase activity (). Third, GTP hydrolysis by ribosome/eEF2 is assumed to take place at the binding site of GTP on eEF2 (), but ribosomes may play a larger role in the hydrolysis of ATP during the dissociation by eEF2. It is known that the 80S ribosome possesses intrinsic ATPase activity (,), which is higher than the intrinsic GTPase activity. This ATPase is presumably associated with the 5S RNP complex and is stimulated by eEF2 (). The central protuberance, which involves the 5S RNP complex and forms B1a and B1b/c inter-subunit bridges (), undergoes substantial conformational changes upon binding of eEF2/sordarin (). These data suggest that eEF2/ATP-dependent splitting of the 80S ribosome might be triggered by the conformational change of the central protuberance of the 60S subunit.
It is known that ATP is required for the initiation (,) and elongation () steps of polypeptide synthesis. In this study, we show that ATP is also consumed for splitting of 80S ribosomes into subunits, which may be important for utilization of 80S ribosomes for the initiation step. Thus, in contrast to prokaryotic protein synthesis, ATP plays critical roles in eukaryotic translation.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Viral infection often entails the hindering of the host cell's translation machinery in order that the viral genome is expressed more efficiently relative to the expression of the host's proteins. In the family of single-stranded RNA viruses, a viral protease cleaves the scaffolding translation initiation factors of the eIF4G family, thereby reducing the efficiency of the host cell's cap-dependent translation initiation and favouring the alternate internal ribosome entry site (IRES) mechanism used by the virus (). The importance of this mechanism for viruses has been best demonstrated in the attenuated Sabin poliovirus strains used for worldwide polio vaccination, which contains point mutations in the IRES () that allow efficient translation in the gut but not in neuronal cells.
Cap-dependent translation initiation is also rendered less efficient under several cellular conditions, but specific cellular mRNAs are still translated with relative efficiency using the IRES mechanism for translation initiation. Many fine reviews have been written on IRES in cellular messages (), viruses (,), stress and apoptosis (). Typical translation of cellular messages in eukaryotes begins with the association of translation initiation factors with the cap-binding protein factor, eIF4E, on the 5′end of the message where the ‘cap’, a methylGDP nucleotide resides. This complex, which includes eIF4E, eIF4G, eIF4A, eIF3, the 40S ribosomal small subunit and an activated start codon tRNA, will scan along the untranslated region (UTR) of the mRNA until it finds a suitable start codon where the large ribosomal subunit will join and protein translation will begin. During mitosis, cellular perturbation or stress, and apoptosis, canonical initiation factors like eIF4E, 4E-BPs, eIF2α and the eIF4G family of proteins are either modified by changes in phosphorylation state or by protein cleavage () and are no longer available for efficient cap-dependent translation initiation. The ribosome may also undergo some modifications as well (). At these times, other protein factors (), many of which are part of the ribonucleoprotein complex, are required to enhance translation initiation through the IRES mechanism. An IRES is a stretch of sequence usually upstream of the AUG start codon in the 5′ untranslated region (UTR) of the messenger RNA that along with the IRES -acting protein factors (ITAFs), recruit the ribosome. It is not clear whether these factors need a sequence motif or a RNA secondary structure/sequence motif combination to bind to IRES-containing mRNAs. In viral IRESes, like HCV and EMCV, RNA secondary structure has been shown to be crucial for IRES function (). These structures are also conserved in other viruses that do not share primary sequence similarity. For example, the HCV IRES structure is similar to CSFV and BVDV (,), whereas the EMCV IRES structure is shared with other cardio- picornaviruses ().
The secondary structure of the cellular IRES of c-Myc (), L-Myc (), Apaf-1 (), FGF-2 (), FGF1 (), Kv1.4 (), Bag-1 (), Igf2 (), cat-1 (), Mnt and MTG8a () have been empirically determined using enzymatic and chemical probing, but no similarities between the structures of these cellular IRES were identified. This could be due to a much wider regulatory range of translation initiation that is needed in distinct cellular contexts relative to a virus's need to translate its messages more efficiently than the host's transcripts. Therefore, the possibility exists that there are structural motifs that are shared in co-ordinately regulated as yet undiscovered IRES.
Determination of viral IRES structures can often benefit from the comparison of many sequences from the same virus, using the variation of sequence to determine which bases pair together. When a structure is preserved, a mutation in a base will be coupled with a second site mutation that preserves the base pairing, and therefore the IRES structure. This co-variation analysis is not always possible with a lower number of available mammalian sequences, and therefore secondary structure determination requires enzymatic or chemical probing to determine the secondary structure in lieu of tertiary structure analysis with NMR or X-ray crystallography. Computational secondary structure prediction has still not achieved the accuracy needed to skip empirical structure determination, but it can be used for comparative purposes when at least one structure is known ().
The X-linked inhibitor of apoptosis protein (XIAP) is the key inhibitor of apoptosis by virtue of binding to and inhibiting distinct caspases (). It was shown that XIAP mRNA is translated by an IRES-dependent mechanism (), and that this mode of XIAP translation is absolutely required for maintaining protective levels of XIAP protein in cells undergoing various forms of cellular stress (). Thus, it is likely that functionally similar IRES exist that govern the expression of cellular genes involved in the control of cellular growth, proliferation and death. In this work, the secondary structure of the XIAP IRES was determined using enzymatic probing. This structure was then used to search a 5′UTR database using the RSEARCH program (), which predicted that several mRNAs had some similar structure features. When tested in a bicistronic reporter construct, two of these UTRs from Aquaporin4 and the uncharacterized ELG1 exhibited IRES activity, while the 5′UTR of NRF was shown previously to contain an IRES element (). Further structural and biochemical probing showed that XIAP, AQP4 and ELG1 share only limited RNA structure similarity; however, additional biochemical analyses demonstrated that they have several IRES -acting factors in common. These data prompt us to propose that, unlike viral IRES elements, the cellular IRES are primarily defined not by an overall common structure but rather by common short RNA motifs and shared -acting factors.
Human embryonic kidney (HEK) 293 and 293T cells were maintained in standard conditions in Dulbecco's modified Eagle's medium supplemented with heat-inactivated 10% fetal calf serum, 2 mM -glutamine and 1% antibiotics (100 units/ml penicillin–streptomycin). Transient transfections were performed using Lipofectamine Plus reagent (Invitrogen) according to the manufacturer's protocol. Briefly, cells were seeded at a density of 3 × 10 cells/ml in 6-well plates and were transfected 24 h later in serum-free Opti-MEM medium (Invitrogen) with 2 µg of DNA and 4 µl of lipid per well. The transfection mixture was replaced 3 h later with fresh Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, 2 mM glutamine and 1% antibiotics.
The pβgal/CAT bicistronic vector containing the minimal human XIAP IRES was described previously as pβgal/5′(−162)/CAT (). The empty vector was modified to include an NheI site upstream of the original XhoI site to enable directional cloning of UTR sequences. The new UTRs were isolated by reverse transcribing total RNA from HEK 293T cells using the First-Strand cDNA Synthesis kit (Amersham Biosciences) with random hexamer primers and then PCR amplifying the UTRs with the following specific primers: AQP4-Nhe-F cgctagcAAGGACAGTTTGGATAAT, AQP4-Xho-R cctcgagCCACCATGATGTTCTCT, ELG1-Nhe-F cgctagcACTTTTGGTGGGCATTTA, ELG1-Xho-R cctcgagCTGGGCGGGGAATA, PCD10-Nhe-F cgctagCCTCAGTTGCTGGTAAG, PCD10-Xho-R cctcgagAAGCCAACTACAGTTGAA, VEGF-D-Nhe-F cgctagcACTTCTCTGCATTTTCT, VEGF-D-Xho-R cctcgagTTTCAATATCCACTGATT, ZINCF-Nhe-F cgctAGCCTGAGAAGATGATGC, ZINCF-Xho-R cctcgaGCTTTCCCACATTCACA. The bicistronic construct with the ATF4 5′UTR was provided as a gift from Dr Jamie Blais. The XIAP-PPT construct with the −47 to −1 sequence removed was previously described (). The AQP4-PPT construct was created using the AQP4-Nhe-F primer and the AQP4-59-Xho-R primer CCTCGAGAGAAACAAATCAGCA and cloning into the pβgal/CAT bicistronic vector. The FLAG-tagged PTB-overexpressing construct was generated by inserting reverse transcription-PCR (RT-PCR) amplified PTB-1 cDNA into the pcDNA 3 vector (Invitrogen). The FLAG epitope was incorporated into the N-terminus of PTB using the following primers: 5′ –gacggcattgtcccagat (BamHI recognition site is underlined; FLAG-coding sequence is italicized) and 5′–ctagatggtggacttggagaag (XhoI recognition site is underlined). The monocistronic constructs were made by deleting out the β-galactosidase open reading frames from the bicistronic constructs using NotI. The hairpin sequence, which was previously described (), was inserted in the NheI site, which is 5′ to the UTRs.
RNA secondary structures were determined using enzymatic probing with RNase T1, RNase A, RNase T2 and RNase V1 according to the protocol supplied with the Ambion enzymes. RNA was transcribed using the MaxiShortScript kit (Ambion) following the manufacturer's protocol. All RNA structures were examined from both ends. Radioactive 5′ end labelling was performed on uncut RNA with Ambion's KinaseMax Kit. Instead of 3′ end-labelling, cold, RNase digested RNA was reverse transcribed using a fluorescent tagged primer from the 3′ end and the reaction was separated on a Licor IR DNA Sequencer alongside a control sequence reaction of the same sequence. RNase cut sites were used as constraints in either MFOLD () or RNASTRUCTURE () to predict secondary structure models. The XIAP mutant structures were probed with -methylisatoic anhydride (NMIA) following the protocol of Wilkinson . () with the modification that reverse transcription took place using a fluorescent tagged primer from the 3′ end and the products were separated on a Licor IR DNA Sequencer alongside a control sequence reaction of the same sequence. RNA samples were reacted with 160, 65, 32 and 16 mM NMIA. Concentrations of 32 and 16 mM NMIA gave the best results for the amount of RNA used.
A human 5′UTR database was generated from the ENSEMBL genome annotation site for NCBI build 34. Exact duplicate UTRs were removed along with shorter UTRs that fully overlapped other entries. The database was supplemented with full-length UTRs of a published IRES dataset (). CT structure files were converted to Stockholm format and used to search the UTR database using the default parameters of RSEARCH on a grid of 6 Sun workstations.
Transiently transfected cells were harvested 24 h post-transfection in CAT enzyme-linked immunosorbent assay (ELISA) kit lysis buffer (Roche), and cell extracts were prepared using the protocol provided by the manufacturer. β-Galactosidase enzymatic activity in cell extracts was determined by the spectrophotometric assay using ONPG (-nitrophenyl-β--galactopyranoside) (), and the CAT levels were determined by using the CAT ELISA kit (Roche) and the protocol provided by the manufacturer. Neomycin phosphotransferase II levels were detected using neomycin phosphotransferase II ELISA (Agdia) following the protocol of the manufacturer. Transient transfections were performed in HEK 293 cells where neomycin phosphotransferase II was being used because HEK 293T cells contain the NPTII gene. Relative IRES activity was determined as a ratio of CAT to β-galactosidase. Unless otherwise noted, the data represent mean ±S.E.M. of three independent experiments performed in triplicates.
Total RNA was isolated from cells that were previously transfected with the βgal/CAT bicistronic constructs as described above. For quantitative RT-PCR, reverse transcription was carried out using the First-Strand cDNA Synthesis kit (Amersham Biosciences) with NotI-d(T) primers. The quantitative PCR was performed using the QuantiTect SYBR green PCR kit (Qiagen) and analysed on an ABI Prism 7000 sequence detection system using the ABI Prism 7000 SDS Software. Quantitative PCRs were carried out to detect β-galactosidase (5′-ACTATCCCGACCGCCTTACT-3′ and 5′-CTGTAGCGGCTGATGTTGAA-3′) and CAT (5′-GCGTGTTACGGTGAAAACCT-3′ and 5′-GGGCGAAGAAGTTGTCCATA-3′) as described previously ().
Isolation of IRES-binding proteins was performed using a modified RNA-affinity chromatography protocol (). Briefly, XIAP IRES RNA, AQP4 5′UTR RNA, ELG1 5′UTR RNA, XIAPΔPPT RNA and AQP4ΔPPT RNA were transcribed with the MEGAShortscript transcription kit according to the manufacturer's protocol (Ambion), and were biotinylated at the 5′ end with the 5′ EndTag Nucleic Acid Labelling System according to the manufacturer's instructions (Vector Laboratories). The biotinylated RNAs (15 μg) were conjugated to Avidin-agarose beads (Sigma) in the presence of incubation buffer (10 mM Tris-Cl [pH 7.4], 150 mM KCl, 1.5 mM MgCl, 0.5 mM DTT, 0.5 mM phenylmethylsulfonyl fluoride, 0.05% [v/v] Nonidet P-40) at 4°C for 2 h with continuous rotation. Unbound RNAs were removed by washing beads twice with incubation buffer. Five hundred microgram of 293T protein extract (in incubation buffer) was added to the coated beads, along with 30 μg yeast tRNA (Sigma) and 200 units of Prime RNase inhibitor (Eppendorf). Reactions were incubated at room temperature with continuous rotation for 30 min, followed by incubation at 4°C with continuous rotation for 2 h. Beads were washed five times with incubation buffer, resuspended in 20 μl of 1× SDS-PAGE loading dye, and boiled for 5 min to elute bound proteins. Proteins were then separated by 10% SDS-PAGE and transferred to PVDF membrane for analysis by western blot using antibodies specific for La (), hnRNP C1/C2 (AbCam), hnRNP A1 (Santa Cruz Biotechnology) and PTB (Zymed Laboratories).
The 5′UTR of XIAP mRNA promotes cap-independent translation initiation through the activity of an IRES element located between nucleotides −162 to +3 (relative to the start codon) (,). Our recent investigations have addressed the role of -acting factors in XIAP IRES activity (,,), yet little is known about the contribution of the RNA structure that is formed by this IRES sequence to the function of the XIAP IRES. Several studies have shown the importance of secondary structure for viral IRES activity (); however, the importance of secondary structure for cellular IRES remains unclear. We therefore hypothesized that if secondary structure is an important determinant of cellular IRES activity, we should be able to identify new cellular IRES by searching for RNA sequences that form structures that are similar to a known IRES element (e.g. XIAP IRES). As a first approach to characterize the role of secondary structure in cellular IRES activity, we empirically determined the secondary structure of the XIAP IRES RNA. The sequence from −162 to +3 of the XIAP 5′UTR was transcribed and subjected to digestion with RNases T1, V1 and T2. The relative sensitivities of sites to these RNases are presented in Supplementary Table 1. We then used the information derived from the RNase T1, V1 and T2 digests as constraints for folding the XIAP IRES RNA sequence into a secondary structure using the MFOLD structure prediction webserver () (). The model proposed has two domains with the first domain having two stems, Ia and Ib, flanking an unstructured region and the second domain, which is smaller and contains a polypyrimidine tract in its major stem that has previously been shown to be required for IRES activity ().
We were interested in identifying novel IRES that are structurally similar to the XIAP IRES, since the ability to do so would suggest an important role for secondary structure in cellular IRES function. Additionally, cellular IRES that share a similar secondary structure may also be regulated and/or function in a similar manner. We therefore developed a method for searching a database of 5′UTR sequences for structural features that are similar to a query sequence. A database of human 5′UTR sequences was compiled from the Ensembl database using Ensmart (now known as BioMart) (). The sequences were filtered for redundancy by removing exact duplicate matches, and 5′UTR sequences that existed within longer 5′UTRs were discarded. The 5′UTRs from published cellular and viral IRES were added from an in-lab database () to the Ensembl-created database. To search this database we used the RSEARCH program, which is designed to search a database with a sequence or sequence alignment that has a known secondary structure (). This program uses scoring matrices built with alignments of the small subunit rRNA, called RIBOSUM, to score for putative preserved base pairings in both the query structure and the database entry, as well as specific nucleotide matches in the alignment. RSEARCH outputs a sequence and structural alignment of the query structure and the RSEARCH-predicted structure of the database entry for each match that is above a user-defined cut-off score. We first tested efficacy of the RSEARCH program for our objectives by searching for 5′UTRs that are structurally similar to the well-studied HCV IRES structure, which was defined for this search by incorporating the latest NMR and crystallographic data (). A distribution of the RSEARCH match scores is shown in A. The top matches were the Hepatitis B isolate, GbvB, CSFV and the BVDV 5′UTRs (B), all known to have similar structures to the HCV IRES (,,). The primary sequences of either the CSFV 5′UTR or the BVDV 5′UTR lack sufficient similarity to the primary sequence of the HCV IRES to be found using BLAST (data not shown), and were therefore identified due to structural similarity and not sequence homology. The one high scoring match to the human PLEKHG4 5′UTR is perplexing. Although the structural alignment to domain III of HCV is quite strong, there are a couple of anomalies that suggest this match may not be significant. First, the two gaps of PLEKHG4 at the analogous IIId and IIIe loops of HCV are quite large, 86 and 225 nt, respectively, and the proposed structure may not form as RSEARCH has predicted. Second, the published IRESes so far have been fairly close to the AUG start codons and this structure is over 200 bases upstream of the known start codon for PLEKHG4. Although these differences suggest that PLEKHG4 would not have a structure that matches the HCV IRES, it must be noted that the match of PLEKHG4 to domain IIIb of the HCV IRES is quite good. Domain IIIb of HCV is known to bind directly to eIF3, the adapter protein that normally binds the 40S ribosomal subunit and eIF4G (,).
After confirming the utility of our search process using the HCV IRES structure, the XIAP IRES structure () was next used as a query to search the same database, and the distribution of the match scores for this search is displayed in A. As the minimal XIAP IRES 170 nucleotide sequence we used is less than the 372 nt of the HCV IRES sequence, the likelihood of structure or sequence matches is much less and therefore the overall scores will be much lower. This is apparent, as the exact match to the XIAP IRES is 332 compared to the exact match to the HCV IRES with a score of 701. The top matches for the XIAP IRES structure were the mouse and rat XIAP IRES orthologs, which have similar primary sequence, with scores of 173 and 144, respectively. The highest match score for a 5′UTR of human sequence known to have an IRES was NRF (), with a score of 35.5 and so we examined hits with scores over 35. There were ∼80 matches (in addition to the XIAP IRES orthologs) with scores higher than 35, which represent a range of 11–14% of an exact match score. In comparison with the HCV structure search, the score for CSFV is just under 14% of an exact match score of the HCV IRES structure.
The 5′UTR of NRF was shown previously to contain IRES activity (), confirming that our search could identify functional IRES. We therefore wished to test the identified 5′UTRs that had significant match scores (i.e. >35) for IRES activity. Several criteria were used to select five 5′UTRs (denoted by an asterisk in B) from the list of potential structure matches for additional testing. As all characterized IRES elements are located just upstream of the AUG start codon, only matches that were present in the 3′ end of the 5′UTRs were chosen. Structure matches that had non-aligning inserts greater than 100 bases within the structural alignments were also not chosen for further testing, as these may represent structures unlikely to occur naturally. The sequences for each of the five chosen 5′UTRs were amplified by RT-PCR using RNA isolated from HEK 293T cells as a template, and the resulting products were subcloned into the pβgal/CAT bicistronic reporter vector (). The 5′UTR from ATF4, which has been shown to not have IRES activity (Blais,J., personal communication), was also cloned into the pβgal/CAT bicistronic vector and serves as a negative control. These plasmids were subsequently transfected into HEK 293T cells, and relative IRES activity was determined by assessing the levels of β-galactosidase and CAT expression. Of the five UTRs tested, only the ELG1 and Aquaporin 4 (AQP4) 5′UTRs exhibit IRES activity, although this activity is much lower than that exhibited by the XIAP IRES (A).
To ensure that the IRES activity exhibited by these two 5′UTRs is not due to the function of a cryptic promoter within the cloned UTR sequence, the CMV promoter was excised from the vector. In such a promoterless construct, any expression of CAT must be due to the presence of cryptic promoter activity within the cloned UTR sequence. The β-galactosidase and CAT protein levels in extracts from HEK 293 cells transfected with the promoterless bicistronic reporter constructs was measured, and is expressed relative to the levels of the neomycin phosphotransferase II (NPTII), which is also contained on the bicistronic vector. NPTII expression is driven by the SV40 promoter, and is therefore expressed independently from the βgal/CAT bicistronic RNA, allowing NPTII expression levels to serve as a control for transfection efficiency. None of the 5′UTRs tested displayed any cryptic promoter activity, as evidenced by the absence of CAT expression from the promoterless constructs (B). The 5′UTRs exhibiting IRES activity were also tested for spurious splicing events by performing quantitative RT-PCR (qRT-PCR) analysis of mRNA isolated from transfected cells using oligonucleotide primers specific to each of the βgal and CAT open reading frames (). We did not detect any splicing events as the ratio of CAT to βgal mRNA remained the same for all of the constructs tested, which also demonstrates the equivalent stability of all the messages (C).
In order to confirm cap-independent activity of messages with the 5′UTRs of AQP4 and ELG1, we constructed monocistronic constructs with these UTRs and the CAT open reading frame with and without upstream hairpin structures. Cap-dependent initiation is inhibited in constructs where a thermodynamically stable hairpin is added upstream of the coding region, which hinders ribosome scanning initiated at the 5′ end of the message (,,). The addition of an upstream hairpin in the control pCAT construct significantly inhibited translation of CAT (D). In contrast, translation mediated by AQP4, ELG1 and XIAP 5′UTRs was resistant to the inhibition by the stable hairpin (D). Therefore, these data indicate that the ELG1 and AQP4 5′UTRs exhibit IRES activity.
We have used XIAP IRES secondary structure to conduct a genome-wide search for novel cellular IRES elements that are structurally similar to the XIAP IRES. To determine if the IRES-containing 5′UTRs of AQP4 and ELG1 were indeed identified by the RSEARCH program due to actual structural similarities between these UTRs and the XIAP IRES, the secondary structures of the ELG1 and AQP4 5′UTR sequences were empirically determined by RNase probing (Supplementary Tables 3–6). The secondary structure models that were generated for the AQP4 and ELG1 5′UTRs, and the RSEARCH sequence/structure matches with the XIAP IRES structure, are shown in . Surprisingly, the secondary structure of the ELG1 IRES does not exhibit any similarity to the structure of XIAP IRES (A). The sequences of ELG1 that RSEARCH attributes to the base-paired stems of domain I or II of the XIAP IRES do not anneal in the empirically determined structure. Closer inspection of the ELG1 sequence shows that it contains 35 AUG repeats, which could theoretically allow it to form many structures, such as the one predicted by RSEARCH.
However, the AQP4 5′UTR structure contains a domain that displays significant similarity to the structure observed in the XIAP IRES domain II. Notably, a double-stranded sequence in the AQP4 5′UTR that contains a polypyrimidine tract is similar to the structure of a polypyrimidine tract in the XIAP IRES (B). Both UTRs include the UUCUCUUUU motif. This polypyrimidine tract has been shown previously to be important for XIAP IRES activity, since upon deletion or mutation of these sequences IRES activity is lost (). We therefore wished to determine if this polypyrimidine tract is also important for AQP4 IRES activity. To this end, the sequence between −59 and −3 within the AQP4 5′UTR (containing the polypyrimidine tract) was deleted in the bicistronic vector, and the resulting construct was then tested for IRES activity. We found that deletion of the polypyrimidine tract sequences in both the XIAP and AQP4 5′UTRs completely abrogates IRES activity (). Therefore, the polypyrimidine tract is absolutely required for the IRES activity of both the AQP4 and XIAP 5′UTRs.
The comparison of the AQP4, ELG1 and XIAP IRES structures suggest that they do not share similar RNA structures. This could indicate that perhaps cellular IRES structures are not as important as viral IRES. Indeed, there exists previous evidence in support of this claim. The polypyrimidine tract upstream of position −34 in the XIAP IRES appears to form a stem (). It is known that the deletion of position −1 to −34 relative to the start codon does not affect activity () but would most likely disrupt the stem formed with the polypyrimidine tract. In order to determine if more disruption in XIAP IRES's structure is possible without affecting IRES activity, we tested several XIAP IRES mutants previously made in our laboratory. It was apparent that several mutants had similar or slightly better activity than the wild-type XIAP IRES sequence (A). The position of some of these mutations should not necessarily change the IRES structure as they are in loop regions; other mutants, however, are predicted to result in altered IRES structure (B). Using -methylisatoic anhydride (NMIA) to selectively probe for open regions of the RNA structure () we found that at least two of the mutants, Mut2-34 and 74c74-34, exhibited changes in their structure relative to the wild-type sequence. Importantly, these mutants did not exhibit any loss in IRES activity. Examples of these changes (denoted by either asterisks for mutant 74c74-34 or an ‘o’ for Mut2-34; B and C) showed that the structural changes were not just at the site of sequence mutations but were elsewhere in the IRES as well. These data support our conclusion that overall structure is not of primary importance for the XIAP IRES's activity.
Although the structural similarity among the IRES elements of XIAP, AQP4 and ELG1 was weak, we cannot rule out that these sequences share IRES -acting factors (ITAFs), which allowed the ELG1 and AQP4 5′UTRs to be identified by our search protocol because the ITAFs may have weakly similar binding motifs (structural or sequence similarity) that would contribute to the RSEARCH match score. ITAFs are mRNA-binding proteins that are involved in a variety of processes dealing with mRNA processing, and are therefore quite promiscuous in the mRNA to which they bind. As the identity of ITAF-binding sites may be difficult to assess from the primary sequences, we have used RNA-affinity chromatography to isolate proteins that bind to the IRES RNA sequences and have probed for the presence of known ITAFs by Western blot analysis. It has already been shown that the La autoantigen () and hnRNP C1/C2 () bind to the XIAP IRES and enhance its activity, whereas hnRNP A1 binds to the XIAP IRES and reduces IRES activity (). Western blot analysis showed that the La autoantigen associates with the XIAP, AQP4 and ELG1 IRES, whereas hnRNP A1 was found to be associated with the XIAP and ELG1 IRES elements, and hnRNP C1/C2 were only found to be associated with the XIAP IRES (A). Since we have found that a polypyrimidine tract is required for both XIAP and AQP4 IRES activity, we assessed whether the polypyrimidine tract binding protein (PTB), a known ITAF for several IRES elements (,,), could associate with these IRES sequences. Indeed, we found that PTB can associate with both the XIAP and AQP4 IRES, but not the ELG1 IRES (A). To confirm that the polypyrimidine tracts of the XIAP and AQP4 IRES elements are required for PTB binding we performed RNA-affinity chromatography using XIAP and AQP4 IRES RNAs in which the polypyrimidine tracts have been deleted. We find that deletion of the polypyrimidine tract in either the XIAP or AQP4 IRES RNA abrogates PTB binding, but does not disrupt the binding of other ITAFs (B). Therefore, the polypyrimidine tracts of the XIAP and AQP4 IRES elements are important for ITAF binding. Moreover, the 5′UTRs identified by our search share the ability to bind to a similar cohort of -acting factors.
To elucidate the role of PTB in XIAP and AQP4 IRES function, we tested the effect of PTB overexpression on the activity of these IRES elements. We find that PTB is a ITAF for both the XIAP and AQP4 IRES elements, as it affects their activities (C). Surprisingly however, we find that overexpression of PTB represses XIAP IRES activity, whereas PTB overexpression enhances AQP4 IRES activity ∼2-fold (C).
In this work, we empirically determined the secondary structure of the XIAP IRES and set out to identify novel cellular IRES by searching for human 5′UTR sequences that display a structural similarity to the XIAP IRES. We first tested the efficacy of our search protocol by using the well-characterized structure of the HCV IRES to search a database of human 5′UTRs and UTRs known to contain IRES. This search effectively identified IRES from GbvB, CSFV and BVDV, which have been previously shown to have structural similarity to the HCV IRES (,,). This exercise established that structurally similar UTRs could be found using the RSEARCH program, thus validating our approach. Our search of the human 5′UTR database using the structure of the XIAP IRES resulted in the identification of five 5′UTRs in which the region with structural similarity was in the correct position (at the 3′ end of the UTR, proximal to the AUG codon) and orientation to exhibit IRES activity. Characterization of the IRES activity of these identified sequences showed that the 5′UTRs of AQP4, ELG1 and NRF promote IRES-dependent translation. However, empirical determination of the structure of the AQP4 and ELG1 5′UTRs and subsequent comparison to the structure of the XIAP IRES revealed that the structures of these sequences share little similarity. We then found that a polypyrimidine tract in the XIAP IRES and AQP4 IRES is absolutely necessary for IRES activity, and that similar -acting factors (ITAFs) can bind to these IRES sequences. Our results lead us to propose that, unlike the viral IRES, the overall structure of cellular IRES is not necessarily an important factor in determining IRES activity, but rather small motifs and the cohort of proteins that bind them define cellular IRES activity.
It has been recognized that both viral and cellular IRES do not share primary sequence homology, and therefore it is not possible to identify mRNAs that harbour IRES elements by comparison of sequence data using search programs such as BLAST. This realization has presented a barrier to the genome-wide identification of IRES, resulting in the identification of mRNAs that contain IRES in a piecemeal fashion—a message is suspected to be translated under conditions that repress cap-dependent translation and the 5′UTR is subsequently tested for IRES activity. However, several studies have highlighted the importance of secondary structure for the function of viral IRES (,,) and some viral IRES share structural homology, suggesting that an underlying determinant for function of viral IRES is the presence of specific structural elements (). We therefore hypothesized that cellular IRES may also share structural homology, and that comparison of the secondary structure of 5′UTRs may be a means to identify novel IRES on a genomic scale.
Our searches of a human 5′UTR genome, supplemented with 5′UTRs known to exhibit IRES activity from all species and viruses, with the structure of the XIAP IRES did indeed result in the identification of RNA sequences that exhibit IRES activity. Of the top matches, two were the 5′UTRs of the mouse XIAP ortholog (MIAP) and the rat XIAP ortholog (RIAP) that both display IRES activity () (Holcik,M., unpublished data), one that has previously been shown to have IRES activity [NRF;()], and we have demonstrated that two others (ELG1 and AQP4) also display IRES function. The IRES activity of AQP4 and ELG1 are much less than XIAP. However, we have only tested IRES activity in one cell line and under conditions of normal cell growth. AQP4, a water channel-forming protein expressed in the brain, has recently been shown to have a role in edema during eclampsia, as its protein levels are elevated during pregnancy (). Interestingly, mRNA levels of AQP4 have not been shown to change (), suggesting a possible IRES function. ELG1 is an uncharacterized transcript that exists in the database as a fully sequenced cDNA clone, accession # AK125048 and a portion of the UTR is represented by EST DB217710.1. While the presence of numerous AUGs in its 5′UTRs is unusual, it is nevertheless not completely uncommon as around 1% of all 5′UTRs have 30 to 100 upstream AUGs (). Interestingly, we did not identify the 5′UTRs of p27(Kip1) () or Bcl-xL, both of which exhibit IRES activity (), in our search (matching scores were between 16 and 17). Yoon . () recently showed that the IRES activity of XIAP, p27(Kip1) and Bcl-xL is specifically and severely impaired in cells harbouring mutations in the pseudouridinase gene Dkc1. This mutation was shown to prevent the proper pseudouridylation of rRNA and is thought to disrupt ribosome structure (). The IRES-dependent translation of the XIAP, p27(Kip1) and Bcl-xL messages is specifically sensitive to these changes, as global translation is not affected. It was therefore hypothesized that these IRES may share some common feature, such as secondary structure, that is sensitive to changes in ribosome architecture (). Based on our results, the secondary structures of these IRES elements may not be similar, and other factors, such as changes in the association of -acting factors with the modified ribosome or direct association of the IRES with the ribosome, may account for the coordinated reduction of XIAP, p27(Kip1) and Bcl-xL IRES activity in Dkc1 mutant cells.
To our surprise, comparison of the empirically determined secondary structures of the ELG1 and AQP4 5′UTRs to the secondary structure of the XIAP IRES showed only limited homology, with only small regions of AQP4 displaying some similarity within the overall structure. From the search of HCV IRES structure we learned that the score of CSFV, which is just over 13% of a perfect match, was significant. Our search for a similar IRES structure with a cut-off score of 35 had us examining matches with scores in the range of 11–14% of a perfect match. In retrospect, these values may have been too low to be significant, but represent the top matches available. As the search with the HCV structure validated this search protocol, the lack of results with the XIAP IRES structure search strongly supports the notion that there are no similar structures. This suggests that cellular IRESes do not share globally similar structure as do some viral IRESes. This conclusion is plausible when considering published evidence to date. First, no overall structural similarity has been found among the 11 known cellular IRES structures (). Second, examination of the c-myc IRES has shown that its activity is not dependent on the overall structure but on distinct sequence modules (). Third, mutations which opened up the Bag-1 IRES structure negated the need for the previously required ITAF, PCBP1, while still retaining full IRES activity (). Fourth, non-overlapping segments of IRES sequence that would not preserve the overall structure still retain partial IRES activity (,), suggesting different modules may synergistically act to provide full IRES activity . In this light, it is not surprising that the XIAP IRES mutants we examined exhibited full IRES activity despite structural changes. Conversely, the point mutation of the IRES which attenuates the Sabin strain 3 of poliovirus does not change the structure () but it does effect the binding of PTB (). It would be wrong to over generalize and suggest that structure will never be required for the function of cellular IRESes. Such an exception is the CAT-1 IRES, which is induced during amino acid starvation, where a specific stem is required for functional activity regardless of its sequence makeup ().
Whereas a virus needs to compete efficiently to get its proteins expressed within an infected cell, cellular transcripts with an IRES are more likely translationally regulated, with several -acting factors contributing differently depending upon the cellular context. Studies of the HCV IRES have shown it to directly interact with the ribosome and induce conformational change of the ribosome (). Cricket paralysis virus IRES has an RNA structure which allows it to interact with the ribosome and to initiate translation without an eIF2—initiator Met-tRNA complex in the P-site (). This mechanism may have evolved to overcome the antiviral response of the cell to shut off protein synthesis though PKR phosphorylation of eIF2α (). The importance of structures seen in viruses has yet to be shown for any cellular IRES.
Importantly, we noted the presence of a polypyrimidine tract that was folded into a stem-loop structure in the both XIAP and AQP4 IRES. As we have previously found that the polypyrimidine tract is absolutely necessary for XIAP IRES activity (), we tested whether deletion of this sequence in the AQP4 5′UTR would also abrogate IRES function. Indeed, deletion of the AQP4 5′UTR polypyrimidine tract causes a loss of PTB binding and a complete loss of IRES activity. Because this small region is required for IRES activity, we believe that short motifs may be critical determinants of IRES function. In agreement with this hypothesis, it has been found that a reiterated PTB-binding motif can induce internal ribosome entry (). Moreover, a search by Mitchell . () for 5′UTRs that harbour this PTB-binding motif has resulted in the identification of novel IRES elements. This search, however, would not have found XIAP or AQP4 because the (CCU) motif pattern considered only a small subset of sequence sites that PTB can bind and is not found in XIAP or AQP4 IRES. Even though the polypyrimidine tract region (UUCUCUUUU) is the same in XIAP and AQP4, PTB overexpression has an opposite effect on their IRES activity. This may possibly be due to competing ITAFs for sequences in this region, which can have either a positive or negative effect on translation initiation. Our observations, together with previously published findings, suggest that the recruitment of a particular cohort of -acting proteins is the critical factor in cellular IRES-mediated translation. Therefore, a cataloguing of the -acting factors required by each IRES may allow functional grouping of these elements and aid in the identification of common features required for cellular IRES-dependent translation.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Replicative DNA polymerases are multiprotein complexes that are capable of synthesizing long stretches of DNA. The high processivity of these polymerases is dependent upon accessory proteins called processivity factors that bind to the catalytic subunit of the polymerase and prevent its dissociation from the template. The best-characterized processivity factors are the so-called ‘sliding clamps’, such as proliferating cell nuclear antigen (PCNA) of eukaryotic DNA polymerases δ and ε (,). Sliding clamps possess no inherent DNA-binding capacity, but require clamp-loading proteins to be assembled onto DNA as toroidal homomultimers in an ATP-dependent process (). After they are loaded onto DNA, the sliding clamps can slide freely along DNA and tether the catalytic subunit to the template, thus ensuring processivity without impeding the movement of the polymerase.
The processivity factors of herpesvirus DNA polymerases employ mechanisms that have yet to be completely explained. The most-studied herpesvirus processivity factor is that of herpes simplex virus (HSV), UL42, which together with the catalytic subunit (UL30) composes the viral DNA polymerase. UL42 differs from sliding clamps in that it binds directly to DNA as a monomer with high affinity and in a manner that does not require clamp loaders or ATP hydrolysis (). Despite the high affinity for DNA, UL42 can diffuse linearly along DNA in the absence of UL30 (). It has been proposed that electrostatic interactions between basic residues on the α-helical ‘back’ face of the UL42 structure and the phosphate backbone of DNA () provide a tether that prevents dissociation, but allows UL42 to diffuse along the helical backbone. In support of this idea, it has been recently shown that substitutions of arginine residues on the basic surface of UL42 decrease the DNA-binding affinity of the protein ().
The human cytomegalovirus (HCMV) DNA polymerase is also composed of a catalytic subunit, Pol or UL54, which possesses basal DNA polymerase activity (,), and an accessory protein, UL44 (). UL44 is believed to serve as the processivity factor for the polymerase, as it has been shown to specifically interact with UL54 and to stimulate long-chain DNA synthesis by UL54 (,). The crystal structure of residues 1–290 of UL44 (UL44ΔC290) () revealed that UL44 has a fold remarkably similar to that of HSV UL42 () and monomers of PCNA (,), even though these proteins have no obvious sequence homology. In addition, like HSV UL42, HCMV UL44 possesses a basic, α-helical back face, which may be involved in binding to and diffusion along the DNA backbone via electrostatic interactions. However, in contrast to UL42, which is a monomer (,,), and PCNA, which is a head-to-tail toroidal homotrimer (,), UL44 forms a dimer in solution in the absence of DNA and its crystal structure shows a head-to-head C clamp-shaped homodimer (,). Although it has been shown that UL44 can bind double-stranded (ds) DNA and that mutations which affect dimerization also affect DNA binding (,,), the details of the UL44–DNA interaction, including whether UL44 binds DNA as a dimer, have not been yet investigated.
Compared to our knowledge of sequence-specific DNA-binding proteins, derived from numerous structural, biochemical and thermodynamic studies, relatively little is known about how non-sequence-specific DNA-binding proteins interact with DNA. Thus, investigations of proteins such as UL44 can shed light on this class of DNA-binding proteins. In this study, we measured the binding of UL44 to DNA using several techniques, and tested the dependence of binding on various properties of DNA including strandedness, sequence and length. Thermodynamic analysis indicated that UL44 binds DNA as a dimer and that binding is entropically driven, while the dependence of binding on ionic strength allowed an estimation of the number of monovalent ions released in the interaction of UL44 with DNA. The results are consistent with hypothesized electrostatic interactions between basic residues of UL44 and the phosphate backbone of DNA.
The wild-type UL44ΔC290 and mutant UL44ΔC290 L86A/L87A proteins were expressed in and purified from BL21(DE3)pLysS (Novagen) as described previously (,). Baculovirus-expressed, full-length UL44, purified as previously described (), was generously provided by Howard S. Marsden. The purity of the UL44 preparations used for binding studies was estimated to be >90% as assessed by SDS–PAGE followed by silver-staining of the gel (Figure S1). Concentrations of all proteins were determined using amino acid analysis at the Molecular Biology Core Facility, Dana-Farber Cancer Institute.
Single-stranded (ss) oligonucleotides of varying lengths (8, 10, 12, 15, 18, 30 or 47 nt) and sequence were purchased from Integrated DNA Technologies, Inc. (Coralville, IA, USA) or from MWG-BIOTECH AG (Ebersberg, Germany) as gel-purified products. To generate ds oligonucleotides, equal amounts of complementary strands were mixed in annealing buffer (50 mM NaCl, 10 mM Tris, pH 8.0, 1 mM EDTA), completely denatured at 95°C and annealed by slowly cooling to room temperature. Formation of dsDNA was confirmed on a 20% (for shorter oligonucleotides) or 10% (for longer oligonucleotides) non-denaturing polyacrylamide/0.5× Tris–borate EDTA (TBE) gel. The concentrations were determined via UV spectrometry. Some of the oligonucleotides were used as ssDNA in filter-binding assays. These oligonucleotides were designed to be devoid of intramolecular structure. Additionally, prior to assay, they were heated to 95°C for 5 min, and then quick-cooled on ice. For filter-binding and electrophoretic mobility shift assays (EMSAs), oligonucleotides were radiolabeled with [γ-P]ATP and T4 polynucleotide kinase and purified as previously described ().
Filter-binding assays were performed by a modification of the double-filter method previously described (,). Briefly, 1 fmol of labeled (∼10 000 c.p.m./fmol) ssDNA or dsDNA was incubated at room temperature with various amounts of purified UL44ΔC290 or baculovirus-expressed, full-length UL44 for 10 min in a 10 μl volume in binding buffer containing 20 mM Tris–HCl (pH 7.5), 50 mM NaCl, 4% glycerol, 3 mM MgCl, 0.5 mM DTT, 0.1 mM EDTA and 40 μg of bovine serum albumin (BSA) per milliliter. At the end of the incubation, samples were applied to a nitrocellulose/DE81 filter stack soaked in binding buffer using a vacuum manifold. The DNA–protein complexes were trapped on the nitrocellulose filter (Schleicher and Schuell, Keene, NH, USA) and the remaining, unbound DNA was trapped on the DE81 filter (Whatman, Dassel, Germany) placed under the nitrocellulose filter. Filters were then washed extensively with binding buffer and dried, and radioactivity was measured by liquid scintillation counting. All data were corrected for background (i.e. radioactivity retained on the filter in the absence of UL44). As the concentration of DNA in these filter-binding assays was less than the apparent dissociation constant () values, these constants were calculated as being equivalent to the amount of protein at which 50% of the DNA was bound by using saturation isotherm analysis.
Experiments to test the affinity of UL44 for DNA in different ionic environments were conducted as described above, but with binding buffers containing 1 mM Tris–HCl (pH 7.5), 4% glycerol, 0.5 mM DTT, 0.1 mM EDTA, 40 μg/ml BSA and various concentrations of NaCl, KCl, NaCHCO or MgCl.
EMSA reactions (10 μl) contained 1 nM of [P]-labeled ds oligonucleotide, 20 mM Tris–HCl (pH 7.5), 50 mM NaCl, 4% glycerol, 3 mM MgCl, 40 μg/ml BSA, 0.1 mM EDTA, 0.5 mM DTT and increasing concentrations of either wild-type UL44ΔC290 or mutant UL44ΔC290 L86A/L87A or baculovirus-expressed, full-length UL44 protein. Mixtures of protein and DNA were incubated for 10 min at room temperature. Following addition of 1 μl of loading buffer (0.025% bromophenol blue, 0.5× TBE and 10% glycerol), bound and free DNA were resolved by fractionation on either a 4% (for analysis of binding of full-length UL44 to ds 30-bp template), or a 5% (for analysis of binding of wild-type or mutant UL44ΔC290 to ds 30-bp template) or 10% (for analysis of binding of UL44ΔC290 to ds 18-bp template) native polyacrylamide gel in 0.5× TBE. Gels were pre-run for 1 h at 10 mA at 4°C, using the Miniprotean II system (Bio-Rad), and run under the same conditions. DNA was visualized using dried gels to expose a phosphor storage screen and the data were analyzed using the Quantity One program (Bio-Rad). The apparent value was calculated from EMSAs with 18-bp dsDNA using the equation:
For EMSAs with 30-bp dsDNA, the apparent values for each binding event were calculated by applying the two-site model of Senear and Brenowitz () using the equations:
ITC experiments were performed using a VP-ITC MicroCalorimeter (MicroCal Inc., Northampton, MA, USA) with protein (UL44ΔC290) concentrations of 3–6 μM and DNA (ds 18- or 30-bp oligonucleotide) concentrations of 30–60 μM. Immediately before the experiments were performed, the protein was extensively dialyzed against buffer containing 20 mM Tris–HCl (pH 7.5), 50 mM NaCl, 5% glycerol, 3 mM MgCl, 0.1 mM EDTA and decreasing concentrations of DTT until the concentration of DTT was reduced to 0.5 mM, as higher concentrations would interfere with calorimetric measurements. After annealing of complementary strands, the dsDNA oligomers were dialyzed against the same buffer used to dialyze the protein to avoid heat signals from mixing of nonequivalent buffers. All solutions were carefully degassed before the titrations using equipment provided with the calorimeter. Each titration experiment consisted of 10-μl injections of ds oligonucleotide into the protein-containing sample cell at 25°C with a mixing speed of 270 r.p.m. Heats of dilution were determined by titrating the same ds oligonucleotide into the dialysis buffer and buffer into protein and subtracted from the raw titration data before data analysis. The data were integrated to generate curves in which the areas under the injection peaks were plotted against the ratio of DNA oligomer to protein. Analysis of the data was performed using MicroCal Origin version 5.0 provided by the manufacturer according to the one- and two-binding site models. Changes in the free energy and entropy upon binding were calculated from determined equilibrium parameters using the equation:
Quantitative analysis of the effects of changes in salt concentration on the equilibrium binding constant, , for binding of UL44 to ds 18-bp DNA was performed according to the binding theory of Record . (,). According this theory, when a ligand with positive charges (or a protein with positive charges in its DNA-binding site) binds to a nucleic acid, some number of phosphates (corresponding to ) are effectively neutralized. As a result, the condensed counterions which were associated with the phosphates are released into solution, as well as the ions which were involved in long-range screening interactions. The theory predicts that, in the presence of a monovalent salt MX, the amount of counterion release in a protein–nucleic acid interaction, and therefore the number of ionic interactions involved, may be determined by measurement of the derivative d log /d log [M]. In a sufficiently diluted solution of the salt MX, where differential hydration effects can be neglected, and if there are no complications from anion binding by the ligand, the predicted quantitative dependence of on [M] is ():
When the binding reaction occurs in the absence of multivalent cations, is a constant (0.88 for duplex B-form DNA) and log is a linear function of log [M]; consequently can be determined from the dependence of log on log [M].
The effects on caused by divalent ions such as Mg are qualitatively similar; however the coefficient in Equation () is replaced by φ, which represents the number of divalent counterions thermodynamically associated per phosphate ():
Therefore, in the absence of preferential anion interactions, the relationship in Equation () should hold:
For duplex B-form DNA, φ = 0.47, and φ/ψ = 0.53 (). If Equation () does not hold for a particular ligand–nucleic acid interaction, then this suggests that salt effects other than those due to counterion release from the nucleic acid should be considered. In our experiments, Equation () did hold.
To characterize the DNA-binding properties of UL44, we first performed filter-binding assays to determine apparent affinities for ds and ssDNA. Purified UL44ΔC290, a truncated protein that contains the N-terminal 290 residues of UL44 and retains all known biochemical activities of full-length UL44 (,) as well as the ability to support origin-dependent DNA replication in transfected cells (A.L. and Pari,G., unpublished data), was used in these assays (this truncated protein will be referred to as UL44 below, unless otherwise specified). Increasing amounts of UL44 were added to 1 fmol of 5′-end-labeled template DNA. A variety of ds and ssDNA oligonucleotides of different lengths and sequence composition were used in these assays. To allow comparison with certain previous measurements of the affinity of HSV UL42 for DNA (,), a buffer with relatively low salt concentration (50 mM NaCl) was used in these experiments. The amount of filter-bound radioactivity was measured to determine the fraction of bound DNA, which was plotted against protein concentration (). With these experimental conditions, in which the concentration of DNA was very low, we could then apply a saturation isotherm analysis to calculate apparent s from the concentrations of protein that led to half saturation. The s are apparent, rather than absolute because for any length of DNA longer than the size of the binding site for UL44, there are multiple potential binding sites (,).
In these assays, UL44 bound dsDNA oligonucleotides with lengths of 18 bp or larger with apparent s in the nanomolar range (A and ). With 15- and 12-bp dsDNA templates, the apparent s were markedly higher and UL44 did not bind all of the DNA at the highest concentrations of protein tested, even though binding appeared to plateau (A). Possible explanations for this observation are presented in the Discussion section. Detectable binding was observed only with DNAs of at least 12 bp.
Binding titrations with ssDNA templates of 18–47 nt showed that UL44 possesses ∼3- to 8-fold greater apparent affinity for dsDNA than for ssDNA of the same length ( and ). To test whether the DNA binding of UL44 is affected by the DNA sequence, filter-binding assays were also performed with several sets of ds and ss oligonucleotides of identical length but different sequence. Similar apparent values were measured with DNA templates of the same length, regardless of sequence (Table S1). Thus, DNA binding by UL44 is not meaningfully influenced by DNA sequence.
We also investigated whether the C-terminal one-third of UL44 influences DNA binding by performing filter-binding assays with full-length UL44 purified from insect cells infected with a recombinant baculovirus (). This preparation exhibited an apparent affinity both for dsDNA ( and ) and for ssDNA (data not shown) comparable to that of the -expressed, truncated UL44ΔC290 protein, suggesting that the DNA-binding properties of UL44 fully reside in the N-terminal two-thirds of the protein.
To assess the stoichiometries with which UL44 binds dsDNA, we performed DNA mobility shift assays with radioactively labeled ds 18mer and 30mer oligonucleotides. Although UL44 was able to bind in filter-binding assays to 12-bp DNA (), no binding to DNAs of 12 or 15 bp was detected in EMSAs (data not shown), likely due to the necessity for a slightly longer DNA template (i.e. 18 bp) to ensure sufficient binding stability of the protein–DNA complex during gel electrophoresis. Constant amounts of 5′-end-labeled DNA were titrated with increasing amounts of UL44. After a 10-min incubation, the protein–DNA complexes were resolved on a non-denaturing gel. As shown in A, incubation of increasing concentrations of UL44 with [P]-labeled ds 18mer resulted in increasing formation of a single protein–DNA complex (C1) with reduced electrophoretic mobility relative to DNA incubated in the absence of protein. Incubation of UL44 with [P]-labeled ds 30mer (B) resulted mainly in the formation of two complexes (C1 and C2). With increasing amounts of UL44, the lower mobility complex, C2, predominated. EMSA analysis of binding of baculovirus-expressed, full-length UL44 to [P]-labeled ds 30mer showed a similar pattern (C). Given that UL44 forms a dimer in solution (,), that dimerization is important for DNA binding () and our ITC analysis (see next section), we interpret these results to mean that the faster- and slower-migrating complexes represent 1:1 and 2:1 stoichiometries, respectively, of dimeric UL44 to the ds 30mer. This interpretation is supported by EMSA analysis of an UL44 mutant (UL44ΔC290 L86A/L87A) that is defective in dimerization () in which faster migrating species are observed, consistent with UL44 monomer–DNA complexes (D).
To determine the binding affinity of UL44 for the ds 18mer in these EMSAs, we performed densitometric measurements of free and bound template at different UL44 concentrations. As incubation of this short template at all of the protein concentrations used to calculate affinities resulted in only one complex that migrated more slowly than the unbound template (A), the affinities measured reflect one UL44 dimer bound to one DNA molecule. The calculated apparent for the ds 18mer was 2.7 ± 0.9 nM. This value is comparable, within experimental error, with that obtained in filter-binding assays with the same DNA template ().
We also determined affinities for each of the two binding events on the ds 30mer in these EMSAs from densitometric measurements of free template and the two bound complexes at different UL44 concentrations. In this case, the first binding event had an apparent of 2.1 ± 0.8 nM, while the second had an apparent of 16.2 ± 2.1 nM. Similarly, the apparent of baculovirus-expressed, full-length UL44 was 1.7 ± 0.9 nM for the first binding event and 20.3 ± 6.2 nM for the second. Possible explanations for this difference in affinity are considered in the Discussion section.
We then used ITC to characterize quantitatively the thermodynamics of UL44 binding to dsDNA. ITC measures heat generated or absorbed upon binding (), and provides in a single experiment the values of the binding constant (), the stoichiometry and the change of enthalpy (Δ). The value then permits calculation of the change in free energy (Δ), which together the Δ allows the calculation of the entropic term Δ.
We first applied ITC to investigate the energetics of UL44 binding to a ds 18-bp oligonucleotide. Calorimetric titrations were performed in which fixed amounts of the 18mer in the syringe were sequentially injected into the sample cell containing purified UL44. The raw data (A, left panel) indicate an endothermic interaction, based on the positive values observed for the peaks. With each injection of oligonucleotide, less and less heat uptake was observed until constant values were obtained (corresponding to the heat uptake due to dilution), reflecting a saturable process. Included in A (right panel) is a corresponding blank titration in which the DNA was injected in plain buffer (values corresponding to heats of dilution). The heats of dilution of both protein and ligand (DNA) were determined and subtracted prior to analysis, and the data were integrated to generate curves in which the molar ratio of DNA to protein is plotted against the kilocalories per mole of injected DNA. The integrated heats for both titrations are shown in B and the parameters for the binding of UL44 to the ds 18-bp oligonucleotide are summarized in . The analysis shows that the UL44–DNA interaction is characterized by a = 3.6 × 10 M. This association constant corresponds to an apparent of ∼2.8 nM, a value that corresponds closely to those measured in filter-binding assays and EMSAs with the same DNA template. Moreover, the binding of UL44 to the ds 18mer was characterized by a relatively large and positive enthalpy of binding (+11.0 ± 0.9 kcal/mol, ). Very similar results were obtained when UL44 was titrated with ds oligonucleotides of identical length (18 bp) but different sequence (data not shown), confirming the non-sequence-specific nature of UL44–DNA interaction. The positive heat of formation of the UL44–DNA complex indicates that this interaction is an enthalpically unfavorable (entropically driven) process.
The stoichiometry of binding was also determined from our analysis of the calorimetry data. Analysis using a one-site model indicated a stoichiometry of 0.5 DNA mole/protein mole (that corresponds to the molar ratio of DNA to protein at the inflection point of the binding isotherm in B, left panel), i.e. one molecule of DNA was bound to two molecules of UL44 polypeptide. Since UL44 forms a dimer in solution (,) and as only one UL44–DNA complex was observed with the 18-bp oligonucleotide in EMSA experiments (A), the observed stoichiometry of 0.5 further supports the interpretation that UL44 binds this DNA template as a dimer.
To further dissect UL44–DNA binding energetics, calorimetric titrations were performed in which a ds 30-bp oligomer was titrated into UL44. A shows the heat effects of 29 subsequent injections of ds 30mer into the sample cell containing UL44 (left panel) or plain buffer (right panel). Binding of UL44 to duplex 30-bp DNA gave rise to a more complex binding isotherm with a curvature typically observed in the presence of multiple, non-equivalent binding sites (B, left panel). These additional heat effects at high levels of this DNA were observed repeatedly. Also, in reverse titrations (titrating UL44 into ds 30mer), additional heat effects were observed when DNA was almost completely saturated (data not shown). The biphasic nature of the isotherm suggested two binding sites with markedly different affinities. Indeed, the ITC data were best fit using a two-site model whose thermodynamic parameters are listed in . The first site in this model has a higher affinity ( = 4.1 × 10 M; apparent = 2.4 nM) and larger enthalpy (Δ = 11.8 kcal/mol). Both values are very similar to those measured in titrations with ds 18-bp DNA (). The second site exhibits weaker affinity ( = 1.2 × 10 M; apparent = 83 nM) and a lower enthalpy change (Δ = 4.5 kcal/mol, ). A possible explanation for these differences in affinity and enthalpy is presented in the Discussion section.
Our results from ITC assays, which showed that binding of UL44 to duplex DNA is entropically driven, suggested that binding may entail the release of bound ions. Similarly, studies of the interaction of various charged protein ligands (e.g. oligolysines, RNAse and repressor and RNA polymerase) with nucleic acids have demonstrated that these association reactions are driven by the entropic effect of release of bound cations from the nucleic acid (). In these cases, the analysis of the ion dependence of equilibrium binding constant, , has allowed investigators to measure the number of participating ions, and has provided information about the molecular details of the binding reaction. In particular, plotting the log versus the log [M], where M is a cation, permits the calculation of the number of charge–charge interactions involved in binding.
The dependence of of UL44–DNA interaction on monovalent electrolyte concentration was determined in various buffers, using either Na or K as the cation and either Cl or CHCOO as the anion. In every case, log was a linear function of log [M], as predicted by Equation () (see Materials and Methods section). A and B show the results of series of experiments in Tris buffer (1 mM, pH 7.5) using NaCl and NaCHCO, respectively, as the monovalent salt. The least squares slopes d log /d log [Na] were −3.53 ± 0.76 and −3.76 ± 0.55, respectively, implying that ∼4 ± 1 ions are released in the interaction of each monomer of UL44 with dsDNA in the salt range and buffer studied. Then it can be estimated (by dividing the least squares slopes by ψ = 0.88, the number of Na ions thermodynamically bound per phosphate in dsDNA) that a maximum of 4 ± 1 phosphate groups on the DNA are involved in ionic interactions with positively charged groups on the protein. An additional set of binding data was obtained in the same buffer but with KCl as the variable monovalent salt. Binding constants under these conditions agreed with those in A within experimental error (data not shown). Thus, we conclude that the effects of the K and Na ions are similar and that the Tris cation in the buffer has minimal effect on the interaction at the salt concentrations used in the experiments.
If the contribution of anion release to the total observed ion release in the UL44–DNA interaction is small, then , the number of phosphate groups involved in ionic interactions, corresponds to that estimated from Equation (). If there is detectable anion release, then is less than this value. A rather small anion effect on was apparent in a comparison of A and B. All conditions in these two sets of experiments were identical, except that in one case the anion was C1 and in the other CHCOO. Values of were ∼3 times larger in sodium acetate than in sodium chloride at the same Na concentration. In contrast, a 40-fold difference was observed in the non-specific interaction of repressor with DNA by replacing Cl with CHCOO (). In any case, the fact that the slopes of the lines in A and B differ by only 6% (which is well within the SE of either least-squares line) suggests that there are no major differences in the amount of anion release (if any) in the two systems.
To help remove ambiguities due to anion binding from the determination of , one can perform protein–nucleic acid binding experiments in the presence of MgX only. This will shift the range of X concentrations associated with experimentally accessible values of to substantially lower concentrations, and thereby reduce the importance of anion binding.
The dependence of log of UL44–DNA interaction on log [Mg] is shown in . The log varied with log [Mg] in a linear fashion, within experimental error, as predicted by Equation () (see Materials and Methods section) in the absence of monovalent cations and provided that the anion-binding effect is not large. From the slope of the least-squares line in (−1.93 ± 0.42), we conclude that 2 ± 1 divalent ions were released upon formation of the UL44–DNA complex in the presence of MgC1, consistent with the 4 ± 1 monovalent ions released in sodium chloride. These results are also consistent with our observation of no meaningful anion binding. In this case, the ratio of ion release terms should be 0.53 [see Equation () in Materials and Methods section]; the experimental value is ∼0.5. We can rule out the possibility of a large number of anion sites with low binding affinities. However, we cannot rule out a small number of anion sites with either low or high affinity. At present, the simplest interpretation of our data is that anion release is not a major factor in the thermodynamics of the reaction.
Numerous structural and thermodynamic studies have been carried out in recent years in order to characterize the binding of sequence-specific DNA-binding proteins. In contrast, thermodynamic studies of non-sequence-specific DNA-binding proteins have rarely been reported due to the difficulties in studying those systems (,). To our knowledge, our studies represent the first investigation of the energetics involved in the interactions of a DNA polymerase accessory protein with DNA. In this study, we used filter-binding experiments under various ionic conditions, EMSAs, and ITC to analyze the interaction of UL44 with several different short DNAs. The combined application of these three techniques sketches a picture of how the HCMV DNA polymerase accessory protein interacts with DNA, and provides some insight into its mechanism of processivity. From a more general point of view, these studies could also contribute to the understanding of the physical principles underlying non-sequence-specific DNA binding.
We will first discuss the stoichiometry of UL44 binding to DNA, the DNA length dependence of this interaction, the importance of electrostatic interactions in UL44–DNA binding and how UL44 compares with other processivity factors. Then, we will discuss how our results could shed some light on the processivity mechanisms of UL44.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Eukaryotic promoters driven by RNA polymerase II are composed of two different functional units, the upstream and core promoter sequences. Cell type-specific transcription factors bind to the upstream sequence(s) and activate transcription via interactions with the basal transcriptional machinery that forms at the core promoter sequence (). This machinery consists of a set of general transcription factors, i.e. TFIIA, TFIIB, TFIID, TFIIE, TFIIF, TFIIH and RNA polymerase II (). Among these, only TFIIB and TFIID can recognize core promoter sequence(s) in a sequence-specific manner (). In metazoans, there are several distinct core promoter sequences, including the TATA element, TFIIB recognition element (BRE and BRE), downstream core element (DCE), initiator (Inr), motif ten element (MTE), downstream promoter element (DPE) and X gene core promoter element 1 (XCPE1) (). These elements, with the exception of BRE and XCPE1, are thought to be recognized by TFIID.
Interestingly, previous studies showed that there are some functional compatibilities between the upstream and the core promoter sequences (,). In yeast, for instance, while the upstream sequence of the promoter activates the and core promoter at the same levels as its own core promoter, the upstream sequences of the latter two genes are largely unable to activate the core promoter (). Another example is the two distinct types of upstream sequences that act specifically with DPE- or TATA-dependent core promoters in embryos (). These functional compatibilities indicate that the core promoter architecture plays an important role in the transcriptional regulation of class II genes.
In this context, it is noteworthy that the utilization of the TATA element is altered during development. The shift from the utilization of a TATA-less promoter to a TATA-dependent promoter during early development () could be explained by the presence of DPE- or TATA-specific upstream sequences (). If this were the case, functional compatibilities between the upstream and the core promoter sequences would provide a potential mechanism for the developmental regulation of differential core promoter usage ().
Previously, we showed that an interplay between the upstream CACGTG element and the TATA element may determine the temporally specific expression profiles of the (Hp) OtxE and OtxL promoters during early development of the sea urchin (). Furthermore, the CACGTG element activates the HpOtxE core promoter specifically from a distal position, whereas it activates the (Sp) Spec2a core promoter from a proximal position in undifferentiated P19 embryonal carcinoma cells, which were derived from a mouse embryo (). Intriguingly, these stimulatory effects of the CACGTG element disappear upon treatment with retinoic acid, which triggers the neuronal differentiation of P19 cells (). These observations led us to believe that the HpOtxE and SpSpec2a promoters could provide an adequate model system to further analyze functional interactions between the upstream and core promoter sequences and their spatiotemporal regulation during early development of sea urchin embryos.
In this study, we examined the expression of two different test promoters containing upstream and core promoter sequences derived from HpOtxE or SpSpec2a in living sea urchin embryos. The promoters drove the expression of CFP and YFP reporter genes, which enabled us to monitor the activity of the two promoters concurrently and semi-continuously in single embryos, to verify the functional compatibility between the upstream and core promoter sequences, and to determine whether the interactions between the upstream and core promoter sequences are dynamically regulated during the early development of sea urchin embryos. The results reinforce the recent view that the combination of the upstream and core promoter sequences may determine the specific spatiotemporal expression profiles of a range of genes that are developmentally regulated.
Gametes of the sea urchin were collected and inseminated as described previously (). Fertilized eggs in which reporter constructs had been introduced (as described below) were cultured at 16°C in a petri dish (6 cm in diameter) at a cell concentration of 1.3% [v/v; 0.2 ml of eggs suspended in 15 ml of artificial sea water (ASW)].
The I-I DNA fragment encoding GFP was amplified by PCR using the primer pair TK7264 and TK7265 and the pGreen Lantern-2 (Invitrogen) plasmid as a template. The PCR product was inserted into the RV site of the pBluescript II SK(-) plasmid (Stratagene) to generate pM4642. The oligonucleotides used in this study are summarized in Supplementary Table 1. The I and HI sites within the open reading frame (ORF) of GFP encoded by pM4642 were disrupted by site-specific mutagenesis () using TK7262 and TK7289 oligonucleotides, respectively, to generate pM4645.
Site-specific mutagenesis was conducted on pM4645 using the oligonucleotides TK7488 (T65G, V68L, S72A) and TK7489 (T203Y) to generate pM4682, encoding YFP. Similarly, site-specific mutagenesis was conducted on pM4645 using the oligonucleotides TK7490 (F64L, Y66W), TK7491 (N146I, M153T) and TK7492 (V163A) to generate pM4684, encoding CFP.
Site-specific mutagenesis was conducted on pM1249 () using oligonucleotide TK867 to generate pM1319, a construct with a truncated HpOtxE promoter fragment (base pairs −461 to +180, numbered with respect to the transcriptional initiation site as +1). The I site was created at the initiating codon of the luciferase ORF of pM1319, and the I site located upstream of the HpOtxE promoter was disrupted by site-specific mutagenesis using the TK6391 and TK6448 oligonucleotides, respectively, to generate pM4497. The I-I short DNA fragment containing the HI site of pM4497 was removed to generate pM7448 by enzyme digestion, followed by blunt end ligation.
The I-I DNA fragment of pM7448 encoding luciferase was replaced with the I-I DNA fragments of pM4682 and pM4684 encoding YFP or CFP to generate pM4692 and pM4705, respectively.
The TATTCA sequence at base pairs −28 to −23 in the HpOtxE promoter of pM4692/pM4705 was changed to TATAAA or TATAAAA by site-specific mutagenesis using the TK1150 or TK8923 oligonucleotides to generate two sets of YFP/CFP plasmids, pM4706/pM4694 and pM7071/pM7070, respectively.
pM4731 (YFP) and pM4725 (CFP) were generated from pM4692 (YFP) and pM4705 (CFP), respectively, by site-specific mutagenesis using the oligonucleotides AK312, AK313 and TK7598.
The AATAATA sequence at base pairs −30 to −24 in the SpSpec2a promoter of pM4731 and pM4725 was deleted by site-specific mutagenesis using the TK1488 oligonucleotide to generate pM4734 and pM4728, respectively. Similarly, the GAATAATAC sequence at base pairs −31 to −23 in the SpSpec2a promoter of pM4731 was changed to CTATAAAAG by site-specific mutagenesis using the TK1681 oligonucleotide to generate pM7073.
#text
The putative GATA-binding site (#1) at base pairs −360 to −355 in the HpOtxE promoter of pM4692/pM4705, pM4706/pM4694 and pM4731/pM4725 was changed to a III site (AAGCTT) by site-specific mutagenesis using the oligonucleotide TK7989 to generate pM4938/pM4937, pM7065/pM7062 and pM4944/pM4943, respectively.
The putative Otx-binding site at base pairs −294 to −289 in the HpOtxE promoter of pM4692/pM4705, pM4706/pM4694 and pM4731/pM4725 was changed to a III site (AAGCTT) by site-specific mutagenesis using the oligonucleotide TK7836 to generate pM4940/pM4939, pM7066/pM7063 and pM4946/pM4945, respectively.
The putative GATA-binding site (#2) at base pairs −250 to −245 in the HpOtxE promoter of pM4692/pM4705, pM4706/pM4694 and pM4731/pM4725 was changed to a III site (AAGCTT) by site-specific mutagenesis using the oligonucleotide TK7837 to generate pM4942/pM4941, pM7067/pM7064 and pM4948/pM4947, respectively.
The E-box at base pairs −66 to −61 in the HpOtxE promoter of pM4692/pM4705, pM4706/pM4694 and pM4731/pM4725 was changed to a III site (AAGCTT) by site-specific mutagenesis using the oligonucleotide TK1812 to generate pM4783/pM4782, pM4781/pM4780 and pM4789/pM4788, respectively.
xref
#text
Introduction of reporter plasmids into fertilized eggs was conducted as previously described () with slight modifications. The YFP/CFP reporter plasmids (2.5 μg each) were linearized with HI and mixed with 10 μg of 3A-digested calf thymus carrier DNA and 0.6 μg of insulator DNA (578 bp) excised from pM4932 by HI. Five milligrams of gold particles (1.5–3.0 μm in diameter, Sigma-Aldrich, USA) was coated with the mixture of DNA described above (total 15.6 μg) by co-precipitation in 7.5% polyethylene glycol 6000 and 0.94 M NaCl. After washes with ethanol, an ethanol suspension of 3.3 mg DNA-coated gold particles (∼20 μl) was placed on a projectile holder. Bombardment using a particle gun gene delivery system (GIE-III IDERA, Tanaka Co. Ltd) was carried out on ∼2 × 10 fertilized eggs. Bombarded embryos were cultured at 16°C in a petri dish (6 cm in diameter) of ASW containing penicillin (100 mg/l) and streptomycin (50 mg/l).
Noble agar (0.45%; Difco, USA) in ASW was kept warm at 70°C and then cooled to 37°C just before use. Embryos were gently mixed with this agar solution and then swiftly transferred onto a chilled slide glass to solidify the agar (final concentration of ∼0.4%). The slide glasses with immobilized embryos were immersed in ASW containing penicillin (100 mg/l) and streptomycin (50 mg/l) and then cultured at 16°C. This method enabled us to monitor gene expression semicontinuously in the same embryos by taking photographs every few hours.
To observe the expression of YFP/CFP reporter genes, the movement of embryos was stopped by adding formaldehyde to a final concentration of 0.1% when the embryos were not immobilized with agar. An aliquot of the embryo suspension containing 200–300 embryos was placed onto a slide glass, and the YFP- and/or CFP-expressing embryos were visualized using a Nikon Eclipse 80i microscope with an epifluorescence attachment. Embryos immobilized with agar were inspected using the same system.
We chose the HpOtxE promoter as a model for studying functional interactions between upstream and core promoter sequences during early stages of development. The zygotic expression of HpOtxE begins at the unhatched blastula stage and gradually decreases by the gastrula stage (). In contrast, the expression of SpSpec2a, a well-characterized target gene of SpOtx, begins at the late cleavage stage, exclusively in the aboral ectoderm territory, and is then largely extinguished after the late blastula stage (). Fortuitously, the HpOtxE and SpSpec2a promoters contain the CACGTG element, a putative binding site for USF, at the same position (−66 to −61 bp) from the transcriptional start site (+1) (,). Thus, to investigate core promoter specificity, the region surrounding the transcriptional start site (−60 to +23 bp) of the HpOtxE promoter could be readily replaced with the corresponding region (−60 to +23 bp) of the SpSpec2a promoter without affecting the distance between the upstream and core promoter sequences ().
To compare the spatiotemporal expression patterns of two distinct test promoters in one embryo, two plasmids were introduced concurrently into fertilized eggs by particle gun bombardment. One plasmid carried a promoter connected to a CFP reporter gene, and the second plasmid carried a second promoter connected to YFP. Transfected embryos were then examined at different developmental stages for the expression of these two reporters by fluorescence microscopy and/or quantitative RT-PCR. The expression patterns were classified into four major groups (profiles |
Transcriptome analysis can be performed in a knowledge-based manner [‘closed systems’ such as microarrays (,)] or independent of any assumptions [‘open systems’, ()] as reviewed in (). The advantage of the latter for less well-analysed organisms is obvious. But even when investigating the human transcriptome, analysis of a number of important processes such as alternative splicing, polyadenylation and the generation of antisense transcripts requires technologies that utilize only a few assumptions regarding the transcriptome's composition. Current ‘open systems’ comprise fragment display, tag sequencing and subtractive hybridization (). Fragment display technologies (,,,) rely on the generation of cDNA fragments either by arbitrary priming or by the use of restriction enzymes. Fragments are labelled and subjected to size separation by gel electrophoresis, and corresponding fragment patterns (‘fingerprints’) from different biological sources are compared. Differentially expressed genes are represented and, thus, identified by fragments differing in abundance between samples. These technologies are particularly appealing given their technical simplicity, low costs for primary analysis involving the comparison of fingerprint patterns obtained from different samples and robustness. Nonetheless, secondary analysis for the assignment of corresponding genes to displayed signals indicating differential expression remains a major bottleneck. The traditional approach of physical isolation followed by reamplification and sequencing () is cumbersome and error-prone. Identification of fragments by their mobility and thus their predicted physical length (,) is of little use for unknown transcripts and has proved unreliable since electrophoretic mobility is strongly influenced by a fragment's base composition (), rendering length predictions inaccurate. Very much like fragment display technologies, tag-sequencing approaches rely on the comparative quantitation of cDNA fragments. However, in this case the quantitation is achieved by determining the frequency of occurrence of a given sequence tag (sometimes called a ‘signature’) within a large population of sequenced tags. Sequencing of the tags can occur in a serial manner [SAGE, ()] or in a massively parallel fashion [MPSS, ()]. Tag sequences are then used to identify the corresponding transcripts by database searches.
could be regarded as a hybrid technology which combines elements of both fragment display and tag sequencing. Our approach allows for the identification of displayed fragments by assigning to each an 18-bp sequence tag as a gene identifier. The parallel architecture of avoids the massive oversampling of abundant transcripts typical for tag sequencing methods, thus overcoming the major costs of SAGE and its variants ().
A pair of near-isogenic lines of barley, cv Ingrid and cv Ingrid BC7 5, were grown in pots of compost soil (from the IPK nursery) in a greenhouse with automatic shading and supplementary light (sodium-halogen lamps) with a light period of 16 h. Temperature ranged from 18°C at night to 21°C during the day. Seven-day-old seedlings were used for all inoculation experiments. (DC.) E.O. Speer f.sp. hordei, strain 4.8 carrying AvrMla9, was cultivated by weekly inoculation of 7-day-old seedlings of barley cv Golden Promise. Seven days post-inoculation, conidia were used for inoculating test plants by shaking inoculated plants over test plants in a settling tower of ∼60 × 60 × 60 cm in dimension. A split-plot design was used for the simultaneous inoculation of both barley lines. Control and inoculated plants were incubated at a constant temperature of 20°C and exposed to natural daylight until RNA extraction.
The abaxial epidermis of primary leaves was stripped 12 h post-inoculation and immediately frozen in liquid nitrogen. Care was taken to strip all four samples [two genotypes each with two treatments (control and inoculated)] concurrently to prevent artefacts from genes under circadian regulation. Ground material was suspended in a hot (80°C) 1:1 mixture of phenol and extraction buffer (100 mM LiCl, 10 mM EDTA, 1% SDS, 100 mM Tris pH 9.0). After addition of 0.5 vol. chloroform, samples were mixed for 30 min. at room temperature. Extraction was repeated and then samples were adjusted to 2 M LiCl and precipitated. Digesting RNA preparations with DNase I proved to be not required and thus was omitted from our protocol.
cDNA synthesis, fragmentation and amplification were performed according to () with the following exceptions: fragmentation was achieved with 20 U of AluI (New England Biolabs); and ligation of blunt-ended linker DAL4138 (upper strand: 5′-TGGATAGAGCAGTGGTAGCACACGTGAGCGATGACTATGAG-3′, lower strand: 5′-CTCATAGTCATCGCTCACGTGTCGGCACATCTCATATA-3′) was performed with 1 U of T4 DNA Quick ligase (New England Biolabs) in the reaction buffer. PCR was executed with 1 U of polymerase (Invitrogen) pre-incubated with TaqStart antibody (Clontech) in a total volume of 20 μl using microtitre plates and PE 9700 thermocyclers (Perkin-Elmer). All four extension bases were positioned on the fragments’ 3′-end. For example, subpool ‘GCAT’ was generated using the primer CP28GCAT (5′-ACCTACGTGCAGATTTTTTTTTTTTTTTGCAT-3′). To visualize amplification products, linker primers were labelled at their 5′-end with a fluorescent FAM group. Reactions were mixed with GeneScan 500 length standard (Applied Biosystems) and analysed on an ABI 3100 capillary sequencing apparatus (Applied Biosystems). Separation of cDNA fragments was performed according to the manufacturer's recommendations. Trace file peaks were recognized by the DAX software package (van Mierlo Software, Eindhoven, NL) using a 4-fold local baseline as a general threshold. The Java-based inhouse software imported the detected signals as unique addresses into a Sybase database. A native Java driver allowed for a higher performance than the usual JDBC/ODBC Bridge. After calibration of each fluorogram via a third-degree polynomial function, an automated QC algorithm checked data integrity. Passed fluorograms were then aligned following an algorithm which processed peaks according to the calibrated fragment lengths, peak patterns and normalized signal intensities calculated on the fly. Alignment was accomplished in three stages: (i) display alignment, (ii) indexing alignment and then (iii) merging of aforementioned. Each match (comprising all display and indexing peaks of the same size, from the same subpool, and with assignment to each other) was given a unique match ID and comprised expression data of one particular transcript across all analysed samples as well as the corresponding sequencing data, i.e. the respective signature. Data were then visualized in a signature list linked to a fluorogram view, comprising output data of a BLAST search against the HarvEST database (). Final data analysis and identification of gene regulation events was effected with GeneSpring software (Silicon Genetics). Copies of the software are available upon request at no charge from A.F.
An overview of the workflow and of the interrelation of display reactions and sequencing reactions can be taken from and . Equal volumes of display reactions of all four conditions (/5 and infected/non-infected) each were pooled subpool-wise (i.e. pooling involved reactions from corresponding subpools out of the total 196 subpools) and cleaved with AluI to remove the adaptor sequence. Reactions were split into seven aliquots and subjected to IIs adaptor ligation with each of seven different adaptors per reaction. Each of the adaptors contained a recognition site for restriction endonuclease BpmI at a different position from the ligatable end Adapter ligation was conducted for 1 h at 20°C, using 1 U of T4 DNA ligase (Roche) in a total volume of 20 μl. IIs adaptors were released using 5 U BpmI (New England Biolabs) in a reaction volume of 50 μl for 1 h at 37°C. Reactions were heat-inactivated for 20 min at 65°C. Each digest was split into two aliquots and subjected to ligation of a sequencing adaptor with the first aliquot to be interrogated for each fragment's first (‘inner’) overhang base and the second aliquot interrogated for the overhang's second (‘outer’) base. Ligation of sequencing adaptors also took place for 1 h at 37°C using 1 U of T4 DNA ligase in a volume of 20 μl. Unligated adaptors were removed using Montage PCR plates (Millipore). A 2 μl amount of each sequencing reaction was mixed with 1-μl GeneScan 500 size standard and 14-μl HiDi (Applied Biosystems) and subjected to capillary electrophoresis on an ABI 3100 sequencer. Trace files were handled as described above, except that the detected signals were subjected to an automated base-calling step. Corresponding fluorograms (representing bases 5–18 of the displayed fragments) were aligned and, by adding the already known 4-bp sequence of restriction endonuclease AluI, used to assemble 18-bp signatures of the displayed fragments.
Gene identity was deduced from 3′ signatures with a method that relied on BLAST as a search engine. The procedure avoided BLAST's limitations in searching short fragments but still allowed handling of ambiguities in the signature or mismatches between signature and database sequences. The procedure first expanded all ambiguous nucleotides and thus created a set of non-ambiguous deduced signatures. A BLASTN search against the selected database was then performed with default parameters as well as ‘−F F’, ‘-q −1’ and ‘−W 7’. The first parameter guarantees that low complexity fragments within the signatures do not interfere with the search while the second allows the recognition of ‘close to signature end’ mismatches and the third ensures that sufficient consecutive nucleotides match despite a mismatch in the middle region of the signature. Since mismatches in the first or last bases are not reported by BLASTN, the results here were post-processed by extending the alignment over the length of the signature by adding matches corresponding to the BLASTN alignment. The matching database entries were then sorted according to hit- and database-entry-quality. While we did not allow for gaps to take place in the alignment, up to two mismatches per signature were allowed to compensate for sequence polymorphisms as well as for sequencing errors.
For each fragment to be reamplified, a PCR primer was synthesized corresponding to the respective 18-bp signature sequence. Using 1 µl of a 1:30 dilution of the respective display reaction as a template, 50 µl reactions were assembled containing 10 µM each of fragment specific primer and subpooling primer, 1.5 mM MgCl, and 2.5 U polymerase in 1× PCR buffer (Invitrogen). Amplification was effected over 25 cycles with duration 30 s at 94°C, 30 s at 60°C and 1 min at 72°C for each cycle.
Quantitative RT–PCR was performed using the Roche LightCycler according to the manufacturer's recommendations.
All data were derived from three independent inoculation experiments. Pathogenesis-related genes were selected based on two criteria: (i) up- or down-regulation by at least 3-fold which was based on average signal intensities from control and inoculated samples; and (ii) a statistically significant difference between control and inoculated samples with < 0.05 (student's -test). Resistance- or susceptibility-related genes were respectively selected utilizing the two criteria described above, as well as a difference in the signal intensities (by a factor of at least five) between inoculated barley lines in the presence or absence of the 5 resistance gene.
Comparison of biological samples in employs a fragment display procedure similar to the already published RMDD protocol (). RNA is first converted to double-stranded cDNA which is cut using a frequently cutting restriction endonuclease, AluI. So-called linkers (double-stranded oligonucleotides with one blunt end, which allows unambiguous orientation of the linkers upon ligation) are ligated to the fragment ends, and cDNA 3′-fragments are amplified by PCR. Amplification is achieved by use of PCR primers capable of binding to a sequence corresponding to either one of the linker strands or to the reverse complement of the oligo(dT) cDNA primer. Since linkers are not phosphorylated, only one of the two linker strands is covalently attached to the cDNA fragments. Thus, amplification of a fragment takes place only when, upon denturation, the potential primer binding site remains attached to the respective strand. This is the case only for the reverse complement of the oligo(dT) cDNA primer, thus allowing selective amplification of cDNA 3′-fragments, while internal fragments remain unamplified. In order to reduce complexity, PCR primers carrying additional ‘selective’ 3′-bases are employed. The rationale behind this step is to use primers which consist of (i) a ‘universal’ sequence which is capable of hybridizing to the cDNA primer sequence incorporated into any cDNA 3′ fragment plus (ii) one or more ‘selective’ bases at the primers’ 5′-ends which allow for primer extension only when the corresponding base or bases on the template strand is/are complementary to the selective base(s) of the primer. Thus, subdivision of all cDNA fragments to be amplified into 3 × 4 × 4 × 4 = 192 different reduced-complexity subpools is achieved, corresponding to two 96-well plates per sample. Use of fluorescently labelled primers allows reactions to be ‘displayed’ on automated sequencers (). Thus, each RNA preparation is represented by a set of 192 fluorograms providing a fingerprint of the respective transcriptome. After normalization, comparison of the signal intensities (‘peak heights’) between corresponding reactions allows for straightforward detection of gene regulation events (a). Peak calling, normalization, cross-sample alignment of corresponding signals and calculation of relative expression levels are performed automatically. With the use of current automatic sequencers, expression profiles can be generated with high throughput. A single 96 capillary sequencer allows for analysis of more than 300 biological samples per month. Spiking experiments, employing transcribed, oligo-adenylated RNAs as ‘pseudo-mRNAs’, demonstrated a sensitivity sufficient to detect transcripts at a relative abundance of 1:100 000, which is in line with published fragment display technologies (,).
For gene identification, the basic idea was to implement a so-called ‘orthogonal sequencing approach’ for generating cDNA-specific sequence tags. While, in standard sequencing, fluorescence signals encoding for the identity of a template's consecutive bases are obtained from one capillary and independent templates are analysed by electrophoresis in independent capillaries, 's orthogonal sequencing provides sequence information for one particular cDNA fragment using one capillary per base, while many fragments are analysed simultaneously in the same set of capillaries. For example, determination of 14 bases each for a set of 100 fragments different in size requires no more than 14 capillaries, the first capillary providing the signals encoding for the identity of each of the fragments’ first base, the second capillary disclosing the identity of each of the 100 fragments’ second base, and so forth ( and b). Thus, takes advantage of the power and throughput of the highly sophisticated latest-generation automatic capillary sequencers, without the requirement of any specialized instruments. Employing this orthogonal sequencing approach, each displayed fragment (representing an expressed gene) is assigned an 18-bp long sequence tag or ‘signature’. The first four signature bases are defined by the restriction enzyme used for fragment generation, while the remaining 14 bases are determined by orthogonal sequencing of fragment mixtures.
In practice, each of the 192 display reactions is split into seven identical aliquots (). For each case, one of seven different adaptors carrying a recognition site for a type IIs enzyme is then attached. As IIs enzyme, BpmI was chosen which has the cutting characteristics CTGGAG(N)16/14 (i.e. cutting the ‘upper’ strand 16 bases away and the ‘lower’ strand 14 bases away of its recognition site, CTGGAG, and thereby generating 2-base overhangs). After cutting with BpmI, a defined portion of each fragment (the exact position of which depends on the position of the BpmI recognition site within the attached adaptor) is converted into a 2-base overhang amenable to base identification. Successive adaptor's BpmI recognition sites are displaced by two bases each: The position of the BpmI site within the set of seven adaptors is 13, 11, 9, 7, 5, 3 or 1 from the adaptors’ ligatable ends. Thus, digest of the fragments linked to the ‘first’ IIs-adaptor exposes all the fragments’ ‘first two bases’ as an overhang; digest of the same fragments linked to the ‘second’ IIs-adaptor exposes ‘bases three and four’ of each fragment, and so forth (Figure 3a, Supplementary Data). Identification of the overhanging bases is achieved by sequence-specific ligation of fluorescently labelled ‘sequencing adapters’. A sequencing adapter is characterized by a single-stranded overhang of defined length and sequence, and a fluorescence label specific for a particular overhang sequence. In this case, overhangs are two bases long, which corresponds to the length of the fragments’ overhangs having been generated by the BpmI digest. Sequencing adaptors have the structure Fluo-Core-NX or Fluo-Core-XN, wherein ‘Core’ refers to a double-stranded core sequence common to all adaptors; ‘N’ to equimolar mixture of A, C, G and T; ‘X’ to a sequencing nucleotide (either A, C, G or T); and ‘Fluo’ to fluorophore encoding the sequencing nucleotide. Upon ligation, each fragment is linked to one of four fluorophores, each in turn serving as a base identifier for a given base position within the overhang, which depends on the set of sequencing adaptors employed (Figure 3b, Supplementary Data). Capillary electrophoresis then allows for identification of one base, each at a predetermined fixed position, for all fragments simultaneously (b). Performance of 14 sequencing reactions for a given display reaction (seven digests and two bases per overhang) followed by alignment of corresponding fluorograms and execution of a base-calling step provides a contiguous stretch of sequence information for any cDNA fragment present. With the addition of the known first four bases, an 18-bp signature for each expressed gene can be determined. Performing all enzymatic steps in 96-well microtitre plates allows for automation of the complete indexing protocol by use of liquid handling stations. The determination of all signatures detectable within a given transcriptome requires 28 96-well plates ().
Sixty percent of the barley signatures identified in our study could be assigned to a corresponding HarvEST database entry. To allow for gene identification of the remaining signatures, the full sequence of the corresponding fragments can easily be obtained by using the signature for oligonucleotide primer design and selectively amplifying the corresponding fragment. This approach was tested with 23 randomly chosen fragments which yielded no hit with a signature BLAST search against the HarvEST database (). In 11 cases (48%), a corresponding database entry could be found. In the remaining 12 cases, a PCR product and a clearly readable sequence were obtained, but no BLAST hit found in HarvEST (data not shown).
Five technical replicates were prepared to examine the reproducibility of . From these reproducibility data the confidence of measured regulation factors, depending on the signal intensity and number of replicates, was estimated. For example, signals of average intensity yielded a lower limit for reliably detectable regulation events ( = 0.001) near 1.5-fold if no replicates were utilized ().
was applied in the analysis of barley epidermal cells infected by barley powdery mildew (Blumeria graminis hordei, ). The study included a comparison of two near-isogenic barley lines differing in the absence or presence of the 5 resistance gene, one of several recessive loss-of-function alleles of the gene which mediates durable and race-non-specific resistance to (). Plant material from three independent experiments was included to assess biological variability. Since the primary events of pathogen attack and plant response take place in the epidermis of affected leaves, RNA was extracted from such stripped tissue of treated and untreated leaves.
A large number of genes were detected as either up- or down-regulated upon infection. Response of both lines to infection was similar. In mutant plants, 291 signatures (235 signatures in wild-type plants) indicating genes up- or down-regulated at least 3-fold upon infection were identified, 189 of which (145 in wild-type plants) were detected in the HarvEST database (Tables 2 and 3, Supplementary Data). The lower number of differentially expressed genes in the susceptible line mostly resulted from less robust expression patterns with subsequent removal of several genes by data filtration. The group of genes strongly regulated in resistant plants was further characterized by determining the percentage of genes that were similar to barley expressed sequence tags (ESTs) encoding for known pathogen–response (PR) proteins (Table 4, Supplementary Data). A major fraction of ESTs corresponded to novel defence-related genes not previously described (Table 5, Supplementary Data).
In order to verify results by an independent method, 14 transcripts indicated as up-regulated were tested by real-time PCR in all four experimental conditions (mutant and wild type, each coupled with infected and non-infected). Although for some transcripts regulation factors as determined by deviated more than 2-fold from RT–PCR factors, up-regulation could be clearly confirmed in all cases (Table 6, Supplementary Data). This finding is not surprising since regulation factors determined with different methods (e.g. northern blot versus microarray versus RT-PCR data) frequently deviate from each other, which appears to reflect sequence-specific bias inherent to the particular method employed (). A recent study shows that even within one given microarray platform, the choice of the data processing method has a non-neglectable influence on the gene regulation data obtained ().
Comparing inoculated and resistant plants with those that were susceptible revealed six and 10 marker genes of susceptibility and -mediated resistance that differed in transcript abundance by a factor of at least four (Table 7, Supplementary Data). The transcript abundance of some resistance-related and pathogen-regulated marker genes with the strongest genotype-dependent differences is shown in .
In contrast to SAGE and MPSS, signature sequencing is required only once per species and tissue type, rendering this approach much less expensive than other tag-sequencing technologies. Even if hundreds of different biological samples are to be investigated, the identity of expressed genes can be determined by one single indexing experiment. Thus the deposition of signature catalogues of a specific biological material in a central database would allow scientists to generate their own sets of display data and interpret them on the basis of publicly available indexing data.
One of the major technical hurdles during development turned out to be the extremely high dynamic range of gene expression, i.e. the difference in copy numbers of high and low abundance transcripts. We estimate this dynamic range to be approximately several thousand to one, which clearly exceeds the dynamic range of commercially available sequencing machines. It turned out that a very simple way of solving this problem is to overload the capillaries in a way that the few fragments representing the most abundant transcripts fall out of the detector's linear range. The respective signals get clipped and are excluded from quantitative analysis. Under these conditions, signals from low abundance genes are strong enough to allow for reliable quantitation. It is interesting to note in this context that the sensitivity of regarding low abundance transcripts appears to be higher than the sensitivity of MPSS, since MPSS signature collections are seriously dominated by a small number of very high abundance signatures. On the other hand, low abundance signatures with an abundance of <100 copies per million turned out to be very unreliable due to sequencing errors (our unpublished data). In the past, the only way to improve the sensitivity of MPSS was to collect data from many runs which caused prohibitive costs. It actually was the astronomically high reagent costs and poor data quality of MPSS that triggered development of in our group.
While a high dynamic range still can be easily managed on the level of the display reactions, it is more difficult to cope with when it comes to signature identification of low abundance transcripts. For unknown reasons, a certain reproducible fluctuation of signal intensities is observed upon performing signature sequencing. In other words, there is no strict linearity between signal intensity of a display reaction and signal intensity of the corresponding sequencing reactions. While this does not impair correct signature identification of medium and high abundance transcripts, signature bases of very low abundance transcripts sometimes ‘drop out’, rendering correctly calling the affected signature base(s) impossible. Thus, future modifications of will have to address signature sequencing of low abundance transcripts. One possibility to achieve an improved sequencing of low abundance transcripts might be to subject fragment mixtures to a normalization step before entering the sequencing branch of the protocol.
An important question is the usefulness of our signatures for correctly identifying the corresponding expressed gene. It has been shown that a signature length of 16–17 bp is sufficient for unambiguous gene identification in ∼92% of human genes (), rendering the approach suitable for analysis of any complex eukaryotic organism. However, this does not exclude ambiguity of a certain fraction of signatures, and this fraction may differ from organism to organism. While EST libraries proved valuable for identification of (differentially) expressed genes, other sorts of databases such as, e.g. genomic libraries might be useful as well. Even without any sequence information at all available of the species under investigation, would allow identification of expressed genes by taking the approach described above of reamplifying and sequencing cDNA fragments of interest by use of gene-specific signature primers. The sequences obtained by this route could be subjected to a ‘heterologous BLAST’ search in sequence databases of related species. It might turn out, however, that the lesser degree of inter-species sequence conservation within cDNA 3′-ends, as compared to the degree of conservation within open reading frames, would necessitate the adaptation of to ‘internal’ cDNA fragments with a higher likelihood of containing coding information. Also, due to the effort required for reamplification and sequencing of many individual fragments, such an analysis could not be called a high throughput approach any more.
Also, any expression profiling technology has to address the issue of comprehensiveness. Theoretically, the strength of open systems is their inherent comprehensiveness. In practice, several factors may compromise perfect comprehensiveness of a given technology. For example, SAGE experiments are subject to cost constraints, which in practice reduces the number of detectable genes. The fact that, by nature of SAGE's experimental design, abundant sequence tags have to be sequenced repeatedly when rare tags are to be identified as well, forces the experimentator to define a balance of cost and sensitivity. In a typical SAGE experiment, 10 000–30 000 tags are sequenced, which seriously limits the coverage of low abundance transcripts. While repeated sequencing of the more abundant tags is avoided by our orthogonal sequencing approach, which dramatically reduces the cost per tag, shares a restriction step with all other signature sequencing technologies as well as with the more recent, restriction-based fragment display technologies. This means that those transcripts will escape detection which lack the corresponding restriction enzyme recognition site. We chose AluI as restriction enzyme for cDNA fragmentation since simulations unveiled that, when applied to barley cDNA, ∼85% of cDNAs result in a cDNA fragment in the ‘displayable’ size range between 60 and 750 bp. Coverage could be even increased to ∼97% by repeating the analysis on the basis of another frequent cutter, however, at the cost of doubling the effort for each sample. It should be pointed out that comprehensiveness of is virtually not limited by the size of this technology's fragment space, i.e. the number of different fragments which can be independently displayed. In this context, it has to be kept in mind that, in capillary electrophoresis, fragments are not separated by size , but by mobility. While fragment size is quantized, mobility is a continuum, depending on factors such as a fragment's charge, size, A/T content, secondary structure and others. Our experiments indicated that effective resolution of state-of-the-art capillary sequencers is ∼0.3 bp. In other words, two fragments differing in apparent size (calculated, with the help of an internal size standard, from the measured mobility) by 0.3 bp can be reliably and reproducibly separated and distinguished from each other. Thus, analysing fragments in the apparent size range from 60 to 750 bp provides a fragment space of (750 − 60) × (1/0.3) × 192 = 441 600 ‘theoretical bins’. Even when fragment size distribution is not homogenous, but rather resembles a bell-shaped distribution, the number of theoretical bins greatly exceeds the number of different mRNA species expected to occur in a given cell type [according to () ∼15 000 species]. Accordingly, the overlap of signals representing two similarly sized fragments turned out to be an extremely rare event.
Employing for the analysis of powdery mildew-infected barley plants, we could identify several hundred signatures indicating transcripts up- or down-regulated upon infection. Interestingly, the BLAST hit-rate of these signatures was higher among down-regulated genes (75%) than up-regulated genes (58%). This trend was confirmed when analysing strongly (>10-fold) regulated genes in 5 plants (Table 4, Supplementary Data). It is suggested that, in addition to a number of fungal genes covered by the presented indexing approach, many up-regulated genes escaped detection in EST collections due to a combination of epidermis-specific and pathogenesis-specific expression patterns. The proportion of epidermal RNA in a whole-leaf preparation is estimated at no more than 5%, causing a significant dilution of transcripts involved in the primary response to pathogen attack in preparations typically utilized for leaf library construction.
Remarkably, comparison of regulation patterns of wild-type and mutant plants suggests that several of these transcripts might be qualitative markers for resistance, in contrast to a recent cDNA array-based study on -mediated resistance where only quantitative differences were found (). A possible explanation for this is that the array (complexity ∼3200 unigenes, which is only a small fraction of the barley transcriptome) represents a closed resource, thereby minimizing the chance to identify rare, qualitative resistance markers. Cloning the marker genes for resistance coupled with functional analysis by reverse genetics () should elucidate their function in the response of epidermal cells to . In addition, a group of genes could be identified that show different expression behaviours between barley lines independent of inoculation (Tables 2 and 3, Supplementary Data), allowing a general study of the effect of the gene on cell physiology.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
In the higher eukaryotes frequently more than 90% of the DNA does not code for functional proteins or RNA. Much of this DNA has originated from the action of mobile genetic elements, mostly retrotransposons that propagate in a copy-and-paste mechanism via an RNA intermediate. While these elements can be viewed as molecular parasites that are in an evolutionary race with their host genome, they can also be regarded as essential genomic components for slowly reproducing species to adapt to a changing environment. They generate allelic heterogeneity and create new possibilities for genetic recombination, increasing genomic fluidity ().
Mobile genetic elements integrate into new genomic locations in two fundamentally different ways. DNA transposons and retrotransposons with long terminal repeats (LTR retrotransposons) use a transposase/integrase to insert a double-stranded DNA copy of the element at the target site. In this case, no DNA synthesis takes place at the site of integration. In contrast, non-LTR retrotransposons use a mechanism called target-primed reverse transcription (). This process is initiated by a targeting endonuclease, which specifically binds to the site of genomic integration. It nicks one strand of the DNA and creates a free 3′ hydroxyl end, which is then used as a primer for reverse transcription of the retrotransposon RNA at the site of integration. Endonuclease and reverse transcriptase are two domains of a single retrotransposon-encoded protein. They are thought to rely on the assistance of ‘host’-encoded proteins to complete the integration process ().
Most non-LTR retrotransposons are APE-type non-LTR retrotransposons (). Their targeting endonuclease belongs to a family of metal-dependent phosphohydrolases that includes nucleases like DNaseI (PDB-ID: 1dnk), APE1 (PDB-ID: 1dew), Exo III (PDB-ID: 1ako) and CdtB (PDB-ID: 1sr4) but also sugar phosphatases like I5PP (PDB-ID: 1i9z) and phospholipases like SmcL (PDB-ID: 1zwx) and Bc-SMase (PDB-ID: 2ddt). Members of this family share the same protein scaffold and the same catalytic residues, but a variation of the connecting surface loops has allowed them to develop quite diverse substrate specificities ().
Under the pressure to survive in their respective host species non-LTR retrotransposons have evolved different strategies (). Stringent elements like R1Bm from () and Tx1L from () encode highly specific targeting endonucleases (,). They integrate into unique genomic locations (a specific sequence within 28S rDNA for R1Bm or within the apparent DNA transposon Tx1D for Tx1L) where they do very little or no damage to the host. Promiscuous elements like the human LINE-1 (L1) element () may integrate into several hundred thousand genomic locations. They have a rather short integration-site consensus [5′-TTTT/AA-3′ for L1 ()] that is nicked by the respective targeting endonuclease (,). The host limits the spread of such elements by transcriptional and post-transcriptional silencing mechanisms that reduce activity to tolerable levels ().
Clearly, the respective endonucleases play a major role in target site selection (,,,). The intriguing question of how different targeting endonucleases recognize the DNA substrate and how easily new specificities can arise in the course of evolution remains open. There are indications that retrotransposons can evolve back and forth between a stringent and a promiscuous mode-of-action () and the ability to manipulate and design target specificity would be a crucial step in converting non-LTR retrotransposons into a genetic tool.
Previously, we described the crystal structure of the human L1 endonuclease (L1-EN) (). Based on structure comparisons and sequence alignments we suggested that the prominent βB6–βB5 hairpin loop may insert into the DNA minor groove and may be particularly important for recognizing the DNA target. Here, we combine a mutational approach (specific point mutants and entire loop grafts) with structural and dynamic analyses. We determine minimal size and structural features of the DNA target and we show that size and flexibility of the βB6–βB5 hairpin loop are crucial for activity. Variation of the loop sequence results in an altered DNA nicking profile including novel sites. This indicates that the engineering of novel specificities may ultimately be feasible.
Point mutants and loop variants of L1-EN were generated in the context of the retrotransposition reporter plasmid pCEP4/L1.3/ColE1/ () as described in the Supplementary Data. For the expression of mutated L1-EN domains we PCR-amplified DNA corresponding to residues 1–239 of wild-type L1-EN () from the respective retrotransposition reporter plasmids and inserted the products into the NcoI/XhoI cloning sites of expression plasmid pETM11 (). Proteins with N-terminal poly-histidine tags were overexpressed in Rosetta II cells (Novagen) and purified over Ni-chelating chromatography and heparin affinity columns. Protein was quantified spectroscopically or on denaturing SDS polyacrylamide gels. For Tx1L-EN protein (residues 1–239) DNA was amplified from plasmid pE1EN () using primers Tx1L-EN-N1 and Tx1L-EN-C239. For crystallization the respective proteins were expressed and purified without tag as described in (). Purified protein (>1 mg/ml) was stored frozen at −80°C at NaCl concentrations above 300 mM.
Retrotransposition frequencies of wild-type and mutant L1 constructs were determined by applying the rapid, quantitative transient L1 retrotransposition assay described previously (). HeLa cells (2 × 10) were plated in each well of a six-well dish and grown to 50–80% confluency in DMEM. The following day, triplicate dishes were transfected using 6 μl Fugene-6 transfection reagent (Roche) and 2 μg of a Qiagen plasmid DNA preparation per well. At 24-h post-transfection, the transfection mixture was removed and replaced by DMEM. At 72-h post-transfection, the medium was replaced with DMEM containing 400 μg/ml G418. After 10–14 days, G418R colonies were stained with Giemsa solution and counted. The recovery of integrated L1 elements for sequencing is described in ().
Supercoiled pBluescript plasmid DNA was prepared from DH5α cells. Closed circle plasmid DNA was obtained by simultaneous digestion and re-ligation of supercoiled DNA (15 μg/ml) with 5 U/ml HindIII and 900 U/ml T4 DNA ligase resulting in only trace amounts of dimeric product. DNA was quantified after linearization on agarose gels containing ethidium bromide. Nicking reactions (10 or 60 μl) were done in single tubes or 96-well trays in 20 mM Na-HEPES (pH = 7.5), 100 mM NaCl, 10 mM MgCl, 0.1 mg/ml bovine serum albumin (BSA) and 4 mM dithiothreitol. Final concentrations were 2 nM DNA (3.6 μg/ml) and 2–128 nM protein, which had been previously diluted in protein buffer (20 mM Na-HEPES (pH = 7.5), 300 mM NaCl, 10 mM MgCl, 0.3 mg/ml BSA and 10 mM dithiothreitol). After 30 min at 37°C, reactions were stopped by the addition of DNA loading buffer containing EDTA (17 mM final). Reaction products were separated on 1.0 or 1.4% agarose gels (0.5 × TBE) containing 0.5 μg/ml ethidium bromide. Nicking activity was quantified by densitometry, determining the fraction of supercoiled plasmid DNA converted to the open circle form.
Gel-purified synthetic oligonucleotides were labeled at the 5′ end with radioactive phosphate (P) using [γ-P]ATP and T4 polynucleotide kinase and were re-purified on a gel. Equimolar amounts (450 nM) of unlabeled complementary and substrate strands were mixed with a trace amount of labeled substrate. The mixture was annealed in 5 mM Na-HEPES (pH = 7.5) by heating to 90°C and slow-cooled to room temperature. After testing various pH and salt conditions nicking reactions (50 μl) were done in 50 mM Na-HEPES (pH = 6.5), 150 mM NaCl, 10 mM MgCl, 0.1 mg/ml BSA and 1 mM dithiothreitol. Final concentrations were 180 nM DNA (0.5–7.5 μg/ml) and 20–2000 nM protein, which had been previously diluted in protein buffer [5 mM Na-HEPES (pH = 7.5), 300 mM NaCl, 10 mM MgCl, 0.5 mg/ml BSA and 5 mM dithiothreitol]. After 30 min at 37°C, reactions were stopped by the addition of 175 μl of 380 mM Na-acetate (pH = 7.5), followed by phenol extraction and ethanol precipitation. Reaction products were separated on 10% denaturing polyacrylamide gels and quantified in a phosphoimager. The intensity of each band was converted into the relative abundance of each nicking site (Supplementary Table 1) and used to generate sequence logos ().
Untagged LTx [20 mM Na-HEPES (pH = 7.0), 200 mM NaCl] was concentrated to 15 mg/ml. Sitting drops (200 nl protein plus 200 nl reservoir solution) were set up at room temperature using a Mosquito robot. Single crystals appeared over night from a reservoir (75 μl) containing 160 mM MgCl, 370 mM (NH)SO and 33.8% PEG 6000. Untagged LR1 [20 mM Na-HEPES (pH = 7.0), 200 mM NaCl] was concentrated to 10 mg/ml. Hanging drops (2 μl protein plus 2 μl reservoir solution) were set up manually at 4°C. Crystals appeared after several days over a reservoir (500 μl) containing 10 mM MnSO, 200 mM (NH)SO and 31% PEG 1000. Hair seeding improved reproducibility significantly. In both cases, crystals were transferred to a cryo-solution containing 15% glycerol (mixing reservoir and 80% glycerol stock solution) and flash-frozen in liquid nitrogen.
Diffraction data were collected at beamline ID23-1 at the European Synchrotron Radiation Facility in Grenoble, France. Diffraction images were processed by MOSFLM () and SCALA (). The structures were solved by molecular replacement using MOLREP () with L1-EN (PDB ID: 1vyb) as search model. Automatic model building was done with ARP/wARP () to a completeness of 90% for LTx and 98% of LR1. Models were completed manually and structures were refined using REFMAC () and COOT () iteratively.
For Normal mode analysis, the PDB files of L1-EN, LTx, LR1 and TRAS1-EN were provided to the web-based server () following the standard protocol to calculate and analyze the first six vibrational modes ().
We designed variants of L1-EN that fall into three categories (). The first category includes point mutations (D145A, T192V, H230A) of catalytic and structurally important residues that are highly conserved within the entire enzyme family. The second category comprises point mutants (R155A, S202A, I204Y) of moderately conserved non-catalytic surface residues expected to affect the accommodation and recognition of the nucleotide downstream of the scissile bond (A and B).
In the third category of L1-EN variants we manipulated the βB6–βB5 hairpin loop, which is positioned to insert into the DNA minor groove with the possibility to read out both sequence and structural parameters (). It is well suited for a loop-grafting experiment because the anchoring residues T192 and S202 on either side are well conserved among many metal-dependent phosphohydrolases. Therefore, we replaced the entire βB6–βB5 hairpin loop of L1-EN with the corresponding sequences from the R1Bm and Tx1L retrotransposons (B). The resulting mutants LR1 and LTx, respectively, were accompanied by the loop deletion variant L3G, where we exchanged the entire loop (including S202) for a linker of three glycines.
Initially, the L1-EN variants were tested in the context of a functional, tagged L1 element in a well-established cell culture assay (,). We scored successful retrotransposition events by the appearance of G418-resistant HeLa cell colonies, subtracting background activity caused by complementation or endonuclease-independent retrotransposition.
All variants reduce the frequency of retrotransposition significantly, confirming the relevance of the mutated elements (). The strongest effects are seen with point mutants D145A, T192V, I204Y and H230A and with loop variants LR1 and L3G. To test whether this is directly related to the ability of the enzyme to recognize and nick target DNA we purified the respective L1-EN variants for assays .
Residues T192 and H230 are hydrogen-bonded via D205 (). These interactions are apparently essential for the structural integrity of L1-EN as the respective mutants were inherently unstable, degraded easily or precipitated rapidly. From the first category only the D145A mutant could be purified as a negative control for catalytic activity.
The purified L1-EN variants were first analyzed in a plasmid DNA nicking assay (), where supercoiled plasmid is converted into the open circle form that runs considerably slower on an agarose gel (). A shows a side-by-side comparison of the activities of all L1-EN mutants (32 nM) on 2 nM supercoiled plasmid. Under these conditions wild-type L1-EN converts 95% of plasmid DNA into the open circle form. The three point mutants, R155A, S202A and I204Y, show strongly reduced activity, with S202A being affected the least and I204Y the most. The strong effect of I204Y suggests that L1-EN probably binds double-stranded DNA in an orientation that differs from the one seen in the complex with DNaseI (), because in DNaseI the tyrosine is present and tolerated at this position. This view is supported by the effects of S202A and R155A, which indicate that these moderately conserved amino acids are indeed involved in contacting the nucleotide(s) downstream of the scissile bond, either specifically or non-specifically. For a direct contact with R155A the downstream DNA would have to be distorted or even flipped as in the complex with APE1 (). Among the loop variants, LTx remains most active, at levels similar to the S202A point mutant. In contrast, LR1 and L3G retain little but still detectable activity ().
The structural context of the DNA target is important for its recognition by L1-EN (). When presented with equal amounts of supercoiled and of relaxed, closed circle pBluescript DNA, L1-EN nicks the supercoiled DNA much more efficiently (B). Since the βB6–βB5 hairpin loop may well be involved in the recognition of an unusual DNA structure caused by supercoiling, we tested the L1-EN loop variants also in this respect (C). While LTx still prefers supercoiled DNA, the very inefficient LR1 shows no detectable preference for supercoiled DNA anymore. The same observation holds true for L3G, where the loop is deleted. This experiment shows that the βB6–βB5 hairpin loop of L1-EN may be particularly important for reading out the structural context of a potential new retrotransposon integration site.
Finally, there is a good correlation between the nicking activities and the retrotransposition frequencies , indicating that the activity of the endonuclease is limiting over a considerable range (). Consequently, alterations in the nicking specificity of the endonuclease should lead to changes in integration specificity. To distinguish whether our mutations simply impair catalysis or indeed alter target recognition we verified if and how nicking specificities were affected.
Genomic L1 pre-integration sites have been analyzed statistically and a consensus sequence has been reconstructed. In the 5′ to 3′ direction the substrate strand consists of an upstream tract of four to five strongly conserved thymidines (T-tract) followed downstream by two more moderately conserved adenines, with the integration occurring at the poly(T)-A junction (). In contrast to previous approaches () we chose this type of asymmetric target for a DNA oligonucleotide nicking assay ().
We designed a DNA duplex consisting of 14 T-A pairs, followed by two A-T pairs and a single clamp of four C-G pairs [A (Cwt)]. We find the 5′ labeled substrate strand (the bottom strand in all figures) to be nicked throughout the entire T-tract with very similar relative frequencies and only the first five thymidines are spared. Nicking at the poly(T)-A junction is enhanced not more than 4- to 5-fold [B (wt)]. As shown previously (), the observed nicking patterns result from multiple independent endonucleolytic nicking events and not from a cryptic 3′ to 5′ exonuclease activity of L1-EN.
For a closer analysis of the DNA structural parameters required for efficient nicking, we manipulated the complementary DNA strand (upper strand in all figures). A mismatched adenine (A:C) in position (+1) immediately downstream of the target site diminishes the preference for the poly(T)-A junction, reducing it to the levels observed for nicking within the T-tract [A and C (Cim)]. This suggests that, at least during the initial step of recognition of a poly(T)-A junction by L1-EN, this nucleotide position needs to be base-paired properly with an unobstructed minor groove. Mismatching the complete remainder of downstream DNA in addition to position (+1) does not cause any further reduction of nicking efficiency at the poly(T)-A junction [A and C (C56m)]. Next, we tested to which degree the complementary strand is required downstream of the target site by deleting an increasing number of nucleotides from the 5′ end. The results show that the complementary strand needs to extend downstream by at least one nucleotide. However, for nicking at the poly(T)-A junction to be preferred over the adjacent T-tract, at least three downstream base pairs are required [A and C (C53-, C54-, C55- and C56-)]. Upstream of the target site, L1-EN prefers at least 5 nt to be base-paired. If this is not the case, nicking is significantly reduced [A and C (C35-)]. In summary (D), our data suggest that preferential recognition of a poly(T)-A junction by L1-EN requires 5 nt upstream that should be base-paired at least close to the target site and 3 bp downstream which are just sufficient to form a short independent stem that does not need to stack on the upstream duplex for stability. Thus, the minor groove at the poly(T)-A junction would be flexible and could easily be widened by external strain on the DNA or simply by the insertion of the βB6–βB5 hairpin loop, pushing the downstream DNA into a position to be contacted by S202 and R155.
The relative enzymatic activities of the three L1-EN loop variants are similar in the plasmid DNA nicking and duplex DNA nicking assays with LTx being the most active and LR1 being the least active (C and B). LTx still nicks T-tract DNA, but the preference for the poly(T)-A junction has disappeared. We conclude that the βB6–βB5 loop of LTx is less well suited to recognize a poly(T)-A junction, although it does functionally replace the βB6–βB5 loop of L1-EN to a large degree. In sharp contrast, LR1 and L3G do not show any significant endonucleolytic activity, even at the highest concentrations (B).
To extend the analysis of the respective nicking profiles we designed long DNA oligonucleotides (Dwt and Dhy) with more sequence variation (). Dwt contains the genomic target sequences of human L1-EN, Tx1L-EN and R1Bm-EN on a single DNA duplex. This design assures that the potential target sites are present at equal concentrations and compete for the respective endonuclease under identical conditions. Dhy is identical, except that the upstream sequences of the respective target sites have been replaced by T-tracts (A). In addition to L1-EN, we also used wild-type Tx1L-EN (wTx) as a positive control in this assay (). B demonstrates the difference in nicking specificity between the sequence-specific Tx1L-EN and the promiscuous L1-EN. Tx1L-EN nicks almost exclusively at the expected target site, after nucleotide 29 on Dwt and to a lesser extent on the corresponding hybrid site on Dhy. L1-EN nicks preferentially at the poly(T)-A junctions on Dwt or Dhy, but also within extended T-tracts and non-canonical sequences like after nucleotides 19 and 20. This gives rise to characteristic and reproducible nicking profiles (C).
The nicking profile of the LTx chimera deviates from both the L1-EN and the Tx1L-EN pattern. LTx does not recognize the Tx1L integration site (nucleotide 29 on Dwt). On Dhy there is no specific nicking at this position either, despite the upstream T-tract that was introduced and expected to fit the L1-EN scaffold (B). This suggests that one cannot simply combine and exchange upstream (L1-EN scaffold) and downstream (βB6–βB5 loop) recognition elements in a modular fashion to generate a desired target specificity. In comparison to L1-EN, LTx loses the preference for the poly(T)-A junction. Although many nicking sites remain the same for both enzymes, the relative nicking frequencies change. As a result, the LTx nicking profile is clearly distinguishable from the L1-EN pattern (B and C, Supplementary Table 1).
Additionally, LTx also nicks novel sites that are recognized neither by L1-EN, nor by Tx1L-EN. This is illustrated by the frequent nick of LTx after nucleotide 23 on Dwt, where the downstream sequence (5′ AGCT 3′) resembles the Tx1L-EN target sequence (5′ AGTT 3′) downstream of nucleotide 29. In this particular case, the sequence of the LTx βB6–βB5 hairpin loop may play a role in the recognition of downstream DNA (A and B).
In clear contrast to LTx, the βB6–βB5 loop of LR1 cannot functionally replace the βB6–βB5 loop of L1-EN, as replacement results in a low nicking activity. With respect to specificity, LR1 rather seems to avoid T-tracts and produces a very distinct nicking pattern that is quite similar to the one from the loop deletion variant L3G (B, Supplementary Figure 1). The prominent nick of LR1 on Dwt after nucleotide 14 [the preferred nicking site for R1Bm-EN ()] does not seem to be specific for LR1, since it is present also with L3G (data not shown), L1-EN (B) and other variants (Supplementary Figure 1).
To test whether the altered nicking specificity of the LTx endonuclease is reflected by an altered integration site preference of the respective L1 variant, we determined the genomic pre-integration sequences from several G418-resistant HeLa cell clones obtained in the cell culture assay (). Comparison of the nicking profiles to the integration site consensus sequences confirms that for the wild-type L1 element, the nicking specificity of the endonuclease and integration site selection match. However, in the case of LTx, they differ significantly. Like L1, the chimeric LTx element prefers to integrate into locations with a T-tract upstream of the nicking site and only a subset of nicking sites appears to be used for integration (C and D).
This is very interesting as it points to additional constraints for L1 retrotransposon targeting other than the DNA nicking specificity of the endonuclease that we assayed on straight DNA duplexes the poly(T)-A junction may be preferentially recognized in a pre-bent conformation and hence the rigidity of the T-tract could play a much more important role than in the assay. Furthermore, there may be additional contributions for a successful genomic integration, such as base pairing between the 3′ ends of retrotransposon RNA and target site DNA. As a consequence, only a subset of nicked sites would allow for efficient initiation of target-primed reverse transcription.
The distinct effects of the exchanged βB6–βB5 hairpin loop sequences on DNA target recognition and hence nicking specificity is intriguing and may largely relate to the respective structures (). We therefore determined the crystal structures of LTx (C and D) and LR1 (E and F) at 2.3 and 1.8 Å resolution, respectively () and compared them to the existing structure of L1-EN (A and B) (). According to an analysis with the program ESCET (), the common scaffold and catalytic center of the three enzyme variants are essentially unchanged (G and H), despite some variance in the crystal packing. The exchanged βB6–βB5 loop sequences are well ordered in both variants, forming protruding beta-hairpins as in wild-type L1-EN, and their orientation is similar.
The backbone of the LR1 hairpin loop superimposes well onto the backbone of the L1-EN hairpin loop (G and H). Since the βB6–βB5 hairpin loop of LR1 is two amino acids shorter it lacks the tip (P197 and H198 of L1-EN) that bends towards the minor groove of a putative DNA substrate (D). Furthermore, residue T200 of L1-EN is replaced by a glycine in LR1, eliminating an additional possibility of LR1 to interact with the substrate. Finally, the LR1 hairpin loop lacks the positive charges of the L1-EN and LTx hairpin loops that might mediate initial contacts with the negatively charged DNA backbone (F). The backbone of the LTx hairpin loop is twisted slightly with respect to the βB6–βB5 hairpin loop of L1-EN, especially at the distal end (G and H). There, the RDGH sequence of Tx1L-EN (B) replaces P197 and H198 of L1-EN, forming a more extended tip with side chains that could all make favorable DNA contacts (D).
Normal mode analysis is a powerful molecular modeling approach that is particularly suited for calculating slow, large-scale movements within proteins, which would be too expensive computationally for full-scale molecular dynamics simulations. We used the web-based server () to analyze the C-alpha chains of L1-EN, LTx and LR1. As an additional reference we included TRAS1-EN, which is encoded by the telomere-specific APE-type retrotransposon TRAS1 from . Its structure (I and J) is characterized by a βB6–βB5 beta-hairpin loop that, like Tx1L-EN, contains eleven residues (). We calculated the respective average deformation energies of the lowest vibrational mode and also plotted the normalized squared atomic displacements along the sequence of each protein (Supplementary Figure 2).
DNA target specificity of L1-EN has been studied before with plasmid DNA () and with special DNA duplexes that contained a symmetric junction of two T-tracts (). The present study confirms such junctions to be ideal nicking substrates for L1-EN and corroborates the importance of the DNA structure for molecular recognition. We extend the previous analyses to asymmetric DNA targets and determine minimal substrate requirements for the flanking upstream and downstream sequences. Furthermore, we look at the nicking specificity of L1-EN on more general DNA substrates and compare it to the integration specificity of L1 elements .
We find that with unstrained duplex DNA, L1-EN requires a minimum of 5 bp upstream and 3 bp downstream of the target site for efficient target recognition. On the upstream duplex L1-EN recognizes mainly the T-tract (A-tract) geometry () that is primarily characterized by its very narrow minor groove (). Downstream, the 3 bp are just enough to form an independent stem. In the case of a T-A junction following the T-tract (poly(T)-A junction), the downstream adenine is not stacked on the upstream thymidine () and thus, the downstream stem can more easily be bent away with an associated widening of the minor groove. Most likely, this local flexibility is a feature that is recognized by L1-EN in addition to the narrow minor groove of the T-tract, leading to the enhanced nicking efficiency observed at the junction. On a strained substrate such as supercoiled plasmid DNA, the difference between cleaving T-tract DNA and a poly(T)-A junction would probably be even more pronounced. The torsional strain might widen the minor groove at the junction even further and facilitate the structural recognition of the DNA target.
Although the structure of L1-EN would allow the accommodation of a flipped nucleotide at position (+1) downstream of the scissile bond (), we do not find any evidence for the base-specific recognition of such a nucleotide. At least for the initial target recognition the nucleotide needs to be part of a downstream stem. However, this does not rule out the possibility that the flexibility (or ‘flippability’) of the nucleotide is required in consecutive steps of the integration process.
In conclusion, L1-EN recognizes structural features of the DNA target rather than specific nucleotides in the sequence. The 5′ TTTT/AA 3′ integration site consensus sequence may fulfill these structural requirements in an ideal way, but many alternative sequences seem to have similar structural features and are nicked . The requirements for integration seem stricter than the requirements for nicking. This indicates that although the nicking specificity of the endonuclease is the primary determinant for integration site selection it may not be the only one (). Additional specificity factors could influence the choice of nicking site in the first place (co-targeting factors) or select among already nicked sites the ones that are suitable for integration (post-nicking factors). The latter possibility is favored by reports of endonuclease-independent retrotransposition () and L1-induced chromosomal breaks ().
During DNA target site recognition, the conformational space available to the downstream DNA duplex is probed by the insertion of the βB6–βB5 beta-hairpin loop of L1-EN into the minor groove at a poly(T)-A junction, according to the presented model (D). The presence of the loop is important for nicking activity and both nicking activity and target specificity are very sensitive to structural changes of the loop, especially at its tip. Similar to the situation in TRAS1-EN () a deletion of the tip (LR1) or of the entire loop (L3G) results in an altered specificity and much reduced activity. To examine the importance of the amino acid sequence we exchanged residue H198 in the tip of the loop, which had no impact on the nicking pattern. Even the substitution of the entire loop with a different sequence and an extended reverse turn (LTx) was tolerated rather well. This suggests that the conformational flexibility of the beta-hairpin loop probing the DNA minor groove may be much more important than its sequence, especially if target recognition proceeds via the structural flexibility of the DNA at the poly(T)-A junction. This hypothesis is supported by the presented Normal mode analysis. The βB6–βB5 hairpin loop of LTx may be able to functionally replace the βB6–βB5 hairpin loop of L1-EN because it is flexible enough to insert partially into the minor groove of many L1-EN targets to probe the conformational space of the downstream duplex. The βB6–βB5 hairpin loop of LR1 may be too rigid for this function. In its natural context on R1Bm-EN () it may only be required as a counter bearing for the target DNA, which would then be probed sequence specifically from the side of the major groove by a unique extension of surface loop βB4–αB2, predicted for R1Bm-EN ().
The L1 retrotransposon bears considerable potential as a genetic tool (). It can be delivered to cells by an adenovirus vector () and its suitability for mutagenesis has recently been demonstrated with a synthetic, highly active mouse L1 element called ORF (). The application of similar L1 retrotransposons for gene delivery into defined genomic locations requires engineering of the endonuclease target specificity as one of the most crucial steps. This appears feasible since there are many natural APE-type non-LTR retrotransposon endonucleases with distinct target specificities that all share the same protein scaffold and the same catalytic site (,).
Loop grafting experiments have been shown to mimic evolutionary processes (), allowing novel specificities to be engineered (,). The analysis of the presented L1-EN βB6–βB5 hairpin loop variants shows that the respective grafting experiments worked successfully from a structural point of view and that other surface loops may be manipulated in a similar way in the future. From a functional point of view, we could show that the DNA nicking profile of L1-EN is quite sensitive to structural changes of the studied loop and that novel specificities can indeed be acquired. For further improvements high-resolution structures of retrotransposon endonucleases in complex with their respective DNA targets would be of great help.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Precise temporal and spatial control of gene expression is accomplished by a broad array of sequence-specific transcription factors. Many of these are inefficient transcriptional activators or repressors on their own, but they recruit potent coactivators or corepressors that cannot bind directly to DNA in turn (). Regulatory mechanisms include chromatin-remodeling factors that mobilize nucleosomes and histone-modifying enzymes. The expression of such regulatory factors is controlled by diverse signaling pathways, other transcription factors and regulatory RNAs, building up a highly complex transcriptional network.
The Notch signaling pathway represents a central regulator of gene expression. This cascade controls cell fate determination and differentiation, making it essential for many aspects of embryonic development as exemplified by a variety of mouse knockout studies. In humans, mutations of Notch ligands or receptors are responsible for a number of diseases like Alagille syndrome, CADASIL, T-cell leukemia, aortic valve calcification and other cardiovascular disorders ().
Notch receptors are single-pass transmembrane proteins that become activated upon ligand binding. This leads to two consecutive cleavage events releasing the intracellular domain (NICD), which then translocates to the nucleus. There NICD interacts with the DNA-binding protein RBPJκ (also known as CBF1, Rbpsuh or Su(H) in ), which is associated with corepressors (e.g. N-CoR, SHARP, CtBP). Interaction with NICD replaces these corepressors and allows recruitment of coactivators like Mastermind/MAML and p300/CBP leading to transcription of target genes () (). The most extensively studied and best understood targets are and [] genes in and the related and genes in mammals (). Interestingly, there is a crosstalk between Notch and the BMP/TGF-beta, JAK-STAT, Ras and HIF signaling pathways to enhance activation of Hey/Hes expression (), suggesting that these factors transduce and integrate signals from multiple pathways.
Besides the activation of target genes via RBPJκ, referred to as the canonical pathway, additional, non-canonical functions of Notch have been described that are less well characterized. These include e.g. regulation of the actin cytoskeleton, interaction with the wingless pathway, RBPJκ-independent activation of target genes (), or activation of the RNA-binding protein Musashi ().
In the fruitfly and seven clustered genes (β, γ and δ) control crucial developmental processes like segmentation, myogenesis or neurogenesis. All of these genes encode basic helix-loop-helix (bHLH) proteins (). The DNA-binding basic domain (b) is contiguous with one of two amphipathic α-helices separated by a loop (HLH) that serve as a dimerization domain and as a platform for additional protein interactions (). The HLH region is followed by two additional α-helical stretches (helix3/4), called the Orange domain. This domain is thought to serve as an additional interface for protein interactions and it acts as a transcriptional repressor when fused to a DNA-binding domain (). A further characteristic of Hairy and E(spl) proteins is the invariant proline residue in the basic domain and a highly conserved carboxyterminal tetrapeptide motif WRPW that recruits the corepressor Groucho ().
genes are activated by Notch signaling and their protein products, as well as Hairy, block neuronal differentiation by inhibiting proneural bHLH activators like Atonal, Daughterless and those of the Achaete–Scute complex (). The molecular details of how this is achieved are diverse and in part still controversial. Proposed models include sequestering of activator complexes away from DNA (), direct binding to promoters of target genes and recruitment to promoters without direct DNA binding ().
In the mouse and rat genomes, seven () () and three genes ( also published as ) have been identified (). Hes proteins are highly similar to Hairy and E(spl), especially within the bHLH, Orange and WRPW domains. Similar to their ancestors, Hes proteins are supposed to bind N- and E-box DNA sequences (CACNAG, CANNTG) and they can recruit TLE1-4 corepressors (the orthologs of Groucho) through their WRPW tetrapeptide (). While and can be induced by the Notch pathway (,), () and () appear to be independent of Notch signaling, and further data on are lacking.
All members of the gene family can be induced by Notch (,,,,), they are strongly conserved during evolution () and there is also a gene with hitherto unknown function (,). Especially the bHLH and Orange domains are similar to those of Hes proteins, but the invariant proline residue in their basic domains is replaced by glycine and they do not bind to N-box sequences (). Hey proteins preferentially bind to an E-box sequence that is also recognized by Hes1, Hes6 and E(spl) proteins (,). The most striking difference between Hes and Hey proteins is the lack of the WRPW tetrapeptide in the latter. Instead a related YRPW peptide or a further degenerated YXXW (HeyL) sequence can be found, which cannot bind TLE corepressors (,). The YXXW motif is followed by a conserved TE(I/V)GAF peptide with presently unknown function (B).
There are several additional mammalian proteins that exhibit strong homologies to Hairy and E(spl). Examples are Helt, DEC1 (also known as Stra13, SHARP-2 or BHLHB2) and DEC2. They generally lack WRPW/YRPW motif sequences and there is no evidence for a Notch-dependent expression thus far.
Mammalian Hairy-related proteins are specifically expressed in various tissues and they fulfill important roles during development and adulthood. It is beyond the scope of this manuscript to review all these functions in detail, instead provides a short overview of the phenotypes seen in gene targeting experiments in mice.
plays an essential role in the development of the nervous system, sensory organs (eye, inner ear), pancreas and endocrine cells, as well as lymphocytes. Loss of or is less severe, but combined with deficiency leads to more profound pathologies, as there is partial redundancy among these genes. is important for somitogenesis. In contrast, genes play critical roles in the cardiovascular system. mice and those with a combined loss suffer from severe congenital heart defects. While mutants are viable, a combined deficiency phenocopies the vascular defects of embryos, including impaired angiogenic remodeling and a lack of arterial differentiation. The known overlap in expression sites () suggests that there may be additional genetic interactions to be uncovered in compound and deficient mutants.
fig
table-wrap
#text
xref
table-wrap
#text
Almost thirty years of research on Hairy and related factors, lead to the publication of a few hundred papers, providing a wealth of information. This survey gathers data on the progress on mouse models, protein–protein and protein–DNA interactions to faciliate future studies and to highlight questions that are still unresolved.
Although quite a number of cofactors are now known that are recruited by Hairy-related proteins, little is known about the dynamics of these interactions. While the Hes1/TLE1 interaction is necessary for neural stem cell maintenance, it dissociates during neuronal differentiation and Hes1 is turned from a repressor into an activator, which is at least partly mediated by post-translational modifications like phosphorylation (). In addition, the repressive functions of Hes6 can be blocked by phosphorylation of a specific serine residue by CK2 (). It appears obvious that such strategies might also be employed for tightly regulated complex formation and dissociation of Hes and Hey proteins with other factors dependent on the cellular context.
Furthermore, DNA-binding affinities of Hes and Hey factors may be regulated by kinases. Phosphorylation of a residue in the basic domain of Hes1 decreases DNA-binding affinity (), implying that Hairy-related factors can be influenced by growth factor signaling cascades. During the last years it already became clear that there is a crosstalk between Notch and the BMP/TGF-beta, JAK/STAT, Ras and hypoxia signaling cascades in the regulation of Hes and Hey genes (). Elucidation of the biological effects as well as the associated post-translational modifications of those interactions will be a challenge in the future.
The question of how Hes and Hey factors act as repressors when they bind to other transcription factors is also not fully answered. We have schematically summarized the current models in . Another possible mode of repression, which was not explicitly mentioned so far, is Hes/Hey-mediated protein degradation of transcriptional activators. It has been proposed that the WRWP peptide of Hes6 mediates proteins degradation and Hes1 seems to induce rapid degradation of the Mash1 transcription factor (,,). Furthermore, the half-life of BMP-induced Id1 is reduced upon Hey1 binding (). Again, knowledge of protein modifications like ubiquitinylation or SUMOylation is elusive.
Finally, it will be challenging to analyze the cellular localization of these factors in more detail. It already became clear that Hey1 is located in the cytoplasm and the nucleus of benign prostate hyperplasia, but is excluded from the nucleus in many malignant prostate cancers (). The cytoplasmic functions of bHLH transcription factors are unknown. It will be interesting to see, if these proteins possess additional functions besides transcriptional regulation. |
Promoter recognition by RNA polymerase III, which transcribes small non-coding genes such as tRNA genes and 5 S RNA genes, is mediated by the transcription factors (TFs) TFIIIA, TFIIIB and TFIIIC (,). These factors assemble in a stepwise manner, beginning with recruitment of the multi-subunit TFIIIC to the promoter of a tRNA gene and the single-subunit TFIIIA to the promoter of the 5 S RNA gene. The intragenic promoters of these genes, which are referred to as internal control regions (ICRs), are often multipartite, with the sub-elements contributing to the appropriate architecture of the final transcription complex (,). The protein–DNA and protein–protein interactions that promote assembly of a transcription complex on a tRNA gene have been extensively studied with TFs from (). Binding of the six-subunit yeast TFIIIC to the bipartite ICR of a tRNA gene protects the entire gene against cleavage by DNase I. TFIIIC then mediates stable binding of TFIIIB, which consists of three polypeptides including the TATA-binding protein, upstream of the start site of transcription. TFIIIB is responsible for recruitment of RNA polymerase III and is able by itself to direct multiple rounds of accurate transcription. In contrast to the tRNA gene, the 5 S RNA gene does not bind TFIIIC directly; rather, the TFIIIA–5 S DNA complex provides the platform for TFIIIC recruitment. Formation of the TFIIIC–TFIIIA–5 S DNA complex is a prerequisite for the subsequent recruitment of TFIIIB and RNA polymerase III. In this way, TFIIIA acts as an adaptor protein to assemble an initiation complex at the promoter of the 5 S RNA gene in which the placement of TFIIIC and TFIIIB is similar to that on a tRNA gene ().
In most organisms, TFIIIA contains nine sequential CysHis zinc fingers that are separated by short linker sequences. However, TFIIIAs with longer spacer regions and additional zinc fingers have been identified in some species (,). The interaction of TFIIIA from with the tripartite ICR of the amphibian 5 S RNA gene has been studied in detail. The ICR consists of an A box (nt +50 to +64), an intermediate element (nt +67 to +72) and a C box (nt +80 to +97). The results of DNase I protection and chemical protection/interference studies with full-length and truncated polypeptides bound to DNA suggest that the three amino-terminal and the three carboxyl-terminal fingers of TFIIIA wind around the major groove, making contacts with the C and A boxes, respectively. Fingers 4, 5 and 6 are proposed to bridge these two domains, with fingers 4 and 6 crossing the minor groove and finger 5 making contacts in the major groove of the intermediate element (). Structures of polypeptides containing the amino-terminal 3 or 6 fingers bound to DNA have been determined by X-ray and NMR analyses and support the models derived from biochemical and genetic studies (,,).
Because a polypeptide consisting of the three amino-terminal zinc fingers of TFIIIA was found to bind to the 5 S RNA gene with an affinity similar to the full-length protein, this amino-terminal module was initially considered to be the main contributor to the affinity of the protein for DNA (). However, additional studies with truncated proteins and full-length proteins that contained point mutations, ‘broken’ fingers and ‘swapped’ fingers, as well as insights derived from structural analyses, led to the conclusion that the carboxyl-terminal fingers also make a large contribution to the affinity of the protein–DNA complex (). Accommodating the binding of the full-length TFIIIA to DNA has been shown to involve both neighboring finger–finger interactions as well as longer range interactions that contribute both favorably and unfavorably to the overall affinity of TFIIIA for the promoter (,).
Once bound to DNA, TFIIIA directs assembly of the TFIIIB–TFIIIC–TFIIIA–DNA complex. The observation that the association of TFIIIA with DNA is much more stable in a fully assembled complex and in a TFIIIC–TFIIIA–DNA complex than in a TFIIIA–DNA complex has led to the suggestion that the rate of assembly of the complete transcription complex is largely independent of the equilibrium binding constant of TFIIIA for DNA as measured in the formation of a binary complex (,). For example, a version of TFIIIA that contains a disruption of zinc finger 3 supports wild-type levels of 5 S RNA synthesis despite having a 27-fold reduction in its DNA-binding affinity ().
The carboxyl-terminal portion of TFIIIA, beginning at zinc finger 7, appears to have a role in the formation of the initiation complex that is distinct from its role in contacting the A box of the promoter (). A mutation that prevents proper folding of any one of the first six zinc fingers of TFIIIA does not affect either the or TF activity of the protein (,). A mutation that disrupts the structure of zinc finger 7, 8 or 9 or deletion of a 14-residue sequence in the C-terminal extension of the protein leads to transcriptional defects that cannot be accounted for solely by effects on DNA binding (,). For example, disruption of zinc finger 7 or zinc finger 8 leads to a similar reduction in DNA-binding affinity of TFIIIA. However, disruption of zinc finger 7 leads to only a modest reduction in the activity of TFIIIA whereas disruption of zinc finger 8 abolishes the ability of the protein to support transcription of the 5 S RNA gene and greatly reduces its activity (,,).
Like its counterpart, yeast TFIIIA contains nine CysHis zinc fingers. Yeast TFIIIA also contains a unique 81-amino-acid sequence between zinc fingers 8 and 9 that is essential for its TF activity (). The ICR of the yeast 5 S RNA gene consists of only a C-box element, located between nucleotides +81 and +94 (). DNase I footprinting and methylation protection studies indicate that the three amino-terminal zinc fingers span the ICR, contacting residues from +79 to +98 (). Zinc finger 5 contacts residues 73 and 74 of the 5 S RNA gene, but this interaction does not appear to contribute to the affinity of the protein for the 5 S RNA gene (). Although zinc fingers 6 through 9 are not in close contact with DNA in the TFIIIA–DNA complex (), these fingers are in close proximity to DNA in the fully assembled initiation complex (,,). A truncated version of yeast TFIIIA containing the three amino-terminal zinc fingers interacts with the ICR with high affinity. Although this TFIIIA–DNA complex is able to recruit TFIIIC, the resulting complex is non-functional (,). The recruitment of TFIIIC appears to require zinc finger 1, as a polypeptide that begins with finger 2 is able to bind to the 5 S RNA gene, albeit with reduced affinity, but does not recruit TFIIIC (). A short leucine-rich segment within the 81-amino-acid region present between zinc fingers 8 and 9 of yeast TFIIIA is also required for the assembly of a functional transcription complex (,). This region might serve as a second docking site for TFIIIC and in combination with zinc fingers 8 and 9, which contribute in a redundant manner to the TF activity of TFIIIA, enforce a topography on the TFIIIC-TFIIIA-5 S DNA complex that promotes proper positioning of TFIIIB (,,).
We have examined the roles of zinc fingers 1 through 7 of yeast TFIIIA in the establishment of a functional transcription complex on the 5 S RNA gene. We constructed versions of TFIIIA in which the folding of an individual zinc finger was disrupted by substitution of a zinc-coordinating histidine residue by an asparagine residue. In this study, we compare the ability of these mutant TFIIIAs to bind DNA and to support and transcription of the 5 S RNA gene.
The construction of pXS-TFC2, which contains the coding region for TFIIIA downstream of a T7 RNA polymerase promoter, has been described elsewhere (). This plasmid served as the parental plasmid for the introduction of mutations into the coding sequence of TFIIIA. Site-directed mutagenesis was achieved by recombinant PCR using the overlap extension procedure as described previously () with pXS-TFC2 as the template except as otherwise noted. All PCR amplifications were performed by using the high fidelity Vent DNA polymerase as instructed by the manufacturer (New England Biolabs), and the sequence of all amplified DNA was verified by DNA sequencing.
Another series of plasmids encoding versions of TFIIIA with disruption of two zinc fingers was made by combining the appropriate restriction fragments from parental plasmids described above. The ∼510-bp NcoI–HindIII fragment from pXS-TFC2(H126N) (third finger disruption) or pXS-TFC2(H154N) (fourth finger disruption) was used to replace the corresponding fragment in plasmids pXS-TFC2(H181N) (fifth finger disruption), pXS-TFC2(H214N) (sixth finger disruption), pXS-TFC2(H240N) (seventh finger disruption) and pXS-TFC2(C367Y) [ninth finger disruption; ()] to generate plasmids encoding versions of TFIIIA containing a disrupting mutation in both fingers 3 and 5, 3 and 6, 3 and 9, 4 and 5, 4 and 6, 4 and 7 and 4 and 9. The plasmid encoding the version of TFIIIA containing disruptions of zinc fingers 5 and 6 was constructed by replacing the ∼560-bp NcoI–Bsu36I fragment from pXS-TFC2(H214N) (sixth finger disruption) with the corresponding fragment from pXS-TFC2(H181N) (fifth finger disruption).
The construction of a plasmid for bacterial expression of wild-type TFIIIA was described previously (). Plasmids for the expression of mutant versions of TFIIIA described above were generated by inserting the 2.3-kb NcoI–BamHI fragment of the appropriate pXS-TFC2 plasmid between the corresponding sites of pET-11d (). This places the coding region of TFIIIA under the control of a bacteriophage T7 RNA polymerase promoter.
Two series of plasmids were used for expression of wild-type and mutant forms of TFIIIA. The yeast shuttle vector ΔpG-3, a pUC18-derived plasmid that contains a 2 μ origin of replication and the selectable marker , has been described previously (). KpnI–BamHI fragments containing the open-reading frames of the wild-type and mutant versions of TFIIIA were purified from pXS-TFC2 plasmids and inserted between the KpnI and SalI sites of ΔpG-3 after the BamHI- and SalI-generated ends had been filled in by the Klenow form of DNA polymerase I in the presence of dNTPs. This placed the coding region of TFIIIA between the constitutive promoter of the glyceraldehyde-3-phosphate dehydrogenase (GPD) gene and the transcription terminator of the phosphoglycerate kinase (PGK) gene.
pRS314-TFC2, a low-copy plasmid containing the wild-type coding sequence for TFIIIA under the control of its own promoter, was constructed as follows. The promoter sequence of the TFIIIA gene was amplified by PCR using pJA230 () as a template. The upstream primer annealed 230-bp upstream of the translation start site and introduced a KpnI site, whereas the downstream primer annealed at the translation start site and introduced an NcoI site. The 230-bp PCR product was digested with KpnI and inserted between the KpnI and EcoRI sites of the plasmid pRS314 after the EcoRI site had been blunted with Klenow in the presence of dNTPs. This generated plasmid pRS314. This /-containing plasmid carries the selectable marker and a yeast autonomously replicating sequence (). NcoI–BamHI fragments containing the open reading frames of wild-type and mutant versions of TFIIIA were purified from pXS-TFC2 plasmids and inserted between the corresponding sites of pRS314.
Wild-type and mutant versions of TFIIIA were synthesized using the TnT (Promega) coupled transcription–translation system essentially as described previously (). pXS-TFC2 and its variants were used as template unless otherwise indicated. Five microlitrers of each 20 μl reaction was loaded on an SDS–15% polyacrylamide gel, transferred to nitrocellulose for western blotting and probed with anti-TFIIIA as described below to confirm that each protein was synthesized and stable. The remainder of the synthesized protein was used in electrophoretic mobility shift assays (EMSAs) and transcription assays.
The pET-11d-derived plasmids were transformed into the strain BL21(DE3), and TFIIIA was purified as described ().
EMSAs were performed as described (). A 20 μl reaction contained a 2 μl aliquot of TFIIIA, 4 μl of partially purified TFIIIC where indicated, and a 270-bp end-labeled DNA fragment, which was excised from p19-5 S and contains the yeast 5 S RNA gene (). TFIIIA for these experiments was either purified from bacteria or synthesized in an transcription–translation reaction programmed to produce the indicated version of TFIIIA. The partially purified TFIIIC-containing fraction derived from yeast was prepared as described for fraction j in Taylor and Segall. transcription assays were performed as described () using the yeast 5 S RNA gene (p19-5 S) as template. A 50 μl reaction contained a 4.5 μl-aliquot of either bacterially expressed or synthesized TFIIIA and 12.5 μl of a yeast-derived heparin-agarose fraction (fraction h) () that contained TFIIIC, TFIIIB and RNA polymerase III.
The haploid yeast strain YRW1 (
and harboring pJA230), which was used for analysis of wild-type and mutant versions of TFIIIA, has been described previously (). pJA230 is a / plasmid with a selectable marker and a 10-kb insert of yeast DNA containing , the gene encoding TFIIIA (). Since is an essential gene, viability of YRW1 depends on the presence of pJA230 ().
ΔpG-3 and pRS314 plasmids directing expression of wild-type or mutant forms of TFIIIA were transformed into YRW1. After transformants had been selected on medium lacking uracil and tryptophan, the ΔpG-3- and pRS314-containing strains were grown on medium lacking tryptophan and containing uracil to allow for loss of pJA230. Cells were then streaked on plates containing medium supplemented with 5-fluoro-orotic acid (5-FOA) and uracil and lacking tryptophan. Because 5-FOA kills cells containing the gene, only those cells which have lost pJA230 and contain a ΔpG-3 or pRS314 derivative encoding a functional version of TFIIIA will grow on the 5-FOA-containing medium.
YRW1 and strains of YRW1 containing ΔpG3-derived plasmids that directed expression of mutant versions of TFIIIA were grown overnight in minimal medium lacking tryptophan to an absorbance at 600 nm of ∼4.0. The cells from 10 ml of culture were harvested, washed twice with cold water and re-suspended in 200 μl of lysis buffer (30 mM HEPES, pH 7.9, 150 mM NaCl, 1 mM DTT and 0.1% NP-40). Cells were vortexed in the presence of 200 μl glass beads for six 1-min bursts interrupted by chilling on ice. Samples were spun at 4° for 10 min at 23 500 , and the supernatant transferred to a fresh tube and re-spun for an additional 10 min. Protein concentration was determined using the Bradford assay with bovine serum albumin as the standard. Protein (3.75 μg) was loaded on an SDS 15%–polyacrylamide gel, transferred to nitrocellulose for western blotting and probed with anti-TFIIIA antibody as described previously ().
In this study, we have assessed the requirements for zinc fingers 1 through 7 of yeast TFIIIA in the establishment of a productive transcription complex on the 5 S RNA gene. A previous assessment of the activities of truncated versions of yeast TFIIIA suggested that the first zinc finger contributes to DNA binding and to recruitment of TFIIIC (). Analysis of the effect of substitutions in zinc fingers 8 and 9 that are predicted to disrupt the folding of the finger indicated that these two zinc fingers contribute in a redundant manner to the TF activity of yeast TFIIIA (). To extend this analysis, we generated a series of plasmids for and expression of proteins in which the second zinc-coordinating histidine residue of each of zinc fingers 1 through 7 was replaced by an asparagine residue (A). We refer to this substitution, which is predicted to prevent stable folding of the finger, as a disrupting mutation and to each disrupted finger as zf#, with # identifying the disrupted zinc finger. The substituted residue for disruption of zinc finger 1 through 7 was H69, H98, H126, H154, H181, H214 and H240, respectively (B). As a first test of the effect of these mutations on the activity of TFIIIA, we compared the ability of approximately equal amounts of synthesized protein (data not shown) to bind to the 5 S RNA gene as assessed by an EMSA (C). Because we have used this assay to give a qualitative assessment of protein–DNA interactions and have not determined values, we use the term relative affinity in describing our interpretation of the EMSAs. TFIIIA(zf6) was the only mutant protein that retained the ability to bind to the 5 S RNA gene with a relative affinity similar to the wild-type protein (C, compare lanes 2 and 8). TFIIIA with a disruption of zinc finger 1, 4, 5 or 7 appeared severely compromised for DNA-binding activity, with disruption of zinc finger 7 being less deleterious than disruption of zinc finger 1, 4 or 5 (C, lanes 3–7 and 9). TFIIIA(zf2) and TFIIIA(zf3) had no detectable DNA-binding activity (C, lanes 4 and 5). Despite the qualitative nature of this assay, the data suggest that within the context of the entire protein, only zinc finger 6 of the first 7 fingers can be disrupted without compromising DNA-binding activity. This analysis also revealed the presence of a minor amount of protein–DNA complexes of higher mobility than typically observed for the TFIIIA–DNA complex, particularly for reactions containing TFIIIA(zf4) and TFIIIA(zf5). It is possible that these complexes contained a truncated protein generated by proteolytic cleavage within the disrupted finger.
We next tested the TF activity of the synthesized proteins. TF activity, which refers to the ability of TFIIIA to assemble a functional transcription complex, was monitored in reactions that contained partially purified TFIIIC, TFIIIB, RNA polymerase III and 5 S DNA (see Materials and Methods section). Comparison of TF activity of the mutant TFIIIAs, as measured by the amount of 5 S RNA synthesized in transcription reactions (D), with the EMSA results (C) showed that there was little correlation between the relative ability of a mutant TFIIIA to bind to the 5 S RNA gene and to support transcription of the gene. For example, TFIIIA(zf4) and TFIIIA(zf5), which appeared severely compromised in their ability to interact with the 5 S RNA gene, were nonetheless able to support near wild-type levels of transcription (C and D, lanes 6 and 7); similarly, TFIIIA(zf3), which had no detectable DNA-binding activity, was able to support a low level of transcription (C and D, lane 5). This suggested that in the presence of yeast TFIIIC, TFIIIB and RNA polymerase III, versions of TFIIIA with low affinity for the ICR could nonetheless be captured on DNA to form transcriptionally active complexes. Indeed, previous studies indicated that TFIIIC influences the stability of the TFIIIA–DNA complex (). Detailed studies with TFIIIA led to the conclusion that the activity of TFIIIA in an transcription reaction is independent of its apparent affinity for 5 S DNA as determined in a binary reaction and that TFIIIA molecules can be sequestered in higher order, essentially irreversible transcription complexes (). It is also possible that the activity of TFIIIA(zf4) and TFIIIA(zf5) increased because potentially proteolytic sensitive sites in these proteins became protected. In contrast, TFIIIA(zf7) did not support transcription of the 5 S RNA gene despite having a readily detectable DNA-binding activity (C and D, compare lanes 9). TFIIIA(zf1) and TFIIIA(zf2), which had minimal or no DNA-binding activity, respectively, also appeared unable to support transcription of the 5 S RNA gene (C and D, lanes 3 and 4). Finally, TFIIIA(zf6), which had near wild-type DNA-binding activity, supported a high level of transcription (C and D, lane 8). In summary, synthesized TFIIIAs with a disruption of zinc finger 4, 5 or 6 retained a high level of TF activity whereas TFIIIAs with a disruption of zinc finger 1, 2, 3 or 7 appeared to have defects in TF activity.
We also expressed wild-type TFIIIA and TFIIIA with a disruption of zinc finger 1, 2 or 3 in bacteria so that we could test the activities of these proteins at higher concentrations than was possible with synthesized protein. We found that we could readily detect TFIIIA–5 S DNA complexes by EMSA with these three mutant TFIIIAs purified from bacteria. Relative to wild-type TFIIIA, however, all three mutant TFIIIAs showed reduced affinity for the 5 S RNA gene. The interaction of TFIIIA(zf2) and TFIIIA(zf3) with DNA appeared to be compromised to a greater extent than the interaction of TFIIIA(zf1) (E, lanes 3, 5, 7 and 9). We emphasize that we have not carried out quantitative measurements to determine values and it is possible that these reactions contained close to saturating amounts of TFIIIA. Of note, however, is that TFIIIA(zf1) was clearly able to bind to the 5 S RNA gene. We also monitored the effect of the addition of partially purified TFIIIC to the EMSA reactions. As observed previously (), recruitment of TFIIIC to a wild-type TFIIIA–DNA complex generated a complex that had greatly reduced mobility (E, lane 4). None of the mutant TFIIIA–DNA complexes appeared, however, to be able to efficiently recruit TFIIIC (F, lanes 6, 8 and 9).
We next tested the ability of these bacterially expressed versions of TFIIIA to support transcription. TFIIIA(zf1) was unable to support transcription (F, lane 3) whereas TFIIIA-zf2 and TFIIIA-zf3 were able to support transcription of the 5 S RNA gene (F, lanes 4 and 5). These results support the idea that zinc finger 1 plays an essential role in the assembly of an active initiation complex ().
We used a plasmid-shuffling protocol to test the mutant versions of TFIIIA for their ability to support cell viability. For this study we used two sets of plasmids: a high-copy ΔpG3-derived series of plasmids in which expression of was directed by the strong, constitutive GPD promoter () and a low-copy series of plasmids in which expression of was under the control of its own promoter (see Materials and Methods section). Plasmids encoding the mutant proteins were introduced into a Δ strain that harbored pJA230. This is a low-copy plasmid that contains , which encodes TFIIIA and is an essential gene, and as a selectable marker (). The transformed cells were then tested for their ability to grow on 5-FOA-containing medium; growth on this medium requires that the -marked pJA230 plasmid has been lost from cells and is therefore diagnostic of the newly introduced mutant version of being able to support expression of the 5 S RNA gene. The only essential role of TFIIIA is in directing expression of the 5 S RNA gene ().
We found that TFIIIA containing a disruption of zinc finger 4, 5 or 6 supported cell growth when expressed at either a low or a high level. TFIIIA containing a disruption of zinc finger 2 or 3 was able to support cell viability, but only when expressed from the high-copy vector. TFIIIA containing a disruption of zinc finger 1 or 7 failed to support cell viability irrespective of expression level (G, top panel). Our anti-TFIIIA antibody is insufficiently sensitive to detect endogenous TFIIIA or TFIIIA expressed from a low-copy vector in our standard western blot analysis () (G, lane 1). However, TFIIIA expressed from the high-copy vector can be readily detected () (G, lane 2). We therefore used western blot analysis to test for cellular accumulation of mutant TFIIIAs expressed from the high-copy plasmid in cells prior to growth on 5-FOA-containing medium. These cells were also expressing wild-type TFIIIA from the low-copy plasmid pJA230. We found that all mutant proteins were sufficiently stable to be readily detected (G, bottom panel) and were thus present in cells at a higher level than chromosomally expressed TFIIIA.
In summary, this analysis indicated that zinc fingers 1 and 7 are essential for the assembly of a functional transcription complex on the 5 S RNA gene. The observation that the deleterious effect of disruption of zinc finger 2 or zinc finger 3 could be compensated for by a higher level of expression of the mutant protein was consistent with the observations made in comparing the activities of synthesized and bacterially expressed proteins (C–F). Disruption of zinc finger 4, 5 or 6 did not appear to compromise the activity of TFIIIA.
Because cells expressing TFIIIA with a disruption in any one of zinc fingers 4, 5 or 6 from a low-copy plasmid were viable, we investigated the possibility that the tertiary structure of this central trimeric unit of fingers might be dispensable. For example, this region of the molecule might serve as a flexible linker to allow appropriate positioning of the amino-terminal and carboxyl-terminal regions of the molecule. We therefore tested the effect of simultaneous disruptions of zinc fingers 4 and 5, 4 and 6, and 5 and 6 on the activity of TFIIIA. For comparison, we also monitored the activity of TFIIIA with simultaneous disruption of zinc fingers 3 and 5, 3 and 6, 3 and 9 and 4 and 9. When these double mutant versions of TFIIIA were synthesized and tested for their TF activity, all were found to be defective in the ability to promote 5 S RNA synthesis (A). However, versions of TFIIIA that combined a disruption of zinc finger 4 with a disruption of zinc finger 5, 6 or 9 supported cell viability when expressed from the high-copy plasmid , although they failed to do so when expressed from the low-copy plasmid (B, lanes 6–8). TFIIIAs that combined a disruption of zinc finger 3, which by itself reduced the activity of TFIIIA such that it could only support cell viability when expressed from a high-copy plasmid (G), with a disruption of zinc finger 5, 6 or 9 were stable, but inactive (B, lanes 3–5). TFIIIA(zf5, zf6) was also unable to support cell viability even when expressed at a high level (B, lane 9). In summary, we found that the activity of TFIIIA(zf3) was sufficiently compromised such that it lost all activity on disruption of a second zinc finger. In contrast, the structure of zinc finger 4 appears to be less important for the overall function of TFIIIA, and zinc fingers 5 and 6 may contribute in a redundant manner to an essential function of yeast TFIIIA.
Our previous () and current data suggest that the first zinc finger of yeast TFIIIA contributes not only to DNA binding, but also serves a role in recruitment of TFIIIC to the TFIIIA–DNA complex. To identify residues in the first zinc finger that might contribute to an interaction with TFIIIC, we mutated amino acids that would be expected to be surface exposed and not directly involved in DNA binding. The solved structures of several zinc finger modules bound to DNA show that such surface-exposed residues are present in the short antiparallel β-sheets and in the β-turn connecting the β-strands (,). We therefore carried out alanine-scanning mutagenesis of the corresponding region of the first zinc finger of yeast TFIIIA. Residues from Y49 to F60 were mutated with the exception of the two zinc-liganding cysteines, C51 and C56 (A). A59 was mutated to valine; all other substitutions were to alanine. For comparison, we also mutated residues L71 and V73, which are situated between the two zinc-coordinating histidine residues that follow the DNA-binding α-helix.
We first tested approximately equivalent amounts of each mutant TFIIIA that had been synthesized (data not shown) for its ability to bind to the 5 S RNA gene and to recruit TFIIIC. All mutant proteins bound to DNA with a relative affinity similar to wild-type TFIIIA with the exception of TFIIIA(Y49A) and TFIIIA(F60A), which appeared to have a modest reduction in affinity for DNA (B, odd numbered lanes, open arrowhead). TFIIIA(K58A) was slightly compromised in its ability to recruit TFIIIC to the TFIIIA–DNA complex (B, lane 34) and TFIIIA(Y49A), TFIIIA(F50A), TFIIIA(Y53A) and TFIIIA(F60A) were severely compromised in their ability to recruit TFIIIC (B, lanes 6, 8, 14 and 38). As both Y49 and F60 are conserved residues of the zinc-finger motif and contribute to overall folding, the first zinc-finger of TFIIIA(Y49A) and TFIIIA(F60A) may be unable to adopt a stable tertiary structure and this may lead to the apparent defect in recruitment of TFIIIC. In the TFIIIA homologs of and other vertebrates, the two positions equivalent to F50 and Y53 are represented by isoleucine and phenylalanine, respectively (,). This conservation of hydrophobic residues in a region of the molecule that is predicted to be surface exposed is consistent with the notion that this region, and particularly F50 and Y53 of yeast TFIIIA, participates in a protein–protein interaction. In a homology model of the first zinc finger of yeast TFIIIA, these residues are indeed surface exposed (A). Additionally, we found that reducing the charge in this region, by mutation of D52 or D54 to alanine, did not affect the ability of the TFIIIA–DNA complex to recruit TFIIIC (B, lanes 12 and 24).
We next tested the synthesized proteins for their ability to support transcription of the 5 S RNA gene in the presence of a yeast fraction containing TFIIIC, TFIIIB and RNA polymerase III. All versions of TFIIIA were active in this assay (C, lanes 1–4, 6, 7 and 10–17). Because both TFIIIA(F50A) and TFIIIA(Y53A) were defective in recruitment of TFIIIC as assessed by EMSA (B, lanes 8 and 14), we next tested the activity of versions of TFIIIA in which these residues were mutated to glutamate, anticipating that the introduction of a charged residue might be more disruptive to the function of the putative hydrophobic interaction interface and thus have a more pronounced effect on the activity of TFIIIA than the alanine substitutions. We also examined the activity of TFIIIA in which the F50A and Y53A substitutions were combined. All three mutants retained the ability to bind to the 5 S RNA gene as assessed by EMSA (B, lanes 9, 19 and 21). TFIIIA(F50E) retained the ability to support transcription of the 5 S RNA gene, but TFIIIA(Y53E) was severely compromised and TFIIIA(F50A/Y53A) did not (C, lanes 5, 8 and 9).
With the use of the plasmid-shuffling protocol, we found that TFIIIA containing the double mutation F50A/Y53A was the only mutant from the collection of first finger mutants that was unable to support cell viability when expressed at either high or low levels (D, lane 9). As assessed by western blot analysis, this mutant protein accumulated to approximately the same level as did wild-type TFIIIA expressed from the high-copy plasmid (E, lanes 2 and 7). TFIIIA(Y53E), which was the only form of TFIIIA other than TFIIIA(F50A/Y53A) that was severely compromised for its ability to support transcription (C, lane 8), was able to support cell viability when expressed from the low-copy plasmid, but not from the high-copy plasmid (D, lane 8). This toxic effect of high-level expression of TFIIIA(Y53E) occurred only when cells were challenged to grow on 5-FOA-containing medium. This unusual observation is difficult to explain. It is possible that TFIIIA(Y53E) is modestly unstable and when present at high levels is inactivated by aggregation in a process that does not affect wild-type TFIIIA present in the same cells.
The ability of TFIIIA(Y53E) and TFIIIA(F50A/Y53A) to bind to the 5 S RNA gene, their inability to support transcription and the failure of TFIIIA(F50A/Y53A) to support cell viability suggested that residues F50 and Y53 were involved in the assembly of the transcription complex on the TFIIIA–DNA complex. For comparison, we also tested the effect of combining two other mutations, D54A and K58A, with F50A and Y53A; we also combined K58A with the more distant mutation L71A. These five double mutant versions of TFIIIA (F50A/K58A; Y53A/K58A; Y53A/D54A; F50A/D54A; K58A/L71A) retained the ability to bind to the 5 S RNA gene (A, odd numbered lanes; TFIIIA shift indicated with open arrowhead). Although all proteins, with the exception of TFIIIA(K58A/L71A), were defective in recruitment of TFIIIC (A, even numbered lanes; TFIIIC shift indicated by closed arrowhead and shown at higher exposure above the main figure), all of the double-mutant proteins were nonetheless able to direct transcription of the 5 S RNA gene in the presence of a yeast fraction containing TFIIIC, TFIIIB and RNA polymerase III (B) and were able to support cell viability when expressed at either a high or low level (C). The observation that the TFIIIA molecules containing these additional combinations of mutations retained activity further supported the conclusion that residues F50 and Y53 play a direct role in the recruitment of TFIIIC to the promoter.
Although it is well established that TFIIIA binds to the ICR of the 5 S RNA gene and recruits TFIIIC to the TFIIIA–DNA complex, the roles of individual zinc fingers in this process are less clear. Whereas specific contacts between zinc fingers and the ICR are a prerequisite for high-affinity, site-specific binding of TFIIIA to DNA, lower-affinity interactions could be important contributors to the overall architecture of the fully assembled transcription complex. Similarly, protein–protein interactions, in addition to being required for the recruitment of TFIIIC to the TFIIIA–DNA complex, could also promote conformational changes in the interacting proteins that are essential for the activity of the transcription complex. In this study, we have presented our initial analysis of the roles of individual zinc fingers of yeast TFIIIA in the assembly of an active transcription complex. We have used broken-finger versions of TFIIIA in which the third zinc ligand (i.e. the first of the two zinc-coordinating histidine residues) of a finger motif is mutated to asparagine in the context of the full-length protein (). This mutation is predicted to disrupt the folding of the finger by interfering with its ability to coordinate zinc () and this would abolish its ability to bind to DNA (). Examination of the products of partial proteolysis of broken-finger versions of TFIIIA indicates that a finger containing such as histidine-to-asparagine mutation is indeed unstructured (). In some cases, however, the fourth zinc-coordinating residue has been shown to be dispensable for folding. The ninth zinc finger of the recently characterized TFIIIA from , which contains an unprecedented 10 potential zinc fingers, lacks the carboxyl-terminal Zn-coordinating histidine residue. This module is nonetheless able to fold and function in DNA binding and TF activity (,). Based on studies with synthetic, truncated zinc -finger polypeptides that showed that these peptides are able to fold in the absence of a carboxyl-terminal histidine residue (), Schulman and Setzer speculate that a water molecule or a dithiothreitol-provided thiolate acts as the fourth zinc ligand in finger 9 of the homolog of TFIIIA.
The effects of disrupting a zinc finger could in part reflect an influence of the mutation on the properties of neighboring zinc fingers. Studies with TFIIIA suggest that disruption of any one zinc finger does not perturb the folding or function of adjacent fingers (,,). However, NMR-derived structures of a polypeptide containing the first three zinc fingers of TFIIIA bound to its cognate DNA site indicate that substantial packing interactions do occur between these zinc fingers when bound to DNA (,). Thus, without more rigorous analysis of the properties of the broken-finger versions of TFIIIA used in this study, we cannot be certain that the histidine-to-asparagine mutations are effectively and uniquely disrupting the structure and function of the mutated finger.
In this study, we compared the and activities of mutant TFIIIAs of . Both these assays showed that zinc fingers 1 and 7 play an essential role in assembly of an active transcription complex. Although our previous () and current analyses with synthesized proteins suggest that the absence of zinc finger 1 leads to a major reduction in affinity of the protein for the ICR, we were able to show with the use of bacterially expressed protein that TFIIIA(zf1) does indeed bind the 5 S RNA gene. This finding supported the idea that the essential role of finger 1 is not its contribution to DNA binding, but rather its recruitment of TFIIIC (). Indeed a comparison of bacterially expressed TFIIIA(zf1), TFIIIA(zf2) and TFIIIA(zf3) suggested that TFIIIA(zf1) bound DNA as efficiently as TFIIIA(zf2) and TFIIIA(zf3). The latter two versions of TFIIIA, however, but not TFIIIA(zf1), were able to support transcription of the 5 S RNA gene and cell viability when expressed from a high-copy plasmid under the control of a strong, constitutive promoter. The observation that TFIIIA(zf2) and TFIIIA(zf3) were unable to support cell viability when expressed from a low-copy plasmid with the gene under the control of its own promoter suggests that the function of TFIIIA that is compromised by disruption of finger 2 or finger 3 can be overcome by increasing the concentration of these proteins in the cell.
Our EMSA analysis of the bacterial proteins suggested that each of the first three zinc fingers contributed to recruitment of TFIIIC, with the contribution of the first finger being essential. Using an alanine-scanning mutagenesis approach, we identified residues within zinc finger 1 that are unlikely to participate in DNA binding but were necessary for the assembly of an active transcription complex. In particular, TFIIIA(F50A/Y53A) bound to the 5 S RNA gene, but did not recruit TFIIIC and did not support transcription nor cell viability, even when expressed at a high level. We speculate that the hydrophobic residues F50 and Y53 are surface-exposed, but become masked by an interaction with TFIIIC. Charged residues in this region of TFIIIA appeared less important for this predicted protein–protein interaction as the mutations D52A, D54A, D57A and K58A were without effect. Based on the known topography of the multisubunit TFIIIC on the 5 S RNA gene (,), the 138-kDa subunit of TFIIIC is a likely candidate for an interaction with the first zinc finger of TFIIIA. Other potential interacting partners are the 91-kDa subunit which has also been positioned at the 3′-end of the gene in close proximity to DNA around the transcription terminator, and the 60-kDa subunit, which has not been detected in photocrosslinking studies on either the tRNA gene or the 5 S RNA gene (,,). Although a double mutation within finger 1 was required to completely eliminate the TF activity of TFIIIA, we note that several other mutations that were combined with F50A or Y53A were without effect.
The transcriptional defect of TFIIIA(H240N), containing a disruption of zinc finger 7, is unlikely to be in its binding affinity to 5 S DNA or in the initial recruitment of TFIIIC; synthesized TFIIIA(zf7) showed some ability to bind to DNA and the first three zinc fingers of TFIIIA suffice for recruitment of TFIIIC (,). It is possible that a non-specific interaction of zinc finger 7 with DNA is essential for the appropriate positioning of TFIIIA in the fully assembled transcription complex. Site-specific DNA–protein photocrosslinking studies show that the interaction of TFIIIA with DNA is more extensive in the fully assembled transcription complex than in the TFIIIA–DNA complex itself (,). Another potential role for zinc finger 7 is in interactions with other TFs. TFIIIC can be recruited to the 5 S RNA gene by a TFIIIA that has mutations within the 81-amino-acid region that separates fingers 8 and 9, but the resulting TFIIIA–DNA complex is non-functional (). We have previously suggested that the 81-amino-acid region provides a second docking site for TFIIIC and that this is important in adjusting the topography of the complex. Thus, it is possible that the essential function of zinc finger 7 from yeast TFIIIA is in proper presentation of the 81-amino-acid domain for its putative interaction with TFIIIC. However, zinc finger 7 could also be playing a more direct role in the TF activity of the protein. For example, it could interact directly with TFIIIC. Such an interaction would most likely be with the τA domain of TFIIIC, which consists of the 131- the 95- and the 55-kDa subunits. This domain is essential for transcription activation and start-site selection by virtue of its role in positioning TFIIIB upstream of the transcribed region (,).
We found that both the and TF activity of yeast TFIIIA was tolerant to disruption of zinc finger 4, 5 or 6. Thus, as suggested previously (), the contacts that are made between zinc finger 5 and guanines +73 and +74 of the 5 S RNA gene are not essential. TFIIIA containing disruptions of both zinc fingers 4 and 5, zinc fingers 4 and 6 or zinc fingers 4 and 9 also supported cell viability, albeit only when expressed at high copy. In contrast, TFIIIA with a disruption of zinc fingers 5 and 6 as well as TFIIIAs that combined disruption of zinc finger 3 with disruption of finger 5, 6 or 9 were unable to support cell viability even when expressed at a high level. Although it is possible that some of these deficiencies reflect redundant functions shared between fingers, we consider it more likely that the inactivity of these proteins reflects a cumulative crippling effect.
In summary, our analysis of the effect of mutations in yeast TFIIIA on its activities and have shown that F50 and Y53 of the first zinc finger play a role in recruitment of TFIIIC and have revealed a requirement for the seventh zinc finger in assembly of a functional transcription complex. The fact that neither of these zinc fingers has an essential role in the amphibian TFIIIA (,) is consistent with the high rate of sequence divergence of TFIIIA () and the observed species specificity in the RNA polymerase III transcriptional machinery (). |
Translation of most eukaryotic cellular mRNAs is initiated by the binding of the eIF4F complex to the 7-methylguanosine cap. This is followed by recruitment of the 43S small ribosomal preinitiation complex and then scanning of the complex to the start codon. An alternative mechanism of translational initiation involves an internal ribosome entry site (IRES). An IRES is an RNA element that can facilitate ribosome recruitment independent of the 5′ cap structure of an mRNA. IRESs were first characterized in viral RNAs (,). Viral infection often leads to inhibition of cellular mRNA translation by inactivating the cap-dependent mechanism. For example, polio virus infection leads to cleavage of eIF4G (), an essential component of the cap-binding eIF4F complex, thereby inhibiting cap-dependent initiation. However, the presence of an IRES in the viral RNA allows efficient synthesis of viral proteins to proceed under these conditions.
A number of reports have described cellular mRNAs with IRES activity (). In many cases, these IRESs are thought to allow continued translation under conditions of cellular stress in which cap-dependent initiation is repressed. The mechanisms by which translation is initiated through IRESs in cellular mRNAs have not yet been characterized in detail. Moreover, for many of the putative cellular IRESs, it has been questioned whether or not the analysis was stringent enough to prove their existence (). For example, it is common to use a bicistronic vector system in which the putative IRES sequence is inserted between the upstream and downstream coding regions. Expression of the downstream cistron is indicative of an IRES encoded within the inserted sequence. However, possible artifacts could arise using this system if the insert has cryptic promoter activity or a cryptic 3′ splice site (). In either case a monocistronic mRNA encoding the downstream coding region would be produced that would be translated in a cap-dependent manner. It is clear that stringent tests need to be performed in order to conclude that a cellular mRNA has a functional IRES.
p27 is an inhibitor of cyclin-dependent kinases (CDKs) and is best understood for its role in blocking the G1 to S transition through inhibition of CDK2. Its expression is often downregulated in tumor cells and for many types of cancer there is a correlation between loss of p27 expression and poor prognosis (). Conversely, p27 is upregulated under conditions of growth factor deprivation and many other types of cellular stress. This is known to involve enhanced translation of the p27 mRNA (). Therefore p27 translation is typically highest under conditions where cap-dependent translation rates are reduced. This suggested the possibility that the p27 mRNA contained an IRES that could mediate cap-independent translation initiation. We tested this using a variety of methods, including bicistronic vectors, strong upstream stem-loops to block scanning and inhibition of cap-dependent translation by active 4EBP (). All of the results supported the conclusion that an IRES resides in the 5′-UTR of the p27 mRNA. These results have been independently verified by several other groups (). However, none of these reports completely ruled out the possibility of a cryptic promoter or splice site residing within the p27 5′-UTR. In fact, a recent publication has suggested that the p27 5′-UTR does display cryptic promoter activity and that this accounts for the putative IRES activity (). These authors used transfection of promoterless constructs and RNase protection assays to support their conclusions, which are clearly in disagreement with previous reports supporting cap-independent translation of the p27 mRNA. Given the conflicting conclusions reached by this report and previous reports, we have performed further experiments to examine the possibility of cap-independent translation through the p27 5′-UTR. The results do not support the existence of cryptic promoter activity or splice sites in the region, indicating the presence of an authentic IRES capable of mediating cap-independent initiation of translation.
The dual luciferase bicistronic construct pTKLL was constructed by ligating an XbaI/BglII fragment from pRL-TK (Promega) with an AvrII/BglII fragment from pGL2CAT. pTKLL-472, carrying the human p27 5′-UTR between the Renilla and firefly luciferase coding regions, was constructed by ligating the XbaI/BglII fragment from pRL-TK (Promega) with an AvrII/BglII fragment from pGL2CAT/Luc-472 (). To create the promoterless bicistronic constructs, pLL and pLL-472, pTKLL and pTKLL-472 were digested with MluI to remove the entire TK promoter from the constructs. The digested plasmids were then recircularized by ligation. To insert an IRE into pTKLL-472, a two step PCR strategy was used. First, two PCR reactions were performed using pTKLL-472 as a template. One reaction used primers TKLL-IRE1 and TKLL-9-5′ and the other reaction used primers TKLL-IRE2 and TKLL-1080-3′ (sequences available upon request). The products of these reactions were mixed as a template for the subsequent PCR amplification using primers TKLL-9-5′ and TKLL-1080-3′. The product of this reaction was digested with NheI and ligated with the large fragment resulting from NheI digestion of pTKLL-472. The final product (pTKLL-472-IRE) has an IRE 35 nt downstream of the transcriptional start site of the TK promoter. Monocistronic constructs based on the pGL4 backbone were prepared by ligating the p27 5′-UTR sequence, in either the forward or reverse orientation, into the HindIII site of PGL4.10[luc2] (Promega). Bicistronic constructs based on the pGL4 backbone were prepared by first ligating the p27 5′-UTR sequence, in either the forward or reverse orientation, into the HindIII site of pGL4.70[hRluc] creating pGL4.70[hRluc]-472 and pGL4.70[hRluc]-472R. The NheI/XbaI fragments from these two constructs were then ligated into the XbaI site of pGL4.13[luc2/SV40] creating pGL4.13-F/R-472 and pGL4.13-F/R-472R. Promoterless bicistronic constructs, pGL4.10-F/R-472 and pGL4.10-F/R-472R, were prepared by ligating the NheI/XbaI fragments from pGL4.70[hRluc]-472 and pGL4.70[hRluc]-472R into the XbaI site of pGL4.10[luc2].
Breast cancer cell lines were grown in Dulbecco's modified Eagle Medium (DMEM) containing 10% fetal calf serum, 100 U/ml penicillin and 100 μg/ml streptomycin. NIH3T3 cells were grown in the same medium but with newborn calf serum. All cells were incubated in a humidified atmosphere containing 5% CO. DNA constructs were transfected using GenePORTER (Gene Therapy Systems) or Dreamfect (Boca Scientific) according to the manufacturers’ instructions. Renilla and firefly luciferase levels were tested using the Dual-Glo™ Luciferase assay system (Promega). Where indicated, DFO was added to the cultures at a final concentration of 100 μM.
DNA templates for transcription were made by PCR using the Advantage 2 PCR kit (BD Biosciences). Vectors containing the human p27 472 nt 5′-UTR inserted in the forward and reverse orientations between the Renilla and firefly coding regions were used as templates for PCR. The 5′ primer contained a T7 5′ overhang and the 3′ primer contained a 30 nt polyA 3′ overhang. The templates were gel purified after PCR.
Messenger RNA (G-capped, polyA-tailed) was made using the mMessage Machine kit (Ambion) using 1 μg of the gel-purified templates as per kit protocol. The mRNA was checked on a formaldehyde gel and the concentration calculated. The mRNA was then transfected into the different cell lines using DMRIE-C transfection reagent (Invitrogen). Cells in 35 mm dishes were rinsed to remove any antibiotics. DMEM (1 ml/dish) was mixed with the DMRIE-C and then the RNA was added at a ratio of 2 μl DMRIE-C per 1 μg RNA. The mixture was vortexed and immediately applied to the cells. After a 4h incubation at 37°C with 5% CO, the medium was replaced with DMEM containing 10% FCS. The cells were incubated six additional hours and then harvested in 150 μl 1× Reporter Lysis Buffer (Promega).
MCF7 cells were transfected with pTKLL or pTKLL-472 one day before harvesting. Cells were lysed in RLN buffer containing 50 mM Tris-HCl pH 8.0, 140 mM NaCl, 1.5 mM MgCl, 0.5% NP-40, 1 mM dithiothreitol and SUPERase.in (Ambion). The cell lysates were centrifuged at ∼600 × for 10 min and the supernatants were used to extract RNA using the RNeasy Mini Kit (Qiagen). To remove DNA contamination, RNA samples were further treated with DNase and then DNase was inactivated using TURBO DNA-free (Ambion). The isolated RNA was then used as a template for reverse transcription using AMV reverse transcriptase and a 3′-specific primer and amplified by PCR using specific primers. The PCR product spans the entire Renilla luciferase coding region and the first 70 bp of the firefly luciferase coding region (see ).
DNA templates for preparing radioactive RNA probes were prepared by PCR using the Advantage PCR system (Clontech). The 5′ primer for each reaction included a T7 promoter sequence. The regions amplified for each probe are shown in . The templates were gel purified by agarose gel electrophoresis, ethanol precipitated, and then redissolved in sterile water at a concentration of 0.5 μg/μl. Labeled probes were synthesized using the T7 Maxiscript kit (Promega) by including 32P-UTP in the reaction. Probes were purified by electrophoresis using denaturing 4% polyacrylamide gels. Probes were estimated to have a specific activity of ∼5 × 10 c.p.m./μg.
One of the most common methods of testing for IRES activity is to use a bicistronic vector in which the putative IRES is inserted between two protein coding regions. We (,) and others () have used this assay for analyzing the activity of the p27 5′-UTR and the results have supported the conclusion that this region contains an IRES. Transfection of the bicistronic construct into cells should result in the transcription of a single mRNA that encodes both the upstream and downstream cistrons. Transcription of this message is dependent on the promoter sequence upstream of the first cistron. However, if the putative IRES harbors a sequence that can function as a promoter, then a monocistronic mRNA encoding only the downstream cistron will be produced. Since the reporter gene encoded by the downstream cistron is monitored by a very sensitive assay, even a low level of expression through a cryptic promoter in the putative IRES-containing sequence will be detectable.
One way to test for cryptic promoter activity in the putative IRES-containing sequence is to delete the authentic promoter upstream of the first cistron. This should eliminate or greatly reduce expression of the bicistronic message. Expression of both the upstream cistron, which is translated in a cap-dependent manner and the downstream cistron, which is translated in an IRES-dependent manner, should be affected. However, if expression of the downstream cistron is dependent on a cryptic promoter, then its expression should not be affected. Indeed, Liu . (), based on the use of promoterless bicistronic constructs together with RNase protection assays, have concluded that the sequence encoding the p27 5′-UTR has promoter activity. This conclusion is inconsistent with a substantial body of work suggesting the presence of an IRES in the 5′-UTR of the p27 mRNA.
To further explore the possibility of cryptic promoter activity, the sequence encoding the full-length p27 5′-UTR () was inserted into a bicistronic construct between an upstream cistron encoding Renilla luciferase and a downstream cistron encoding firefly luciferase (A). Expression of this construct (pTKLL472) is driven by the viral thymidine kinase (TK) promoter. A second construct was prepared in which the TK promoter was deleted (pLL472). Similar constructs lacking the p27 5′-UTR sequence were also prepared (pTKLL and pLL). These constructs were transfected into the breast cancer cell line MCF7 (B). Deletion of the TK promoter resulted in considerable loss of Renilla luciferase expression (compare pTKLL to pLL and pTKLL472 to pLL472). However, there was not a complete loss of activity, with both promoterless constructs retaining 15–20% of the activity of the upstream cistron relative to the construct containing the TK promoter.
As expected, neither construct lacking the sequence encoding the p27 5′-UTR (pTKLL and pLL) had substantial levels of firefly luciferase expression. In agreement with previous publications (,), the TK promoter-driven construct containing the p27 5′-UTR sequence (pTKLL472) showed high-level expression of the downstream cistron. However, deletion of the TK promoter resulted in only a ∼24% decrease in expression of firefly luciferase. This indicates that a major portion of the expression of the downstream cistron is dependent on the p27 5′-UTR but not on the TK promoter.
One possibility for the results above is that the p27 5′-UTR has promoter activity resulting in production of a monocistronic message encoding firefly luciferase. To examine this possibility we performed RNase protection assays. Four different probes, each complimentary to a different region of the mRNAs encoded by the bicistronic constructs, were utilized (A). Probe 1 is 180 nt in length and protects a fragment of ∼100 nt from cells transfected with pTKLL (C, lane 5) and pTKLL472 (C, lane 6). This corresponds to the size expected if the TK promoter is used to initiate transcription. A second major band of ∼130 nt is also detected in these lanes. This fragment could only result from transcriptional initiation in the vector somewhere upstream of the TK start site. Interestingly, there are also protected fragments from cells transfected with the promoterless constructs pLL (C, lane 7) and pLL472 (C, lane 8). These bands indicate a substantial level of transcription initiation within upstream vector sequences, even in the absence of any known promoter.
Probe 2 is 256 nt in length and protects a fragment of 227 nt within the upstream cistron encoding Renilla luciferase. This probe hybridizes with RNAs from cells transfected with each of the bicistronic reporter constructs, including those in which the TK promoter has been deleted (D, lanes 5–8). The intensity of the protected band is similar in all four samples. An identical band is observed in the control sample using transcribed bicistronic RNA (IVT RNA, lane 3), while no band is observed in the control lane using RNA from untransfected cells (lane 4). In titration experiments, ∼0.024 fmol of the transcribed RNA was easily detectable, indicating the high sensitivity of this assay. The results again support the conclusion that substantial transcription initiation occurs in vector sequences independent of the TK promoter.
Probe 3 is 390 nt in length and encompasses the 3′ end of the Renilla luciferase gene and 200 nt of the p27 5′-UTR. It protects a band of 166 nt in cells transfected with bicistronic constructs lacking the p27 5′-UTR (pTKLL or pLL, E, lanes 5 and 7). A band of approximately equal intensity is observed with both constructs, indicating a significant level of transcription independent of the TK promoter. Probe 3 should protect a fragment of 367 nt from cells transfected with bicistronic constructs carrying the p27 5′-UTR and a band of this size is detected both with (lane 6) and without (lane 8) the TK promoter. However, two additional strong bands of ∼210 and 125 nt are detected in these samples. Importantly, exactly the same pattern of bands is observed in the control sample using transcribed RNA. Thus, these bands do not arise from transcriptional initiation at a cryptic promoter within this region and are most likely to be the result of strong secondary structures in the RNA.
Probe 4 is 632 nt in length and spans the entire p27 5′-UTR and the 5′ end of the firefly luciferase coding region. This probe is expected to protect a 95 nt fragment from cells transfected with the bicistronic reporters lacking the p27 5′-UTR. A band of approximately this size is detected for both pTKLL and pLL (F, lanes 5 and 7). For full-length transcripts from the bicistronic constructs containing the p27 5′-UTR, the probe should protect a fragment of 584 nt. For cells transfected with either pTKLL472 or pLL472 a faint band with a calculated size of 588 nt is observed along with two stronger bands with calculated sizes of ∼531 and ∼494 nt and a series of minor bands of smaller size (F, lanes 6 and 8). The identical pattern of bands is observed when transcribed bicistronic RNA is used for the hybridization reaction (F, lane 3 and G, lane 4). Thus, the observed bands cannot be the result of alternative transcription initiation sites within the sequence encoding the p27 5′-UTR and there is no evidence of significant cryptic promoter activity within this region.
The results presented above indicate that the reporter constructs have significant transcriptional activity that is independent of the authentic TK promoter. Cryptic transcriptional regulatory sequences and promoter elements in vector sequences have been observed previously and are a common problem in reporter gene assays (). In an attempt to minimize this problem we generated a series of reporter constructs base on the vector pGL4. This vector (Promega Corporation) has been extensively modified to remove cryptic regulatory sequences and promoter elements in the vector backbone as well as in the luciferase coding region (). An upstream synthetic poly(A) signal/transcription pause site was also incorporated to impede read through from any transcripts initiated in vector sequences. We prepared both monocistronic and bicistronic reporters based on the pGL4 backbone. For the monocistronic constructs, the p27 5′-UTR was cloned upstream of firefly luciferase either in the forward or reverse orientation (A). The forward 5′-UTR construct expressed slightly higher luciferase activity than either the reverse 5′-UTR construct or the parent vector (pGL4.10, B). However, the activity of the forward 5′-UTR construct was expressed at a level nearly 600-fold less than a similar construct in which luciferase expression is driven by the SV40 promoter [pGL4.13(SV40)]. These results indicate that the p27 5′-UTR has minimal, if any, promoter activity in this setting.
pGL4-based bicistronic constructs were prepared in which the p27 5′-UTR, either forward or reverse orientation, was cloned between the firefly luciferase (upstream cistron) and Renilla luciferase (downstream cistron) coding regions. Constructs with and without the SV40 promoter upstream of firefly luciferase were prepared (A). Expression of the upstream cistron was almost entirely dependent on the presence of the SV40 promoter regardless of the orientation of the p27 5′-UTR (B). For the downstream cistron, elevated expression was only observed for the construct carrying the SV40 promoter and the p27 5′-UTR in the forward orientation. There was some expression of the downstream cistron in the construct carrying the p27 5′-UTR in the absence of the promoter, but this was only ∼15% of that observed with the promoter. With the promoter-driven construct carrying the 5′-UTR in the reverse orientation the downstream cistron was expressed at ∼6% of the level of same construct with the 5′-UTR in the forward orientation. Thus expression of the downstream cistron is dependent on the orientation of the p27 5′-UTR and the presence of the SV40 promoter upstream of the first cistron.
To further support the conclusion that both cistrons can be expressed from a single mRNA molecule, bicistronic vectors containing the p27 5′-UTR were cotransfected with siRNAs targeted to either the Renilla luciferase sequence or the firefly luciferase sequence. Two different pGL4-based bicistronic vectors were used in these experiments. One construct has the Renilla luciferase sequence as the upstream cistron and the firefly luciferase sequence as the downstream cistron (A) while the other has the firefly luciferase sequence upstream and the Renilla luciferase sequence downstream (B). With both constructs, siRNAs targeting either of the luciferase genes substantially reduced expression of both enzymes. It is especially important to note that siRNA inhibition of the upstream cistron also led to inhibition of the downstream cistron, further supporting the conclusion that the p27 5′-UTR mediates translation in a cap-independent manner.
Another means of testing for IRES activity is to directly transfect bicistonic mRNAs into cells. In this way, the processes of transcription and RNA processing are bypassed and there is no possibility of cryptic promoter activity or cryptic splicing through the putative IRES encoding region. Two bicistronic mRNAs were produced by transcription (, top). One mRNA included the p27 5′-UTR in the forward direction and the other mRNA included the complement of the p27 5′-UTR. Thus both mRNAs were of equal length and the distance between the two protein coding regions was identical. In addition, a 7-methylguanosine cap and a polyA tail of 30 nt was incorporated into both mRNAs. The mRNAs were transfected into two different breast cancer cell lines (, bottom). In both lines expression of the downstream cistron (firefly luciferase) was enhanced by the p27 5′-UTR in the forward direction, supporting the conclusion that this sequence is able to promote cap-independent translation.
Using northern blotting, we () and others (,) have previously characterized the transcripts produced by transfecting cells with bicistronic constructs containing the p27 5′-UTR. These experiments did not reveal any small transcripts that might have resulted from a cryptic splice site in the 5′-UTR. However, northern blotting may not be sensitive enough to detect low levels of short transcripts. Van Eden . () have utilized an RT-PCR strategy for detecting transcripts produced from bicistronic vectors. We have employed a similar procedure using one primer from just downstream of the TK promoter transcriptional start site and the other primer from within the firefly luciferase coding region (, top). Using this primer pair should result in the amplification of a ∼1635 bp fragment from the authentic bicistronic message from the reporter construct carrying the p27 5′-UTR (pTKLL-472) and a ∼1160 bp fragment from the control construct lacking the UTR (pTKLL). MCF7 cells were transfected with these bicistronic reporter constructs. As a control, another culture was left untransfected. Total RNA was isolated, and then RT-PCR was performed (, bottom). To control for contaminating plasmid DNA, a separate reaction was performed in which the reverse transcriptase was left out. A major band of the expected size was observed only in the lanes which included reverse transcriptase. Also, slightly larger RT-PCR products, of unknown origin, were observed for both the pTKLL and pTKLL-472 transfected cells. Importantly, in cells transfected with the p27 5′-UTR-containing construct, we did not detect any amplification products of a smaller size that were also not detected in the untransfected cells and the cells transfected with pTKLL. These results do not support the existence of a cryptic splice site within the 5′-UTR.
To further explore the possibility of cryptic splicing sites within the p27 5′-UTR we inserted an IRE 35 nt downstream of the transcriptional start site of the TK promoter in pTKLL-472 (, top). There is no potential 5′ splice site within this 35 nt sequence. This, and the proximity of the IRE to the transcriptional start site, makes it extremely unlikely that any splicing event would occur within this region. Thus any splicing event involving a cryptic 3′ splice site in the p27 5′-UTR would have to use a 5′ splice site downstream of the IRE. This, in effect, would transfer the IRE to a position upstream of the firefly luciferase coding region. If this occurred, expression of firefly luciferase would become sensitive to iron chelation, which leads to activation of iron-response protein (IRP). Active IRP binds to IREs and blocks cap-dependent translation.
The bicistronic construct containing the IRE, as well as the control construct lacking the IRE, were transfected into NIH3T3 cells (, bottom). The cells were treated with or without deferoxamine (DFO), a cell permeable iron chelator, and then harvested for luciferase assays. For the control construct lacking an IRE, treatment with an iron chelator had no effect on either Renilla or firefly luciferase activity. In contrast, treatment of cells transfected with the IRE-containing construct caused ∼50% inhibition of Renilla luciferase expression while having no effect on firefly luciferase expression. These data provide further support for the conclusion that expression of the downstream cistron does not result from the presence of cryptic splice acceptor site within the p27 5′-UTR.
Bicistronic reporter constructs are commonly used to study cap-independent translation through IRESs. However, this approach has been criticized because of possible artifacts from cryptic promoter activity (,) or a cryptic splicing site (,) residing within the test sequence. At least five previous publications have analyzed p27 5′-UTR sequences using bicistronic vectors and all have concluded that an IRES is present within this region (,). However, these findings were recently contradicted by the experiments of Liu . (). This group used promoterless bicistronic reporter constructs to support the conclusion the p27 5′-UTR has an absence of IRES activity, but rather that the DNA sequence encoding the p27 5′-UTR has cryptic promoter activity. This group further supported this conclusion using RNase protection assays to detect putative transcription start sites within the sequence encoding the p27 5′-UTR. In this report, we also used bicistronic constructs with and without a promoter upstream of the first cistron. Like Liu . () we found that significant expression of the downstream cistron was retained even after deletion of the TK promoter in the bicistronic construct carrying the p27 5′-UTR. However, our RNase protection assays demonstrated that there is considerable transcriptional initiation within vector sequences that is independent of the TK promoter. This vector-initiated transcription would be expected to inhibit expression of the upstream cistron because of numerous upstream AUG start codons and stop codons. The downstream cistron should still be actively expressed through cap-independent translational initiation mediated by the IRES. This is what was observed in our experiments.
Another important observation of our experiments is the detection of ‘extra’ bands in RNase protection assays using probes complimentary to the p27 5′-UTR. These bands could be interpreted as being the result of transcriptional initiation through cryptic promoters in the 5′-UTR sequence, as suggested by Liu . (). However, exactly the same pattern of bands was observed when transcribed RNA was used in the RNase protection assay. Thus, the bands cannot be due to promoter activity in this region and are most likely the result of very strong secondary structure in the p27 5′-UTR. The findings highlight the importance of including the transcribed RNA control when using RNase protection assays to map potential transcription initiation sites. This control was not included in the experiments described by Liu . () and it is likely that the putative transcription start sites in the p27 5′-UTR reported by this group are actually due to RNA secondary structure rather than transcriptional initiation.
Our experiments indicate a significant level of vector-initiated TK promoter-independent transcription in our original bicistronic constructs. A number of reports have previously described promoter activity in vector backbone sequences (). To surmount this problem, Promega Corporation recently developed the pGL4 series of luciferase reporter vectors (). This was done by removing dozens of consensus transcription factor binding sites and promoter modules. We used these redesigned reporter vectors to create monocistronic and bicistronic constructs carrying the p27 5′-UTR. With the p27 5′-UTR, monocistronic construct expression of luciferase was ∼600-fold less than that driven by the SV40 promoter. With the bicistronic constructs, expression of the downstream cistron was largely dependent on both the p27 5′-UTR and the presence of the SV40 promoter upstream of the first cistron. Thus, these experiments suggest that the sequence encoding the p27 5′-UTR has very minimal promoter activity and that the p27 5′-UTR is able to mediate cap-independent translation. This is further supported by the observation that siRNAs targeted to either the upstream or downstream cistron of these pGL4-based bicistronic reporter constructs leads to decreased expression of both cistrons.
A second potential problem associated with bicistronic vectors is the possibility of cryptic splice sites within the putative IRES sequence. Van Eden . () described an RT-PCR procedure that was able to detect products from cryptic splice sites with high sensitivity. In this report we used a similar strategy for analyzing the transcripts produced from the bicistronic reporter carrying the p27 5′-UTR. The results provided no indication that monocistronic mRNAs were produced by splicing within the 5′-UTR. Further support for a lack of cryptic splicing at sites within the p27 5′-UTR was provided by experiments in which an IRE was inserted just downstream of the transcription start site for the bicistronic mRNA. A splicing event involving the 5′-UTR and removing the upstream cistron (Renilla luciferase) would result in linking the IRE to the downstream cistron (firefly luciferase). If such an event was responsible for expression of the downstream cistron then this expression should be sensitive to treatment with iron chelators. However, we found that only the upstream cistron was inhibited by DFO.
Another test for IRES activity is to directly transfect bicistronic mRNAs rather than bicistronic DNA vectors. This approach completely avoids the processes of transcription and splicing. However, a potential problem with this approach is that the transfected RNAs are not processed in the nucleus and may not have access to all of the factors necessary to promote efficient translation. Nevertheless, we have observed significant expression of the downstream cistron that is dependent on the presence of the p27 5′-UTR (). Similar RNA transfection experiments have been carried out by Cho . () and their results are in agreement with ours.
In addition to the results described above, several additional findings indicate that the p27 5′-UTR is able to mediate cap-independent translation. Both Cho . () and Shi . () have demonstrated p27 IRES-dependent expression in a cell-free translation system. Kullmann . () found that the presence of the p27 5′-UTR makes expression of the downstream cistron resistant to the effects of a phosphatidylinositol-3 kinase inhibitor that represses global cap-dependent translation. Similarly, we () have found that p27 IRES-driven expression is not sensitive to the effects of rapamycin while Shi . () have actually observed an increase in p27 IRES activity following treatment with rapamycin. Rapamycin blocks mTOR kinase activity and specifically inhibits cap-dependent translation because of accumulation of active, hypophosphorylated 4EBP, an inhibitor of the cap-binding protein eIF4E. Finally, we have demonstrated that overexpression of a constitutively active 4EBP enhances cap-independent expression through the p27 5′-UTR as well as expression of the endogenous p27 gene ().
Finally, a highly pertinent finding was recently reported by Yoon . (). They found that cells expressing a mutant form of the gene had impaired translation of p27 mRNA and other mRNAs having IRESs. is mutated in the X-linked disease dyskeratosis congenita which is associated with a number of abnormalities including increased susceptibility to cancer. The DKC1 protein is a pseudouridine synthase that modifies ribosomal RNA. Ribosomes derived from cells carrying the mutation have lower levels of pseudouridine and are defective in translational initiation through cellular and viral IRESs but show no loss in global cap-dependent translation. Reduced expression of endogenous p27 protein and impaired ability of the p27 5′-UTR to mediate cap-independent translation, as shown by direct transfection of transcribed bicistronic mRNA, were observed in these cells ().
In summary, our results uncover a shortcoming in the use of ‘promoterless’ bicistronic constructs as a means of demonstrating cryptic promoter activity in a 5′-UTR coding sequence. They also demonstrate the importance of using transcribed RNA as a control when mapping putative transcription start sites by RNase protection assays. The data presented here and other recent publications strongly support the conclusion that the p27 5′-UTR has an authentic IRES. It will be important to determine the molecular mechanism of translation initiation through this element, how its activity is controlled, and if it has a role in the downregulation of p27 that is often observed in tumor cells. |
The standard version of the genetic code includes 61 sense codons and 3 stop codons. The 61 sense codons in an mRNA molecular are translated to code 20 amino acids in the course of protein synthesis. Except for methionine and tryptophan, the 18 most common amino acids are coded by two to six codons called synonymous codons. Previous studies have clearly indicated that synonymous codons are used with unequal frequency in a protein-coding sequence and choice of the synonymous codons is far from random both within and between organisms (). Consequently, the patterns of codon usage vary considerably among organisms, and also among genes from the same genome (). Codon usage has been known to present a potential impediment to high-level gene expression in and yeast (,). In humans, codon-mediated translational controls may play an important role in the differentiation and regulation of tissue-specific gene products ().
Virus genomes frequently have their codon usage significantly different from their host species (,). We observed that papillomaviruses (PVs), like many mammalian DNA viruses, have a significantly greater usage of codons with third position of A/T, manifesting a A+T rich genome (). Meanwhile, PV late (L1 and L2) genes frequently use the codons such as UUG, CGU, ACA and AUU that are rarely used in mammalian genes (,). Thus, when PV L1 and L2 genes are transfected into a wide range of eukaryotic cells, large amounts of the mRNAs can be transcribed, but no protein product is detected (), which indicates that expression of the late genes is subject to post-transcriptional regulation (). In HPVs, generalized substitution of isoencoding codons (mammalian preferred codons) with a higher G+C content allows expression of L1 and L2 proteins in different types of eukaryotic cells (,). The poor expression of several viral and other proteins has also been attributed to the unfavourable codon usage of their genes (). The codon usage of the native GFP gene is not adapted to mammalian ‘consensus’ and unmodified GFP is poorly expressed in mammalian systems(,). Modification of the codon usage of HIV-1 genes to those used by highly expressed human genes has been found to significantly increase HIV-1 structural protein expression ().
It is clear now that expression of hetero genes at translational levels can be significantly increased by synonymous codon substitution (). However, we observed that the effects of synonymous codon substitution on translation of the PV L1 mRNAs in the KC cultures were dependent on the cell differentiation (). In our study, the transiently transfection assay of the PV L1 gene expression in KC cultures was conducted at 48 h post-transfection as expression of a targeted gene has been generally overlooked in favour of early time points in other studies (,). In a recent study, we observed that hm variants in which were introduced with different sets of six consecutive codons down stream the AUG codon show different mRNA translation efficiency and have different duration of the GFP expression in transiently GFP plasmids-transfected mammalian cells. (,). Thus, it is possible that generalized substitution of synonymous codons may improve not only the translation efficiency of the viral genes, but also change the duration of its expression in KCs post-transfection. In this article, we address the following issues: (i) how long expression of the PV L1 genes can be detected in primary mouse and human KC cultures following transient transfection of the and PV L1 gene expression constructs; (ii) whether generalized substitution of isoencoding codons affects the duration of the L1 gene expression due to cell differentiation; (iii) if cell-free lysates prepared from the mouse KC cultures can be used for translation of both and PV L1 mRNAs and (iv) whether supplementation of exogenous aa-tRNAs affects expression of and L1 DNAs and mRNAs in different cell-free translation systems.
The four PV L1 gene expression plasmids: pCDNA3 HPV6b L1, pCDNA3 HPV6b L1, pCDNA3 BPV1 L1 and pCDNA3 BPV1 L1 used in the experiments have previously been described (,). Briefly, both BPV1 and HPV6b wt L1 ORFs are approximately 1.5 kb in length encoding 500 amino acids. The PV wt L1 genes show a strong codon usage bias, amongst degenerately encoded amino acids, toward 18 codons mainly with T at the third position that are otherwise rarely used by mammalian genes (,). We artificially modified BPV1 and HPV6b L1 genes such that the L1 ORFs were substituted with codons having G or C at the third position, which are preferentially used in the mammalian genome (). All of the native and codon modified PV L1 ORFs were sequenced and cloned into the mammalian expression vector pCDNA, which contains the simian virus 40 (SV40) ori (Invitrogen, Australia), to give four expression plasmids pCDNAHPV6b L1, pCDNAHPV6b L1, pCDNABPV1 L1 and pCDNABPV1 L1. Correct orientation of the ORFs relative to the plasmids was confirmed by enzyme restriction analysis. All the PV L1 constructs were sequenced to confirm the desired sequences before use.
Mouse KCs were isolated from new born mouse skin as originally described () with some modifications. Briefly, isolated KCs were grown as adherent cultures at a density of 6.5 × 10 cells/cm in freshly prepared 3:1 medium as previously described () for one day. The KC cultures were then transfected with the PV L1 gene expression constructs (pCDNA3 HPV6b L1, pCDNA3 HPV6b L1, pCDNA3 BPV1 L1 and pCDNA3 BPV1 L1) (,) using lipofectamine (Invitrogen, Australia) according to the manufacturer's protocol with some modifications. Following transfection, the DNA-transfected KCs continued to grow in KC-SFM medium with low calcium (0.09mM) and was collected for RNA and protein preparation at days 3 (D3), 6 (D6), 9 (D9) and 12 (D12) post-transfection.
Human KCs isolated from neonatal foreskins were cultured in KC serum-free medium (Life Technologies, Australia) up to confluence. The KCs were then passaged at a density of 6.5 × 10 cells/cm. After one day, the human KCs were transfected with the four PV L1 gene expression constructs as described above. Following transfection, the DNA-transfected KCs continued to grow in KC-SFM medium with low calcium (0.09mM) and collected for RNA and protein preparation at days 3 (D3), 6 (D6) and 9 (D9) post-transfection.
Here. 0.5 µg of RNA purified from cultured KCs transfected with the different PV L1 gene expression constructs was converted to complementary DNA using Oligo-dT primers and PowerScript reverse transcriptase (Promega, Australia) according to the manufacturer's protocol. We used 50 ng of each cDNA sample in 20 µl RT-PCR reactions using a Promega kit (Promega, Australia) supplemented with 3 mM MgCl and Platinum Taq Polymerase (Invitrogen, Australia). RT-PCR was undertaken using the Taqman system (AB Applied Biosystems, Australia).
The L1-DNA-transfected KCs were lysed in cell lysis buffer containing 2 mM of PMSF and sonicated for 40 s. Forty microgram (40 μg) protein samples from whole-cell extracts were separated by SDS-PAGE and blotted onto PVDF membrane. The blots were first probed with monoclonal antibody () against PV L1 protein, and then incubated with horseradish peroxidase-conjugated goat anti-mouse IgG antibody (Sigma, Australia). A chemiluminescence system (ECL kit; Amersham, Australia) was used to visualize the L1 signals. After probing for the L1 protein, the blots were stripped and relabeled with antibody against β-tubulin.
synthesis of the PV L1 proteins in cultured mouse KCs was studied by labeling with [S]-methionine. KCs cultured in six-well plates for one day (D1) were transfected with the PV L1 gene expression constructs (pCDNA
HPV6b L1, pCDNA
HPV6b L1, pCDNA
BPV1 L1 and pCDNA
BPV1 L1) using lipofectamine (Invitrogen, Australia) as previously described (). After transfection, L1 gene-transfected KCs continued to be grown in KC-SFM medium and at D2, D5 and D8 post-transfection were incubated in 2 ml of medium supplemented with 10 µCi of L-[S]Met (370 kBq) overnight. L1 gene-transfected KCs labelled with [S]-methionine at D3, D6 and D9 post-transfection were collected for protein analysis. Forty microgram (40 μg) protein samples, each in 1 ml of immunoprecipitation buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM phenylmethane sulphonyl fluoride (PMSF) (Sigma, Australia), 2 µg/ml benzamidine (Sigma, Australia), aprotinin 2 µg/ml and 1 µg/ml leupeptin (Auspep, Australia)) were immunoprecipitated using antibody against the L1 protein and then separated by SDS-PAGE. The SDS-PAGE gel was dried and autoradiographed.
In parallel experiments, intracellular localization of the PV L1 proteins in mouse KC cultures were examined. KCs cultured for one day were transfected with each of the four PV L1 plasmids, and then fixed and permeabilized with 85% ethanol at 3 and 9 day post-transfection. The fixed KCs were then blocked with 5% skim milk-PBS, probed with monoclonal antibody against the PV L1 protein, and followed by Cy3-conjugated anti-mouse immunoglobulin G (IgG) (Sigma, Australia). The L1 labeled KCs were further blocked with 5% skim milk PBS and probed with polyclonal antibody against involucrin (Sigma, Australia), followed by FITC secondary antibody. Nuclei were countstained by 4′,6′-diamidino-2-phenylindole (DAPI). Antibody-labeled KCs were examined by immunofluorescence microscopy.
One microgram of each plasmid DNA was incubated with 10 μCi of [S]methionine (Amersham, Australia) and 20 μl of T7 DNA-polymerase-coupled rabbit reticulocyte lysates (RRL) (Promega, Australia), with or without additional aa-tRNAs as indicated. Translation was performed at 30°C from 15 to 45 min and stopped by adding 2× SDS sample loading buffer. The L1 proteins were separated by SDS-PAGE on a 10% gel and blotted onto PVDF membrane. The blots were imaged using a phosphor screen and quantified by PhosphorImager analysis using the Imagequant program (Molecular Dynamics, USA).
Cell-free lysate was prepared from the mouse primary KCs cultured for one, three and eight days as described previously () with some modifications. Briefly, the cultured KCs were scraped into an extraction buffer containing 100 mM Hepes-KOH (pH 7.4), 120 mM potassium acetate (pH 7.4), 2.5 mM magnesium acetate, 1 mM dithiotreitol, 2.5 mM ATP, 1 mM GTP, 100 μM S-adenosyl-methionine, 1 mM spermidine, 20 mM creatine phosphate, 40 U of creatine phosphokinase (Sigma, Australia) per ml, and, importantly, an additional 100 mM sucrose. The collected KCs were then passed through a 25-gauge needle ten times, and the lysate was centrifuged at 4°C and 100 for 2 min. The supernatant was collected to prepare the mRNA-dependent KC cell-free lysate. The endogenous mRNAs in the prepared KC cell-free lysate were hydrolysed by incubating the lysate at 20°C for 10 min in the presence of 10 U/per ml of micrococcal nuclease () and 1 mM CaCl. The enzyme was inhibited by adding 2.5 mM ethylene glycol-bis(β-aminoethyl ether)-′,′-tetraacetic acid (EGTA, pH 7.0). The cell-free lysate was supplemented with 0.02 mM hemin before use.
italic
#text
italic
sub
#text
Total tRNAs were extracted and purified from KC cultures using a Qiagen kit as instructed by the supplier (). aa-tRNAs were produced for addition to either RRL system or KC lysate system as previously described (). tRNAs (2.5 × 10 M) purified from D0 and D8 mouse and human KC cultures were added to a 50 µl reaction mixture containing 10 mM Tris-acetate (pH 7.8), 44 mM KCl, 12 mM MgCl, 9 mM β-mercaptoethanol, 38 mM ATP, 0.25 mM GTP and 15 µl of rabbit reticulocyte extract. The reaction was carried out at 25°C for 20 min and 30 µl of HO was then added to the reaction to dilute the tRNAs to 10 M. The aa-tRNAs were aliquoted and stored at −70°C for later use.
Recently, we reported that expression of the PV L1 proteins from PV L1 genes is only detected in differentiated KCs and that the codon composition of the L1 genes determines the differentiation-dependent expression of the L1 proteins (). Here, we have examined whether continuous expression of L1 protein, as observed , can be recapitulated and how the gene codon composition mediates the continuous expression of the L1 genes. We transiently transfected mouse KCs that had been cultured for one day (D1 cultured KCs) with the PV and L1 gene expression constructs. The L1-tranfected mouse KCs were collected for RNA and protein preparation at D3, D6, D9 and D12 post-transfection. Progressive changes of the cell morphology of the L1-transfected mouse KCs from D3 to D12 post-transfection were examined by light microscopy (data not shown). The morphological changes observed for the L1-transfected KCs over time reflected cell differentiation that was confirmed by the increased expression of involucrin that is a marker of keratinocyte terminal differentiation (,) using immunofluorescence microscopy () and [S]-methionine labeling experiments. We have also used immunofluorescence microscopy to investigate whether and how the L1 proteins encoded by the HPV6b and L1 genes were expressed in the L1-transfected KCs at D3 and D9 post-transfection (). The L1 protein expressed from the L1 gene was detected in the L1-transfected KCs at both D3 and D9, but from the L1 genes only at D3 post-transfection (). Meantime, expression of involucrin in L1-transfected KCs at both D3 and D9 post-transfection was examined (). 20–30% of the L1-transfected KCs at D3 post-transfection exhibited weak involucrin signal, while all the L1-transfected KCs at D9 post-transfection had strong involucrin expression, suggesting that they proceed cell differentiation. ().
The transcripts of both and PV L1 genes were then analysed by quantitative RT-PCR. Gel electrophoresis revealed that the L1 mRNAs can be detected in KCs up to 12 days post-transfection (A). Given that the reported half life of the PV L1 mRNAs was approximately 5 h in mammalian cells () and mouse KCs (), the results suggest that the cultured KCs can continuously transcribe the transiently transfected PV L1 genes over the time period examined. Quantitative RT-PCR analysis revealed that transcription of both and PV L1 genes was distinctly down-regulated in the L1-transfected KCs as the cell differentiated. Meantime, the levels of L1 mRNAs from the L1 genes were significantly higher than those from the L1 genes over the time course (B), suggesting that replacement of the GC- ending codons in the PV L1 genes can improve their transcription efficiencies in KCs. In contrast, western blot analysis showed that the detected signals of the L1 proteins expressed from the two PV L1 genes increased significantly over time, with the highest levels of the L1 proteins detected at D9 post-transfection (B). Steady expression of the L1 proteins from the two PV L1 genes was detected at D3 and D6 (B) after which expression of the L1 proteins decreased dramatically and was not detectable in the HPV6b L1 transfected KCs at D12 post-transfection (B). The results suggest that the PV L1 proteins can be continuously detected in L1-transfected KCs for a certain period, and that codon composition of the L1 mRNA plays a key role in regulation of the continuous expression of the L1 proteins.
Considering that human KCs are the host cells of HPV infection, we also investigated the expression of the four PV L1 gene expression constructs in primary human KC culture. Both L1 mRNA transcription examined by quantitative RT-PCR and protein expression analysed by western blot were investigated. Similar results were obtained (A and B). Thus, the data produced from both mouse and human primary KCs suggest that the continuous expression of the PV L1 proteins is also post-transcriptionally regulated and associated with cell differentiation.
To investigate whether the PV L1 proteins continuously detected from both and L1 genes in L1-transfected KCs over the time course are due to continuous synthesis of the L1 proteins, rather than their intrinsic stability in cultured KCs. The L1-transfected mouse KCs maintained in normal KC medium were labeled overnight with [S]-methionine at D2, D5 and D8 post-transfection, respectively. The extended labeling time is required to incorporate sufficient radioactivity into the newly synthesized proteins to be detected. Following [S]-methionine labeling, the L1-transfected KCs were collected at D3, D6 and D9 post-transfection and analysed for the synthesis of involucrin, β-tubulin and L1 protein (A and B). Immunoprecipitation of the methionine-labeled involucrin and β-tubulin showed that the level of methionine-labeled involucrin steadily increased in cultured KCs over time accompanied by a slight decrease of the β-tubulin synthesis observed in the D9 KC culture (A). The increasing expression of the involucrin over time reflects the L1-transfected KC cultures’ proceeding cell differentiation.
Immunoprecipitation of the methionine-labeled L1 proteins showed levels of the newly synthesized L1 proteins expressed from the two PV L1 genes were low at D3 (B), but were significantly up-regulated at D9. In contrast, the L1 protein synthesis in the KCs transfected with the two PV L1 expression constructs was active at D3 post-transfection (B). The synthesis of the L1 protein by the HPV6b L1 expression construct was active in the L1-transfected KCs at D6, but not detectable at D9 post-transfection (B). The L1 protein synthesis in the KCs transfected with the BPV1 L1 expression constructs was significantly down-regulated from D3 to D9 post-transfection. To compare the relative levels of the newly synthesized L1 proteins from the PV L1 gene expression constructs to those from the PV L1 gene expression constructs over the time course, we examined the ratio of the newly synthesized L1 protein to the tubulin (L1/tubulin) by densitometric analysis of L1 protein and β-tubulin signals (C). The ratios of the L1/tubulin produced from the PV L 1-transfected KCs were 0.66 ± 0.15 for HPV6b and 0.66 ± 0.04 for BPV1 at D3 post-transfection and 0.92 ± 0.25 for HPV6b and 0.76 ± 0.01 for BPV1 at D6 post-transfection, but were significantly increased up to 1.65 ± 0.16 for HPV 6b and 1.75 ± 0.55 for BPV1 at D9 post-transfection. In contrast, the ratios of the L1/tubulin produced from the PV L1-transfected KCs were 1.70 ± 0.02 for HPV 6b and 1.34 ± 0.03 for BPV1 at D3 post-transfection and 1.87 ± 0.39 for HPV 6b and 0.82 ± 0.05 for BPV1 at D6 post-transfection, but were significantly decreased to 0.09 ± 0.01 for HPV 6b and 0.29 ± 0.02 for BPV1 at D9 post-transfection. The ratios between PV L1/tubulin and PV L1/tubulin were significantly different over the time course, except for BPV1 L1 at D6 post-transfection (C). The data suggest that the detected continuous expression of the L1 proteins from the PV L1 genes (C) was due to the continuous synthesis of the L1 proteins and that codon composition of the L1 mRNA sequences was a major determinant of the continuous synthesis of the L1 proteins.
Recently, we showed that translation of L1 from PV L1 genes in a RRL cell-free translation system was preferentially enhanced by aa-tRNAs from differentiated KCs (). Here, we have examined whether and how aa-tRNAs prepared from the mouse KCs cultured for one (D0) or eight (D8) days affect translation of L1 proteins from L1 genes in the RRL cell-free translation system. We showed that translation of the L1 proteins from PV L1 genes was much slower than that from the PV L1 genes over a time course of 15–45 min (A). Supplementation of the aa-tRNAs from D0 or D8 KCs had no distinct effects on the translation of the L1 genes (B). In contrast, introduction of exogenous aa-tRNAs from D0 KCs did not enhance translation of PV L1 genes, but addition of exogenous aa-tRNAs from D8 KCs significantly enhanced translation of PV L1 genes in a dose-dependent manner with optimum efficiency at 10M (C). Addition of 10M aa-tRNA from D8 cultured KCs produced strong L1 bands from both PV L1 genes at 15 min, while the corresponding L1 bands from the control reactions without added aa-tRNAs were weak (C). It was observed that 10M aa-tRNAs from D8 cultured KCs gave enhanced translation of L1, 1.59 ± 0.6-fold for HPV6b and 2.09 ± 0.8 for BPV1 (C, lower panel). The results suggest that the improved translation of native L1 mRNA and inhibition of the translation of codon modified L1 mRNA in differentiated epithelial cells is probably due to an association of gene codons with the cellular aa-tRNAs as the tRNA gel profiles were different between D0 and D8 cultured KCs (unpublished data).
To prove that aa-tRNAs from the human KCs would have similar effects, we have examined further how the tRNAs (10 M) prepared from both D0 and D8 human KC cultures affected translation of the L1 proteins from the PV and L1 genes in the RRL cell-free translation system. Again, introduction of exogenous 10 M aa-tRNAs from D8 human KCs dramatically increased the production of L1 protein from the two PV L1 gene constructs while they had few effects on expression of the two PV L1 genes (B). Surprisingly, supplementation of 10 M aa-tRNAs from D0 human KCs could significantly enhance translation of the two PV L1 genes, but not of the two PV L1 genes (A). The results indicate that aa-tRNAs from D0 and D8 human KC cultures differentially enhance translation of the PV and L1 genes in RRL translation system.
Next, we developed a cell-free lysate translation system from the primary mouse KC cultures to investigate whether and how the PV L1 mRNAs that were transcribed could be translated . We observed that the PV L1 mRNAs, whether they were transcribed from the or PV L1 genes, could be translated in the KC cell-free lysates when the L1 mRNAs were capped with m7 GpppG and the endogenous mRNAs in the cell-free lysates were completely hydrolysed (). We observed further that the PV L1 mRNAs could be translated in the D0 lysate, but the translational efficiencies were distinctly lower than those in the D3 and D8 lysates (). In contrast, although the PV L1 mRNAs could be translated both in the D0 and D8 lysates (), the translational efficiencies of the L1 proteins from the PV L1 mRNAs in the D8 lysate were dramatically decreased (). The data suggest that the PV L1 mRNAs have the translational preference to the cell-free lysates prepared from the differentiated mouse KC cultures, while the PV L1 mRNAs prefer to translate the L1 proteins in the cell-free lysate prepared from less-differentiated mouse KC cultures. The results are consistent with the western blot analysis of continuous expression of the PV L1 proteins from both and L1 genes in L1-transfected KCs (B).
We have also examined whether additional supplementation of the aa-tRNAs (10 M) prepared from D0 and D8 mouse KC cultures could enhance translation of the PV and L1 mRNAs in the D0 and D8 lysates (). Introduction of exogenous aa-tRNAs from D0 KCs only enhanced significantly the translation of the BPV1 L1 mRNAs in the D0 KC lysate at the 2 h time point (A and B). In contrast, addition of exogenous aa-tRNAs from D8 KC culture significantly enhanced the translation of both HPV6b and BPV1 PV L1 mRNAs in the D0 lysate at the two time points (1 or 2 h) (A and B). The aa-tRNAs from D0 KC culture significantly enhanced translation of the L1 mRNAs from both HPV6b and BPV1 in D8 lysate, while the aa-tRNAs from D8 KC culture only enhanced translation of the BPV1 L1 mRNA in D8 lysate at the 1 h time point (C and D). The results support the findings for the translation of the PV L1 genes in RRL translation system ( and ).
The PV life cycle is tightly linked to epithelial cell differentiation. Translation of the virus capsid proteins (L1 and L2) is restricted to the differentiated KCs although a range of polycistronic mRNAs including late open reading frames (L1 and L2 genes) can be detected in undifferentiated PV-infected KCs (). Thus, different posttranscriptional mechanisms that likely prevent capsid protein translation have been proposed to explain why measurable late gene mRNAs are not associated with production of late proteins in less differentiated cells. First, an AU-rich element in the 3' untranslated region of the late gene polycistronic mRNA of several HPVs has been reported to inhibit translation of the late genes in undifferentiated epithelial cells (,,). However, a range of strong constitutive promoters including retroviral LTRs, CMV and SV40 have been used to construct many authentic sequence L1 gene expression constructs from different HPV types in various laboratories (,). These L1 expression constructs that do not contain the inhibitory elements can easily produce the L1 transcripts when they are transiently or stably transfected to different mammalian cells. But efficient translation of these authentic sequence L1 gene constructs has not been reported. Second, a splicing enhancer or silencer in HPV genome is involved in the expression of the late genes. A splicing enhancer in the E4 coding region of HPV16 is required for inhibition of premature late gene expression (). Zhao . () reported further that an hnRNP A1-dependent splicing silencer in the HPV16 L1 coding region prevents premature expression of the L1 gene. Unfortunately, whether the reported splicing enhancer or silencer can induce protein expression of the late genes has not been reported. Third, blockage to translation of L1 mRNA has been attributed to sequences within the L1 ORF (,). Tan . reported first that most of the sequence of L1 ORF is inhibitory for L1 translation () A major inhibitory element located within the first 129 nucleotides of the L1 gene was then reported (). Again, expression of L1 and L2 proteins from HPV native L1 and L2 gene expression constructs has not been observed in transiently transfected mammalian cells (). Evidently, the mechanism of the tight control of PV late gene expression is not understood well.
Recently, by establishing mouse primary KC culture, we were successful, for the first time, to express the L1 proteins of two PV types (HPV6b and BPV1) by transient transfection of authentic or codon modified L1 gene expression plasmids (). KC cultures are heterogeneous, which contain two cell populations: replicating and terminally differentiating cells (). The terminally differentiating KCs that do not divide in culture progressively proceed differentiation and increase their cell size (). Considering the specific characteristics of the KC cultures, we used lipofectamine to transfect the L1 gene plasmid DNAs to the cultured KCs at 80–90% confluence instead of 50–70% confluence, which is recommended by the manufacturer. We then used flow cytometry analysis to examine the transfection efficiency of the KC cultures transfected with a GFP expression construct and observed that 30–40% of the cultured KCs expressed GFP (unpublished data). The L1 gene expression constructs used in our experiments were driven by a strong constitutive CMV promoter. Therefore, it is possible that the levels of expression of L1 mRNA and protein in the present experimental systems would be much higher than those ever achieved during a real infection. It is also clear that the claimed inhibitory elements in the L1 ORFs are not the active inhibitors of translation of L1 mRNAs in our previous and present studies.
In the present study, we have investigated continuous expression patterns of PV and L1 genes in cultured mouse and human KCs following a single transfection of the L1 plasmid DNAs. Regarding expression of the PV genes in different mammalian cells, previous studies have reported that synonymous codon substitution can overcome the translation blockage of the viral capsid gene mRNAs () and increase levels of E7 protein up to 100-fold due to highly efficient translation of the codon-replaced viral mRNAs (). A deeper understanding of the quantitative relationship between codon usage and protein levels must include studies on the relationship between codon usage and mRNA levels, and subsequently also the relationship between mRNA levels and protein productivity. Previous studies have reported that the mRNAs of and genes can be continuously expressed in primary cultured KCs, irrespective of their proliferative or differentiated state (). However, many small vectors with heterologous promoters are prone to vector loss and transcriptional silencing although continuous expression of a transgene at therapeutic levels is required for successful gene therapy (). Only the viral vectors containing U1 snRNA promoters may be an attractive alternative to vectors containing viral promoters for continuous high-level expression of therapeutic genes or proteins (). Therefore, we examined first whether and how mRNAs of the PV L1 genes, whether or not they are codon modified, were continuously transcribed in primary KC cultures over twelve days following a single transfection of the L1 gene expression constructs in D1 mouse and human KC culture. Both the and PV L1 genes cloned into the pCDNA3 vector were driven to express by the CMV promoter, they exhibited continuous mRNA transcription in mouse and human KC culture. But the levels of the PV L1 mRNAs were significantly higher than those of the PV L1 mRNAs in the cultured KCs over the time course investigated, irrespective of their differentiation status, indicating that codon usage can improve transcription efficiency of the viral genes examined in the cultured KCs. The PV L1 genes were substantially substituted with the codons ending with G or C, exhibiting GC manifest of their mRNA sequences because the PV L1 genes have a strong bias of using codons ending with A or T (). Thus, from the gene transcription perspectives, the PV L1 genes favour the codons ending with G or C in their gene sequences.
An interesting observation in the present study is that the expression of the PV L1 proteins from the two PV L1 genes was scarcely detected at D9 post-transfection. This expression pattern is distinctly different from that of the two PV L1 genes, which showed continuously up-regulated expression of the L1 proteins and increased the duration of translation of their mRNAs in both mouse and human KC cultures. It has been reported that genes known to be expressed during latency display codon usage strikingly different from the genes that are expressed during lytic growth in the Epstein-Barr virus (). In particular, the percentage of codons ending with G or C is persistently lower (about 20%) in all latent genes than in non-latent genes (). Also, the latent genes have codon usage substantially different from that of host cell genes to minimize deleterious consequences to the host of viral gene expression during latency (). Thus, the principal explanation to account for the disparity of continuous expression of the L1 proteins from and PV L1 genes in the present study appears to be their synonymous codon usage. The strong A/T codon usage bias of the PV L1 genes due to natural selection may be a main determinant for the increased duration of translation of their mRNAs in differentiated KCs.
Codon usage bias as the main determinant of efficient mRNA translation has been observed in many genes of (). In the present study, we showed that up- and down- regulated translations of either or PV L1 mRNAs in cultured mouse and human KCs were dependent on cell differentiation. Generally, differentiation-dependent translation of genes is determined by interaction of regulatory mRNA sequences with translational regulators (). Two proteins, hnRNPK and hnRNP E1/E2 from rabbit reticulocytes, mediate translational silencing of cellular and viral mRNAs in a differentiation-dependent way by binding to specific regulatory sequences (). Expression of 15-lipoxygenase (LOX) gene is typically regulated by a translational regulatory process. LOX mRNA accumulates early during differentiation, a differentiation control element in its 3′ untranslated regions confers translational silencing until late stage erythropoiesis (). The human cytomegalovirus (HCMV) major immediate-early (MIE) genes, encoding IE1 p72 and IE2 p86, are activated by a complex enhancer region that operates in a cell type- and differentiation-dependent manner (). Enzymes that are responsible for the translational control of gene expression have also been observed (,). A novel mechanism of translational gene regulation by which silenced mRNAs can be translationally activated by the c-Src kinase was reported (). In HPV, raft cultures treated with activators of protein kinase C can induce post-transcriptional changes in the late gene expression, which may occur through inactivation or down-regulation of splicing factors that inhibit use of the late region polyadenylation site, resulting in increased instability of late region transcripts (). As discussed above, different translational inhibitory mechanisms have been proposed to explain the block to PV L1 and L2 protein expression (,). However, neither the ‘block’ nor the inhibitory element is incomplete as how else would new infections/particles arise. Based on the systematical analysis in the present study, it is evident that the primary determinant mediating differentiation-dependent translation of the PV L1 mRNAs in KCs is the synonymous codon usage of the PV L1 genes, supporting that codon-mediated translational controls may play an important role in the differentiation and regulation of tissue-specific gene products in humans ().
A question arising from the present study is why the gene codon usage can differentially turn translation of the PV L1 genes on and off by a post-transcriptional regulatory process. In the bacterium and in yeast, the use of synonymous codons is strongly biased, comprising both bias between codons recognized by the same transfer RNA (tRNA) and bias between groups of codons recognized by different synonymous tRNAs. Therefore, highly expressed genes use a subset of optimal codons in accordance with their respective major isoacceptor tRNA levels (,). A model was developed for the co-evolution of codon usage and tRNA abundance explaining why there are unequal abundances of synonymous tRNAs leading to biased usage between groups of codons recognized by them in unicellular organisms (). Recently, a significant correlation was found between tRNA relative abundances and codon composition of Buchnera genes (). Evidently, the match between codon usage and tRNA that mediates translational regulation is a widely used mechanism of translational control. However, in multicellular eukaryotes, there is limited experimental data on tRNA abundance () and its function in regulating gene expression. Although a relationship between tRNA abundance and codon usage in has been reported (), no experimental evidence of how codon usage parallels tRNA content to regulate gene translation has been provided. Therefore, it is not clear how tRNA abundance in cells reflects the corresponding codons in mRNAs of the target genes to mediate mRNA translation. Recently, we have reasoned that if the PV L1 protein production by the native L1 genes, which is blocked in less-differentiated epithelial cells, is due to the restricted availability of the appropriate tRNAs, the tRNA pools should be different between proliferative and differentiated KCs. Consequently, tRNA profiles of the differentiated KCs , obtained by high pressure liquid chromatography (HPLC) separation, were distinct from those of undifferentiated cells in mouse and bovine (). Meanwhile, aa-tRNAs from differentiated mouse KCs significantly enhanced translation of native PV L1 genes in a rabbit reticulocyte lysate cell-free translation system (). In the present study, we observed that the tRNA profile of D0 KCs differs from that of D8-cultured KCs using tRNA gel electrophoresis (unpublished data). Functionally, aa-tRNAs from D8 cultured KCs significantly enhanced translation of the PV L1 genes or mRNAs in both RRL and D0 lysate translation systems, but only aa-tRNAs from D0 human KC cultures significantly enhanced translation of the PV L1 DNAs in RRL translation system and from D0 mouse KC cultures significantly enhanced translation of the PV L1 mRNAs in D8 cell-free lysate translation system. Previous studies have shown that tRNAs from uninfected cells can rescue the translation of global proteins in nucleopolyhedrovirus-infected Ld652Y cells (). Stimulating expression of GTase with tRNA and tRNA resulted in a 5-fold increase in GTase production in in which codon usage is highly biased and the tRNAs specific to codons AGA and AGG are extremely rare (). A correlation between tRNA content constraints and gene expression levels has been clearly demonstrated using the concept of optimal codons (). Here, the inability to express PV L1 protein from the native L1 gene sequence in undifferentiated KCs and from modified L1 gene sequence in differentiated KCs appears to be due to availability of the cellular tRNAs. Native PV L1 mRNAs may translate in undifferentiated KC lysate in which endogenous mRNAs (host mRNAs) were completely hydrolysed, which may have evolved to exploit the tRNA resources available for their efficient translation in differentiated KCs. The argument is strongly supported by our early work that showed that the problem of translation blockage of PV L1 genes in non-differentiated mammalian cells can be alleviated by gene codon modification () or by cellular tRNA population alteration (). |
One lesson of comparative genomics has been that numerous intricate and interdependent processes underlie organismal evolution. Even attempts to obtain an unambiguous picture of bacterial evolutionary relationships—organisms which reproduce in the absence of genetic exchange—have often been confounded with the emergence of the data that contradict accepted beliefs. For example, while bacterial phylogenies have historically used the highly conserved sequences of the small subunit ribosomal RNA, more complete genome sequence data has documented significant levels of gene transfer between the distantly related organisms, a strongly confounding influence on the elucidation of taxonomic relationships (). Beyond obfuscating the tree form of life, lateral gene transfer (LGT) mobilizes ecologically important genes among taxa, making it a potent force in the diversification and speciation of prokaryotes (,).
Change in gene inventory is a historical process. In the absence of experimental means to determine the evolutionary history of a gene, several complementary methods have been developed to infer the occurrence of gene transfer events, categorized as phylogenetic incongruency tests and parametric methods. The former identifies single gene topologies that deviate significantly from consensus relationships; aberrant phylogenies are considered to be the most reliable indicator of ancestral gene transfer events. Caveats for their use include biased mutation rates, improper clade selection, gene loss, segregation of paralogs and long branch length attraction (). More importantly, the success of phylogenetic methods depends entirely on the breadth and depth of the sequence database, which is especially evident in the inability to use these approaches to identify orphan genes of foreign origin. Lastly, phylogenetic studies may yield ambiguous results. For example, a recent survey of 13 species of γ-proteobacteria concluded that few LGT events took place among them, since organismal relationships inferred from the sequences of most genes failed to reject the consensus topology (); however, it was later reported that these same data failed to reject any topology, not only the consensus one (), suggesting that the phylogenetic signal was insufficiently robust to either accept or reject hypotheses regarding gene transfer.
In contrast, parametric approaches are based on the hypothesis that sequence features are similar within a genome but differ significantly between genomes. Genes which share a common set of features —that is, typical genes—are classified as native. In contrast, putatively foreign genes have atypical features inconsistent with the patterns reflected by the bulk of the genome; the features of these genes are posited to reflect the mutational proclivities of their donor organisms. While ancient gene transfer events would be difficult to detect as their atypical features ameliorate (,), genes of recent foreign origin are of special interest to microbiologists due to their role in recent changes in their ecological niche and/or metabolic repertoire. However, the sets of foreign genes detected by parametric approaches often differ significantly (); this may result from the different metrics being utilized, or different thresholds used to discriminate between ‘typical’ and ‘atypical’ genes. Such conflicts between methods have not been easy to resolve; until recently, the efficacy of parametric methods had been difficult to assess due to the lack of benchmark protocols ().
Yet these caveats are somewhat minor compared to an intrinsic weakness shared by nearly all parametric methods. Rather than falling into totally discrete groups corresponding to typical and atypical genes, compositional features of genes lie along a continuum (A). That is, there is no easily defined threshold beyond which atypical genes are clearly of foreign origin. Native genes may also be strongly atypical; for example, highly expressed genes have codon usage bias patterns that distinguish them from the majority of chromosomal genes (). As a result, arbitrary thresholds must be employed for declaring atypical genes to be of likely foreign origin, where choice of threshold balances Type I and Type II errors (). Conservative thresholds lead to rare misclassification of native genes as foreign, at the expense of more falsely declared native genes. Liberal thresholds detect more foreign genes, but also incur more false predictions (i.e. native genes misclassified as foreign). Advances in the efficacy of parametric methods critically depend upon a decoupling of Type I and Type II errors, so that genes that lie in the twilight zone (the somewhat atypical native genes or weakly atypical foreign genes) may be robustly classified as either native or foreign. To accomplish this goal, we use two features of gene transfer in bacterial genomes. First, many alien genes are introduced in genomic islands; here, large number of genes arrive from a single donor genome and are physically adjacent. Second, the non-random distribution of donor genomes for any one recipient () increases the likelihood that foreign genes may resemble each other even if they arrived in separate transfer events.
Using this information, we have implemented here a 2-fold approach for foreign gene identification. First, we employ a novel gene clustering method based on Jensen–Shannon (JS) divergence measure. Contrary to the arbitrary thresholds used by existing parametric methods, this approach segregates genes into distinct classes within a hypothesis testing framework. In this way, we identify foreign genes not solely by their incongruence with the majority of genes in the genome but also by their similarity to each other (B). Yet even here, we would expect that somewhat atypical native genes may be misclassified as alien, and vice versa. To escape the limitations imposed by any single threshold in classifying genes with ambiguous features, we use genome position information to reassign genes between native and foreign classes based on the characteristics of physically adjacent genes (C). The performance of this approach was assessed on a test platform of artificial chimeric genomes () and then applied to well-understood K12 genome.
The complete genome sequences of the prokaryotes DSM4304, 168, R1 chromosome I, SCRI1043, K12, Rd KW20, DSM2661, FA1090, GMI1000, serovar Paratyphi A str. ATCC 9150, 1021, . PCC6803 and MSB8, O1 biovar eltor str. N16961 chromosome I, and KIM were obtained from GenBank. Protein-coding genes were extracted using the coordinates provided in the annotation.
The JS divergence between two probability distributions and of a discrete random variable is defined as (),
(.) is the Shannon information entropy defined as
DNA sequences are represented by alphabet = (A,C,G.T). To measure the compositional difference between two DNA sequences and of length and , respectively, the probability distribution ( = 1, 2) is represented by the relative frequency vector {(), ∈ }, () = ()/() is the count of nucleotide in sequence .
To assess the statistical significance of this measure under the null hypothesis that sequences and are similar, that is, both sequences are generated from the same probability distribution, we use the analytical approximation of the probability distribution of JS that was shown to follow a χ distribution function [Arvey,A., Raval,A., Azad,R.K. and Lawrence,J.G., unpublished data; (see also Section IV(C) in ()]. For asymptotically large values of ,
The -value for the test is thus obtained as 1-Pr{JS ≤ }.
We employed the JS divergence in an agglomerative hierarchical clustering method to measure the dissimilarity (or similarity) between genes or gene classes. The clustering algorithm begins with single gene classes. For each iteration, each pair of classes is considered and the JS distance between classes is measured. If the -value computed for the JS distance between closest classes is less than pre-set significance threshold, the distinction between the two classes is deemed statistically significant precluding the merger of these classes; otherwise the classes are merged. The algorithm is repeated recursively until the distinction between all classes is statistically significant, preventing any further class merger. The frequency vector for multigene classes is the mean frequency vector of the constituent genes and its size is the mean size.
To quantify the compositional difference between genes, the DNA sequence of a gene is represented by a 12-symbol alphabet = { = 1–3} accounting for nucleotide identity and the three codon positions. A 48-symbol alphabet representation of DNA sequence accounting for the dinucleotide identity and the codon positions was also used. We termed the respective JS(, ) as JS-N and JS-DN.
The JS() for codon usage bias is termed JS-CB.
To evaluate the performance of our proposed method, we constructed artificial genomes using generalized hidden Markov models (HMMs) (). Briefly, genes making the core of a genuine genome—those representing the spectrum of mutational signatures native to that genome—are obtained by a gene clustering algorithm based on Akaike information criterion [AIC ()]. These genes are segregated into distinct classes using a -means clustering algorithm employing relative entropy as distance measure to decide the algorithm convergence. Multiple gene models trained on these gene classes are then used in the framework of a generalized HMM to generate an artificial genome representing the variability found among genuine core genes. A chimeric artificial genome is obtained as the mosaic collection of genes sampled from different artificial genomes. To a chosen recipient artificial genome, we inserted at a random position one or more contiguous genes selected randomly from a sample of donor artificial genomes. Insertion is carried out recursively until a chimeric genome of a desired composition is obtained. Because the evolutionary histories of genes are known precisely in these genomes, and because the genes fairly represent the variability seen in genuine genomes, chimeric artificial genomes serve as valid test beds for assessing the parametric methods of gene transfer detection ().
Two sets of 4000-gene artificial genomes were created. Artificial Genome I had a core of 3000 genes (75%) representing an artificial genome; the remaining genes were acquired from five different donors— (7%), (5%), Rd (3%), (6%), (4%). Artificial Genome II had a core of 3400 genes (85%) representing an artificial genome with the remainder acquired from 10 donors— (1%), (1%), (2%), Rd (2%), (1%), (1%), (2%), (2%), PCC6803 (1%) and (2%).
Compositional properties of genes rarely lie as points about a single defining set of parameters; rather, they fall along a range of parameters (for example, of codon usage bias). At high stringency (significance threshold), the JS clustering algorithm may cause native genes, or the genes from a donor organism, to be sorted into more than one class representing this spectrum; relaxing the stringency may raise the misclassification error and lead to the undesirable merger of classes of genes. Gene-context information can be used to identify classes of genes that may have originated from the same source organism. If a gene belongs to class whereas the two flanking genes are grouped in class , we define this adjacency as a link between classes and .
If () exceeds an established threshold, the genes comprising the two classes are physically associated within the genome, perhaps due to common origin; the genes from these two entropic classes are assigned to a single logical class.
In the next post-processing step, we again use the genome context information of genes to refine the composition of gene classes. Here, a gene is reassigned to the class of its neighbors only if it plausibly lies within that class. Specifically, if a gene belongs to logical class whereas the immediate neighbors of this gene are grouped in logical class , this gene is reassigned to class , if and only if it is either not atypical or only slightly atypical with respect to class (determined by slightly relaxing the stringency) as inferred within a hypothesis testing framework.
Other parametric methods for foreign gene detection were coded as follows. Karlin () suggested dinucleotide bias as a genome signature, = /, assessed through the odds ratio, is the frequency of the dinucleotide XY and is the frequency of the nucleotide X. If the dinucleotide average relative abundance difference between gene and genome (average over all genes) defined as exceeds an established threshold, the gene is classified as foreign. The Karlin's Codon Usage Difference () between gene and genome was quantified as , is the frequency of codon normalized in the respective synonymous codon group is the normalized frequency of amino acid . If (|) exceeds an established threshold, is classified as a foreign gene.
Hayes and Borodovsky () developed a -means gene clustering algorithm using Kullback–Leibler distance, , as a measure of codon usage difference between gene and cluster to decide the algorithm convergence ( is the size of the th group of synonymous codons, denotes the normalized frequency of codon as described above). Initial seeds for typical and atypical clusters were obtained from GeneMark predictions, each gene was reassigned to the cluster with the closest cluster center determined through , cluster centers were recomputed and this process was repeated until convergence. Our recently developed AIC-based gene clustering algorithm is similar in spirit to our proposed JS divergence based gene clustering method, gene classes are populated in a hierarchical agglomerative clustering fashion, however, here clustering is decided in a model selection framework. We used a generalized version of AIC, , as a stopping criterion for clustering [ is the maximum likelihood, is the number of free parameters, is the sample size and is the tuning parameter ()]. Garcia-Vallve . () used multiple metrics, namely G+C content, codon and amino acid usage to compile putative horizontally transferred genes in their HGT-DB database. The machine-learning method Wn-SVM uses a one-class support vector machine for identifying alien genes (). Alien-Hunter detects putative alien genes using variable order motif distributions ().
For assessing the performance of the parametric methods in identifying the foreign genes, we obtain the misclassification error rates as Type I error = FN/(TP + FN) and Type II error = FP/(TP + FP), where TP = true positives, FN = false negatives and FP = false positives (note that conventionally TP, FN and FP are interpreted in accordance with a null hypothesis testing, here without loss of generality, positives and negatives respectively mean the genes declared as foreign and native by a method). Type I error is the percentage of foreign genes that were misclassified as native, whereas Type II error is the percentage of predicted foreign genes that were actually native. The average value of Type I error and Type II error was used as a single error rate parameter. JS- or AIC-based clustering methods yield one class comprising the majority (60–95%) of genes in the genome; the remaining genes are distributed among several smaller classes. Native genes are represented by the largest class while the foreign genes are, by definition, identified as the residents of all other classes. Artificial genomes contain ‘foreign’ gene with known sources; donor-specific misclassification error rate is defined as the percentage of genes from a donor organism misclassified as native genes. Classes generated by the clustering methods were assessed using two parameters: class abundance and class purity. Class abundance is the percentage of genes from a source organism identified correctly in a respective class (the sensitivity with respect to the class). Class purity is the percentage of genes in the class correctly assigned to that class (the specificity with respect to the class).
We posit that both native and foreign genes in bacterial genomes will fall into multiple classes. That is, foreign genes will not only be atypical, but they may also be segregated into groups of similar genes (). As a result, the identification of atypical genes can rely both on their dissimilarity to native genes as well as on their shared characteristics. These features may help delimit the boundaries between typical genes and sets of atypical genes. We employed our proposed JS gene clustering methods to segregate genes in bacterial genomes into classes. As described in the Materials and Methods section, all genes from a genome were initially assigned to single-gene classes (, row 1). The most similar classes merged recursively until the classes were distinct from each other at a given significance threshold. A trade-off between Type I and Type II errors is evident by changing the stringency used to discriminate the gene classes. As the number of classes decrease, more native genes are identified correctly, but more foreign genes are incorrectly deemed native (). Clustering stops prematurely at high significance thresholds, generating numerous potentially similar classes; at low significance thresholds the distinction between classes is high, however, the likelihood of undesirable merger of classes increases. Optimum performance is defined as the threshold setting which minimizes the mean error.
We used three criteria for class merger: codon position specific nucleotide composition (JS-N) and dinucleotide composition (JS-DN) as well as codon usage bias (JS-CB); their relative performance is shown in . Depending on genome composition and the threshold parameters, between 6 and 11 major classes were typically obtained; additional classes contained very few genes. For all methods, decrease in Type I error caused an increase in Type II error and vice versa (). JS-based clustering methods, which form many atypical gene classes, generally outperform other methods which sort genes into a single foreign gene class (; ), including Karlin's dinucleotide () and codon usage bias methods (), and Hayes and Borodovsky's -means method (). Gene classification methods based on the AIC, which also allow for the assignment of atypical genes to more than one class (), also performed well.
Because native genes show a spectrum of compositional properties, we must decrease the significance threshold to allow them to join the large class of native genes (). Yet this simple change in threshold may also allow foreign genes to be included, leading to increased Type I error. This coupling of Type I and Type II errors can only be circumvented if other information is used to perform class merger. That is, we must only merge classes of weakly atypical native genes to the largest class, while leaving classes of weakly atypical foreign genes separate. To do this, we rely on gene position to perform a differential class merger, termed class reassignment.
For reassigning foreign genes misclassified as native and native genes misclassified as foreign, we used genome context information, a technique developed by Lawrence and Ochman (,). There, reassignment of native or foreign genes was performed through human intervention by examining the class identity of genes flanking ambiguously assigned, weakly atypical genes. If a small number of genes from an otherwise contiguous set of foreign genes were identified as native, they were reassigned into the foreign class by invoking the rule of adjacency. In our case, positional information can be used to map the JS methods’ generated gene classes originating from the same source organism and vet the incorrect assignments of genes to the classes. We used the class linking measure () to merge classes obtained at strict stringency on the basis of relative positions of their constituent genes and not on their entropic divergence. If () exceeded an established threshold, the classes and were merged; this process was iterated until the merger of any two classes was not legitimate. The threshold was set to 0.3 after testing on a number of data sets.
The composition of classes was also refined using gene context information. We examined genes that were flanked by genes both belonging to a different class (B); if such a gene was reasonable member of that different class—that is, if it had sufficient affinity for that class inferred within a hypothesis testing framework—the gene was reassigned to that class. We repeated this process until no gene reassignment was significant. This process would serve to purify classes, enabling them to add members that were sufficiently different so that they were misclassified, an inevitable result of classes of genes which overlap in sequence features (B). For Karlin's methods, the refinement of the predictions using positional information was done in a multi-threshold approach, where genes whose features lay between the ‘clearly typical’ and ‘clearly atypical’ boundaries were reclassified in this way. Although Hayes and Borodovsky's clustering method does not use a threshold to discriminate between typical and atypical genes, we used the distance from class center to discriminate between genes which are strongly associated with the class and those which are weakly associated.
The use of gene context information reduced remarkably both the Type I and Type II errors for all three JS methods, visualized in as curves that approach the intersection of the axes; the JS-CB algorithm showed the most improvement. The JS-CB method also balanced the Type I and Type II errors better than other methods at the optimal thresholds, and the variances in the errors generated by JS-CB were much lower compared to other clustering methods (). Note that the inclusion of positional information makes the JS-CB method more efficient than the JS-DN approach which had earlier yielded consistently lower misclassification error rates; the decrease in the misclassification error rates of the JS-CB method is nearly 3-fold on Artificial Genome I and 2-fold on Artificial Genome II. Results were not improved if class refinement preceded class reassignment (data not shown).
Using positional information, we also see that the margin of improvement in JS-based classification methods was higher than in AIC-based gene classification methods. The variances in error rates of both AIC-CB and AIC-DN methods were also much higher. The AIC methods are thus sensitive to the thresholds used; as a result, an optimal threshold, one which minimizes the error values with significantly low variances, is difficult to realize. The use of positional information also increased the accuracy of Karlin's, and Hayes and Borodovsky's methods, although not to the same degree as gene-clustering algorithms (). That is, positional information became more useful when atypical genes were assorted into multiple classes.
We would anticipate that genes with markedly different compositional properties would be the easiest to detect as foreign. We examined the performance of JS methods as a function of donor genome and found that all three discriminant criteria served well in detecting gene transfer from four of the donors in Artificial Genome I, where the misclassification error rate (percent of genes from a donor genome misclassified as native) was less than 5% (). Genes from the artificial genome were misclassified at much higher rates by all methods, with the JS-CB method performing best. These results show that the error rates are also functions of the discriminant criterion and the gene number. We anticipate that combining the methods using more than one discriminant criterion may compensate for any one metric weaknesses, as was seen for AIC-based methods ().
By their very nature, the accuracy of JS methods in detecting classes of atypical genes increases as the number of genes in each class increases. To determine if the strong performance of the JS methods in detecting most atypical genes was a result of gene numbers, we compared results for Artificial Genome I to those for Artificial Genome II, wherein fewer genes were selected from each of a greater number of donor genomes. Here, all JS methods performed well (<10% misclassification error) in discriminating foreign genes from six donors; only JS-CB was most consistent in classifying correctly genes from and and none of the methods performed well on genes from (). In most cases, the methods generated higher misclassification error rates than those for Artificial Genome I. The performance of the AIC gene clustering methods in classifying the genes of donor genomes is shown in Supplementary Table 1. While AIC-DN performed better than AIC-N and AIC-CB, it could not classify the majority of the and genes correctly. Overall, JS-CB emerged as the most effective method in classifying the genes as foreign or native, being least affected by the identity of the donor genome.
To assess the efficiency of JS methods in grouping genes contributed by different donor organisms, we examined two accuracy parameters—class abundance and purity—after the gene classes were refined using positional information (). While all the JS methods grouped the genes that have arrived from 4 donors in the Artificial Genome I into distinct classes, JS-CB performed the best as measured by both accuracy parameters. For the most part, all JS methods generated classes with a very high degree of purity (>90%); class purity was highest where genes arrived from compositionally distinct donors. Only JS-CB could group well the genes (class abundance and purity were both greater than 80%). Even when Type II error was relatively high—for example, when many genes were misclassified as ‘native’ by JS-N and JS-DN—those genes that were identified as ‘foreign’ were placed into a relatively pure class (∼80% genes). With Artificial Genome II, the performance of JS-N dropped significantly, with only three gene classes having class abundance and purity above 70%. The JS-N method grouped and genes (class abundance and purity in the range of 60–70%), but it failed to cluster genes originating from five genomes, even if they were identified as foreign ( and ). JS-DN performed better, grouping genes from seven donor genomes. The JS-CB method performed even better, classing the genes of eight donors very well (class abundance and purity both exceeded 70%), less efficiently, and genes poorly. These results show that the JS methods, particularly the JS-CB, can be powerful tools for identifying genes that originate from the same donor organism.
The performance of the AIC gene clustering method in classing the genes from donor genomes is shown in Supplementary Table 2. None of the AIC methods seemed proficient in grouping the genes from Artificial I. On Artificial II, the performance went worse with AIC-DN grouping together majority of the genes of only four genomes. The performance of AIC-CB was no better; AIC-N, however, performed comparably with JS-N. Overall our analysis shows that JS methods are most consistent and efficient in classing the genes in genomes.
As expected, classification accuracy increases with the number of genes in each class. In artificial genomes, classes are represented by genes from a single donor species. Yet genuine genomes will likely not receive multiple transfer from any single donor, although it may experience multiple events from related donors. Because LGT is believed to occur more frequently among evolutionary closely related organisms (,), the array of donor genomes may indeed be non-random. More importantly, differences in genome composition increase as a function of the evolutionary distance between species, and related genomes are compositionally similar. As a result, JS methods should sort genes from related donors into one or few classes.
To assess how the effectiveness of the JS entropy clustering depends on the evolutionary distance between donor and recipient genomes, we performed simulated gene transfers into an artificial genome using genes from genomes modeled after the related γ-proteobacteria as donors (). , being the closest to the among the five donors, had most of its genes (∼80%) misclassified by JS-CB. That is, JS methods could not distinguish genes from genes, so that these genes would form a single class if they were both introduced into a foreign genome. Genes from artificial genomes constructed from other members of the Enterobacteriaceae were also found to be compositionally similar to genes (), while genes from more distantly related γ-proteobacteria were distinguished more efficiently. Thus, JS methods do not form species-specific classes; rather, genes from any member of a bacterial family would be placed into a single compositional class.
Although artificial genomes mimic the genic complexity of genuine genomes, it is difficult to model the positional distribution of foreign genes. Therefore, it is unclear if the advantages of using positional information will be seen in genuine genomes. To examine this, we applied the best performing method, JS-CB, to the well-characterized K12 genome. A set of putative horizontally transferred genes (‘HT’ genes henceforth) was defined as those present in the K12 genome but absent from the LT2 genome. This yielded 891 HT genes (very short genes, those with length <300 nt, were not considered). Given the vagaries of genuine data, this set is known to be imperfect in two ways. First, foreign genes acquired before the divergence of and will be excluded, leading to some foreign genes being mislabeled native. Second, homologs of native genes lost from the genome will be included in our test set.
Using this set of HT genes as a guide, we observed that the JS-CB method performed well (). At a baseline significance threshold of 0.05 (Supplementary Table 3), the number of false predictions was high due to several small classes of native genes misclassified as foreign (mean error ∼49%). Class reassignment caused a significant drop in the number of false predictions at the cost of fewer true predictions (mean error 42%). Upon class refinement, the mean error decreased further to 39.6%. An equivalent number of predicted foreign genes was obtained at a significance level of 10 without using positional information, but the mean error was 59.5%. Thus the use of positional information results in remarkable improvement in the HT gene detection in genuine genomes, as predicted from the artificial genome simulations.
Using this benchmark set of putative foreign genes, the JS-CB method also outperformed other parametric methods for foreign gene detection (). Karlin's codon usage method () identified only 50 genes (>600 nt) as the laterally transferred candidates; while specificity was high, sensitivity was very low. Garcia-Vallve . () have compiled 306 putative HT genes of K12 in their HGT-DB database, their method was also not found to be sensitive. We also tested two recently proposed parametric methods for LGT detection, Wn-SVM () and Alien-Hunter (). Alien-Hunter had a comparatively lower Type I error as it identified highest number of foreign genes among all methods but this came at the cost of very high number of false predictions. Wn-SVM generated less false predictions but could identify fewer foreign genes We also tested the best performing method among the AIC methods (AIC-DN) which performed slightly better than Wn-SVM, generating less of false predictions at equivalent number of true predictions. JS-CB achieved much better accuracy than other methods, identifying correctly 449 HT genes at the cost of 190 false predictions. While the mean errors of other methods were close to 50%, it was remarkably less by more than 10% for JS-CB. While these numbers are a function of the data set analyzed, they suggest that the JS-CB method is a promising approach when compared to other commonly used approaches.
Current parametric methods select a threshold to discriminate between foreign and native genes. While these thresholds are often arbitrary, our proposed entropic clustering method discriminates between the gene classes in the framework of statistical significance. As a caveat, there are multiple hypothesis testing problems involved, namely the repetition of the test in each iteration step and over the hierarchy. Therefore, appropriately stringent thresholds must be chosen to compensate for multiple tests. Although sporadic rejection of the null hypothesis when using multiple tests results in failure to merge classes, these classes may be merged in subsequent steps using positional information. Although the AIC-based approach we introduced earlier () also has a strong theoretical underpinning, the thresholds in the generalized AIC cannot be rigorously described. Among the parametric methods of foreign gene detection, to our knowledge, the JS clustering methods are the only methods that classify atypical genes in the framework of statistical significance.
A shortcoming of parametric methods is their difficulty in identifying weakly atypical genes. The trade-off is clear: classifying only strongly atypical genes as foreign decreases false predictions, however, this comes at the expense of many foreign genes misclassified as native, a more relaxed criteria increases the sensitivity of a method at the expense of false predictions. This inherent weakness limits the abilities of this class of methods. Through this study, we propose gene context information as a means to address this issue. The utility of positional information increases when the confidence of typical and atypical gene classes increases. That is, optimal assignment occurs at higher stringencies ensuring the purity of both typical and atypical gene classes, at the expense of creating a larger number of classes. In a two pronged approach (class reassignment followed by class refinement), the misclassification of foreign genes was reduced by allowing weakly atypical native genes to join the native gene class by virtue of their positions, not by relaxing the criteria for class merger. This also serves to reduce the misclassification of native genes as weakly atypical foreign genes join their classes in a similar fashion (Supplementary Table 4).
We also observed that positional information works synergistically with gene clustering methods reducing the classification errors better than for methods which classify the genes only as native and foreign (). To examine this further, we carried out numerical experiments where genes from all the small classes generated by JS-CB were pooled as a single foreign class and the largest class represented native genes. Class refinement was then done using the positional information of genes. By minimizing the mean error over the parameter space of the method, comparison was made with cases when class refinement was done for all method-generated classes and also when full power of positional information (both class reassignment and refinement) was used for these classes (Supplementary Table 5). The Type II error decreased significantly causing a decrease in mean error when class refinement was performed on all method- generated classes as opposed to two classes (typical and atypical). Both Type I error and Type II error decreased remarkably when class reassignment followed by class refinement was done at strict stringencies. In addition, since JS methods effectively group genes from common donors (), they may be useful in helping identify potential donors for foreign genes in bacterial genomes.
Hayes and Borodovsky () developed a -means algorithm for partitioning a gene-set into primarily two classes ( = 2). The gene models trained on these classes were then incorporated in a prokaryotic gene finder, GeneMark-genesis, where the use of two gene classes improved considerably the identification of genes, particularly those with atypical composition. The success of such prediction algorithms critically depends on the purity of the gene classes. The value of ‘’ is not known and = 2 may not be best option to model genic complexity, as shown by our experiments on chimeric artificial as well as genuine genomes. Our hierarchical agglomerative gene-clustering algorithm provides a solution: gene classes grow logically starting with single genes and the process is halted when the distinction between the gene classes is deemed statistically significant. Native genes are identified as belonging to the single largest class that has typically ∼60–95% of the total genes, and foreign genes are divided into several small classes. It should be possible to build a gene model for each gene class, which will likely improve the accuracy of gene identification.
c
o
m
p
a
r
i
s
o
n
t
o
t
h
e
f
r
e
q
u
e
n
t
l
y
u
s
e
d
p
a
r
a
m
e
t
r
i
c
m
e
t
h
o
d
s
f
o
r
f
o
r
e
i
g
n
g
e
n
e
d
e
t
e
c
t
i
o
n
,
a
s
w
e
l
l
a
s
o
u
r
p
r
e
v
i
o
u
s
l
y
d
e
v
i
s
e
d
A
I
C
-
b
a
s
e
d
m
e
t
h
o
d
s
,
o
u
r
p
r
o
p
o
s
e
d
J
S
g
e
n
e
c
l
u
s
t
e
r
i
n
g
m
e
t
h
o
d
s
w
e
r
e
f
o
u
n
d
t
o
b
e
m
u
c
h
r
o
b
u
s
t
a
n
d
c
o
n
s
i
s
t
e
n
t
i
n
c
l
a
s
s
i
f
y
i
n
g
f
o
r
e
i
g
n
g
e
n
e
s
i
n
a
r
t
i
f
i
c
i
a
l
a
s
w
e
l
l
a
s
g
e
n
u
i
n
e
g
e
n
o
m
e
s
.
A
m
o
n
g
t
h
e
t
h
r
e
e
J
S
m
e
t
h
o
d
s
,
J
S
-
C
B
p
r
o
v
e
d
t
o
b
e
m
o
s
t
e
f
f
i
c
i
e
n
t
n
o
t
o
n
l
y
i
n
i
d
e
n
t
i
f
y
i
n
g
f
o
r
e
i
g
n
g
e
n
e
s
,
b
u
t
a
l
s
o
i
n
g
r
o
u
p
i
n
g
g
e
n
e
s
c
o
n
t
r
i
b
u
t
e
d
b
y
d
i
s
t
i
n
c
t
d
o
n
o
r
o
r
g
a
n
i
s
m
s
a
s
d
i
s
t
i
n
c
t
c
l
a
s
s
e
s
.
I
n
p
u
r
s
u
a
n
c
e
o
f
o
u
r
l
o
n
g
-
t
e
r
m
g
o
a
l
o
f
q
u
a
n
t
i
f
i
c
a
t
i
o
n
o
f
l
a
t
e
r
a
l
g
e
n
e
t
r
a
n
s
f
e
r
i
n
p
r
o
k
a
r
y
o
t
e
s
,
w
e
i
n
t
e
n
d
t
o
e
x
p
l
o
i
t
t
h
i
s
a
b
i
l
i
t
y
o
f
t
h
e
J
S
m
e
t
h
o
d
s
i
n
i
d
e
n
t
i
f
y
i
n
g
t
h
e
s
o
u
r
c
e
s
o
f
g
e
n
e
t
r
a
n
s
f
e
r
i
n
p
r
o
k
a
r
y
o
t
e
s
.
D
e
v
e
l
o
p
m
e
n
t
o
f
t
h
e
h
i
g
h
l
y
a
c
c
u
r
a
t
e
g
e
n
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
m
e
t
h
o
d
s
h
a
s
b
r
o
u
g
h
t
u
s
c
l
o
s
e
r
t
o
r
e
a
l
i
z
i
n
g
t
h
e
g
e
n
o
m
e
s
c
a
l
e
q
u
a
n
t
i
z
a
t
i
o
n
a
n
d
c
h
a
r
a
c
t
e
r
i
z
a
t
i
o
n
o
f
l
a
t
e
r
a
l
g
e
n
e
t
r
a
n
s
f
e
r
e
v
e
n
t
s
i
n
p
r
o
k
a
r
y
o
t
e
s
.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
i
s
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
PrrC, the optional anticodon nuclease (ACNase) (,) represents potential antiviral devices widespread among bacteria (). PrrC's activity is silenced by the genetically linked type Ic DNA restriction endonuclease EcoprrI () and is unleashed by Stp, the phage T4-encoded peptide inhibitor of EcoprrI (). The activation of PrrC causes specific cleavage of tRNA 5′ to the wobble base () and, consequently, could block T4-late translation and contain the infection (). However, the T4-coded RNA healing and sealing enzymes 3′-phosphatase/5′-polynucleotide kinase (Pnk) and RNA ligase 1 (Rnl1) () normally restore the intact form of tRNA (,), exercising perhaps their intended functions (,). Known PrrC homologs are invariably linked to EcoprrI homologs and, hence, could also act as secondary defenses mobilized when an associated DNA restriction endonuclease is compromised (,,).
When PrrC is expressed by itself it exhibits overt (core) ACNase activity () that purifies with a homo-oligomeric PrrC form, possibly a dimer of dimers (). PrrC's N-proximal ∼265 amino acids are thought to constitute an NTPase domain that mediates the activation of the latent ACNase (,) (A). This region contains ABC ATPase-like motifs (), albeit, sufficiently different from the typical to justify classifying PrrC's N-domain as a distinct subtype. The divergent sequence of PrrC's N-domain could account for the unusual nucleotide requirements of the ACNase activation reaction and PrrC's idiosyncratic nucleotide binding attributes. Namely, the activation depends on the cooperation of GTP and dTTP and is inhibited by ATP. Moreover, dTTP exhibits higher affinity for PrrC than GTP or ATP (μM versus mM-range); and dTTP but not GTP or ATP stabilizes PrrC's core ACNase activity. PrrC differs from the typically dimeric ABC ATPases () also in its apparent dimer of dimers structure (). The remaining ∼130 amino acid region of PrrC harbors residues implicated in tRNA recognition () and cleavage () and does not resemble a known protein structure. The main cues PrrC recognizes in tRNA map to the anticodon stem-loop (ASL). They comprise the anticodon sequence, base modifications and base-pairing interactions ().
Peptide mimicry and Cys-mediated intersubunit cross-linking data reported here suggest that the tRNA-binding motif of PrrC is shared by C-domain portions interfacing in parallel. The proposed parallel dimerization of the C-domains can be reconciled with the expected head-to-tail dimerization of the N-domains () by further suggesting that the regulatory and functional sites of PrrC arise at distinct assembly stages of its dimer of dimers form.
Purified tRNA labeled with P at the 33p34 junction (,) and synthetic tRNA ASL analogs (,) were prepared as previously described. The analogs included ASL-3 that matches mammalian tRNA in RNA sequence and base modifications but contains an extra 3′-dT that facilitated its synthesis; and ASL-C that has the same RNA sequence but base modifications of tRNA (); both were [5′-P] labeled as described (). The synthetic end-protected peptides used () were purchased from Genscript Corporation and were over 80% pure. Diazenedicarboxylic acid (diamide) was purchased from Sigma.
The PrrC forms used contained a C-terminal His tag and, except where indicated, also the leaky D222E mutation that allows high level expression of the protein (). The triple mutant D222E/C268A/C385A termed PrrC*, its Cys replacement derivatives and the H356A/F292S double mutant were generated by Quick Change (). Other PrrC mutants have been described (,).
The PrrC proteins were expressed under the control of the T7-Lac promoter and Shine–Dalgarno sequence of plasmid pRRC11 () in Rosetta (DE3)pLysS (Novagen, UK) encoding T7 RNA polymerase, T7 lyzozyme () and rare tRNAs from plasmid pRARE. The cells were grown in LB medium at 37°C to a density of ∼6.10 cells/ml. They were shifted then to 30°C and PrrC's expression induced by adding 1 mM IPTG. After further incubation for 2 h, the cells were harvested and the PrrC proteins purified by immobilized metal affinity chromatography as described ().
cells induced to express the indicated PrrC forms were harvested by high-speed centrifugation, re-suspended in 0.01 vol of 10 mM diamide in water and incubated at 10°C for the indicated time. The cells were then lyzed by heating them for 1 min at 100°C in SDS–PAGE sample buffer, their proteins were separated by SDS–PAGE and PrrC monitored by immunoblotting using purified polyclonal anti-PrrC antibodies ().
The standard core ACNase form used in the assays was a His-tagged derivative of the leaky PrrC mutant D222E. It was purified by TALON® (Clontech) affinity chromatography followed by Superdex-200 gel filtration (). The standard ACNase reaction mixture (10 μl) contained 1 fmol of the P-labeled tRNA or the indicated [5′-P] labeled ASL, both at 3000 Ci/mmol; 2 μM dTTP, 4 mM Na-HEPES buffer, pH 7.5; 0.5 mM MgCl, 15 mM NaCl, 5% glycerol and 0.5 M trimethylamine--oxide (TMAO). The reaction was performed at 10°C. It was initiated by adding the enzyme and terminated at the indicated time by adding 1.5 vol of 10 M urea, 0.01% bromphenol blue and 0.01% xylene cyanol. The products were separated by polyacrylamide–urea gel electrophoresis and quantified by densitometry or counting.
Peptide-RNA UV-cross-linking was performed in mixtures (10 μl) containing ∼0.1 pmol of [5′-P] labeled ASL-3, 0.1–2.0 nmol of Lys-anticodon recognizing peptide (LARP) or indicated LARP derivative, 5 mM Na-HEPES buffer, pH 7.5; 1 mM MgCl, 30 mM NaCl, 0.5–1.0 M TMAO and 10% glycerol. The mixture was placed in a 35-mm polystyrene culture dish and irradiated for 5 min at 4°C over a TFX-20 M transilluminator providing radiation spectrum with a peak at 312 nm. Aliquots were mixed with 1.5 vol of 10 M urea containing 0.01% each of xylene cyanol and bromphenol blue and separated by denaturing gel electrophoresis.
ThePrrC mutant proteins H356A and H356A/F292S were cross-linked to ASL-3 in 12 μl ACNase reaction mixtures essentially as described above but containing 0.3 pmol of ASL-3 and ∼1 μg of the indicated PrrC allele purified by TALON cobalt affinity chromatography. The mixture was irradiated for 5 min at 4°C as described for the LARP-ASL-3 cross-linking. An equal volume of 2× SDS–PAGE gel sample buffer devoid of reducing agent was added and the products separated by 10% SDS–PAGE. The proteins were transferred then to a nitrocellulose membrane, PrrC monitored by immunoblotting using anti-PrrC antibodies () and the PrrC/ASL-3 cross-linking product detected by autoradiography.
The indicated PrrC forms were subjected to GA-mediated protein–protein cross-linking as previously described and the products monitored by immunoblotting ().
Mutations that alter ACNase cleavage specificity have implicated PrrC's Asp in tRNA recognition (). Asp maps to a consensus sequence shared by a subset of the known PrrC homologs. This sequence comprises a predicted coil region (284–92) and the essential aromatic residues Phe and Tyr () (,). Its possible role in tRNA recognition was investigated using an end-protected mimic (Ac-KYGDSNKSFSY-NH). This mimic was termed LARP to indicate its anticipated function. LARP inhibited PrrC's ACNase activity in a dose-dependent manner when assayed with tRNA (A and B) or ASL analogs (C–E). An IC value of ∼10M was obtained in all cases. The inhibition decreased with increasing substrate level within the range examined (F and G). These results hinted that LARP acted by occluding the RNA substrate and/or competing with the substrate over the protein.
The possible interaction of LARP with the ACNase substrate was tested by attempting to UV-cross-link the peptide to ASL-3. The cross-linking was performed at 312 nm where the tRNA wobble base is highly photoreactive (). Several cross-linking products were obtained in amounts proportional to the dose of LARP (A and B). The most abundant (designated ) was retarded in gel electrophoresis relative to ASL-3 by the equivalent of ∼4 nt, possibly due to a single LARP adduct. Slower migrating products designated could contain additional LARP moieties, judged from their incremental retardations and dependence on LARP. Presumably, they arose through peptide–peptide cross-links since data shown below suggested that LARP could also self-interact ( and ). The non-irradiated mixture yielded weaker bands over a continuous distribution trailing behind ASL-3 (lane 2), probably non-covalent complexes that partially dissociated during the fractionation.
LARP was mutated to evaluate the importance of its sequence to the ASL-cross-linking and ACNase-inhibiting activities. Trimmed derivatives lacking Lys (Δ1) or Tyr (Δ11) formed UV-cross-links to ASL-3 ∼ 7- or 3-fold less efficiently than LARP, respectively (A). The conjugates obtained with the trimmed mutants (designated a′, a*, ′ and ) were less retarded than those of LARP, possibly due to their lower mass. Δ1 and Δ11 also hardly inhibited ACNase (B).
LARP's sequence was also changed by PrrC missense mutations that confer an ACNase null phenotype. These mutations: F292S (), Y294S or Y294F () were respectively termed F9S, Y11S and Y11F in the peptide context. They were introduced in a LARP derivative containing an extra N-terminal Cys (CLARP, described in more detail later). CLARP was far more efficient than LARP both in cross-linking ASL-3 and inhibiting ACNase () and using it as the reference peptide facilitated the detection of residual mutant activities.
F9S and Y11S impaired the two activities in a correlated manner. The milder Y11S mutation reduced the yield of the primary UV-cross-linking conjugate with ASL-3 nearly 2-fold (A, lane 3 versus 4) and partially impaired the inhibition of ACNase (B, lane 3 versus 4). It is noteworthy that this mutation also prevented the formation of a stable secondary conjugate, as inferred from the absence of the expected band and appearance instead of a smear trailing behind band . The more drastic mutation F9S reduced the cross-linking to ASL-3 8-fold (A, compare lanes 8 and 9) and severely impaired ACNase inhibition (B, compare lanes 8 and 9). The corresponding PrrC mutation F292S could also abrogate the UV-cross-linking of PrrC to ASL-3. This was inferred from the observation that the active site PrrC mutant H356A, which lacks ACNase activity () but efficiently binds tRNA (data not shown), formed UV-cross-links to ASL-3 ∼ 30-fold more efficiently than F292S/H356A (C).
The Y11F mutation conferred a different phenotype. It nearly abolished the ACNase-inhibiting potential of CLARP (B, compare lanes 2, 3 and 5) yet doubled the yield of the primary UV-cross-linking conjugate obtained with ASL-3 (A, compare lanes 3 and 5). In part this increase could be ascribed to the failure of Y11F to form stable higher conjugates, as with Y11S. Nonetheless, the overall cross-linking yield obtained with Y11F suggested that this mutant bound ASL-3 at least as efficiently as CLARP. The discrepant behavior of Y11F hinted that the ACNase-inhibiting forms of LARP exerted their effect not only by occluding the RNA substrate but also by forming a binary complex with PrrC. Since LARP was employed in excess over PrrC and ASL-3, it could exert these two ACNase-inhibiting modes independently. The possibility that LARP interacted with PrrC directly was reinforced by data shown below suggesting that both LARP and its matching PrrC sequence can self-interact ( and ). The failure of the Y11S and Y11F mutants to form stable conjugates containing additional peptide moieties could not be ascribed to a critical role of Tyr in peptide–peptide UV-cross-linking since deleting this residue did not elicit such an effect (A, lane 5).
LARP derivatives extended at the N- or C-end with Cys (CLARP and LARPC, respectively) were intended for further modification with Fe-EDTA that rendered them artificial nucleases (M. Amitsur, D. Klaiman and G. Kaufmann; unpublished data). Unexpectedly, CLARP and, to a lesser extent, LARPC were far more potent than LARP in UV-cross-linking ASL-3 and inhibiting ACNase. Thus, when CLARP, LARPC or LARP were UV-cross-linked to ASL-3 at 1 mM DTT and identical peptide levels, similar product patterns were obtained. However, the product yields with CLARP or LARPC were about an order of magnitude higher than with LARP (A). Employing the Cys-containing peptides at a level 10-fold lower than that of LARP resulted in comparable yields (B). CLARP and LARPC UV-cross-linked ASL-3 more efficiently than LARP also without DTT, LARPC yielding under these conditions a relatively high proportion of the secondary conjugate (C). However, the cross-linking efficiency of CLARP and LARPC to ASL-3 was relatively weak at 10 mM DTT (D). CLARP and LARPC inhibited ACNase more strongly than LARP at 1 mM DTT (E) but were less effective than LARP at 10 mM DTT (F).
The suspicion that LARP exerted its ASL-3 cross-linking and ACNase inhibiting activities as a parallel dimer prompted us to examine if the matching PrrC sequence dimerizes similarly. To this end, a Cys residue was introduced in this sequence instead of Ser, Ser or Ser. We expected that the mutant Cys residues will self-interact and form intersubunit disulfide cross-links if the region containing them dimerizes in parallel. Each of the three mutations was placed over the D222E/C268A/C385A background termed PrrC*. This facilitated the isolation of the mutant proteins (,) and precluded the formation of non-specific S-S cross-links due to the wild-type Cys residues. The PrrC* control and to-Cys derivatives exhibited comparable ACNase activities and protein levels (data not shown). However, only PrrC* and S288C/PrrC* retained ACNase activity (A).
To induce the formation of S-S cross-links the cells expressing the to-Cys mutants or the PrrC* control were exposed to diamide (). The cellular proteins were separated then by SDS–PAGE under non-reducing conditions and PrrC visualized by immunoblotting. As shown, the three mutants, but not PrrC* yielded products that migrated in SDS–PAGE slower than the ∼47.5-kDa PrrC monomer (B). These products included a major form designated X that migrated below the 175-kDa size marker as well as several less pronounced and faster migrating forms. These various products appeared also without diamide, perhaps due to partial oxidation. However, their amounts increased considerably with diamide, consistent with their formation by disulfide cross-linking.
To determine if any of the diamide-dependent products arose by intra-PrrC cross-links, the four PrrC forms were isolated by immobilized metal affinity chromatography in the absence of a reducing agent. Subsequently they were fractionated by SDS–PAGE as such, after prior treatment with 1 mM DTT or also with 2 mM diamide. Western analysis revealed that the purified Ser → Cys mutants, but not PrrC*, yielded a single product with mobility similar to that of band X of the diamide treated cells (compare C, lanes 1, 4, 7 and 10 to B lanes 4, 8, 12 and 16). Moreover, this band was abolished by DTT (C. lanes 2, 5, 8 and 11) and restored by excess diamide (lanes 3, 6, 9 and 12), indicating it depended on disulfide cross-linking. Corresponding protein staining (D) ascertained that the various PrrC forms were not contaminated by significant amounts of other proteins and, consequently, that band X was unlikely to contain proteins other than PrrC. The failure of a considerable fraction of the to-Cys mutant proteins to generate band X could be attributed to its misfolding since PrrC is thermally unstable. Alternatively, this fraction could represent a functionally relevant conformation incompatible with the S-S cross-linking; e.g. due to a substrate or inhibitor bound to the tRNA site.
Band X migrated in SDS–PAGE less than expected of a cross-linked PrrC dimer of ∼95 kDa. This fact could be attributed to its particular shape or irreversible entanglement of two such dimers within the PrrC tetramer. To distinguish between these possibilities, we subjected PrrC* and S293C/PrrC* to GA-mediated protein–protein cross-linking. PrrC* yielded the familiar pattern () where the monomer is gradually converted into apparent GA-cross-linked tetramer forms via dimeric and less pronounced trimeric intermediates (E, lanes 1–4). The monomeric fraction of the mutant behaved similarly (lanes 5–8). Band X, which coincided with the most retarded GA-cross-linked dimers, was also gradually converted into higher forms superimposed over those generated by the fraction that failed to form the S-S cross-links. These data suggested that band X was a dimer that was retarded due to its particular shape.
We analyzed in a similar fashion PrrC* derivatives containing Cys in the predicted α-helix found immediately downstream to the LARP-like region (). It replaced L298, I302 or I306 of the α-helix hydrophobic face or the hydrophilic Q300. These mutants also yielded the S-S cross-linked form designated X, L298C yielding the highest proportion (F, lanes 1–3), possibly due to closer contact of the dimerizing α-helices at position 298. Q300C yielded a variant (designated Xb) that migrated slightly faster than the form X generated by the other mutants designated here Xa (F, compare lanes 3 and 4, 6 and 7). The slight mobility difference may be attributed to the presence of Cys at the hydrophilic and, hence, outward-pointing face of the dimerizing α-helix. Consequently, an S-S cross-link mediated by it could constrain the dimer into a more compact, faster migrating form. The PrrC mutant D222E containing the wild-type residues Cys and Cys failed to generate band X (F, lanes 16–18). This result underscored the specificity of the intersubunit cross-links triggered by the mutant Cys residues placed in the 288–306 range.
LARP formed UV-cross-links to ASL analogs of tRNA and inhibited their PrrC-catalyzed cleavage ( and ). Moreover, the two activities were impaired by trimming LARP () or placing in it missense mutations that confer a null ACNase phenotype in the PrrC context (). Importantly, the CLARP mutation F9S and the corresponding inactivating PrrC mutation F292S severely inhibited the respective UV-cross-linking of ASL-3 to the peptide (A) or protein (C). These results could be taken to indicate that LARP mimics its matching PrrC sequence in tRNA recognition. However, even in that case LARP would probably represent only part of PrrC's tRNA-binding motif since the level at which it effectively inhibited ACNase () far exceeds the of efficient ACNase substrates (). Moreover, known RNA-binding motifs including that of colicin E5-ACNase are several fold longer than LARP (,).
The functional relevance of LARP to its matching PrrC region was suggested also by the apparent ability of both to dimerize in parallel. We deduced that LARP acted as a parallel dimer in forming UV-cross-links to ASL-3 and inhibiting ACNase from the ability of a single Cys residue tethered to either the N- or C-end of LARP to dramatically enhance these activities and abolition of these enhancements under highly reducing conditions (). These coincidences could be accounted for by the stabilization of a common functional dimeric structure of LARP by a disulfide bond formed at one or the other end of its protomers, a requirement met by a parallel but not anti-parallel dimer. Alternatively, the similar effects of the N- and C-proximal Cys extensions reflected the increased probability of non-specific peptide–RNA interactions caused by mere dimerization of the peptide, whether caused by the N- or C-terminal S-S bond. However, we favor the first explanation in view of the intersubunit S-S cross-links triggered by all seven Cys substitutions placed along the C-proximal PrrC 288–306aa region overlapping the sequence matching LARP (). Such an outcome is consistent with parallel dimerization of the mutated region and lends support to the assumption that LARP could also dimerize in this orientation. The PrrC region that dimerizes in parallel could be confined to a portion of the ACNase domain since the regulatory N-domain is expected to dimerize in a head-to-tail fashion, like typical ABC/ATPases where the nucleotide binding pocket arises by the interaction of the Walker A motif of one subunit with the ABC signature motif of the second (). Moreover, the wild-type Cys at the edges of the C-domain (A) did not trigger intersubunit cross-links (F). The relation of the PrrC region that dimerizes in parallel to the tRNA-binding motif may be revealed by accurately mapping and functionally characterizing it.
Finally, as illustrated in B, the opposite orientations in which PrrC's N- and C-domains dimerize dictate a phosphofructokinase-like topology () where the NTPase and tRNA sites arise at distinct assembly stages of the PrrC dimer of dimers (). |
Recent discoveries of novel structured RNAs () indicate that such RNAs are common in cells. To assist in discovering additional structured RNAs, we have developed an automated pipeline that can identify conserved RNAs within bacteria (). This pipeline assembles the potential 5′ untranslated regions (UTRs) of mRNAs of homologous genes, and uses the CMfinder () program to predict conserved RNA structures, or ‘motifs’, within each set of UTRs. Automated homology searches are then employed to find additional examples of these motifs, which CMfinder uses to improve the secondary structure model and sequence alignment of each motif. The output is a set of alignments with a predicted RNA secondary structure, and these alignments are subsequently analyzed manually to make improvements to the model and to assess which motifs merit further study.
Although other automated searches for RNAs have been performed (), our pipeline () is distinguished from these by three features. First, our pipeline uses CMfinder, which can discover a motif even when some input sequences either do not contain the motif or include motif representatives carrying unrelated sequence domains. CMfinder also can produce a useful alignment even with low sequence conservation, however, the algorithm will exploit whatever sequence similarity is present. Second, our pipeline integrates homology searches to automatically refine the alignment and structural model for each motif. Third, since our pipeline aligns UTRs of homologous genes, it is well suited to find -regulatory RNAs, and its dependence on sequence conservation is further reduced.
Indeed, using the bacterial phylum Firmicutes as a test case, we previously demonstrated that this pipeline makes useful predictions for virtually all known -regulatory RNAs (). The pipeline also finds motifs that are likely to be -encoded, or ‘non-coding RNAs’ (ncRNAs), when these happen to be upstream of homologous genes.
In the present report, we describe the use of this pipeline to find structured RNAs amongst all bacteria whose genomes have been sequenced. We describe 22 novel motifs that are likely to be conserved, structured RNAs. We were particularly interested in discovering riboswitches, a type of structured RNA usually found in mRNAs that directly senses a specific small molecule and regulates gene expression (,). Subsequent experiments have confirmed that two of these motifs are novel riboswitches. The first binds SAH (-adenosylhomocysteine) (J.X.W., D. Rivera, E.R.L., R.R.B., in preparation; ) and the second is a SAM (-adenosylmethionine)-binding riboswitch in and related species (Z.W., E.E. Regulski, R.R.B., unpublished data). Experimental evidence with another riboswitch candidate indicates that it senses molybdenum cofactor or ‘Moco’ (E.E. Regulski, R. Moy, R.R.B., unpublished data).
A candidate of particular interest is the Genes for the Environment, for Membranes, and for Motility (GEMM) motif, which has properties that are typical of riboswitches. For example, GEMM is a widespread and highly conserved genetic element that, in 297 out of 309 cases, is positioned such that it is likely to be present in the 5′ UTR of the adjacent open reading frame (ORF). The genes putatively regulated by GEMM are typically related to sensing and reacting to extracellular conditions, which suggests that GEMM might sense a metabolite produced for signal transduction or for cell–cell communication.
Some characteristics of all 22 predicted RNA motifs are summarized in , and we discuss the features and possible biological roles of some candidates in the ‘Results’ section. Additional information on all candidates, including annotated multiple sequence alignments, and collections of taxonomy and nearby genes, are presented in Supplementary Data. Raw pipeline predictions are accessible at .
The potential 5′ UTRs of genes classified in the Conserved Domain Database () version 2.08 were used as input for our computational pipeline (). Completed genome sequences and gene positions were taken from RefSeq () version 14, but we eliminated genomes whose gene content was highly similar to other genomes. Our UTR extraction algorithm () accounted for the fact that a UTR might not always be immediately upstream of the gene due to operon structure.
For each conserved domain, the collected sequences of potential UTRs were given as input to CMfinder () version 0.2, which produced local multiple sequence alignments of structurally conserved motifs within the UTR sets. These alignments were then used to search for additional homologs within annotated intergenic regions, except that intergenic regions were extended by 50 nt on both ends to account for misannotated ORFs. This homology search was performed with the RNA program version 0.2f () with ML-heuristic filters (), Covariance Model () in global mode implemented by Infernal () version 0.7 and an E-value () cutoff of 10. CMfinder then refined its initial alignment using these new homologs. Known RNAs were detected based on the Rfam Database () version 7.0. Predictions were scored based on phylogenetic conservation of short sequences (), diversity of species and structural criteria. The pipeline algorithm is described in more detail elsewhere (). The Supplementary Data includes a list of software and databases used.
We split bacterial genomic sequence data into groups, and performed UTR extraction, motif prediction and homology search on each group separately. This was motivated by the fact that UTRs in different phyla often do not contain the same motif. However, phyla with few sequenced members were coalesced based on taxonomy to try to assemble sufficient sequence data to predict a motif. Phyla with only one or two sequenced members (e.g. Chloroflexi) were ignored. The following eight groups were used: () Firmicutes, () Actinobacteria, () α-proteobacteria, () β-proteobacteria, () γ-proteobacteria, () δ- and ε-proteobacteria, () Cyanobacteria and () Bacteroidetes, Chlorobi, Chlamydiae, Verrucomicrobia and Spirochaetes.
We then selected promising motifs, further analyzed them by performing additional homology searches and edited their alignments with RALEE (). We used NCBI BLAST (), Mfold (), Rnall () (to identify rho-independent transcription terminators), CMfinder and RNA to assist in these analyses. Information on metabolic pathways associated with motifs was retrieved from KEGG (). To expand alignments by identifying additional structured elements, we extended alignments on their 5′ and 3′ ends by 50–100 nt, and realigned as necessary using either CMfinder or by manual inspection.
The additional searches for homologs used RNA in several ways. We found both global- and local-mode () Covariance Model searches gave complementary results. Sequence databases used in manually directed searches were the ‘microbial’ subset of RefSeq version 19 (), and environmental shotgun sequences from acid mine drainage () (GenBank accession AADL01000000) and Sargasso Sea () (AACY01000000). Additional marine shotgun sequences were used for the and ATPC motifs ().
Because the full set of sequences is roughly 3.2 billion nucleotides, searches can report many false positives, especially for shorter motifs. When appropriate, we searched four kinds of subsets of these sequences. First, we did not always use the environmental sequence data. Second, we sometimes searched only genomes in the bacterial group (e.g. β-proteobacteria) from which the motif was originally derived. Third, we sometimes searched only intergenic regions (extended by 50 nt as before). Fourth, we used the BLAST program tblastn to search for genes homologous to those associated with the motif. RNA searches were then conducted on the 2 kb upstream of these matches (which is expected to contain the 5′ UTR), and 200 nt downstream (since apparent coding homology might extend upstream of the true ORF, causing BLAST to misidentify the start codon). Using the motif's bacterial group as the BLAST database facilitated the discovery of highly diverged homologs. For example, a search upstream of genes in ε-proteobacteria revealed homologs of the motif (see later) with a truncated stem. Additionally, with such small databases, we can forego the ML-heuristic filters in RNA. When the full sequence set is used as the BLAST database, it can help to find homologs in other phyla.
To establish the extent of conservation reflected in consensus diagrams (e.g. in ), sequences were weighted to de-emphasize highly similar homologs. Weighting used the GSC algorithm (), as implemented by Infernal (), and weighted nucleotide frequencies were then calculated at each position in the multiple sequence alignment. To classify base pairs as covarying, the weighted frequency of Watson–Crick or G–U pairs was calculated. However, aligned sequences in which both nucleotides were missing or where the identity of either nucleotide was uncertain (e.g. was ‘N’, signifying any of the four bases) were discarded. Classification as a covarying position was made if two sequences had Watson–Crick or G–U pairs that differ at both positions amongst sequences that carry the motif. If only one position differed, the occurrence was classified as a compatible mutation. However, if the frequency of non-Watson–Crick or G–U pairs was more than 5%, we did not annotate these positions as covarying or as compatible mutations.
Promising structured RNA motifs predicted by the CMfinder pipeline were examined manually to refine the consensus sequence and structural models (see ‘Materials and methods’ section) and to provide information on possible function. Key findings for each candidate are summarized in . Of the 22 motifs identified, seven are depicted in and the remainder are depicted in Supplementary Data.
-regulatory RNAs such as riboswitches are regions of mRNAs that regulate gene expression. In bacteria, most -regulatory RNAs occur in the 5′ UTR of the mRNA under regulation. Although it is not possible to reliably predict the transcription start site, we declare representatives of a motif as positioned in a ‘5′ regulatory configuration’ to a gene when the element could be in the 5′ UTR of an mRNA (if the transcription start site is 5′ to the element). When most or all representatives of a motif are in a 5′ regulatory configuration to a gene, this is evidence that the motif might have a -regulatory function.
-regulatory RNAs often have one of two noteworthy structural features: rho-independent transcription terminators, or stems that overlap the Shine–Dalgarno sequence (bacterial ribosome-binding site) (). Rho-independent transcription terminators usually consist of a strong hairpin followed by four or more U residues (). Regulatory RNA domains can control gene expression by conditionally forming the terminator stem. Similarly, conditionally formed stems can overlap the Shine–Dalgarno sequence, thereby regulating genes at the translational level ().
GEMM is widespread in bacteria and appears to have a highly conserved sequence and structure suggestive of a function that imposes substantial biochemical constraints on the putative RNA. We found 322 GEMM sequences in both Gram-positive and Gram-negative bacteria. It is common in δ-proteobacteria, particularly in and related genera. Within γ-proteobacteria, it is ubiquitous in Alteromonadales and Vibrionales. It is also common in certain orders of the phyla Firmicutes and Plantomycetes. Prominent pathogens with GEMM include the causative agents of cholera and anthrax.
Out of 309 GEMM instances where sequence data includes gene annotations, GEMM is in a 5′ regulatory configuration to a gene in 297 cases, implying a -regulatory role. Genes presumably regulated by GEMM display a wide range of functions, but most genes relate to the extracellular environment or to the membrane, and many are related to motility.
GEMM consists of two adjacent hairpins (paired regions) designated P1 and P2 (). P1 is highly conserved in sequence and structure, and consists of 2- and 6-bp stems separated by a 3-nt internal loop and capped by a terminal loop. The internal loop is highly conserved, and the terminal loop is almost always a GNRA tetraloop (). The P1 stem exhibits considerable evidence of covariation at several positions, and is highly conserved in structure over a wide range of bacteria. This fact, and the more modest covariation and variable-length stems of P2, provide strong evidence that GEMM functions as a structured RNA. The sequence linking P1 and P2 is virtually always AAA, with only two exceptions in 322 examples.
The P2 hairpin shows more modest conservation than P1. When the P1 tetraloop is GAAA, a GNRA tetraloop receptor usually appears in P2. This receptor is often the well-known 11-nt motif, which might be favored by GAAA loops (), but some sequences could be novel tetraloop receptors. When P1 has a GYRA tetraloop, the receptor-like sequence is almost never present, although a bulge nearer the P2 base is sometimes found ().
Many instances of GEMM include a rho-independent transcription terminator hairpin. The 5′ side of the terminator stem often overlaps (and presumably competes with) the 3′ side of the P2 stem (B). If GEMM is a riboswitch, ligand binding could stabilize the proposed P1 and P2 structure, thus preventing the competing transcription terminator from forming. In this model, higher ligand concentrations will increase gene expression. One third of GEMM representatives in δ-proteobacteria, and some in other taxa, are in a ‘tandem’ arrangement, wherein one instance appears 3′ and nearby to another in the same UTR. Such arrangements of regulatory RNAs are implicated in more sophisticated control of gene expression than is permitted by a single regulatory RNA configuration ().
An understanding of the biological role of GEMM will likely shed light on the broad variety of microbial processes that it appears to regulate. In fact, GEMM is implicated in two systems that are already the object of several studies in the species and in causes cholera in humans, but spends much of its lifecycle in water, where it can adhere to chitin-containing exoskeletons of many crustaceans. Chitin, a polymer of GlcNAc (-acetylglucosamine), has been shown to affect expression of many genes (). GEMM appears to regulate two of these chitin-induced genes. The first, , is important for adhering to chitin beads () and human epithelial cells (), as well as infection of mice ().
The second chitin-induced gene is . Remarkably, chitin induces natural competence in (), and expression is essential for this competence. has two genes that match the CDD models COG3070 and pfam04994 that correspond to separate domains. Both domains yield RPSBLAST () E-values better than 10. One of these is (locus VC1153). GEMM appears to regulate the other, which we call (VC1722). Thus, in and related bacteria, GEMM might participate in chitin-induced competence, or even regulate competence in environments not containing elevated chitin concentrations.
and related δ-proteobacteria can generate ATP by oxidizing organic compounds, using metal ions such as Fe(III) as electron acceptors (). GEMM is associated with pili assembly genes in species. Pili in have been shown to conduct electricity (), and are thus a part of the process of reducing metal ions.
Moreover, GEMM appears to regulate seven cytochrome genes in . Although this bacteria has 111 putative cytochrome genes, five of the seven GEMM-associated genes have been identified in previous studies, and might have special roles. OmcS (Outer-Membrane Cytochrome S) is one of two proteins that are highly abundant on the outer membrane of , and is required for reducing insoluble Fe(III) oxide, but not for soluble Fe(III) citrate (). OmcG and OmcH are necessary for production of OmcB, an essential cytochrome in many conditions (). OmcA and OmcT are associated with OmcG, OmcH or OmcS. Only four other Omc annotations remain in that have no direct GEMM association: OmcB, OmcC, OmcE and OmcF.
Unlike known riboswitches, GEMM is associated with a great diversity of gene functions (). This observation indicates that, if GEMM is a riboswitch, it is not serving as a typical feedback sensor for control of a metabolic pathway. Rather, GEMM more likely senses a second-messenger molecular involved in signal transduction or possibly cell–cell communication (). In this model, different bacteria use GEMM and its signaling molecule to control different processes. The fact that many GEMM-associated genes encode signal transduction domains could suggest a mechanism by which many of the signal transduction proteins are regulated. Preliminary biochemical results indicate that GEMM RNAs indeed serve as aptamer components of a new-found riboswitch class (N.S., E.R.L, R.R.B., unpublished data).
The SAH motif is highly conserved in sequence and structure (), showing covariation within predicted stem regions, including modular and variable-length stems. The SAH motif is found in a 5′ regulatory configuration to genes related to SAH (-adenosylhomocysteine) metabolism, primarily in β- and some γ-proteobacteria, and especially the genus . SAH is a part of the -adenosylmethionine (SAM) metabolic cycle, whose main components include the amino acid methionine. SAH is a byproduct of enzymes that use SAM as a cofactor for methylation reactions. Typically, SAH is hydrolyzed into homocysteine and adenosine. Homocysteine is then used to synthesize methionine, and ultimately SAM.
High levels of SAH are toxic to cells because SAH inhibits many SAM-dependent methyltransferases (). Therefore cells likely need to sense rising SAH concentrations and dispose of this compound before it reaches toxic levels. The genes that the SAH motif associates with are -adenosylhomocysteine hydrolase (), cobalamin-dependent methionine synthase () and methylenetetrahydrofolate reductase (), which synthesizes a methyl donor used in methionine synthesis. This genetic arrangement of the SAH motif and its high degree of conservation are consistent with a role in sensing SAH and activating the expression of genes whose products are required for SAH destruction. Indeed, biochemical and genetic evidence supports the hypothesis that this motif is an SAH-sensing riboswitch (J.X.W., D. Rivera, E.R.L. and R.R.B., in preparation).
This motif is found upstream of COG4708 genes in some species of and in , although some instances of the COG4708 gene family in lack the putative RNA motif. COG4708 genes are predicted to encode membrane proteins.
Although the COG4708 motif is highly constrained phylogenetically and has only six unique sequences, it shows covariation, modular stems and variable-length stems (). The motif has a pseudoknot that overlaps the putative Shine–Dalgarno sequences of COG4708 genes, which suggests that the motif encodes a -regulator of these genes.
We recently characterized a riboswitch that senses the modified nucleobase preQ (). Since this riboswitch is associated with COG4708, we proposed that COG4708 is a transporter of a metabolite related to preQ. Therefore, we hypothesize that the COG4708 motif is also a preQ-sensing riboswitch. Preliminary experiments support this hypothesis (M. Meyer, A.R. and R.R.B., unpublished data). The COG4708 motif shares no similarity in sequence or structure with the previously characterized preQ-sensing riboswitch ().
The motif is only found in a 5′ regulatory configuration to genes, which are likely co-transcribed with the related downstream genes and . The products of these three genes synthesize succinyl-CoA from 2-oxoglutarate in the citric acid cycle. All detected instances of the motifs are in β-proteobacteria in the order Burkholderiales. Although many nucleotides in the motif are strictly conserved, those that are not show covariation and contain very few non-canonical base pairs (). The motif has stems that overlap the putative Shine–Dalgarno sequence, so the motif probably corresponds to a -regulatory RNA. Note that the exact position of the putative Shine–Dalgarno sequence is inconsistent among motif instances, so is not well reflected in (see alignment in Supplementary Data). The relatively complex structure of the motif suggests that it might be a riboswitch. However, it is difficult to evaluate its degree of sequence and structure conservation since the motif is not broadly distributed.
The 23S-methyl motif consists of two large hairpins. The second hairpin ends in a run of Us and appears to be a rho-independent transcription terminator. Both stems have considerable covariation, providing strong evidence that they are part of a functional RNA. Although the structural model shows that many paired positions sometimes have non-canonical base pairs, each instance of the motif consists predominantly of energetically favorable pairs, as shown in Supplementary Data. The presence of a putative transcription terminator suggests that this is a -regulatory RNA. Since 23S rRNA methyltransferase interacts with an RNA substrate, it might autoregulate its expression using the 23S-methyl motif, in a manner similar to autoregulation of ribosomal protein genes ().
This motif is found in a variety of β-proteobacteria, especially . There is some ambiguity as to the DNA strand from which it might be transcribed, because its structure exhibits comparable covariation and conservation in both directions. In one direction, it is often upstream of genes. It could be a -regulatory RNA in this direction, but there are two genes, not homologous to each other, that are immediately downstream of motif representatives and these are positioned on the wrong strand to be controlled in the usual manner of -regulatory RNAs. In the other direction (anti-), the motif is not typically in the 5′ UTRs of genes. The anti- motif ends in a transcription terminator hairpin. Many genes downstream of anti- instances are on the opposite strand, therefore we propose that anti- could encode a non-coding RNA.
This motif consists of a single hairpin with several conserved positions (). It is widespread in , a genus of β-proteobacteria. It typically occurs multiple times in succession (2–6 copies) with conserved linker sequences, but ranges to as many as 12 copies in two instances. In 141 occurrences of single or repetitive MAEB motifs, 132 are in a 5′ regulatory configuration to a gene. In fact, many of these genes are directly involved in primary metabolism (e.g. genes involved in biosynthesis, catabolism or transport of small molecules), and not genes such as DNA repair, replication, signaling or motility. Out of the 46 conserved domains (excluding hypothetical genes) downstream of MAEB in more than one instance, at least 42 are annotated as participating in primary metabolism (see Supplementary Data for list). There are many bacterial genes not involved in primary metabolism, so these data suggest a functional association with metabolic gene control.
There is a possible relationship between MAEB and cellular response to abundant glycine. MAEB is frequently associated with and , which are part of the glycine cleavage system, wherein excess glycine feeds into the citric acid cycle. MAEB is also associated with several citric acid cycle genes (see Supplementary Data). However, MAEB is associated with some other genes with a more tenuous relationship to glycine or the citric acid cycle. It is tempting to infer a relationship to the glycine cleavage system because the highest number of MAEB repeats are associated with the genes in this system. Moreover, there exists at least one riboswitch class that binds glycine (), but this class is present in only one copy per genome in organisms with MAEB, possibly leaving a role in glycine regulation for MAEB.
Although we cannot rule out the possibility that MAEB could be a repetitive element, its association with metabolic genes argues against this hypothesis. It is also possible that MAEB is part of a protein-binding RNA like CsrB. CsrB is an RNA with roughly 18 hairpins, each of which can bind one CsrA protein subunit ().
The mini- motif consists of two tandem hairpins whose stems show considerable covariation and whose loops have characteristic ACGR motifs (). Mini- is widespread in α-, β- and γ-proteobacteria, with additional examples in other taxa. We named this motif mini- because it appears to be a -regulator of a set of genes similar to that of the previously described () (hereafter termed ‘’). The Supplementary Data lists all eight conserved domains common to and mini-. However, the structures of and mini- appear to be unrelated.
The motif is a highly structured and broadly conserved motif that was proposed previously to be a promising riboswitch candidate. The simple structure of mini-, however, is uncharacteristic of most other riboswitches, though its broad phylogeny suggests a function that dictates broad conservation. Mini- appears to be a -regulatory element because it is associated with a relatively narrow set of gene functions and it is near to their coding sequences (90% are within 33 nt of the Shine–Dalgarno sequence). We propose that mini- serves the same (but currently unknown) role as the motif, although the mechanisms used to control gene expression could be different.
We note that there might be instances of mini- with only one hairpin. However, we did not explore this possibility because the simplicity of the single hairpin would lead to a prohibitively high false positive rate in genome-scale searches. This issue is not a problem for the full, two-hairpin motif (see Supplementary Data).
The motif is found upstream of all genes in fully sequenced ε-proteobacteria (e.g. and ). The gene encodes GAR (phosphoribosylglycinamide) synthetase, which is involved in purine biosynthesis. The motif shows covariation and modular stems, although it also exhibits some mutations that disrupt base pairing ().
To test the hypothesis that the motif represents a riboswitch aptamer, we used in-line probing assays () to test for binding of the RNA against a panel of available purine compounds, including GAR (see Supplementary Data for list). Our assays showed no evidence of structural modulation induced by any of these compounds (data not shown). Although these data fail to support the hypothesis that the motif is a riboswitch, its consistent association with the gene at least implies a -regulatory role.
This motif is widespread among Actinobacteria, and consists of two hairpins, where the loop of each contains a run of at least 6 Cs. The 6C motif exhibits significant covariation in its stems. 6C motif instances are usually moderately close (200–300 nt) to genes predicted to be related to chromosome partitioning and pilus assembly. However, given its distance, it is not clear whether 6C is functionally related to these genes.
However, both motifs could also be dsDNA sites recognized by protein dimers, where each subunit binds to sites on opposite strands. Alternately, either motif could conceivably function as a structural dsDNA element. A hairpin element, combined with other factors, could favor a structure with two intra-strand hairpins embedded in dsDNA, a ‘cruciform’-like structure that is the preferred target for proteins in distinct, though related contexts (,). DNA-binding motifs in Xis were also described (), although no motifs containing hairpins were reported.
The ATPC motif occurs in some Cyanobacteria in an ATP synthase operon, between genes encoding the A and C subunits. ATPC motif instances are found in all sequenced strains of , and certain species of . The motif consists primarily of a three-stem junction. Previous studies have proposed hairpin-like structures in Cyanobacterial ATP synthase operons, but not more complicated shapes, and in different locations from the ATPC motif ().
Given the gene context, we expect that this motif mimics the ligand of the downstream ribosomal protein gene (), and that the product of this gene thereby controls its own expression. Although we commented on ribosomal protein gene autoregulation previously in Firmicutes (), we generally ignored ribosomal-gene-associated RNA motifs in the present study because many have already been characterized. However, the identification of the cyano-30S motif supports the view that such RNAs are found in a wide variety of phyla.
The lacto-1 and lacto-2 motifs are confined to the order Lactobacillales. The lacto-1 motif has some covariation, but some mutations disrupt base pairing, so its assignment as a structured RNA is uncertain. Some instances intersect a variable region of the S(MK) (or SAM-III) () riboswitch between the main hairpin and the Shine–Dalgarno sequence. The lacto-2 motif consists of a large hairpin with many internal loops, some of which have highly conserved sequences. Although some mutations disrupt pairing, there is a considerable amount of covariation, which suggests that the lacto-2 motif instances are probably structured RNAs.
Two predicted motifs have several instances in , but are not found in any other sequenced bacteria. The motifs, TD-1 and TD-2 have 28 and 36 representatives, respectively. Seven TD-1 motif representatives overlap reverse complements of instances of TD-2, and share the two 5′-most hairpins. Although it is possible that the two motifs could be merged, it is not obvious how, because there is significant variation in the non-overlapping instances.
Both motifs show covariation and either variable-length or modular stems. However, the modest but noticeable number of mutations that disrupt pairing reduces confidence that they are functional RNAs. The TD-1 motif is usually in a 5′ regulatory configuration to genes, although the wide array of poorly characterized genes makes it difficult to suggest a coherent -regulatory function. The TD-2 motif does not share the 5′ regulatory configuration, so it could correspond to a non-coding RNA.
Using the CMfinder-based comparative genomics pipeline, we found 22 novel putative RNA motifs. Two have already been experimentally confirmed as riboswitches. For several others, covariation and other characteristics suggest that they are functional structured RNAs, and we have proposed possible functions for many of the motifs. Thus, our pipeline appears to be useful for discovering novel RNAs, which in turn will contribute to our understanding of RNA biochemistry and bacterial gene regulation.
Our findings here and previously () demonstrate that the CMfinder-based pipeline is usually able to recover RNAs that are widespread, possess a highly conserved and extensive secondary structure, are roughly 60 nt or more in length, and are associated with homologous genes. Three candidate riboswitches have these characteristics (GEMM, Moco and SAH). The remaining three candidates, SAM-IV, the COG4708 motif and the motif are more narrowly distributed than most known riboswitches in that none of these motifs is found outside a single order in taxonomy level. This observation suggests that many of the undiscovered riboswitch classes have more narrow phylogenetic distributions than those discovered previously.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Mitochondria are primary ATP-producing organelles which originated from an ancestral alpha-proteobacterial endosymbiont (,). The oxidative phosphorylation (OXPHOS) pathway, consisting of the mitochondrial respiratory chain and the ATP synthase complex, is the final biochemical pathway in energy conversion (). However, mitochondria have been found to perform pivotal roles in many other metabolic, regulatory and developmental processes. During the evolution of the mitochondria, most genes of the ancestral endosymbiont have been transferred from the mitochondrial to the nuclear genome. The genes that reside on the mitochondrial DNA (mtDNA), e.g. 13 OXPHOS-related genes in human, are translated into their respective protein products by the mitochondrial translation machinery, which resembles the prokaryotic translation system (), comprising a mitochondrial ribosome (mitoribosome) and several translation factors. In mammals, all proteins that constitute the mitochondrial translation machinery are encoded by the nuclear genome. Also the mitochondrial genomes of fungi generally encode only very few mitochondrial ribosomal proteins (MRPs). In contrast, the mitochondrial genome of many protozoa and plants encode numerous MRPs. Most notably, the ‘primitive’ mitochondrial genome of the freshwater protist was found to contain as many as 27 MRP-encoding genes ().
Mitoribosomes have undergone major remodeling during their evolution. For instance, it has been found that, despite the fact that their rRNA content is only half the content of bacterial ribosomes (), mitoribosomes generally exceed the bacterial ribosomes both in molecular mass as well as in physical dimensions (). Mass spectrometry studies of the bovine and yeast mitoribosomes, which have served as model systems during the past years, revealed that mammalian mitoribosomes comprise about 80 MRPs (). In yeast, about 70 MRPs have been identified thus far, although protein–protein interaction data and mutational analysis suggest that this number might be substantially higher (). Interestingly, a recent proteomics survey of mitochondrial ribosome-related complexes in the kinetoplast-mitochondria of revealed an additional protein complex which was found to be specifically associated with the small subunit (SSU), but not with the large subunit (LSU) of the mitoribosome (). The biological role of the additional SSU-associated complex, which mostly consists of proteins unique to kinetoplastida, remains unknown.
Apparently, mitoribosomes have expanded their protein content in the course of evolution by acquiring numerous extra ‘supernumerary’ MRPs. Currently information regarding the functions of these supernumerary MRPs is rather limited. In addition, loss of MRPs that are normally part of the ‘bacterial core ribosome’ is also observed.
Recently, several attempts have been made to identify the MRPs in other eukaryotes, such as (), () and (), however, a comprehensive study of the evolution of the mitoribosomal proteome remains to be established. Such a study bears relevance for the potential identification of genes that cause mitochondrial dysfunctions in human patients. Since mitochondria perform many fundamental functions, mitochondrial dysfunction results in a wide variety of multisystemic diseases, predominantly affecting tissues with high metabolic energy rates (). These disorders are mostly caused by the dysfunction of one or more enzyme complexes of the OXPHOS system and several mutations have been identified in mtDNA () as well as in nuclear DNA (). Not only mutations in structural OXPHOS genes, but also mutations in genes involved in the mitochondrial transcription or translation, or in the assembly of OXPHOS complexes can result in OXPHOS disease. Although the vast majority of components of the mitochondrial translation system are nuclear encoded, thus far most mutations associated with mitochondrial translation defects have been reported in mtDNA-encoded tRNAs and rRNAs (,). Recently, mutations in four different nuclear gene products involved in mitochondrial protein synthesis have been reported in patients with mitochondrial disease: elongation factors EFG1 (), EFTs () and EFTu (), and small ribosomal subunit protein MRPS16 (). Additional information regarding which MRPs could be implicated in disease could be obtained by studying the evolutionary conservation of the MRPs.
The aim of the present study is 2-fold, from an evolutionary and a disease point of view: (i) gain insight into the evolution of the mitoribosome and its protein content in various eukaryotic species; (ii) prioritize MRPs as candidates for their involvement in mitochondrial disease. We have performed a comprehensive comparative genomics analysis of the MRPs in 18 eukaryotic species from which both the complete nuclear and the organellar genome sequences were available. Representatives are included from different phylogenetic groups: six metazoa ( and ), two fungi ( and ), one microsporidian (), one mycetozoan (), one plant (), one alga (), one apicomplexa (), one ciliophora (), two kinetoplastida ( and ), one diplomonad () and the mitochondrial genome of the freshwater protist . In addition, we included a balanced, phylogenetically diverse subset of completely sequenced prokaryotic genomes in our survey.
Our analysis retrieved multiple previously unidentified MRPs in several species, two of which constitute potential novel human MRPs. Furthermore, the current study established orthology relationships between seven MRPs reported to be fungi specific and ribosomal proteins from other eukaryotes. The establishment of these homology relationships enabled us to trace the origins of some of the supernumerary MRPs and to predict their molecular functions. In addition, we have investigated all MRPs in the presence of additional, newly acquired protein domains that might point at mitochondria-specific adaptations of the mitoribosome, and we have reconstructed the evolutionary history of the mitoribosomal proteome in terms of gains and losses of the MRPs along the different eukaryotic lineages. The newly detected mitochondrial ribosomal genes constitute, in addition to the set of evolutionary conserved MRPs, excellent new screening targets for human patients with unresolved mitochondrial oxidative phosphorylation disorders.
Genome data (both proteomes and genomes) of the eukaryotes described in this study were obtained from the Integr8 database (, Release 41) at EBI (), except for the genomes of (version 2 obtained from TIGR, ), (version 4 obtained from Sanger, ), (version 7.0 obtained from Broad institute at MIT, ) and (version 3.0 obtained from JGI, ). For most eukaryotic species, the organelle genomes were included in the gene set per species, otherwise those were downloaded separately from GOBASE () ().
In addition to the eukaryotic genomes mentioned above, we have also included a balanced, phylogenetically diverse subset of completely sequenced prokaryotic genomes. We have explicitly chosen not to include the complete set of completely sequenced prokaryotic genomes available, since the large number of prokaryotic proteins might cause an imbalance in the PSI-BLAST analysis described below. Sequence data (proteomes and genomes) of the selected prokaryotic genomes were all downloaded from the Integr8 database at EBI, and included: (gamma-proteobacteria), (firmicutes), (spirochetes), WashU, (alpha-proteobacteria), sp. PCC 7920, sp. PCC 6803, SS120 (cyanobacteria), (Deinococcus-Thermus group), (aquificae), (crenarchaea), (euryarchaea).
For each of the experimentally verified proteins of the (,), fungal () and mammalian () mitochondrial ribosomal proteins we have performed iterative PSI-BLAST searches () against a locally assembled protein database comprised of the proteomes of the organisms indicated above. For each of these searches, an E-value inclusion threshold of 0.005 was applied, and the proteins that were retrieved after five iterations were aligned using MUSCLE ().
For those cases where a candidate MRP, from a species from which the query MRP was still missing, appeared in the PSI-BLAST hit list, but with an E-value between 0.005 and 10, a reciprocal search was performed. If this search successfully retrieved a MRP within five iterations (E-value inclusion threshold of 0.005), it was added to the initial set of homologous proteins, and a new alignment was created. Whenever a given MRP was still not found in the predicted proteome of a eukaryotic organism, a hidden Markov model (HMM) of the respective protein was created from the above-mentioned alignments using the HMMER program (). Subsequently, the HMM was used to screen the genomic sequences at the DNA level. The latter screen served as a final control for missing MRPs that represent unpredicted genes in the genome annotation process. To delineate orthologous groups, neighbor-joining trees were derived from these alignments using Kimura distances (). A bootstrap analysis using 1000 samples was conducted for each phylogenetic tree. The selection of the orthologous group for each MRP was subsequently done by manually examining all trees. Assignment of orthologous relationships among the eukaryotic sequences was based on the species-phylogeny for eukaryotes (,) and the presence of these sequences within a monophyletic branch of the tree that contained known MRPs. Homologies between different families of MRPs, if not already found during the PSI-BLAST searches described above, were detected using the hhsearch algorithm (), by comparing HMM profiles constructed from aligned MRPs with each other, using an E-value cut-off of 1e−4.
Maximum-likelihood (ML) trees were created using PHYML (version 2.2.4) () based on alignments of the above-mentioned orthologous groups. Prior to the analyses, global and pairwise gaps were removed from the alignments. The appropriate model of protein evolution was selected using MODELSELECTOR (). Bootstrap support values were derived from 100 replicates.
All MRPs homologous to bacterial core RPs that were identified in the current research were examined for additional, potentially newly evolved domains. Using HMMER (), each MRP was analyzed with an HMM profile that was built from an alignment of its ancestral bacterial counterpart. These parts of the MRPs that were not covered by the respective HMM profile (minimum length 30 amino acids), were searched against the PFAM () and CDD () databases using HMMER () and RPS-BLAST () respectively. Hits below an E-value threshold of 0.01 were regarded significant.
We have employed a systematic analysis of nuclear and mitochondrial genomes in order to identify the mitoribosomal proteome content encoded in a wide variety of eukaryotic species. To avoid confusion regarding different nomenclatures that are used for ribosomal proteins, we have adopted the nomenclature for human MRPs, approved by the Human Genome Nomenclature Committee (). All proteins are designated MRPSxx or MRPLxx, where S and L stand for small and large subunit, respectively, and xx is the number of the corresponding bacterial ribosomal protein. The MRPs lacking bacterial orthologs (supernumerary proteins) are assigned higher numbers, starting at MRPS22 and MRPL37 for the small and large subunit, respectively. In yeast, unfortunately, the nomenclature is as variable as the methods employed for their characterization. Therefore, in order to prevent further confusion about proteins with the same name in human and yeast that are not part of the same orthologous group, we have indicated all human MRPs in capital letters throughout the manuscript (e.g. MRPS28), whereas for the yeast proteins we only capitalized the first letter (e.g. Mrps28, which corresponds to MRPS15 in human, also see and Supplementary Data).
Starting from a curated list of MRPs that have been identified in various studies, we first searched for homologs using several search algorithms (see Materials and Methods). Subsequently, the list of homologs was subjected to a phylogenetic analysis in order to define orthologous groups of MRPs. Not only does our analysis distinguish MRPs from their cytosolic counterparts, it also enabled for differentiation between ‘true’ MRPs and homologous ribosomal proteins of other organelles, such as the chloroplast (in plants) or apicoplast (in the apicomplexa ). We realize that this approach has its limitations, as it does not account for potential retargeting events that have been shown to occur, such as in , where a chloroplast-type S13 protein encoded by a diverged duplicated nuclear gene has been demonstrated to be targeted to the mitochondrion (). In order to account for such cases, we included such an experimentally verified protein in our list, whenever our analysis did not suggest a candidate MRP.
Altogether, our survey resulted in the identification of highly variable numbers of MRPs across the genomes of the investigated eukaryotic species, several of which had not been identified as MRPs previously (see , and Supplementary Data). The total number of identified MRPs ranged from 81 in most metazoan species, to 80 in yeast, to 63 in plants, down to a mere 39 MRPs in the apicomplexan . We recognize the fact that the latter numbers are probably an underestimate of the actual numbers of MRPs in these species, as there are no experimental studies to identify potential lineage-specific supernumerary MRPs. Moreover, since protein sequences of these species have been found to evolve relatively fast (), it is at least feasible that we failed to identify some MRPs that have evolved beyond the detection limits of our search algorithms.
Apart from mapping MRPs that are encoded by eukaryotic genomes, we were able to find hitherto undetected homology between experimentally verified supernumerary MRPs and bacterial ribosomal proteins that were presumed to have been lost from the primitive eukaryotic mitoribosome. Using sensitive profile–profile searches for detection of distant homology between protein families, we connected the MRPS24 orthologous group to the bacterial RPS3 ribosomal protein family (E-value 2.2e−5) and similarly, we connected the MRPL47 orthologous group to the bacterial RPL29 ribosomal protein family (E-value 8.6e−7) (see ). The complementary phylogenetic distribution patterns of the respective genes are in support of this suggestion. For example, MRPL47 is present in all eukaryotes analyzed (except for and ), but not in bacteria. The opposite phylogenetic pattern is observed for bacterial ribosomal protein RPL29, for which no homolog was identified in any eukaryotic genome. Apparently, these MRPs have diverged almost beyond homology detection limits during the course of evolution (see ), possibly as an adaptation to specific mitoribosomal functionality. Given the detected homology between the above-mentioned MRPs and bacterial ribosomal proteins, we have placed the MRPs in the orthologous group of their respective bacterial counterparts (see and ).
We also encountered some cases where we could not detect an expected MRP ortholog in some species, which might be caused by incompleteness of the respective genomes. For example, despite the fact that we could identify orthologs of MRPL50, MRPS28 and Mrp10 in all metazoa analyzed in this survey, as well as in the genome (data not shown), we were unable to detect these MRPs in the genome of . Likewise, MRPL56 was found in all metazoa, including , but not in (see ). In addition, in some instances our analysis proposed different proteins than have been reported in earlier reports. The discrepancies between our study and that of others are outlined in the Supplementary Data.
Not unexpectedly, MRPs were not detected in the genomes of the microsporidian and the diplomonad . These organisms do not contain full-fledged mitochondria, but instead they contain mitosomes () or mitochondrial remnants (), which typically lack organellar DNA. Consequently, these organisms have no need for mitochondrial translation machineries, explaining the observed loss of MRP-encoding genes.
Yeast Rsm22 is homologous to an rRNA methyltransferase and has been shown to be part of the SSU of the yeast mitochondrial ribosome (). The association of Rsm22 with the SSU has also been confirmed in a large-scale affinity-capture mass spectrometry study (). We have detected an ortholog of Rsm22 in most eukaryotes, as well as in some prokaryotes. Interestingly, the STRING database () shows that the gene encoding the Rsm22 ortholog in some prokaryotes is clustered on the genome with genes that encode ribosomal proteins (not shown), suggesting that the Rsm22 orthologs in these species are also associated with the ribosome. In , the genes encoding Rsm22 and Cox11 are fused. Cox11 is an assembly factor of the OXPHOS enzyme complex cytochrome c oxidase and in , Cox11 has been shown to associate with the mitoribosome and is postulated to function in co-translational insertion of copper ions into Cox1, a subunit of cytochrome c oxidase (). Finally, Rsm22-deficient yeast strains exhibit a growth defect on non-fermentable (respiratory) carbon sources, indicative of mitochondrial respiratory dysfunction (). All these lines of evidence suggest that Rsm22 is a protein of the mitoribosomal small subunit that is essential for proper mitochondrial function. As Rsm22 is a predicted rRNA methyltransferase, it might be involved in methylation of mitochondrial rRNA, and as such play a role in the regulation of protein translation efficiency.
Yeast Mrp10 was shown to be a component of the SSU of mitoribosomes as well and disruption of the gene resulted in a mitochondrial translation defect and a tendency to accumulate deletions in mtDNA (). Jin . observed low sequence similarity to yeast Mrpl37, a constituent of the LSU (). However, we identified yeast Mrpl37 as an ortholog of human MRPL54 and did not detect any significant sequence homology between Mrp10 and Mrpl37. Nevertheless, we did find Mrp10 orthologs in many eukaryotes, all of which were found to contain a CHCH domain (oiled coil 1—elix 1—oiled coil 2—elix 2, PFAM domain 06747). Within each helix of this CHCH domain, two invariant cysteine residues are present in a C–X–C motif. Besides in Mrp10, CHCH domains have also been detected in Cox19 and in the respiratory Complex I (NADH-ubiquinone oxidoreductase) 19 kDa subunit (NDUFA8) (). In yeast, Cox19 is present in the cytoplasm as well as in the inter membrane space of the mitochondria, where it is suggested to play a posttranslational role in the assembly of cytochrome c oxidase through copper transport to mitochondria. It has been proposed that the conserved cysteine residues of the CHCH motif in Cox19 mediate as ligands for the copper ions (,). Most likely, yeast Mrp10 and their orthologs in other eukaryotes, including human, constitute an essential component of the mitochondrial translation machinery and we anticipate that the CHCH domains that are present in this orthologous group, as well as in Complex I subunit NDUFA8, are involved in trafficking of metal ions.
We also detected a potential third novel human MRP, Ppe1, however this case is less clear-cut than these described above. Ppe1 is a protein, which has been shown to be part of the SSU of the yeast mitoribosome (). We detected Ppe1 orthologs in all eukaryotes analyzed, except for and (and the mtDNA lacking eukaryotes). However, the localization of Ppe1 in the mitochondrion is questionable, as the report by Kitakawa . is the only indication for this (). Moreover, the yeast and human Ppe1 have previously been reported to function as protein phosphatase methylesterases of Protein phosphatase 2A (,), which are mainly localized in the nucleus and cytoplasm (). Additionally, yeast Ppe1 mutant strains do not exhibit any growth defects (), while mutations affecting synthesis or function of respiratory chain components generally result in growth impairment on a non-fermentable carbon source. Thus currently there is, besides the reported association with SSU of the mitoribosome in yeast, no clear indication that Ppe1 is implicated in mitochondrial translation. Therefore, additional research is needed to corroborate the cellular localization and function of yeast Ppe1 and its orthologs in other eukaryotes.
The protein composition of mitoribosomes is variable, not only in terms of the number of MRPs, but also in terms of the specific MRPs contained (see and ). By mapping the distribution of the orthologous groups of MRPs onto a reconstructed tree of the eukaryotes () while always considering the most parsimonious scenario, we obtained an overview of the history of gains and losses of MRPs during the divergence of eukaryotes (see ). Mitoribosomes originated from the bacterial ribosome present in the alpha-proteobacterial ancestor of mitochondria, consisting of a microbial core of 54 ribosomal proteins in total. Already in the earliest stage of eukaryotic evolution, prior to the major divergence of the fungal, plant and metazoan lineages, several supernumerary proteins were recruited to the mitoribosome, while only one bacterial core protein was lost at this stage (RPS20, see ), resulting in a primitive eukaryotic mitoribosome of 68 MRPs. Subsequent to the gain of this primary set of 15 proteins, the mitoribosome diversified along the various eukaryotic lineages and a number of lineage-specific gain and loss events can be observed. Another gain of MRPs occurred at the level of the metazoa, prior to the bifurcation of nematodes and eumetazoa. At this stage, 14 proteins were recruited to the metazoan mitoribosome, whereas seven bacterial core proteins were lost. Major protein gains at these two time points have also been observed in the evolutionary history of Complex I (). Additionally, 11 fungi-specific MRPs were recruited after the animal-fungi radiation. In the lineage towards the kinetoplastida, a major surge of 29 proteins is observed. These proteins have been shown to comprise a protein complex that is associated with the SSU of kinetoplastid-mitoribosomes (). The role of this additional protein complex and its association with the SSU is still unclear. Apart from this major gain in the kinetoplastida, we observe loss of several proteins in the AVE (alveolates, viridiplantae and excavates) lineage, predominantly consisting of bacterial core proteins, but also including supernumerary MRPs in some lineages. Possibly, some of these proteins remained undetected due to sequence divergence. The fact that the AVE lineages are dominated by loss events of MRPs is likely due to the lack of experimental data of mitoribosomal content in these species.
Compared to the 54 ribosomal proteins present in the alpha-proteobacterial ancestor, 81 MRPs in human represent a substantial gain in complexity (see B). However, the observed increase in complexity following the endosymbiotic event is not uncommon, as it has also been reported for electron transport chain complexes (). It has been proposed that the recruited supernumerary proteins are mainly involved in the assembly and stabilization of the complexes (,). The mammalian supernumerary MRPs are shown to be mostly localized to the peripheral regions of the mitoribosome (). Since all proteins synthesized by metazoan mitoribosomes are inserted into the inner mitochondrial membrane, at least some of the metazoa-specific supernumerary MRPs are presumed to be associated with positioning the mitoribosome during co-translational insertion of nascent polypeptides into the membrane. Interestingly, a protein that has been proposed to be essential for co-translational insertion in yeast, Mba1 (MRPL45, see ‘Origins from outside of the mitoribosomal proteome—MRP recruitment’), was already part of the primitive eukaryotic mitoribosome and has been recruited from the alpha-proteobacterial ancestor. Similar to the evolution of Complex I, we observe the recruitment of proteins of bacterial origin to the mitoribosome (MRPL45 and Rsm22), however, this number is considerably lower (only two, compared to six in Complex I), especially considering that in the evolution of the mitoribosome more proteins were added compared to Complex I (37 and 32 in the human mitoribosome and Complex I, respectively).
As outlined above, the protein content of the mitoribosome has significantly increased in most species during the evolution of the mitochondria. Our analysis indicates that in several stages of the mitoribosomal evolution various extra proteins, the supernumerary MRPs, have been added to the mitoribosomal proteome. Some of these supernumerary MRPs were found to be present in all eukaryotes that were investigated, but not in bacteria, such as MRPS29 (see ). Clearly, these MRPs have been recruited in an early stage of the evolution of the eukaryotes, before the formation of major eukaryotic kingdoms. In contrast, other MRPs have originated relatively recently, such as MRPL37, which is specific for the metazoa. However, it is not only interesting to know ‘when’ certain MRPs were added to the mitoribosome, it is also enticing to trace their origins. We found that several of the supernumerary MRPs are homologous to other proteins, some of which are also part of the mitoribosome (see ). The latter finding implies that gene duplication events have played a substantial role in the expansion of the mitoribosomal proteome (see ). Below we will discuss some of the most protruding cases.
Some supernumerary MRPs are homologous to MRPs that are part of the bacterial core of the mitoribosome. These proteins have seemingly been duplicated from within the mitoribosomal proteome. One of the most straightforward examples is the emergence of three variants of MRPS18 in the metazoan lineage resulting from gene duplication events (see ). In contrast with earlier reports (), we identified three MRPS18 variants in and we found that all three metazoan variants emerged from the same ancestral metazoan sequence, which is unlikely to be derived from chloroplast S18, as was reported by Cavdar Koc and co-workers () (see ). All three MRPS18 variants seem to be essential in embryonal development, as targeted disruption of these genes result in an embryonic lethal phenotype (). Independent from the observed MRPS18 duplications in metazoa, it is interesting to note that most sequenced actinobacterial genomes encode two variants of the S18 gene, probably the result of a gene duplication event in these species. Currently, experimental evidence regarding the functions of the different MRPS18 versions in metazoa is lacking, but it is believed that each mitoribosome contains only one copy of MRPS18, giving rise to a heterogeneous population of mitoribosomes ().
Another example can be found in the duplication of MRPS10 in the lineage of the metazoa, which gave rise to MRPL48 (see and ). This duplication event constitutes one of the cases where the duplicate gene product has become part of the other mitoribosomal subunit, in this case from the SSU (MRPS10) to the LSU (MRPL48). Likewise, a duplication event at the base of the metazoan radiation seems to have given rise to the supernumerary MRPs MRPS30 and MRPL37, both of which are only present in metazoa (see ). Interestingly, MRPS30 is also known as programmed cell death protein 9 (PDCD9 or p52) (,). Another supernumerary protein, MRPS29, is also implicated in apoptosis, as it has been shown to be identical to death-associated protein 3 (DAP3) (,). Given the dual roles of these MRPs, the processes of mitochondrial translation and apoptosis are apparently closely coupled ().
We found that some supernumerary MRPs have probably been recruited to the mitoribosome, as they display significant homology to other mitochondrial proteins, some of which are also implicated to play a role in mitochondrial translation (see ). Of course, it cannot always be resolved if an MRP has been part of the mitoribosome prior to the duplication event. For some instances, we can use the phylogenetic distribution patterns of the copies in order to point out which was first. One of such examples, where a pre-existing protein has been recruited to the mammalian mitoribosome, is the acquisition of MRPL39, which is homologous to the -terminal domain of threonyl-tRNA synthetases and which has a universal phylogenetic distribution. This homology has been noticed before and it is suggested that MRPL39 can bind tRNA in a similar manner as the threonyl-tRNA synthetase (). Conceivably, the -terminal tRNA-binding domain of a mitochondrial threonyl-tRNA synthetase was recruited to the mitoribosome in order to compensate for the loss of bacterial ribosomal proteins that are involved in tRNA binding, such as RPS13 and RPL5 (see and ).
The most striking example of MRP recruitment is that of MRPL45 (see ). First we found that MRPL45 and yeast protein Mba1 belong to the same orthologous group, which in addition comprises proteins from all metazoa, the plant , the alga as well as from the mycetozoan . Mba1 is a mitochondrial protein associated with the matrix face of the inner membrane, which associates with the highly conserved inner membrane protein Oxa1. Mba1 and Oxa1 orchestrate the insertion of both mitochondrial- and nuclear-encoded proteins from the mitochondrial matrix into the inner membrane (,). Moreover, we detected homology between MRPL45 (including Mba1) and Tim44 (see ), a peripheral mitochondrial inner membrane protein that is an essential component of the import motor PAM of the TIM23 complex, which is implicated in the translocation of proteins from the inter membrane space to the mitochondrial matrix [for reviews see (,)]. The apparent sequence homology, as well as functional homology (i.e. protein translocation), suggests that MRPL45 and Tim44 share a common alpha-proteobacterial origin. Many alpha-proteobacteria contain two distinct types of Tim44/MRPL45-like genes (see ).
A close examination of MRPL45 and Tim44 proteins reveals that the -terminal region of Tim44 is absent in MRPL45 (not shown). This region is supposed to be located in the mitochondrial matrix where it interacts with mitochondrial Hsp70 chaperone, which is also a component of the PAM complex (). Apart from its association with Oxa1, MRPL45 has been shown to be a constituent of the mitoribosomal LSU in multiple studies (,). Moreover, it has been shown that in yeast, transcription of MBA1 is tightly co-regulated with that of genes encoding MRPs (), implying a functional link between the respective gene products. The anticipated association of the mitoribosome and membrane-bound constituents of the mitochondrial protein insertion machinery would explain the fact that up to 50% of the bovine mitoribosomes are found to associate with the inner membrane fraction of mitochondria (). Taking all this in consideration, we postulate that MRPL45 functions as a bridging factor between the mitoribosome and Oxa1, and as such constitutes an essential component for successful co-translational insertion of proteins into the inner membrane of mitochondria in eukaryotes. Interestingly, the MRPL45 orthologous group, which also includes the bacterial Tim44-like proteins, shows a similar phylogenetic distribution pattern as the orthologous group of Cox11, an assembly factor of cytochrome oxidase, which is involved in the co-translational insertion of copper ions into the nascent Cox1 protein (). The observed pattern of co-presence and absence of these genes across a phylogenetically diverse set of species indicates that their gene products might be involved in the same process (). As Cox11 has been shown to associate transiently with the mitoribosome as well, it is feasible that MRPL45 and Cox11 act together in order to functionally insert Cox1 into the inner mitochondrial membrane ().
Other examples of MRPs that have been recruited to the mitoribosome are the supernumerary proteins MRPL43 and MRPS25 (see ), which are homologous to each other as well as to the B8 subunit (NI8M) of respiratory Complex I (NADH-ubiquinone oxidoreductase) of the OXPHOS system (,). Presumably, MRPL43 and the B8 subunit of Complex I originated from a single gene that was present in the primitive eukaryote, as they are present in a phylogenetically wide range of organisms. Most likely, another duplication of MRPL43 then gave rise to MRPS25, which is only found in metazoa and fungi. More examples of MRP recruitment include yeast MRPs Mrp1 and Rsm26, which are both part of the iron/manganese superoxide dismutase family (), and MRPL44, which is homologous to a wide range of double-stranded RNA-binding proteins (see ).
Several comprehensive experimental studies on the yeast mitochondrial translation machinery have significantly extended its mitoribosomal proteome (). However, for a large number of MRPs that have been characterized in yeast thus far, no significant or conclusive sequence similarity has been reported to ribosomal proteins from other sources. In the current study, we were able to link seven of these seemingly fungi-specific MRPs to known ribosomal proteins in other species, as will be outlined below.
Apart from the homology between yeast Mba1 and MRPL45, which has already been discussed above, we identified Mrp49 as the ortholog of MRPS25. It is, however, surprising that the orthologs of MRPS25 found in as well as in are reported to be located in the large instead of the small subunit of the mitoribosome (,), in contrast with the location of bovine MRPS25 (). It is possible that the subunit localization of the MRPS25 orthologs is different between metazoa and fungi; however, further experimental evidence is needed in order to clarify this.
We also connected Rsm27, a yeast protein of the small subunit (), to the MRPS33 protein family. In addition, we detected MRPS33 orthologs in all metazoa as well as and . Moreover, we found that yeast MRPs Mrpl28 (), Mrpl44 (), Mrpl50 () and Mrpl40 () are in fact the orthologs of MRPL40, MRPL53, MRPL9 and MRPL24, respectively. Both MRPL24 and Mrpl40 were found to contain a KOW motif (PFAM domain pfam00467), which is present in a variety of ribosomal proteins as well as in the bacterial transcription elongation factor NusG (). The homology between Mrpl50 and bacterial RPL9 has been noted before (), but it was dismissed as being insignificant, since the sequence similarity was not conserved throughout the full length of Mrpl50. In our analysis, however, we did detect significant homology between Mrpl50 and the MRPL9 family (including bacterial RPL9). Our finding is supported by the fact that Mrpl50 and MRPL9 representatives reside within the same protein superfamily (SSF55658, ‘L9 N-domain-like’) ().
Finally, we were able to confirm the distant homology between Var1 and Rps5, two mtDNA-encoded homologs of bacterial RPS3 in fungi, as was previously reported by Lang and colleagues (). Var1 and Rps5, which are the only mtDNA-encoded MRPs in yeast and , respectively, have both been identified as essential components of the SSU of the mitoribosome, showing similar phenotypes in mutant strains (). Despite the fact that RPS3 has been shown to play an essential role in bacterial ribosome assembly, and that Var1 and Rps5 are both essential for assembly of the mitochondrial SSU in fungi, we were unable to detect candidate mitochondrial RPS3 homologs in the metazoa. In the latter organisms, this essential role in ribosome assembly could be performed by MRPS24, which displays distant homology to RPS3 (see ). Moreover, we could detect mtDNA-encoded RPS3 homologs in and (,), the latter of which constitutes a peculiar case where the RPS3 encoding gene is split and both the 5′ and 3′ parts are fused to an ORF of unknown function (ORF425 and ORF1740, respectively). In any case, the fact that genes encoding RPS3 homologs are present on several mitochondrial genomes implies an alpha-proteobacterial origin of this protein.
Compared to bacterial ribosomes, mammalian mitoribosomes contain scarcely half as much rRNA and over twice as much protein, due to the presence of enlarged bacterial core proteins and supernumerary proteins (). Previously, it has been hypothesized that these proteins might compensate for the loss of rRNA segments in the mitoribosome (,). Recently however, it was shown that many supernumerary MRPs occupy new quaternary positions in both the large and small subunits of the mitoribosome (), inconsistent with this hypothesis. We compared the sequences of all core MRPs with their bacterial counterparts in order to explore the possibility that core MRPs contain additional domains that perform extra, mitochondria-specific functions. Indeed, we observe that MRPs are significantly larger (on average almost twice as large) than their bacterial ancestral sequences (see Supplementary Data, Figure S8). In some cases, we could identify functional domains within these newly evolved regions, which we expect to be linked to ribosome assembly or function (see ). For instance, we identified a RNA recognition motif (RRM_1, pfam00076) and a copper-binding domain (Cu_bind_like, pfam02298) in the -terminal regions of MRPS19 and MRPL22, respectively (see ). The RRM motif that is present in MRPS19 () is found in a wide variety of RNA-binding proteins, which are implicated in a wide variety of cellular processes. Here, the motif is most likely involved in increasing the binding affinity to and/or stabilization of the tertiary structure of rRNA. The role of the copper-binding domain in MRPL22 is less clear. It might be involved in the co-translational insertion of copper ions into proteins that require copper for activity, in a similar way as has been observed for the cytochrome c oxidase assembly factor in yeast, Cox11 () (also see above). The functions of other functional domains that are fused to MRPs are still obscure, such as a bacterial-type membrane protein (COG5373) to MRPL27 in (see ).
In the present study, we have mapped the evolution of the mitochondrial ribosomal proteome using a comparative genomics approach that combined the most recent experimental data and computational techniques. The results of our study hint at a complex evolutionary scenario in which an ancestral ribosome of alpha-proteobacterial descent doubled its amount of protein in most eukaryotic lineages by elongation of its proteins and by recruiting new proteins of diverse origins. We observe that the mitoribosome has a complex history in terms of MRP gene gain and loss events during evolution. A major protein gain of the mitoribosome in eukaryotic evolution occurred prior to the divergence of the main eukaryotic domains, resulting in a primitive eukaryotic mitoribosome of 68 MRPs. Subsequently, the mitoribosome diversified along the different eukaryotic lineages with a continuous recruitment of new proteins, including a large gain after the bifurcation of the animal-fungi clade in each lineage. Another notable surge has taken place in the kinetoplastida lineage, where 29 new proteins were found to form a new, distinct protein complex with unknown function, which associates with the SSU of the mitoribosomes ().
In addition to the gain of MRPs, we also observe that numerous MRPs have been lost during the course of evolution, some of which have occurred independently in different lineages, as guided from ribosomal protein content that is encoded by the ‘ancestral’ mitochondrial genome of (). The extensive loss of genes encoding core MRPs in some lineages contrasts with the evolutionary scenario that is observed for another mitochondrial protein complex with an alpha-proteobacterial ancestry, respiratory Complex I (NADH:Ubiquinone oxidoreductase). During the evolution of this protein complex, which has approximately tripled its size in most eukaryotes, hardly any gene loss was observed, except for these lineages in which the complete respiratory complex was lost (). Moreover, the bacterial core of Complex I has been restricted from gene loss events in all eukaryotic genomes analyzed thus far, which contrasts with the scenario of the mitoribosome. What is the underlying reason for the observed differences in the evolutionary trajectories? Evidently, some of the core ribosomal proteins are not as essential as the core subunits of Complex I for proper functioning of the protein complex, and are therefore expendable. This is the case for RPS20 and RPL25, which are absent from most mitoribosomes, for example. RPS20-deficient strains were found to be viable, but showed an increased misreading ability of nonsense codons (); RPL25-defective and wild-type ribosomes were found to translate at the same rate , albeit that the mutant type was less efficient (). In contrast, some MRPs that are missing in most eukaryotic species, such as MRPS1 and MRPL5, are indispensable for translation in and result in lethal phenotypes when mutated (,). Moreover, MRPL5-deficient yeast strains display a respiratory-deficient phenotype, probably caused by a defect in the mitochondrial translation machinery (). Apparently, the evolution of the mitoribosomal proteome involved several drastic events that altered the absolute necessity of some ribosomal proteins.
In addition, the present study was performed in order to prioritize MRPs for their potential involvement in mitochondrial respiratory diseases. Based on the fact that essential genes appear more evolutionary conserved than non-essential genes, the bacterial core ribosomal proteins and translation factors that are conserved in most eukaryotes constitute excellent screening targets for human mitochondrial disorders. Mutations in non-essential MRPs could be risk factors for these diseases. Thus far, mutations in patients with mitochondrial respiratory disorders have been reported for the gene encoding small ribosomal subunit protein MRPS16 (). But what about the supernumerary MRPs? In general, they are phylogenetically less conserved than the bacterial core proteins; however, their recruitment bears functional relevance, as is indicated by knock-out studies in yeast and . As such, the two novel human MRPs, orthologs of yeast Rsm22 and Mrp10, are good candidates for unresolved mitochondrial disorders, since they show a wide phylogenetic distribution, have not been investigated before and yeast mutants show clear respiratory-deficient phenotypes. Translation-related mitochondrial disorders are not solely caused by mutations in mitoribosomal genes, but also by mutations in genes encoding factors that functionally interact with the mitoribosome, such as mitochondrial translation elongation factors EFG1 () and EFTs (). In addition to the newly discovered human MRPs, these interactors constitute excellent screening targets for human patients with unresolved mitochondrial OXPHOS system disorders.
Altogether, our study reveals that the mitoribosomal proteome has, after initially having been acquired via the alpha-proteobacterial endosymbiont, been subjected to a complex evolution that involved numerous gain and loss events, resulting in a large variety in mitoribosomal protein content between different eukaryotic lineages. Potentially, the increased amount of proteins that seem to be associated with all modern eukaryotic mitoribosomes might reflect specific adaptations to mitochondria-specific functions, or, it might even reflect the increased complexity of eukaryotes themselves. Finally, we expect that additional experimental studies of mitoribosomal proteomes, such as the one that was recently performed on mitoribosomes, will reveal an even more dramatic picture of the evolutionary trajectory of this protein complex. It will be a major challenge to gain more insight in the functional roles of the newly acquired MRPs.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
With complete genome sequencing now routine, biology faces the fundamental problem of large-scale automatic annotation of gene function. Local alignment tools () predominate in automatic annotation, because many of them have the heuristics and accurate -values required for screening large databases rapidly and evaluating search results. In some motif searches, however, including searching for complete domains within a protein sequence, local alignment has two shortcomings. First, it is distracted by strong but incomplete motif matches. Second, it does not align domains over their entire length and does not define their boundaries. Ideally, therefore, complete domains should be aligned to protein subsequences, in a ‘semi-global alignment’ (). Accordingly, this article compares the conserved domain (CD) retrieval of a new semi-global alignment tool GLOBAL (GLObal Blocks Aligned Locally) with local and semi-global versions of HMMer (a Hidden Markov model alignment tool) (,) and to RPS-BLAST () [a local alignment tool, and the current default search tool for NCBI's conserved domain database (CDD) ()].
To elaborate on the CDD, it offers a comprehensive classification of the CDs composing proteins, modeling each CD as a multiple sequence alignment (MSA). Certain MSAs have been manually curated, to designate a contiguous subset of columns as the ‘footprint’ of the corresponding CD. For present purposes (Discussion section), we further partitioned the CD footprint into contiguous ‘blocks’ of conserved columns, separated by ‘spacers’ of poorly conserved columns. Manual curation constantly refines the MSA to increase the consistency of the blocks with collateral information such as structural alignments or function. Tools related to PSI-BLAST () then convert the CD footprint into a position-specific scoring matrix (PSSM).
A benchmarking test set based exclusively on the VAST protein structure alignment program () permitted us to compare the CD retrieval performance of GLOBAL, HMMer and RPS-BLAST (Results section). To summarize our main findings, the semi-global alignment tools, GLOBAL and HMMer_semi-global (i.e. HMMer in ‘global’ mode), essentially had indistinguishable CDD retrieval performance, and both outperformed the local alignment tools, HMMer_local (i.e. HMMer in ‘local’ mode) and RPS-BLAST.
Presently, GLOBAL's main advantage over HMMer_semi-global is that GLOBAL has unusually accurate -values. Programs for building protein profiles through iterative search, e.g. PSI-BLAST (,), require accurate -values to avoid corrupting their profiles with false positives. GLOBAL's accurate -value therefore opens up the possibility of an iterative program for finding ‘complete’ domains, either within a single protein or a group of proteins. Moreover, GLOBAL aligns individual blocks to a query protein sequence with gapless local alignment, so it is readily amenable to the word-match heuristics that accelerate RPS-BLAST searches through the CDD (). We are currently investigating word-match heuristics for GLOBAL.
The layout of the article is as follows. The ‘Materials and Methods’ section describes the test set for benchmarking CDD retrieval, the LROC score used to measure retrieval performance, the GLOBAL algorithm and -value, and the implementation of the retrieval tools.
The ‘Results’ section assesses first the retrieval performance of the tools, and then the -value accuracy of the tools with the best CDD retrieval (GLOBAL and HMMer_semi-global). Finally, the ‘Discussion’ section examines the implications of our findings. In particular, the ‘independent alignments approximation’ in the ‘Materials and Methods’ section provides an -value for many types of global and semi-global alignments, in response to the statement: ‘There is no theory for [the statistical significance of] global alignment.’ (). The independent alignments approximation is not as simple as an extreme-value approximation, but it can be orders of magnitude more accurate ( and ).
The Supplementary Data present the mathematical analysis relevant to the basic concepts in the main article. Interestingly, the Supplementary Data shows that GLOBAL can be viewed either as a classical alignment technique or an unusually simple HMM. Thus, GLOBAL provides a convenient bridge between HMMs and the wealth of statistical and computational techniques available for classical alignment (as aforementioned, e.g. GLOBAL is susceptible to the word-match heuristics used in RPS-BLAST).
Our standard of truth for comparing different CDD retrieval methods used two databases. First, single-linkage clustering based on BLAST -values of ≤10 yielded a non-redundant set of 10 185 proteins from the protein structure database PDB. The non-redundant database, ‘DB_10185’, is available at: . Second, the CDD [version 2.02, available at: ()] contained a set of 331 MSAs, each not contained within any larger MSA in the familial hierarchy of the CDD, each manually curated and each containing at least one sequence with a known structure. Thus, each of these 331 MSAs could be structurally aligned through that known structure to the members of DB_10185 from PDB. We extracted the set from CDD, to create a database, ‘DB_331_CD’.
VAST is a local structural alignment program (), whose complete list of alignments is available at: . From the list, we found all pairs in DB_331_CD and DB_10185 (from PDB) whose structural alignments had a VAST -value of ≤10. Each structural alignment had a first and last column in the corresponding MSA from DB_331_CD, the columns between forming a ‘VAST footprint’, analogous to the CD footprint described in the Introduction. To discount strong but incomplete local structural similarities, our standard of truth considered an MSA to contain a ‘complete’ CD, only if the VAST footprint occupied at least 80% of the CD footprint. The two databases (DB_10185 and DB_331_CD), the intersection of the VAST and CD footprints, and the test set containing the structurally related pairs, are available at: .
The following receiver operating characteristic (ROC) analysis is an established method for measuring the performance of a database search method (,). If the protein query contains a CD, the CD is ‘relevant’; otherwise, it is ‘irrelevant’. Let the total number of irrelevant CDs be . In response to a protein query, a CDD search tool produces a retrieval list, which ranks all CDs in the database. The ‘ROC curve’ plots the fraction of relevant CDs preceding the -th irrelevant CD against the fraction . The ‘ROC score’ is the area under the ROC curve. Analogously, the ‘ROC curve’ is the ROC curve truncated on the -axis after the first irrelevant CDs, with the ROC score being the area under the ROC curve divided by (,). The normalization by ensures that an ideal retrieval method (which returns all relevant CDs before any irrelevant CD) receives a ROC score of 1.0.
The Supplementary Data describe the ‘Localization-Response Operating Characteristic’ (LROC) curve (), which accounts for alignments and is therefore slightly preferable to the ROC in the context of CD retrieval. LROC curves appear in the figures; the ROC curves are similar.
To evaluate CDD retrieval over all protein queries, we merged the retrieval lists for each protein query into a single ‘pooled list’, by sorting the CDs on their -values (). The ROC procedure was carried out on the pooled retrieval list.
GLOBAL exploits the block structure of the CDD directly. Call any (possibly empty or full) subset of contiguous columns in a block, a ‘sub-block’. To identify a CD within a query protein sequence, GLOBAL aligns one sub-block from each CD block, in order, against the sequence (). GLOBAL only aligns block columns to the sequence, never spacer columns.
On one hand, GLOBAL sometimes aligns all columns in a block or even all blocks to the sequence (e.g. alignment π in ). On the other hand, it sometimes leaves unaligned arbitrary numbers of columns at the ends of blocks (e.g. alignment π) or even entire blocks (e.g. alignment π leaves the purple block unaligned). Moreover, aligned block columns may overlap with unaligned columns from other blocks (e.g. alignment π makes the purple block overlap with some unaligned columns from the blue block ).
In aligning a CD of blocks ( = 1, … ,) against a sequence = , … ,, GLOBAL aligns the CD blocks in order to the sequence, applying gapless local alignment to each block (). GLOBAL alignments have the usual 1-1 correspondence with paths through an alignment graph (e.g. ). A GLOBAL alignment π has weight equaling the sum of scores from the sub-block columns it aligns to the sequence letters. The GLOBAL score is the maximum weight over all possible alignments π. A GLOBAL alignment with weight = is ‘optimal’. (see Supplementary Data for details.).
Dynamic programming can find the optimal GLOBAL alignment, as follows. Initialize by aligning against with the Smith–Waterman algorithm () for gapless local alignment. To induct, note that for > 1, an optimal GLOBAL alignment of , … , against decomposes into: (i) an optimal GLOBAL alignment of , … , against some subsequence … ; and (ii) an optimal local alignment of against , … ,. Thus, with the optimal alignment scores of , … , against all subsequences , … , in hand, optimize the alignment score of , … , against each subsequence , … , by maximizing over = 1, … , in the decomposition above. As usual, discover optimal alignments by backtracking through an alignment matrix. The GLOBAL algorithm requires () time, where CD footprint has length and the protein sequence has length . (See Supplementary Data for details.)
GLOBAL alignments have some desirable properties. GLOBAL assigns score 0 to unaligned sequence, regardless of its length, between aligned block columns (). [In HMM terminology, the blocks are ‘free modules’ ()]. GLOBAL therefore respects CDD curation, by freely permitting insertions of arbitrary length between conserved blocks in a protein. GLOBAL assigns score 0 to unaligned columns at the block ends (). Thus, it recognizes that in evolution, a secondary structure is frequently conserved at its center but not at its end. GLOBAL assigns score 0 to entire unaligned blocks. It therefore recognizes that in evolution, protein domains sometimes experience large deletions.
One evolutionary event does disadvantage GLOBAL, however. In a CD with a long block, unusual insertions into a protein might split the block into sub-blocks. GLOBAL can then align at most one sub-block to the protein sequence correctly. The ‘disadvantage’ has a trivial remedy: before retrieving with GLOBAL, just split long blocks in the CDD arbitrarily.
DB_331_CD contains = 331 CDs. For a random protein sequence, the -value = is the expected number of CDs with a -value not exceeding . GLOBAL calculates its -value under an ‘independent letters model’, in which a random protein sequence consists of independent, identically distributed amino acids. To optimize empirical retrieval performance, in the random model for each CD, the amino acid frequency was ‘composition corrected’ (,), to match empirical frequencies within the corresponding CD footprint.
Although the Supplementary Data give a mathematical analysis of the GLOBAL -value, the basic concepts appear here.
Before proceeding, note the following concept, ‘Markov computation’, explained formally in the Supplementary Data and used repeatedly below. A dynamic programming algorithm has a state that changes in response to successive inputs. If the inputs are random and independent of previous states and inputs, the successive states of the dynamic programming computation form a Markov chain. Variants on matrix multiplication can therefore compute the distribution of the successive dynamic programming states. Many articles have been written on special cases of Markov computation (,).
The following argument makes many assumptions and is ultimately justified by the success of the resulting -value ().
An optimal GLOBAL alignment respects block order. Thus, it usually aligns the -th block within some subsequence of of ‘effective length’ (). For convenience, assume that does not depend on , so = . (In practice, the resulting approximation is both accurate and relatively simple.) Let be the optimal gapless local alignment of against a random sequence of length . Assume the ‘independent alignments approximation’, that for some length , the GLOBAL score has about the same distribution as the sum of independent variates .
The first task is to determine the effective length . Let the block have length ( = 1, … ,). On one hand, GLOBAL locally aligns the blocks ( = 1, … ,) in order against . The starting points of the block alignments within the alignment matrix can be chosen in … {!/[( − )!!]} different ways. (; There, the -coordinates of the starting points can be chosen in … ways; the y-coordinates, in !/[( − )!!] ways.)
On the other hand, under the independent alignments approximation, each of the blocks ( = 1, … ,) is ‘independently’ aligned locally against a random sequence of effective length . The starting points of the block alignments within the sequences can be chosen in , … ,() different ways. (See . There, the -coordinates of the starting points can be chosen in , … , ways; the -coordinates, in ways.)
Then, the number of ways of choosing the starting points is the same if
if = {!/[( − )!!]}. To make the formula for apply in the case < , we replaced with + − 1, to produce = [( + − 1)!/{( − 1)!!}], the effective length used throughout this article. (The expression ( + − 1)!/{( − 1)!!} is the number of ways of choosing objects from objects with replacement, so any object can be chosen several times.)
The next task is to find the distributions of . [Approximations based on Gumbel distributions () for were consistently inferior to the following ‘independent diagonals approximation’, data not shown]. In the Smith–Waterman alignment matrix () for computing , let the maximum local alignment score on a diagonal be . The independent diagonals approximation for gapless local alignment () assumes that the diagonals within the alignment matrix are probabilistically independent. Because (the maximum being over all diagonals ), it follows that (the product being over all diagonals ). Thus, the distribution of can be computed from the distributions of .
To compute the exact distribution of , note that in ‘gapless’ local alignment, the Smith–Waterman algorithm applies a dynamic programming recursion along each diagonal in its alignment matrix. On each step of the recursion, the input for updating the dynamic programming state is an amino acid. In the independent letters model, the amino acid constitutes a random input independent of previous states and inputs. A Markov computation therefore yields the exact distribution of .
Finally, the distribution of can be determined through the usual convolution algorithm (which is itself a Markov computation, with inputs ). The distribution of serves as an approximation to the distribution of the GLOBAL score . The GLOBAL -value calculation is complete.
Unless otherwise indicated, all implementations used default parameters. The accuracy of an MSA is known to influence retrieval performance (). All implementations therefore used the same public resource and the same MSAs. If a tool required PSSM input, publicly available tools at NCBI derived the required PSSMs from the MSAs.
The implementation of GLOBAL had no free parameters except the ones inherent in the PSSMs derived from MSAs. (The Supplementary Data support the claim by showing that GLOBAL is a special HMM.) The curated blocks publicly available at: provided the primary input for GLOBAL's blocks (as described in the ‘Discussion’ section).
The implementation of HMMer used version hmmer-2.3.2.bin.intel-linux. First, the search database was constructed from the MSAs in DB_331_CD (described above). Second, hmmbuild with the default option hmmls built models for the semi-global alignments; and hmmbuild with the option – built models for the local alignments. We call these variants ‘HMMer_semi-global’ and ‘HMMer_local’. Finally, hmmcalibrate fitted the Gumbel parameters for the HMM -value, and hmmpfam searched the CDD.
The implementation of RPS-BLAST used the NCBI standalone version at: .
Our test set emphasized subtle protein relationships with pairwise sequence identity mostly <25% (Supplementary Data). The LROC curve and LROC score (see ‘Materials and Methods’ section) measured the CDD retrieval performance of GLOBAL, HMMer_semi-global, HMMer_local and RPS-BLAST on the test set.
As retrieval performance improves, the LROC curve for a tool moves higher on the LROC plot. The LROC plot in was truncated at a 5% false-positive rate, because users rarely examine a CD retrieval list farther. shows that the semi-global methods (GLOBAL and HMMer_semi-global) had comparable retrieval efficacy and dominated the local methods (HMMer_local and RPS-BLAST) throughout CDD retrieval. The domination increased toward the ‘twilight zone’ at the right of the LROC plot, where relationships are most difficult to detect.
Similarly, as retrieval performance improves, the LROC score increases. The LROC scores displayed in (corresponding to ∼1, 5, 10 and 20 unrelated CDs per protein query) were chosen because most users examine a CD retrieval list at least up to the first irrelevant CD, but usually not farther than the 20th. confirms that the semi-global methods dominated the local methods. As assessed by the bootstrap (), the small numerical difference in LROC scores between GLOBAL and HMMer_semi-global was statistically insignificant in early retrieval, up to about the 10 unrelated CDs per protein query, after which GLOBAL showed a statistically significant improvement over all other tools. Although late retrieval is not as important as early retrieval, it does enter some applications [e.g. SAM T99 uses WU-BLAST retrieval up to -values of 300 (,)].
GLOBAL and HMMer_semi-global had similar computation times, seconds to minutes, for: (i) the -value pre-computation for 1 CD; and (ii) the CD retrieval for 10 protein queries. To emphasize the computational differences, GLOBAL's pre-computation involves dynamic programming, not simulation, whereas HMMer_semi-global's pre-computation involves simulation.
Theoretically, the -value estimates the number of errors (false positives) preceding a CD in a CD retrieval list. To evaluate the accuracy of an -value, the ‘EPQ plot’ plots the empirical average number of retrieval errors per query against the -value (). If an -value were to estimate the errors perfectly for each query, the EPQ plot would place the corresponding point on its diagonal line ( = ). The borderline for statistical significance is usually placed somewhere between about 0.01 and 1 error per query (e.g. () and ). As an initial assessment of -value accuracy, the EPQ plots in for GLOBAL, HMMer and RPS-BLAST were unremarkable, except to indicate that on average, our objective (structural) standard of truth misclassified about 2% of the positives per query as negatives. Although a misclassification rate of 0.02 has little impact on the LROC assessment of relative retrieval performance, it could have a large impact on an EPQ assessment of -value accuracy, particularly for -values less than about 0.02.
Rather than make logically circular, subjective judgments to ‘correct’ the EPQ plot, we used simulations to compare the -value accuracies of GLOBAL and HMMer_semi-global, the two tools with the best CDD retrieval performance.
HMMer derives its -values from fitting the two parameters of a Gumbel distribution. By default, it fits its Gumbel parameters from 5000 random sequences with lengths normally distributed about a mean of 350. To make the conditions of comparison favorable to HMMer_semi-global, the Gumbel parameters in HMMer were fit from 100 000 random sequences instead of the default 5000. Similarly, because HMMer's -value approximation should be most accurate for sequences of length 350 (the mean length used in its Gumbel fit), to favor HMMer further, the -values in the two semi-global tools were tested by aligning CDs against 1 000 000 random sequences of length 350 from a standard (Robinson and Robinson) background amino acid distribution ().
For a random protein sequence, the -value = is the expected number of CDs with a -value not exceeding (see ‘Materials andMethods’ section). For a fixed number of CDs in the CDD, the - and -values therefore have the same relative error. Accordingly, displays -value accuracies for GLOBAL; , for HMMer_semi-global. Although the three CDs shown represent the full range of GLOBAL accuracies, from best to worst, they are otherwise arbitrary. Compared with all other CDs in the CDD, the calculated GLOBAL -values for cd00288 were the least accurate, probably because cd00288 has many short blocks. The GLOBAL -values were generally quite accurate, however, whereas the Gumbel -value approximations in HMMer_semi-global were not. The Gumbel -value approximations have increasing errors as the -value decreases, likely reflecting the notorious difficulties in fitting the Gumbel scale parameter λ ().
The GLOBAL -value is typically an underestimate for sequences longer than about 400, an overestimate for sequences shorter. In our experiments, the -value accuracy usually (but not always) improved with decreasing -values and the number of blocks in a CD. The percentages of the 331 CDs in our test set where the calculated GLOBAL -value = 0.1 differed from simulation estimates by less than a factor of 2.0 were as follows: 60% at sequence length 200; 93% at sequence length 400; and 62% at sequence length 1000. The ‘twilight’ -value = 0.1 corresponds to the -value = 0.1/331 ∼ 3 × 10 in and .
Intuitively, classification methods should be ‘global’, in the sense that they should exploit all available information. Correspondingly, in the identification of ‘complete’ protein domains within protein sequences, shows that the semi-global tools (GLOBAL and HMMer_semi-global) dominate the local alignment tools (HMMer_local and RPS-BLAST). HMMer_local is competitive among local alignment tools based on HMMs (,), so potentially, semi-global alignment could dominate local alignment in applications requiring the identification of complete protein domains within protein sequences.
Domain classification methods should be global in the sense above, but they must also maintain flexibility, to handle common evolutionary events like deletions at the ends of a conserved secondary structure. Tools with a strong tendency to align complete blocks (or motifs) might lack such flexibility. In fact, we tested the default mode of several such tools: MAST (,), various implementations of META-MEME () and the global implementation of SALTO (). In our hands, according to our benchmarking test set and throughout all of CDD retrieval, none of these tools performed as well as any tool appearing in (data not shown).
In general, the quality of MSAs noticeably influences retrieval performance (). In the CDD, curators define ‘curated blocks’ very restrictively, e.g. curated blocks do not contain gap characters. Moreover, within the ‘curated spacer’ between a consecutive pairs of curated blocks, each MSA sequence is padded in its middle with inserted gap characters, up to the length required. Because CDD curators do not actively align sequence in the curated spacers, the curated spacer alignments are adventitious.
To test whether the curated spacers contain information relevant to domain identification, the ‘blocks’ used throughout this article were defined by augmenting the curated blocks with all contiguous MSA columns having fewer than 50% gap characters. The adventitious alignments within curated spacers were left unchanged. Thus, curated blocks corresponded to subsets of our blocks, and curated spacers corresponded to supersets of our spacers.
After modifying GLOBAL to use the (shorter) curated blocks, retrieval performance degraded (data not shown), so the curated spacers do indeed contain relevant information (). Our findings therefore suggest that careful curation of alignments between the curated CDD blocks might noticeably improve the identification and alignment of complete domains within query proteins.
Like HMMer_semi-global, GLOBAL is an HMM (see Supplementary Data). In the GLOBAL HMM, transition probabilities are determined by the block sizes in a CD and are effectively fixed (indeed, they take rather counter-intuitive values). Thus, the GLOBAL HMM fits only emission probabilities, whereas HMMer fits both transition and emission probabilities. The retrieval performances for GLOBAL and HMMer_semi-global were almost indistinguishable, however (). Within limits, therefore, the retrieval performance of an HMM probably depends more on its emission probabilities than on its transition probabilities.
HMMs sometimes estimate their -values rather poorly () (). The manual for HMMer, e.g. warns that the HMMer_semi-global Gumbel -value approximation is sometimes very inaccurate (). Some authors even question the theoretical foundations for fitting Gumbel distributions in an HMM (). In contrast, GLOBAL calculates its -values by dynamic programming, not by simulating or fitting distributional parameters. Consequently, it estimates -values for its null model of independent, identically distributed amino acids quite accurately.
In addition, GLOBAL uses gapless local alignment to align each CD block to a protein sequence (). It is therefore amenable to the same heuristics accelerating local alignment computations in RPS-BLAST.
To summarize, GLOBAL is a new semi-global alignment tool for finding complete domains within protein sequences. It has competitive retrieval performance, an accurate -value and the possibility of heuristic acceleration, all of which enhance its potential as a high-throughput tool. The implementation of GLOBAL as the default tool at NCBI for searching the CDD is underway; the current version of GLOBAL is available at: .
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Many restriction endonucleases (REases) bind to their target sequences and hydrolyze both strands of the duplex DNA simultaneously, cleaving the molecule in a single catalytic cycle. In general, hydrolysis reactions for the two strands proceed simultaneously, thus requiring the presence of two catalytic sites. Evolutionarily, the simplest way for a protein with one catalytic site to acquire the second one is to dimerize with one another—as a homodimer—and this is the strategy that many REases appear to have adopted. As a result, and in spite of their considerable diversity, REases usually accomplish double-stranded (ds) cleavage of DNA using two identical catalytic sites. Not all do so, however. Some enzymes use a single catalytic site to cleave the two strands of DNA sequentially in a single binding event (e.g. BfiI) or in consecutive binding events (e.g. MspI and HinP1I) (). Here we describe a family of unusual REases from the bacterial genus that accomplish ds cleavage using two different catalytic sites. Some members of this family are heterodimers with the two catalytic sites located in separate polypeptide chains. Other members are monomeric with the two catalytic sites located in a single polypeptide chain.
One common mode of homodimerization is exhibited by the familiar ‘Type IIP’ REases such as EcoRI (GAATTC −5/−1) and BglI (GCCNNNNNGGC −4/−7) that recognize palindromic sequences. These enzymes bind to ds DNA as homodimers and recognize their target sequences by a concerted process in which both subunits participate in equal and opposite measure: one subunit recognizes certain features of the sequence in one orientation, while the other subunit recognizes the same features but in the other orientation. In consequence, the overall recognition sequence is symmetric, and so are the cleavage positions within the sequence. Heterodimeric REases have been discovered in the forms of Bpu10I (CCTNAGC −5/−2), BbvCI (CCTCAGC −5/−2) and BspD6I (GAGTC 4/?), but they are relatively rare. Bpu10I and BbvCI each comprise two similar but non-identical subunits. Both subunits contain a single catalytic site, and both are required for ds cleavage activity. Mutating either catalytic site produces ‘half-active’ heterodimers that cleave one strand of the recognition sequence only; that is, ‘nick’ the DNA instead of cleaving it (Janulaitis,A. ., 2005, US patent 6867028) (). Another example of a heterodimeric REase is BseYI (CCCAGC −5/−1) consisting of two non-identical subunits that can be expressed separately in . BseYI restriction activity can be reconstituted by mixing the two subunits (Nkenfou,C., Morgan,R. and S.-Y.X., unpublished data).
An alternative mode of homodimerization is exemplified by FokI (GGATG 9/13), the archetypal Type IIS REase. FokI binds to DNA as a monomer but cleaves cooperatively with another monomer () because its recognition sequence is asymmetric, it binds in one orientation only and cleaves asymmetrically, to one particular side of the recognition sequence. FokI comprises an N-terminal DNA-binding domain and a discrete C-terminal catalytic domain with a single catalytic site. Cleavage is accomplished by transient homodimerization between the catalytic domains of neighboring DNA-bound monomers (). At low enzyme concentrations, solitary DNA-bound FokI molecules neither nick nor cleave the DNA. The truncated catalytic domain of FokI can dimerize transiently with such solitary molecules to stimulate ds cleavage, although this domain on its own is inactive (). Cooperativity implies that the extent of cleavage increases sigmoidally with increasing enzyme concentration rather than monotonically, and that DNA molecules with a single recognition sequence are cleaved poorly compared with those containing multiple recognition sequences. These two mechanistic characteristics have been shown experimentally for FokI, suggesting that transient dimerization is a strategy commonly adapted by these REases. Steric considerations suggest that this form of dimerization is feasible only for enzymes in which the amino acid residues responsible for catalysis and DNA recognition are distinct and well separated; for enzymes that cleave well outside of their recognition sequences.
A number of REases cleave within asymmetric sequences or very close to the recognition sequences. These enzymes do not appear to engage in transient homodimerization. Instead these enzymes use two different catalytic sites for cleavage (). We describe here four enzymes of this kind from thermophilic isolates of : BsrDI, BtsI, BsmI and BsrI. The four appear to be distantly related: they exhibit limited amino acid sequence similarities; they recognize similar asymmetric five- and six-base sequences in ds DNA; and they all cleave very close to their recognition sequences and produce fragments with 2-base, 3′-overhangs.
Of interest is our finding that these enzymes possess two oligomeric organizations: BsrDI and BtsI endonucleases are heterodimers; BsmI, Mva1269I (a BsmI isoschizomer) and BsrI are monomers that might have evolved from heterodimeric ancestors by gene fusion. Of interest, too, is our finding that BsrDI and BtsI act as sequence-specific DNA nicking enzymes, or DNA cleaving enzymes, depending on their subunit compositions. Each enzyme comprises one large subunit and one small subunit. The large subunits, reminiscent of Type IIS enzymes, contain one catalytic site and all the elements for DNA recognition. On their own, they bind to DNA and specifically nick the bottom strand of their recognition sequence. We found that the small subunit of BsrDI and BtsI contain a catalytic site for top-strand cleavage. The small subunit is inactive on its own but in combination with the large subunit the heterodimer becomes a REase that cleaves both strands. We refer to the large subunits of BsrDI (Nb.BsrDI) and BtsI (Nb.BtsI)—natural, as opposed to mutationally engineered, DNA nicking enzymes—as ‘hemidimers’. By combining the catalytic-deficient large subunit of BsrDI and BtsI with its small subunit partner, we have successfully created top-strand-specific nicking enzymes Nt.BsrDI and Nt.BtsI.
The BsrDI- and BtsI-producing strains were from NEB's strain collection. strains ER2502, ER2566, ER2683 (Km, ) and ER2848 (Tc, ) were provided by M. Sibley and E. Raleigh (NEB). Plasmids pRRS (Ap, a pUC19 derivative), pLG339 [pSC101 origin, Km, Tc ()], pAGR3 [Ap, a pBR322 derivative with P promoter, ()] and pAII17 (Ap, a pBR322 derivative with T7 promoter, Bill Jack, NEB) were used as the expression vectors.
D70 genomic DNA was digested with REases and ligated into pRRS with compatible ends. Plasmid library was challenged with BsrDI and the digested DNA was used to transform ER2683 competent cells. Plasmids from cultures of individual transformants were screened for resistance to BsrDI digestion. The inserts in the methylase-positive clone and subclones were sequenced using a kit from ABI. A DNA fragment carrying genes were amplified in PCR, digested with BamHI and SphI and ligated to pLG339 with compatible ends.
Inverse PCR was carried out to amplify the DNA adjacent to the M genes. Inverse PCR products were purified and directly sequenced by primer walking. The pre-modified host ER2848 (pLG-) was used for endonuclease expression. Cells were cultured to late-log phase at 37°C and IPTG induction (0.5 mM) was carried out for 3 h. Cell extracts were prepared by sonication. In order to improve the expression level, and genes were amplified in PCR and independently cloned into the expression vector pAII17. Recombinant BsrDI activity was reconstituted by mixing cell extracts containing BsrDI A and B subunits or by mixing purified BsrDI A and B subunits. IPTG-induced cell extracts were used to digest λ or pUC19 DNA. Nicking activity was assayed on a single-site substrate pGPS2.1 or pUC-Km (a pUC19 derivative with a Km gene).
Cell extract containing BsrDI A subunit was heated at 65°C for 30 min and the denatured proteins were removed by centrifugation. The BsrDI A subunit in the supernatant was further purified by chromatography through Heparin Sepharose FF, DEAE Sepharose and Q Sepharose columns (GE HealthCare). Cell extracts containing BsrDI B subunit were prepared from ∼20 g of IPTG-induced cells. BsrDI B subunit (Nb.BsrDI) was purified by chromatography through a Heparin Hyper D column (BioSepra), a Q Sepharose column and a Heparin column (Tosoh Bioscience).
ApoI, NlaIII and Sau3AI partially digested genomic fragments from were ligated to EcoRI, SphI and BamHI digested, CIP-treated pUC19, respectively. The primary DNA libraries were challenged by BtsI digestion. After retransformation, plasmids were isolated and screened for resistance to BtsI digestion. The inserts in the resistant plasmids (M clones) were sequenced with pUC universal primers and custom-made primers. ORFs adjacent to genes were obtained by inverse PCR and direct sequencing of the PCR products. The and genes were amplified in PCR and cloned in two steps into pACYC184 to generate pre-modified host ER2683 (pACYC-). The and genes were amplified in PCR, digested with appropriate REases and ligated to pUC19 separately and transferred into pre-modified host ER2683 (pACYC-).
BtsI A subunit was purified by heat denaturation of proteins and chromatography through a hydroxyapatite column. Cell extracts containing BtsI B subunit were heated at 65°C for 30 min and then centrifuged at 26 000 for 20 min at 4°C to remove most of proteins. BtsI B subunit was purified by Heparin and hydroxyapatite column chromatography.
Cell extracts containing BtsI A (small) and B (large) subunits or purified A and B subunits were mixed and incubated with 1 μg of ϕX174 DNA (RF form with a single BtsI site) at 55°C for 1 h. The cleavage products were analyzed by agarose gel electrophoresis and compared with the cleavage pattern of the native BtsI REase. The nicking activity was also assayed on ϕX174 ds DNA. To confirm the nicking strand specificity of the BtsI B subunit, the nicked circular product was subjected to run-off sequencing.
The Asp, Glu and Lys catalytic residues in the active site for top-strand nicking (PDXEXK) and the Asp, Glu, Gln and Arg residues in the active site for bottom-strand nicking (PDXEXQR) were mutated to Ala residue by site-directed mutagenesis with Phusion DNA polymerase and inverse PCR. Following DpnI digestion, plasmids coding for inactive proteins (inverse PCR products) were transformed into methylase-protected host for expression. The entire alleles were sequenced to confirm the desired mutation. Cell extracts were prepared from IPTG-induced cultures. Catalytically inactive A subunit was combined with wild-type (wt) B subunit or catalytic-deficient B subunit was mixed with wt A subunit. Reconstituted restriction or nicking activity was examined using the appropriate DNA substrates.
Several BsrDI-resistant clones were isolated following the methylase selection procedure (). An M clone with a 4-kb insert was sequenced and three potential ORFs were found (A). The aa sequence of one ORF was found to resemble a putative DNA transposase. The other two ORFs contain conserved motifs of adenine N6 and cytosine N4 methyltransferases and were named M1.BsrDI and M2.BsrDI, respectively. The two M genes were amplified in PCR and cloned into the plasmid vector pLG339 to produce pre-modified host ER2566 (pLG-). After several rounds of inverse PCR and sequencing, three additional ORFs (ORF1, ORF2 and ORF3) were found (A).
We fortuitously discovered that the gpORF1 is a strand-specific DNA nicking endonuclease. When IPTG-induced cell extracts containing gpORF1 was used to digest λ DNA at 65°C, no apparent ds DNA cleavage activity was detected. However, when the cell extracts of gpORF1 were incubated with supercoiled pUC19, the DNA was converted to nicked-circular form. The nicking is site specific and not due to contaminating nucleases since a pBR322 derivative with BsrDI site deletion was not nicked (data not shown). ORF2 was amplified in PCR and cloned into pAII17. IPTG-induced cell extracts lack site-specific restriction or nicking activity (data not shown). The negative result was not due to the absence of the expressed protein since a ∼25-kDa protein band was clearly detected by SDS–PAGE. In addition, no mutations had been introduced into the cloned insert.
We confirmed that both the BsrDI large and small subunits are required for restriction activity. The BsrDI small subunit (A subunit) was partially purified and its purity was estimated to be 90% by SDS–PAGE analysis (A, lane 3). BsrDI REase activity was reconstituted by mixing the cell extracts containing gpORF1 and gpORF2 (data not shown) or by mixing purified large and small subunits (B, lanes 4–8). When 0.8 nM BsrDI A subunit was incubated with 8 nM pUC19 (lane 3) there was no change in DNA mobility. However, when 0.1 nM of the large subunit (B subunit) was added to the same reaction containing A subunit, complete cleavage of pUC19 was achieved (lane 4). The data from B demonstrate that restriction activity requires the presence of both subunits. Thus, BsrDI is a heterodimeric REase encoded by (ORF1) and (ORF2).
We further demonstrated that BsrDI large subunit (B subunit) is a bottom-strand specific DNA nicking endonuclease. The BsrDI large subunit (B subunit) was purified by chromatography such that >95% homogeneity was achieved (A, lane 2). Most striking is the nicking activity displayed by the large subunit (C, lanes 3–9). Complete conversion of 8 nM pUC19 to open circular form can be achieved with 4 nM of the large subunit in 30 min at 65°C. To determine the strand and sequence specificity, the plasmid pGPS2.1 with a single BsrDI site was treated with the large subunit. The nicked DNA product was purified and subjected to run-off sequencing. The sequencing peaks drop off at 5′ GCAATG(a) 3′, indicating a nick at 5′↓CATTGC 3′. The sequence was continuous on the opposite strand, indicating that no nicks have been introduced to the top-strand template (data not shown). The gene product of ORF1 was thus named Nb.BsrDI (Nb for bottom-strand nicking) with the specificity of 5′ ↓CATTGC 3′ (↓ indicating the nicking site). These results indicate that in addition to residues responsible for binding specifically to the recognition sequence, all catalytic residues necessary for bottom-strand cleavage (C) are located in the BsrDI large subunit (B subunit). In contrast, top-strand nicking and thus ds cleavage activity requires the assembly of A and B subunits. Amino acid sequence alignment of BsrDI large subunit with the N-terminus of BsmI/Mva1269I, and BsrI revealed a non-canonical catalytic site C with conserved residues P/A/F D-X-E-X–QR, whereas the BsrDI small subunit contains a canonical Type II catalytic site C with conserved amino acid residues PD-X-EXK. The BsmI/Mva1269I and BsrI REases are monomeric proteins with large molecular weights (78.1, 78.4 and 69.4 kDa, respectively) () (S.-Y.X., Z.Z. and G.G.W., unpublished data).
The BtsI M genes were successfully cloned by the methylase selection procedure. The inserts of the M clones were sequenced and the assembled sequence generated a 4986-bp sequence with four potential ORFs. The predicted aa sequences of two ORFs contain conserved motifs of amino-methyltransferases and were named M1.BtsI and M2.BtsI. The two M genes were amplified separately in PCR and cloned in two steps into the plasmid vector pACYC184 to generate a pre-modified expression host ER2683 (pACYC-). We surmised that the third ORF (ORF S, 495 bp) flanked by the two M genes might encode a regulatory protein or one subunit of the BtsI REase. A partial ORF of 657 bp was also found adjacent to the gene. Following inverse PCR and direct sequencing, a complete ORF of 987 bp was derived (ORF L, B).
Similar to BsrDI, we found that the BtsI large subunit (B subunit, gpORF L) is a strand-specific DNA nicking endonuclease. Both the large and small subunits are required for BtsI restriction activity. The four genes in the BtsI R-M system are organized in the order of and (B). ORF S and ORF L are oriented in the opposite direction and are separated by the gene. The ORF S and ORF L were independently expressed in . No specific endonuclease activity was detected on λ or supercoiled ϕX174 DNA with cell extracts of gpORF S. However, cell extracts containing the large subunit prepared from ER2683 (pACYC-, pUC19-ORF L) displayed DNA nicking activity when ϕX174 RF was used as the substrate (data not shown). A shows the purified BtsI A/B subunits. The A subunit displayed aberrant migration on the protein gel. The apparent molecular mass is 27 kDa, while the predicted size is 18.6 kDa. Purified BtsI A subunit forms oligomers with three distinct species in native PAG electrophoresis (data not shown). The significance of this self-assembly is not known. The apparent molecular mass of the B subunit on the protein gel was as predicted (38 kDa). DNA nicking activity was detected when purified BtsI B subunit was incubated with ϕX174 ds DNA (B, lanes 4–12). Ds cleavage activity of BtsI was reconstituted by mixing the two purified subunits (C, lanes 3–7). The A subunit alone did not show nicking or ds DNA cleavage activity (C, lane 8). To determine the nicking strand and sequence specificity of the BtsI large subunit (B subunit), the nicked-circular product was purified and sequenced. The run-off sequencing results show sequencing peaks drop off after the sequence 5′ GCAGTG(a) 3′, indicating that BtsI large subunit (B subunit) is a bottom-strand NEase (Nb.BtsI) that specifically nicks the 5′ ↓CACTGC 3′ site. DNA sequence from the top-strand template is continuous, indicating no nicks are introduced on the top strand (data not shown). We also found that Nb.BtsI nicks a miscognate site (star site) 5′ ↓CACTG 3′ that is one base (underlined ) different from the cognate site when 10-fold over-digestion was carried out. BtsI endonuclease also displays strong star activity in over-digestion. However, the entire spectrum of BtsI star sites has not been fully characterized (data not shown).
BsrDI GCAATG (2/0) and BtsI GCAGTG (2/0) share limited amino acid sequence similarity with several other Type IIS REases that cleave DNA in the same general manner, that is, close to their recognition sequences with a 2-base 3′-cohesive end. These related enzymes include BsmI/Mva1269I (GAATGC 1/-1), and BsrI (ACTGG 1/-1). Aligning the recognition sequences of these enzymes reveals that each of them cleaves DNA two bases downstream from an invariant ANTG tetranucleotide in the top strand of the recognition sequence and immediately before the complementary CANT tetranucleotide in the bottom strand (). Interestingly, while BsrDI and BtsI are heterodimeric, the other three enzymes appear to be monomeric. These latter might have arisen by gene fusion between ancestral subunits that remain separated in BsrDI and BtsI.
Amino acid sequence analysis and experimental observations suggest that all of these enzymes harbor two catalytic sites. The first, N-terminal, catalytic site (C) in each enzyme is somewhat non-canonical in sequence: PD-X-E-X-QR, whereas the second catalytic site (C) in each is canonical: PD-X-EXK () (). One site (C, catalytic site for bottom strand) is located in the large subunit of BsrDI and BtsI, or near the N-terminus of the monomeric proteins (BsmI/Mva1269I, BsrI) and catalyzes the cleavage of the bottom strand of the recognition sequence. The other site (C, catalytic site for top strand), is located near the C-terminus in the monomeric enzymes, and within the small subunit in the heterodimeric ones, and probably catalyzes hydrolysis of the top-strand (see mutagenesis results section). The region between the catalytic sites is variable and perhaps comprises the DNA target-recognition domain (TRD) of each protein. To a first approximation, the domain/subunit architecture of the heterodimers BsrDI and BtsI can be described as C∼TRD:C, where ‘∼’ represents covalent polypeptide linkage and ‘:’ represents non-covalent association. Likewise, the architecture of monomers BsmI/Mva1269I, and BsrI can be described as C∼TRD∼C. As described above, the isolated large subunits of BsrDI and BtsI (composition C∼TRD), in the absence of their C-containing small subunit, nick only the bottom strand of the target sequence.
The Asp, Glu, Gln and Arg in C the putative catalytic site for bottom-strand nicking in BsrDI B subunit (Nb.BsrDI) were mutated to Ala. The mutant cell extracts were assayed for DNA nicking or ds cleavage activity. A shows that [B]-D83A, [B]-E99A and the double mutant [B]-Q112A/R113A lack DNA nicking activity as anticipated. Similarly, the Asp, Glu and Lys in C, the putative catalytic site for top-strand nicking in the A subunit were changed to Ala by site-directed mutagenesis. BsrDI A subunit mutants [A]-D67A, [A]-E82A/K84A, [A]-K86A, as well as the wt A subunit do not show DNA nicking or restriction activity. Surprisingly, when the catalytic-deficient B subunit mutants were combined with wt A subunit, DNA nicking activity was reconstituted. To determine the nicking strand specificity of the newly restored nicking activity, the nicked circular DNA was gel-purified and subjected to run-off sequencing. B shows that [wt]-A/[B]-D83A or [wt-A]/[B]-E99A are top-strand nicking enzymes with substrate specificity of GCAATGNN↓. [wt-A]/[B]-Q112A/R113A gave rise to the same result (data not shown). Combining BsrDI wt B subunit with A subunit mutants [A]-D67A or [A]-E82A/K84A resulted in bottom-strand nicking activity only, further confirming that [A]-Asp67, [A]-Glu82 or [A]-Lys84 are involved in top-strand catalysis. Mixing BsrDI wt B subunit with [A]-K86A produced BsrDI restriction activity, indicating [A]-Lys86 is not a catalytic residue although this Lys residue is conserved in four of the five endonucleases (see amino acid alignment in ).
Alanine scanning was also applied to BtsI A/B subunits. The Asp, Glu, Gln and Arg in C, the putative catalytic site for bottom-strand nicking in BtsI B subunit (Nb.BtsI) were changed to Ala by site-directed mutagenesis. The mutant cell extracts were assayed for DNA nicking or restriction activity. A shows that [B]-D58A, [B]-E72A and the double mutant [B]-Q85A/R86A lost DNA nicking activity as expected. Similarly, the Asp, Glu and Lys in C, the putative catalytic site for top-strand nicking in the A subunit were substituted by Ala. BtsI A subunit mutants [A]-D65A, [A]-E78A/K80A, [A]-K82A as well as the wt A subunit do not show nicking or restriction activity. However, when the catalytic-deficient B subunit mutants were combined with wt A subunit, DNA nicking activity was restored. Run-off sequencing of the gel-purified nicked-circular DNA products showed that [wt]-A/[B]-D58A or [wt-A]/[B]-E72A are top-strand nicking enzymes with substrate specificity of GCAGTGNN↓ (B). [wt-A]/[B]-Q85A/R86A is also a top-strand nicking enzyme (data not shown). Combining BtsI wt B subunit with A subunit mutants [A]-D65A or [A]-E78A/K80A produced only bottom-strand nicking activity, further confirming that [A]-Asp65, [A]-Glu78 or [A]-Lys80 are involved in top-strand hydrolysis. Similar to [A]-Lys86 of BsrDI, [A]-Lys82 of BtsI is not a critical residue for top-strand cleavage because mixing BtsI wt B subunit with [A]-K82A generated BtsI restriction activity (A).
We describe here cloning and expression of BsrDI and BtsI R-M systems, and the discovery of two natural nicking enzymes Nb.BsrDI and Nb.BtsI. In addition, top-strand nicking variants, Nt.BsrDI and Nt.BtsI, were created by mixing catalytic-deficient B subunits with the respective wt A subunit. Two independent catalytic sites were implicated in BsrDI/BtsI large and small subunits for top-strand and bottom-strand cleavage.
Only a few naturally occurring DNA nicking endonucleases (NEases) have been described from bacterial and viral sources. Two strand-specific and sequence-specific DNA NEases were found in the lysates of Chlorella viruses NYs-1 and NY2A (,). These are Nt.CviPII (previously named CviNYSI nickase, ↓CCD, D = A, G or T) and Nt.CviQII (previously named CviNY2A nickase, R↓AG, R = A or G). The CviPII and CviQII N-M systems have been cloned and expressed in heterologous hosts (,). Nt.BstSEI and isoschizomer Nt.BstNBI (GAGTCN↓) were found in two strains of (,). Nt.BspD6I, an isoschizomer of Nt.BstNBI with identical amino acid sequence has also been cloned (). Nt.BstNBI was thought to be a nicking enzyme that had lost the ability to dimerize because the enzyme purified from the native strain was a NEase. Recently, however, the gene encoding its small subunit partner has been identified adjacent to the gene coding for the nicking subunit (Heiter,D. and G.G.W., unpublished data). Presumably, the small subunit of the BstNBI endonuclease was dissociated from the large subunit and was separated from the large subunit during purification, resulting in a natural top-strand nicking enzyme. Similarly, a gene encoding a small subunit of 186 aa residues had been found downstream of the Nt.BspD6I gene. BspD6I restriction activity was reconstituted by mixing Nt.BspD6I with the purified small subunit (). Based on the mutagenesis results of BsrDI and BtsI, it should be possible to engineer bottom-strand nicking variants from BstNBI/BstSEI/BspD6I REases by mixing the catalytic-deficient variants of Nt.BstNBI/Nt.BstSEI/Nt.BspD6I with their wt small subunit, assuming that two independent catalytic sites reside in two subunits.
Strand-specific NEases have been used in DNA strand displacement amplification (SDA), EXPAR DNA amplification (), in DNA fragment assembly and cloning, and in preparation of nicked-duplex or gapped DNA for studying DNA repair and DNA base stacking (,). Recently, a nicking endonuclease-mediated DNA amplification (NEMDA) using Nt.CviPII and Bst DNA polymerase in the absence of input primers has also been described (). Nt.CviPII-digested genomic DNA into small partial duplex DNA that serves as the templates and primers for primer extension. Recently, nicking enzymes have been used to nick and label DNA for single-molecule barcoding system (,).
The relatively small number of NEases identified in nature has prompted efforts to engineer NEases from existing Type IIA/Type IIS/Type IIT REases by domain swapping or mutagenesis (Type IIA, REases that cleave asymmetric sites; Type IIT, heterodimeric REases). Strand-specific nicking variants have been created from AlwI, Bpu10I, BbvCI, BsaI, BsmBI, BsmAI, BsmI, BspQI, MlyI, Mva1269I and SapI (REBASE) (). For Mva1269I, mutating the critical Asp, Glu, and Lys residues of C produced enzymes that nicked the bottom strand of the recognition sequence. Mutating C of Mva1269I produced enzymes that nicked the top strand specifically, but at very low efficiency unless the bottom strand was already nicked (). The catalytic sites of Mva1269I appear to act sequentially, with C cleaving the bottom strand first, and in so-doing creating the substrate upon which C can act to cleave the top strand (). Comparable results were observed with BsmI: bottom-strand nicking derivatives (Nb.BsmI, E546V) have been generated when the Glu residue of C was substituted for Val or Ala (Z.Z., Meixsell,T. and S.-Y.X., unpublished data). Mutating the Arg residue of the PD-X-E-X-QR motif in C produces a variant R123D with low nicking activity. Although the native BsrDI and BtsI prefer to nick the bottom strand first and then break the top strand in restriction of ds DNA (data not shown), the isolation of Nt.BsrDI and Nt.BtsI implies that there is no stringent requirement for nicking the bottom strand first, i.e. top-strand hydrolysis can take place without a pre-nicked bottom strand. |
GW bodies, also called dcp bodies, or P-bodies in yeast, are recently described cytoplasmic structures involved in mRNA metabolism. They were first implicated in mRNA degradation. In eukaryotes, mRNA degradation is initiated by the removal of the polyA tail followed by either 3' to 5' degradation by the exosome or decapping of the 5' extremity and 5' to 3' degradation by Xrn1. GW bodies contain all the proteins of the 5' to 3' mRNA degradation machinery, including the decapping complex Dcp1/2, its cofactors LSm1-7 and Rck/p54 (also known as Dhh1 in yeast, Me31 in drosophila and Cgh1 in Caenorhabditis) and the exonuclease Xrn1 (). In yeast, experiments designed to slow down the final steps of mRNA degradation enhance P-bodies in size and number. This has been observed following mutations of Dcp1 or Xrn1 (). Moreover, when a poly(G) tract is introduced into a reporter mRNA to block exonucleolysis, it accumulates in P-bodies, indicating that they are active sites of mRNA degradation (). Similarly, in mammals, polyadenylated RNA are detected in GW bodies following the stable depletion of Xrn1 by RNA interference (). Furthermore, inhibiting translation with a drug which releases free mRNA, such as puromycin, leads to an increase of the GW body number (). Conversely, when mRNAs are frozen on polysomes by a translation inhibitor such as cycloheximide, GW bodies disappear (,). Taken together, these data indicate that GW bodies are formed from a pool of untranslated mRNAs available for degradation.
GW bodies also contain the post-transcriptional gene silencing machinery, including both proteins of the RNA-induced silencing complex (RISC), such as Argonaute (Ago), and short RNAs, whether short interfering RNAs (siRNAs) or micro-RNAs (miRNAs) (). One of the GW body markers, GW182, which was initially identified as a human autoantigen, turned out to be a direct Ago partner (). These observations have led to the proposal that GW bodies are the sites of RNA interference activity. This issue is controversial, as some studies report the inhibition of both mi-RNA- and si-RNA-mediated interference in the absence of GW bodies (), while others report a clear inhibition of mi-RNA-mediated interference and a slight inhibition of si-RNA-mediated interference (,); a few even indicate no inhibition of the si-RNA-mediated interference (,). The presence of RISC in the GW bodies is consistent with the need to degrade the fragments generated by the initial siRNA-mediated mRNA cleavage (). At first glance, the presence of miRNAs was more puzzling, as they guide translation inhibition and not mRNA cleavage. Recently, it has become clear that miRNA-mediated silencing can result in RNA decay, initiated by deadenylation and decapping rather than endonucleolytic cleavage (). In addition, GW bodies can also play a role in mRNA storage. In Huh7 hepatoma cells, the CAT1 mRNA is repressed by miR122 and localized in GW bodies. Following amino acid deprivation, its translation resumes and it disappears from the GW bodies ().
A more general role for GW bodies in mRNA storage is still uncertain in mammals but has been established for P-bodies in yeast. When glucose deprivation leads to translation arrest and accumulation of mRNAs in P-bodies, glucose re-addition leads to mRNAs leaving the P-bodies and resuming translation (). Therefore, P-bodies have a dual role in mRNA degradation and storage, albeit it is still unclear how these two opposite functions are coordinated. In addition, mammalian cells harbor distinct cytoplasmic structures involved in mRNA storage, the stress granules, which are induced by stress. They contain their own set of proteins, including translation initiation factors, such as eIF3, and translation repressors, such as TIA1/TIAR, as well as some of the GW body proteins, such as Xrn1 (). Stress granules are frequently in contact with GW bodies. In some cases, they recruit GW bodies and seem to fuse with them. We have proposed that this intermingling between stress granules and GW bodies could trigger the transition from mRNA storage to mRNA degradation during stress ().
The number of GW bodies per cell is variable, even within a clonal cell population. How this number could impact on their presumed function is unknown. First, it is unclear whether a GW body needs to be macroscopic in order to fulfill its function. Second, the GW body assembly pathway is still unknown. One of the reported parameters is the position of the cells in the cell cycle. Most GW bodies disappear before mitosis, reassemble in the G1 phase and enlarge in late S and G2 phases (). It has been proposed, as a consequence, that the efficiency of RNA interference may vary during the cell cycle (). In addition, several proteins of the GW bodies have been reported to be essential for the GW body assembly, because their depletion by RNA interference leads to the disappearance of GW bodies. This has been observed for Ccr4, Rck/p54, Lsm1 and eIF4ET (); GW182 (), Ge1 () and Rap55 (). One interpretation of this data is that some of these proteins have a scaffolding role in GW body assembly. Alternatively, all of them may be necessary to ensure a certain level of repressed mRNA required for the GW body assembly ().
We have previously shown that the translational regulator CPEB1 is enriched in GW bodies. This raised the possibility that it could play an active role in the degradation of the mRNA to which it is bound. Alternatively, it could be passively attracted by its target mRNA as a consequence of the translation repression (). While investigating CPEB1 function using RNA interference, we found that siRNAs targeting CPEB1 also led to GW body disassembly. Surprisingly, this property was shared with several other siRNAs and appeared to be independent of the gene targeted. The siRNA-guided RNA interference had the same efficiency in the absence of GW bodies, demonstrating that these bodies were not the sites of action of the RISC machinery. Finally, except in the case of si-p54, arsenite could counteract the effect of these siRNAs, despite the fact that this drug prevents protein synthesis, thus confirming that CPEB1 protein is not involved in GW body assembly whereas p54 is truly required.
Transfections were performed with 3 μg si-RNA or polyI/C per 35-mm diameter dish using a standard calcium phosphate procedure, as described previously (). siRNAs and polyI/C were purchased from MWG (MWG Biotech, France) and Sigma (Sigma Aldrich, France), respectively. The siRNA concentration was checked both by spectrophotometry and on gel (data not shown). In , control cells were transfected with 1.3 µg of a human CPEB1-long expression vector (IMAGE 6047179). Si-Eg5 was transfected as previously described ().
IFN dosage was performed as previously described (). Each sample was tested in duplicate and the dosage was repeated twice. Briefly, UPIL reporter cells were plated in 100 µl in a 96-well plate and cultured overnight following the addition of 25 µl of each culture medium sampled from siRNA-transfected cells. The standard curve was obtained by culturing the cells in the presence of serial dilutions of IFN-α2 at 100 000 units/ml. Luciferase activity was then measured using a Steady-Glo luciferase assay system (Promega, France).
Monoclonal anti-CPEB1 and anti-Kif15 antibodies were raised in our laboratory (,). Rabbit polyclonal anti-p54 and mouse monoclonal DM1 anti-α-tubulin were purchased from Bethyl Laboratories Inc. (Texas, USA) and Sigma Aldrich (France), respectively. The anti-GW182 human index serum was a kind gift from Theophany Eystathoy (University of Calgary, Alberta, Canada), the anti-hDcp1 rabbit antibody from Bertrand Séraphin (Centre de Génétique Moléculaire, Gif, France) and the anti-eIF3 goat antibody from John Hershey (University of California, Davis, CA). Secondary antibodies conjugated to rhodamine, FITC and horseradish peroxidase were purchased from Jackson Immunoresearch Laboratories (Immunotech, France). Monoclonal TS9 anti-CD9 and TS81 anti-CD81 antibodies () were obtained from Diaclone (Besançon, France).
Cells were scraped in PBS, resuspended in a lysis buffer (50 mM Tris-HCl pH8, 150 mM NaCl, 1 mM EDTA, 1% Nonidet P-40) supplemented with Complete Protease Inhibitor cocktail (Roche Diagnostics, France) and incubated on ice for 30 min. Soluble proteins were recovered after centrifugation at 15 000 and 4°C for 10 min and quantitated by the Bradford method (Bio-Rad, France). Proteins were separated on a 7.5% polyacrylamide SDS-PAGE gel along with the Prestained Protein Ladder 10–180 kDa (MBI Fermentas, France) and transferred to a nitrocellulose Hybond-C-Extra membrane (Amersham Pharmacia, France). All incubations were then performed at room temperature. Non-specific protein-binding sites were blocked by incubation in PBS-T (PBS, 0.1% Tween-20) containing 5% (wt:vol) non-fat dry milk for 1 h. The membrane was successively incubated with primary antibody diluted in PBS-T containing 5% non-fat dry milk for 1 h, washed in PBS-T, incubated with horseradish-peroxidase-conjugated secondary antibody for 45 min and washed in PBS-T. Immune complexes were detected using the SuperSignal® Chemiluminescent Substrate detection reagent (Perbio Science, France), and quantified by densitometry of the X-Ray film. The membrane was dehybridized for 30 min by incubation in 50 mM Tris-HCl pH 6.7, 10 mM β-mercaptoethanol, 0.5% SDS at 52°C before a second hybridization.
Cells were grown on glass coverslips and fixed in −20°C methanol for 3 min. Cells were incubated with the primary antibody for 1 h, rinsed with phosphate buffered saline (PBS), incubated with the secondary antibody for 30 min, rinsed with PBS and stained with 0.12 µg/ml DAPI for 1 min, all steps being performed at room temperature. Slides were mounted in Citifluor (Citifluor, UK) and observed on a Leica DMR microscope (Leica, Heidelberg, Germany) using a 63X1.32 oil immersion objective. Photographs were taken using a Micromax CCD camera (Princeton Instruments).
Total RNA was extracted using the SV Total RNA Isolation kit (Promega, France) and quantified by spectrophotometry. Reverse transcription reactions were performed with 1 µg RNA using random primers and Mu-MLV reverse transcriptase (Invitrogen, France). One-tenth of the reaction was used for PCR amplification with OAS2 primers [(GCTTTGATGTGCTTCCTGCCTT) and (ACCCCTTTGGCTTCAGTTTCCTT)], which amplified a 249-bp product or β-actin primers [(AGAGCTACGAGCTGCCTGAC) and (AGTACTTGCGCTCAGGAGGA)], which amplified a 300- bp product. PCR conditions were 35 cycles (94°C for 45 s, 60°C for 60 s and 72°C for 45 s) and 17 cycles (94°C for 30 s, 55°C for 30 s and 72°C for 30 s) for OAS2 and β-actin, respectively. The amplification products were separated on a 1.2% agarose gel, stained with ethidium bromide and quantified using a Fluor-S-Max imaging system (Bio-Rad, France). Each primer set targeted two separate exons in order to distinguish reverse transcribed mRNAs from residual genomic DNA.
We have previously shown the presence of the translational inhibitor CPEB1 in GW bodies. In order to gain insight into its function at that location, we used RNA interference to repress endogeneous CPEB1. HeLa cells were transfected with two siRNAs directed against human CPEB1 (si-CPEB1.1 and si-CPEB1.2), and the efficiency of the depletion was checked after 44 h by western blot using 4A2 monoclonal anti-CPEB1 antibody (A). Because this antibody detects two bands in the 50–70 kD region, cells transfected with a human CPEB1 expression vector were used as a control of CPEB1 migration. A siRNA targeting rabbit beta-globin (si-Glo.1) was used as an irrelevant siRNA. Both si-CPEB1s led to the disappearance of endogeneous CPEB1 protein. In a parallel culture dish, cells were fixed and GW bodies were immunostained with anti-Dcp1 antibodies. The number of GW bodies was reduced after si-CPEB1.1 and si-CPEB1.2 transfection, compared to mock or si-Glo.1 transfected cells (B). The same result was obtained using an antibody directed against GW182 as a second marker of the GW bodies (C). This effect persisted 72 h after transfection (D). We repeated the experiment in a second human cell line, RPE, and observed a similar effect (E). The phenomenon was quantified by measuring the distribution of GW bodies per cell (F). Compared to si-Glo.1, si-CPEB1.1 led to a similar decrease in HeLa and RPE cells, which intensified between 44 and 72 h. The effect was general rather than concerning a subpopulation of cells.
Because several targets of the CPEB1 regulatory pathway play a role in cell division during early development (Cyclin B1, Histone H3, kinesin Eg5, etc.) the depletion of CPEB1 protein could have an impact on cell proliferation. As the GW body number was reported to vary during the cell cycle, HeLa and RPE cells were transfected with si-CPEB1.1 or si-Glo.1 and stained 44 h later with propidium iodide in order to analyze the cell cycle by cytofluorimetry (G). No difference was observed, indicating that the disappearance of GW bodies was not due to the disturbance of the cell cycle by CPEB1 repression. Taken together, these observations suggested that CPEB1 is required for GW body maintenance.
HeLa cells and RPE cells were transfected with equal amounts of these siRNAs and analyzed 48 h later. The pattern was identical in the two cell lines, with the number of GW bodies being either similar to what is seen in untransfected cells, or strongly decreased, depending on the siRNA. Si-CPEB1.3 and si-CPEB1.4 had little effect whereas si-p54 led to the complete disappearance of the GW bodies, as previously described () (A). PolyI/C partially decreased the number of GW bodies (B). In addition, one of the three siRNAs targeting CD81, si-CD81.2, and one of the irrelevant siRNAs, si-Glo.2, led to the disappearance of GW bodies (C and D). Taken together, poly I/C showed an intermediate effect, while 5 of the 15 siRNAs tested led to a strong diminution of the GW bodies, with no obvious link with the function of their expected target: si-CPEB1.1, si-CPEB1.2, si-p54, si-Glo.2 and si-CD81.2.
Because some siRNAs are more potent in RNA interference than others, we investigated whether the effect on GW bodies correlated with the efficiency of silencing. In fact, si-CPEB1.1 and si-CPEB1.2 are active both in decreasing the GW body number and in depleting CPEB1 (A), whereas si-CPEB1.3 and si-CPEB1.4 are inactive on both (A and A). We therefore systematically measured the activity of the siRNAs, except for the irrelevant β-globin and LT-α siRNAs, which have no predictable mRNA targets expressed in HeLa or RPE cells.
The activity of si-Kif15 and si-p54 was measured by western blot 44 h after transfection (A). Both siRNAs had similar silencing activity, whereas only si-p54 led to the disappearance of GW bodies. The activity of the siRNA targeting the tetraspanins was measured by cytofluorimetry using antibodies directed against CD9 and CD81 (B). The four siRNAs were active on their respective targets, with 70% silencing efficiency for si-CD9 and 84–95% efficiency for CD81 siRNAs. Strikingly, the only siRNA which led to the disappearance of GW bodies, si-CD81.2, had the same silencing activity as the two other si-CD81s. This data is summarized in . There is clearly no correlation between the silencing activity of the siRNAs and their effect on the GW body number.
Having found that the effect of siRNAs on the GW bodies is related to their sequence, as not all of them decrease the GW body number, but neither to their expected target nor to their interference activity, as illustrated by the three si-CD81s, we hypothesized that they could activate a pathway distinct from the silencing machinery. Although siRNAs are short double-stranded RNA molecules 21 nt long, it has been reported that some of them do activate the interferon (IFN) pathway in certain cell lines in a sequence-dependent manner (,). We therefore considered this pathway as a candidate mechanism for the reduction of GW body number.
We first investigated whether IFN was produced in response to the transfection of the various siRNAs. RPE cells were transfected as described previously and the culture medium was sampled 48 h later. The presence of IFN was assayed using the UPIL5 reporter cell line which expresses luciferase under the control of an IFN responsive promoter (). Despite the sensitivity of this assay, no IFN could be detected, indicating that its production, if any, is less than 50 units per ml (A). In parallel, we analyzed the expression of OAS2 which has been shown to be one of the IFN-responsive genes that is the most induced by siRNAs (15-fold at 48 h in RCC1 cells, as reported in Sledz . ()). RNA was extracted and the OAS2 expression was analyzed by RT-PCR along with a β-actin control. Overall, siRNA transfection rather led to a slight decrease of OAS2 expression with respect to β-actin (B), including si-CPEB1.2 and si-Glo.2 and both of these decrease the GW body number. This suggested that the IFN pathway is not induced in these transfection conditions.
Finally we analyzed directly the effect of IFN on GW bodies. HeLa cells were cultured in the presence of 10 u/ml IFN-α for up to 48 h and GW bodies were analyzed by immunofluorescence. At this time, the IFN treatment had no effect on the GW body number (data not shown). The full activation of the IFN pathway involves the presence of double-stranded RNA which, in our siRNA transfection experiments, could be provided by the siRNA itself. We used polyI/C as a conventional co-activator of the IFN response. Cells were cultured in the presence of 10 u/ml IFN-α for 40 h and then transfected with polyI/C in the presence of IFN-α. GW bodies were analyzed by immunofluorescence 8 h later, at a time where polyI/C transfection alone does not affect the GW body number. Even under these conditions, IFN had no effect on GW bodies (data not shown). In conclusion, we find no evidence for an involvement of the IFN pathway in the disappearance of GW bodies following siRNA transfection.
We first used si-Eg5 as a reporter of siRNA activity. The Eg5 kinesin is required for mitosis and its depletion leads to a phenotype of prometaphase arrest which can be quantified, as previously described (). HeLa cells were mock transfected, transfected with si-Glo.1 or si-LT.1, which do not change the number of GW bodies, or transfected with si-Glo.2, si-CPEB1.1 or si-p54, which do decrease their number. After 40 h, cells were divided in two and transfected 8 h later with either buffer alone or si-Eg5. The cell cycle was analyzed 40 h later by cytofluorimetry after propidium iodide staining (A). A G2/M blockage of about 60% was observed in all cultures, indicating that the si-Eg5 is as efficient in the absence as in the presence of GW bodies.
In case this phenotype was dependent on an Eg5 protein threshold and therefore not sensitive to small variations in silencing efficiencies, we performed a similar experiment in RPE cells using si-Kif15 as a reporter of siRNA activity and testing its efficiency directly by western blot analysis. RPE cells were mock transfected, transfected with si-Glo.1, which does not change the number of GW bodies, or transfected with si-Glo.2, si-p54 or si-CPEB1.1, which do decrease their number. After 24 h, cells were divided in two and transfected 7 h later with buffer alone or si-Kif15. Proteins were analyzed 24 h later by western blot (B). At this time, Kif15 depletion was incomplete and similar in all samples, confirming that the silencing efficiency does not depend on the presence of GW bodies.
Similar experiments were previously reported using a siRNA targeting GW182 to decrease the GW body number, and subsequently testing interference of lamin A/C (). The authors observed a strong inhibition of lamin silencing, suggesting the requirement of either intact GW bodies or GW182 for an efficient interference (). To enable the comparison with these results, we investigated the effect of the same si-GW182 on Eg5 and Kif15 silencing. Despite a clear reduction of GW bodies (C), we did not observe any significant inhibition of further silencing (D and E). Neither did si-GW182 or si-p54 have any significant effect on the subsequent silencing of lamin A/C (F), confirming that si-RNA-driven interference does not require intact GW bodies.
The disappearance of GW bodies could be due either to the degradation of its components or to their dispersal. To address this issue, Dcp1 protein was analyzed in HeLa cells 48 h after siRNA transfection (A). The protein had the same abundance in cells transfected with si-CPEB1.1, which strongly decreases the GW body number, as in mock or si-Glo.1 transfected cells. This indicated a redistribution of GW body components toward the cytoplasm rather than a degradation.
In order to determine if the disassembly of GW bodies by siRNA is reversible, we investigated the effect of arsenite after siRNA transfection. Indeed, beside its role as an inducer of stress granules, arsenite has been shown to promote the assembly of GW bodies (). HeLa cells were mock transfected or transfected with si-CPEB1.1, si-Glo.2, si-p54 or si-GW182 for 48 h and then treated or not with arsenite for 30 min. GW bodies and stress granules were analyzed by immunofluorescence using anti-Dcp1 and anti-eIF3 antibodies, respectively. Arsenite fully induced stress granule assembly in the five samples (B–E and data not shown). The number of GW bodies strongly increased in mock transfected cells, as previously described, with 100% of the cells having numerous GW bodies (B). The same increase was observed in si-CPEB1.1, si-Glo.2 and si-GW182 transfected cells (C and D and data not shown). By contrast, only a few cells recovered GW bodies after si-p54 transfection (D). Instead, the Dcp1 protein relocalized in the newly assembled stress granules (E). The same results were obtained in RPE cells. This indicated that, in most cases, arsenite is sufficient to counteract the siRNA action on GW bodies. Interestingly, at the dose used, arsenite is a strong translation inhibitor (). Therefore, the GW body reinduction occurs in the absence of neo-synthesized protein, confirming that neither the CPEB1 nor the GW182 protein are necessary for GW body assembly. However, the p54 protein is truly required for the GW body assembly, at least in response to arsenite.
GW bodies contain proteins involved in mRNA degradation, in RNA interference as well as in mRNA storage. CPEB1 is known as a specific translation regulator acting within a large multiprotein complex bound to the 3' untranslated region of its target mRNAs. We have used RNA interference to investigate its function in human cells and found that the depletion of the CPEB1 protein is accompanied by the disappearance of GW bodies. While we had previously shown that the CPEB1 protein accumulates in GW bodies, it seemed unlikely that the depletion of a specific regulator present at low abundance could induce GW body disassembly. We therefore tested a larger set of siRNAs, including a siRNA targeting Rck/p54, which is a CPEB1 partner also found in GW bodies, siRNAs targeting mRNAs encoding proteins that should not be involved in mRNA metabolism, such as kinesins Kif15 and HSET, and tetraspanins CD9 and CD81, which are transmembrane proteins expressed at the cell surface, as well as irrelevant siRNAs designed against mRNAs that are not expressed in the cells under study, such as β-Globin and LT-α. Out of 15 siRNAs, five shared this property of making GW bodies disappear: two targeting CPEB1 and three targeting Rck/p54, CD81 and β-Globin, respectively. The effect on GW bodies was therefore not restricted to the siRNAs depleting proteins obviously involved in mRNA metabolism.
The main parameter described so far to affect GW body number in mammals is the cell cycle, and the CPEB protein is clearly involved in mitosis (). In Xenopus oocytes, CPEB-regulated genes include cyclins A and B, Mos and Aurora-A kinases, Eg5 kinesin, histone H4, all of which are involved in the resumption of meiosis (). In mammals, cyclin B1 translation is subject to a similar regulation in the MCF7 cell line (). The depletion of CPEB1 in HeLa and RPE cells could therefore lead to cyclin B1 misregulation and cell cycle abnormalities. However, in spite of a drastic depletion of the CPEB1 protein, we did not observe any change in the cell cycle, indicating either that CPEB1 is not responsible for cyclin B1 regulation in these cells or that it can be replaced by a functionally identical protein. Consequently, the disappearance of the GW bodies was not related to a particular accumulation in M or early G1 phase, which are the two cell cycle phases where GW bodies are sparse.
It has also been reported that the number of GW bodies decreases in 3T3 quiescent cells (). RPE cells, which have been immortalized by expression of the telomerase, are highly sensitive to contact inhibition and their proliferation is down-regulated at confluence, yet the number of GW bodies increased in confluent cells compared to that in exponentially growing cells (data not shown). Similarly, the number of GW bodies increased at confluence in HeLa cells (compare 44- and 72-h panels in F). Thus, confluence and quiescence enhance GW bodies in these cell lines, which is in agreement with the situation in yeast, where P-bodies increase in number and brightness with cell density (). However, we observed no marked difference of cell density 44 h after transfection of the various siRNAs (data not shown).
The fact that an siRNA designed against β-globin led to the disappearance of GW bodies raised the possibility that it was due to off-target silencing. First, as there is no sequence similarity between the various siRNAs, they would be expected to silence different off-target genes, yet lead to a common phenotype, which seems very unlikely. Second, in the case of β-Globin and CPEB1 siRNAs, GW bodies could be efficiently reinduced in 30 min by arsenite, despite the fact that arsenite fully inhibits protein synthesis. This indicated that whatever proteins might be involuntarily depleted, they were not required for the GW body assembly. This strongly suggests that the disappearance of the GW bodies in these cases is not due to the depletion of a particular protein but rather to some deregulation of GW body genesis.
One potential consequence of introducing double-stranded RNA in mammalian cells is the activation of the IFN pathway. Although short molecules are less potent than long ones, siRNAs can nevertheless induce part of the IFN response. This has been reported following transfection of synthetized siRNAs (), plasmid-encoded shRNAs () and chemically synthetized siRNAs (,). In the last case, some siRNAs were found to be more potent than others, depending on their sequence. Although these data were obtained in cells from the immune system, it could have provided an explanation for our observations. We tested this hypothesis by measuring both IFN in the culture medium of the transfected cells and OAS2 expression. Indeed OAS2 was reported to be highly induced 48 h after siRNA transfection in RCC1 cells (). Production of IFN was undetectable and OAS2 mRNA was not induced after siRNA transfection, providing no evidence of IFN pathway activation. Moreover, IFN addition to the culture medium did not affect the GW body number.
In the literature, the role of GW bodies in RNA interference is still unclear. On the one hand, in one study, GW182 depletion using RNA interference simultaneously decreased the GW body number and the silencing by miRNAs and siRNAs (). In other studies, GW182 depletion led to a clear inhibition of miRNA-mediated silencing but only to a slight inhibition of siRNA activity (,). On the other hand, two studies reported that cells lacking GW bodies could still perform efficient siRNA-mediated silencing. These cells were obtained either by stably depleting Drosha protein using shRNA () or by transiently silencing Lsm1 using siRNAs (). The latter study also reported that Rck/p54 protein is associated to RISC but is not required for siRNA-mediated cleavage activity. Our data are in full agreement with these two studies, as no inhibition of siRNA activity was observed following the disappearance of GW bodies. We therefore conclude that RNA interference can take place outside of the GW bodies.
A possible explanation for our observations could be that, in certain cases, the large amount of transfected siRNA molecules saturates RISC and progressively clears it out of the GW bodies. Our functional data do not support such a saturation effect. First, the effect on GW body number is not related to the silencing efficiency of the siRNA, which might be expected to be related to its affinity for RISC. Second, the silencing of Eg5 or Kif15 remained intact after the disappearance of GW bodies, indicating no saturation of the silencing machinery. Alternatively, siRNAs could sequester, outside of the GW bodies, RNA-binding proteins that are not involved in silencing activity but that are required for GW body assembly. The arsenite experiment argues against this possibility, as a large number of GW bodies reappear in 30 min while protein synthesis is inhibited. Finally, a third hypothesis would be that some siRNAs alter mRNA metabolism through a mechanism still to be identified, leading to a decrease of mRNAs directed to GW bodies. These would be restored when translation is inhibited by arsenite.
Whatever be the explanation, it remains clear that the disappearance of GW bodies following siRNA transfection does not obligatorily indicate that the targeted gene is important for GW body assembly. RNA interference is now a straightforward strategy to investigate protein function in mammalian cells, and it has led to the conclusion that many proteins of the GW bodies are essential for their assembly: Ccr4, Rck/p54, Lsm1, eIF4ET (), GW182 (), Ge1 () and Rap55 (). How so many proteins could all be required has not been much discussed, and our results open up the possibility that some of these proteins are not really essential for GW body assembly. Our data indicate that GW182 is not the GW body scaffold protein, as previously hypothesized (). It can be noted that Lsm1 depletion by a genetic approach in yeast leads to more numerous P-bodies rather than to their disappearance (,); it would be interesting to use arsenite treatment to confirm its importance for GW body assembly. We found that depletion of Rck/p54 protein not only leads to the disappearance of GW bodies but also prevents the induction of GW bodies by arsenite. This protein is therefore strictly required for the assembly of GW bodies, at least in response to arsenite. This is reminiscent of the situation in yeast, where Dhh1 works redundantly with Pat1, and the double Dhh1/Pat1 mutation prevents P-body formation in response to glucose deprivation (). Rck/p54 is an RNA-binding protein of the DEAD box helicase family, with multiple properties: it associates to Dcp1 and enhances decapping in yeast (), it associates to RISC and is required for miRNA-mediated silencing in mammals () and it belongs to a translation repressing complex along with CPEB in Xenopus oocytes (). It is an abundant protein which could play a scaffolding or remodeling role in GW bodies due to its helicase activity (). Alternatively, it could be part of the machinery which drives quiescent mRNA to the GW bodies for storage and/or degradation.
Although GW bodies and stress granules are related mRNP in terms of both structure and composition, Rck/p54 silencing had absolutely no effect on the stress granule assembly. Interestingly, we found that Dcp1, which is normally localized in the GW bodies independently of arsenite treatment, relocated to stress granules in the absence of Rck/p54. We have previously reported a similar behavior when stress granules are induced by CPEB1 expression (). These two data suggest that, when GW bodies are missing, stress granules can be a second default localization of Dcp1 protein. Our data, obtained in the same cellular context during arsenite treatment, establish that, despite strong structural and functional similarities between GW bodies and stress granules, their assembly follows distinct pathways. |
Nucleic acids are more plausible than proteins as the components of a self-contained replicating system in the very first stages of the emergence of life on Earth (). Watson–Crick base pairing provided a very plausible mechanism by which a polynucleotide could direct the synthesis of its complement from mononucleotides or short oligonucleotides, while no equivalent mechanism is known for the replication of a polypeptide. The discovery of ribozymes (,) and then the demonstration that ribosomal peptide synthesis is a ribozyme-catalysed reaction () strengthened the case for an early RNA World. In this context, the central problem for origin-of-life studies is to understand how this seminal world became established on the primitive Earth. Plausible scenarios for the prebiotic chemistry have been proposed (,), but the problem is far from being solved. A recurrent theme is that RNA may have emerged from an earlier world under extreme conditions of pressure and/or temperature and pH (). In all scenarios, molecules with backbones forming stable double helices held together by Watson–Crick base pairing appear as crucial intermediates or crucial building blocks. This fact led us to undertake a programme on the behaviour of such molecules under high pressure.
Some studies were recently devoted to the analysis of the effects of high hydrostatic pressures on the stability of living cells (), the quaternary structures of proteins and viruses (), protein–protein and protein–DNA interactions (), catalytic RNA () and tRNA (). Concerning nucleic acids, they were investigated through their single-strand/duplex equilibrium (), their B/Z transition () and their hydration networks () all using spectroscopic techniques in solution. Following Le Chatelier's principle, pressure tends to favour states with smaller specific volumes. In DNA, structural adaptation to high pressure results from several contributions, including the helical structure integrity (helix-to-coil transition), the stacking of bases, the Watson–Crick association and the hydration network around the duplex. Aromatic-ring stacking is favoured by pressure while Coulombic or hydrophobic interactions are disfavoured (). Hydrogen bonds and associated networks should be, as in protein and protein assemblies (), stabilized by pressure. How these various parameters influence the structure and stability of the DNA duplex is not easy to predict. Clearly, besides spectroscopic data, accurate 3D structures were required in order to describe the behaviour under high pressure at the molecular level and provide a firm starting point for simulations. With respect to pioneering work based on high-pressure beryllium cell (), our technical developments combining diamond anvil cell and synchrotron radiation of ultrashort wavelength () have considerably extended the possibilities of high-pressure macromolecular crystallography (HPMX) in a pressure range increased by one order of magnitude (from 0.2 to 2 GPa). After studies on monomeric (), then multimeric () proteins and a complex protein assembly (), this article reports the first application of HPMX to nucleic acids.
The d(GGTATACC) oligonucleotide was selected owing to the high stability of its crystalline A-DNA crystalline form, its strong base stacking (2.9 Å compared to the 3.3 Å in B-DNA duplex), and its particularly well-defined hydrogen-bonding network, mostly located in the major groove (). The last reason for this choice is an interesting feature related to the crystal packing of d(GGTATACC) crystals. Besides Bragg reflections of the A-form crystal, a pattern was observed on diffraction pictures of nucleotide crystals, similar to fibre diagrams that led Watson and Crick to their interpretation of the DNA structure (). Main features of this pattern include a characteristic diffuse crossed X-ray pattern, a set of streaks perpendicular to the c* direction of the reciprocal lattice and two strong elongated meridian reflections (). The following interpretation for the origin of this pattern was proposed (). Oligonucleotide molecules pack in infinite super-helices of duplexes down the 6-fold axis of the P6 space group (). The central channel of the super-helix can trap oriented molecules of DNA. Simulations of streaks allow excluding the A-form for occluded molecules and favour the B-form. The meridional reflections are consistent with the base-pair stacking in B-form molecules. According to these results, we have in hand a system in which both A and B forms of DNA can be simultaneously monitored against external hydrostatic pressure, although information derived on the B-form is obviously quite limited.
Crystals of the nucleotide were compressed up to 2 GPa at 295 K. We describe the high-resolution crystal structures of A-DNA at four pressures from ambient to 1.39 GPa. Meridional streaks attributed to B-DNA are observed up to at least 2 GPa. These results highlight the remarkable adaptation of the base-paired double-helix architecture base-paired architecture to high pressure.
The d(GGTATACC) sequence was synthesized by standard phosphoramidite chemistry on solid support (Applied Biosystems 391 DNA Synthetizer). Crystals were obtained following the batch method (). Ten milligrams of lyophilized octamer were dissolved in 200 μl of a 15% methyl-pentane-diol (MPD) solution, buffered by sodium cacodylate (5 × 10 M, pH 7) and containing additives: sodium azide, spermine tetrahydrochloride and MgCl. To this solution, was added each day 5 μl of the same solution but containing 50% of MPD, followed by a rapid mixing. After 5 days, crystals began to appear as elongated hexagonal rods and grew easily to a size of 0.2 × 0.2 × 1.0 mm in a week. They were stabilized by increasing the final MPD concentration to ∼35%.
Crystals were hydrostatically compressed in a diamond anvil cell. The compression chamber (diameter 400 μm, height 200 μm) was drilled in a copper or stainless steel (for pressure above 0.7 GPa) gasket (). Gasket preparation and sample loading were performed as described (). The crystallization solution with 35% MPD was used as compression medium. Pressure monitoring was performed by using the wavelength shift of laser-excited fluorescence from a small ruby sphere loaded in the compression chamber. Two different diamond anvil cells were used for data collection. The first one had a useful aperture of 62° and a standard diamond mount (). The second one was of a novel design to provide both a larger useful aperture (82°) and a pressure range up to ∼2.5 GPa ().
Diffraction data from crystals compressed in the diamond anvil cell were collected at the ESRF (Grenoble, France), on the ID27 beamline with a MAR CCD 165 mm detector. The wavelength was calibrated and set to 0.3738 Å (Iodine K absorption edge). Four data sets were collected at room temperature and ambient pressure, 0.55, 1.04 and 1.39 GPa. The crystal-to-detector distance, calibrated using the diffraction rings of a reference silicon powder, was 352.9 mm for the 0.55 GPa data set and 302.9 mm for the others. The X-ray beam was collimated to 50 × 50 μm. Exposure times were 30 or 45 s, depending on the sample, for a rotation step of 1°. As data collection was performed at room temperature, crystals were translated several times, every 10 to 35° of rotation, during exposure in order to irradiate fresh zones ().
Diffraction frames were integrated using XDS (). All data were independently put on an absolute scale using SCALA ().
The starting molecular model was either the deposited coordinates ref. 115D () or 1VJ4 () from the Protein Data Bank. In the case of 115D, the isomorphous di-bromo derivative of 1VJ4, thymine residues were reconstructed from original bromo-uracils and all water molecules were removed. A first round of rigid body refinements was done with AMoRe (), then refinements proceeded at the maximum resolution with individual isotropic thermal factors and bond distance and angle restraints, as used in SHELXL (). Water molecules were localized during the course of refinement cycles by analysing density peaks in − Fourier difference maps. They were accepted when they met standard criterions like correct bond distances and angles toward polar atoms of the model, and B thermal factors below a given threshold arbitrarily fixed to 65 Å. All analyses of the geometric parameters were done using NEWHEL93 (). All molecular figures were created using the program PyMol ().
Single crystals of d(GGTATACC) were gradually compressed from atmospheric pressure to 2 GPa (). Unit cell parameters of the A-DNA crystal were measured at each step of the pressure ramp from diffraction data recorded over a rotation range of 2°. In spite of the elongated (anisotropic) shape of the DNA cylinder, they decrease isotropically up to a pressure value of ∼1.5 GPa. The isothermal compressibility of the crystal is defined as = −1/(∂/∂), where is the unit-cell volume. The variation of is shown in . Least-squares fit up to 1.5 GPa is () = (∞) + .exp(−.) with V(∞) = 62290 Å, = 11030 Å and = 1.25 GPa. The derived isothermal compressibility is χ() = . . . exp(−.). The largest value is at ambient pressure (χ = 0.215 GPa). We recall that the compressibility of bulk water is 0.35 GPa; the compressibility of water in the vicinity of polar atoms is similar to that of an ‘ice-like’ structure, 0.18 GPa (). The mean value of χ() between ambient pressure and 1.5 GPa is 0.088 GPa, similar to the compressibility of tetragonal hen egg-white lysozyme crystal (,). Above ∼1.6 GPa, the cell volume increases (the compressibility is negative) and the crystal quality, as monitored by mosaicity and resolution of diffraction data, gradually deteriorates. The diffraction that extends to 1.6 Å at ambient pressure falls off to 3 Å at ∼1.8 GPa and is completely lost at ∼2.0 GPa. Four high-resolution structures of the A-DNA form were determined at ambient pressure, 0.55, 1.04 and 1.39 GPa. In each case, one or two crystals were sufficient to acquire high-completeness data () in spite of data collection at room temperature. Refinements were performed at 1.60–1.65 Å resolution to R-factors (R-free factors) of 15.2 (19.1), 16.9 (20.1), 19.3 (22.7) and 18.8 (22.6)%, respectively ().
The overall duplex structures shown in evidence axial compression of the helix, which reacts to pressure like a molecular spring. The base-stacking shrinkage is 2.6 Å for the full octamer length (from 23.5 Å at ambient pressure to 20.9 Å at 1.39 GPa, i.e. a relative contraction of 11%). The average base-pair step varies from 2.92 Å down to 2.73 Å (). This spectacular plasticity associates only small changes in phosphodiester backbone angles for accounting the denser base stacking. The sugar puckering parameters (Taum and P) describing the sugar conformations were analysed. All of them belong to the canonical C3′endo (or northern) conformation, a behaviour already mentioned earlier (). The pseudo-rotation P remains remarkably constant over the whole pressure range between standard values of 1 to 30°, including the O3′ end sugar of chain A, which moves from 30.9 to 33.3°. The only exception is the O3′ end sugar of chain B that adopts at ambient pressure a C2′/C1′ twist (nearly a southern conformation) and goes back to 36.8° (northern) at 1.04 GPa and above.
The variation of base-pair spacing versus pressure () is approximately linear up to 1.0 GPa and becomes steady beyond, which means that the molecule becomes progressively unable to accommodate increasing compression. The gradual degradation of the crystal order beyond ∼1.6 GPa may be due, at least partly, to this effect. Contrary to common B-DNA crystal packing arrangements, where duplexes stack on the top of each others building infinite helices along crystal axes, the hexagonal packing of d(GGTATACC) builds infinite super-helices around the 6-fold crystallographic axis with a close-contact zone between neighbouring helices (,). In this region, the wedge effect of hydrostatic pressure produces a hinge point in base stacking between the fifth and the sixth base pairs (). This steric effect might also contribute to crystal degradation.
The Watson–Crick type of base-pair association, which represents the foundation of the genetic code transmission, could be another way for the helix to ‘breathe’ transversally. In proteins, in the elastic compression regime, salt bridges and H-bond lengths are usually shortened by ∼0.1 Å.GPa (). In the case of A-DNA, evolutions of polar atom distances within G–C and A–T base pairs () can be interpreted in terms of small variations of rise, buckle, propeller-twist and other parameters that contribute to the adaptation to high pressure (). Accordingly, the geometry of Watson–Crick base pairings remains essentially invariant in the pressure domain up to 1.39 GPa. The lengths of vectors C1′–C1′, that may be used to quantify the DNA cylinder transversal squeeze, are all identical within their SDs, i.e. 10.5 Å, whatever the applied pressure.
It was suggested from modelling and docking simulations that water hydration under high hydrostatic pressure would tend to build octahedral arrangements () around DNA, to account for a negative ▵V. These arrangements have been invoked in the B/Z transition observed under high pressure (). The A-form of DNA shows a heavily hydrated major groove and a poorly hydrated minor groove. These features are distinct with respect to other DNA forms, and as such are worth to be investigated under pressure. The hydration scheme of d(GGTATACC) has been extensively analysed because it presents regular pentagonal arrangements () in the major groove (), a typical motif also observed in some well-ordered proteins at atomic resolution such as crambin (). The particular conformation of the A-form of DNA allows water molecules to directly bridge phosphate groups along each individual strand. At ambient pressure, 75 direct polar contacts are established through 51 water molecules belonging to the first shell of hydration, including nine over the 12 possible phosphate bridges. As mentioned earlier, the remaining water molecules build an ordered network from rim to rim thus completely filling most of the available space in the major groove. Under pressure, more water sites become apparent. This is illustrated in , which shows the same G–C base pair at one end of the duplex superimposed to the 2− electron-density maps calculated at different pressures. At 0.55 GPa, the number of observed isolated peaks in the first shell of hydration increases to 69, although some of them were suspected not to correspond to water sites as they fill parts of elongated electron densities that cannot be connected to the model. When pressure is increased above 1.04 GPa, one of the elongated densities resolves as a linear chain, ideally fitting a folded spermine molecule. The stabilization of the hydrogen-bond network is made evident by a slow but continuous decrease of the average normalized 〈 〉 thermal factor of water molecules common to all structures () down to a limit of ∼34 Å. The number of direct polar contacts from the first shell of hydration to the oligomer also increases rapidly with pressure up to 0.55 GPa, then more slowly up to 1.39 GPa. Beyond 1 GPa, the first shell of hydration is gradually compressed thus leading to more direct contacts towards the DNA (). Nevertheless, the pentagonal network located in the major groove remains conserved in the whole pressure range.
The four X-ray data sets were recorded with exposure times appropriate for collecting the A-form diffraction data but too short to get a complete picture of the diffuse scattering pattern described in (). The series of narrow streaks distributed in the pattern of a cross was barely observable, but the extended meridional streaks associated with the base-pair stacking were clearly observed up to at least 2 GPa. According to the interpretation given in (), the diffuse scattering pattern is produced by occluded B-DNA molecules in the packing channels of the A-DNA crystal structure. A suitable orientation of crystals in the high-pressure cell allowed us to record these meridional streaks () while ramping pressure, which provided data to determine the average stacking distance. A continuous and smooth shortening is observed, that can be monitored even in the range 1.5–2 GPa where the A-DNA crystal order falls off (). The average base-pair step is 3.34 Å at ambient pressure and 3.07 Å at 2 GPa.
We have shown that d(GGTATACC) in the crystalline A-form can withstand very high pressures, up to ∼1.9 GPa. The gradual loss of long-range order between 1.6 and 1.9 GPa may be related partially or completely to packing effects as mentioned previously and the molecule in solution might be stable even beyond 1.9 GPa. The four high-resolution structures show that geometry of base pairs is well preserved under compression up to at least 1.39 GPa. The B-form molecules occluded in channels of the crystal packing are in a solution-like environment. The information derived on this form from diffuse scattering is limited. Nevertheless, the smooth variation of the period of stacking derived from the evolution of meridian streaks reveals that the B-form is probably stable up to at least 2 GPa. We shall consolidate and extend this preliminary result by performing another single-crystal study on a dodecanucleotide that crystallizes in the B-DNA form.
The remarkable adaptation of d(GGTATACC) to high pressure is clearly associated to the base-paired double-helix topology of the molecule, by which it behaves as a molecular spring. These properties are probably shared by molecules featuring similar topology, with sugar-phosphate or polypeptide backbones. At the prebiotic stage, the base-paired double-helix architecture was crucial in the emergence of molecules with catalytic properties and able to store genetic information. Such architectures could withstand not only pressure in the deepest sea trenches but also much higher pressures found in Earth's interior or in the context of rare events such as impact of a meteorite. We suggest that this remarkable adaptation to harsh conditions may have played an important role during the sequence of events that led to the seminal RNA World. |
Methylation of cytosines is one of the marks of transcriptionally inactive chromatin (). Constitutively silenced chromatin, like pericentromeric heterochromatin, is heavily methylated in all cells throughout an organism's lifespan, whereas most promoters of genes are methylated only when permanent silencing is needed during development (). Clearly, different loci in the genome are treated distinctly by the DNA methylation machinery and more generally by the heterochromatin assembly pathway. To what extent those differences are mediated by differences in composition of the DNA sequence, the genomic context or by DNA binding proteins remains generally unclear.
One class of sequences illustrating the complexities of methylation patterns is comprised of transposable elements (TEs). TEs are thought to be silenced by heavy CpG methylation in mammalian genomes (). However, a number of reports suggest that the silencing of repeated sequences is not homogeneous throughout the genome or in all cells of an organism. For example, the human endogenous retrovirus (HERV) family HERV-K displays different methylation levels between copies and between cell lines (). In , differences in methylation are also observed between random selected copies of S1, a SINE element (). In humans, the promoter region of , a HERV-W copy encoding Syncytin-1 is specifically hypomethylated in placenta and methylated in other tissues (). Syncytin-1 is a retroviral protein with fusogenic properties involved in the development of human placenta (). However, it is unknown whether hypomethylation in this tissue is a characteristic specific to this co-opted copy or general to the HERV-W family. It is notable that DNA methylation, and the heterochromatic structure in general, have the property of spreading from the silenced target to the surrounding DNA (,,). Such observations have led to the assumption that transposable elements should generally have the same methylation level as their flanking sequence (). If this is true, different methylation levels between endogenous retroviral copies should reflect the methylation status of the surrounding DNA. This assumption, however, has not been extensively tested.
Placenta is of particular interest with respect to expression of endogenous retroviruses (ERVs) because many ERV families are transcribed in this tissue (). One of the hypotheses proposed to explain this relatively high degree of ERV expression is that placenta has lower DNA methylation compared to other tissues (,). Indeed, in mouse blastocysts, the trophectoderm seems to be less methylated than the inner cell mass (), whereas in human blastocysts the opposite situation has been reported (). On the other hand, human term placentas show about 20% lower overall DNA methylation compared to brain, liver and peripheral blood lymphocytes (PBL) (,) and 10–27% lower methylation of retroelements compared to that observed in spleen (). Moreover, compared to other tissues, placenta has a higher frequency of HERV copies acting as alternative promoters to cellular genes (,). These gene promoters of retroviral origin initiate transcription in their long terminal repeat (LTR) and, in most cases; their activity is tissue-specific. Indeed, about 40% of the tissue-specific LTR-derived gene promoters which have been documented are active in placenta. This could be the result of the general hypomethylation of this tissue allowing the expression of many HERV copies and consequently of LTR-derived promoters as well. The main questions addressed in this study were (i) whether the methylation level of LTR-derived gene promoters is representative of the general methylation of HERVs in placenta and (ii) whether differences in methylation between copies can be explained by differing genomic environments. To answer to these questions, we analyzed the methylation status of three HERV-E LTRs active as alternative gene promoters and of six other highly related LTRs in different genomic contexts.
Two term placenta samples were used for all bisulfite sequencing and combined bisulfite and restriction enzyme analysis (COBRA) experiments: P24 from a pre-eclamptic pregnancy, gestational age 32 weeks and P25, from a normal placenta, gestational age 40 weeks. For LTR-Mid1 only, two additional placental samples were used: P15 from an eclamptic pregnancy, gestational age 38 weeks and P33, from a pre-eclamptic pregnancy, gestational age 40 weeks. P21, a normal term placenta sample, gestational age 41 weeks, was used for the Southern blot. Different samples of normal PBLs were used for bisulfite methods (sequencing and COBRA) and Southern blotting.
Bisulfite conversion of DNA was performed using the EZ DNA methylation kit (Zymo Research), according to the manufacturer's protocol, with the following modification: genomic DNA was incubated with CT conversion reagent at 50°C for a total of nearly 4 h with 15 s pulses to 95°C every 15 min. Converted DNA was eluted in 20 μl elution buffer and 2 μl was used for PCR. All primers used are shown in Supplementary Table S1. To ensure that only the LTR of interest was amplified, the forward primer was located in the 5′ flanking sequence of this LTR. A first PCR (PCR1) was performed under the following conditions: 94°C for 5 min, followed by 40 cycles of 94°C for 30 s, variable annealing temperatures depending on the primers for 1 min, and 72°C for 30 s. A final elongation step of 10 min was included in all reactions. In most cases, a second PCR (PCR2) was necessary to obtain enough PCR products for subsequent analysis. PCR2 was performed either by using the same primers as for PCR1 or with nested primers (Table S1). For PCR2, 2 μl of the PCR1 product was used as template. PCR2 conditions were the same as for PCR1 except that only 35 cycles were performed. PCR1 and PCR2 were performed on each sample in three independent reactions to control for PCR bias in the following COBRA. The products were analyzed by gel electrophoresis and purified with MiniElute (Qiagen), then digested individually with restriction enzymes which contain CpG dinucleotides in their recognition sites. If the CpG was methylated in the original sample, then the restriction enzyme site is maintained during bisulfite treatment and the PCR product will be digested. Enzymes used and the number of restriction sites in each LTR is given in Supplementary Table S2. In some cases, several CpG sites were tested by a single restriction enzyme (e.g. four CpG sites of LTR-EBR were tested by TaqI, see Table S2). In these situations, it is impossible to distinguish among methylation states of the different CpGs. Therefore, the same results of COBRA are indicated for all those CpGs in and and they are marked by a connector line. Consequently, COBRA results when connected should be regarded as a sum of several CpG sites and not separately for each CpG site. All enzymes were purchased from New England Biolabs, and digests were performed following manufacturer's instructions. No PCR bias was observed between different PCRs performed on the same sample. Hence, the three PCR products from the same individual sample were pooled and cloned using the pGEM-T Easy kit (Promega). Sequencing was performed by McGill University and Genome Québec Innovation Centre Sequencing Platform. Only unique sequences (as determined by either unique CpG methylation pattern or unique non-conversion of non-CpG cytosines) are shown, and all sequences had a conversion rate >96%.
To compare the methylation of HERV-E LTRs between placenta and PBL in a genome-wide way, the LTR2B-cons-fw: 5′-TGAGGGAAGAGAGAGATTTTT 3′ and the LTR2B-cons- rev: 5′- ATTATAAAAAAAAAAACTTTATTCAACTAA-3′ primers were chosen in a well-conserved region of the LTR2B consensus given by Repbase (). (LTR2B is the Repbase nomenclature for a major subfamily of HERV-E LTRs to which LTR-MID1 and LTR-PTN belong). An PCR was performed with the UCSC In-silico PCR program (UCSC Genome Browser) using as primer sequences the corresponding LTR2B-cons-fw and -rv sequences for non-bisulfite converted DNA. Seventy-eight HERV-E LTRs were predicted to be amplified. A multiple alignment of the 78 sequences showed presence of a SnaBI restriction site (TACGTA) in 42 sequences, a ClaI (ATCGAT) in 36 sequences and a DraI (TTTAAA) in all 78 sequences (two sites present in most of them and at least one was maintained). For band intensity quantification the ImageQuant 5.0 (Molecular Dynamics) program was used. To deduce the percent of methylated CpGs, the theoretical number of CpG-carrying sequences out of the total of amplified sequences as predicted by the PCR was used.
Genomic DNA from placenta (sample P21) and PBL was digested with the methylation sensitive enzyme (HpaII) and the methylation insensitive isoschisomer (MspI) overnight. Then, the ApoI restriction enzyme was added for 5 h. All digestions were performed according to the manufacturer's instructions (New England Biolabs). Digested DNA was separated in a 1% agarose gel and transferred to a Zeta-Probe GT Blotting Membrane (Bio-Rad). The probe corresponding to the gag region of HERV-K was amplified by PCR with probe-K-fw: AGCGTGGTCATTGAGGACA and probe-K-rev: AAAGCTGAGATAAGAGGCATATTT primers. The 150 bp long PCR product was cloned into the pGEM-T Easy vector, confirmed by sequencing and labeled with 32-P. Hybridization was performed in ExpressHyb Hybridization Solution (BD Biosciences) at 60°C. Washing conditions were: 2 × 1 min and 2 × 20 min at room temperature in 2 × SSC; 0.1% SDS and 3 × 20 min at 60°C in 0.5 × SSC; 0.1% SDS. Quantification of band intensity was performed as for the genome-wide COBRA. To calculate the percent of methylated CpGs, the percent of sequences not carrying CpGs was deduced from the non-digested band in the MspI digest.
Individual LTRs were arbitrarily chosen as follows: the consensus sequence of the three LTR-derived promoters was derived by hand and placed upstream of the first 1260 bp of the HERV-E internal consensus sequence termed ‘Harlequin’ by Repbase (). This file was used as a query to search the human genome for similar sequences using the BLAT alignment tool (). Among the output sequences, LTRs were chosen based on the following criteria: their total length had to match the query and they had to be in the 5′extremity of a provirus. The genomic contexts also were taken into account: three LTRs located in introns of genes expressed in placenta were chosen (group B). Genes were considered to be expressed in placenta if corresponding placental expressed sequenced tag (ESTs) were reported in the UCSC Genome Browser (). Three LTRs were chosen to be the furthest away from any known genes and EST-rich regions (group C). Group A comprises the three LTR-derived promoters. The percent identity between all LTRs analyzed here is shown in Supplementary Table S3.
To determine which HERV-E copies are transcribed in placenta, we used two different strategies to deal with the EST data from the UCSC annotation database and the NCBI dbEST database, respectively. In the first case, the genomic coordinates of both HERV-Es and ESTs from placenta are known, and by comparing these coordinates, we identified all those HERV-Es which overlapped with ESTs in placenta and assigned them as specific HERV-E copies expressed in placenta. In the second case, there was no information on genomic coordinates available for those ‘simple ESTs’ from NCBI. However, since HERV-E copies are relatively old and diverged from other copies (e.g. Table S3), we successfully mapped most of these ESTs to the human genome with significant mapping results by using BLAT (standalone version: v. 33) (). In order to reduce the computational workload, we did not try to map all ‘simple ESTs’ to the human genome, but only those derived from HERV-Es by repeat-masking them with the HERV-E consensus sequence and the CENSOR program (v 4.1) downloaded from RepBase () After the mapping, genomic coordinates of those ‘simple ESTs’ were obtained, and the same strategy described in the first case was applied to find specific HERV-E copies expressing these ESTs in placenta.
Distribution free Mann–Whitney U-tests and Kruskal–Wallis H-tests were used for the statistical analysis.
The three LTR-derived gene promoters analyzed here are alternative promoters specifically active in placenta, keeping the coding sequences of the genes intact, and they are all derived from 5′ LTRs of HERV-E proviruses () ().
The first is the LTR-derived promoter of Pleiotrophin (LTR-PTN) (), a secreted heparin-binding cytokine with diverse functions involving mitogenic activity in fibroblasts, endothelial and epithelial cells, lineage-specific differentiation of glial progenitor cells, neurite outgrowth and angiogenesis (). A promoter of genomic origin called hereinafter ‘native promoter’ drives expression in developing brain and in the adult testis, uterus, glia and neurons. The provirus providing the LTR-PTN promoter is located around 70 kb downstream of the native promoter and 7 kb upstream of the first coding exon. Although Schulte . () reported that the LTR-PTN was the only promoter responsible for placental expression, native-promoter driven transcripts in placenta are present in Genbank. Experiments with PTN-depleted choriocarcinoma cells suggest that placental expression of PTN might be important for trophoblast growth, invasion and angiogenesis .
The second LTR-derived gene promoter studied here is associated with the endothelin B receptor gene (LTR-EBR) (). EBR is a G protein-coupled receptor of endothelin (). In human placenta, endothelins are involved in fetoplacental circulation and can also act as growth factors (,). The HERV-E provirus providing the alternative promoter is located 57.5 kb upstream of the native promoter which has a ubiquitous expression pattern. The LTR-EBR is responsible for about 15% of the total amount of placental transcripts ().
The last HERV-E LTR-derived promoter is linked to the Midline 1 gene (LTR-MID1) (). MID1 encodes a microtubule-associated protein (). Mutations in this gene are involved in the pathogenesis of Opitz syndrome, a genetic disorder affecting midline structures (). This gene has five alternative promoters, three of them being widely expressed, one adipose tissue specific and the LTR-MID1 placental and fetal kidney specific (). The provirus driving placental expression is located between two ubiquitous promoters: at 13 kb from the downstream promoter and 30 kb from the upstream promoter (). The retroviral promoter drives 25% of the placental transcripts and 22% of the fetal kidney transcripts ().
As mentioned earlier, LTR-MID1, LTR-EBR and LTR-PTN are alternative promoters specifically expressed in placenta (,,). To assess whether their expression correlates with their methylation levels, methylation of the three LTRs was determined by COBRA and bisulfite sequencing in placenta and PBL samples. Prior to this analysis, the expression profile of the three LTR-derived promoters was confirmed by RT-PCR in both tissues using the same samples as for the methylation analysis. All three promoters are active in placenta and silent in PBL (data not shown). The results of the COBRA and the bisulfite sequencing in placenta and PBL are shown in . In placenta, these three LTRs are almost completely unmethylated. In contrast, they are heavily methylated in PBL. The results obtained by COBRA are in good agreement with those obtained by bisulfite sequencing demonstrating that the sequenced clones are representative of the pool of amplified sequences. The low methylation of all three LTR-derived promoters in placenta and their high degree of methylation in PBL indicates that methylation and expression profiles of the HERV-E LTR-derived promoters correlate. These results suggest that DNA methylation is involved in regulation of the three LTR-derived gene promoters.
To clearly address the question of whether the strong hypomethylation of LTR-derived promoters in placenta reflects an overall hypomethylation of HERVs in this tissue, a Southern blot analysis with methylation-sensitive enzymes was attempted. However, use of a HERV-E specific probe resulted in smears and lack of any defined bands (data not shown). This is presumably due to the relatively high divergence (10–30%) between HERV-E copies. Since HERV-Es are too divergent for Southern blot analysis, a genome-wide COBRA analysis was performed. For this, primers corresponding to relatively conserved regions of the consensus LTR2B—a subfamily of HERV-E LTRs— were chosen allowing the amplification of 78 copies, as predicted by the UCSC PCR program (UCSC Genome Browser). Out of the 78 copies, 54% have a SnaBI restriction site and 46% a ClaI site, both sites including a CpG dinucleotide. A DraI site (TTTAAA) present on all 78 sequences was used to control that, when methylation is not a factor, the digestion profile was as expected according to the predictions and was identical for both samples from placenta and PBL. COBRA of this heterogeneous amplicon (A) yields a double band of lower size (bands number 2; A) corresponding to the methylated sequences. The undigested band (band 1; A) corresponds to LTR copies with either unmethylated CpG sites or those not containing the CpG site. Measurement of the relative intensity of bands 1 and 2 for each of the digestions of three independent experiments with SnaBI and one with ClaI shows 4–25% lower methylation in placenta compared to PBL. The average of the different measures suggests 14% lower methylation of HERV-E in placenta than in PBL. However, to correctly calculate these percentages, the precise proportion of sequences lacking the tested CpG sites is necessary. With this method, this proportion is theoretical, based on the prediction of the PCR program. This prediction could be somehow biased, since the primers used correspond to non-bisulfite treated DNA and are different from the actual primers used in the COBRA. The experimental estimation of the proportion of the actual CpG lacking sequences in the pool of amplified sequences is impossible. Hence, another experimental procedure allowing a more accurate estimation of the percent of methylation was necessary, in order to confirm our results obtained by the genome-wide COBRA.
For that purpose, the most recent family of HERVs, HERV-K [HML-2 subfamily, 1–2% divergence ()], was chosen to carry out a Southern blot with methylation sensitive enzymes. MspI and its methylation-sensitive isoschizomer HpaII were used and the results are shown in B and C. The majority of HERV-K copies seem to be methylated at the tested CpG site in the LTR both in placenta and PBL. Measurement of the relative intensity of bands 1 and 2 from two independent experiments and different exposure times, suggests 9–14% less methylation of HERV-K in placenta than in PBL with an average of 11%. This percentage is very similar to the one obtained by the genome-wide COBRA on HERV-E sequences.
To compare the methylation levels of other HERV-E copies inserted in similar or different genomic contexts with those of the LTR-derived promoters, we chose six HERV-E proviruses without any known gene-promoting function (see Materials and Methods section) and the methylation of their 5′ LTRs in placenta was determined by both COBRA and bisulfite sequencing. We define ‘group A’ as the three LTR-derived gene promoters described previously and ‘group B’ as the three out of the six random LTRs located in introns of genes expressed in placenta (). LTRs of groups A and B have similar genomic context: they are all located in introns of genes expressed in placenta with the exception of LTR-EBR which is located 57.7 kb upstream of the first exon of EBR gene (). LTR-VI is located in an intron of the Ankyrin S1 (ANKS1) gene expressed in many tissues including placenta. LTR-II is located in an intron of the Tissue Factor Pathway Inhibitor 1 gene (TFPI1) expressed mainly in endothelial cells and also in placenta. LTR-IV-2 is located in an intron of the TMEM144 gene highly expressed in the central nervous system and providing a few ESTs in placenta. The other three LTRs are located far away from any annotated gene and are ‘group C’ (). The results of methylation analysis of these LTRs are shown in . LTRs of group B present 22.6–41.2% methylation and LTRs of group C 30.5–91% methylation. COBRA results are in good accordance with bisulfite sequencing results reflecting the global methylation of each LTR. Group A and B LTRs have no significant intra-group differences in their methylation levels in placenta whereas group C is very heterogeneous in terms of methylation. Hence, it cannot be considered as a homogeneous group regarding methylation. This could be due to the fact that the definition of this group as ‘far from genes’ is inexact: the presence of unknown genes or transcriptional activity in the vicinity of such LTRs cannot be excluded. For instance, LTR-III shows a methylation level (30.5%) similar to the levels displayed by LTRs of the group B. LTR-III could be located nearby an un-annotated transcriptionally active region. However, according to the UCSC Genome Browser, no obvious difference is observed between the insertion sites of group C LTRs with regard to EST occurrence in nearby regions or the conservation of the surrounding sequences between mammalian species. The other two LTRs of group C, LTR-IV and LTR-IV-3, have significantly higher methylation levels when compared to group B and group A (). The comparison between groups A and B, both located in introns of genes expressed in placenta, shows that group A displays highly significant lower methylation than group B ( = 2e). This shows that groups A and B, although located in regions with similar transcriptional activity, differ in their methylation status. In conclusion, LTR-derived promoters have significantly lower methylation levels than any arbitrarily chosen LTR even when located in similar genomic contexts.
Our study shows that HERV-E copies display widely variable methylation levels in placenta. It was unknown how many of the approximately 294 copies of the full-length HERV-E proviruses [out of ∼1200 total HERV-E copies ()] are transcriptionally active in placenta and we chose the six random LTRs based on similarity and location criteria but with no knowledge of their expression status. To gain an estimate of the number of HERV-E copies transcribed in placenta and to assess whether the difference of methylation between copies correlates with expression, we extracted placental ESTs homologous to HERV-E derived sequence and then identified the specific copies of HERV-Es producing each of these ESTs (see Materials and Methods section). The results are shown in . The 162 placenta-specific ESTs were clearly identified as corresponding to HERV-E sequences but only 74 could be unambiguously mapped to 26 HERV-E loci. ESTs comprising only HERV-E sequences are described as ‘simple’ and those comprising non-HERV-E sequences as well are denoted as ‘chimeric’. ESTs starting inside the 5′LTR are likely to correspond to transcripts initiated by the LTR itself. Only 10 of the 26 copies are associated with this type of EST and are classified as ‘likely’ transcriptionally active copies. Among these, the proviruses corresponding to the three LTR-derived promoters were identified. ESTs corresponding to HERV-E's internal region could be the result of transcription initiated by the provirus itself or by upstream sequences. Therefore nine copies associated with these types of placental ESTs are classified as ‘possibly’ transcriptionally active. Among these copies, LTR-II and LTR-VI were identified, both members of group B. ESTs comprising the upstream flanking sequence of a 5′LTR or corresponding to copies deleted in their 5′ extremity are probably transcribed from a genomic promoter upstream of the provirus. Consequently, they are considered as ‘unlikely’ transcriptionally active copies. LTR-II provides five ESTs corresponding only to its internal sequence and four spliced ESTs composed of exons corresponding to the provirus and the flanking gene. Hence the LTR-II copy is considered as possibly active but also involved in read-through transcription. Interestingly, one of the ‘likely’ active copies (chromosome 1 : 166098952–166105765) seems to act as an alternative promoter for the IQWD1 gene of unknown function. Therefore, over one third of the ‘likely’ transcriptionally active copies correspond to LTR-derived alternative gene promoters.
As reported by others (), this study suggests that only a limited number of HERV-E loci are transcriptionally active. Notably, only two out of the 1200 HERV-E copies present in the genome, those at chromosome 17: 23581671–23590492 and chromosome 20: 24847861–24861663, provide 36% of the HERV-E placental ESTs (). Moreover, solitary LTRs represent 75% of all HERV-E copies but only 16% of the ‘likely’ or ‘possibly’ active copies correspond to this type of element. It seems that full length copies are more expressed than solitary LTRs as previously observed (). It is also surprising that two LTRs (II and VI) out of the 294 full length copies identified in this screening as possibly active happened to be chosen at random for our methylation study of individual LTRs. This is very likely due to the criteria used for the choice of the individual LTRs: the sequence similarity to the LTR promoters (LTR-VI is the closest related to the LTRs MID-1 and PTN and LTR-II to LTR-EBR, see Table S3), the fact that they are part of full length proviruses and their localization in introns of genes expressed in placenta. All these criteria seem to increase the likelihood for an LTR to be active in placenta.
We attempted to determine whether a clear relation existed between level of methylation and level of expression as reflected by the number of ESTs found in placenta. Among the copies studied here, only ESTs corresponding to unmethylated copies (group A) or to copies with intermediary levels of methylation (group B) were found. Although these two groups differ by their methylation levels, no obvious difference was observed in the number of ESTs that they provide and the amount of data does not allow any statistical analysis. Therefore, experimental analysis of their expression is necessary to determine if differences in methylation levels as low as 20% affect expression. It has been reported that the same Intracisternal-A-Particle (IAP)-LTR promoter can be active or silenced in different individual mice when its methylation differs only by 20–30% (). This could be due to the fact that, in some cases, LTR promoters seem to be active not because of their overall hypomethylation in a given tissue but because of their complete demethylation in a limited number of cells of this tissue. Indeed, the IAP-LTRs mentioned above, show not only a strong variability in methylation between mice but also a mosaic methylation between cells of the same mouse. A similar phenomenon could explain why our study identified ESTs corresponding to LTR-VI and LTR-II but not LTR-IV-2. The latter is the only one from group B not presenting any completely demethylated sequences. In addition, when different LTRs are compared, it is very likely that methylation or chromatin context is not the only factor responsible for the transcription levels; sequence divergence and heterogeneity for transcription factor binding sites will be involved as well (). For example, when the promoter activity of two unmethylated HERV-K LTRs was compared in transient transfection assays, they displayed different levels of transcription (). However, when their endogenous expression was tested in human teratocarcinoma cells, they displayed different levels of transcription but in the opposite sense, since the stronger LTR promoter is more methylated in the cell.
To determine if methylation of HERV-E LTRs is the same as their flanking sequence and if the difference in methylation between LTR-derived promoters (group A) and random LTRs located in similar genomic contexts (group B) is related to differences in methylation of their flanking regions, we determined the methylation status of flanking sequences of two LTRs from group A (LTR-MID1 and LTR-EBR), one LTR from group B (LTR-IV-2) and one LTR of group C (LTR-IV). Methylation of the CpG sites contained in 500 bp adjacent to the LTRs was determined by bisulfite sequencing and results are shown in . LTR-IV (group C) has the same methylation level as its flanking sequence, whereas LTR-IV-2 (group B) and LTR-MID1 have much lower methylation levels compared to their flanking sequence ( < 0.001, ). For LTR-EBR and its flanking sequence, no significant difference in methylation level was revealed by bisulfite sequencing. However, a COBRA of LTR-EBR and its flanking region suggests a low methylation level of the flanking sequence and a complete absence of methylation for the LTR-EBR itself. The flanking sequences of LTR-IV-2 and LTR-MID1 show similar, high methylation levels (92.3% and 70.4%, respectively) even though they are located in introns of genes expressed in placenta. This discrepancy between levels of methylation of the LTRs and their flanking sequences is striking since it is believed that the main determinant of ERV methylation is the chromatin state of their insertion site (). To confirm this pattern and eliminate the possibility that this observation results from eventual PCR biases, we amplified and sequenced 500 bp containing five CpGs of the LTR-IV-2 and three CpGs of its flanking sequence (). Indeed, these clones confirm the pattern observed previously. For LTR-IV-2, methylation from the flanking sequence seems to spread over the first CpGs of the LTR but drops gradually towards the internal CpG sites. Methylation of LTR-MID1 seems to drop more abruptly at the border of its 5′ extremity resulting in a clear-cut difference of the methylation level between LTR-MID1 and its flanking sequence. This difference between LTR-IV-2 and LTR-MID1 could be explained by a less efficient spreading of methylation through LTR-MID1 because of a lower density of CpGs in the flanking sequence or/and to transcription factors binding to the LTR and protecting it from methylation (,). Indeed, LTR-MID1 is transcriptionally active, but no ESTs corresponding to LTR-IV-2 were identified by our screening of the NCBI database, suggesting this LTR is not active. Moreover, no difference in methylation between flanking sequences of different groups was observed: they all present high methylation levels except for one of the two members of group A (LTR-EBR). Considering CpG sites located in other repeats present in the flanking sequences, there is no evidence for more methylation of repeated DNA compared to unique sequences. In conclusion, very different levels of methylation are observed in flanking sequences, an observation that cannot be explained by the nature of the sequence (repeat or unique) or the transcriptional activity (intron or inter-genomic sequence). Our results suggest that there is no systematic correlation between methylation levels of HERV-E LTRs and their flanking sequences. Furthermore, the hypomethylation of the LTR-derived promoters compared to random LTRs cannot be explained by a difference in methylation of their flanking sequences.
In this study, we report methylation levels of nine HERV-E LTRs. We show that methylation levels of LTRs in placenta are widely variable, ranging from 4% to 91%. It seems that the genomic environment has an impact on the methylation level of the LTRs, since those located in well-defined genomic environments such as introns of genes expressed in placenta (groups A and B) have similar levels of methylation in intra-group comparisons. However, genomic environment alone cannot explain other differences observed; for example, the difference between LTR-derived promoters and LTRs located in introns of genes expressed in placenta. This statement is even clearer when flanking sequences of LTRs are compared. The methylation level of an LTR is not systematically similar to that of its flanking sequence, whether it is an LTR-derived promoter or a random LTR. Clearly, spreading of heterochromatin is not the only determinant of the degree of methylation of LTR promoters. Other factors such as transcription factors binding the LTRs are likely involved. Expression levels of individual LTRs will depend on both methylation/chromatin context as well as the retention of transcription factor binding sites. Indeed, our screening for transcriptionally active HERV-Es in placenta show that very few of them are active, very likely because expression is dependent on the combination of factors mentioned earlier.
LTR-derived promoters occur more frequently in placenta than in any other tissue (,). This does not seem to be solely due to a general hypomethylation of the DNA in this tissue, since the LTR-derived promoters studied here are almost completely unmethylated and therefore do not reflect general methylation levels. Moreover, LTR-MID1 is located in a highly methylated flanking region showing that its lack of methylation does not result from a general ‘protection’ of the surrounding sequence against methylation. Consequently, the expression of LTR-derived promoters in placenta cannot be explained only as a secondary effect of the genomic environment. The HERV-E LTR-derived gene promoters studied here present tissue-specific methylation in accordance with their expression profile. Their epigenetic status and activity make them more analogous to tissue-specific promoters of genomic origin than to transposable elements.
Overwhelming evidence for essential roles of ERV-encoded proteins in placenta morphogenesis and evolution has been reported in recent years (). However, there may be additional impact of ERVs in the placenta due to host gene regulation effects. It is clear that most genes essential for placental development are not specific for this tissue (), suggesting that placenta specific regulatory networks may play an important role in morphogenesis and evolution of this organ.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
ELK-1 belongs to the ternary complex (TCF) subfamily of the ETS-domain transcription factors. It is a 428 amino acid (aa) protein (52 kDa) containing the four functional domains characteristic of all members of the TCF family. The TCFs are major nuclear targets for the RAS–MAPKs (mitogen-activated protein kinases) of the extracellular-signal regulated kinase (ERK) subfamily, and the closely related SAPK/JNK and p38 stress-activated protein kinases. They therefore appear to act as an integration point for both growth and stress signals. In cultured cells, ELK-1 functions as a transcriptional activator via its association with serum response factor (SRF) in a ternary complex on the serum response element (SRE) of many immediate early genes (IEGs: e.g. and ) ().
De-regulation of ELK-1 expression has been associated with cancers of the prostrate () and breast (), as well as certain T-cell malignancies (). Mitogenic regulation of cell proliferation is accompanied by up-regulation of immediate early genes such as that play a critical role in activation of downstream AP1-dependent genes required for growth. With regards to breast cancer, recent studies on the MCF-7 cancer cell line have shown that 17β-Estradiol (E2) activates both the RAS–MAPK and the PI3K/AKT pathways. The former leads to phosphorylation of ELK-1 and the latter phosphorylation of the SRF, both of which activate the SRE in the proto-oncogene promoter (,). Both these pathways also impact directly on the translational read-out of the cell. However, the extent to which the E2-mediated altered growth phenotype was related to changes in the translation of the elk-1 mRNA (or other mRNAs encoding components of these signalling pathways) was not investigated.
Our lab is interested in the regulation of ELK-1 expression at the level of translation initiation, a step which is normally rate-limiting in protein synthesis. In the majority of eukaryotic mRNAs, translation begins with the binding of the eukaryotic initiation factor 4F (eIF4F) to the capped 5′ end, an interaction that is mediated via its eIF4E subunit. The eIF4F cap-binding complex is also composed of eIF4A, a member of the ‘DEAD-box’ family of RNA helicases, and eIF4G. Binding of eIF4F leads to recruitment of the ribosomal 43S preinitiation complex (containing the eIF2-GTP-tRNA ternary complex). The ribosome, and associated factors, is then thought to linearly scan the RNA unwinding intermolecular structure in an ATP-dependent fashion until an AUG codon is encountered and a 48S initiation complex forms. Codon–anticodon base pairing in the ribosomal P site is then thought to trigger hydrolysis of the eIF2 bound GTP, and joining of the 60S ribosomal subunit (). This multi-component, coordinated series of events is strictly regulated and responds to both intra- and extra-cellular signals ().
Very few studies have been performed on the translational regulation of the gene. However, it was reported that AKT negatively regulated ELK-1 expression at the level of translation and that a region within the first 279 nt of the ELK-1 open reading frame (ORF) was necessary and sufficient for this control (,). These latter studies were performed using GST–ELK-1 fusion constructs; i.e. the 5′ untranslated region (UTR) of the elk-1 mRNA was removed and GST was fused N-terminally. However, the elk-1 5′ UTR exhibits many of the features associated with cellular mRNAs whose expression is tightly controlled at the level of translation initiation, including upstream open reading frames (uORFs) and predicted thermodynamically stable RNA structures positioned close to the cap. In addition, an alternatively spliced transcript with a 106-nt deletion within the 5′ UTR has been reported, opening the possibility that these alternatively spliced isoforms of the elk-1 mRNA could be linked to the translational read-out. Therefore, in this manuscript we have gone back and examined the role of these 5′ UTRs in the regulation of ELK-1 expression. The results provide intriguing insights into how the different elements identified, namely uORFs and RNA structure, impact on ribosomal access to the ELK-1 AUG initiation codon.
HEK293T were cultured in Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal calf serum (Brunschwig), 1% penicillin/streptomycin, in a humidified atmosphere containing 5% CO. For the polysome analysis, cells were taken in the growing phase. They were hypertonically shocked by shifting to medium containing 300 mM NaCl for 50 min. The cells were then placed in normal isotonic medium for 30 min. When rapamycin (Rap) was used, 100 nM Rap (LC laboratories) or 0.01% DMSO (the negative control) was added during the hypertonic shock, 20 min before the transfer back to isotonic conditions. Rap and DMSO were kept on the cells throughout the recovery period.
Calcium phosphate-mediated DNA transfections were performed essentially as described in (). All transfections were performed in triplicate. The activity of firefly (FLuc) and renilla luciferases (RLuc) in the lysates prepared from transfected cells were measured using the Dual-luciferase reporter assay system (Promega) and light emission was measured over 10 s using a TD-20/20 luminometer (Turner Designs).
After treatment, cells were scraped into the culture medium and pelleted for 4 min at 800 r.p.m. The pellets were lysed for 15 min on ice in 100 mM KCl, 50 mM Tris–Cl pH 7.4, 1.5 mM MgCl, 1 mM DTT, 1 mg/ml heparin, 1.5% NP-40, 100 μM cycloheximide, 1% aprotinin, 1 mM AEBSF and 100 U/ml of RNasin. Nuclei were pelleted by centrifugation, 10 min at 12 000 r.p.m. The supernatant was loaded on a 15–50% sucrose gradient (in 100 mM KCL, 5 mM MgCl, 20 mM HEPES pH 7.4 and 2 mM DTT). Extracts were fractionated for 3 h 30 min at 35 000 r.p.m. at 4°C in a Beckman SW41 rotor. The gradient was analysed and the fractions collected with the ISCO UA-6.
The RT–PCR was performed using the One Step RT-PCR kit (Qiagen) according to the manufacturer's instructions. The number of amplification cycles was first determined for each set of primers, and corresponded to the exponential phase of the different products.
The bicistronics constructs were made by cloning three inserts into the pBS–KS+ multiple cloning site. The 5′ insert corresponding to was generated by PCR from the pRL-SV40 Vector (Promega) with the primers (5′-GAGCTCGCTAGCCACCATGACTT-3′) and (5′-TCTCGAGCCCCTAGAATTATTGTTC-3′), and digested with SacI/XhoI. The central insert contained the 5′ UTRs of Elk-1 generated by PCR with the primers (5′-TGGGTCCATGGCTGGGGGAGTGC-3′) and (5′-GCTCTAGAGCTCGAGACATTGGGCTCCTCCTCCTCGGGCCCACGTGAGCTGTAGGGAAACGC-3′) with XhoI as the 5′ site and NcoI as the 3′ site. The 3′ FLuc insert was excised from pGL3-Basic (Promega) as an NcoI/XbaI fragment. The monocistronics constructs were generated by removing . The ΔSL clones were made by removing a SmaI/BstUI fragment in the monocistronic constructs. All clones were transferred from pBS–KS to pEBS–PL () as SacI/Acc65I fragments. The ΔSL constructs expressed from the pEBS-PL vector retain 59 nt upstream of the first AUG codon (uAUG1) whose Kozak context also was unaltered. This expression vector was selected because its intron is positioned downstream of the cistrons. Consequently this eliminates the risk of generating a FLuc monocistronic transcript by alternative splicing ().
The pBS 5′ UTR clone was linearized with Acc65I (position 208 on the mRNA) and a T7 polymerase RNA run-off transcript generated. This was 3′ end labelled using P-pCp (Amersham) and RNA ligase (Amersham). The probe was purified from a 5% acrylamide–urea gel by passive diffusion. Structural probing studies were performed by titrating the RNase V1 (Ambion) and RNase T1 (Roche) precisely as outlined in the Ambion kit. Products were resolved on 5% acrylamide–urea gels and visualized by autoradiography.
The 15.2-kb gene spans seven exons (I to VII) and six introns, with the coding region encompassing exons III to VII (A). However, an alternatively spliced form which removes the non-coding exon II was detected in a human hippocampus cDNA library (). The longer form has a 5′ UTR of 346 nt [based upon the +1 position determined by (); hereafter referred to as 5′ UTR] and spans exons I to III. It contains two uORFs (A, middle panel), a feature characteristic of many cellular messengers whose expression is tightly controlled at the level of translation initiation (). RNA-folding programs predict considerable secondary structure. Indeed, within the first 168 nt (which are 75% G/C) there is a highly stable stem-loop (referred to hereafter as SL), with a Δ° approaching −40 kcal/mol (C). Earlier results indicated that such structural elements could serve as effective barriers to scanning ribosomes (,) and this was recently confirmed in live cell studies (). The AUG1 is located within the loop of SL and its ORF (an ORF collinear with that of ELK-1) terminates after 54 codons. It is then followed by a second short uORF of only two codons (+1 relative to ELK-1), which terminates 17 nt before the ELK-1 AUG. The shorter transcript (5′ UTR) is 240-nt long and is generated by the removal of the 106-bp exon II (A, lower panel) (). Like 5′ UTR, it is G/C rich, and has conserved the SL and the two uAUGs. However, both are now in the same frame (i.e. there is now only a single uORF of 26 codons). RNA structural probing performed on the 5′ 208 nt (a region common to both spliced forms) clearly demonstrated V1 sensitive regions corresponding to the major part of the stem of SL and T1 sensitive sites within the loop regions (). Therefore, the major features of the predicted fold exist in solution.
We initially asked if we could detect these alternatively spliced transcripts in a range of human cell lines and human tissues. The elk-1 mRNA is generally expressed at very low levels and is difficult to detect in most cells even by RNase protection (data not shown). Therefore, we used semi-quantitative RT–PCR performed with oligos flanking exon II. To ensure the quantitative nature of this approach, we went to considerable effort to analyse the read-out during the linear phase of amplification (Supplementary ). Both long and short forms were readily detected (the intermediate band indicated by a star is actually a DNA heteromer composed of one light strand and one heavy strand generated during the PCR) (B and C). In the tissue samples, the relative amounts of both isoforms were similar in colon, lung and testis (B). However, the short form was major in the brain. Both forms were present at similar levels in the MRC-5 (lung fetal primary cells) and 293T (kidney) cell lines; however, the short transcript was dominant in HeLa S3 (cervical carcinoma), SK–NAS (neuroblastoid) and GM 06990 (lymphoblast) (C). Attempts to perform real-time quantification proved frustrating, probably due to the extensive secondary structure around exon II. In addition, such an approach required the utilization of a primer specific for 5′ UTR that traversed the exon I/III boundary, and one specific for 5′ UTR (within exon II). The former in particular gave very non-reproducible read-outs. We next examined if the levels of elk-1 transcript correlated with intracellular protein levels. Immunoblots performed on extracts prepared from the five cell lines revealed no clear transcript:protein correlation (D). Although ELK-1 was abundant in 293T cells and readily detectable in HeLa S3, it was barely visible in the SK–NAS and GM 06990 cells and totally undetectable in the MRC-5 cells (even when the blot was over-exposed). We also observed no difference in the ELK-1 protein half-life in 293T and HeLa cells (data not shown). These results indicate that ELK-1 protein levels can be regulated in a cell-specific manner at the level of translation.
The 5′ UTR represents a key element in translational regulation. Therefore, despite the fact that both 5′ UTR and 5′ UTR retain the two uAUGs and the major structural elements, it is not inconceivable that the role of alternative splicing is to modulate the translational read-out. Such a scenario would be consistent with the variability of the two forms both in human tissue and cell lines. We decided to examine the translational behaviour of the endogenous transcripts. Translation initiation on messengers with long structured 5′ UTRs is considered to be highly sensitive to the activity of eIF4E (,). Therefore, we examined the recruitment of the elk-1 mRNAs onto polysomes in the presence or absence of the drug rapamycin (Rap). Rap is a lipophilic macrolide that binds mTOR preventing its signalling downstream to S6K1 and 4E-BP (). This leads to the sequestration of eIF4E into an inactive complex with hypophosphorylated 4E-BP.
High salt (300 mM Na) provokes a rapid inhibition of protein synthesis, disaggregation of polysomes (compare A and B), dephosphorylation of eIF4E, 4E-BP1, rpS6 and an increased association of eIF4E and 4E-BP1 (,). Upon restoration of isotonic conditions these effects are reversed and the polysomal fraction reconstituted (C). The technique is in effect an competition experiment in which one creates a pool of free ribosomes stripped from their mRNAs by hypertonic shock then relieves this block and follows the ability of the endogenous mRNAs to compete. Using this approach, it is possible to examine what effect Rap has on the re-recruitment of RNA populations onto polysomes. In particular, one can directly examine the ability of the two endogenous elk-1 mRNA transcripts to compete for normal levels of eIF4E (Rap), or limiting amounts (Rap). In addition, by analysing the polysomal distribution rapidly after the restoration of isotonic conditions (i.e. 30 min) one can effectively eliminate the secondary effects of Rap on the cellular transcriptional program ().
In 293T cells, 30 min after the restoration of isotonic conditions the polysomes had started to reform (C). This process was clearly delayed in the Rap-treated cells (D). We used RT–PCR to investigate how the two elk-1 transcripts had behaved on these gradients. Reactions were performed with 50 ng of total RNA per fraction. As a control, we followed the distribution of the β-actin mRNA (a house-keeping gene with a short 73-nt poorly structured 5′ UTR) (C and D, lower panels). In cells treated with Rap, we observed that the 5′ UTR was enriched in the ‘monosomal-light polysomal’ fraction of the gradient compared not only to the 5′ UTR but also actin (D, lower panel). This suggests that the 5′ UTR elk-1 transcript is less sensitive to Rap and therefore preferentially re-recruited (and therefore proportionally enriched within the 50 ng of RNA). The fact that we do not see these effects in the heavy polysomal region of the gradient may reflect the short recovery time that was tested (i.e. the pattern in this region of the gradient essentially reflects the small fraction of mRNAs that remained heavy polysomal after the initial hypertonic shock). Nonetheless, these results showed a different behaviour of the 5′ UTR and the 5′ UTR. We decided to confirm this using an alternative approach. The two elk-1 5′ UTRs were fused to a FLuc reporter in the PolII expression vector pEBS-PL. As controls, we also fused the EMCV IRES and the 5′ UTR from the SeV P/C mRNA (AUG81: it is 110-nt long, has no detectable IRES activity, no uAUGs and little predicted structure) (A) (). These constructs were co-transfected into 293T cells with a vector expressing LacZ (pHR′CMV-LacZ), which served as an internal control against which FLuc activity was normalized. Cells were then treated or non-treated with Rap between 4 and 24 h post-transfection, and the effect of the drug confirmed with immunoblots against 4E-BP1 (B). As expected, the EMCV-FLuc construct was resistant to the drug, consistent with its IRES activity. The fact that the AUG81-FLuc normalized values were <1 indicates that its 5′ UTR is more sensitive to Rap than that of the internal control (this may be a consequence of the controls shorter 5′ UTR). However, the 5′ UTR gave normalized values nearly 2-fold lower than the 5′ UTR, a result consistent with its increased sensitivity to the drug (C).
One possible source of eIF4E independent translational control mediated via the 5′ UTR (particularly those that are long, structured and contain multiple uAUGs) is an internal ribosomal entry site (IRES). IRES activity has been reported in the 5′ UTRs of cellular mRNAs that play key roles in growth control (). We therefore tested for this activity using the now classical Rluc–FLuc bicistronic assay. The 5′ UTR and 5′ UTR were inserted into the intercistronic region of a bicistronic construct generated in pEBS-PL. As positive controls, we used the viral IRES of encephalomyocarditis virus (EMCV) and the 5′ UTR of Dap-5 (a reported cellular IRES). A negative control was provided by an inverted form of the EMCV IRES [referred to as RLuc-(VCME)-FLuc] (C). Plasmids were transfected into 293T cells and reporter assays were performed at 24 h post-transfection (A). Both EMCV and Dap-5 gave significant second cistron FLuc activity above the VCME negative control. Likewise, small but significant FLuc activity was detected with the elk-1 constructs. However, the bicistronic assay is in itself an insufficient test to unambiguously assign IRES activity (). In addition, when these constructs were transfected into cells as bicistronic RNAs or translated in RRLs only the EMCV construct screened positive (data not shown). This opened the possibility that second cistron activity was derived from a shorter monocistronic construct. Northern blot analysis of total RNA from the transfected cells using either a FLuc or RLuc probe confirmed the presence of a single bicistronic transcript in all transfections (B). However, this technique is considered relatively insensitive. We therefore decided to test for cryptic promoter activity by subcloning the (5′ UTR)-FLuc, (5′ UTR)-FLuc and (Dap-5)-FLuc regions into the promoter-less vector pGL3. Upon transfection into 293T cells both elk-1 5′ UTRs gave FLuc activity that was significantly higher than that observed in the negative control (pGL3-Basic FLuc) consistent with a promoter activity (D). Reporter activity from the Dap-5 construct was consistently lower than the negative control. These results were confirmed by combining the pEBS-PL bicistronic constructs with a siRNA generated against RLuc as outlined in (). The increased resistance of FLuc activity to a siRNA directed against RLuc in the bicistronic constructs carrying the 5′ UTR and 5′ UTR is consistent with the expression of a shorter monocistronic FLuc transcript (E). The cryptic reporter activity derived from both 5′ UTRs is weak (we estimate 5% of the activity of an SV40 minimal promoter) explaining why a short monocistronic FLuc transcript was not detected by northern blot (B). However, cellular IRESes are frequently reported to have activities significantly weaker than those of viral IRESes such as EMCV. In this context, weak promoter activity could have a major impact on the interpretation. Therefore, experiments failed to provide conclusive evidence for an IRES within either the 5′ UTR or 5′ UTR.
Since no IRES activity could be demonstrated within the 5′ UTRs we next analysed the role of the uAUGs as translational modulators. Although both UTRs retain two uAUGs (referred to as AUG1 and AUG2), the 5′ UTR contains two uORFs whereas both AUGs are in frame in the short form (i.e. there is a single uORF that shares a common UGA stop codon with the second uORF in the 5′ UTR: ). It has previously been reported that the two uORFs in the human oncogene mdm2 mRNA act synergistically to repress oncoprotein expression (). Additionally, since ribosomes enter the elk-1 mRNA via the 5′ cap, the recognition of these uAUGs by the scanning pre-initiation complex and the fate of the ribosomes after termination of these uORFs (i.e. are they released from the mRNA or does a fraction continue to scan to subsequently reinitiate downstream?) become central elements in the regulation of ELK-1 expression (). The question we therefore posed was ‘Does these alternative uORF organizations have a consequence for initiation at the downstream ELK-1 AUG codon?’ Starting with the (5′ UTR)-FLuc and (5′ UTR)-FLuc monocistronic constructs, we mutated AUG1/AUG2 both individually and together, as well as the UGA stop codon (). Both 5′ UTRs functionally repressed reporter gene expression (∼12-fold for the 5′ UTR and ∼11-fold for the 5′ UTR compared with the AUG81 control). Within the context 5′ UTR, the AUG1 change was largely neutral whereas AUG2 and AUG1 + 2 produced a marked and progressive increase in the reporter read-out which reached levels >30% of AUG81-FLuc. Curiously, the AUG1 + 2 double mutation in the context 5′ UTR had a much more modest effect suggesting that uAUG-mediated repression may be less important in this background. Changing the UGA stop codon reduced reporter activity by . 2-fold for both 5′ UTR and 5′ UTR. This mutation extends the uORF so that it overlaps with that of FLuc by six codons (). This drop in reporter activity suggests that a fraction of the initiation events at the ELK-1 start codon arise by re-initiation. Since it occurs in both 5′ UTR contexts, indicates that it derives from ribosomes that initiate at AUG2. Small uORFs (in this case only two codons) are thought to be favourable for re-initiation probably because the 40S subunit post-termination remains associated with the mRNA and retains many of the initiation factors that are normally lost progressively during the elongation phase (,).
We have frequently evoked RNA structure as an element in the translational repression mediated by the 5′ UTRs. Indeed, the relatively modest effect of the AUG1 + 2 mutations within the 5′ UTR led us to postulate that the structure was playing a central role in modulating the translational read-out. We decided to examine this directly by deleting all the structural elements in the FLuc monocistronic constructs described earlier (referred to as ΔSL). This leaves AUG1 in place with the same Kozak consensus and does not perturb the organization of the uORFs in both 5′ UTR and 5′ UTR (C). When total RNA was isolated from cells transfected with these constructs and expressed in a rabbit reticulocyte lysate (RRL), the removal of the SL region had a marked positive effect on the reporter read-out (Supplementary ). systems are known to be highly sensitive to RNA structure close to the cap (,), and this result confirms that such elements exist in the two 5′ UTRs of elk-1. However, this change had only a very modest effect when the same constructs were assayed (). Why should the disruption of structural elements within the 5′ UTR have such a small effect? One possible answer arises from an examination of the structure. The SL fold positions the AUG1 in the loop region (C), and the results from the AUG mutation studies demonstrated that it is largely silent. We postulated that AUG1 was not always seen by the ribosome because SL promoted ribosomal shunting (or discontinuous scanning) thereby rendering it invisible (). Thus in the ΔSL mutation the reporter read-out is the sum of the positive effect due to the removal of structural elements and the negative effect of the now accessible uAUG1. Therefore, mutation of AUG1 in the ΔSL background should now have a more marked positive effect on the reporter read-out. The ΔSL AUG1 double mutation was therefore introduced into the pEBS-PL 5′ UTR-FLuc and 5′ UTR-FLuc constructs and these were analysed by transient expression in 293T cells (). The effect of the AUG1 mutation was clearly enhanced in the ΔSL background, particularly within the 5′ UTR (. 2.5-fold). These results would suggest that at least a fraction of the scanning ribosomes bypass AUG1 by shunting, an event promoted by the stable SL (). However, should SL be unfolded, the continued repression of ELK-1 protein levels would be assured by the uAUG1. A dual mode of pre-initiation complex displacement (i.e. linear scanning versus shunting) that responds to physiological signals within the cell has already been reported on an mRNA ().
To independently confirm that ribosomal access to the uAUG1 is modulated by structure and to examine directly initiation events at the start codon, the following experiment was performed. Starting with the pEBS-PL constructs 5′ UTR-FLuc and ΔSL 5′ UTR-FLuc, the uORF was fused to that of the reporter (A upper panels). These were transfected into 293T cells alongside a 5′ UTR-FLuc control plasmid. Immunoblots performed on the 5′ UTR-FLuc fusion transfected cell extracts produced two bands migrating slightly slower than the FLuc protein control, consistent with initiation events at the uAUG1 and uAUG2. These N-terminally extended FLuc products were expressed at equal amounts (lanes 3 and 4). However, upon deletion of the SL the protein product derived from uAUG1 became nearly 2.5-fold more abundant than that derived from uAUG2 (lanes 5 and 6). This confirms that structural elements around the uAUG1 render it less accessible to the scanning ribosome. Note also that the total amount of FLuc translation products expressed in both fusion constructs was essentially identical; indicating that at least in the 5′ UTR background, ribosomes are very efficiently shunted past the upstream structural elements. Furthermore, the failure to detect a third band corresponding to initiation at the authentic AUG for FLuc in the 5′ UTR fusions suggests that the FLuc protein detected in the non-mutated wild-type background (lanes 1 and 2) is the product of a re-initiation event after translation of the uORF.
Generally, the rate-limiting step in translation initiation is the recruitment of the eIF4F complex to the 5′ cap because the eIF4E component is a limiting initiation factor in many cell types and is itself subject to regulation. Not all mRNAs compete equally for eIF4E, and this appears to be related in large part to the extent of RNA secondary structure within the 5′ UTR. Around 10% of mRNAs contain atypically long 5′ UTRs which frequently have the potential to form stable structures. These regions appear to serve as translational control elements in the expression of a number of proteins that are key players in the regulation of cell growth and differentiation (,,). They function as thermodynamic barriers to the scanning ribosome and are most effective when positioned close to the 5′ cap (). When 5′-proximal they probably also impede recruitment of the eIF4F cap-binding complex and hence the 40S ribosome. De-regulated expression of such mRNAs, due either to mutations within the 5′ UTR sequence itself or due to alterations in signalling pathways that impinge on the activity of the pre-initiation complex, can play a role in neoplastic transformation ().
ELK-1 is a key player in the integration of mitogenic and stress-mediated signalling pathways, and its mRNA contains many features that would indicate tight translational regulation. An additional layer of complexity arises due to the presence of a shorter, alternatively spliced form of the 5′ UTR. The relative ratio of the two forms exhibits both cell type and tissue variability. Alternative splicing occurs in ∼35% of human genes, but is most frequently observed in the 5′ UTRs (). Since such changes do not affect the coding potential they must play a role in regulating the read-out, an interpretation consistent with the key role of the 5′ UTR in ribosomal recruitment. Indeed, alternative splicing was reported to introduce a translational control element (a putative stem-loop) within the 5′ UTR of the human neuronal nitric-oxide sythase (nNOS) mRNA (). Likewise, the coupling of tissue-specific alternative splicing with the translational read-out was reported for the human dicer gene (), and the bovine growth hormone receptor (GHR) gene ().
One attractive function for alternative splicing within the 5′ UTR is the generation, or the regulation of, an IRES. IRESes permit recruitment of the pre-initiation complex independent of the 5′ cap and therefore require a more limited set of initiation factors (in particular eIF4E). They have been reported on mRNAs whose expression continues when the activity of the eIF4F complex is compromised (e.g. during conditions of cellular stress or stages of the cell cycle). Alternative splicing coupled with the regulation of IRES activity has been reported for the mRNA encoding the neurotrophin receptor TrkB (). However, although both the elk-1 mRNA 5′ UTR and 5′ UTR gave weak but positive read-outs in the bicistronic assay, this was readily attributable to cryptic promoter activity from the transfected DNA plasmids. It also explained the inability to detect ‘IRES-like’ activity when bicistronic RNAs were transfected into cells or translated in RRLs. Once again our results highlight the importance of using stringent controls when interpreting this type of assay, an observation already cited by others (,,).
Having eliminated IRES elements within the 5′ UTRs we turned our attention to the role of RNA structure and the organization of the uORFs as elements modulating expression. The major secondary structural feature, namely SL, is conserved in both 5′ UTRs. This is highly G/C rich (85%), and with a predicted stability of −37 kcal/mol could be a significant barrier to the 40S scanning ribosome. Indeed, live cell studies have recently demonstrated that the steepest drop in translation efficiency occurred when stem-loop stability increased from −25 to −35 kcal/mol (). The observation that the uAUG1 + 2 double mutant was much less effective at relieving repression in the 5′ UTR () suggested that structure was a more important feature in global translational down-regulation in this background. Such a scenario would also explain the delayed re-recruitment of the endogenous 5′ UTR elk-1 transcript onto polysomes in the presence of Rap (and hence limiting eIF4E). Nonetheless, the differences we observed between the two 5′ UTRs of the elk-1 mRNA was considerably more subtle than the >80-fold variation reported between the 5′ UTR spliced forms of the bovine gene (). This may arise because the uAUG1 is rendered invisible to the scanning ribosome due to shunting.
Both splice variants have retained the two uAUGs, and alignment shows that they are also conserved in the mouse elk-1 5′ UTR (data not shown). Not surprisingly, AUG codons are generally underrepresented in the 5′ UTR and their conservation across species is generally an indication of function (). However, a noticeable difference between the two elk-1 transcripts is that although both have retained the two uAUGs, only the 5′ UTR has two uORFs. Although multiple uORFs have been reported to synergistically repress translation (), our studies failed to demonstrate any major differences in the two elk-1 5′ UTRs. However, mutation of both AUG1 and AUG2 had a more important effect in the 5′ UTR background (reporter activities in the AUG1 + 2 double mutant were more than 2-fold higher in 5′ UTR than in 5′ UTR), and mutation of AUG2 had generally a more important effect than AUG1. This may simply reflect the better ‘Kozak context’ of AUG2 (..GGGAUGG…...) relative to AUG1 (..CGUAUGG..), which will consequently sequester more of the scanning ribosomes. However, structural probing studies confirm that AUG1 is positioned in the loop of the highly stable SL, a configuration that has been reported to facilitate ribosomal shunting, a mechanism that would render AUG1 less accessible to the pre-initiation complex (), and thereby reduce its activity as a repressor of downstream initiation (B). Consistent with such a model was our demonstration that AUG1 served as a much more efficient initiation codon in the ΔSL background, and that its repressive effect was also more marked when the structural elements were disrupted.
The observation of an altered distribution of the two elk-1 mRNA isoforms in different tissues and cell lines, coupled with differences in sensitivity to Rap, suggests a 5′ UTR-mediated regulation of protein expression. However, in our assays both 5′ UTRs were equally effective translational repressors. What then can be the basis of this regulation? Repression arises both as a consequence of RNA structure and uAUGs, recurrent features in the mRNAs that encode growth regulatory proteins (). However, the relevant contribution of each of these elements to the repression phenotype may not be the same in the 5′ UTR and 5′ UTR. The uAUGs appeared to be more effective negative elements in the latter since their removal had a more marked effect on the reporter read-out (3-fold increase as compared with ∼1.5-fold in 5′ UTR; ). This led us to postulate that RNA structure may be a more important feature in the 5′ UTR (assuming that repression is the sum total of the effect of uAUGs and global RNA structure). However, there is an important caveat in the interpretation of our assays in that they were performed under optimal growth conditions. This does not reflect the physiological context, in which alterations in initiation factor activity occur. These changes have important consequences for the translational read-out. For example, the activity of eIF2 which forms part of the eIF2-GTP-tRNA ternary complex associated with the scanning ribosome is tightly regulated by cell growth and stress signals. Alterations in the eIF2 activity are known to modulate translational re-initiation, an event that appears to play a role in ELK-1 expression based upon the UGA mutation () (). Re-initiation-mediated ELK-1 expression is probably coupled with ribosomal starts at the upstream AUG2. This small ORF, comprising only two codons, is conserved in many mammalian elk-1 mRNAs (e.g. human, mouse, rat, cow and dog). Additionally, the relevant amounts and activity of different key components of the initiation apparatus, in particular the eIF4 family (eIF4E, eIF4G, eIF4A, eIF4B), vary in a tissue-specific manner (). Variations in the amount/activity of eIF4E (the cap-binding protein) and eIF4A/4B (the ATP-dependent helicase and its cofactor) will impact directly on the expression of mRNAs with structured 5′ UTRs, and will therefore have consequences for the translational expression of key cellular proteins. This in turn may impose variations in the 5′ UTR of genes whose expression must be tightly controlled in all environments, a scenario that could explain the variation in the ratio of the two spliced forms observed. For example, when eIF4E activity/amount is low, ELK-1 expression would arise mainly from the 5′ UTR transcript since in this context the uAUGs rather than RNA structure are the major regulatory elements. Such a scenario would explain the preferential re-recruitment of the 5′ UTR transcript onto polysomes in the presence of Rap, and hence reduced eIF4E activity. It is therefore not inconceivable that under certain physiological conditions, the differences in the activity associated with the two spliced variants may be more apparent than observed in our assays. An additional order of complexity arises from the recent observation that other members of the DEAD-box helicase family can play a role in the selective translation of cellular mRNAs () (B, lower panel). Many of these helicases have RNA chaperone activities and are key components of the splicing and nuclear export machinery. They provide a tantalizing explanation for how nuclear events, such as alternative splicing, can impact profoundly on the translational read-out (), and may also impact on the altered polysomal recruitment of the two elk-1 transcripts. Finally, the cell lines we employed, namely 293T, are tumoural. However, most (if not all) tumoural cell lines have major perturbations in the pathways that regulate protein synthesis (a characteristic that contributes to the tumoural phenotype) (), a background that may also serve to mask the functional differences between the two spliced variants.
An earlier study on ELK-1 expression performed without the 5′ UTRs identified a region within the first 279 nt of the ORF that was responsible for translational down-regulation upon AKT activation (). This region was not included in our studies. Since the AKT pathway directly impinges on eIF4E activity, via mTOR, it remains possible that this element and the 5′ UTRs may interplay in regulating ELK-1 expression (). Additionally, a shorter form of ELK-1 generated from an internal AUG initiation codon was reported (sELK-1) (). How the ribosome accesses this site (it is the seventh AUG codon on the mRNA), and the origin of its tissue specificity (it is found mainly in neuronal cells), remains unclear. The internal 279 nt AKT-responsive site did not appear to play a role in its expression. Nonetheless, we are currently investigating if this ORF element in the context of the different 5′ UTRs plays a role in modulating both ELK-1 and sELK-1 expression in response to growth/differentiation signals.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Base or nucleotide flipping is the displacement of a base in regular B-DNA from the helix into an extrahelical position. First observed by X-ray crystallography for the bacterial C5-cytosine methyltransferases M.HhaI () and M.HaeIII (), nucleotide flipping (base extrusion) has been documented later for other methyltransferases (), glycosylases (), glycosyltransferases (,) and various DNA repair enzymes (). Some enzymes, e.g. the methyltransferases, flip a nucleotide of only one DNA strand (). Others, like endonuclease IV, alter the backbone conformations of both strands flipping the deoxyribose and nucleotide at an abasic site (). Either way, nucleotide flips occur because enzymes need access to a DNA base to perform chemistry. For example, DNA methyltransferases transfer the methyl group to the extruded base, while glycosylases involved in DNA repair excise the extrahelical base (). Typically, an amino acid side chain is intercalated into the DNA to fill in the ‘hole’ introduced after the base flipping event (,,).
Nucleotide flipping in the co-crystals of restriction endonuclease Ecl18kI with cognate DNA came as a surprise (). In a functional sense, Ecl18kI is a ‘standard’ Type II restriction endonuclease (REase): it recognizes pentanucleotide sequence CCNGG and cuts phosphodiester bonds on the 5′ sides of the outer cytosines to generate 5 nt 5′-overhangs (). Although the endonuclease does not subject the central bases to any kind of modification, in the crystal structure these bases were clearly extrahelical and accommodated in pockets on Ecl18kI made by the side chain atoms of Arg57 on one face and the indole ring of Trp61 on the other face (). Unlike in other complexes with flipped nucleotides, there was no ‘hole’ in the DNA and no amino acid intercalation. Instead, the DNA was compressed, so that the base pairs adjacent to the flipped nucleotides stacked directly against each other. The resulting DNA compression reduced the length of the interrupted 5 bp stretch CCNGG to the length of a 4 bp stretch CCGG and made the distance between the scissile phosphates in the Ecl18kI–DNA complex equal to the distance between the scissile phosphates in the NgoMIV complex with a continuous sequence GCCGGC (,). Therefore, we suggested that Ecl18kI uses base flipping to adapt the conserved sequence readout machinery for the interrupted target site and predicted that the evolutionary related REases EcoRII and PspGI that cut the related CCWGG sequence before the first C, might also flip nucleotides ().
Nucleotide flipping in solution by Ecl18kI, PspGI and EcoRII remains to be established. So far, it is only supported by the observation that PspGI accelerates deamination of the central cytosine in the incorrect CCCGG sequence, which differs from the canonical sequence at the center (). 2-Aminopurine (2-AP) has often been used as a fluorescence probe to detect base flipping in solution (). The 2-AP fluorescence is highly quenched in polynucleotides due to the stacking interactions with neighboring bases () and therefore increases strongly when the base is flipped out of the DNA helix (,). Here, we use 2-AP as a fluorescence probe for base flipping and provide the first direct evidence in solution that Ecl18kI, the C-terminal domain of EcoRII (EcoRII-C) and PspGI extrude the central base pair while interacting with their recognition sites.
2-AP containing oligodeoxynucleotides were obtained from Integrated DNA Technologies (HPLC grade, Coralville, USA), non-modified oligodeoxynucleotides were from Metabion (HPLC grade, Martinsried, Germany). In order to assemble oligoduplexes, appropriate oligodeoxynucleotides () containing 2-AP or non-fluorescent control strands were mixed with a 1.05-fold molecular excess of complementary strands in the reaction buffer A (33 mM Tris-acetate, pH 7.9 at 25°C, 66 mM potassium acetate), heated to 85°C and allowed to cool slowly over several hours to room temperature. For the DNA binding and cleavage studies one strand of the 25 bp duplexes was 5′-end labeled with [γ-P]ATP (Hartmann Analytic, Braunschweig, Germany) using a DNA labeling kit (Fermentas, Vilnius, Lithuania).
The wt Ecl18kI, Ecl18kI mutant W61A, EcoRII-C, PspGI and MvaI proteins were purified and concentrations were determined by measuring absorbance at 280 nm as described in (,,,). All protein concentrations are indicated in terms of the dimer, except for MvaI, which is a monomer in solution ().
The W61A mutant of Ecl18kI was obtained by the modified QuickChange Mutagenesis Protocol (). Plasmid pET21b(+)_R.Ecl18kI [Ap] () was amplified by PCR using DNA polymerase (Fermentas, Vilnius, Lithuania) and two complementary (partially overlapping) primers obtained from Metabion (desalt grade, Martinsried, Germany) containing the desired mutation. After PCR the methylated parental (non-mutated) plasmid was digested with DpnI (Fermentas, Vilnius, Lithuania). BL21(DE3) cells carrying the plasmids pVH1 [Kn] (with lacI) and pHSG415ts [Cm] bearing the gene () were transformed with the PCR product by the CaCl method. Plasmid DNA was isolated by the alkaline lysis procedure and purified using the GeneJET™ Plasmid Miniprep Kit (Fermentas, Vilnius, Lithuania). Sequencing of the entire gene of the mutant confirmed that only the designed mutation had been introduced.
Gel shift analysis of DNA binding by wt proteins and Ecl18kI W61A mutant protein was performed by titrating P-labeled 25 bp oligoduplex (see ) at 0.1 nM concentration with increasing amounts of protein. values were evaluated as described elsewhere ().
The DNA cleavage activities of wt Ecl18kI, Ecl18kI mutant W61A, EcoRII-C and PspGI were monitored using a 25 bp oligoduplex containing a P-label either in the top or the bottom DNA strand (). Cleavage rates of both strands were evaluated separately. Ecl18kI cleavage reactions were conducted at 20°C in the reaction buffer A, containing 10 mM MgCl and 0.1 mg/ml BSA using 200 nM of oligoduplex and 300 nM of protein. EcoRII-C and PspGI cleavage reactions were performed in the same buffer A at 25°C using 200 nM of oligoduplex and 1000 nM of protein. Aliquots were removed at timed intervals and quenched by mixing with loading dye [95% (v/v) formamide, 0.01% (w/v) bromphenol blue, 25 mM EDTA] before denaturing gel electrophoresis. The samples were analyzed and quantified as described in ().
All fluorescence measurements were acquired in photon counting mode on a Fluoromax-3 (Jobin Yvon, Stanmore, UK) spectrofluorometer equipped with Xe lamp. Sample temperatures were maintained at 25°C by a circulating water bath. Oligoduplexes I or II were used as 2-AP labeled DNA (). Emission spectra (340–420 nm) were recorded at an excitation wavelength λ = 320 nm with excitation and emission bandwidths of 2 and 8 nm, respectively. At least two scans were averaged for each spectrum. Sample emission spectra were collected in reaction buffer A in the presence and absence of calcium acetate on 250 nM DNA alone or 250 nM DNA mixed with a 5-fold excess of the protein to ensure saturation of the fluorescence signal. Control spectra used for the background subtraction corrections were collected under identical conditions except that oligoduplex III was used instead of the fluorescent DNA. The fluorescence emission value of the corrected spectrum was determined at the emission maximum (see Supplementary Table S1) for each sample. For the oligoduplex titration experiment, emission spectra of the 250 nM oligoduplex I with protein in a 0–2000 nM range were collected.
We have used the 2-AP fluorescence assay to monitor base flipping in the Ecl18kI–DNA complex in solution. A number of 25 nt oligoduplexes that contain the fluorescent base analog 2-AP at different positions were designed (). In the oligoduplex I, 2-AP was incorporated within the CCNGG sequence instead of A in the central base position. In the oligoduplex II, 2-AP was introduced immediately adjacent to the target site. Like most restriction enzymes, Ecl18kI requires Mg ions for DNA cleavage. In the absence of divalent cations, it forms a rather weak complex with cognate DNA (Supplementary Table S2). Addition of Ca ions that do not support cleavage significantly increased Ecl18kI–DNA complex stability (). Gel shift experiments revealed that 2-AP incorporation into the target sequence had no effect on the affinity of Ecl18kI for cognate DNA in the presence of Ca ions (). In the buffer supplemented with Mg ions, Ecl18kI cleaved 2-AP containing and lacking oligoduplexes at identical rates (data not shown).
We titrated the 2-AP containing oligonucleotides with Ecl18kI in the binding buffer supplemented with Ca ions and monitored the change of the 2-AP fluorescence intensity at 367 nm (A, Supplementary Table S1). The free oligoduplex containing 2-AP at the central position showed low signal because the fluorescence was quenched due to base stacking interactions (B). When Ecl18kI bound at saturating concentrations to oligoduplex I, which contains 2-AP in the central position, the fluorescence intensity increased 6.5-fold. In contrast, only small changes were observed with oligoduplex II, which carries 2-AP outside of the target site (B and C). The change of fluorescence intensity for the oligoduplex I suggests that the 2-AP stacking with DNA bases is disrupted. It is compatible with nucleotide flipping, which has been shown to enhance 2-AP fluorescence to varying extents in different systems ().
Gel shift analysis showed highly decreased binding of Ecl18kI to cognate DNA in the absence of Ca ions (see, Supplementary Figure S1). However, we found that at much higher enzyme and DNA concentrations used in the fluorescence titration experiments, Ecl18kI formed a binary complex with cognate DNA in the absence of Ca ions (D). The value obtained from the titration data was 52 ± 12 nM. In the Ca-free buffer, Ecl18kI binding to the oligoduplex I containing the 2-AP in the central position increased the fluorescence intensity ∼28.5-fold at saturating protein concentrations, while only small changes were observed with oligoduplex II (E and F). The 2-AP signal in the buffer without Ca ions was ∼4 times higher than the signal in the buffer supplemented with Ca (D). These results suggest that the structure of the complex formed in the presence of Ca ions may differ from that formed without Ca.
2-AP fluorescence is often quenched in the hydrophobic environment of a protein (,,,). In the crystallographic complex of Ecl18kI with DNA, the flipped nucleotides are accommodated in pockets that are lined by tryptophan Trp61. In order to test whether Trp61 quenches 2-AP fluorescence, we replaced this residue with alanine. The Ecl18kI W61A variant did not cleave cognate oligoduplex III or the 2-AP containing oligoduplex I radioactively labeled at either strand, but it retained the ability to bind both oligoduplexes albeit at ∼10-fold decreased affinity according to the gel shift assay (see, Supplementary Table S2). Binding of Ecl18kI W61A to oligoduplex I in the presence of Ca ions increased the 2-AP fluorescence intensity ∼125-fold () suggesting that the mutant was able to flip out the central nucleotide. 2-AP fluorescence in the ternary W61A–DNA–Ca complex was ∼20 times higher than in the wt Ec18kI–DNA–Ca complex (). Thus, the W61A mutant data support the assumption that low 2-AP fluorescence in the ternary complex with the wt protein is due to the quenching of the extruded base by stacking interactions with the Trp61 residue. However, one cannot exclude that increased space in the binding pocket of the W61A mutant may allow a different orientation of the extrahelical 2-AP and affect the fluorescence intensity.
The C-terminal domain of the EcoRII restriction enzyme and the PspGI restriction enzyme are specific for the CCWGG sequence (where W stands for A or T) and cleave it before the first C. It was suggested that Ecl18kI and EcoRII-C/PspGI may be evolutionarily related (). This raises the intriguing question of whether EcoRII-C and PspGI also flip the central W nucleotides while interacting with their target sites.
The EcoRII structure, which was solved in the absence of DNA () shows a very similar fold to Ecl18kI () except that it has an extra N-terminal regulatory domain (). Structural comparison between Ecl18kI and the C-terminal domain of EcoRII reveals that Arg57 and Trp61, which sandwich the flipped bases in the Ecl18kI–DNA cocrystal structure, spatially coincide with the Arg222 and Tyr226 of EcoRII suggesting that EcoRII may flip the central base similarly to Ecl18kI (). Therefore, we analyzed base flipping by EcoRII-C in solution using the 2-AP fluorescence assay. EcoRII-C turned out to bind to the oligoduplex I containing the 2-AP in the central position (see, Supplementary Table S2) and binding was accompanied by ∼12-fold increase of fluorescence (), suggesting that the base is extruded from the double helix.
The structure of the PspGI enzyme which recognizes the same CCWGG sequence as EcoRII but lacks the extra N-terminal domain is not yet known, but modeling studies () suggest significant similarities to Ecl18kI (). Moreover, genetic studies support the PspGI model and provide indirect evidence that PspGI may flip central nucleotides within a sequence that matches its target site except for the presence of a G-C pair instead of the A-T pair at the center (). We found that 2-AP at the central position of the recognition site does not change PspGI binding affinity (see, Supplementary Table S2). PspGI titration of 2-AP-containing oligonucleotide I showed an increase of 2-AP fluorescence and resulted in a 64-fold increase of the signal at saturation in comparison to the free oligoduplex (). Thus, according to the 2-AP assay, PspGI should flip central nucleotides while interacting with its recognition site like Ecl18kI and EcoRII-C.
Recently, we have solved the crystal structure of MvaI restriction enzyme that recognizes the CC/WGG sequence identical to that recognized by EcoRII and PspGI but cleaves it before the W nucleotide as indicated by the ‘/’ (). In the MvaI–DNA complex structure, the DNA conformation does not deviate essentially from the canonical B-form and there is no evidence for base flipping. Binding studies in solution revealed that MvaI binds 2-AP-containing oligonucleotide I (see, Supplementary Table S2), however, this did not lead to an increase of 2-AP fluorescence ().
z
y
m
e
s
t
y
p
i
c
a
l
l
y
f
l
i
p
n
u
c
l
e
o
t
i
d
e
s
t
o
g
a
i
n
a
c
c
e
s
s
t
o
o
t
h
e
r
w
i
s
e
p
o
o
r
l
y
a
c
c
e
s
s
i
b
l
e
b
a
s
e
s
.
B
a
s
e
d
o
n
c
r
y
s
t
a
l
l
o
g
r
a
p
h
i
c
i
n
f
o
r
m
a
t
i
o
n
a
n
d
s
e
q
u
e
n
c
e
a
n
a
l
y
s
i
s
,
w
e
h
a
v
e
s
u
g
g
e
s
t
e
d
t
h
a
t
t
h
e
r
e
s
t
r
i
c
t
i
o
n
e
n
d
o
n
u
c
l
e
a
s
e
s
E
c
l
1
8
k
I
,
E
c
o
R
I
I
a
n
d
P
s
p
G
I
e
m
p
l
o
y
b
a
s
e
f
l
i
p
p
i
n
g
i
n
a
n
o
v
e
l
w
a
y
t
o
a
c
h
i
e
v
e
s
p
e
c
i
f
i
c
i
t
y
f
o
r
t
h
e
i
r
t
a
r
g
e
t
s
a
n
d
t
o
a
d
j
u
s
t
t
h
e
i
r
c
l
e
a
v
a
g
e
p
a
t
t
e
r
n
s
.
#text
S
u
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
The transcription factor AraR controls the utilization of carbohydrates in . The control exerted by AraR is modulated by the presence of the effector molecule arabinose leading to induction of expression of at least 13 genes, comprising the arabinose () regulon, which includes the gene (). The products of these genes ( and ) include extracellular and intracellular catabolic enzymes involved in the degradation of arabinose, galactose and xylose containing polysaccharides, uptake of these sugars into the cell and further catabolism of -arabinose and arabinose oligomers (,).
A key property of AraR is its ability to bind specific DNA sequences in the absence of the inducer -arabinose, as determined by DNAse I footprinting analysis (,,). AraR recognizes and binds at least eight palindromic operator sequences, located in the five known arabinose-inducible promoters. Three of these promoters contain two boxes: the promoter of the (boxes OR and OR), of (OR and OR) and of (OR and OR). In the cases of the genes and , a single box is present (OR and OR). AraR binding to the promoters displaying two boxes is cooperative, requiring in phase and properly spaced operators, and involves the formation of a small loop in the DNA. These two mechanistically diverse modes of action of AraR result in distinct levels of transcriptional regulation, as cooperative binding to two boxes results in a high level of repression while interaction with a single operator allows a more flexible control (,,).
AraR is a 362 amino acid homodimeric protein that shows a chimeric organization, consisting of two functional domains with different phylogenetic origins (,,): a small N-terminal DNA-binding domain (DBD) comprising a winged helix–turn–helix (HTH) motif belonging to the GntR family of transcriptional regulators () and a larger C-terminal domain homologous to that of the GalR/LacI family of bacterial regulators and sugar-binding proteins (). AraR typifies one of the six GntR-subfamilies of proteins (,). Currently, there are 54 members of this rapidly growing class of proteins, which can be found in prokaryotes [CDART database; ()].
Previously, a model for AraR was derived using comparative modelling based on crystal structures of FadR (DBD) and PurR (COOH domain) from (). We have used random and site-directed mutagenesis to map the functional domains of AraR required for DNA binding, dimerization and effector binding. The arabinose-binding pocket is composed of polar and charged residues, whereas the dimerization interface has a hydrophobic nature. In both cases, the residues are distributed along the primary sequence of the C-terminal domain (). Based on crystallographic studies of structurally and functionally related proteins, binding of the effector to the COOH region in AraR is predicted to elicit a conformational change in the N-terminal region, leading to inhibition of binding to operator sequences, and allowing transcription from the arabinose-responsive promoters. This allosteric signal involves a switching mechanism for communicating structural changes triggered in the sensor domain to the regulatory domain, decreasing the affinity of the latter for DNA.
Winged helix motifs are functionally and mechanistically versatile (). They are primarily involved in DNA binding, but cases have been reported in which they participate in protein–protein interactions. Monomeric, homo- or heterodimeric protein–DNA complexes have been characterized and revealed quite distinct modes of binding to DNA, which can involve interactions between the recognition helix and the wing with the major and minor groove (). Although the level of amino acid identity for the DBD of all members of the GntR superfamily is low (∼25%) they share this conserved structural topology (). Global analysis of the conservation of amino acid sequences in DNA-binding proteins concluded that residues interacting with the DNA backbone establish a set of core contacts that provide stability for homologous protein–DNA complexes, and consequently are well conserved across all protein families. On the other hand, residues that interact with DNA bases have more variable levels of conservation (). Previous mutagenic studies showed that AraR residues in the N-terminal region were required for DNA binding because mutations in these residues abolished its regulatory function (). However, the precise contribution of the mutated amino acids to DNA-binding activity was unclear.
To understand the specific properties of the interaction AraR-operator sequences, we substituted amino acids, in or near the HTH motif, which according to the model were predicted to contact DNA. We determined the effects of these substitutions on the ability of AraR to function and on the DNA-binding affinities . Conversely, mutational analysis of the AraR-binding sites was used to determine the base-specific requirements for transcriptional regulation and DNA binding . These experiments gave both expected and unexpected results, which together showed that specific AraR residues and operator bases are crucial to achieve a high level of regulatory activity, while others display variable contributions to DNA binding. In addition, an AraR mutant was isolated, which partially suppresses the loss of regulation observed in certain mutated DNA operators.
strains used in this work () were grown in Luria–Bertani (LB) medium () or C minimal medium () and solid sugar-free agar (SFA) medium (LabM) or LB broth solidified with 1.6% agar. Chloramphenicol (5 μg ml), kanamycin (10 μg ml) and erythromycin (1 μg ml) were added when appropriate. The Amy phenotype was tested by detection of starch hydrolysis on tryptose blood agar base medium (Difco) plates, containing 1% of potato starch, with a I–KI solution as described previously (). DH5α (Gibco BRL) or XL1-blue were used for routine molecular cloning work and BL21 DE3 pLysS () for overexpression of mutant AraR proteins. strains were grown on LB medium, with ampicillin (100 μg ml), chloramphenicol (20 μg ml), kanamycin (30 μg ml) and IPTG (isopropyl-β--thiogalactopyranoside) (1 mM) added as appropriate. The and cells were transformed as described previously ().
DNA manipulations were carried out as described previously (). Restriction enzymes were purchased from MBI Fermentas, New England Biolabs or Roche, and used according to manufacturer's instructions. DNA was eluted from agarose gels using the GeneCleanII kit (Bio101) or the GFX DNA purification kit (GE Healthcare). PCRs were performed in a GeneAmp PCR system 2400 (Perkin-Elmer) and PCR products purified using QIAquick PCR purification kit (QIAgen). DNA was sequenced using an ABI PRIS BigDye terminator ready reaction cycle sequencing kit (Applied Biosystems).
Amino acid substitutions in AraR were made by the QuikChange (Stratagene) site-directed method using as template plasmid pLS30 () and mutagenic oligonucleotides carrying the modified codon in the centre (listed in Supplementary Table 1). For R41A and H42A, a 486-bp region containing the mutations (BglII–MluI fragment) was subcloned in pLS30, originating plasmids pSC16 and pIF41, respectively. The linearized plasmids were used to transform strain IQB350, leading to the integration of the mutant alleles into the chromosome at the locus via double recombination (Strains are listed in ). Substitutions K4A, Y5F, E30A and Q61A were generated by site-directed mutagenesis of pLS30, the amplified products were ligated, linearized and used to transform (). Substitution R45L was obtained by chance when attempting to create mutation R45A () using an identical procedure. Presence of the mutations was verified by sequencing the allele in the resulting plasmids or strains.
Plasmid pLM51 () a pBluescript II KS (+) derivative carrying the wild-type promoter, was used as template for generating single-nucleotide substitutions in OR or OR, using the QuikChange (Stratagene) site-directed method and pairs of mutagenic oligonucleotides (listed in Supplementary Table 1). Resulting plasmids contained the following mutations in OR: A→C (pLM61), G→T (pLM62), T→G (pLM58), C→A (pLM63), G→T [pLM54; ()], T→G (pLM59), A→C (pLM57), T→G (pLM60); or G→T in OR [pLM77; ()]. The 204-bp BamHI–EcoRI DNA fragment from these plasmids, containing the mutagenized operator region, was then subcloned in the same sites of pSN32 (). This procedure generated respectively pLM68, pLM69, pLM65, pLM70, pLM56 (), pLM66, pLM64, pLM67 and pLM78 (), which bear transcriptional fusions of the promoter with single-point mutations to After linearization, plasmids were used to transform IQB215, giving rise to strains where these fusions were integrated at the locus (). To analyse the repression exerted by AraR on the boxes, integration of the allele at the locus was accomplished by transformation with pIF76.
The insertion of a 1446-bp RI–HI fragment from pLS30 (), containing the allele, into pDG1664 (EcoRI–BamHI) () yielded plasmid pIF76. This plasmid was used as DNA template in random PCR mutagenesis, according to the method described by Leung . (). Random PCR mutagenesis with oligos ARA6 and ARA73 amplified a 650-bp 5′-end region of the allele. After digestion with BamHI–Eco47III, the fragment was subcloned in the same plasmid leading to the replacement of the equivalent region of the wild-type allele, from sites −227 to +251 relative to the transcription start site (containing the promoter and the first 75 codons of the gene). The recombinant plasmids were transformed into DH5α yielding a library of mutations (contained in ∼2200 transformants). A plasmid pool was used in separate experiments to transform strains with a Δ Δ background, in which the OR operator sequence carried the mutations described above (), leading to integration of AraR mutants at the locus via double crossing-over. The constitutive expression of due to the presence of the mutated boxes leads to a Lac phenotype in the receptor strains, reflected by a blue colour in SFA medium with X-gal. To isolate mutant alleles suppressing the deleterious effect of the operator mutations, we screened for colonies displaying a weaker Lac phenotype (white/light blue phenotype in the same medium). Chromosomal DNA from these colonies was used as template to amplify the mutagenized region of the allele, which was subsequently cloned back into pIF76 as described above and sequenced. The resulting plasmid, pIF85, bears a mutation leading to a single amino acid substitution, M34T.
strains were grown in C minimal medium supplemented with 1% (w/v) casein hydrolysate in the presence and in the absence of -arabinose 0.4% (w/v) as previously reported (). Samples of cell culture were collected and analysed 2 h after the addition of -arabinose. The ratio of β-galactosidase activity, determined as described () from cultures grown for 2 h in the presence and absence of inducer was taken as a measure of AraR repression in the analysed strains (Repression Index).
strains were grown as for β-galactosidase assays. Preparation of cell extracts and immunoblotting were performed as described (). Blots were developed with anti-AraR-MBP2* serum () using the ECL detection system (Amersham Biosciences). Protein concentration was determined using a Bio-Rad kit.
Fusion of the C-terminus of AraR variants to six histidines in the plasmid pET30a(+) (Novagen) was engineered, placing the genes under the control of a T7 promoter. The construction of plasmid pLS16, carrying the wild-type allele, was described previously (). Construction of plasmids carrying the AraR substitutions F37S, Q61R and L33S was accomplished in a similar manner. Briefly, the alleles containing these mutations were amplified with oligos ARA50 and ARA51 from the pLS30-derivatives pIF1, pIF2 and pIF3 obtained previously (). The 1112-bp PCR product was separately digested with AvaI–NdeI and AvaI–HindIII, the resulting 282-bp and a 805-bp fragments were inserted in pET30a(+) restricted with NdeI and HindIII, yielding plasmids pIF5, pIF6 and pIF7, respectively. To introduce mutations S53P, H42A or M34T, regions BglII–KpnI were obtained from plasmids pIF17, pIF41 and pIF85 described above, and used to substitute the same region in pIF7 generating pIF111, pIF123 and pIF121. For K4A, Y5F, E30A, R45A and Q61A chromosomal DNA from strains IQB712, IQB571, IQB568, IQB563 and IQB564 was used (). PCR products were digested with appropriate enzymes (NdeI–KpnI, BglII–KpnI or BglII–HindIII) and used to substitute the corresponding region in pIF7. These procedures yielded pIF124, pIF112, pIF78, pIF74 and pIF75, respectively. The presence of the mutations was verified by sequencing the alleles. For the purification of these AraR-his6 variants, BL21 (λDE3) pLysS () cells transformed with the corresponding pET30 derivatives were grown at 37°C to an optical density at 600 nm of 0.6 in 1 l of LB medium, and then expression of the fusion proteins was induced by addition of IPTG to 1 mM. Incubation in the same conditions continued for additional 2 h. All subsequent steps were carried out similarly to the method described previously ().
A DNA fragment carrying the OR–OR region was amplified from pLM51 using primers ARA262 and ARA263 (Supplementary Table 1). After purification, the 126-bp PCR product was labelled with T4 Polynucleotide Kinase (MBI Fermentas) and [γ-32P]dATP, followed by extraction with phenol/chloroform and precipitation with ethanol. Binding reactions contained 12 mM HEPES-KOH pH 7.6, 10 mM MgCl, 0.5% [w/v] BSA, 1 mM DTE, 10% Glycerol (v/v), 200 mM NaCl, 4 mM NaHPO4, 4 mM NaHPO4, 0.4 mM EDTA, a 200-fold molar excess of competitor DNA (polydIdC), 1 nM of labelled DNA and increasing concentrations of wild-type or mutant AraR proteins. After incubation for 30 min at room temperature, the mixture was loaded onto a pre-run 8% polyacrylamide gel in 25 mM Tris 200 mM Glycine (pH 8.9) and run at 100 V for ∼1 h. Gels were dried under vacuum and exposed to a Phosphorimager screen before analysis with a Molecular Dynamics Storm 860 Imager and ImageQuant version 5.0. To determine the dissociation constants, protein concentrations were used according to previous results () and values were obtained using the GraphPad Prism software. For competition DNA-binding experiments, various amounts of cold double-stranded oligonucleotides containing single mutations in the operator sequences (Supplementary Table 1) were added to the reaction in the presence of 40 nM AraR. As controls, we used oligonucleotides carrying the wild-type operator (ARA288 and ARA289) or a non-specific DNA sequence with the same length (ARA244 and ARA245). The following procedures were made as described above. The percentage inhibition in the presence of competitor DNA was determined similarly to the method described by Bera . (). The radioactivity of bound DNA was quantified in the control without competitor, and in samples containing 500-fold molar excess of the distinct competitors. Inhibition (%) = 100 × [(bound)− (bound)/(bound)].
We have previously established that AraR interaction with DNA is achieved via a 70 amino acid N-terminal domain. These results were obtained by random and site-directed mutagenesis based on a 3D model of the DBD derived from the crystal structure of the regulator FadR (). However, many of these mutations resulted in changes that could alter the protein structure or interfere with DNA binding. In order to conduct a more clarifying characterization of the role of specific AraR residues on its DNA-binding activity, we made several amino acid substitutions in or near the HTH motif. Residues were chosen based on the 3D model and/or primary sequence alignment of AraR-like proteins (). The majority of the substituted amino acids was predicted to contact directly the bases of the DNA and consequently, would account for the specificity of the interaction with operator sequences. Positions K4, E30, R41, H42 and Q61 were exchanged to alanine and Y5 to phenylalanine (A and B). The substitutions were designed to minimize local structure disruption and probe loss of contact with the DNA. In addition, in this work we analysed a mutation R45L generated by chance during the construction of mutant R45A, which was characterized in a previous work (). Plasmid pLS30, carrying an wild-type allele, and its derivatives harbouring the mutated alleles, obtained after site-directed mutagenesis, were integrated as single copy at the locus of receptor strain IQB350. This strain bears an null mutation (Δ) and a transcriptional – fusion (see Materials and Methods section) and therefore expresses constitutively β-galactosidase. In the resulting strains (), the levels of β-galactosidase expressed from the operon- fusion reflect the regulatory activity of the AraR variant encoded by the allele integrated at the locus.
The effect of each substitution was analysed by determining the levels of accumulated β-galactosidase in strains grown under inducing (presence of arabinose) and non-inducing conditions (absence of arabinose). The results are summarized in A. The regulatory activity was quantified as the repression index (A). The receptor strain IQB350 Δ-) and IQB352, a derivative carrying the wild-type allele at the locus, were used as controls, yielding repression index values of 1 and 99, respectively, which correspond to the absence and maximal regulation exerted by the protein. Mutation R41A had no effect on the regulatory activity when compared to the wild type. Therefore, although amino acid R41 is conserved among all members of the GntR family (A) and establishes interactions with bases in the major groove according to FadR-DNA data [R45 in FadR; ()], it is dispensable for AraR binding . Variant H42A displayed a minimal decrease on repression activity (1.6-fold). Moderate effects were observed with Y5F and E30A, 3.3- and 4.8-fold, respectively. The more drastic effects were seen with K4A, a 30-fold decrease in repression, and with R45L and Q61A the regulatory activity was completely abolished. The lack of regulation of R45L is identical to that observed with R45A (). Together, the results suggest that these three residues play the most important roles in DNA binding of the ones analysed here.
Because the observed decrease in repression could be the result of deficient accumulation of the mutant proteins, as a consequence of lower stability and proteolysis, we measured the abundance of each AraR variant. The strains were grown as for the β-galactosidase assays and the level of AraR estimated by western immunobloting using equivalent amounts of their soluble cell extracts (B). The cellular level of all mutant proteins was comparable to that seen with wild-type AraR, ruling out the possibility of deregulation originated by degradation of the repressor.
The apparent affinity constants () of AraR mutants for operator sequences were determined by EMSAs using a P-labelled 126-bp DNA fragment, which carried both operators of the metabolic operon, OR and OR (depicted in ). Binding of AraR to this DNA fragment was specific, as the presence of the inducer arabinose but not xylose prevented the formation of the protein–DNA complex (A). Titration of 1 nM of DNA with increasing concentrations of wild-type repressor (B) allowed the determination of an apparent 3.9 × 10 M, which is defined as the amount of protein necessary to shift 50% of the labelled probe (). This value is comparable to that previously calculated for each individual box using DNase I quantitative footprinting experiments, 3.4 × 10 M and 4.7 × 10 M for OR and OR, respectively ().
All the mutant proteins that displayed an effect were overexpressed in and purified to homogeneity (see Materials and Methods section). Binding to DNA was assayed by EMSA and their respective apparent affinity constants determined (C). Variant H42A, which showed the minimal loss of repression (2-fold) bound DNA with an apparent of 4.1 × 10 M, similar to the wild-type protein. The most severe effects were displayed by Q61A ( ∼ 5.5 × 10 M), K4A and R45A, both showing an apparent > 1.5 × 10 M. These three mutants were also unable to perform a regulatory activity . Residues K4 and H42 are completely conserved among AraRs () and in a contact distance of the DNA according to the model (A and B), however, mutation of these residues had different outcomes. Since these two residues are not conserved among the members of the entire GntR family they may contribute in very different extents to the DNA-binding specificity of AraR-like proteins.
R45 is a conserved amino acid in the GntR family members. In the regulator FadR from , the substitution of the corresponding residue (R49) has also a drastic effect (). Moreover, the crystal structure of the FadR–DNA complex (,) shows that R49 locates in the recognition helix of the winged HTH and interacts with a phosphate group, not specifically a base. According to the predictions of the tertiary structure of AraR, R45 is also located in the recognition helix (B), which is generally more responsible for the interaction with DNA, in particular the positively charged residues. Q61 belongs to the predicted wing of the DNA-binding motif (A and B). The corresponding residue in both FadR and in GntR from is also positively charged (A) and substitutions led to loss of DNA-binding ability (,). In FadR, H65 is part of the wing and makes specific contacts with an adenine (,).
Intermediate decreases of DNA binding were observed with Y5F and E30A, with 2.0 × 10 M and 2.3 × 10 M, respectively, similarly to that seen . Therefore, these exchanges led to a comparable effect both and ( and ). Moreover, the nature of the mutation Y5F revealed the importance of the OH group in the interaction with DNA. Both residues are conserved in the GntR-family proteins, and the corresponding residues in FadR, A9 and E34, were shown to contact the DNA backbone (,). The latter also contacts nearby amino acids, contributing presumably to the stabilization of residues that interact specifically with the DNA bases.
Additionally, four other AraR mutants, L33S, F37S, S53P and Q61R, obtained by random mutagenesis and characterized in a previous work [(); B), were studied by EMSA. Both substitutions F37S and S53P had led to derepression , of ∼24- and 2-fold, and in mutant L33S the regulatory activity was almost completely abolished (). These values could be explained by the observed instability of the proteins (). However, purified mutants F37S and S53P showed decreased DNA-binding affinities, 7.4 × 10 M and 6.5 × 10 M, respectively, that may also contribute to the deregulation . These results could be explained by the nature and localization of these substitutions, which suggest implications in the folding of the DBD. Overexpression and purification of L33S yielded only small amounts of protein. Nevertheless, at the maximal concentration that we could use in EMSA assays, 50 nM of the mutant, no DNA binding was observed (data not shown).
Interestingly, while the Q61A substitution completely abolished regulation and DNA binding , the change to arginine in the same position showed only a 1.6-fold decrease of regulatory activity (), and the affinity to the DNA probe 3.1 × 10 M, was even slightly higher than that displayed by wild-type AraR. Noteworthy, in GntR the inverse of AraR mutation Q61R (i.e. GntR R75Q) leads to a significant loss of regulation (). Based on these observations, we may speculate that the rise of positive charge as a result of AraR substitution Q61R increased the overall (non-specific) affinity for DNA, leading to a titration of the protein.
In summary, K4, R45 and Q61, were the most critical AraR residues in achieving specific DNA binding, and Y5 and E30 also play an important role, and overall there was a good correlation between the effects of the mutations in the binding affinities to the operon promoter and in the regulatory activity .
AraR recognizes and binds at least eight palindromic operator sequences, located in the five known arabinose-inducible promoters. Three of these promoters contain two boxes: the promoter of the metabolic operon (boxes OR and OR), of (OR and OR) and of (OR and OR). In the cases of the genes and , a single box is present (OR and OR). AraR binding to the promoters displaying two boxes is cooperative and involves the formation of a small loop in the DNA. In fact, for full repression, communication between repressor molecules bound to two properly spaced operators is required, as shown by the analysis of mutations designed to prevent cooperative binding of AraR (,). An alignment of the eight boxes, identified by DNase I footprinting and/or mutagenesis, showed the 16-bp consensus sequence 5′-ATTTGTACGTACAAAT-3′ and highlighted the conserved nucleotides at each position (A). This operator consensus presents the typical signature for -acting elements recognized by GntR family members 5′-(N)x--N--(N)x-3′ ().
In a previous work, we showed that G is important for AraR binding because the substitution G→T in both boxes OR and OR caused defect in the regulatory activity of AraR and prevented cooperative binding (). To further investigate which nucleotides within the consensus sequence were necessary for protein binding, single-nucleotide exchanges were made in OR at the promoter of the operon. The most conserved bases were substituted and mutations were designed to introduce transversions from AT to CG and CG to AT: A→C, G→T, T→G, C→A, T→G, A→C and T→G. The mutated promoters transcriptional fused to the gene were independently integrated at the locus of the receptor strain (see Materials and Methods section and ). Strain IQB572, bearing a transcriptional fusion to the wild-type operator (Δ Δ- was used as a control to assay the repression exerted by AraR. The levels of accumulated β-galactosidase activity measured in all strains are shown in B. Mutations having the most drastic effect on AraR binding were G→T, both in OR and OR [as previously determined; ()], A→C and T→G, leading to a decrease in the regulatory activity >9-fold compared to the control. A moderate effect of deregulation, varying from 2.4- to 4.4-fold, was observed for A→C, G→T and T→G, and substitution T→G had no effect .
Surprisingly, C→A abolished expression both in inducing and non-inducing conditions. One possibility to explain this result would be an increase in the affinity of AraR for the mutated operator leading to a tight binding of the repressor (even in the presence of arabinose), thus preventing transcription by RNA polymerase. To test this hypothesis, we investigated the Lac phenotype in -null mutants (Δ) bearing the transcriptional fusion – and , strains IQB257 and IQB530 (), respectively. The results obtained in solid medium with X-gal and arabinose indicated that the lack of expression in mutant C→A (Position +4, relative to the transcriptional start site) was independent of the presence of AraR (data not shown) suggesting that the mutation may affect transcription initiation by RNA polymerase.
The effect of the operator mutations was also analysed by EMSA competition assays. The experiments were performed in the presence of 1 nM of the OR–OR DNA probe described above, 40 nM of AraR, and increasing concentrations of a double-stranded 38-bp competitor oligonucleotide (50–500 nM) containing the wild-type or the described mutations in OR or OR. In addition, oligonucleotides carrying all possible substitutions at the highly conserved base pairs, G, A and T were also used. We compared the ability of these cold DNAs to titrate binding of AraR to the labelled probe, reflected in the decrease of the intensity of the protein–DNA complex band. The wild-type OR box was able to compete for AraR binding in a concentration-dependent manner, with a 79% loss of band shift at 500-fold excess competitor DNA (A). In contrast, a non-specific oligonucleotide (equivalent in length; Supplementary Table 1) used as control disrupted only 18% of the binding (data not shown). Inhibition of binding in the presence of 500 nM of the different competitors was quantified and the results are summarized in B. The DNA containing mutation T→G in OR competed in levels similar to that obtained for the wild-type box (68 and 79% inhibition, respectively). In contrast, AraR was unable to bind the boxes with single base-pair substitutions in G, either to T (the mutation tested , previously), to A and C (inhibition values between 21 and 16%). A notorious decrease in binding to A→C was also observed, which was more pronounced when A was exchanged for G or T. However, oligonucleotides containing a mutation at T (either to G, A or C) were still able to partially compete for the repressor. The three mutations, A→C, G→T and T→G, leading to a partial de-repression also showed an intermediate effect. Taken together, the results indicate a good correlation between the and , but the exchanges at T and the mutation A→C, comparatively to the regulatory activity were expected to bind to AraR less tightly. This could be due to the more reduced sensitivity of the competition assays. Interestingly, the mutation C→A, that affected transcription even in the presence of inducer, did not compete (23%) for AraR, indicating that the mutation has an effect on AraR binding that could not be tested .
In previous work, a search for AraR operator sequences in the genome (,) identified a putative binding sequence in the open reading frame [identified as a myo-inositol transporter, ; ()]. The sequence 5′-TTTACGTACAATT-3′ [+27 relative to the translation start site; ()] displayed only two deviations (underlined) from the consensus sequence: A and G. Construction and analysis of transcriptional fusions of the promoter region and 5′-end of to showed that expression is not AraR dependent, thus the potential operator is not functional (Inácio,J.M. and I.S.-N., unpublished data). To determine the ability of AraR to bind this sequence , competition assays were performed as described above but no competition was detected (B). Since T at Position 1 is present in functionally active AraR boxes (A) these observations suggest, in accordance to the mutagenic analysis described above, that G at Position 5 plays an important role in AraR binding.
In conclusion, we found that bases in both arms of the palindrome of the AraR boxes are involved in AraR–DNA contacts. The and studies together with sequence analysis of the eight functionally active AraR boxes indicated that bases G, A and T are crucial for AraR binding, and that A and T play also an important role. Furthermore, Position 5 required a purine (Pu) for functionality , and sequence analysis suggested that the corresponding mirror base (Position 12) in the other arm of the palindrome is always a pyrimidine (Py). An alignment of all putative AraR-binding sequences based on a search of the consensus 5′-ATTTGTACGTACAAAT-3′ in genomes of bacteria from the / group that contain AraR orthologues also highlighted the majority of the invariable positions: Pu, T and the correspondent mirror A, Py and G at the centre of the palindrome (Supplementary Figure 1).
In order to isolate AraR mutants that could suppress the loss of regulation caused by the single nucleotide substitutions in the boxes, an screening method was developed (see Materials and Methods section). Briefly, random mutagenesis of the 5′-end of the allele was performed by PCR and the resulting library of plasmids carrying the mutated alleles was used to transform strains, allowing its integration at the non-essential locus via a double-crossover event. The receptor strains possessed a Δ Δ background, and carried different mutations in the OR operator sequence (). The constitutive expression of due to the inability of the wild-type AraR to bind the mutated operator leads to a Lac phenotype in the absence of inducer. However, if the integrated mutant allele encodes a protein that suppressed the deleterious effect of the operator mutation a Lac phenotype is displayed indicating recovery of regulation. Thus, we screened for transformants with decreased β-galactosidase production in the absence of arabinose.
One transformant of strain IQB533, containing the fusion –, displayed a gain-of-function phenotype, and the sequencing of the allele revealed the substitution M34T, located in the HTH motif of the protein. To determine if this effect was specific or also affected AraR binding to the other mutated promoters, the allele was integrated at the locus of the corresponding strains and β-galactosidase activities were measured (A). Interestingly, the repression exerted by mutant M34T was higher than that exerted by wild-type AraR in almost all operators, in particular with mutant boxes G→T (both in OR and OR), T→G, A→C and T→G. Since a higher level of intracellular accumulation of this AraR variant could explain this phenotype, we determined the accumulation of the wild-type AraR and mutant M34T in strains IQB572 and IQB583, respectively (). The observed cellular levels of protein were similar in both strains (B), indicating that the phenotype displayed by M34T was not due to increased concentration of protein. In fact, EMSA assays performed as described above showed that the mutant displays an increased affinity to the operon promoter probe, with an apparent of 1.0 × 10 M (C), which is almost 4-fold lower than that of the wild-type protein ( 3.9 × 10 M). The substitution M34T is located in the first helix of the winged HTH motif (A). According to a study on protein–DNA interactions based on structures of 129 complexes (), threonine is responsible for a far larger number of protein–DNA bonds than methionine, although almost all are made with the DNA backbone and not the bases. This is in agreement with the results we obtained, that show an increase in the repression exerted over all mutated boxes, suggesting an increased DNA affinity of M34T through non-specific contacts.
italic
#text
S
u
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
The influenza A viruses are members of the family Orthomyxoviridae with a genome consisting of eight single-stranded RNA segments of negative polarity. Each one is encapsidated by binding to the polymerase complex and to a number of nucleoprotein (NP) monomers (), forming a ribonucleoprotein (RNP) complex (). Transcription and replication of each virus RNP take place in the nucleus of the infected cell [reviewed in ()] and hence viral mRNAs are potential substrates for the cellular splicing machinery and need to be exported from the nucleus to express the viral proteins. Indeed, the collinear transcripts from segments 7 and 8 can be directly expressed or can become spliced to generate M1 and M2 or NS1 and NEP(NS2) proteins, respectively (). In productively infected cells, these splicing events are regulated in such a way that the steady-state ratios of spliced versus non-spliced mRNAs is only a few percent [reviewed in ()]. These relative levels can be modified in non-productive cells () or by virus mutation () and are not constant along virus infection (), suggesting the involvement of virus and cellular factors in the regulation of virus mRNA splicing.
The NS1 protein is a crucial regulatory factor during virus infection [reviewed in ()], affecting cellular and viral gene expression and virus counteraction of the interferon response. It accumulates in the nucleus early during virus infection () and when expressed from cloned DNA (), but it can be found in association with polysomes later in the infection (,). NS1 is a RNA-binding protein that interacts with dsRNA (,), vRNA (,), poly-A-containing RNAs () and U6 snRNA (). The RNA-binding domain is located within the N-terminal half of the protein (,), and the rest of the protein appears to contain an effector domain (). NS1 protein interacts with many viral and cellular factors. These include the virus RNP and/or polymerase (), cellular proteins involved in translation, like hStaufen, PABPI and eIF4G (,) and cellular factors involved in post-transcriptional processing of RNA, like CPSF (), PABPII () and NS1–BP, a potential splicing-related factor ().
The splicing of NS1 mRNA has been studied . Early work indicated that it is a non-spliceable transcript under normal conditions (), due the block after formation of a 55S pre-splicing complex (). Subsequent mapping experiments led to the identification of RNA sequences within NS1 mRNA that may down-regulate its splicing efficiency (). studies have also been carried out to analyze the splicing of NS1 mRNA. When expressed from a RNA PolII-driven cDNA it can be spliced (,) and its splicing can be blocked in trans by the expression of the encoded NS1 protein (,).
In addition of the splicing inhibition, PolII-driven expression of NS1 from a cDNA led to a general retention of mRNAs in the nucleus (,,). The interaction of NS1 with the CPSF polyadenylation factor () suggested a mechanism for this transport block, as NS1–CPSF interaction would inhibit the 3′-processing of cellular transcripts. Furthermore, it was proposed that CPSF interaction would indirectly down-regulate the splicing of cellular, but not viral, single-intron containing transcripts ().
The main limitation of these studies is that NS1 expression was carried out by transfection and the ‘viral’ mRNAs analysed are indeed PolII-driven transcripts. To avoid these problems we have established a transient virus replication-transcription system in which a recombinant NS RNP is amplified and transcribed by the viral polymerase. Using this more physiological system we show here that NS1 protein down-regulates the splicing and block the nuclear export of its own viral mRNA. Furthermore, we report a genetic analysis of the roles of the various NS1 domains for such regulation.
The origin of the HEK293T cells () has been described previously (). The cells were cultivated as described (). The pHH plasmid containing the NS cDNA of A/Victoria/3/75 strain RNA under control of the polI promoter has been described (). The antisera specific for NS1 and NEP(NS2) proteins were generated by hiperimmunization of rabbits or rats with purified proteins.
To generate mutant plasmid pHHΔNS1, a site-directed mutagenesis was carried out on plasmid pHHVicNS () using the QuickChangeTM kit from Stratagene and the following mutagenic oligonucleotides: NS1mut1 (5′GCTTCCTTTGGCATGTCTGATAACAAATTGTAGACCAAGAACTAGG3′) and NS1mut2 (5′CCTAGTTCTTGGTCTACAATTTGTTATCAGACATGCCAAAGGAAGC 3′). In this way, two successive stop codons were introduced in the NS1 reading frame within the intron sequence. Likewise, a NS1 mutant plasmid was generated in which RNA binding is abolished (pHHVicNSR38A/K41A) (,), using the following mutagenic oligonucleotides: NS1mut3 (5′TCGGCTTCGCGCAGATCAGGCGTCCCTAAGG3′) and NS1mut4 (5′CCTTAGGGACGCCTGATCTGCGCGAAGCCGA3′).
For protein labelling , cultures were washed, incubated for 30 min in a DMEM medium lacking met-cys and labelled for 1 h at 36 hours post-transfection with S-met-cys to a final concentration of 200 μCi/ml (). Total extracts were prepared in Laemmli sample buffer and processed by polyacrylamide gel electrophoresis and autoradiography.
Immunoprecipitation was carried out as described (). Protein-A-Sepharose immune matrices were prepared with rabbit anti-NS2 IgG and were incubated with soluble cell extracts prepared in 100 mM NaCl-1 mM EDTA-50 mM Tris–HCl-1% NP40, pH 7.5. After washing with the same buffer, the bound material was eluted in Laemmli sample buffer and processed by polyacrylamide-gel electrophoresis and autoradiography.
The transfected or mock-transfected cells were washed with PBS and collected by centrifugation for 5 min at 2500 r.p.m. The isolation of total RNA was carried out using ULTRASPEC RNA Isolation Reagent (Biotecx, Houston, TX, USA) according to the manufacturer's instructions. For cell fractionation, the cultures were resuspended in isotonic buffer (150 mM NaCl-1.5 mM MgCl-10 mM Tris–HCl-0.5% NP40-2mM VRC-1 mM DTT, pH 8.5) and the nuclei were separated by centrifugation for 5 min at 3000 r.p.m. and 4°C. After an additional wash in the same buffer, the supernatants were pooled as cytoplasmic fraction and the pellet was used as nuclear fraction. The extraction of RNA from the nuclear fraction was carried out as indicated above for total RNA. For RNA extraction from the cytoplasmic fraction it was incubated with proteinase K (1 μg/ml) in 100 mM NaCl-50 mM Tris–HCl-5 mM EDTA-0.5% SDS, pH 7.5) for 1 h at 37°C, extracted with phenol–chloroform–isoamylalcohol and recovered by ethanol precipitation. The nucleic acids were digested with RNAse-free DNAse (1 U/μg RNA) and further extracted with phenol-chloroform-isoamylalcohol and precipitated with ethanol as above. Quantitation of ribosomal RNAs and ribosomal RNA precursors in the subcellular fractions indicated that around 80% of total cell rRNAs were present and no rRNA precursors could be detected in the cytoplasmic fraction.
Purification of polyA-containing RNA was carried out by oligo-dT cellulose chromatography. RNA samples were resuspended in 0.5% sarkosyl in DEPC-treated HO and boiled for 3 min. After dilution in 0.5 M NaCl-10 mM Tris–HCl-0.5% sarkosyl, pH 7.5, the samples were applied to the column and incubated for 1 min. The flow-through was re-applied again and the column was washed with the same buffer. The polyA-containing RNA was eluted in two steps by washing first with 0.5% sarkosyl and then with DEPC-treated HO. Finally, the polyA-containing RNA obtained was boiled again and used as input for a second step of oligo-dT cellulose chromatography as above, to ensure the removal of viral genomic RNAs complementary to the virus mRNAs. To control the recovery of the oligo-dT cellulose chromatography a known amount of a polyA-containing labelled riboprobe was added and the radiactivity present in the various fractions was determined. An average recovery of 75% was obtained in the experiments carried out.
The aim of this work was the analysis of the splicing and nucleo-cytoplasmic transport of influenza virus mRNA, independent of other virus processes but nevertheless in a virus-like environment. To this aim we set up a transient viral replication–transcription system in which the generation of the viral transcripts would be driven by the normal virus machinery. As previous work of several groups has implicated NS1 protein in the control of mRNA maturation and transport, we expressed the polymerase and NP by RNA PolII-driven transcription and a segment 8 genomic RNA under control of an RNA PolI promoter (A). In this way, the consequences of NS1 expression on the splicing and nucleo-cytoplasmic transport of segment 8-specific mRNAs could be analysed under conditions more similar to a virus infection. Indeed, when the synthesis of NS proteins was determined in cultures transiently replicating and transcribing a recombinant NS replicon a clear signal for both NS1 and NEP(NS2) proteins was observed (; wt) and the accumulation of NS1 and NEP(NS2) proteins were verified by western-blot or immunoprecipitation using specific antibodies (A, B; wt). Both the synthesis and accumulation levels were similar to those obtained in a normal infected cell (data not shown). This experimental setting allowed the genetic analysis of the implication of NS1 protein in viral mRNA maturation and transport by simply mutating the replicon included in the system. Thus, a number of altered replicons in which NS1 had been mutated or replicons derived from NS1 mutant viruses (,) were prepared and their genotype is described schematically in B. Mutant ΔNS1 contained two stop codons close to the N-terminus of the protein but 22 nt away from the splicing 5′-donor site, leading to the synthesis of a 18 amino acids long peptide. Accordingly, neither synthesis nor accumulation of NS1 protein was observed in the system with this replicon ( and A), while the synthesis of NEP(NS2) protein was still detectable ( and B). Likewise, NS1 mutants lacking the effector domain (NS1 81) or devoid of RNA-binding activity (NS1 R38A/K41A) and mutants with ts phenotype (NS1 81 and NS1 11C) could be expressed in the system and showed expression levels according to the described phenotypes (, A–C) (,). Furthermore, immunofluorescence studies indicated that a large fraction of the cultures in which the NS replicons had been transiently established showed NS1 expression (), allowing biochemical analysis of the transcripts generated. In these cultures, the levels of NS1 expression were similar to those observed during infection and most of the signal was found in the nuclei of the cells ().
To study the potential role of NS1 protein in the maturation and transport of NS segment transcripts we first determined whether the expression of NS1 could alter the stability of the collinear or spliced mRNAs, encoding NS1 and NEP(NS2) proteins, respectively. Total cell RNA was isolated from cultures in which wt or ΔNS1 NS replicons had been established, prior to and at various times after actinomycin D chase. The polyA-containing fraction of these preparations was analysed by RNAse-protection assay using a probe that could distinguish collinear, unspliced mRNA from the spliced transcript. The results are shown in and indicate that the presence or absence of NS1 did not alter the half-lives of NS1 or NEP(NS2) mRNAs. Therefore, we can use the relative accumulation of NS1 versus NEP(NS2) transcripts as a measure of the extent of NS collinear mRNA splicing.
To determine the extent of splicing of NS collinear transcript we establish the various NS replicons by transfection of cultures of HEK293T cells as described above. Total cell RNA was isolated and the polyA-containing RNA fraction was purified and analysed by quantitative RNAse-protection assay. The conditions of the assay were set to ensure an excess of probe for all samples. The results of a representative experiment are presented in and summarizes the data obtained from three independent experiments. When wt NS replicon was used, both NS1 collinear and NEP(NS2) spliced mRNAs were detected and an extent of splicing of around 45% was obtained (, wt; ). However, deletion of NS1 led to dramatic change in the splicing efficiency, as almost no collinear transcript could be detected (, ΔNS1; ). These results indicate that NS1 protein down-regulates the splicing of its own mRNA.
To test whether the N-terminal RNA-binding or the C-terminal effector domains in NS1 are involved in this down-regulation we determined the splicing phenotypes of mutants affected in either one. Mutant NS1 81 () lacks the sequences 81-238 in NS1 protein and hence does not contain the effector domain. The NS1 81 mutant virus is temperature-sensitive and shows a major defect in late gene expression and minor reduction in vRNA replication. Therefore, we examined the splicing of transcrips from a NS1 81 replicon at both permissive (32°C) and restrictive (39°C) temperatures. The extent of splicing at permissive temperature was similar to wt (, NS1 81; ), and therefore we can conclude that the N-terminal region of NS1 protein is sufficient for down-regulation of the splicing of its own mRNA and the C-terminal effector domain is not required for such function. Although splicing efficiency was reduced at restrictive temperature (, NS1 81; ), the wt replicon showed the same decrease (, wt; ), indicating that this effect is not due to the mutation. Such a reduction for the wt replicon could be the consequence of a low-splicing efficiency of the cell machinery at high temperature, as the corresponding wt recombinant virus showed reduced replication at 39°C (), or may be connected to the temperature-sensitivity of vRNA replication reported earlier ().
In view of these results we tested whether the NS1 RNA-binding activity, which has been mapped to the N-terminal half of the protein (,,), is required to down-regulate NS1 mRNA splicing. Mutations R38A and K41A, known to abolish NS1–RNA binding (,) were introduced in the NS replicons and the ratio of NS1 versus NEP(NS2) mRNAs was determined in cells in which the mutated replicon had been established. The results are presented in and and show that essentially no splicing of the NS collinear transcript is observed. Thus, we can conclude that RNA-binding by NS1 protein is not necessary for it to down-regulate the splicing of its own mRNA and a mutant lacking RNA-binding activity behaves as an highly efficient blocker of NS1 mRNA splicing. The splicing efficiency of NS1 ts mutant 11C (V18A, R44K, S195P) was similar to that observed for mutant NS1 81 (data not shown).
The transcription, maturation and nucleo-cytoplasmic transport of mammalian mRNAs are linked processes and influenza virus transcription depends functionally and structurally on the cellular transcription apparatus (). In view of the alterations induced by NS1 in the splicing of NS transcripts we seeked for potential changes in the nucleo-cytoplasmic transport of these mRNAs. The various NS replicons described above were established in HEK293T cell cultures and nuclear and cytoplasmic fractions were prepared. The polyA-containing RNAs from these fractions were isolated and the recovery of NS1 and NEP(NS2) mRNAs was determined in each fraction by quantitative RNAse-protection assay. The results of a representative experiment are presented in and the data obtained from three independent experiments is summarized in . The distribution of these mRNAs in both cellular fractions indicated that nucleo-cytoplasmic export was more efficient for the spliced NEP(NS2) than for the collinear NS1 mRNA (, wt). When NS1 expression was abolished, most of the transcript was spliced and efficiently exported to the cytoplasm (, ΔNS1). However, the collinear NS1 mRNA generated by the R38A/K41A mutant replicon, which constitutes most of the NS transcript produced (see ) was exported very efficiently (, R38A/K41A). Therefore, we can conclude that NS1 protein down-regulates the export of its own mRNA and that this regulation requires the RNA-binding activity of NS1. A temperature sensitive phenotype was found for transport of NS1 mRNA with mutants NS1 81 and 11C. Thus, while the phenotype obtained at 32°C was indistinguishable from wt, the export of NS1 mRNA was very efficient at restrictive temperature (, NS1 81; 11C). Contrary to the results for NS1 mRNA splicing (), the transport of NS mRNAs generated from a wt NS replicon was not affected at high temperature (; ), thereby verifying that mutants NS1 81 and 11C are temperature-sensitive for the block of NS1 mRNA export.
In this report, we have analysed the consequences of NS1 protein expression in the splicing and nucleo-cytoplasmic transport of NS transcripts using a transient system for the replication and transcription of a recombinant influenza virus NS replicon. This is a convenient system that can be used as a model of the situation of an infected cell. Indeed, the experimental approach used is a development of the amplification of transcriptionally active viral mini-RNPs employed in the past for structural studies (,,). Furthermore, full-length RNPs generated in a similar way have been rescued into infectious virus (,). Here we show that NS replicons can be transiently amplified and express both NS1 and NEP(NS2) proteins in proportions and to levels similar to those produced in the virus infection (). We felt that this should be a convenient system to carry out a genetic analysis of the role of NS1 in the maturation and export of viral transcripts as the studies could be made independent of other possible phenotypes described for NS1 mutants. Thus, mutants NS1 81 and 11C show predominant alterations in late virus gene expression and particle formation (,), aspects not relevant in the experimental setting used here. On the other hand, a ΔNS1 virus mutant is difficult to handle in a cell line with a normal interferon response () but can be studied normally in a transient situation as the one described here. Although it has been reported that the presence of NEP(NS2) protein inhibits viral RNA replication in a similar recombinant system (), we consider this effect non-relevant to our studies, that deal with the post-transcriptional processing and export of viral mRNAs.
The results presented here indicate that NS1 protein inhibits the splicing of the NS collinear transcript in a RNA-independent manner, as mutant NS1 R38A/K41A behaves as wt (; ), and preferentially blocks the nucleo-cytoplasmic export of NS1 mRNA by a mechanism dependent on NS1–RNA interaction, since mutant NS1 R38A/K41A permits an efficient export of this mRNA (; ). Although cells in which a mutant NS R38A/K41A has been established accumulate large amounts of NS1 mRNA and very little NEP(NS2) mRNA in their cytoplasms ( and ), quasi-normal synthesis of NEP(NS2) protein is produced ( and B). This apparent contradiction can be explained by the fact that, at early times in the replication and transcription of the NS R38A/K41A replicon, no NS1 protein is present in the system. Until NS1 protein is accumulated to the required intracellular levels, efficient splicing of the collinear transcript is to be expected, as this is the situation with a NS ΔNS1 replicon (). Nevertheless, to account for the observed levels of NEP(NS2) synthesis it would be necessary to invoke a preferential translation of NEP(NS2) mRNA. It is conceivable that NEP(NS2) mRNA is efficiently incorporated into the cell translation machinery because it is generated at the early phase of the replicon gene expression. On the contrary, NS1 mRNA is produced at later times, when NS1 protein has accumulated. At that point a large fraction of the cell-derived mRNA is being used for virus-induced cap-snatching and incorporation of new mRNAs into the cell translation machinery might be affected. In addition, the level of NS1 protein present in each setting may influence the translation ability of the NEP mRNA present (,,). Thus, when NS1 is not expressed (ΔNS1 replicon), translation of NEP mRNA might be much less efficient than in those cases in which NS1 protein is abundantly expressed, as for instance in cells harbouring the wt NS replicon.
The effects of NS1 in the splicing and nucleo-cytoplasmic transport of mRNAs have been studied previously by several groups, including ours, but the analyses were carried out or using as targets cellular mRNAs or viral ORFs expressed as RNA polymerase II transcripts, but not true virus transcripts. On the basis of these studies, several models have been put forward along the years for the inhibition of pre-mRNA splicing by NS1: (i) From the results of experiments it was first proposed that the binding of NS1 to a specific motif in U6 snRNA would impede proper formation of U6-U2 and U6-U4 snRNP interactions, leading to the block of the splicing reaction (,). Such inhibitory interaction was postulated to occur in splicing reactions on cellular pre-mRNAs but not on viral mRNAs (). This model would not explain the alterations induced by NS1 in the use of alternative splice sites () and is not supported by the results presented in this report, as we show that NS1 protein indeed inhibits the splicing of a true viral pre-mRNA (the NS collinear transcript) and the RNA-binding activity of NS1 is not required for such inhibition (). (ii) After reporting that the effector domain of NS1 protein binds the 30 kDa subunit of CPSF and blocks cellular mRNA cleavage and polyadenylation (), it was proposed that the splicing inhibition by NS1 would be an indirect consequence of the failure in 3′-end processing, as the definition of the exon in a single-exon containing mRNA depends on the 3′-end polyadenylation (). This model would explain the inhibition of cellular mRNA splicing and the lack of such inhibition for intron-containing viral mRNAs (), since the 3′-polyadenylation of the latter is carried out by polymerase stuttering (). Again, our results do not support the proposed model, as NS1 protein indeed blocks the splicing of the NS collinear transcript and a NS1 mutant lacking the effector domain fully inhibits the splicing of its own mRNA ().
We cannot rule out the possibility that NS1 inhibition of cellular mRNA 3′-end formation would indirectly affect the splicing of cellular pre-mRNAs, but the results presented here suggest that NS1 protein would inhibit splicing by a more general mechanism. The interaction with NS1–BP, a human protein involved in splicing () is one possibility, while a de-regulation of the phosphorylation state of SR proteins is an alternative that would explain the modification in alternative splice site usage by NS1 expression (). A general inhibition of the splicing machinery would be in line with the changes in the localization of snRNPs induced by NS1 expression or virus infection () and could be considered as part of the viral response to inhibit the expression of cellular antiviral factors (). Whatever the mechanism involved, our data indicate that inhibition would be mediated by the N-terminal region of NS1 but would not require NS1–RNA interactions ().
As indicated earlier, the nucleo-cytoplasmic transport of mRNAs is inhibited by expression of NS1 from cloned DNA (,,) and a model has been proposed for such inhibition based upon the interaction of NS1 with the 30 kDa subunit of the CPSF and the subsequent inhibition of 3′-end mRNA processing (). In addition, the interaction of NS1 with nuclear PABPII would block the export of mRNAs that could partially escape from the inhibition of 3′-end formation (). This model suggested that the export of true viral mRNAs would not be affected by this transport block because their poly A-tails are produced by the viral polymerase in a way independent of normal cell mRNA polyadenylation (). In addition, the interaction of NS1 with the TAP/p15 pathway for mRNA export has been recently described (), leading to the block of transport of cellular mRNAs that have undergone splicing. However, the results shown in this report indicate that a preferential block of NS1 mRNA export is produced when NS1 is expressed from a NS replicon and further suggest that the RNA-binding activity of NS1 is required for the inhibition of viral mRNA export (). These results suggest that, in addition to the block of cellular, spliced mRNA export, mediated by the inhibition of cellular mRNA 3′-end processing () or by targetting the TAP/p15 mRNA export pathway (), NS1 knocks down the export of its own virus mRNA, a non-spliced mRNA, by RNA-binding. The mechanism for this nuclear retention is not clear at this point in time but a recent publication showing that nuclear export of some virus mRNAs, including NS1 mRNA, requires their association to the cellular transcription machinery might shed some light (). Thus, NS1 might act by avoiding the coupling of the viral transcription events to the RNA polymerase II complex after virus RNP-mediated cap-snatching and hence interfering with the association of shuttling hnRNP proteins to the virus transcripts. This interference would not take place on the spliced NEP(NS2) mRNA because it would associate to the standard EJC/TAP-p15-dependent mRNA export pathway ().
The nuclear retention of NS1 mRNA could be a way to improve the possibilities for this collinear transcript to form a spliceosomal complex in the context of general splicing inhibition described above. Thus, complete inhibition of the splicing of the NS collinear transcript would be deletereous for virus infection, as no NEP(NS2) protein would be formed, and the virus might have evolved an escape mechanism to avoid a complete block of viral mRNA splicing. Such scenario is compatible with the phenotypes observed in the genetic analysis presented here. When NS1 protein is not expressed (ΔNS1), no inhibition of splicing and export is produced, the collinear NS1 transcript is almost fully spliced and NEP(NS2) mRNA is efficiently exported. If a RNA-binding mutant of NS1 is expressed, splicing inhibition persists but the export block is eliminated because no RNA recognition can take place, leading to an efficient export of collinear NS transcript.
In summary, the results presented in this report demonstrate that NS1 regulates the splicing and nucleo-cytoplasmic export of its own collinear mRNA, in addition to the previously reported inhibitions of cellular mRNA processing and export. The simplest interpretation of these results implies that the mechanisms for splicing and export inhibition are distinct and at least in part different from the way NS1 appears to alter cellular mRNA metabolism. |
Recent developments indicate that -nucleosides and -nucleoside analogues are potent anti-viral, anti-tumour and even anti-malarial agents (). The -derivative, lamivudine (β--2′,3′-dideoxy-3′-thiacytidine, 3TC), has been approved for treating both HIV and HBV, while emtricitabine (β--2′,3′-dideoxy-5-fluoro-3′-thiacytidine, FTC) is used in HIV therapy. Telbivudine (β--thymidine, -dT) was recently approved for treating hepatitis B by the US Food and Drug Administration. Many -nucleosides are currently in advanced clinical trials for the treatment of a variety of virus diseases. These include clevudine [1-(2-fluoro-5-methyl-β--arabinofuranosyl)uracil, -FMAU], elvucitabine (2′,3′-didehydro-2′,3′-dideoxy-β--5-fluorocytidine, β--d4FC), valtorcitabine (val-β--2′-deoxycytidine, val--dC), pentacept (2′,3′-didehydro-2′,3′-dideoxy-3′-fluoro-β--cytidine, β--3′-Fd4C) and β--2′-Fd4C (2′,3′-didehydro-2′,3′-dideoxy-2′-fluoro-β--cytidine). Most -enantiomers considered for treating virus diseases have similar activities to their -counterparts, but are less sensitive to degrading enzymes and have better safety profiles (,). These properties are due mainly to the enantioselectivity of the enzymes that interact with these substrates (). -nucleosides must be phosphorylated by cellular or viral kinases before they can reach the targeted virus enzymes. The enantioselectivity of the enzymes involved in nucleoside synthesis and salvage pathway is governed by no general rule. Each individual metabolic enzyme must therefore be studied. Among the four deoxyribonucleoside kinases in human cells, cytosolic thymidine kinase 1 (TK1) is strictly enantioselective while mitochondrial thymidine kinase 2 (TK2) is less specific (). Deoxycytidine kinase (dCK) and deoxyguanosine kinase (dGK) are poorly enantioselective (,). The phosphorylation of nucleoside monophosphates and their analogues to their diphosphate derivatives is then carried out by NMP kinases in both and salvage pathways. The NMP kinases in human cells include one dTMP kinase, one UMP-CMP kinase, six isoenzymes of adenylate kinase and several guanylate kinases (). Only the abilities of hUMP-CMP and dTMP kinases to phosphorylate some -deoxynucleoside monophosphate analogues has been studied to date (). Finally, NDPK is strictly enantioselective (). The final step can be carried out by phosphoglycerate kinase, which has a broad substrate specificity (). Other ATP-synthesizing enzymes such as creatine kinase can also be involved in this step (). Triphosphorylated -derivatives can interact with viral polymerases, acting as competitive inhibitors or alternate substrates, usually leading to chain termination. Viral polymerases incorporate -derivatives more readily than do the human ones (). Nucleoside degrading enzymes, such as nucleoside deaminases and phosphorylases, tend to be strictly selective for -nucleotides although only limited data is available (,). The lack of degradation of -nucleosides has however been demonstrated in mice, where -nucleosides were gradually excreted in the unchanged form following intraperitoneal administration (). 5′-Nucleotidases also seem to be moderately to highly enantioselective (,). The ribonucleotide reductases, which catalyze the conversion of ribonucleotides to 2′-deoxyribonucleotides, are also key enzymes in DNA replication and repair. , these enzymes are inhibited by 2′-azido-2′-deoxynucleoside 5′-diphosphates (). However, N--dUrd is not cytotoxic due to its poor intracellular phosphorylation. The mechanism of enzyme inactivation and the enzymatic monophosphorylation of the parent nucleosides has been thoroughly studied (). More recently, we have shown that ribonucleotide reductase is enantiospecific with respect to the natural configuration of the sugar moiety (). This report examines the phosphorylation of the β- and β- stereoisomers of 2′-azido pyrimidine monophosphates by human UCK, TMPK and AKs and characterizes the enantioselectivity and the possible cross-activities of these three human kinases for natural (d)NMP ().
Part of this work was presented during the XVIIth Round Table for Nucleosides, Nucleotides and Nucleic Acids in Bern, in September 2006.
#text
The activities of the major adenylate kinases hAK1 (cytosolic) and hAK2 (in the inter-membrane space of mitochondria) were measured as a function of the - and -(d)AMP concentrations as well as of pyrimidine nucleoside monophosphates. Human AK1 and AK2 were both specific for the -enantiomers. They did not phosphorylate -(d)NMP, except for hAK2 having very little action on -dAMP corresponding to a ‘catalytic efficiency’ about 10 smaller than for the -stereoisomer (). The rate of -(d)AMP phosphorylation increased with the substrate concentration, as expected for Michaelis curves. The maximum rate V of -AMP phosphorylation by hAK2 was not reached due to substrate inhibition at concentrations above 0.3 mM (data not shown). This inhibition is generally attributed to unproductive binding of the ligand at the same site or at a secondary site. The reaction rates for hAK2 were slightly below those for hAK1 but were almost compensated for by the K values, resulting in somewhat similar catalytic efficiencies as might be expected from the similarities of their active site sequences. -dAMP was a slightly poorer substrate than -AMP for both hAK1 and hAK2. The deoxyribonucleotides, -dCMP and -dUMP, were not substrates for the AKs although these enzymes phosphorylated the pyrimidine nucleotides, -CMP and -UMP, to a minor extent () ().
-AMP was bound by both enzymes. It competitively inhibited the phosphorylation of -AMP and -dAMP by hAK1 and hAK2, with an estimated K of 200 µM for hAK1 and 18 µM for hAK2 (data not shown). -dAMP was used because the mitochondrial AK2 was inhibited by excess -AMP. The binding of -nucleotides to the AMP site of hAK1 and hAK2 was thus unproductive. Additionally, -(d)AMP was also not a substrate for hUCK or hTMPK (data not shown).
Despite this poor phosphorylation, -dA reduced the virus load in the woodchuck model of chronic B hepatitis although less efficiently than -dC or -dT (). The other isoenzymes of AK, i.e. hAK3, 4, 5 and 6 could contribute to the phosphorylation of -(d)AMP and explain this activity ().
The recent approval of -dT for treating hepatitis B has prompted us to study the phosphorylation of -dTMP by hTMPK. The catalytic efficiency of the enzyme for -dTMP was around 0.7% that -dTMP (). The catalytic turnover number k was 10 times smaller than that for -dTMP, while the K was 20-fold higher, resulting in a lower k/K (about 900 Ms) (A, ). -dTMP also gave rise to substrate inhibition at 1 mM and the K value was estimated to be 5 mM (). -dUMP was a good substrate with an efficiency in the 10 Ms range, as already reported (), but the phosphorylation of -dUMP very slow (k/K 60 M s).
The relative binding constants for the - and -isomers of dTMP were measured using the fluorescent competition assay based on the fluorescent probe MABA-dTDP (). The dissociation constants for -dTMP and -dTMP from hTMPK were 2 µM and 45 µM, respectively (B). The K ratio was 22, comparable to the K ratio, i.e. 19 (). Using the same assay, the nucleosides, - and -dT, also competed with MABA-dTDP and had K values of 40 and 450 µM, respectively, indicating that the -nucleoside also binds to the active site less efficiently than its -counterpart. The affinity of the enzyme for -dTMP (K = 45 µM), was greater than its affinity for -dUMP (K = 150 µM), which is a good substrate of the enzyme (B).
-dUMP and -dCMP were substrates of hUCK, but they were phosphorylated much more slowly than were the natural nucleotides, despite the K of the enzyme being slightly more favourable for the -compounds. The K values were 1.3 mM for -dUMP and 1.0 mM for -dCMP, compared to 0.70 mM for -dUMP and 0.73 mM for -dCMP (, ). But the reaction rate was 20 times slower for -dUMP than for -dUMP and 100 times slower for -dCMP than for its -counterpart. -CMP was also phosphorylated by hUCK but it was a very poor substrate for this enzyme, with a catalytic efficiency of 100 M s, 10 times lower than that for -dCMP and 10 times lower than for -CMP. This decreased efficiency is due to a much lower k, as the K was similar to that for -dCMP (0.75 mM). Excess of -CMP was inhibitory, unlike the natural substrate. The -derivate that was the best substrate for hUCK was -dCMP, but it was far less efficient than the monophosphate form of the antiviral analogue -3TC. -3TCMP is indeed phosphorylated to -3TCDP by hUCK with a catalytic efficiency k/K of 2.8 × 10 M s ().
The activity of recombinant hUCK was evaluated using the nucleoside analogues, N--dCMP and N--dUMP (). Both compounds were substrates of the enzyme and followed Michaelis–Menten kinetics. High concentrations did not give rise to substrate inhibition of hUCK, unlike the natural substrates, -CMP and -UMP. The K values were 0.36 mM for N--dCMP and 0.9 mM for N--dUMP compared to 20 μM for -CMP and 50 μM for -UMP (). Despite the higher K values, the catalytic efficiencies for phosphorylation of the 2′-azido derivatives were in the same order of magnitude (10–10) as those for the natural substrates of the enzyme. The reaction rate with N--dUMP was similar to that with -UMP (, ). The reaction rate of hUCK with N--dCMP was almost 4 times faster than that with -CMP, demonstrating that N--dCMP is currently the best-known substrate for hUCK (B). Other enzymes hAK1, hAK2 and hTMPK did not significantly phosphorylate these substrates.
The enantioselectivity of hUCK for 2′-azido-2′-deoxynucleoside monophosphates was also evaluated. The K value for N--dCMP was increased (1.2 mM) while the V was lower (1.1 U·mg), leading to a 4000-fold lower catalytic efficiency than for the -enantiomer (). N--dUMP was also a poor substrate of the enzyme, with a K of 1.3 mM and a k of 0.34 s(). The 2′-azido group was thus only favourable for the -analogues. Overall, in the -series, the pyrimidine 2′-deoxyribonucleotides were better substrates than the pyrimidine ribonucleotides and the 2′-azido-2′-deoxyribonucleotides, indicating that 2′-substitutions may cause steric hindrance for -NMP binding to hUCK.
The NMP kinases all have a highly conserved structure with a central CORE domain that contains an ATP binding loop (P-loop) and two mobile domains: an NMP binding domain and an LID domain, which provide the catalytic residues for the reaction (). Both the NMP and LID domains are extremely mobile and undergo large ‘hinge bending’ motions (). The same conformational changes are believed to occur in all NMP kinases when they switch from their opened to closed conformation upon substrate binding. The X-ray structure of hAK1 complexed with the bisubstrate inhibitor Ap5A was recently solved: it showed the interactions of AMP with the active site in the closed conformation (). The structure of hTMPK complexed with various ligands has been thoroughly explored (), but the structure of the free apoenzyme is not known. In contrast, the structure of hUCK is only known in its open conformation (). Substrate-free hTMPK probably also exists in an opened conformation and substrate-bound hUCK probably adopts a closed conformation as does the homologous enzyme from ().
-AMP and -dTMP were tentatively docked in the closed conformation of hAK1 and hTMPK, and -dCMP was docked in an enzyme model based on the UCK (). The docking of -(d)AMP in hAK1 failed to provide a model (data not shown): this could be due to the rather specific interactions of -AMP with the protein, which are not mediated by water molecules, thus limiting its capacity to accommodate other substrates. The docking of -deoxypyrimidine monophosphates in hTMPK and UCK was more successful ( and ). The dTMP and CMP binding sites are represented as four interacting motifs: (i) the LID domain, (ii) the P-loop, (iii) the mobile part of NMP binding domain and (iv) the stable part of NMP binding domain. The binding of -dTMP differed from that of -dTMP, especially at the deoxyribose moiety (). The H-bonds between the first layer residues and the thymine ring involved several water molecules, providing a ‘flexible’ NMP binding domain that readily accommodates modified substrates. The thymine ring was stacked onto Phe72 in -dTMP binding, but it was not optimal, as the base was shifted through 15°. The interaction of the phosphate group with Mgion and Asp15 in the P-loop, via a water molecule, was conserved, as was its interaction with Arg97 from the mobile NMP domain. However, the deoxyribose 3′OH, which interacts with the LID Gln157 via a water molecule in the -dTMP/hTMPK complex, was modelled as being H-bonded to the NH of the highly conserved Arg45 of the immobile NMP domain in the -dTMP/hTMPK model.
The major interactions responsible for the binding of -CMP to hUCK involved the base and phosphate (). The interaction of Asn100 with the 4-amino group of the cytidine base played an important role in the base specificity and explains why -(d)CMP is a better substrate than -(d)UMP (). The phosphate does not interact with the P loop as in hTMPK but strongly connects the LID domain (Arg134 and Arg140) to the immobile NMP domain (Arg39, Arg96 and Glu36) (A). These residues still interacted with -dCMP phosphate group in the -dCMP/UCK model, except for Arg134 (B). The cytidine moiety for both -CMP and -dCMP was H-bonded to Val63 and Asn100 from the mobile NMP domain, resulting in similar positioning of the base for both enantiomers. The -nucleotide did not have the major interaction of the -CMP 2′OH with the main chain Lys61 carbonyl, contained within the NMP mobile domain. -dCMP was a better substrate for hUCK than was -CMP, probably due to 2′-OH steric hindrance. However, the ring oxygen in -dCMP interacted with Arg96, anchoring the sugar to the immobile NMP domain (B).
The model obtained with N--dCMP bound to hUCK had favourable interactions as the azido group interacts with both Asp142 in the LID domain and the α-carbonyl of Gly60 in the mobile NMP domain (C). The 2′OH of the ribonucleotide has been shown to contribute to the LID closure of hUCK by its interaction with the carbonyl of Lys61 located in the mobile NMP domain (). This interaction also explains the higher affinity of the enzyme for D-ribonucleotides compared to -deoxyribonucleotides. The replacement of 2′-hydroxyl by a bulkier azido group prevents complete closure of the LID, which results in higher K values. However, the opening of the LID, which promotes product release, is proposed to be the rate-limiting step of the reaction catalyzed by hUCK. A recent NMR experiment on hyperthermophilic and mesophilic homologues of AK also suggested that the opening of the AMP binding and/or LID domains upon product release was the rate-limiting step influencing the catalytic turnover (,). The azido group might thus facilitate the LID opening or could connect the LID and mobile-NMP domains and synchronize their opening, explaining the higher rate of phosphorylation of N--dCMP compared to -CMP. Dynamic studies should help to understand these higher rates.
We have compared the activities of the kinases on - and -nucleotides and shown that both hTMPK and hUCK have relaxed enantioselectivities for dNMP, while hAK 1 and 2 are strictly devoted to -(d)NMP. The pyrimidine kinases phosphorylated only -derivatives of the deoxy series. UCK was active with -dCMP but far less efficiently than with -3TCMP (). hTMPK, hUCK, hAK1 and hAK2 all phosphorylated their respective -(d)NMP with some minor cross-reactivity. For example, -CMP was phosphorylated efficiently by hUCK (k/K = 6.5 × 10 M s) and more slowly by hAK1 and hAK2 (k/K about 10 M s). -dUMP was a better substrate for hTMPK (k/K = 2.8 × 10 M s) than for hUCK (k/K = 6 × 10 M s). N--dCMP was the best -series substrate for hUCK with a k 4-fold higher than that for the natural substrate (-CMP). This could be due to a tighter interaction between the LID domain and the NMP domain as explained by the structural analysis.
RNAi studies have shown that hTMPK is implicated in the activation of -FMAU (). There was generally a good correlation between kinetic studies and the antiviral effect, emphasizing the need for efficient cellular activation of the antiviral drug to its triphosphate form, as shown for -3TC (). The present study demonstrates the activity of hTMPK on -dTMP, even if the catalytic efficiency was relatively low (k/K = 900 M s). -dTMP had the highest relative phosphorylation efficiency of all the -(d)NMP tested. The phosphorylation of -dTMP correlates with studies on the intracellular metabolism of -dT in HepG2 cells and primary cultures of human hepatocytes. The conversion of -dTMP to -dTDP clearly appears to be the rate-limiting step in both cell types (). The first phosphorylation of -dT is believed to be carried out by either hdCK or hTK2 with catalytic efficiencies of 1.6 × 10 M s and 1 × 10 M s, respectively (). The recent crystallization of dCK with -3TC and troxacitabine shows how the nucleoside binding site of dCK can conserve all essential interactions with these analogues or -dC and thus maintain productive substrate positioning for phosphoryl-transfer (). Given the great similarity between dCK and hTK2, we can assume that they are similarly flexible for the productive positioning of -dT, explaining the high catalytic efficiency. The binding of -dTMP to hTMPK is not optimal and may be the rate-limiting step in the pathway. NDPK does not recognize -(d)NDP as substrates and phosphoglycerate kinase is probably involved in the conversion of -dTDP into -dTTP, with a catalytic efficiency of 500 M s (). From these data, the second and third phosphorylation steps have low catalytic efficiencies and appear rate-limiting in the activation pathway. This study demonstrated the relaxed enantioselectivity of hTMPK, which is most likely essential to the formation of -dTTP and thus to the antiviral properties of -dT. |
In most animal models, including , , zebrafish and , the onset of zygotic gene activation is delayed until the midblastula transition (MBT). (; ). Whereas there is no MBT in mammals, here zygotic gene activity is also delayed after fertilisation (). In the zebrafish blastula, the general delay in zygotic gene activity is followed by the sudden and broad activation of a large number of genes representing all main gene ontologies (; ) leading to gastrulation. The activation of the zygotic genome is parallelled by an equally significant process, the differential degradation of maternally inherited mRNAs (; ; ). Whereas little is known about the mechanisms of degradation of maternal mRNA, they are known to involve both transcription-dependent and -independent pathways (; ; ; ). Dynamic changes in expression of maternally and zygotically activated genes are observed during zygotic gene activation also in the mouse (). Not all maternally inherited mRNAs degrade during early embryogenesis and many maternal mRNAs continue to influence embryo development until later developmental stages (; reviewed by ).
The initiation of zygotic transcription during MBT is believed to be regulated by a competition between chromatin and the assembly of the transcription machinery (; ; ). The TATA-binding protein (TBP) has been implicated as a key regulator of transcription initiation in early embryo development in vertebrates (; ; ). TBP protein levels have been shown to be limiting for transcription before MBT and are dramatically upregulated at the initiation of zygotic transcription (; ; ). TBP, together with TBP-associated factors (TAFs) are components of the TFIID complex, a key point at which activators can control transcription through the core promoter. Until recently, it was argued that TBP is required for the correct initiation of all RNA polymerase (Pol I, II and III)-mediated transcription in eukaryotes. However, recent reports have revealed the contrary: the composition of Pol II core promoter-binding complexes varies and is likely to represent a point of differential gene expression regulation (reviewed by ). Consistently, whereas TBP is essential for early embryo development, it is not required for all Pol II transcription as demonstrated by studies on a small number of vertebrate genes (; ; ). The apparent redundancy of TBP in vertebrates is probably due to the function of TBP-like factors (TLF/TRF2) (; ) and the recently described second set of TBP paralogue genes TBP2/TRF3 (; ; ). The functional requirement for different TBP family proteins in embryogenesis suggests specific nonoverlapping roles for these factors in regulating subsets of genes (; ; ; ).
Our objective was to investigate the transcriptional regulatory mechanisms that involve core promoter recognition proteins such as TBP in the whole organism. The transition of gene activity from maternally inherited mRNAs to zygotic gene expression provides an ideal model for the analysis of the control of transcription initiation (). By using Morpholino (MO) knockdown and microarray expression profiling, we have addressed which genes require TBP for their activity and what is the function of TBP in regulating the transition from maternal to zygotic regulation during early vertebrate embryo development. We show that TBP is preferentially required for genes that exhibit dynamic changes in their expression during ontogeny. Furthermore, we provide evidence for a previously undocumented negative regulatory role of TBP in zygotic gene activation. Importantly, we also describe a novel biological function of TBP: a role in the degradation of a subset of maternal mRNAs after MBT.
In the early embryo, the steady-state levels of mRNA result from a dynamic process of gradual degradation of maternal mRNAs and the delayed initiation of zygotic gene expression at the MBT (). To investigate the role of TBP in regulating genes expressed in the early zebrafish embryo, we carried out a microarray analysis of 10 501 genes at the dome stage in embryos in which TBP function was blocked using MO antisense oligonucleotides as described previously (). The dome stage occurs 1.3 h after the start of the global initiation of zygotic transcription at the MBT () and any TBP-dependent changes in gene activity detected at this stage are still expected to be mostly direct transcriptional effects. Loss of protein was confirmed by Western blot (). A total of 1927 genes from the 10 501 represented on the microarray were selected, having applied stringent but commonly used criteria (FDR cut off of 0.05) to eliminate potential false positives and false negatives from the analysis (see Materials and methods). Three distinct response groups of genes were identified: downregulated genes (⩽−2-fold change), genes with low variability in expression (values between >−2 and <+2-fold-change); upregulated genes (⩾+2-fold change) ( and ). The three groups thus identified were further validated by semiquantitative RT–PCR experiments; out of a total of 39 genes representing the above groups, 37 showed comparable activity to that observed in the microarray experiments (). The specificity of the effects detected was confirmed by microinjecting a second MO targeting TBP mRNA (TBP MO2), which resulted in comparable gene expression changes to the above TBP MO injection when analysed by RT–PCR of 28 genes (). Furthermore, the gene expression changes caused by TBP MO injection could be reverted by injecting a form of TBP mRNA that could not be targeted by the MO ().
To characterise further the genes affected by loss of TBP function, we tested whether genes in the three response groups described above showed discrete expression dynamics during zebrafish ontogeny. To this end, we compared the data set described above to an ontogenic stage-dependent expression profiling experiment on the zebrafish transcriptome (Konantz M, Otto G-W, Weller C, Saric M, Geisler R. Microarray analysis of gene expression in zebrafish development, manuscript in preparation). The two gene sets share 717 genes () and the proportions of the three TBP morphant response groups among these 717 genes are similar to those of the total TBP microarray data set ().
Meta-analysis of the ontogenic stage-dependent gene expression array was carried out to define two classes of genes. Genes showing stage specific peaks of expression activity during zebrafish ontogeny were classified as ‘stage-dependent' and genes that showed no significant variation in gene expression during ontogeny were considered as constitutively active genes (, see Materials and methods). Nearly half of the 717 genes (46.9%) that overlap between the two microarray data sets were shown to be constitutively expressed genes. The remaining genes belonged to the stage-dependent class (53.1%) showing dynamic activity during zebrafish ontogeny ( and ).
Applying the ‘constitutive versus stage-dependent' classification to the TBP microarray gene response groups revealed that genes that require TBP for their activation were predominantly stage-dependent (77%, ), whereas upregulated genes in TBP morphants showed the opposite tendency. The low-variable group of genes did not show a bias to either stage-dependent or constitutively expressed genes. These results indicate that TBP-dependent activation tends to be a property of genes that show dynamic activity during ontogeny. Moreover, TBP tends to negatively regulate steady-state levels of constitutively active genes.
TBP could influence steady-state levels of mRNA in zebrafish embryos both through transcriptional as well as post-transcriptional processes. To address the former, we tested 23 GFP constructs using promoters of zebrafish genes expressed at the sphere/dome-stage and representing various gene ontology classes (). TBP-dependent promoter activation was evident for seven promoters, including the gene promoter ( and and ). This result is consistent with the proposed role of TBP in activating zygotic transcription of many genes during development. On the other hand, 12 promoters, including the gene promoter, did not show significant changes of activity upon loss of TBP function ( and ). TBP independence of transcription was further confirmed by its mRNA levels () and the utilisation of its TSS (data not shown) in TBP morphants. No correlation was found between known promoter motifs (such as TATA boxes, CpG islands, etc.) and TBP response (data not shown).
Several promoters (4 out of 23) showed a clear increase of promoter activity upon loss of TBP, including the 1.4-kb promoter of the gene ( and ). This finding suggests negative regulatory role of TBP on the gene promoter and is in line with the inverse correlation between mRNA and TBP protein levels at the late blastula and early gastrula stages (; ). Co-injection of a synthetic TBP MO-resistant () mRNA, but not of bacterial control mRNA rescued the epiboly movements of the animal cap (, bright field view) and activity (, fluorescence views). Finally, the injection of TBP MO2 resulted in comparable effects to TBP MO both in blocking epiboly movements and in the increased activity of the promoter construct (). These results demonstrate that the specific loss of TBP protein is the reason for the observed upregulation of the promoter in TBP MO-injected embryos.
We classified genes as being maternal or zygotic through another microarray experiment utilizing mRNA pre- and post-MBT; those showing a decrease of mRNA levels from pre- to post-MBT (MBT down) were classified as prevalently maternal and vice versa (MBT up) for prevalently zygotic ones ( and ). We then compared this experiment with the TBP morphants data set, which resulted in an overlap of 131 genes ( and ). The overlap showed that maternal mRNAs were enriched among the upregulated genes of TBP morphants (, MBT down) and the inverse was observed for zygotic mRNAs (, MBT up). A side by side hierarchical clustering analysis of gene activity fold changes in the MBT experiment versus the TBP MO experiment demonstrates further the inverse correlation between the levels of mRNAs before or after MBT as compared to mRNA levels in TBP morphants versus controls ().
We further verified our findings by intersecting the TBP MO microarray experiment with an independent set of 622 maternal mRNAs (), which resulted in an overlap of 143 genes (). As shown in , maternally inherited transcripts were significantly (-value 1.043e−11) enriched among mRNAs upregulated in TBP morphants and underrepresented in the downregulated gene set (-value 3.483e−5). Together, these results suggest that the upregulation of genes observed in TBP morphants could be in large part due to the specific loss of degradation of many maternal mRNAs.
To validate the predicted involvement of TBP in the degradation of maternal mRNAs, we investigated the fate of individual maternal mRNAs. is a maternally expressed gene (), which is upregulated 2.51-fold in TBP MO embryos. We analysed the expression of in wild-type and TBP-morphant embryos at regular intervals for the first 6 h of development by whole-mount hybridisation (WISH). We found high levels of expression in fertilised wild-type eggs and early embryos before MBT, followed by a sharp decrease soon after the MBT, followed by a slight increase at the dome stage (). In contrast, in TBP morphant embryos, mRNA levels showed similar levels throughout early development, consistent with the assumption that degradation of maternal mRNA was impaired. We verified that the lack of degradation of mRNA in TBP morphants was not due to a general delay in embryo development by observing the expression of two zygotically expressed genes: the TBP-independent gene (), which correctly initiated transcription after the sphere stage in TBP morphants (); and the TBP-dependent (), whose activity was lost in TBP morphants (). These results suggest efficient depletion of TBP at dome stage. To further verify the defect in maternal mRNA degradation, RT–PCR analysis was carried out on several maternally expressed genes that showed upregulation in TBP morphants. and (expressed both maternally and zygotically; ) showed elevated levels at dome stage in comparison to c MO-injected embryos, suggesting loss of degradation of maternal mRNA, as opposed to the control gene β, which did not show a change in its steady-state levels (, lanes 1–4). RT–PCR analysis of zygotic genes was also carried out; the TBP-independent showed no change in its mRNA levels, whereas zygotic activity of dropped (data not shown) as shown previously by WISH.
To test directly the fate of mRNAs deposited in the egg, we utilised a synthetic mRNA microinjected into the fertilised eggs (, ()). This mRNA could be readily distinguished from endogenous by reducing the cycles in the RT–PCR reaction (, compare lanes 1–2 to 5–6 of ()). Microinjected mRNA was more efficiently degraded in c MO- than in TBP MO-injected embryos (, compare lanes 6 and 8) and similar results were obtained by WISH (data not shown). Thus, the apparent increase of mRNA levels in TBP morphants is not due to premature activation of zygotic expression, but due to the loss of degradation of mRNAs.
To verify the specificity of the maternal mRNA degradation phenotype to loss of TBP protein function, the ability of a MO-insensitive TBP mRNA to rescue the phenotype in TBP MO-injected embryos was tested. TBP MO and () co-injected embryos were split after injection and separate batches were injected for a second time either by mRNA or mRNA. Expression of recombinant TBP resulted in increase of degradation of (, compare lanes 7, 8 with 9, 10) as well as that of microinjected synthetic mRNA. In contrast, control mRNA did not result in rescue of the degradation phenotype of TBP morphants (, compare lanes 8 and 10). These results demonstrate that the effect of TBP MO on maternal mRNA degradation is directly attributable to the loss of TBP protein function.
Little is known about the mechanisms of maternal mRNA degradation in zebrafish, however, it is likely to involve several maternal as well as zygotic transcription-dependent mechanisms. Not all maternal mRNAs were degraded in TBP morphants (). This may be due to different regulatory mechanisms acting in parallel during maternal mRNA degradation. To investigate this further, we verified the kinetics of mRNA degradation by exploiting a published microarray data set on maternal mRNAs () and compared it to our TBP morphant data set. We established three classes of mRNAs based on the time of their degradation (see and ): a ‘fast' group of mRNAs, which degrade transcription independently or in a transcription dependent manner immediately after initiation of zygotic transcription; a ‘medium' group which is mostly degraded after MBT by early gastrula stage; and a ‘late' group degraded during neurulation and somitogenesis. The comparison of these groups with the TBP-morphant experiment () showed that maternal mRNAs upregulated in TBP morphants follow the pattern of expression dynamics of the ‘medium' group (-value=2.205e−06). TBP-dependent maternal mRNAs showed minimal degradation until MBT (3 h post-fertilisation (hpf)) and accelerated degradation by early gastrulation (4.5 hpf), suggesting that zygotic transcription-dependent mechanisms are involved in their degradation.
These results suggest that TBP is only acting on a subset of mRNAs and that these mRNAs require transcription for their degradation. To test the transcriptional requirement for degradation of maternal mRNAs, we treated embryos with α-amanitin at a concentration that inhibits Pol II activity () and carried out RT–PCR analysis of gene expression. High levels of , and mRNAs were retained after MBT in α-amanitin-injected embryos similarly to TBP morphants (), demonstrating that these mRNAs require transcription and TBP for their degradation. These results taken together with the results obtained using synthetic mRNAs () suggest that the sustained levels of maternal mRNAs in TBP morphants are due to the inhibition of maternal mRNA degradation rather than ectopic activation of zygotic transcription of the respective genes. Not all maternal mRNAs require zygotic transcription (and TBP) for their degradation as confirmed by the fact that degradation of the gene is unaffected by α-amanitin and TBP MO () and its degradation is primarily mediated by mechanisms acting before MBT (). Thus, TBP appears to function within a transcription dependent mechanism directing the degradation of a subset of maternal mRNAs eliminated in a tight time window after MBT during early gastrulation.
Recently, a novel mechanism for maternal mRNA degradation has been described, which involves the zygotically transcribed miR-430 microRNA (). Importantly, the maternally inherited and transcripts have been shown to be targets of miR-430 regulation (), suggesting a potential link between TBP and miR-430 function in mRNA degradation. Therefore, we explored the relationship between miR-430- and TBP-dependent mRNA degradation mechanisms.
We first addressed the question whether TBP-dependent maternally inherited transcripts represent targets of miR-430-mediated mRNA degradation in general. We compared the overlap between experimentally verified miR-430 target genes () and our TBP morphant microarray gene sets (see , and Materials and methods). As shown in , a significant enrichment of miR-430 target genes among the upregulated genes of TBP morphants was observed (=0.002074). The proportion of miR-430 target genes was found to be higher among TBP-upregulated genes (20%) than among maternal genes in general (14%, =0.0276). This difference is significant also after 100 randomisation experiments in which we randomly selected 100 maternal genes (15% s.d.=3, =0.0507). This suggests that the enrichment of miR-430 targets among the upregulated genes of TBP morphants is not simply reflecting the high proportion of maternal genes among miR-430 targets and upregulated genes in TBP morphants.
Subsequently, we have checked the degradation patterns of miR-430 target genes by analysing the overlap between the maternal genes with known degradation kinetics () and miR-430 target genes (). The degradation kinetics of miR-430 target genes largely but not exclusively overlap with that of the maternal genes degraded by TBP-dependent mechanisms (). Taken together, these results indicate a correlation between miR-430- and TBP-dependent mRNA degradation.
We next tested whether miR-430-dependent mRNA degradation requires TBP function. Embryos were injected at the zygote stage with a combination of synthetic mRNAs containing UTR sequences with or without miR-430 target sites () together with TBP MO or c MO, and mRNA distribution was detected by WISH before and after the MBT. A synthetic mRNA containing the 3′ UTR region from SV40 that lacks miR-430 target sequences was not degraded until the 50% epiboly stage ( and ), suggesting that they are not degraded by the miR-430 pathway. In contrast, mRNA linked to the UTR from the gene containing a miR-430 target site is degraded by the 50% epiboly stage in c MO-injected embryos, but not in TBP MO-injected embryos (). Similar results were obtained with mRNA fused to the 3′ UTR sequences of the and genes containing miR-430 target sites or when only the miR430 target site sequences were added to (). These results suggest that TBP is required for miR-430-dependent degradation of several mRNAs. Upon injection of a transcript containing a mutated 3′ UTR sequence of lacking the miR-430 target site, the mRNA became insensitive to degradation until the 50% epiboly stage (), indicating that the degradation of these synthetic mRNAs are indeed miR-430-dependent. The defects in miR-430-dependent degradation of mRNA in TBP morphants was also rescued by overexpression of recombinant TBP. Co-injection of synthetic mRNA with and TBP MO resulted in reversal of the mRNA degradation phenotype of TBP morphants, indicating that the mRNA degradation effects relate directly to loss of TBP ( and ). As miRNA genes are known to be transcribed by polymerase II () miR-430 may be a candidate target of TBP-dependent transcription regulation. However, miR-430 expression in TBP morphants is unaffected (), suggesting that TBP functions downstream of miR-430 production in the mRNA degradation process. We investigated further whether TBP is involved in general microRNA function or the mir-430 pathway specifically. Thus, we co-injected miR-430 and miR-1 with their respective target mRNAs () with TBP and c MO. miR-430-mediated mRNA degradation was blocked in TBP morphants, as opposed to that by miR-1, confirming the specificity of TBP function to a subset of miRNA-dependent processes ().
Given the tight temporal control of TBP-dependent mRNA degradation, we hypothesised that those miR-430 target mRNAs are degraded in a TBP-dependent manner, which are eliminated at a ‘medium' rate during late blastulation/early gastrulation. To test this hypothesis, we utilised the overlap between miR-430 target genes and maternal genes and identified those, which show either medium or fast degradation (). Then we tested both fast- and medium-degrading miR430 target genes for TBP dependence of their degradation. The results indicate that medium-degrading mRNAs are more likely to be TBP-dependent (increased accumulation in TBP MO injected embryos) than fast-degrading mRNAs (5 of 7 versus 1 of 6, respectively, ). Taken together, our results indicate that TBP is specifically required for the degradation of a subset of miR-430-dependent mRNAs that are eliminated at late blastula/early gastrula stages.
s
u
m
m
a
r
y
,
w
e
h
a
v
e
d
e
m
o
n
s
t
r
a
t
e
d
a
d
i
f
f
e
r
e
n
t
i
a
l
r
e
q
u
i
r
e
m
e
n
t
f
o
r
T
B
P
i
n
t
h
e
r
e
g
u
l
a
t
i
o
n
o
f
m
R
N
A
l
e
v
e
l
s
i
n
t
h
e
e
a
r
l
y
z
e
b
r
a
f
i
s
h
e
m
b
r
y
o
a
n
d
p
i
n
p
o
i
n
t
e
d
t
h
r
e
e
l
e
v
e
l
s
o
f
r
e
g
u
l
a
t
o
r
y
a
c
t
i
v
i
t
i
e
s
a
s
s
o
c
i
a
t
e
d
w
i
t
h
T
B
P
f
u
n
c
t
i
o
n
.
T
h
e
a
p
p
r
o
a
c
h
t
o
b
l
o
c
k
T
B
P
f
u
n
c
t
i
o
n
b
y
M
O
o
l
i
g
o
n
u
c
l
e
o
t
i
d
e
s
r
e
s
u
l
t
e
d
i
n
t
h
e
e
f
f
i
c
i
e
n
t
d
e
p
l
e
t
i
o
n
o
f
T
B
P
i
n
t
h
e
e
m
b
r
y
o
b
e
f
o
r
e
t
h
e
M
B
T
,
a
n
d
t
h
u
s
a
l
l
o
w
e
d
t
h
e
d
e
t
e
c
t
i
o
n
o
f
t
h
e
e
a
r
l
i
e
s
t
e
f
f
e
c
t
s
o
f
l
o
s
s
o
f
T
B
P
p
r
o
t
e
i
n
o
n
z
y
g
o
t
i
c
t
r
a
n
s
c
r
i
p
t
i
o
n
a
n
d
a
s
s
o
c
i
a
t
e
d
m
a
t
e
r
n
a
l
d
e
g
r
a
d
a
t
i
o
n
p
r
o
c
e
s
s
e
s
.
T
h
e
e
x
t
e
n
s
i
v
e
u
s
e
o
f
a
s
e
c
o
n
d
T
B
P
t
a
r
g
e
t
i
n
g
M
O
a
n
d
r
e
s
c
u
e
e
x
p
e
r
i
m
e
n
t
s
w
i
t
h
a
M
O
-
i
n
s
e
n
s
i
t
i
v
e
r
e
c
o
m
b
i
n
a
n
t
T
B
P
i
n
t
h
i
s
s
t
u
d
y
p
r
o
v
i
d
e
d
s
t
r
o
n
g
e
x
p
e
r
i
m
e
n
t
a
l
v
e
r
i
f
i
c
a
t
i
o
n
o
f
t
h
e
s
p
e
c
i
f
i
c
i
t
y
o
f
o
u
r
k
e
y
f
i
n
d
i
n
g
s
t
o
l
o
s
s
o
f
T
B
P
f
u
n
c
t
i
o
n
.
Wild-type embryos (Tubingen AB) were collected after fertilisation and dechorionated by pronase treatment as described previously (). The TBP-specific MO (TBP MO) used was described previously (). TBP2 MO was CAAAAGACGTAAACGATAATTCGCA. Embryos were injected also with a c MO with five mismatches relative to the TBP-specific MO (GACGTACGCTGTTCTTCTCCTCGAT) or a standard c MO (CCTCTTACCTCAGTTACAATTTATA) provided by Gene-tools LLC (Phylomath, USA). MOs were injected into the cytoplasm of zebrafish embryos at the one cell stage. A total of 1200 embryos were collected for microarray analysis at the dome stage. Embryos were co-injected with combinations of MOs and mRNAs for analysis of gene expression. For synthetic mRNA production, the was used containing as a I–I fragment of the biologically active core domain (aa 104–297) () of the TBP (). The plasmid in a pCS2+ vector with an gene was used to make control mRNA (). Both mRNAs were injected at 100 ng/ml concentration. mRNAs were produced by transcription of linearised plasmids using the mMESSAGE mMACHINE Kit (Ambion, UK). Plasmids for the production of synthetic mRNAs with various miR-430 target sequences were described by . In co-injection experiments, MOs were co-injected with 20 ng/ml mRNAs followed by injection with or mRNAs.
Twenty-three promoter constructs were made which contained 1–2 kb upstream and 30–300 bp downstream sequence around the predicted (9 promoters) or 5′RACE verified (14 promoters) transcriptional start site (TSS) linked to a GFP reporter. Promoter fragments were amplified from genomic DNA using TripleMaster Enzyme Mix (Eppendorf, Germany) and cloned into the vector (Invitrogen, Germany). The promoter constructs were injected into the cytoplasm of one-cell stage embryos. At 50% epiboly, embryos were fixed overnight in 4% PFA. TBP-dependence analysis of promoters was carried out by co-injecting MOs in a concentration of 1 mM with 100 ng/μl of mRNA. For recombinant TBP rescue experiments, the promoter was PCR-amplified and subcloned into the + replacing the CMV promoter by using I and I sites. Yellow fluorescence protein was visualised by using a Leica MZ16F fluorescent stereo microscope and digital image recording. Western blotting with the 3G3 antibody against TBP was carried out as described previously ().
Fixed embryos were incubated overnight at 4°C with wild-type GFP (1:500) (Torrey Pines Biolabs, USA) or cycle3 GFP antibodies (1:200) (Invitrogen, Germany). Embryos were incubated with goat-anti-rabbit HRP secondary antibody in PBT (1:500) (DakoCytomation, Denmark) for detection by diaminobenzidine solution according to manufacturer's instruction (Vector, USA). An LNA oligonucleotide probe was used in WISH to detect miR-430 miRNA as described previously (). WISHs were carried out with standard protocols ().
Analysis of a selected set of genes by RT–PCR was carried out using 50 embryos for each treatment group. Microinjected and wild-type embryos were collected at the 512-cell stage and shield stage. Total RNA was extracted using Trizol (Invitrogen) following manufacturer instructions. A 1 μg portion of total RNA was reverse transcribed (M-MLV Reverse Transcriptase, Promega, Germany) and PCR was carried out from several targets using the oligonucleotide primers specified in . PCR products were separated in a 2% agarose gel.
Three developmental stages around MBT were analysed on microarrays containing the primary set of 16 177 oligos (16 k set). RNA was obtained from freshly laid eggs (zygotic target), 1k-stage embryos (1k target) and 30% epiboly embryos (30% epiboly target). The following hybridisations were performed with dye swap each: zygotic versus 1k, zygotic versus 30% epiboly and 1k versus 30% epiboly. Microarray chips representing 10 501 nonredundant Genbank ESTs of the primary 16 416 set of 65-mer oligonucleotides were used in the subsequent hybridisations. Two independent hybridisations were carried out for three biological repeats of treatment groups (TBP MO- versus c MO-injected embryos), resulting in a maximum of 12 data points per gene. The reliability of the fold changes was assessed using a regularised -test () and the adjustment of -values to control the false discovery rate (FDR) () selecting genes with FDR smaller than 0.05 (minimum 5 data points). This cut-off value of FDR is the minimal widely used () and results in a good balance of quality versus quantity in the selected gene lists.
Several meta-analyses of different, unrelated microarray experiments were carried out as has been described previously (). Gene sets of the TBP morphant microarray were compared with existing gene sets () generated on the same platform (Compugen microarray), with the experiment GSE4201, () and the experiment E-TABM-33 from (Konantz M, Otto G-W, Weller C, Saric M, Geisler R. Microarray analysis of gene expression in zebrafish development, manuscript in preparation) executed on the Affymetrix platform (See for details). The outdated gene annotation by Compugene () was updated by converting all Entrez Nucleotide identifiers of the probes to their respective Unigene identifiers using a custom Perl script and release 91 of Unigene. The heatmap and hierarchical clustering of microarray experiments was created using the R programming language version 2.2.0. The clustering of fold changes is based on the ‘complete linkage' method provided by R. See for details.
Maternal genes present in the TBP-knockdown and miR-430 targets gene sets () were identified by utilizing an existing data set of transcripts accumulated in the unfertilised egg (). The variation of steady-state mRNA levels over developmental time was determined by comparing fold changes as described in () and was visualised by plotting their values in the unfertilised egg and at 3, 4.5 and 6 hpf). For a clustering of maternal mRNAs into fast-, medium- and slow-degrading subgroups the following criteria were used: fast-degrading, >200% decrease of fold change from 0 to 3 hpf; medium–degrading, <100% from 0 to 3 hpf, >100% from 3 to 4.5 hpf; slow-degrading, <50% decrease of fold change till 4.5 hpf).
We downloaded the raw data of the experiments (; wild type, MZ Dicer and MZ Dicer miR-430-injected) by the GEO database (). The probe summaries were generated by using RMA from the BioConductor collection of packages. The Limma package was used to perform the statistical analysis and select differentially expressed genes. The following comparisons were made: MZ Dicer versus wild type; MZ Dicer versus MZ Dicer miR-430–injected; wild type versus MZ Dicer miR-430-injected. We selected as differentially expressed genes all the genes resulting in at least one comparison with a corrected -value ⩽0.05. All the genes significantly upregulated in MZ DICER compared to both wild type and MZ DICER miR-430-injected were considered as potential miR-430 target (1650 probes). |
Prions are the causative agents of neurodegenerative diseases, which include bovine spongiform encephalopathy (BSE) in cattle, scrapie in sheep, chronic wasting disease in mule deer and elk, and Creutzfeldt–Jakob Disease (CJD) in humans. The infectious agent is believed to consist of improperly folded forms of a host-encoded protein, the cellular prion protein (PrP). Conversion of PrP into the disease-associated isoform, PrP, is thought to be the primary pathogenic event, although the mechanisms by which PrP causes disease are poorly understood. PrP is absolutely required for disease progression, as PrP knockout () mice do not succumb to disease and do not propagate infectivity following intracerebral challenge with infectious prions (, ).
The mammalian prion protein family currently consists of two proteins: PrP which is expressed at high levels in the central nervous system (CNS), and Doppel (Dpl), a molecule with a similar three-dimensional structure, whose postnatal expression is normally confined to the testis (; ). Whereas a role for Dpl in the proper functioning of the male reproductive system has been confirmed in two lines of Dpl knockout mice (; ), the function of PrP, a well-conserved neuronal glycoprotein, comprises a conundrum, in part because phenotypic alterations in mice have been subtle or disputed. One emerging area of consensus concerns a protective effect of PrP against neuronal insults. In particular, PrP is upregulated following ischemic brain damage, in both humans and mice (; ). PrP deficiency in mice increases infarct size following cerebral artery occlusion and increases caspase 3 activation (), and PrP overexpression improves neurological behavior and reduces infarct volume in a rat stroke model (). Another widely observed and perhaps related phenomenon involves the ability of PrP to protect against a variety of proapoptotic stimuli (; ; ; ; ; ; ). Strong evidence for a neuroprotective activity for PrP against apoptosis has come from studies of transgenic (Tg) mice expressing internally deleted forms of PrP or wild-type (wt)Dpl within the CNS. The presence of Dpl in the brain of mice leads to a neurodegenerative syndrome characterized by a profound apoptotic loss of cerebellar cells (; ; ). A similar phenotype is observed when N-terminally truncated versions of PrP (ΔPrP) are expressed in the brain (). Remarkably, both syndromes are abrogated by the coexpression of wt PrP. Recently, it has been shown that a smaller deletion restricted to the well-conserved central domain of PrP is sufficient to elicit a highly toxic phenotype in mice (; ). The above studies have led to a model in which Dpl and ΔPrP initiate aberrant signaling through a hypothetical prion ligand termed L, a process which is blocked by PrP binding (). Assuming that the interaction between PrP and L represents an essential physiological event, the authors also proposed the existence of a PrP-like protein termed π, which binds to L and is capable of compensating for the absence of PrP in mice. To this date, no candidates for π (or L) have been put forward. However, an open reading frame was discovered which, when translated, exhibits homology to the central hydrophobic domain in PrP. This gene, denoted (‘shadow of the prion protein'), is present from zebrafish to humans and is predicted to encode a short protein, Shadoo (Sho) (). is located on chromosome 7 in mice, away from the gene complex on chromosome 2.
Building on the genetic interaction between PrP and Dpl or ΔPrP, we have established an assay for PrP activity in primary cultures of cerebellar granule cells (, and references therein). Here, cerebellar granule neurons (CGNs) cultured from mice are transfected with plasmids encoding Dpl or PrP alleles of interest and individual apoptotic events scored. This assay recapitulates the phenotypes produced by multiple PrP alleles in Tg mice, including neurotoxicity of both Dpl and ΔPrP, and neuroprotective activity of PrP against the toxicity elicited by either Dpl or ΔPrP. In conjunction with biochemical and histological analyses, we have used the CGN assay to explore the properties of Sho. We now demonstrate that Sho is a GPI-anchored neuronal glycoprotein present in the CNS from early postnatal life. Not only is Sho PrP-like in its ability to protect against both Dpl and ΔPrP toxicity in the CGN assay, but it is also strikingly reduced in prion infections.
Motivated by the absence of obvious phenotypic defects in adult mice, we considered proteins that might overlap functionally with PrP. One criterion was evolutionary conservation. In this regard, bioinformatic analyses by Premzl and co-workers have yielded an interesting candidate in the shape of the open reading frame present in genomic DNA of species from mammals to fish (, ; ) and which, unlike the hypothetical gene (; ), is present in the mouse genome. The architecture of the predicted Sho protein loosely resembles the flexibly disordered PrP N-terminus (). Analyses of expression by RT–PCR () and interrogation of expressed sequence tag databases (UniGene) imply expression in embryonic and adult neuronal tissue as well as the retina and visual cortex, but have yet to document spliced mRNAs. Consequently, we evaluated Sho as a putative third member of the prion gene family.
To address expression at the protein level, we raised antibodies against Sho. Two antisera (‘04SH-1' and ‘06SH-3') were against a mouse Sho peptide consisting of residues 86–100, whereas a third (‘06rSH-1') was against full-length recombinant mouse Sho(25–122) expressed in () and recognizes an N-terminal epitope contained within residues 30–61 (). Assessed by Western blot of tissue lysates, 06rSH-1 was virtually devoid of cross-reactive species (). Cross-reactive species of molecular weights incompatible with authentic Sho were present in analyses with antisera 04SH-1 and with 06SH-3, but these had varying intensities and/or different molecular weights for the two antisera. Consequently, the following comments are restricted to signals detected by two or more varieties of α-Sho antibodies.
Similar to PrP, murine Sho is revealed as being expressed at the cell surface, -glycosylated and sensitive to the GPI anchor cleaving enzyme phosphatidylinositol-specific phospholipase C (PI-PLC) in transfected N2a neuroblastoma cells (). Following PNGaseF treatment, full-length Sho has a molecular weight of 9.1 kDa as assessed by SDS–PAGE (predicted 9.5 kDa). In addition to the full-length protein, a fraction of the protein in transfected N2a cells has a faster electrophoretic mobility (). In PrP, a well-documented physiological ‘C1' cleavage occurs just before the hydrophobic tract at residues His111 and Met112 (human PrP numbering scheme, underlined; ) both in cultured cells and in the adult brain (; ; ). Since cell lysates for these analyses were prepared in the presence of protease inhibitors, it is possible that an analogous endoproteolytic processing could figure in the biogenesis of Sho. A detailed description of truncated forms of Sho will be presented elsewhere (Coomaraswamy , in preparation).
Analysis of mouse tissue by Western blotting defined a predominant glycosylated protein species of similar molecular weight to transfected full-length Sho (approximately 18 kDa), and one that is developmentally regulated, appearing at embryonic day 16 and persisting in early postnatal life and in the brains of adults (). PNGaseF-sensitive bands of identical molecular weights were observed with two independent Sho antibodies, confirming the authenticity of the bands and the presence of Sho in the mouse CNS. Sho signal was emphasized in membrane-enriched preparations from adult mouse brains, corroborating its membrane anchorage (). In Western blots performed with α-Sho86–100 antisera, Sho signal increased following treatment with PNGaseF, suggesting that -glycosylation at Asn107 partially occludes antibody binding at the adjacent epitope for the peptide-directed antiserum.
hybridization using antisense-strand riboprobes prepared against the mouse open reading frame (but not sense-strand controls) yielded signals in the adult mouse CNS. Analyses of the hippocampus and cerebellum revealed prominent signals in the cell bodies of pyramidal cells and Purkinje cells, respectively ( and ). By way of comparison, has a broader pattern of neuronal expression (; ). Immunohistochemistry for Sho protein yielded prominent signals in the same cell types defined by hybridization ( and ), that is, hippocampal neurons and cerebellar Purkinje cells. In the case of antisera 04SH-1, these signals were absent when antibodies were preincubated in a solution containing the Sho86–100 peptide immunogen ( and ). Besides defining Sho as the ‘second' cellular prion protein present in neurons of the adult CNS, these data define intracellular transport phenomena, as immunohistochemical signals were present in cell processes in addition to the cell bodies detected by antisense riboprobes (i.e., predicted to contain Sho mRNA). In the case of Purkinje cells, immunostaining was present not only in cell bodies but also prominent in their processes, specifically in the dendritic arborizations present within the molecular layer of the cerebellum ( and , signals detected with all three antisera). A related phenomenon was observed in the hippocampus, notably in CA1 pyramidal neurons. Here, Sho immunoreactivity was absent from axonal projections (with all three α-Sho antibodies), present in cell bodies (seen by all three α-Sho antibodies), and notable in the apical dendritic processes located in the stratum radiatum of the hippocampus (strong signals with 04SH-1 and 06SH-3, and a less intense signal with 06rSH-1) ( and ).
Since PrP has been reported to possess an unusual property for a GPI-linked protein, the ability to undergo basolateral (dendritic) sorting in polarized cells in distinction to the more typical apical (axonal) sorting, parallel analyses were undertaken for PrP and Sho. PrP was examined at basal levels from the (endogenous) gene, or expressed from a cosmid transgene of the hamster PrP gene including 25 kb of sequences 5′ to the transcriptional start-site (; ). In the case of wt mouse PrP, analyses were performed with 7A12 antibody () using mice as negative controls ( and ), whereas non-Tg wt mice served as negative controls for the use of the hamster PrP transgene-specific 3F4 antibody ( and ). Besides the anticipated differences in signal levels between wt and Tg(SHaPrP)7 mice, the PrP-directed antibodies yielded similar expression patterns, with widespread staining throughout the brain. PrP was underrepresented in the cell bodies of the pyramidal neurons of the hippocampus but abundantly present in the stratum oriens containing the axonal projections, and also in the radiatum, lacunosum moleculare and molecular layers (). In the cerebellum, mouse PrP was absent in wt Purkinje cells (as described previously (; )) and hamster PrP was represented in some but not all Purkinje cells of Tg(SHaPrP)7 mice, and at a level lower than that of the surrounding neuropil (). On the other hand, PrP was present in the granule cell layer and abundant in the molecular layer. Strikingly, the abundant signal present in the radiatum layer was not totally uniform, and—by virtue of a negative staining effect—revealed the outlines of Purkinje cell dendritic arborizations in the cerebellum and the apical dendritic of CA1 neurons (): these are structures with marked immunostaining with α-Sho86–100 (04SH-1 and 06SH-3) and α-rSho (06-rSH1); ). These data therefore define a complementary ‘interlocking' aspect to PrP/Sho protein expression in the cerebellum and suggest a similar effect in the hippocampus.
Although the most prominent Sho signals were obtained in the hippocampus and cerebellum (), signals for both mRNA and Sho protein were also present in other areas of the brain including the cerebral cortex, the thalamus, and the medulla (; ). Coincidence with neurofilament staining indicates that neurons are the prominent site of Sho expression (). PrP signal was abundant in these regions (, and data not shown), confirming that Sho and PrP expression profiles overlap in certain areas of the brain. A further example of this phenomenon is apparent in the adult retina (; Chishti , in preparation). In overview, Sho expression within the adult mouse brain is either less widespread than PrP, or basal Sho levels in certain areas fall below the detection threshold of our current procedures. Lastly, to assess whether there is a physiological cross-regulation effect between PrP and Sho, we performed a number of analyses on mice (). These failed to reveal a clear upregulation of Sho in response to PrP deficiency.
Two N-terminal PrP modules, the charged region and the copper-binding octarepeats, contribute to neuroprotective activity in a neuronal assay (). However, while others analyzed the same N-terminal modules, they highlighted a trafficking effect (), so we sought determinants beyond these potential delivery signals. The C-terminus of PrP comprises a three-helix bundle () similar to Dpl (), and in the form of alleles such as PrPΔ32–121 (), exhibits Dpl-like toxicity. We therefore focused on the ‘remaining' area of PrP, residues 91–121 between the octarepeats and the structured domain () containing the hydrophobic domain (HD) with Sho homology. The deletion alleles were tested for their ability to engender stable forms of PrP. As anticipated from prior analyses of in-frame deletions (; ; ), these forms of PrP were synthesized at levels similar to wt PrP, and also underwent similar maturation (; ) ().
Residues 100–105 are poorly conserved () and in the erebellar ranule euron (CGN) cellular assay (), a PrPΔ100–105 allele exhibited protective activity like wt PrP (<0.001) when measured against Dpl. Conversely, PrPΔ91–99 and PrPΔ106–112 alleles had decreased protective activity (neither is significantly different from Dpl alone; ). PrPΔ106–121 transfected alone exhibited partial neurotoxic activity approaching that of a toxic PrPΔ32–121 control allele (), but different from that of wt PrP (<0.01) (). These data offer a parallel to assays in Tg mice. Here, deletions invading the PrP central region from residues 91 to 121 produce proteins that are progressively toxic: no significant pathological abnormalities are noted in Tg mice expressing PrP Δ32–80, Δ32–93, and Δ32–106, but mice expressing Δ32–121 (or Δ32–134) exhibit profound loss of cerebellar cells (; ), as do mice expressing PrPΔ94–134 or Δ105–125 (; ). From this we infer that a third determinant of neuroprotection lies between PrP residues 91 and 99, and a complex or modular activity determinant (‘determinant 4') lies between residues 106–121, whereby a Δ106–112 deletion loses protective activity and a larger Δ106–121 deletion exhibits partial neurotoxic activity. These data extend observations made in hippocampal neurons, where PrP residues 95–124 were implicated in neuroprotective action versus Dpl ().
Sequence identity between Sho and PrP occurs in the central hydrophobic region. Since this area is highly conserved in PrP () and is predicted to overlap or be immediately adjacent to determinants of neuroprotective action ( and ), we tested Sho in a functional assay in CGN cultures. Remarkably, a wt mouse Sho cDNA was PrP-like in its ability to block the proapoptotic action of cotransfected Dpl () (<0.001 versus Dpl alone, no significant difference between wt PrP alone versus Dpl+Sho). Since the putative PrP-like activity called π was inferred from the action of ΔPrP, rather than the action of Dpl (), further experiments were carried out to assess the effect of Sho upon the toxic action of a ΔPrP allele (). In these cotransfections, wt Sho reduced the toxic activity of singly transfected PrPΔ32–121 to a baseline defined by ‘empty vector' (pBUD.GFP) controls (<0.001), and overlapping expression and colocalization was apparent in transfected neurons (). Besides adding to the similarities between Sho and π, these data underscore the notion that ΔPrP and Dpl exert toxicity through the same pathway.
Given the similarity between Sho and PrP in the HD, this domain was a strong candidate for the ‘active site' of Sho. Consequently, we assessed the protective activity of Sho bearing a precise HD deletion. Here, the cotransfected ShoΔ62–77 allele was not protective against Dpl (not significant versus Dpl alone) (). wt and Δ62–77 Sho alleles are expressed at similar levels in bulk-transfected cells, and with immunoreactivity present upon the surface of transfected N2a cells (; ). In contrast to a wt control allele, the ShoΔ62–77 allele also did not exert significant protection against a toxic PrPΔ32–121 allele (<0.001 versus baseline of pBUD.GFP empty vector; ).
As an extrapolation from results presented above and from prior analyses (; ; ), the ability of Dpl to cause cell death in CGNs implies lack of a compensatory neuroprotective activity (e.g., Sho) in these cells. To test this inference, and as an additional control for immunohistochemical analyses presented in and , we assessed Sho expression in the lysates of purified CGNs. Although Sho was readily detected in brain lysate, the analysis failed to detect a comparable signal in a normalized loading of CGN lysate ().
To assess Sho in a disease setting, Western blots were performed on brain homogenates from wt mice infected with the RML isolate of mouse-adapted prions (). Two independent cohorts produced in different laboratories were used for this analysis. Assessed in the clinical phase of disease, animals from both cohorts demonstrated a striking reduction in Sho protein levels (). Notably, this decrease in signal is apparent using two distinct Sho antibodies with independent epitopes. Normalized against Thy-1, another GPI-anchored neuronal protein, Sho signals of RML-inoculated brain extracts versus brain extracts from healthy controls were reduced eight-fold (). A variety of other neuronal proteins were probed as internal controls (). Although mild loss of signal was apparent for some of these—perhaps a direct consequence of neuronal damage and death—in no instance did these reductions approach those seen for Sho. As a further control we assessed Sho levels in a widely used animal model of Alzheimer's Disease, TgCRND8 mice expressing a mutant β-amyloid precursor protein transgene (). Using mice in clinical phase of the disease (marked by overt amyloid deposits, normalized Aβ levels equaling those of sporadic Alzheimer's disease, and profound cognitive impairment), no difference was apparent between TgCRND8 and non-Tg mice with regards to Sho levels (). Thus, depletion of Sho is absent in non-infectious CNS amyloidosis.
Our first conclusion is that Sho is a bona fide neuronal glycoprotein and the third member of the mammalian prion protein family. Beyond mere sequence homology, Sho demonstrates a number of biochemical and cell biological properties also exhibited by PrP. These include addition of a GPI anchor, -glycosylation at one or two sites and a cleavage event likely positioned N-terminal to the hydrophobic tract ( and ; ). Furthermore, there are functional redundancies between PrP and Sho measured in CGN cells. Although Sho and PrP have dissimilar N-terminal regions that may tailor their activities in different neuroanatomic sites, we infer that the center of both PrP and Sho comprises a functionally conserved and ancient activity domain contributing to neuroprotection. In PrP-deficient areas of the brain, Sho may be particularly important in filling-in a PrP-like activity. Although the exact relevance of PrP/Sho neuroprotection against exogenous stimuli versus brain physiology remains to be deciphered, a role against neuronal insults such as ischemic damage seems plausible. Indeed, increased PrP expression following stroke has been observed in both humans and mice (; ).
A second conclusion is that the study of Sho may help decipher the molecular mechanisms underlying physiological signaling from the prion protein family. Here, some initial hints can be inferred from mutational analysis. As deletions invade the central region of PrP, not only is protective activity lost, but the mutant PrPs gain spontaneous proapoptotic activity, as shown by loss of the granule cell layer in Tg mice (), and echoed here by the divergent performance of PrPΔ106–112 and Δ106–121 alleles in the CGN assay. It is possible that a gain of proapoptotic function for PrP within CA1 neurons when it is sterically blocked and dimerized by antibodies against residues 95–105 proceeds by a related mechanism (). In any event, the PrP central region defined by genetic mapping (analyzed by CGN assays in this paper, see also ) is natively unstructured in NMR studies (; ) and the homologous region in Sho is likely to be natively unstructured as well (). This region may serve as a site for protein–protein interactions, and in this regard, our results parallel and extend a model for signaling from PrP protein complexes. Besides PrP, this model involves a hypothetical ligand, L, and a conjectural PrP-like protein denoted ‘π' (; ). Two docking sites are posited between L and PrP, and one docking site between Dpl or ΔPrP and L. Based on sequence similarity and domain structure (), biochemical properties (), a low level expression of Sho in the cerebellar granule cells ( and ), and similar effects of wt PrP or Sho versus Dpl or PrPΔ32–121 in CGN assays () (), Sho can be seen to resemble π. In this scheme, sequences within or around the hydrophobic domain of Sho or PrP would comprise a docking site for L, a site associated with a protective effect. In PrP, partial loss of this site may confer spontaneous neurotoxic behavior, consistent with the enhanced toxicity observed in Tg mice expressing PrP with HD deletions (; ). A second site lying within the α-helical domain present in PrP, ΔPrP and Dpl is associated with a toxic effect (i.e., as per ΔPrP and Dpl) unless the first site is also present (as in wt PrP) (; ; ). Even though the original L–PrP–π model may prove incorrect in a cell biological sense (i.e., that signaling may have more to do with protection from toxic stimuli rather than regulation of endogenous trophic signals), it provides a useful framework for deciphering molecular interactions. Certainly the ability to now perform studies with PrP, ΔPrP, Dpl and Sho and assess responses within a cellular assay should aid the identification of authentic PrP interactors.
A third conclusion is that Sho may open a window to pathological signaling events in prion disease. Studies of scrapie-infected rodents support the concept of neuroanatomic ‘target areas' that give rise to the clinical features of disease (). Thus far, the mechanisms by which disease associated forms of PrP (PrP or ‘PrP') might cause dysfunction in these target areas have proven elusive (). For example, whereas a requirement for PrP was inferred from grafting tissue from mice overexpressing PrP into mouse brains and performing prion challenge of chimeric mice (), a genetic reconstitution experiment employing an astrocyte-specific hamster PrP transgene demonstrated an opposite effect, namely neuronal damage in PrP-deficient neurons upon prion challenge (). The concept of pathogenesis , that is, independent of PrP expression, is underlined by prior studies and by experiments herein, which demonstrate that the dendrites of CA1 neurons, an early ‘clinical target area' in prion infections, are largely devoid of PrP (). Susceptibility of dendritic structures in this neuroanatomic area has been demonstrated by several techniques in different models of infectious prion disease (; ; ). We note that these structures are strongly labeled by both Sho peptide directed antisera in wt mice (), and are distorted and attenuated in prion-infected mice (not shown), as already described by others (; ; ).
Strikingly, Western blot analyses of brains from clinically ill prion-infected mice revealed a dramatic reduction of Sho protein (). Because Sho has neuroprotective properties (), it seems reasonable to consider whether a loss of Sho may underlie some of the clinical or pathological features of prion disease. In this regard, future studies utilizing mice should prove informative. The mechanism governing the loss of Sho protein during prion infections remains to be determined. While a transcriptional effect seems unlikely (not shown), protein structure may play a role. Since Sho's conserved region coincides with the epicenter of PrP misfolding to disease-associated forms such as PrP (, ; ), conformational alterations in unstructured, cell-anchored Sho (or in Sho shed into the parenchyma, like other GPI-anchored proteins; ) may occur upon interaction with PrP. However, this does not explain how altered Sho proteins, should they exist, might be more prone to degradation. Expanding the view to include the concept of aberrant signaling may provide a more useful take on this problem. Assuming that Sho exerts a neuroprotective function, perturbations in this protective pathway could destabilize Sho or lead to a loss of Sho-expressing cells and a commensurate drop in protein level. According to this hypothesis, parenchymal PrP might interfere with the physiological protective activity of Sho by virtue of both its aberrant fold and cell biological resemblance to Sho. While new experiments will be required to probe the validity of these concepts and the mechanisms controlling Sho expression, we suggest that Sho is a potential modulator for the biological actions of normal and abnormal PrP.
Protein alignments were created using the T-COFFEE algorithm and EST searches conducted using UniGene. Data sets for all transfected constructs were assessed for normality (>0.10) by the Kolmogorov–Smirnov test and analyzed by one-way ANOVA, Tukey pairwise comparisons, with significance set at <0.05, using Prism software (v4.0c, GraphPad Inc.).
mice and littermates were maintained on a C57/B6 congenic background. Mice for inoculations, either C3H/B6 hybrids (Toronto) or B6 inbred (McLaughlin Research Institute) were inoculated intracerebrally with 30 μl of 0.1% RML-infected brain homogenate diluted in phosphate-buffered saline (PBS) containing BSA (50 mg/ml), penicillin (0.5 U/ml) and streptomycin (0.5 μg/ml). Age-matched non-inoculated or mice inoculated with normal brain comprised negative controls. Clinically ill mice were killed and 10% brain homogenates in PBS were prepared. Proteinase K digestions (50 μg/ml) were performed at 37°C for 1 h. Sho levels were calculated by densitometry (Scion Image) and comparison to a standard curve created from serial dilutions of non-infected brain homogenate. TgCRND8 mice () on a C3H/C57BL6 outbred background were killed at clinical illness (∼8 months) and 10% brain homogenates prepared in 0.32 M sucrose containing protease inhibitors.
The Sho open reading frame was amplified from mouse genomic DNA and inserted between the dIII and I sites of either pcDNA3 or pBUD.CE4. GFP (Invitrogen). PrP and Sho mutants were generated using methods previously described (), with constructs verified by DNA sequencing (details available on request). cDNA encoding residues 25–122 of the mouse protein was amplified from pcDNA3.Sho and inserted between the I and HI sites of the pET-19b vector (Novagen) and expressed in BL21(DE3)/RIL (Stratagene). Following induction and lysis, filtered lysate was subjected to Ni affinity chromatography (Qiagen), eluted with 250 mM imidazole and dialyzed into 20 mM Tns–Hcl pH8.0, 100 mM NaCl. The polyhistidine tag was removed with enterokinase (Roche). Purity was assessed by SDS–PAGE and concentration estimated using bovine serum albumin standards. Far-UV circular dichroism spectra of recombinant Sho in PBS pH 7.0 and a protein concentration of 0.1 mg/ml were collected at 22°C using a Jasco J-715 spectropolarimeter.
The peptide CRRTSGPGELGLEDDE (Sho residues 86–100 plus an N-terminal cysteine) was conjugated to maleimide-activated KLH (Pierce) and injected into New Zealand White rabbits. Alternatively, recombinant Sho was conjugated to KLH using EDC (1-ethyl-3-[3-dimethylaminopropyl] carbodiimide hydrochloride) chemistry (Pierce) and injected into rabbits. Antibodies were precipitated from serum using ammonium sulfate and then affinity purified using the respective peptide (SulfoLink column) or polypeptide (AminoLink Plus Coupling Gel column) immunogens. Epitope positions are shown in .
N2a cells, HEK293 and CGN cells were cultured, transfected and lysed as described previously (), except that staining for an activated caspase 3 neo-epitope (Cell Signaling Technology) was used to quantify apoptotic events and conditioned medium from adjacent wells replaced medium on transfected cells. CGN assays represent two or more independent triplicate transfections into cells, with observers being blind to the genotype of transfected test plasmids. For PIPLC treatment, 24 h post-transfection, Sho-transfected N2a cells were washed three times with PBS, and then treated with PI-PLC (Invitrogen) diluted in PBS for 40 min at 4°C. Cells were then washed twice and lysed as above.
N2a or HEK293 cells 24 h post-transfection were washed with PBS, fixed with 4% paraformaldehyde, washed with PBS, blocked with 2% goat serum, then incubated with either Sho antibody (04SH-1, 6 μg/ml) or PrP antibody (8H4, 1 μg/ml), overnight at 4°C. Following PBS washes, cells were incubated with AlexaFluor488- or AlexaFluor594-conjugated secondary antibodies (Invitrogen; 1:300) for 2 h and then washed three times with PBS. For tissue analysis, mice were saline-perfused. Brains were fixed in Carnoy's fixative (10% glacial acetic acid, 60% methanol, 30% chloroform) bisected and processed. Six-micrometer sections were dried, deparaffinized and taken to TBS (pH 7.2) for incubation with primary antibody at 4°C (anti-Sho 04SH-1 (1:100, this paper), anti-Sho 06SH-3 (1:200, this paper), anti-Sho 06rSH-1 (1:50, this paper), anti-PrP 3F4 (1:20 000; Signet), anti-PrP 7A12 (1:10 000; ), and anti-calbindin D-28K (1:2000; Chemicon). Other antibodies used include anti-Thy-1, R194 (a generous gift from Roger Morris); anti-NSE, rabbit polyclonal from Polysciences Inc.; anti-synaptophysin, SY38 from Chemicon; anti-actin, rabbit polyclonal from Sigma (20–33); and anti-neurofilament H, SMI-32 from Sternberger Monoclonals Inc. For peptide blocking of antibody reactivity, peptide was added in four-fold mass excess to Sho antibody and incubated overnight at 4°C before use. Sections were rinsed in TBS and processed in DakoCytomation EnVision Labelled Polymer, visualized with DAB (3,3′-diaminobenzidine) and counterstained with Harris' hematoxylin. Images were obtained with a Leica DM6000B microscope using a Micropublisher 3.3RTV camera (Q Imaging Inc.) and OpenLab4.0.4 software (Improvision). Images in and were de-convoluted using the iterative restoration function in Volocity. For fluorescent double labeling experiments, the secondary antibodies used were AlexaFluor-488 (green) and AlexaFluor-594 (red)-conjugated antibodies (Invitrogen). For immunocytochemistry on CGNs (), cells were fixed 24 h post-transfection in ice-cold methanol for 5 min at −20°C, and then processed and stained.
pcDNA3.Sho plasmid was linearized with dIII or II to generate antisense and sense probes, respectively, using the DIG RNA labeling kit (Roche). Sections (6 μm) of formalin-fixed brains were cut using an RNase-free blade, mounted and dried overnight at 63°C. Tissue processing and hybridization was performed as described by . DIG-labeled RNA probes were diluted 1/200–1/400 in hybridization buffer. Alkaline phosphate substrate NBT/BCIP was added, color development performed for 16 h in the dark and slides mounted without counterstaining.
Mouse whole-brain homogenates were prepared in 0.32 M sucrose containing complete protease inhibitor cocktail (Invitrogen). For preparation of crude membranes, homogenates were spun at 700 for 10 min at 4°C, pellets washed with 1 volume of homogenization buffer, spun again as above, supernatants were pooled then spun at 100 000 for 1 h at 4°C and pellets resuspended in cold 50 mM Tris–HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100. After spinning (100 000 , 1 h, 4°C) the supernatant was stored at −80°C. Protein was separated on either NuPAGE 4–12% gels (Invitrogen) using the MES buffer system, or by conventional SDS–PAGE using 14% polyacrylamide gels and transferred to either nitrocellulose (PrP blots, 3% BSA block) or PVDF (Sho blots, 5% non-fat skim milk block). Blots were incubated overnight with primary antibodies (D13 or D18 at 0.5 μg/ml for PrP (InPro Inc.), or 04SH-1 or 06rSH-1 Sho polyclonals at 1:500 or 1:2000, respectively, incubated with HRP-conjugated secondary antibody and developed using ‘Western Lightning' ECL (Perkin-Elmer). The embryo blot was supplied from Zyagen (San Diego, CA), probed with 04SH-1 and then stripped using 0.2 M glycine pH 2.2, 1% Tween-20, 0.1% SDS, followed by reprobing with D13 antibody |
Acetylcholine binding proteins (AChBPs) have been identified from different snails, including () (; ) and (). AChBPs are homologous to the ligand-binding domains (LBDs) of the nicotinic acetylcholine receptors (nAChRs) and pharmacological characterization has demonstrated that their properties most closely resemble those of the α nAChRs (), which also function as homopentamers. Crystallization and structure determination of AChBPs (; , ) from these different species has revealed a highly conserved architecture, despite the relatively low sequence identity between different AChBPs. A similar level of sequence homology is found with the LBDs of members of the ligand-gated ion channel family, comprising the nAChRs, GABA-A/C receptors, 5-HT3 and glycine receptors. AChBP has been co-crystallized with prototype ligands that are known to bind to nAChRs, thus establishing the structural determinants for ligand recognition of agonists such as nicotine and carbamylcholine (), and antagonists such as different α-conotoxins (; ) and long-chain snake neurotoxins (), and partial agonists such as lobeline (). Comparison of these different crystal structures has revealed conformational changes occurring upon ligand binding, and has allowed predictions as to how these conformational changes may be coupled to channel opening through the loops that form the interface with the transmembrane domain in the nAChR.
The venoms from cone snails are a rich source of peptides with high affinity for several voltage- and ligand-gated ion channels, including nAChRs (). Recently, we and others have solved crystal structures of Ac-AChBP in complex with two different α-conotoxins, namely PnIA(A10L D14K) () and ImI (; ). These two conotoxins greatly differ in their selectivity among AChBPs and comparison of the toxin–receptor interface in both complexes provided structural insight into the molecular determinants of ligand selectivity ().
In addition to being remarkable probes for structural studies, α-conotoxins also have therapeutic potential (). For example, Vc1.1, the first α-conotoxin being developed to treat neuropathic pain also caused an accelerated recovery of injured neurons (). Recently, Vc1.1 was shown to specifically target α/α nAChRs, providing a rationale for its analgesic property (). Determining the specific roles of the multiple nAChR subtypes under physiological or pathological conditions and the development of drugs to treat nAChR-related disorders, such as Parkinson's disease and nicotine addiction, requires further subtype-selective ligands.
In this study, we used Ls-AChBP as a bait to discover the novel α-conotoxin TxIA from . Pharmacological characterization shows that TxIA binds with very high affinity to AChBPs from different species, as well as selectively to certain subtypes of neuronal nAChRs. The co-crystal structure of Ac-AChBP with a more potent analog TxIA(A10L), revealed that this α-conotoxin adopts a different binding orientation to that observed for other α-conotoxin–AChBP complexes. A salt bridge between Arg5 in TxIA(A10L) and Asp195 of Ac-AChBP was visible in the structure and the importance of this interaction for AChBP binding and nAChR selectivity was established using binding assays, surface plasmon resonance (SPR) and electrophysiological experiments with mutant receptors and conotoxin analogs. These results highlight the potential of an AChBP screen to discover novel ligands acting at the nAChR and provide a new pharmacophore for the design of ligands with improved subtype selectivity.
We tested the activity of crude venoms obtained from more than 30 species of Australian cone snails against Ls-AChBP in a competitive binding assay with radiolabeled α-bungarotoxin (I-Bgt). We chose the venom of cone snails as our ‘combinatorial library of ligands', as all species tested so far were shown to contain at least one nAChR ligand among the 50–200 unique conopeptides known to occur in each venom (). Accordingly, the venoms from all species showed some competition in our Ls-AChBP-binding assay (). We focused on venom, as full competition with I-Bgt was observed, indicating the presence of a high affinity or abundant ligand (). Upon isolation the active compound was found to be in low abundance, indicating that it was relatively potent (). Mass spectrometry revealed a monoisotopic mass of 1656.68 Da, similar in size to previously isolated α-conotoxins. N-terminal sequencing revealed a novel 16 amino-acid peptide belonging to the 4/7 α-conotoxin family, which we named α-conotoxin TxIA (). The calculated mass (1661.67 Da) was consistent with two disulfide bonds (–4 Da) and an amidated C terminus (–1 Da), two post-translational modifications common in this class of conotoxins ().
Synthetic analogs of TxIA were assembled using Boc chemistry for further analysis and to determine the cysteine connectivity. The ‘native' conformation (connectivity 1–3, 2–4) and the ‘ribbon' fold (connectivity 1–4, 2–3) were both able to displace I-Bgt bound to Ls-AChBP. However, whereas the native fold displayed a of 1.7 nM, the ribbon fold was 380-fold less potent (). Therefore, the connectivity is assumed to be 1–3, 2–4 in the venom-isolated α-TxIA, as was found in all other α-conotoxins identified to date. Comparison with other synthetic α-conotoxins known to bind nAChRs revealed that α-TxIA is 50-fold more potent than the α nAChR-selective PnIA(A10L) in displacing I-Bgt at Ls-AChBP (). The muscle nAChR-selective EI displayed an intermediate value of 496 nM, followed in decreasing order of potency by the αβ nAChR-selective PnIA (1.0 μM) and the α/αβ nAChR-selective MII (1.1 μM), whereas EpI had the lowest affinity (7 μM). Previously, pH was shown to influence the potency of the histidine-containing α-MII in functional assays (). This effect also applies to MII binding to Ls-AChBP (Spearman's test, <0.01), whereas the other toxins were most active at physiological pH (). In a saturation-binding experiment, addition of an ∼IC concentration of native α-TxIA increased the of I-Bgt from 3.5 to 40 nM without affecting the maximum binding , indicative of a competitive binding interaction (). α-TxIA was also tested on rat brain membranes using I-Bgt and H-epibatidine as tracers to determine the affinity for α and mostly αβ neuronal nAChRs, respectively. α-TxIA displaced I-Bgt with a of 1.2 μM (α nAChR), but failed to displace H-epibatidine using up to 10 μM of peptide. Finally, α-TxIA was tested in a functional assay on heterologously expressed mammalian nAChRs (). α-TxIA potently inhibited nicotine-induced current at the αβ nAChR (IC=3.5 nM) and α nAChRs (IC=392 nM), but had no activity at the αβ nAChR and muscle nAChR at concentrations up to 10 μM. Thus, α-TxIA is among the most αβ-selective toxins identified.
The sequence of α-TxIA was compared to previously identified 4/7 α-conotoxins (). Interestingly, only three residues are different from α-conotoxin PnIA, yet PnIA is 600-fold less potent than TxIA at Ls-AChBP. To identify which of these residues conferred the high affinity at Ls-AChBP, we synthesized PnIA and TxIA mutants covering two of the three differences ( and ). A third difference at position 15 was not investigated, as it was located outside the binding site identified in the co-crystal structure of Ac-AChBP with PnIA(A10L D14K) (). For the mutants tested, PnIA(A10L) had 12.5-fold higher affinity at Ls-AChBP, 20-fold higher affinity at the α nAChR, but 10-fold reduced affinity at the αβ nAChR (; ; ) (). In contrast, TxIA(A10L) had similar potency to native TxIA at Ls-AChBP ( and ), suggesting that Ile9 (Ala in PnIA) is able to substitute for Leu10 in PnIA(A10L) in the conserved hydrophobic patch that we have shown previously interacts with the complementary binding site of the nAChR (). Despite the lack of effect on Ls-AChBP affinity, TxIA(A10L) was 12- and 2-fold more potent at the α and αβ nAChRs, respectively (). Thus, a long-chain hydrophobic residue (Leu or Ile) at position 9 or 10 is important for high-affinity binding of TxIA and PnIA to Ls-AChBP and the α nAChRs, but not αβ nAChRs. A second difference is a Leu in position 5 of PnIA compared with an Arg in TxIA. As this position is well placed to directly interact with the receptor (), this change in the physical property and length of the side chain could be an important contributor to high-affinity binding to Ls-AChBP. In support of this hypothesis, EI also has high affinity for Ls-AChBP and a positive charge in the equivalent position, whereas the low affinity [Y15]-EpI has a negatively charged aspartic acid at this position. In agreement with these observations, PnIA(L5R A10L) had 220-fold increased affinity for Ls-AChBP compared to PnIA, clearly demonstrating the important role of Arg5 for high-affinity binding to Ls-AChBP ( and ). The role of this residue appears also important for αβ nAChR binding (10-fold increased affinity of PnIA(L5R A10L) compared to PnIA(A10L) at αβ nAChR), but not for the α nAChR (same affinity compared to PnIA(A10L)), suggesting that Arg5 interacts with Ls-AChBP and αβ nAChRs in a similar manner ( and ).
To gain further insight into the nature of the interactions of α-TxIA with AChBP and their contribution to high affinity binding, we solved the crystal structure of Ac-AChBP in complex with the most potent TxIA analog, TxIA(A10L). Co-crystals were initially obtained with Ls-AChBP, Bt-AChBP and Ac-AChBP, but the latter gave diffraction data of better quality (see for statistics). TxIA(A10L) has comparable affinities to displace H-epibatidine from and AChBPs (data not shown). The structure of the complex was determined at 2.4 Å resolution () and solved by molecular replacement. The asymmetric unit contains two pentamers and all binding sites were occupied by TxIA(A10L). The structure of Ac-AChBP in complex with TxIA(A10L) is very similar to other α-conotoxin complexes, with r.m.s.d. of 0.73 Å (1023 C atoms) upon superposition with the complex of Ac-AChBP with PnIA(A10L D14K) () and 0.67 Å (1024 C atoms) for the Ac-AChBP complex with ImI (). The r.m.s.d. between monomers in the Ac-AChBP–TxIA(A10L) complex is 0.34±0.04 Å. TxIA(A10L) binds with loop C displaced outward by a distance of 10.87±0.50 Å as measured between the Cys188 C atom in the Ac-AChBP complex with TxIA(A10L) and the HEPES-bound Ac-AChBP structure (), similar to values measured for other α-conotoxin complexes. We previously observed ∼3° rigid body rotations of the monomers in the Ac-AChBP complex with PnIA(A10L D14K) (), whereas these rotations were not seen in the Ac-AChBP complex with ImI (). The Ac-AChBP complex with TxIA(A10L) displays intermediate rotations (1–2°) of the monomers with respect to each other, which are most likely due to crystal contacts. The structure of TxIA(A10L) itself is very similar to other α-conotoxins and has an r.m.s.d. of 0.64±0.10 Å upon superposition with PnIA (PDB accession code 1PEN). TxIA(A10L) covers a surface area in the binding pocket of 788±18 Å, which is intermediate between the receptor–toxin interfaces formed with PnIA(A10L D14K) (827±32 Å) and ImI (679±15 Å) ().
We previously observed that PnIA(A10L D14K) and ImI share a similar orientation in the binding pocket, but differ dramatically in the nature of interactions formed within the binding site (; ). Surprisingly, we see that TxIA(A10L) adopts an orientation that is different from those seen in other α-conotoxin–AChBP complexes. TxIA(A10L) is tilted 20° downward by a pivotal reorientation of the conotoxin around Pro7 (), which results in displacement by a distance of 4.72±0.83 Å, as measured between the C atoms at position 14 in the TxIA(A10L) and PnIA(A10L D14K) complexes. This different orientation of the conotoxin compared to those seen in other complexes () is sustained by Arg5, which projects deep onto the principal face of the binding site and forms a hydrogen bond with Tyr186 and a salt bridge with Asp195 (), an interaction not seen in any of the other α-conotoxin complexes (). The structural data thus confirm a key role of Arg5 in the high-affinity binding of TxIA(A10L) to Ac-AChBP. The interface of TxIA(A10L) with the principal binding site is further characterized by the formation of four additional hydrogen bonds between TxIA and Pro7-Trp145 (loop B), TxIA and Asn12-Glu191 (loop C), and TxIA Pro7 and Asn11-Tyr193 (loop C) (). TxIA Pro7 also seems to play a dominant role in forming extensive van der Waals interactions with residues of the principal face that are not involved in contacts with PnIA(A10L D14K), namely Tyr91 (loop A), and Ser144, Trp145, Val146 and Tyr147 (loop B). In contrast, interaction of TxIA(A10L) with residues of the complementary face are similar to those seen in the Ac-AChBP complex with PnIA(A10L D14K) and are mostly hydrophobic in nature.
Comparison of the different AChBP–conotoxin complexes reinforces the notion that α-conotoxins can use different surface contacts to interact with the principal binding site (), whereas the surface area that contacts the complementary face of the binding site remains relatively conserved (). The calculated surface area of conotoxin ImI (1221±13 Å), for example, is smaller than PnIA(A10L D14K) (1508±17 Å) and TxIA(A10L) (1488±29 Å), but they all share a hydrophobic patch on one face of their surface that projects on the complementary binding site. On the other hand, ImI has two arginine residues (Arg7 and Arg11) that protrude into the principal binding site, a surface property not present in PnIA(A10L D14K). TxIA(A10L) seems to have surface properties that are intermediate between ImI and PnIA(A10L D14K), even though TxIA and PnIA are the same length and TxIA has the relatively exposed Arg5.
Structure–activity relationships between TxIA analogs indicated an important role for Arg5 in the high-affinity binding of TxIA to AChBPs. Indeed, incorporation of an Arg residue in PnIA(A10L) at the equivalent position enhanced the affinity for Ls-AChBP that approached TxIA(A10L) affinity, with the tight electrostatic interaction between Arg5 of TxIA(A10L) and Asp195 of Ac-AChBP seen in our co-crystal structure explaining this effect. To further confirm this interaction, we mutated Asp195 in Ac-AChBP to Ala, Asn and Lys (unfortunately, insufficient Ac-AChBP D195K was expressed to allow a complete pharmacological characterization). In a competition assay, D195A and D195N showed binding properties for nicotine and acetylcholine that are comparable to wt Ac-AChBP ( nicotine=0.68±0.10 μM and 0.81±0.13 μM for D195A and D195N, respectively). However, TxIA and TxIA(A10L) showed a 30- to 50-fold reduction in affinity for the D195A and D195N mutants (). A similar reduction was also observed for PnIA(L5R A10L). In contrast, the affinity of PnIA(A10L D14K), which lacks an Arg at position 5, for D195A and D195N remained virtually unchanged. In addition, we compared the kinetic behavior of TxIA analogs on Ac-AChBP and the D195A mutant using SPR. In agreement with results from the binding assay, we observe an approximately sixfold acceleration in the dissociation rate for Arg5-containing α-conotoxins TxIA, TxIA(A10L) and PnIA(L5R A10L), whereas the dissociation kinetics for PnIA(A10L D14K) were unaffected by the D195A mutation (). These results provide strong evidence that the observed drop in affinity for Arg5-containing analogs at D195A can be directly attributed to the loss of an energetically favorable interaction with Asp195, which stabilizes Arg5-containing α-conotoxins in their bound position.
To extrapolate the functional importance of this electrostatic interaction to mammalian nAChRs, we introduced the equivalent D197A and D195A mutations in αβ and α nAChRs, respectively, and compared the potency of TxIA analogs by two-electrode voltage clamp analysis on the oocyte expressed mutant receptors (). In agreement with the results obtained with the Ac-AChBP mutants, we see that Arg5-containing conotoxins had 200- to 500-fold decreased activity at D197A-αβ. Together, these results show that the electrostatic interaction between Arg5 of TxIA(A10L) and Asp195 in Ac-AChBP and Asp197 in αβ nAChRs provides an important energetic contribution to their enhanced affinity at these receptors. In contrast, the activities of TxIA, TxIA(A10L) and PnIA(L5R A10L) at D195A-α were little affected, suggesting that other interactions, probably involving a the hydrophobic patch around position 9/10, dominate α-conotoxin interactions with this subtype.
It was an intriguing finding that D195A and D197A mutations had a dramatic effect on TxIA binding at AChBP and αβ nAChR, respectively, whereas the corresponding mutation (D195A) in α nAChR barely affected the binding of Arg5-containing conotoxins including TxIA. To help understand this difference, we constructed homology models of both nAChRs and docked TxIA(A10L) into its binding site. Analysis of conotoxin binding in the αβ receptor clearly shows that TxIA(A10L) adopts a backbone orientation that is similar to the binding mode observed in our Ac-AChBP co-crystal structure (). This interaction specifically allows an electrostatic interaction between Arg5 and the conserved Asp residue D197, in agreement with our experimental results. To further confirm the existence of this novel binding mode in native αβ nAChRs, we utilized a β-subunit mutant possessing enhanced hydrophobic contacts between the toxin and receptor (). As expected, PnIA(L5R A10L) binds to α-[V109A]β with an affinity over 10 times higher (0.34 nM, CI 0.30–0.38) than observed at the wild-type receptor (4.6 nM, CI 3.69–5.83) (see ). Interestingly, when this β-subunit mutant is coexpressed with the low affinity [D197A]α (2300 nM, CI 1780–2966), we observed a dramatic (>100-fold) rescue of PnIA(L5R A10L) affinity (20.5 nM, CI 17.4–24.1). This rescue effect can be explained if the C-terminal hydrophobic half of toxin repositions to interact with [V109A]β, as seen in the Ac-AChBP–PnIA(A10L D14K) co-crystal structure () compensating for the loss of the energetically favorable Arg5–D197 interaction. Due to conservation in the binding pocket, docking of TxIA(A10L) in the α homology model gave a similar overall result to that obtained using a αβ model. However, our experimental results show that, for this subtype, Arg5 is unlikely to interact with D195. TxIA(A10L) was therefore placed in the α binding pocket using the PnIA(A10L D14K) binding mode () (). This orientation of the conotoxin does not allow a direct interaction between Arg5 and Asp195, explaining the lack of effect the α-D195A mutation. Instead, hydrophobic interactions with the complementary side (equivalent of the β-subunit) dominate the binding interaction at α.
Finally, mapping the α sequence to the AChBP crystal structure reveals an additional positive charge, Lys184 positioned where it could form an internal salt bridge with Asp195, thus reducing the likelihood of an interaction with Arg5 of TxIA(A10L). In addition, the downward tilt of TxIA(A10L) toward Asp195 might be prevented by a H-bond expected between the TxIA-Leu10 main chain and Gln115 in the α nAChR. Together, these observations support the possibility that TxIA(A10L) binds to α in a conformation observed for ImI and PnIA mutants in the AChBP subtypes. In such a conformation, the relevance of an Arg5-Asp195 salt bridge would be minimized, as has been observed for ImI binding to AChBP. In contrast, the binding to the αβ nAChR subtype is likely to be similar to that observed in our AChBP/conotoxin complex, including an important contribution to binding from the Arg5-Asp195 salt bridge.
In this study, we demonstrate that an AChBP screen can be used to discover and guide the isolation of new α-conotoxins in crude venom. This rapid and sensitive assay has advantages over fluorescent and electrophysiological methods used previously, and is amenable to high-throughput applications. As all α-conotoxins tested bound to Ls-AChBP in the micromolar to mid nanomolar range, despite having distinct nAChR preferences for muscle and homomeric and heteromeric neuronal nAChR subtypes, it appears that Ls-AChBP has retained ancestral nAChR features that allow a broad range of nAChR ligands to bind. In support, an Ls-AChBP screen of over 30 different cone snail venoms revealed that all significantly displaced I-Bgt binding. This high hit-rate supports the hypothesis that each cone snail venom contains at least one nAChR antagonist (). Given that many cone snails are molluscivorous, and the apparent broad distribution of AChBP in molluscs, AChBP could represent a previously unrecognized molecular target for α-conotoxins with the potential to disrupt molluscan neurotransmission.
The sensitivity of an Ls-AChBP screen is demonstrated with the isolation of α-TxIA, a trace component of venom. The complete pharmacological characterization at Ls-AChBP and nAChRs was subsequently achieved using an identical synthetic form. α-TxIA was found to be the most potent α-conotoxin acting at Ls-AChBP reported to date and also had high affinity for certain mammalian nAChRs. Low nanomolar concentrations of α-TxIA were sufficient to inhibit αβ nAChR, whereas 100-fold higher concentrations were needed to block α nAChR current. In contrast, α-OmIA, which also displays high-affinity binding to AChBPs, is equipotent at αβ and α nAChRs (). Sequence comparison of TxIA with other α-conotoxins and synthesis of selected TxIA analogs revealed important contributions from hydrophobic residues at position 9 and 10 and the key role of Arg5 for high-affinity interactions at Ls-AChBP. Indeed, substitution of Arg5 in PnIA(A10L) produced a gain-of-function analog with similar properties to TxIA. The co-crystal structure of Ac-AChBP in complex with the higher potency analog TxIA(A10L), confirmed the crucial role of Arg5, which is coordinated by Asp195 and Tyr186, an interaction not seen in two other Ac-AChBP–conotoxin complexes (; ; ). The presence of this salt bridge contributed to a 20° downward tilt of the toxin backbone when compared to PnIA(A10L D14K). An electrostatic interaction was also observed between Arg7 of conotoxin ImI and Asp195 in only one of the two available Ac-AChBP co-crystal structures (PDB accession codes 2BYP and 2C9T). However, mutant cycle analysis has shown that the interaction between Arg7 and Asp197 in nAChRs does not contribute to high-affinity binding of conotoxin ImI (), most likely due to the different orientation of the toxin in the binding site. By mutating Asp195 in the background of Ac-AChBP, we confirmed the functional importance of the electrostatic interaction between Arg5 and Asp195 as observed in the co-crystal structure. Importantly, we observed that the affinity of PnIA(A10L D14K) remains unaffected by Asp195 mutations. This result provides strong evidence that the enhanced affinity of Arg5-containing conotoxins can be attributed to an energetically favorable electrostatic interaction with Asp195 in Ac-AChBP. In combination with electrophysiological recordings on mutant nAChRs, we demonstrated that this conclusion extrapolates to αβ nAChRs, but not to α nAChRs. As Asp195 is highly conserved among AChBPs and nAChRs, Arg5 TxIA may share a common set of binding interactions in AChBPs and certain nAChR subtypes.
The reorientation of TxIA(A10L) in the binding pocket of Ac-AChBP is a surprising observation in light of the two other α-conotoxin complexes that we and others have previously determined (; ; ). Such differences highlight that docking simulations based on the backbone orientation of PnIA(A10L D14K) () or contacts analysis based on superposition of the α-conotoxin backbone () should be interpreted with caution, as these approaches may not adequately address residue changes or the reorientation of side chains in the toxin or receptor, which could generate backbone reorientations as observed for TxIA(A10L). However, from our analysis of homology models, it appears likely that TxIA(A10L) adopts the same orientation in αβ nAChRs as seen in our crystal structure, whereas it may adopt a PnIA-like binding orientation in the α nAChR binding pocket. Given this result, it is somewhat surprising that TxIA has no affinity for αβ nAChRs despite the presence of an equivalent Asp residue and the close similarity to αβ nAChRs across the principal binding face. As no obvious clash could be identified during docking to an αβ homology model, it appears that AChBP is not a strong predictor of αβ structure, perhaps reflecting different docking pathways or an altered binding site structure.
In conclusion, we have demonstrated that Ls-AChBP can be used to rapidly identify new ligands for nAChRs. By screening the crude venom of different cone snails, we discovered a new α-conotoxin TxIA with enhanced subtype selectivity for αβ nAChRs. Co-crystallization of the most potent analog TxIA(A10L) with Ac-AChBP identified a new α-conotoxin binding mode that was stabilized by a critical salt bridge between Arg5-TxIA(A10L) and a highly conserved Asp residue on the principal face of the binding face of Ac-AChBP. The functional importance of this interaction was confirmed through mutagenesis studies in different nAChR subtypes and AChBP. These results establish a structural framework for developing ligands with enhanced selectivity for the αβ nAChR. Engineering AChBPs with the ligand-binding sites of specific nAChR subtypes, or with the ligand-binding sites of other ligand-gated ion channels, is expected to identify specific pairwise interactions underlying nAChR selectivity and further expand the potential of this protein scaffold to discover novel pharmacological probes.
Venom from specimens was extracted and fractionated as described previously for other cone species (). Biological activity was tested using the Ls-AChBP radioligand-binding assay as described previously (). This assay directed the final purification of the active compound on analytical RP-HPLC (C Phenomenex column). A 20 pmol portion of pure peptide was Edman sequenced (Biomolecular Research Facility, Newcastle, Australia) on an Procise HT (Applied Biosystem). Molecular mass analysis of the native and synthetic peptides and LC-MS analysis of crude venom were performed on a C Phenomenex column (2.1 × 150 mm, 5 μm) eluted with 0 to 60% B in 60 min (A=0.05% TFA; B=0.045% TFA, 90% ACN). Eluant was monitored with a PE-Sciex API III triple quadrupole mass spectrometer (Thornhill, Ontario, Canada) over 400–2000 and data analyzed using Analyst software (Agilent, CA, USA).
Competitive binding assays with His-tagged Ls-AChBP and I-radiolabeled α-bungarotoxin (specific radioactivity 5.5 TBq/mmol) were carried out as described previously (). A binding assay was established using P rat brain homogenate as described previously (), with a final protein concentration of 6 mg/ml as determined by a BCA protein assay (Pierce, Rockford, IL, USA). Various concentrations of conotoxins diluted in 200 μl incubation buffer (50 mM HEPES at pH 7.4, 100 mM NaCl, 0.2% BSA) were incubated for 90 min in 100 μl P membrane with a final concentration of 3 nM I-α-bungarotoxin to measure binding to α nAChRs, or 1 nM H-epibatidine (Amersham Biosciences, Castle Hill, Australia) plus 2 μM cold α-bungarotoxin for non-α nAChRs.
Binding assays with untagged Ac-AChBP and D195 mutants were performed using 5 nM H-epibatidine. Ligands were incubated in a white Optiplate-96 (Perkin Elmer) with radioligand and protein (30–120 ng) in binding buffer (PBS, 20 mM Tris at pH 8, 0.05% Tween 20) in a final volume of 100 μl. FlashBlue GPCR beads (2 mg/ml) (Perkin Elmer) were added and after 90 min incubation and 16 h standing in the dark, the radioactivity was measured with a Wallac 1450 MicroBeta liquid scintillation counter. Binding-data were evaluated by a nonlinear, least squares one-site competition curve fitting procedure using Prism 4.01 (GraphPad Software Inc., San Diego, CA, USA).
Untagged Ac-AChBP was expressed from baculovirus in SF9 insect cells and purified from medium as described previously (). Ac-AChBP D195 mutants were constructed using a QuikChange approach (Stratagene, La Jolla, CA, USA) and verified using DNA sequencing. All peptides were synthesized using Boc chemistry with neutralization protocols as described previously (). The oxidized peptides were purified by RP-HPLC and analyzed by electrospray mass spectrometry.
cDNAs encoding neuronal nAChRs were provided by J. Patrick (Baylor College of Medicine, Houston, TX, USA) and subcloned into the oocyte expression vector pNKS2 (). Site-directed mutagenesis was performed with the QuikChange mutagenesis Kit (Stratagene, La Jolla, CA, USA) and primers were synthesized by MWG Biotech AG (Ebersberg, Germany). Sequences were verified by didesoxynucleotide sequencing (MWG Biotec AG). cRNA was synthesized from linearized plasmids with SP6 RNA polymerase using the mMessage mMachine kit (Ambion, Austin, TX, USA). frogs were purchased from Nasco International (Fort Atkinson, WI, USA). Oocytes were prepared as described previously () and injected with 50 nl aliquots of cRNA (0.5 μg/μl). Two-electrode voltage-clamp recordings were performed as described previously (). Current responses to acetylcholine or nicotine were measured 1–10 days after cRNA injection at a holding potential of −70 mV using a Turbo Tec 05X Amplifier (NPI Electronic, Tamm, Germany) and Cell Works software. Currents were filtered at 200 Hz and digitized at 400 Hz. The perfusion medium was automatically switched using a custom-made magnetic valve system. A fast and reproducible solution exchange (<300 ms) was achieved using a 50 μl funnel-shaped oocyte chamber combined with a fast solution flow (∼150 μl/s) fed through a custom-made manifold mounted immediately above the oocyte. Agonist pulses were applied for 2 s at 4 min intervals. Peptides were applied for three minutes in a static bath when responses to three consecutive agonist applications differed by less than 10%. EC values were calculated from a nonlinear fit of the Hill equation to the data (GraphPad version 3.0, San Diego, CA, USA). Data are presented as mean±s.e. from at least four experiments.
SPR spectroscopy was performed at 25°C on a Biacore T100 (Uppsala, Sweden). Proteins were immobilized on a CM5 chip using the amine coupling procedure where the proteins at 0.1 mg/ml concentration were flown over the chip at 5 μl/min in 10 mM sodium acetate at pH 4.0. Approximately 5000 response units (RUs) of Ac-AChBP D195A and wild-type Ac-AChBP were immobilized on the chip and the empty flow cell was used as the control. Conotoxins (5–600 nM) in running buffer (25 mM sodium phosphate at pH 8.0, 100 mM sodium chloride) were injected across the chip at 30 μl/min. Biacore T100 evaluation software was used for analysis of the experimental data. GraphPad Prism (GraphPad version 4.0, San Diego, CA, USA) was used to generate final figures.
Crystals of Ac-AChBP in complex with TxIA(A10L) were grown in nanoliter drops by mixing 200 nl protein solution with 200 nl reservoir solution composed of 200 mM sodium malonate, 20% polyethyleneglycol 3350 and bistrispropane at pH 8.5. The crystals belong to spacegroup 1 and have the following unit cell dimensions: a=72.54 Å, b=85.75 Å, c=121.67 Å, α=90.14°, β=80.01°, γ=70.64°. Glycerol was used as a cryoprotectant and incrementally added to the mother liquor to a final concentration of 30% before flash-freezing the crystals by immersion in liquid nitrogen. Diffraction data were collected at beamline X06SA at the Swiss Light Source, Villigen. The resolution of observed reflections rapidly decayed from 1.8 to 2.8 Å during a 180° sweep due to radiation sensitivity of the crystals. In addition, the diffraction data showed signs of nonmerohedral twinning. The main crystal lattice was indexed with MOSFLM and data reduction and scaling was carried out with SCALA using the CCP4 program suite (). The protein structure was solved by molecular replacement with PHASER () and the open C-loop structure of Ac-AChBP as the search model (PDB accession code 2C9T). The initial model was refined to =27% and =31% with REFMAC () using NCS and TLS restraints (). Difference electron density maps clearly indicated the occupancy of all binding sites by TxIA(A10L). The α-conotoxins were built into density with COOT () using the structure of PnIA(A10L D14K) as a template (PDB accession code 2BR8). The model of Ac-AChBP with TxIA(A10L) bound was then further refined to =24% and =30%. The issue of nonmerohedral twinning was addressed by reprocessing the diffraction data with EVAL14 () and deconvoluting overlapping reflections from the interfering lattice. The resulting electron density map was characterized by a lower noise level, but higher -values for the refined protein model, which is expected due to lack of profile fitting in EVAL14 (). Further refinement of the structure was therefore carried out using MOSFLM-processed data. Difference electron density peaks near most of the disulfide bridges in Ac-AChBP and TxIA(A10L) indicated severe radiation damage. Therefore, diffraction data over a 360° sweep were collected on multiple segments of the crystal at beamline EH23-2 of the European Synchrotron Radiation Facility, Grenoble. The beam was attenuated to minimize radiation sensitivity and a merged data set with improved statistics was obtained to 2.4 Å resolution including data to I/σI=1. The model was automatically rebuilt using pyWARP () and has a =23% and =25% with good geometry after iterative cycles of manual rebuilding and refinement. Structure validation was carried out using WHATIF () and MOLPROBITY (). Full molecular replacement and refinement of the data to 2.7 Å, using a more conservative data cut-off (I/σI=2), showed that the electron density did not change, but automatic rebuilding failed, indicating that the weak data contained significant information and that any errors in their measurement were sufficiently reduced by the use of maximum likelyhood refinement. Coordinates have been deposited in the Protein Data Bank with accession code 2UZ6. AREAIMOL and CONTACT were used to analyze interaction surface areas and contacts (). Interaction surface areas are reported as average±s.d. of binding sites occupied by conotoxin and in the context of the number of pentamers present in the asymmetric unit. Interactions between residues of conotoxins and Ac-AChBP were only considered if present in at least three of five binding sites. Figures were prepared with PYMOL (DeLano Scientific, San Carlos, CA, USA).
The crystal structure of TxIA(A10L) bound to Ac-AChBP was used as a template to build homology models of rat α, αβ and αβ nAChRs following the method described by . |
Cohesin is member of the family of Smc (structural maintenance of chromosomes) containing protein complexes. Smc complexes are conserved from prokaryotes and archea to eukaryotes, and play important roles in chromosome structure and segregation in all organisms studied (; ). Eukaryotes contain at least three distinct Smc complexes that have partly overlapping functions, but each of which is essential for cell viability. The cohesin complex associates with chromosomes during G1 phase of the cell cycle, and ensures that the sister chromatids produced during DNA replication in S-phase remain paired with each other after their synthesis. This pairing allows recognition of the replication products in mitosis by the spindle apparatus and their bipolar alignment on the mitotic spindle. Cleavage of the cohesin subunit Scc1 by the protease separase liberates sister chromatids to trigger chromosome segregation at anaphase onset. The other Smc complexes are condensin, required for chromosome compaction during mitosis, and an Smc5/Smc6 containing complex with a role in DNA repair. Smc complexes are involved in several additional aspects of chromosome biology, including transcriptional regulation, chromatin boundary formation, and the DNA replication checkpoint. The mechanistic basis by which Smc complexes act on chromosomes is still poorly understood.
Cohesin forms a large proteinaceous ring whose circumference is largely composed of the coiled coils of the Smc1 and Smc3 subunits. Electron micrographs of vertebrate cohesin illustrate the ring shape, and interaction studies with recombinant subunits expressed in baculovirus-infected insect cells have demonstrated the subunit arrangement to form this ring (; ). It is thought that the cohesin ring binds to chromosomes by topologically embracing DNA (). Mutant analysis has furthermore suggested that ATP bound to the Smc head domains must be hydrolysed for cohesin to load onto chromosomes (; ). Cohesin associates with chromosomes initially at sites bound by the Scc2/Scc4 cohesin loader, an essential cofactor in the loading reaction. From these sites, cohesin appears to translocate away towards sites of convergent transcriptional termination (; ).
How DNA enters the cohesin ring, and what effect binding and hydrolysis of ATP has on the conformation of the complex is poorly understood. Recent results suggest that in addition to the ATPase heads the Smc hinge, where Smc1 and Smc3 interact at the opposite side of the ring, plays an important role in DNA binding of Smc complexes (; ; ). In particular, it has been suggested that the interface between Smc1 and Smc3 at the hinge may need to open up to let DNA enter the ring. How energy derived from ATP hydrolysis at the heads is transferred to open up the hinge is unclear. Atomic force microscopic images of both the fission yeast condensin complex, as well as the cohesin Smc1/3 dimer, have shown that the Smc hinge bends back towards the heads (; ), but biochemical confirmation of a possible head–hinge interaction is missing. Once chromosome-bound, cohesin's behaviour during DNA replication, when it links the nascent sister chromatids, is also poorly understood. An ATPase motif that is important during binding of cohesin to chromosomes in G1 is no longer required (). How therefore the two replication products are trapped by cohesin, is not known. The replication fork might pass through the cohesin ring, leaving replication products trapped inside, without participation of cohesin's ATPase. In addition, binding of cohesin to human chromosomes becomes more stable at the time of S-phase (). It has remained controversial whether after replication individual cohesin complexes embrace both sister chromatids, or whether interactions between more than one cohesin complex establish sister chromatid cohesion.
The geometry of the cohesin complex bound to chromosomes , and possible conformational changes during DNA binding and the establishment of sister chromatid cohesion, are difficult to study using conventional biochemical techniques. As a step towards addressing these questions, we have analysed fluorescence resonance energy transfer (FRET) between cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) fluorophores fused to cohesin subunits in live budding yeast. This allows determination of relative distances between the fluorophores and their possible changes during the cell cycle. Systematic FRET measurements between pairwise fluorophore combinations allowed us to refine the picture of the complex and place Pds5 as a possible matchmaker between opposite sides of the complex. Our results suggest a stable geometry of cohesin throughout the cell cycle, and that any conformational changes in response to ATP hydrolysis are likely to be transient. We also used FRET to search for interactions between more than one cohesin complex. Together, this allows us to present an updated view on the behaviour of the cohesin complex .
To analyse FRET between cohesin subunits , we utilised a recently developed simple and robust method based on the FRET ratio (FRET; ). In this approach, fluorescent intensities of CFP and YFP are measured with an epifluorescent microscope, and FRET is seen after excitation of the CFP fluorophore as increased emission in the YFP channel. Even without FRET, fluorescence is detected after CFP illumination in the YFP channel due to spectral spillover between the channels. Therefore, spillover factors are first determined, and FRET is measured as the ratio of the observed FRET intensity over the expected spillover (see Materials and methods, and , for details).
FRET gives a measure for FRET that is independent of fluorophore concentration, but sensitivity of the measurements is greatest for equimolar fluorophore concentrations. We therefore compared the concentration of the cohesin subunits within budding yeast by measuring fluorescent intensities of CFP fusions expressed at their genomic loci. For these, and all following experiments, we used homozygous diploid yeast strains, which yield increased fluorescent intensities over haploid strains. was deleted and the growth medium supplemented with adenine to reduce background fluorescence from intermediates of the adenine biosynthesis pathway. All cohesin subunits are essential genes in budding yeast, and the fluorophore-tagged subunits in all cases were the only copies of the proteins present in the cells analysed. There were no growth defects, indicating that the tagged proteins were all functional in sustaining all essential aspects of cohesin activity. Fluorescence intensities were measured in an area of fixed size within the nucleus of all strains containing the cohesin subunit-CFP fusions ( and ). This showed that in budded G2 cells, all cohesin subunits were present at approximately the same concentration. Approximately 5000 copies of cohesin are present in haploid G2 cells (; ), and we expect approximately double this number in our diploid strains.
While observing fluorophore-tagged cohesin subunits, we noticed that in cells with small to medium-sized buds cohesin was enriched in a distinct focus within the nucleus. At higher resolution, the focus appeared to acquire the shape of a ring. Dual-colour imaging, including spindle pole body (SPB) and kinetochore markers, showed that the foci likely represent centromeres, clustered around the SPB, where cohesin is enriched () (; ). While centromeres remain attached to the SPB throughout the cell cycle, the timing of foci appearance correlated well with cohesin binding to chromosomes, from the G1/S transition when buds emerge until in mitosis. In early-anaphase nuclei, when cohesin dissociates from chromosomes after Scc1 cleavage, the foci disappeared. Between anaphase and G1, most subunits appeared diffuse throughout the nucleus. As an exception, Scc3-CFP was enriched along the nuclear membrane during this time, but we do not know the reason or possible consequence of this ().
Because of the greater signal intensities, the fluorescence intensity measurements above were made within the nuclear foci. FRET measurements in G2 cells were also routinely made within the foci. Analyses within the diffuse nuclear region that were performed in parallel gave similar results (). As a positive control for FRET we attached a tandem fusion of CFP with YFP, separated by a short glycine–alanine linker, to the C-termini of both Scc3 and Pds5. FRET for these strains was 2.15±0.2 (=32) and 2.12±0.24 (=34) (). This provides an upper limit for FRET expected from closely juxtaposed CFP and YFP fluorophores. If there is no FRET between CFP and YFP, the signal intensity in the FRET channel is expected to be equal to the spillover from both fluorophores, resulting in a FRET value of 1. Values close to this negative baseline were observed, for example, when the C-terminal Scc3-CFP tag was combined with Pds5-YFP, FRET had a value 1.07±0.09 (=52), or vice versa Scc3-YFP with Pds5-CFP, FRET had a value 0.99±0.14 (=31).
Of particular interest within the cohesin complex are the two Smc ATPase head domains. The Smc ATPase is part of the family of ‘ATP binding cassette (ABC)' ATPases, whose function is thought to involve ATP-dependent dimerisation. Structural and biochemical analysis of the Rad50 ABC ATPase has shown how ATP is sandwiched between the two head domains to promote their dimerisation (). Crystal structures of bacterial Smc heads, and of the budding yeast Smc1 head bound to ATP, suggest that they use a similar mode of ATP-dependent dimerisation (; ). It is unknown, however, when during the cell cycle cohesin ATPase heads dimerise or dissociate. Interaction studies have suggested that Smc1 and Smc3 heads can bind each other directly, but also that Scc1 might play an important role in bridging or stabilising their interaction. Electron micrographs of cohesin show the Smc head domains separated from each other, bridged by the non-Smc subunits (; ; ; ).
To examine the interaction between the Smc heads , we measured FRET between Smc1 and Smc3 tagged at their C-termini with YFP and CFP, respectively. FRET in budded exponentially growing cells was 2.06±0.14 (=38; ). This value is close to FRET observed for the covalent CFP-YFP fusion controls, indicating close proximity between the Smc heads. Exchanging the tags (i.e., Smc1-CFP, Smc3-YFP) produced similar results (FRET=2.12±0.19, =47). We next asked whether the close association of the Smc heads was regulated during the cell cycle, and if it depended on Scc1. We monitored FRET between Smc1-CFP and Smc3-YFP after release of small unbudded G1 cells, obtained by centrifugal elutriation, into progression through the cell cycle (). Thirty minutes after release, cells were still in G1 and the Smc fluorescence diffuse in the nucleus, probably because Scc1 was not yet present. FRET was 2.03±0.44, indicating close association of the Smc heads. As cells progressed through S, G2 and M phases, no significant change to FRET was observed.
Despite the high FRET value, indicating close proximity between the Smc heads, it is difficult to estimate their actual physical distance. The distance between the termini must be less than 10 nm, the limit of FRET between CFP and YFP, and because of the high value is probably closer to 3 nm, the minimum distance between the fluorophores. Because FRET was similarly high in G1, when Scc1 is absent, it is likely that it represents direct dimerisation of the ATPase heads. To confirm that the observed FRET represents direct, Scc1-independent, association of the Smc heads, we repeated the measurements in cells in the presence of, or depleted for Scc1. For this, we replaced the Scc1 promoter with the galactose-inducible promoter. Cells were grown in medium containing galactose, then the culture was split and one half was transferred to medium lacking galactose to repress Scc1 expression. Two hours later, Scc1 was largely depleted from the culture lacking galactose (). As expected, in the absence of Scc1, cohesin failed to associate with chromosomes, and nuclear foci were not observed (data not shown). FRET between Smc1-CFP and Smc3-YFP was similar with or without Scc1 (with Scc1: FRET=1.77±0.12, =53; without Scc1: FRET=1.85±0.27, =47). Note that in this experiment a microscope with greater spectral spillover resulted in lower absolute FRET values. These results suggest that the two Smc heads dimerise, probably in an ATP-bound state, in the absence of Scc1, and that they retain close association after Scc1 joins the complex and cohesin is loaded onto chromosomes.
We next analysed whether we could detect Smc head disengagement when Scc1 is cleaved and the cohesin ring dissociates from chromosomes in anaphase. We measured FRET in early anaphase cells displaying dumbbell-shaped nuclei selected from the 120 min time point of the experiment shown in . Foci of cohesin had dispersed, as expected, but we did not find evidence for a greater distance between the Smc heads (FRET=1.92±0.26, =63). FRET in our experiments is a population average of all cohesin molecules present in the observation area. Transient, asynchronous dissociation of the Smc heads during cohesin loading or unloading from chromosomes would not be detectable with our technique. Alternatively, ATP hydrolysis during cohesin loading, and Scc1 cleavage in anaphase, may lead to conformational changes in the cohesin complex that do not alter the distance between the fluorophores attached to the Smc1 and Smc3 C-termini.
We next tried to enrich for ‘open' cohesin complexes in anaphase by ectopic overexpression of a C-terminal Scc1 cleavage fragment (Scc1(met269-566)). This fragment resembles that normally produced by separase cleavage of Scc1. It associates with the Smc1 head and weakens its interaction with Smc3 (; ). Cells were arrested in metaphase by nocodazole treatment, and expression of Scc1(met269-566) induced for 2 h. Nuclear foci disappeared, consistent with cohesin dissociation from chromosomes (). Nevertheless, FRET between the Smc heads did not diminish (FRET=1.99±0.22, =71), similar to control cells not expressing the Scc1 fragment (FRET=1.92±0.18, =58). This suggests that even when the interaction between the Smc heads is weakened by the Scc1 cleavage product, they remain associated. The high local concentration of the two heads, connected at the Smc hinge, might promote their association. Weakening of the Smc head interaction may facilitate transient head dissociation during anaphase, too short-lived to be detected under our conditions. Alternatively, Scc1 cleavage may promote another conformational change within cohesin that leads to its removal from chromosomes.
We next used FRET to study the interaction of Scc1 with the Smc heads. We constructed strains harbouring fluorophores at the Scc1 N- or C-termini, in combination with fluorophores at the Smc heads. In the following, we use YFP-Scc1 to indicate YFP fused to the Scc1 N-terminus, and Scc1-YFP for the fluorophore at the C-terminus. Many current models of the cohesin complex draw Scc1's N- and C-termini in considerable distance from each other, bridging a gap between the Smc1 and Smc3 heads (; ). The Scc1 C-terminus is thought to contact only Smc1, while the N-terminus associates with Smc3. If this arrangement was correct, we would expect strong FRET between fluorophores at the Scc1 C-terminus and Smc1, but weak or no FRET with Smc3. Inversely, we would expect FRET between fluorophores at the Scc1 N-terminus and Smc3, but not Smc1. In contrast to these expectations, we observed equally strong FRET between the Scc1 C-terminus and both Smc1 and Smc3. FRET between Scc1-YFP and either Smc1-CFP or Smc3-CFP was 1.87±0.25 (=52) and 1.86±0.16 (=45), respectively (). We confirmed this observation after exchanging the fluorophore tags, FRET between Scc1-CFP and either Smc1-YFP or Smc3-YFP was 1.88±0.17 (=42) and 1.93±0.15 (=42), respectively. This suggests that the Scc1 C-terminus is placed close and equidistant from both Smc1 and Smc3 heads. These results are inconsistent with models in which Scc1 bridges a gap between the Smc heads. Instead they support our finding that the two Smc heads are closely juxtaposed, and suggest that the Scc1 C-terminus is placed between the two heads. The arrangement of the Scc1 C-terminus in its crystal structure with an Smc1 head is consistent with our FRET results (), if we consider that Smc3 adopts the position of the second Smc1 head in the homodimer structure.
We next analysed the positioning of the Scc1 N-terminus relative to the Smc heads. FRET of CFP-Scc1 with Smc1-YFP or Smc3-YFP was 1.23±0.13 (=49) and 1.58±0.13 (=48), respectively. This suggests that the Scc1 N-terminus is positioned closer to the Smc3 head, but that it retains proximity also with Smc1. The association of the Scc1 N-terminus with Smc3 is thought to be less stable than that of the C-terminus with Smc1 (). We therefore analysed whether we could see any evidence for a change or regulation of this interaction during the cell cycle. Plotting FRET as a function of cell cycle progression showed that the interaction remained constant (). From these results we suggest that the two Smc heads remain in contact for most of the cell cycle, and that Scc1 binds the two heads in an orientation that is largely perpendicular to the axis that connects the two heads (see schematic representation in ).
We next focused our attention on the Scc3 subunit. Biochemical evidence suggests that Scc3 interacts with the Scc1 C-terminal half, and associates with the cohesin complex in an Scc1-dependent manner (). To map Scc3 with respect to the other subunits , we carried out FRET experiments with fluorophores attached to either terminus of the subunit. The closest association was found between Scc3-YFP and Scc1-CFP (FRET=1.4±0.14, =44), consistent with the characterised biochemical interaction (). We also observed FRET between Scc3-YFP and both Smc1-CFP and Smc3-CFP (FRET=1.35±0.16, =55 and 1.37±0.14, =55, respectively), suggesting equidistant positioning between the two Smc subunits. Significant FRET also occurred between fluorophores attached to the Scc3 N-terminus and Scc1 or the Smc heads. This confirms association of Scc3 with Scc1 , and suggests that Scc3 is positioned symmetrically with respect to the Smc heads.
We also carried out FRET measurements between intramolecular fluorophore pairs attached to the N- and C-termini of cohesin subunits. This revealed significant FRET between the Scc1 N- and C-termini (FRET=1.45±0.14, =32), consistent with a relatively close positioning of the two termini with respect to each other within the complex (). We also observed weaker FRET between the tagged N- and C-termini of Scc3 and Pds5, respectively, suggesting that the ends are in relative proximity to each other. We then wanted to probe the geometry of the Smc1 subunit in more detail. The Smc1 head is separated from the Smc hinge by an approximately 50-nm-long stretch of coiled coil. As the Smc hinge has been suggested to show functional interactions with the ATPase heads (), we assessed the distance between the Smc1 head and hinge by FRET . To this end, we inserted a YFP fluorophore into a predicted surface loop at the Smc1 hinge, and a CFP fluorophore was added to the C-terminus of the same protein. FRET between these fluorophores was close to the negative baseline (FRET=1.05±0.19, =55). This suggests that a direct interaction between the Smc1 head and hinge, if it exists, only occurs transiently .
Budding yeast Pds5 is essential for cohesin association with chromosomes, and binds together with cohesin to the same chromosomal sites in late G1 (; ; ). Fission yeast and human Pds5 have been shown to be part of the cohesin complex, but appear to be dispensable for a basal level of sister chromatid cohesion (; ; ). Little is known about the association of Pds5 with cohesin, so we sought to determine with which subunits Pds5 interacts. We failed to detect significant FRET between N- or C-terminal Pds5 fluorophore tags and most cohesin subunits (). Combination of CFP-Pds5 with Scc1-YFP and Pds5-YFP with CFP-Scc1 yielded the highest among the very low FRET values (FRET=1.08±0.16, =47 and FRET=1.1±0.13, =42, respectively). A -test to evaluate the significance of these values suggested that they are greater than those obtained for the Pds5-YFP/Smc1-CFP pair (<0.01). These very weak FRET values should be regarded equivocal. To our surprise we found a clear FRET signal between both N- and C-terminally tagged Pds5 and the Smc1 hinge-YFP insertion (FRET=1.15±0.23, =48 and FRET=1.21±0.19, =40, respectively, which is greater than the Pds5-YFP/Smc1-CFP pair at <0.0001). This suggests that Pds5 is in contact with the Smc1 hinge.
Because of the very weak FRET of fluorophore-tagged Pds5 with Scc1, we searched for independent evidence whether Scc1 might be involved in Pds5's interaction with cohesin. We first asked whether an interaction of Pds5 with cohesin can be detected by co-immunoprecipitation. Scc1 co-precipitated with Pds5 from yeast extract and, as has been observed with human cohesin (), this interaction was salt sensitive (). This confirmed that Pds5 is part of, or interacts with, the budding yeast cohesin complex. We then tested whether the association of Pds5 with cohesin depended on Scc1. Using a strain in which Scc1 could be repressed under control of the promoter, we observed an interaction between Pds5 and the cohesin subunit Smc1 in the presence of Scc1, which was abolished in the absence of Scc1 (). This suggests that the interaction of Pds5 with cohesin depends on Scc1. Pds5 might contact Scc1, and once bound to cohesin, engages in an interaction with the Smc hinge. Alternatively, association of Scc1 with cohesin could introduce a conformational change that allows Pds5 binding to the Smc hinge.
Atomic force microscopy of Smc1/Smc3 heterodimers showed an apparent interaction of the Smc heads with the hinge (), so we wondered whether we could find biochemical evidence for this association. We overexpressed in yeast an Smc1 head construct consisting of the N- and C-terminal head domains connected by a short peptide linker (), as well as the two Smc1 and Smc3 halves of the hinge. Immunoprecipitation against the Smc1 half-hinge demonstrated stable complex formation with the Smc3 part of the hinge (). Moreover, this Smc hinge complex efficiently co-precipitated the Smc1 head, suggestive of a direct head–hinge interaction. The quantities of overexpressed hinge and heads precipitated in this experiment exceeded the level of the endogenous cohesin complex, and we could not detect other cohesin subunits in the immunoprecipitate (data not shown). Therefore, the interaction between the Smc head and hinge observed in this experiment is most likely direct. This evidence for a direct Smc1 head–hinge association is in contrast to our failure to detect physical proximity between the two by FRET (). A possible solution to this apparent paradox is that our biochemical results reveal an interaction that occurs only transiently , either because of a conformational equilibrium biased towards complexes with separated heads and hinge, or because the interaction occurs only as an intermediate, for example, during cohesin loading onto chromosomes.
Several models have been put forward to explain how cohesin might link two replication products after DNA synthesis (; ; ; ). One important question is whether one cohesin ring encircles and holds together both sister chromatids, or whether individual cohesin complexes bind both sister chromatids, and linkages are established by interactions between pairs of cohesin complexes. characterisation of cohesin isolated from yeast chromosomes has so far not found evidence for higher order interactions between more than one cohesin complex (; ; ). Nevertheless, the existence of such interactions is difficult to exclude. We therefore utilised our FRET assay to search for interactions between two cohesin complexes. We first analysed two copies of Smc1 that were tagged in a diploid yeast strain at their C-termini with CFP and YFP, respectively. The existence of cohesin dimers in ‘head to head' orientation should result in FRET between the two tagged Smc1 termini. However, no FRET was detected (FRET=1.05±0.24, =51) (). A similar experiment with CFP- and YFP-tagged copies of Smc3 again detected no interaction (FRET=1.03±0.13, =34). A corollary of this experiment is that FRET observed between fluorophore-tagged Smc1 and Smc3 is due to head interaction within the cohesin complex, and not due to higher-order cohesin interactions or crowding between adjacent cohesin complexes at chromosomal association sites.
The above experiments would not detect interactions between cohesin complexes if they occurred in a ‘hinge to hinge' orientation. To test this possibility, we constructed CFP and YFP insertions within the hinge in the two Smc1 copies of a diploid strain. Again, no FRET was observed (FRET=1.0±0.22, =40). While these results cannot exclude association between more than one cohesin complex at sites different from the ones here tested, our observations pose limitations on how such interactions could occur .
Cohesin is a ring-shaped multi-subunit protein complex with several intriguing architectural features. Key to cohesin's function are the two Smc ATPase heads that close the ring on one side. Structural and biochemical evidence has been obtained both for direct head–head dimerisation as well as for a role of the Scc1 subunit in bridging a gap between the two heads. Furthermore, structural and functional evidence for an interaction of the Smc heads with the Smc hinge at the opposite side of the ring has been obtained. Which of, and when during cohesin's function in sister chromatid cohesion, these interactions occur remained largely uncharacterised. We have now used FRET to analyse the behaviour of the cohesin complex in live budding yeast.
FRET is a powerful technique to assess the proximity of interacting proteins . Recently, a practical method to measure FRET using CFP and YFP fluorophore fusions to budding yeast proteins expressed from their genomic loci has been introduced (). In the first instance, this technique was used to obtain a structural image of the core components of the yeast SPB. These measurements were facilitated by the concentration of many copies of each subunit within the small volume of the SPB. We now show that FRET measurements are also possible on a protein complex of moderate abundance and a more dispersed localisation in the yeast nucleus. We also use this technique to analyse possible conformational changes in the cohesin complex during the cell cycle. The results have shed new insight into the architecture of cohesin.
We found that the two ATPase heads are in constitutive close contact with each other throughout the cell cycle. Once synthesised at the G1/S transition, the Scc1 subunit maps to these heads in an unexpected configuration. In contrast to many models in which Scc1 bridges a gap between the Smc heads, we find that the subunit is more likely to lie across two interacting heads, perpendicular to what would be expected from a bridge. Scc1 is predicted to consist of folded domains at the N- and C-termini, connected by less structured central sequences. The N-terminus is thought to contact Smc3, and the C-terminus Smc1. In contrast, our FRET results suggest that the Scc1 C-terminus lies between and equidistant from both Smc heads. This is consistent with crystallographic analysis of the Scc1 C-terminus bound to the Smc1 head (), if we assume that Smc3 takes the place of the second Smc1 head in the homodimer crystal structure. Indeed the Scc1 C-terminus is so close to the predicted position of the Smc3 head, that it would be surprising if no contact existed between the two. Interaction studies with recombinant Scc1 fragments and the Smc head domains are not inconsistent with such a notion ().
This geometry could have important implications as to the role of Scc1 in the cohesin complex. The observed location of Scc1 is reminiscent of the C-terminal regulatory domain of the MalK ABC transport ATPase (). This domain stabilises the interaction between the ATPase heads, while those are undergoing a tweezers-like opening motion. In the case of cohesin, such a motion could be relayed onto the Smc hinge that might contact the ATPase heads opposite to Scc1 from inside the ring. The direct head–hinge interaction that we observe is consistent with this possibility. As suggested, this interaction could mediate opening of the hinge dimer during cohesin loading onto DNA (). The Smc heads might in this way never separate very far, at least until Scc1 is cleaved during anaphase. How exactly Scc1 cleavage leads to cohesin dissociation from chromosomes is not clear. Integrity of Scc1 might be important to mediate its stabilising function between Smc1 and Smc3, and after cleavage, its ability to connect the Smc heads might be disrupted (). An additional, not mutually exclusive, possibility is that the C-terminal cleavage product produced actively interferes with the Smc head interaction, thereby further reducing their affinity (). While this may well allow the Smc heads to separate and leave DNA to exit the ring during anaphase, our FRET results suggest that separation of the heads even in anaphase, if it occurs, is transient.
We also provide evidence that the Pds5 subunit contacts cohesin in an Scc1-dependent manner, and binds to the Smc hinge at the opposite side of the cohesin ring. This opens the possibility that Pds5 acts as a molecular matchmaker for an interaction between the Smc heads and hinge, as previously seen on atomic force microscopic images (). While we could biochemically demonstrate a robust interaction between the Smc1 head and hinge, this conformation may occur only transiently during the process of DNA binding (; ). Pds5 could facilitate the interaction by bringing together Scc1 and the Smc hinge from opposite sides of the ring. Such a matchmaker role could increase the efficiency of the head–hinge interaction, but may not be essential in all circumstances. This could explain why Pds5 is a dispensable subunit of cohesin in fission yeast, and tolerates reduction by RNA interference in human cultured cells. Pds5 may serve an additional role in maintaining the structural integrity of the cohesin complex during longer periods in G2 (; ).
In the future, two advances could allow such reactions to be studied. Ideally, FRET experiments with single molecules in reconstituted DNA binding reactions should allow a more detailed analysis of cohesin's behaviour. This approach is so far limited, in that cohesin loading onto DNA in a purified system has not yet been successfully reconstituted. In the interim, it could become possible to take advantage of the genetic amenability of budding yeast to engineer situations in which cohesin accumulates in intermediates of loading or unloading reactions. This could involve the analysis of mutant cohesin complexes, or of the wild-type complex in different mutant strain backgrounds. In an attempt to trap cohesin during DNA loading, we tried to analyse cohesin in yeast strains mutant for the cohesin loader subunit Scc2 (; ). However, increased background fluorescence at the higher restrictive temperatures required to inactivate Scc2 prevented us from analysing these strains further. This obstacle could be overcome by the generation of cold-sensitive mutant alleles. The analysis of mutations in cohesin subunit themselves, for example the ATPase motifs, poses a similar challenge. Mutant subunits that do not sustain cell viability have been studied after ectopic expression in addition to the endogenous copy (; ). This means that FRET analysis would be limited to a corresponding subset of cohesin complexes with accordingly lower fluorescent and FRET signals. The introduction of more sensitive imaging equipment might open such possibilities in the future.
Our studies so far have provided new insight into the architecture of the cohesin complex , and its behaviour during the cell cycle. Future studies will analyse the mechanism of cohesin, and that of related Smc protein complexes, at higher resolution, to understand their molecular activities in chromosome structure and dynamics.
Strains used in this study were diploid, homozygous for all genetic features, unless otherwise stated, and of the W303 background (a/α, Δ , , ). YEP medium was supplemented with either 2% glucose or 2% raffinose. To induce expression from the promoter, YEP+raffinose medium was supplemented with 2% galactose. Arrest in G2/M phase was achieved by addition of the spindle poison nocodazole at 5 μg/ml for 2 h. Small G1 cells were isolated by centrifugal elutriation, as described ().
N- and C-terminal tagging of genes at their genomic loci was performed by gene targeting using polymerase chain reaction (PCR) products (; ). Details of the YFP and CFP variants, the protocols used, as well as the plasmid templates themselves are available from the University of Washington Yeast Resource Center (). To construct a fluorophore insertion at the Smc1 hinge, we predicted surface loops in the hinge region by sequence alignment with the Smc hinge, for which a crystal structure has been determined (). Of two locations tested, insertion of YFP after proline yielded a Smc1 hinge-YFP derivative that fully complemented cell growth as the sole source of Smc1. The open reading frame until proline was cloned using PCR as an I/I fragment into YIplac128, and fused to a I/I fragment encoding the remainder of Smc1. Inserted into the I site was PCR amplified YFP (or CFP) flanked by linker peptides of the sequence VDGSTG on both sites. Next a 472-bp promoter PCR fragment was added upstream using I/I sites, and finally an additional 470-bp sequence upstream of the promoter fragment was amplified but cloned behind the open reading frame using I and III. This construct was linearised by I restriction for integration at the locus. Constructs for the expression of an Smc1 head and the Smc1/Smc3 hinge were as described (). The hinge domains included 50 amino acids of flanking coiled coil sequence.
Immunoprecipitation was performed by the addition of precleared yeast extracts to α-Pk (clone SV5-Pk1, Serotec)- or α-myc (clone 9E10)-conjugated protein-A–Sepharose beads (Sigma) for 90 min. The beads were extensively washed in extraction buffer EBX (50 mM HEPES/KOH pH 7.5, 100 mM KCl, 2.5 mM MgCl, 0.25% Triton X-100) containing protease inhibitors. For co-immunoprecipitation of Pds5 with Smc1, the concentration of KCl was reduced to 50 mM. Bound proteins were eluted in SDS–PAGE sample buffer and analysed by Western blotting.
Cells for FRET analysis were grown overnight at 30°C on YPD plates supplemented with 150 μg/ml adenine. Pinhead-sized colonies were scraped from the plate and resuspended in 12 μl SC medium. A 3 μl volume of the cell suspension was mounted on an agarose patch (1% SeaPlaque GTG agarose, Cambrex Bio Science, in SC medium) and covered with a coverslip. Cells were observed on a DeltaVision RT system (Applied Precision) based on an Olympus IX71 microscope. CFP excitation and emission filters used were 440AF21 and 480AF30, YFP excitation and emission filters used were 500AF25 and 545AF35 (the first number indicating the wavelength of maximum transmission and the second the bandwidth of the filters), and the dichroic mirror used was 436-510DBDR (all from Omega Optical). We used an × 100 UPUplan Apochromat (NA=1.4) objective, and images were captured with a CoolSNAP HQ camera (Roper scientific).
CFP and YFP spillover factors were measured in strains Y1967 and Y1970, expressing Smc1-CFP and Smc1-YFP, respectively. The CFP spillover factor () is the intensity of the Smc1-CFP signal in the FRET channel, divided by its intensity in the CFP channel, and was found to be 0.34±0.04 (=48). The similarly derived YFP spillover factor () was 0.09±0.04 (=46). Spillover factor measurements were repeated throughout the course of our studies and did not change significantly. FRET in the experimental strains containing fluorophore pairs was derived from three background-corrected intensity measurements, YFP, FRET and CFP, as follows.
This was then used to determine the FRET ratio (FRET). |
Granulocytes in mouse BM can be categorized into increasingly mature subsets on the basis of nuclear and granular morphology, or by the expression of CD11b and the RB6-8C5 antigen (Ag; , ). mAbs that are specific for CD11b and the RB6-8C5 Ag (Gr-1) identify three distinct double-positive populations in naive BL/6 mice: CD11bGr-1, CD11bGr-1, and CD11bGr-1 ( A). The CD11b Gr-1 cell population expressed significantly higher levels of c-Kit and CD16/32, both markers of developmental immaturity ( A). The neutrophil-specific mAb NIMP-R14 () efficiently labeled the CD11bGr-1 and CD11bGr-1 cells and bound a subset (60%) of the CD11bGr-1 compartment (unpublished data).
After sorting to high purity, CD11bGr-1 cells were found to be primitive, with little nuclear condensation and azureophilic cytoplasm ( B). The morphology of CD11bGr-1 cells is consistent with that of promyelocytes and myelocytes (with occasional metamyelocytes; reference ), the mitotic progenitors of all granulocytes. CD11bGr-1 cells were identical to mature peripheral neutrophils, whereas CD11bGr-1 cells represented an intermediate stage of neutrophil maturation (metamyelocytes and band forms) with faintly stained cytoplasm and incompletely condensed, ring-shaped nuclei ( B; reference ).
Proliferating cells in each neutrophil population were identified by determining the DNA content in fixed and permeabilized cells ( C; reference ). Consistent with its expression of c-Kit and morphology, the CD11bGr-1 population contained a high frequency (24 vs. 3%) of proliferating cells ( C). As controls, 19% of IgMB220 pro-/pre-B cells were in cycle compared with 6% in the mature, IgMB220 population ( C). Thus, CD11b and Gr-1 expression reveal three increasingly mature granulocyte populations. CD11bGr-1 cells are mitotically active promyelocytes and myelocytes; CD11bGr-1 cells are nondividing, immature neutrophils; and CD11bGr-1 cells are mature neutrophils.
Inflammation mobilizes lymphocytes from the BM without reducing Gr-1 cell numbers (). The subsequent granulocytosis that was elicited by inflammatory stimuli could result from a specific expansion of granulocytic progenitors or a general enhancement of the ability of BM to support hematopoiesis (). To investigate these possibilities, we followed the dynamics of developing and mature neutrophils and lymphocytes in the BM after immunization ().
4 d after immunization, the proliferating CD11bGr-1 neutrophil compartment grew three- to fourfold, whereas the mature CD11bGr-1 populations changed little ( A); coincidentally, B220 and B220 cell numbers decreased approximately threefold ( A; reference ). The reciprocal behavior between CD11bGr-1 neutrophils and B220 lymphocytes was repeated subsequently; as B220 BM cell numbers returned to normal, the expanded, immature neutrophil compartments diminished (days 8 and 12; A).
In contrast with the rapid changes in B220 lymphocyte and CD11bGr-1 neutrophil numbers, the mature neutrophil compartment of BM is more stable. In naive mice, a tibia and femur contain ∼7 × 10 CD11bGr-1 neutrophils; 4 d after immunization that number increases ∼40%, and by day 12 doubles to ∼1.4 × 10 ( A).
Adjuvant-induced expansion of the BM's primitive granulocytic compartment includes the c-KitCD11bGr-1 cell subset and occurs even as c-Kit pro–B cells are lost from BM ( B; reference ). This reciprocity suggests a regulatory mechanism that acts by reducing lymphopoietic resources while sparing factors that are important for granulocyte development.
The inflammation-induced emigration and subsequent recovery of BM lymphocytes correlates with changes in CXCL12 mRNA and protein expression (). Mindful that CXCL12 is crucial for granulocytic and lymphoid progenitors to colonize the BM (–), we used quantitative RT-PCR to identify differentially regulated factors that might expand granulopoiesis in BM that is transiently deficient in CXCL12.
mRNA specific for B cell–activating factor (BAFF), CXCL2, CXCL12, GAPDH, GM-CSF, hypoxanthine-guanine phosphoribosyltransferase (HPRT), IL-1β, IL-3, IL-7, and SCF were quantified relative to GAPDH message at various times after immunization ( and A). As expected, CXCL12 message decreased ∼10-fold at days 4 and 8 after immunization (); recoveries to 50% of naive levels were achieved by day 12; and 16 d after immunization, CXCL12 mRNA levels were ≥10-fold higher than naive controls ( A). SCF mRNA followed a similar kinetics, and decreased 10-fold by day 4 and recovered to supranormal levels by day 16. Immunization decreased BAFF and IL-7 message levels by not >50%, a decrease that also was observed in the HPRT control ( A). However, although HPRT message returned to naive levels, mRNA that was specific for IL-7 and BAFF became elevated (170% and 400% of naive controls, respectively) 16 d after immunization. Overexpression of CXCL12, SCF, BAFF, and IL-7 in BM coincides with recovery of central B lymphopoiesis and a return to normal levels of granulocyte production ( A; reference ).
In contrast, mRNAs that are specific for IL-1β and CXCL2, a potent activator of granulopoiesis () and neutrophil chemokines (), respectively, were not changed significantly by immunization until day 12; both then decreased to 30–40% of naive levels. mRNA levels in BM for two important myelopoietic growth factors, IL-3 and GM-CSF, were below the detection limit (≤10 GAPDH levels; ) of our PCR assay.
Inflammatory reductions of CXCL12 message in BM correlate with lower levels of CXCL12 protein (). To determine if reduced SCF mRNA also decreased BM SCF protein, SCF in the BM plasma (BMp) of naive and immunized mice (day 4; = 4) was quantified by ELISA. SCF levels in naive mice averaged 163 pg/ml BMp; BMp from immunized mice contained only 35 pg/ml SCF, ∼21% of control levels (unpublished data).
To determine whether reductions in SCF and CXCL12 expression might depress lymphopoiesis while sparing granulopoiesis, we determined the effects of reduced c-Kit signaling on the proliferation and differentiation of common lymphocyte progenitors (CLP) and common myeloid progenitors (CMP) in vitro (, ). In brief, ∼100 CLP or CMP were introduced into established cultures of adherent BM cells that were supplemented with lymphoid or myeloid growth factors. PTX or antagonistic mAb specific for LFA-1 or c-Kit were added to replicate cultures. 4 d later, the effects of these treatments on CLP and CMP differentiation were determined by enumerating B220 and Gr-1 cells ( B).
PTX efficiently inhibits proliferation of CLP and CMP, and reduces the production of differentiated B220 and Gr-1 cells to ∼50% of control cultures ( B). Although mAb to LFA-1 had little effect (P > 0.05) on CLP or CMP cultures, addition of c-Kit mAb significantly inhibited (P <0.01) B220 cell production from CLP (27% of controls), but had little activity in CMP cultures (80% of controls; B).
Similarly, whereas B220 and Gr-1 cells migrated to BMp in vitro, lymphocyte migration was inhibited by mAb to CXCL12 far more effectively (13% of controls; P = 0.04) than granulocyte migration (≤50% of controls; P = 0.05), despite the greater sensitivity of Gr-1 cells to the dilute source of chemoattractant ( C; reference ). mAb to CXCL5 had no effect on B220 or Gr-1 cell migration. In these in vitro assays, SCF and CXCL12, two BM growth factors specifically diminished by inflammation (), preferentially affect lymphocyte progenitors.
The different effects of anti-CXCL12 on the ability of developing B cells and neutrophils to migrate toward BMp ( C) suggest distinct sensitivities to G signals. Although PTX depletes the BM of lymphocytes, it increases the numbers of Gr-1 BM cells ().
To determine the relative importance of G signaling in the homing of BM lymphocytes and granulocytes, we incubated BM cells (1.5 × 10) from B6.SJL mice (CD45.1) in media which contained 0, 25, or 100 ng/ml of PTX and then injected them into BL/6 (CD45.2) recipients. The next day, we enumerated CD45.1 neutrophils and B cells in BM to determine homing efficiencies.
CD45.1Gr-1 and CD45.1IgMB220 cells were detected easily in recipient BM 14 h after transfer of untreated cells; treatment with PTX reduced the homing efficiency of mature CD45.1CD11bGr-1 neutrophils and IgM B220 pro–/pre–B cells more than fivefold (). In contrast, pretreatment with PTX resulted in nonsignificant, twofold reductions in the numbers of CD45.1CD11bGr-1 pro-/myelocytes that returned to the BM (). Homing by immature, CD45.1CD11bGr-1 neutrophils also was resistant to PTX. Doses of PTX that abrogated BM reentry by mature, CD11bGr-1 neutrophils and B220 lymphocytes reduced homing by immature neutrophils <20% (). Thus, CD11bGr-1 and CD11bGr-1 granulocytes localize in BM by a mechanisms that is relatively insensitive to disruption of G-coupled signaling; homing by B220 lymphocytes and mature, CD11bGr-1 neutrophils is PTX-sensitive.
The reciprocal expansion of the Gr-1 BM compartment following adjuvant-induced emigration of B220 lymphocytes (; reference ) suggested to us that granulocytes and lymphocytes might occupy a common developmental niche. If so, Gr-1 and B220 cells must co-locate within the BM. Histologic sections of femurs from naive mice ( A) demonstrate that Gr-1 and B220 cell clusters are interspersed closely. 4 d after immunization, expansion by Gr-1 cell clusters is obvious, whereas B220 lymphocyte clusters become rare ( B). The proximity of granulocytes and B lymphocytes in the BM is consistent with a shared developmental niche but does not exclude the possibility of a tight mosaic of sites that is specific for myelo- or lymphopoiesis.
To determine whether granulocytes and B lymphocytes can develop within a common microenvironment, we prepared cultures of BM stromal cells free of hematopoietic progenitors () and seeded them with autologous c-Kit B220 and c-KitGr-1 cells labeled by carboxyfluorescein diacetate, succinimidyl ester (CFSE; green) or Cell Tracker orange (CMRA; red). 1 d later, the numbers and distributions of B220 (CFSE/green) and Gr-1 (CMRA/red) cells in cultures with and without stromal cell layers were identified by fluorescence microscopy.
In the absence of stromal cells, brightly fluorescent, nonadherent B220 and Gr-1 cells were distributed uniformly and relatively infrequent ( C). In the presence of stromal cell layers, B220 and Gr-1 cell numbers were increased substantially (not depicted) and many B220 and Gr-1 cells had migrated beneath stromal cells to form “cobblestone clusters” ( D). The majority of cobblestone clusters contained B220 and Gr-1 cells that were dimly fluorescent and suggestive of cell division; mixed cell populations often were beneath a single stromal cell and demonstrated that c-KitB220 and c-KitGr-1 cells can develop in a common microenvironment ( D).
If the B220 and Gr-1 BM compartments also share a developmental niche in vivo, reductions in one lineage will result in a specific and compensatory expansion of the other (; reference ). To test this prediction without disturbing the cytokine milieu of BM (), we enumerated myeloid (Gr-1), lymphoid (B220), and erythroid (TER119) cells in the femurs of normal and RAG1-deficient mice.
mice ( = 5) were labeled with fluorochrome-coupled mAbs specific for the B220, TER119, or RB6-8C5 Ags; labeled and unlabeled cells from individual femurs were enumerated by flow cytometry.
On average, femurs from control mice contained 155.9 ± 15.9 × 10 nucleated cells comprising (all ×10) 76.1 (±8.5) B220 cells; 42.9 (±3.0) Gr-1 cells; 18.3 (±4.6) TER119 cells; and 18.5 (± 2.6) unlabeled cells ( E).
mice (130.6 ± 13.4 × 10) were significantly (P = 0.002) lower than that of BL/6 controls, with all cell losses confined to B220 compartments ( E). In RAG1-deficient BM, cells with a pre-/pro-B phenotype were increased (from 2 × 10 to 10 × 10), which was consistent with blockade of D-to-J joining (unpublished data; reference ). Despite this expansion, femurs of RAG1-deficient mice contained only one third as many B220 cells (27.5 ± 3.3 × 10 vs. 76.1 × 10) as controls. Without compensating increases in other BM cell compartments, femurs from RAG1-deficient mice should have contained only ∼108 × 10 nucleated cells.
Instead, losses in B220 BM cell numbers were offset by increased numbers of Gr-1 cells. The mean number of Gr-1 femoral cells in RAG1-deficient mice was significantly higher (P < 0.001) than in controls; they increased from 42.9 ± 3.0 × 10 cells to 62.9 ± 9.5 × 10 cells. This increase represented expansions in all neutrophil compartments, including the proliferating CD11bGr-1 cell subset (P = 0.01).
and control mice contained virtually identical numbers of TER119 (20.0 ± 2.6 × 10 vs. 18.3 ± 4.6 × 10, respectively; P = 0.36) and unlabeled cells (20.2 ± 2.8 × 10 vs. 18.5 ± 2.6 × 10; P = 0.21). Removal of BM lymphocytes by a specific developmental blockade results in the compensatory expansion by Gr-1 cells only. The specific and reciprocal coupling between Gr-1 and B220 cell numbers and their co-localization defines a common developmental niche for BM granulocytes and lymphocytes.
Leukocyte production in the BM is controlled by inflammatory cytokines. In concert, TNFα and IL-1β induce the loss of BM lymphocytes by emigration while significantly expanding granulocyte production (). This expansion is realized by increasing CD11bGr-1 pro-/myelocyte numbers () in concordance with lymphocyte losses (). We conclude that the expansion of this generative, CD11bGr-1 compartment fuels the reactive CD11b Gr-1, CD11bGr-1 neutrophilia of inflammation and infection ( and ; references –, ).
Inflammation's reciprocal effects on BM lymphopoiesis and granulopoiesis could be coupled or independent; however, coupled regulation is the more parsimonious explanation to link these leukocyte compartments. To identify an inflammation-dependent mechanism that regulates granulopoiesis and lymphopoiesis we assayed mRNA specific for a variety of growth and localization factors (CXCL 12, CXCL 2, IL-1β, IL-3, IL-7, GM-CSF, SCF, BAFF) that influence lymphocytes and neutrophils (; ). We observed that inflammation's effects on the BM are specific; immunization with adjuvant significantly reduced only CXCL12 and SCF, and these reductions correlated exactly with the depletion of BM lymphocytes and the expansion of CD11bGr-1 cell numbers (; reference ). Although our search for regulated factors was not exhaustive, it is significant that in vitro and in vivo, reductions in CXCL12 and SCF preferentially disturb CLP and B220 lymphocyte development and localization with no or lesser effects on CMP and CD11bGr-1 granulocytes ( and ). Adult viable c-Kit–deficient mice exhibit normal myelopoiesis with little or no B-lymphopoiesis (); administration of AMD3100, a specific antagonist of the CXCL12 receptor, CXCR4, mobilizes CD34 human stem cells but not myelocytes or metamyelocytes (). Given this biased activity, how could reductions of CXCL12 and SCF effect granulopoietic expansions?
An attractive possibility is that the reciprocal control of granulopoiesis and lymphopoiesis reflects competition between the generative CD11bGr-1 and B220 compartments in the BM. Thus, when inflammation reduces the capacity of the BM to retain lymphocytes and promote lymphopoiesis ( and ; reference ), granulocyte precursors acquire increased resources for proliferation and/or survival and expand into areas that are vacated by migrant lymphocytes. A prediction of this hypothesis is that lymphocyte and granulocyte progenitors occupy and compete in a common BM niche.
Histologic examination of femoral BM reveals B220 and Gr-1 cells which are organized in small, adjacent clusters ( A). Immunization with adjuvant alters this distribution by dramatically expanding Gr-1 cell areas and reducing B220 cell zones ( B); these histologic changes match the dynamics of the Gr-1 and B220 cell numbers after immunization (). To determine whether B lymphocyte and granulocyte development can occur in a common microenvironment, we cultured c-KitB220 and c-KitGr-1 cells with autologous BM stromal cells that were depleted of hematopoietic activity (). These cultures supported large numbers of developing leukocytes that had migrated beneath the stromal cell layer to form cobblestone clusters. These clusters contained Gr-1 and B220 cells, often beneath a single stromal cell body ( D). The close association of B220 and Gr-1 cell clusters in the BM ( A) and their adjacent development in vitro ( D) provide strong support for our hypothesis that these cell compartments share a developmental niche in the BM.
If developing granulocytes and lymphocytes compete for a BM niche, reductions in one compartment must result in expansions by the other.
mice and their normal, congenic controls. Losses of B220 cells in the BM of RAG1-deficient mice were compensated only by increased numbers of Gr-1 cells; the numbers of TER119 erythroid lineage and unlabeled cells in RAG1-deficient and control mice were virtually identical ( E).
Our lymphocyte ablation experiment had three possible outcomes (). If B220, Gr-1, and TER119 cells occupy separate BM niches, reduction in any one cellular compartment will not affect the others. Conversely, if all hematopoietic lineages compete generally in BM, losses in any one compartment would be compensated by expansions of all of the others. Neither of these outcomes was observed in RAG1-deficient mice. Instead, reduced B220 cell numbers expanded only the Gr-1 cell compartment. This specific and reciprocal coupling between B220 and Gr-1 BM cell numbers defines a common developmental niche () that may be visualized in vitro and in vivo (). Granulopoiesis and lymphopoiesis compete in the BM, whereas under these experimental conditions, erythropoiesis is independent of both.
The relative abundance of mRNAs for lymphoid growth factors () suggests that lymphopoiesis and granulopoiesis are balanced normally by a surfeit of growth resources that are lymphoid-specific or -preferent ( and ; ). If so, reductions in CXCL12 and SCF would reduce BM lymphocyte numbers directly but promote granulopoiesis indirectly, by removing lymphocyte competitors. Reductions in SCF (mimicked by inhibiting c-Kit signaling) and CXCL12 primarily inhibit the activities of lymphocyte progenitors ( and ). These biased effects provide a mechanism to shift the leukopoietic equilibrium in the BM toward granulopoiesis and the support of neutrophilic responses to infection (, ).
The new leukopoietic equilibrium that is established by inflammation may require additional signals to normalize. We note that reestablishment of normal lymphopoietic levels coincides with supranormal expression of lymphoid growth factor genes and decreased IL-1β and CXCR2 message (). Excess lymphoid support may be necessary to restore lymphoid progenitors to BM niches that are colonized by progenitor and immature granulocytes, and thereby, reverse increased granulopoiesis.
The reductions of BM CXCL12 and SCF elicited by the proinflammatory cytokines, TNFα and IL-1β (), preferentially affect lymphopoiesis and lymphocytes over the other hematopoietic compartments ( and ). These differential effects and competition between the BM's B220 and Gr-1 compartments () offer a simple, but robust, mechanism for innate immunity to regulate granulopoiesis and lymphopoiesis. During infection, acute granulocytic responses are crucial for host protection; in contrast to lymphocytes, mature granulocytes are very short-lived and are incapable of proliferation, and effective granulocytic responses require expanded leukopoiesis in the BM.
Why does the specialization of the BM lead to peripheral lymphopoiesis? We note that neutrophils, especially those elicited by myelopoietic growth factors and/or proinflammatory cytokines, release large amounts of BAFF (, ). Perhaps the reactive neutrophilia that is elicited by TNFα and IL-1β amplifies extramedullary lymphopoiesis by increasing local levels of BAFF to support cell survival and/or differentiation to plasmacytes.
C57BL/6 (BL/6, CD45.2), congenic B6.SJL-Ptprca/BoAiTac (B6.SJL, CD45.1), B6.129S7-Rag1/J were purchased (Jackson ImmunoResearch Laboratories or Taconic Farms). Mice were housed in specific pathogen-free conditions at the Duke University Animal Care Facility and given sterile bedding, water, and food. Mice used in these experiments were 6–18 wk of age; these studies were reviewed and approved by the Duke University Institutional Animal Care and Use Committee.
Mice were immunized by i.p. injections of (4-hydroxy-3-nitrophenyl)acetyl-chicken γ globulin (50 μg) emulsified in IFA (Sigma-Aldrich; reference ); (4-hydroxy-3-nitrophenyl)acetyl-chicken γ globulin contained 10 or 12 mol nitrophenyl/mol chicken γ globulin.
FITC, PE, biotin, or allophycocyanin (APC), PE-Cy5–conjugated mAb specific for mouse B220, IgM, CD11b, CD4, CD8, CD40, TER119, Thy1.2, CD117, CD45.1, F4/80, or RB6-8C5 were purchased from BD Biosciences or eBioscience. The Gr-1 mAb binds RB6-8C5 Ag, a marker of granulocyte development (). CD93-specific 493 mAb () was purified from culture supernatants in our laboratory. Texas red (TXR) conjugates of antibody (Ab) for mouse IgM were purchased from Southern Biotechnology Associates, Inc. We used the neutrophil-specific mAb, NIMP-R14 (), to confirm histologic identification of granulocyte subsets.
BM cells were labeled with FITC-conjugated Gr-1 mAb, PE- or TXR-conjugated anti-IgM, and APC-conjugated anti-B220. CD11bGr-1IgMB220, CD11bGr-1IgMB220, and CD11bGr-1IgMB220 cells were doubly sorted to >98% purity in a FACS Vantage SE flow cytometer.
CMP and CLP were recovered from BM as described previously (, ). In brief, BM cells were labeled with PE-CY5–conjugated mAbs specific for CD4, CD8, TER119, Gr-1, CD11b, and B220. Labeled, Lin cells were depleted by incubation with anti-PE magnetic beads (Miltenyi Biotec; GmbH). To isolate CMP, Lin cell populations were stained with APC-conjugated anti–c-Kit mAb, TXR-conjugated anti–Sca-1, FITC-conjugated anti-CD34, and PE-conjugated anti-CD16/32; CLP were identified by labeling with APC-conjugated anti–c-Kit, TXR-conjugated anti–Sca-1, FITC-conjugated anti-Thy1.2, and PE-conjugated anti–IL-7R. CMP and CLP were recovered by sorting with a FACS Vantage SE flow cytometer.
BM cells that were labeled by various combinations of fluorochrome-mAb conjugates were fixed in 1% paraformaldehyde at 4°C for 30 min and subsequently incubated with 1 μg/ml propidium iodide in 2% saponin/PBS at 4°C for 30 min. The DNA content in various cell populations was determined by quantifying levels of bound propidium iodide by flow cytometry (). Cells containing 2N (G) or >2N levels (G and S) of DNA were enumerated. Doublet cells were excluded by forward and side-scatter gating ().
mRNA from BM cells was precipitated in Trizol reagent (Invitrogen) and reverse transcribed (Superscript II; Invitrogen). Quantitative PCR amplifications of cDNA were performed in an iCycler thermal cycler (Bio-Rad Laboratories) with SYBR Green PCR core reagents (Applied Biosystems) and primers specific for BAFF, CXCL2, CXCL12, GAPDH, GM-CSF, HPRT, IL-1β, IL-3, IL-7, or SCF transcripts (). Standard amplification parameters for the quantitative PCR were as follows: initial denaturation at 94°C for 10 min; amplification cycle; denaturation at 94°C for 45 s; and anneal/extension at 64°C for 45 s. The relative expression levels for growth factor genes were calculated by the comparative C (threshold cycle) method recommended by the manufacturer (Applied Biosystems) normalized to GAPDH message in the same sample. In brief, ΔC values were determined by subtracting C from C. Expression levels relative to GAPDH were defined as: 2
For the individual genes studied, expression data were standardized to averaged, homologous message in naive mice ( = 3–5); mRNA levels in the BM of naive mice for GAPDH, HPRT, BAFF, CXCL2, CXCL12, SCF, IL-1β, IL-7, IL-3, and GM-CSF vary over a ≥10-fold range ().
The ability of cells to localize in the BM was determined by cell transfer. In brief, 2 × 10 BM cells from B6.SJL (CD45.1) mice were incubated at 37°C in media containing PTX (0, 25, 100 ng/ml) for 1 h, washed, and injected i.v. into naive, congenic C57BL/6 recipients (CD45.2). Femoral and tibial BM cells were harvested 14 h later and labeled with mAb specific for CD45.1, the RB6-8C5 Ag, CD11b, B220, or IgM. Labeled CD45.1 cells were analyzed and enumerated by flow cytometry to determine migration efficiencies.
BMp was prepared by flushing femurs and tibias from four mice with 500 μl RPMI 1640 containing 0.5% BSA and 10 mM Hepes (migration media). Suspended cells were transferred into Eppendorf-type centrifuge tubes; cells and debris were removed by repeated (2×) centrifugation at 5,000 for 10 min.
BM cell chemotaxis was quantified in 24-well Transwell plates (Corning Costar) containing cells at 5 × 10/ml in migration media. Mouse CXCL5 (PeproTech) or CXCL12α (PeproTech) or BMp was added to migration media to specific concentrations, and 500-μl aliquots were placed in the lower chambers of plates. In some experiments, neutralizing Ab to mouse CXCL12 (100 μg/ml; R&D Systems) or CXCL5 (50 μg/ml; R&D Systems) also was added. The 5-μm pore inserts were placed on the wells and preconditioned at 37°C for 1 h; BM cells (5 × 10) were loaded (100 μl) into the upper chambers. Cells were allowed to migrate into the lower chamber for 4 h at 37°C. Subsequently, cells in the upper and lower chambers were incubated with antibodies specific for B220 and the RB6-8C5 Ag; labeled cells in duplicate cultures were enumerated by microscopy. In some experiments, the BM cells were pretreated with 100 ng/ml PTX for 1 h at 37°C before their introduction into the migration plate.
c-KitIgMB220 and c-KitCD11bGr-1 BM cells were sorted and labeled with CFSE IMS (Molecular Probes, Inc.) or CMRA (Molecular Probes, Inc.). Labeled cells were cultured in established, long-term BM cultures containing ∼2.5 × 10 stromal cells (); 1 d after co-culture, CFSE- and CMRA-labeled cells were analyzed by fluorescence microscopy.
CMP or CLP were cultured with the OP9 stromal cell line () in the presence of specific cytokines to promote proliferation and differentiation. For CMP cultures, 10 ng/ml IL-3 (R&D Systems) and GM-CSF (PeproTech) were added; CLP cultures contained 10 ng/ml IL-7 (R&D Systems) and Flt3-ligand (R&D Systems). After 5 d of culture, total myeloid or lymphoid cell numbers were determined by microscopy.
SCF levels in BMp were determined in naive, control, and adjuvant immunized mice (day 4) by the ELISA methods described by Ueda and colleagues (). In brief, 96-well plates were coated with goat anti-SCF Ab (AF-455-NA); bound SCF was detected with biotinylated with a second anti-SCF goat Ab (BAF455; both R&D Systems). rMouse SCF (R&D Systems) was used to calibrate the assay and determine its detection limit (1 pg/ml).
Statistical significance (P ≤ 0.05) of data was determined by Student's test. |
Genetic studies in first identified Polycomb group (PcG) genes as regulators that are required for the long-term repression of HOX genes during development (reviewed in ). To date, 17 different genes in are classified as PcG members because mutations in these genes cause misexpression of HOX genes (reviewed in ). All PcG genes are also conserved in mammals and at least some of them are also conserved in plants (reviewed in ; ; ). In all these organisms, PcG gene products function as repressors of HOX and/or other regulatory genes that control specific developmental programs (reviewed in ). Moreover, recent studies that analyzed genome-wide binding of PcG proteins in and in mammalian cells identified a large number of target sites, and thus a whole new set of genes that potentially is subject to PcG repression (; ; ; ; ).
Biochemical purification and characterization of PcG protein complexes has advanced our understanding of the PcG system. To date, three distinct PcG protein complexes have been isolated from : PhoRC (), PRC1 (; ) and PRC2 (; ; ; ). The composition and activities of these different complexes and current views on the mechanisms by which these complexes might repress transcription of target genes have been discussed in recent review articles (; ; ).
Biochemically purified PRC2 contains the three PcG proteins Enhancer of zeste (E(z)), Suppressor of zeste 12 (Su(z)12) and Extra sex combs (Esc) and, in addition, Nurf55, a protein that is present in many different chromatin complexes (; ). PRC2 and the homologue mammalian complex are histone methyltransferases (HMTases) that specifically methylate H3-K27 in nucleosomes (; ; ; ). Chromatin immunoprecipitation (X-ChIP) analyses in showed that PRC2 binds in a localized manner at Polycomb response elements (PREs) of target genes, but that H3-K27 trimethylation is present across the whole upstream control, promoter and coding region of these genes (; ; ; ). Studies that compared the inactive and active state of the HOX gene in developing found that PRC2 is constitutively bound at PREs and, surprisingly, that the whole upstream control region is constitutively trimethylated at H3-K27 (). However, presence or absence of H3-K27 trimethylation in the promoter and coding region correlates tightly with the gene being repressed or active, respectively (). H3-K27 trimethylation is thus a distinctive mark of PcG-repressed chromatin.
Analysis of mutants suggests that E(z) is also responsible for the genome-wide H3-K27 mono- and dimethylation that has been reported to be present on more than 50% of H3 in (). However, biochemical analyses showed that E(z) protein alone does not bind to nucleosomes and is virtually inactive as an enzyme; E(z) needs to associate with Su(z)12 and Nurf55 for nucleosome binding and with Esc for enzymatic activity (; ; ; ). This implies that the genome-wide H3-K27 mono- and dimethylation is generated by PRC2 or another E(z)-containing complex that is able to interact in a non-targeted manner with nucleosomes across the whole genome. Conversely, this raises the question whether H3-K27 trimethylation at PcG target genes is simply a consequence of PRC2 being targeted to PREs or whether additional features such as post-translational modifications or associated factors are required.
Previous studies reported that the PcG protein Polycomblike (Pcl) interacts with E(z) in GST pull-down, yeast two-hybrid and co-immunoprecipitation assays (; ). Like most other PcG proteins, Pcl has also been found to be bound at PREs in (; ). However, to date, no Pcl-containing complexes have been purified and the role of Pcl in PcG repression has remained enigmatic. In this study we report the biochemical purification of Pcl complexes. We show that Pcl exists in a stable complex with PRC2. Our analyses demonstrate that this Pcl complex plays a critical role in generating high levels of repressive H3-K27 trimethylation at PcG target genes.
We used a tandem affinity purification (TAP) strategy () to purify Pcl protein complexes from embryos. To this end, we first generated transgenic strains that express a TAP-tagged Pcl fusion protein (TAP-Pcl) under the control of the a-tubulin promoter. Using a genetic rescue assay, we found that this TAP-Pcl fusion protein is functional and can substitute for endogenous Pcl. Specifically, animals that are homozygous for , a protein-negative allele of (see below), die at the end of embryogenesis but are rescued into viable and fertile adults if they carry the transgene expressing TAP-Pcl protein (see Materials and methods).
Following the TAP procedure, we purified proteins associated with TAP-Pcl from nuclear extracts that we prepared from TAP-Pcl transformant embryos. The purified material was separated on SDS-polyacrylamide gels and silver staining of the gel revealed five protein bands that consistently co-purified with TAP-Pcl (). Sequencing of peptides from these bands by nanoelectrospray tandem mass spectrometry identified these proteins as the PRC2 components E(z), Su(z)12, Nurf-55, Esc and, in addition, heat-shock cognate protein HSC70 (; ). LC-MS/MS analysis of total purified material confirmed that these polypeptides are the main co-purifying proteins (). The observation that PRC2 components E(z), Su(z)12, Esc and Nurf-55 co-purify with Pcl suggests that Pcl and PRC2 constitute a specific form of PRC2 that we name Pcl-PRC2.
To confirm the association of Pcl with PRC2, we also used the TAP-tag strategy to purify proteins associated with TAP-E(z) protein in embryos. Mass spectrometry and western blot analyses revealed that Pcl protein indeed co-purifies with TAP-tagged E(z), albeit clearly at substoichiometric amounts compared to the PRC2 core components (; ). In contrast, we note that the material purified either with TAP-Pcl or with TAP-E(z) did not contain detectable amounts of the PhoRC component Pho, the PRC1 component Pc, or HDAC1/RPD3 ( and ; ), three proteins that have been reported to interact with PRC2 subunits in binding or co-immunoprecipitation assays (; , ; ). Together, these results further strengthen the view that Pcl-PRC2 is a stable biochemical entity. One possible explanation for the presence of substoichiometric amounts of Pcl in the TAP-E(z) purification would be that Pcl-PRC2 is a distinct form of PRC2 and that only a fraction of PRC2 in the cell is associated with Pcl. An alternative possibility would be that Pcl is a less tightly associated component of PRC2 that tends to dissociate from the complex during biochemical purification. In this context, it is interesting to note that the Hsc70-4 protein is present in the material purified with either TAP-Pcl or TAP-E(z). Hsc70-4 has also been reported to co-purify with both and human PRC1 (; ). At present it is not known whether Hsc70-4 is required for assembly, stability or function of Pcl-PRC2, PRC2 and PRC1, and further studies will be needed to address these questions but, intriguingly, and also interact in genetic assays ().
Finally, we explored how Pcl physically interacts with PRC2 subunits. To this end, we used baculovirus expression vectors to coexpress Flag-tagged Pcl with individual PRC2 core components in Sf9 cells, and then used the Flag-epitope for affinity purification from Sf9 cell extracts. We could thus reconstitute a stable dimeric Pcl–E(z) complex (), consistent with previous studies that reported physical interactions between Pcl and E(z) in GST pull-down assays (; ). Interestingly, the baculovirus reconstitution assay revealed that Pcl also forms a stable complex with Nurf55 and, albeit less efficiently, also with Su(z)12 (). In contrast, purification from cells coexpressing Flag-Pcl and Esc resulted in the isolation of Flag-Pcl protein only (). Together, these data suggest that Pcl associates with PRC2 though interactions with E(z), Nurf55 and Su(z)12.
We next compared the HMTase activity of Pcl-PRC2 with the activity of PRC2. For these experiments we used recombinant tetrameric PRC2 (rPRC2) that we reconstituted as previously described (; ). Attempts to generate recombinant Pcl-PRC2 of comparable quality with stoichiometric quantities of the five components have been unsuccessful, and we therefore used Pcl-PRC2 purified from for these assays. As substrate in the HMTase reactions we used reconstituted recombinant mononucleosomes that contained either wild-type histone H3 or mutant forms of H3, in which lysine 9 (H3) or lysine 27 (H3), or both (H3/) had been mutated to alanine. Pcl-PRC2 and rPRC2 both methylated histone H3 at K27. This specificity is demonstrated by the finding that mononucleosomes containing wild-type H3 or H3 are efficiently methylated by either complex, but that no methylation is observed on mononucleosomes containing H3 or H3/ (). However, purified Pcl-PRC2 appears to be more active than rPRC2 because three-fold molar excess of rPRC2 was needed to obtain comparable H3-K27 methylation signals (). At present it is not known whether this relatively higher HMTase activity of Pcl-PRC2 is due to inclusion of Pcl, because it could also be due to post-translational modifications that are present on other subunits of natively purified Pcl-PRC2 but are missing on rPRC2. Mass spectrometry analysis of the H3 protein band isolated from mononucleosomes after HMTase assays independently confirmed H3-K27 as the site being methylated, and it revealed that Pcl-PRC2 and rPRC2 both trimethylate H3-K27 (). Interestingly, reaction with either complex generated comparable summed intensities of mono-, di- and trimethylated H3-K27, suggesting that both complexes are comparably efficient in generating the different methylated states of H3-K27 (). Taken together, these data suggest that Pcl-PRC2 and PRC2 are both H3-K27-specific HMTases that methylate H3-K27 in nucleosomes .
To study the role of Pcl-PRC2 in relation to that of PRC2, we next analyzed mutants that lack Pcl protein. For these studies we used the loss-of-function allele that we recently isolated (see Materials and methods). appears to be a protein-negative allele because no Pcl protein is detected in extracts prepared form 16- to 18-h-old homozygous embryos (). Importantly, the levels of Su(z)12 and E(z) protein in such homozygous embryos are indistinguishable from those observed in wild-type embryos (); other PRC2 subunits are thus stable in the absence of Pcl protein. This suggests that mutants specifically lack the function of Pcl-PRC2 but retain normal levels of PRC2.
We then tested whether Pcl might be required for H3-K27 methylation . In a first set of experiments we generated clones of homozygous mutant cells in imaginal discs of developing larvae and analyzed the global level of H3-K27 mono- and trimethylation in the mutant cells by immunostaining wing discs with antibodies against H3-K27me1 and H3-K27me3, respectively. Antibodies against H3-K27me2 gave strong staining signals in the cytoplasm of imaginal disc cells and this precluded the analysis of H3-K27 dimethylation in these experiments (data not shown). In parallel to analyzing mutant clones, we also analyzed wing discs with clones of cells that were homozygous for null mutations in or , respectively. In all these experiments, discs were analyzed 96 h after clone induction and the clones of mutant cells were identified by absence of a GFP-expressing marker gene. In or mutant clones in wing discs, H3-K27me3 and H3-K27me1 signals are reduced to undetectable levels (). The loss of H3-K27me3 and H3-K27me1 signals in cells lacking these two PRC2 core components is consistent with earlier reports that showed that all H3-K27 methylation in depends on E(z), the catalytic subunit of PRC2 (; ; ). Unexpectedly, we found that H3-K27me3 levels are also reduced in mutant clones. Although the reduction is not as severe as in or mutant clones, it is consistent in all discs examined (). In striking contrast, the level of H3-K27me1 signal in mutant clones is not diminished and is indistinguishable from that of wild-type cells (). Taken together, these results suggest that, in , Pcl-PRC2 is required for H3-K27 trimethylation, but is apparently not required for H3-K27 mono-methylation.
In an independent set of experiments, we compared the levels of total H3-K27me1, H3-K27me2 and H3-K27me3 in wild-type, , and mutant animals by Western blot analysis. Like other PcG genes, , and are all expressed in the female germline, and wild-type products, deposited into the egg by the mother, rescue the homozygous mutant embryos to a certain extent (; ; ; ). and homozygotes thus develop even into larvae, whereas homozygotes die at the end of embryogenesis. The imaginal disc tissues in or homozygous second instar larvae are only poorly developed, and the cells stop proliferating but, remarkably, they still contain substantial levels of H3-K27me1, H3-K27me2 and H3-K27me3. Compared with wild-type control larvae, we only found H3-K27me1 and H3-K27me2 signals to be detectably reduced in these mutants (). This strongly suggests that H3-K27 methylation and in particular H3-K27 trimethylation that was generated by maternally deposited PRC2 or Pcl-PRC2 in the early embryo persists into the larval stages. Similarly, in 16- to 18-h-old homozygous embryos, H3-K27me1, H3-K27me2 and H3-K27me3 Western blot signals are indistinguishable from those in wild-type embryos (), even though maternally deposited Pcl protein is no longer detected in homozygotes at this stage (). The inability to detect changes in H3-K27 methylation levels in mutants in this assay could have different reasons. Like in the case of or homozygotes, it is possible that the H3-K27me3 signals present in mutant embryos represent H3 molecules that were methylated early in embryogenesis by maternally deposited Pcl–PRC2 complexes. Alternatively, it could be that Pcl is not required for H3-K27 methylation in embryos and, finally, it is possible that H3-K27me3 levels are not globally reduced in mutant embryos but are perhaps only reduced at particular target genes.
To distinguish between these possibilities, we performed X-ChIP assays to monitor the levels of H3-K27me3, H3-K27me2 and H3-K27me1 at PcG target genes in wild-type and in homozygous embryos. Recent X-ChIP on chip studies in tissue culture cells and in embryos identified a number of genes to which PRC2, PRC1 and PhoRC components are bound and that also contain chromatin that is trimethylated at H3-K27 (; , ; K Oktaba and J Müller, unpublished). Among those, we used the (), (), (), (), (), (), (), () and () genes for our analysis. We thus prepared chromatin from 16- to 18-h-old wild-type or homozygous embryos and performed X-ChIP assays with antibodies against H3-K27me3, H3-K27me2 or H3-K27me1, and also with antibodies against unmodified H3. Unmodified H3 X-ChIP served as a critical control for the ability to detect nucleosomes at the different chromosomal regions and as reference for comparing the levels of H3-K27 methylation in wild-type and in mutant embryos. In addition we also monitored binding of the three PcG protein complexes PhoRC, PRC1 and PRC2 by performing X-ChIP reactions with antibodies against the PhoRC component Pho, the PRC1 component Ph and the PRC2 component Su(z)12. Real-time quantitative PCR was used to measure the abundance of specific genomic DNA sequences in the immunoprecipitates. At each PcG target gene, we monitored presence of H3, presence of the different methylated states of H3-K27 and binding of the three PcG proteins at (i) the PREs that we had identified by Pho X-ChIP-on-chip in embryos (K Oktaba and J Müller, unpublished) and (ii) at one or more other regions within the transcribed portion of the gene (). Two sequences located in two distinct intergenic regions elsewhere in euchromatin served as controls ().
Comparison of the H3-K27me3, H3-K27me2 and H3-K27me1 profiles between wild-type and mutant embryos revealed that the levels of these three modifications are very differently affected in the absence of Pcl. In particular, mutants show a 2- to 3-fold reduction of H3-K27me3 signals in most regions of the nine target genes and this reduction is accompanied by a 2- to 4-fold increase of H3-K27me1 and H3-K27me2 signals at these regions (). Several specific aspects of these observations should be noted. First, trimethylation of H3-K27 in target gene chromatin is only reduced but not abolished in mutants, and although the reduction is seven-fold in the region of the in or in the coding region of , H3-K27me3 levels are not detectably reduced at the gene (). Together with the observation that global H3-K27me3 levels are not detectably reduced in Pcl mutant embryos (), but that they are clearly reduced in mutant clones in imaginal discs (), this suggests that Pcl-PRC2 is required to generate high levels of H3-K27 trimethylation at many but not all PcG target genes in embryos, but probably becomes required for this methylation at most target genes during larval development. Second, in wild-type animals, H3-K27me1 and H3-K27me2 X-ChIP signals at target genes are lower than in the control regions 1 and 2 but they reach comparable levels in mutant embryos (). This suggests that in wild-type animals, nucleosomes in PcG target gene chromatin are extensively trimethylated at H3-K27 and that in the absence of Pcl, these nucleosomes are mono- and dimethylated similar to the rest of the genome. Third, in Pcl mutants, H3-K27me1 and H3-K27me2 X-ChIP signals are also two-fold increased at control regions 1 and 2 (). One possible explanation for this effect could be that in the absence of Pcl, a larger proportion of ‘free' PRC2 becomes available for the untargeted genome-wide mono- and dimethylation of H3-K27.
One possible explanation for the reduced H3-K27 trimethylation at PcG target genes in mutants could be that Pcl is required for anchoring of PRC2 at PREs. In X-ChIP assays with antibodies against E(z), we were unable to detect E(z) at PREs in wild-type embryos, but antibodies against Su(z)12 revealed enrichment of Su(z)12 at the PREs of all nine genes in wild-type embryos (). We note that the Su(z)12 X-ChIP signals at HOX gene PREs in embryos are substantially lower than in imaginal discs (), and, at all PREs, the signals were lower than the Pho or Ph X-ChIP signals in embryos (). Nevertheless, we found that Su(z)12 X-ChIP signals at most PREs are slightly reduced in mutant embryos compared with wild-type embryos even though at most PREs the reduction was within the experimental error and binding of Su(z)12 was still higher than in the control regions 1 and 2 (). It thus appears that PRC2 binding at PREs is reduced but not abolished in the absence of Pcl. In contrast, Pho and Ph X-ChIP signals are indistinguishable in mutant and wild-type embryos (). Binding of PhoRC and PRC1 to PREs thus seems to be unaffected in the absence of Pcl protein. This observation is consistent with earlier findings that suggest that Pho directly or indirectly targets PRC1 to PREs, independently of H3-K27 trimethylation (, ; ).
Finally, we asked how the lack of Pcl protein and the concomitant reduction of H3-K27 trimethylation affect repression of these target genes. To this end we compared the expression patterns of , , , , and in wild-type and in mutant embryos by staining embryos with antibodies against their protein products. mutant embryos show widespread misexpression of and , whereas is only misexpressed in a few rare cells, and we have been unable to detect misexpression of , or in these mutant embryos (). Similar misexpression phenotypes were observed in embryos homozygous for other PcG mutations: only and show widespread misexpression, and are only misexpressed in a few rare cells, and and show no detectable misexpression (; ; ; J Müller, unpublished observations). Intriguingly, this was also true for embryos that were homozygous for the temperature-sensitive allele and had been reared at the restrictive temperature and therefore lacked detectable levels of K27 di- and trimethylation (data not shown; ; ). Taken together, these observations suggest the following: first, in Pcl mutants, H3-K27 trimethylation levels are evidently also reduced in the chromatin of target genes that are not becoming widely misexpressed in the embryo and it also occurs in upstream control regions. This suggests that the reduction of H3-K27me3 levels is probably directly due to the lack of Pcl and/or the reduction of PRC2 binding at target genes, and is not a secondary consequence of these target genes becoming transcriptionally active. Second, in Pcl and other PcG mutants, not all target genes show the widespread misexpression that is observed in the case of HOX genes. This suggests that removal of PcG function alone does not result in transcriptional activation of these target genes, but that specific transcriptional activators are needed in order for these genes to become activated in cells outside of their normal expression domains. Consistent with this, we find that even though is only subtly misexpressed in mutant embryos, during larval stages, shows widespread misexpression in mutant clones in imaginal discs (). Similarly, we find that , and become misexpressed in PcG mutant clones in imaginal discs (, and data not shown). This suggests that although high levels of H3-K27 trimethylation may not yet be critical during embryonic development, they appear to become critical for repression of these target genes during larval development.
In this study, we show that Pcl-PRC2 is a distinct form of the PRC2 HMTase, with a critical role in H3-K27 trimethylation at PcG target genes. In the following sections we shall discuss in turn the main conclusions that can be drawn from the data reported here.
Biochemically purified Pcl complexes contain Pcl together with the four core subunits of PRC2. In contrast, our biochemically purified E(z) complexes contain only substoichiometric amounts of Pcl and the previously described purifications of PRC2 failed to reveal Pcl in the purified material (; ; ; ). Moreover, fractionation of crude nuclear extracts by gel filtration indicated that Pcl and PRC2 components Esc, E(z) and Su(z)12 co-fractionate in high-molecular-weight assemblies, but that the bulk of these other PRC2 components is present in lower-molecular-weight fractions that do not contain Pcl (; ). Taken together, these observations suggest that only a fraction of PRC2 is associated with Pcl and that Pcl-PRC2 is a distinct complex.
Previous X-ChIP studies showed that Pcl and Su(z)12 colocalize at and PREs (). This suggests that Pcl-PRC2 is bound at these PREs. Here, the analysis of mutants that lack Pcl protein and therefore lack Pcl-PRC2, allowed us to obtain insight into the function of this complex. Our results provide strong evidence that Pcl-PRC2 is needed to generate high levels of H3-K27 trimethylation in the chromatin of PcG target genes. Unlike in or mutants, removal of Pcl in embryos or in imaginal discs only reduces but does not eliminate H3-K27 trimethylation. Nevertheless, repression of several PcG target genes is abolished in mutants. This suggests that not only the mere presence of H3-K27me3, but presence of high levels of H3-K27me3 is crucial for maintaining these PcG target genes in the repressed state. Previous studies on the gene suggested that presence of H3-K27 trimethylation in the promoter and coding region is critical for PcG repression (). One possibility would be that it is the overall density of H3-K27me3-marked nucleosomes across the promoter and coding region that determines whether a PcG target gene is repressed. Another possibility would be that even though a whole chromatin domain becomes trimethylated at H3-K27, only a few H3-K27me3-marked nucleosomes at a particular position (e.g., around the transcription start site) are actually required for repression, and failure to maintain this trimethylation results in loss of repression.
The observation that Su(z)12 binding and H3-K27 trimethylation are reduced but not lost in the absence of Pcl is consistent with the idea that Pcl might help anchoring PRC2 to PREs, but it also suggests that at least some PRC2 must be targeted to PREs independently of Pcl. It seems likely that the residual H3-K27 trimethylation present in mutant embryos and in mutant clones in imaginal discs is generated by PRC2 that is bound at PREs independently of Pcl. In this context it is important to note that we found that not only Pcl-PRC2 but also PRC2 is able to trimethylate H3-K27 in recombinant nucleosomes . Apart from the suggested role in tethering of PRC2 to PREs, it is possible that Pcl also functions in a post-recruitment step to help PRC2 generate high levels of H3-K27 trimethylation at target genes. For example, the tudor domain and PHD fingers of PRE-bound Pcl might interact with modified nucleosomes in the promoter and coding region of target genes to ensure that they become trimethylated at H3-K27 by the associated PRE-tethered PRC2.
Finally, we found no evidence that Pcl-PRC2 would be required for the genome-wide H3-K27 mono- and dimethylation. Our X-ChIP analyses suggest that H3-K27 mono- and dimethylation across the genome might even slightly increase in the absence of Pcl (). In contrast, there is a loss of all H3-K27 methylation in either or mutants (; ). This suggests that PRC2 or another E(z)-containing complex generates the genome-wide H3-K27 mono- and dimethylation. The experiments in mutants thus allowed us to dissect the role of different H3-K27 methylation states in . The selective reduction of H3-K27me3 levels, and the concomitant loss of repression of PcG target genes in mutants, provides compelling evidence that only the trimethylated state of H3-K27 is functional in PcG repression in . Pcl-PRC2 is evidently critically needed to generate the high levels of H3-K27 trimethylation that are required to maintain a Polycomb-repressed chromatin state.
The a-tubulin-TAP-Pcl and a-tubulin-TAP-E(z) transgenes in the transformation vector CaSpeR have the following structure: a 2.6-kb fragment of the a-tubulin 1 gene, containing promoter and 5′ untranslated region sequences () linked to the N-terminal TAP tag (), followed by or cDNA fragments that contain the ORF of Pcl (Pcl) or E(z) (E(z)), respectively (plasmid maps are available on request). Rescue function of the a-tubulin-TAP-Pcl and a-tubulin-TAP-E(z) transgenes was tested by introducing the transgene into a or mutant background, respectively. Specifically, in the case of , we recombined a copy of the a transgene onto an chromosome and found that aa animals are wild type in appearance, viable and fertile. Similarly, we found that the a transgene rescues animals that are trans-heterozygous for two protein-negative -null alleles (); a animals are wild type in appearance, viable and fertile. TAP was performed from embryonic nuclear extracts as previously described (); see for detailed information.
A detailed list of peptide sequences obtained from mass spectrometry analysis of the protein bands shown in is available in . For quantitative mass spectrometry of modified peptides, the samples were separated on a nano-flow 1D-plus Eksigent (Eksigent, Dublin, CA) HPLC system coupled to a qStar Pulsar quadrupole time-of-flight MS (Applied Biosystems, Darmstadt, Germany). The resulting MASCOT search result file and the generic mass spectrometry data of each sample were parsed using MSQuant in a no-label setting (). The quantitation result (peptide ion volumes in thompson·sec) for each peptide was presented in a diagram (). The presented data was an average of a duplicate analysis. To minimize differences between the samples, the ion volume of all modified and unmodified peptides was summed and normalized ().
Staining of imaginal discs and embryos was performed as described ().
X-ChIP on wt and embryos was performed as described in . embryos were collected from an strain by selecting for GFP-negative embryos using an embryo sorter (COPAS™ SELECT, Union Biometrica). Primers used for amplification are listed in .
All antibodies used in this study are listed in .
Recombinant PRC2 complex was expressed and purified as described in .
Assays were performed as described ().
For Western blot experiments shown in , embryos were dechorionated and taken up in ice-cold PBS buffer containing 0.01% Triton. Embryos were homogenized with a glass dounce homogenizer, lysate was cleared at 400 for to pellet the embryonic carcasses. Supernatant was centrifuged at 1100 and the pellet was resuspended in SDS–Laemmli buffer. For Western blots on larval extracts, imaginal disc and CNS tissues were dissected from wild-type or mutant larvae and resuspended in SDS–Laemmli buffer.
The following strains were used in this study:
was isolated in a screen for new PcG mutation; details for the screen will be described elsewhere (). In brief, was induced with EMS on an chromosome; complementation tests with revealed that is an allele of . Sequence analysis revealed a single nucleotide exchange in , changing the codon E701 from AG into a premature stop codon (AG). |
The T cell antigen receptor (TCR) complex comprises variable α/β subunits that recognise peptide/major histocompatibility (MHC) complexes and invariant signal transduction subunits of the CD3 antigen. A key stage in T cell development in the thymus is the selection of cells that have successfully rearranged their TCR-β locus. This occurs in T cell precursors, which do not express of the MHC receptors, CD4 and CD8 (double negative (DN) thymocytes) . The DN stages of intrathymic differentiation can be followed by the sequential pattern of expression of CD44 and CD25. T cell progenitors enter the thymus as CD44CD25 cells (termed DN1); these then express CD25 (DN2) and begin to rearrange T-cell receptor β loci. Cells then lose CD44 expression and continue β chain rearrangements to completion (DN3). Cells that successfully rearrange their TCR-β locus will express a functional receptor complex known as the pre-TCR comprising a TCR-β chain, the p-Tα subunit and the signalling subunits of the CD3 antigen . When the pre-TCR is expressed at the cell membrane it promotes cell survival and entry into the cell cycle ; cells downregulate CD25 and transit to the DN4 pre-T cell subset. DN4 T cells undergo proliferative expansion and differentiate to CD4CD8 double positive (DP) cells. These cells then undergo TCR α-chain gene rearrangements and upon expression of a functional α/β TCR complex are subjected to positive and negative selection to generate CD4 or CD8 single positive (SP) thymocytes .
Normal pre-T cell development requires the coordination of a complex program of gene transcription by signal transduction networks mediated by the pre-TCR complex and cytokine and stromal signals . Mice deficient for the recombinase activating (RAG) genes which are unable to rearrange TCR-β subunits and express a pre-TCR are unable to progress beyond the DN3 stage of thymocyte development . Similarly, the absence of pre-Tα or CD3 subunits blocks thymocyte development at the DN3 stage . The transition of thymocytes beyond the pre-T cell stage of thymocyte differentiation is also dependent on signal transduction networks mediated by tyrosine and serine kinases . In this context, crucial responses are mediated by Rho family guanine nucleotide binding proteins such as RhoA, Rac-1 and CDC42 .
The importance of RhoA for thymocyte development has been demonstrated by studies of transgenic mice that express C3-transferase under the control of T cell specific promoters such as the p56lck and CD2 promoters . This toxin selectively ADP-ribosylates RhoA within its effector-binding domain and abolishes its biological function. Transgenic mice that express C3-transferase under the control of the p56 promoter have a small thymus and severely reduced numbers of peripheral T cells . This phenotype is caused by survival defects in DN2 and DN3 thymocytes that lack Rho function . During embryogenesis no cells progress beyond the DN1 stage but in adult mice a few T cell progenitors survive and develop to DN4s and beyond . This ‘leakiness’ either reflects selection of cells that compensate for loss of Rho function in DN2/3 thymocytes or reflects heterogeneous and asynchronous expression of the lck promoter in adult DN thymocytes. The few DP thymocytes and mature T cells found in adult Lck-C3 transgenic mice have numerous defects including reduced survival, proliferation, integrin mediated cell adhesion and defective cell motility . A complementary strategy to probe Rho function in the thymus used the CD2 locus control region (LCR) to target C3 transferase to T cell progenitors (CD2–C3 mice) . The inhibition of Rho function at the DN2/3 stage in CD2–C3 mice is not leaky and causes T cells to become blocked in differentiation at the DN3 stage of thymus development. CD2–C3 mice thus have a thymic phenotype indistinguishable from the phenotype of Recombinase gene null mice or mice lacking key structural or signaling components of preTCR complex : thymocyte development is blocked at the pre-T cell/DN3 stage .
The basis for the failure of pre-T cell differentiation in CD2–C3 transferase mice is not known. One way to address this issue is to use microarray gene expression profiling to determine the impact of losing Rho function on transcriptional responses in pre-T cells. Previous studies in transformed cell lines have identified a role for RhoA in regulating activating protein-1 (AP1) family of transcription factors but transcriptional targets for RhoA signal transduction in primary non-transformed cells have not been explored. The present data show that RhoA regulates expression of genes encoding members of the Fos/Jun and early growth response (Egr) family of transcription factors but also has an impact on expression of genes regulating diverse biological functions including serine/threonine kinases, protein phosphatases, enzymes that regulate protein biosynthesis and proteins that regulate nuclear structure and function.
Mice were bred and maintained under specific pathogen-free conditions in the transgenic animal unit. RAG2 and C3 transgenic mice, which selectively express the bacterial toxin C3-transferase under the control of CD2 promoter and locus control region in the thymus, have been described in detail elsewhere .
Fluorescein isothiocyanate (FITC), phycoerythrin (PE), allophycocyanin (APC), and biotin-conjugated antibodies were obtained from Pharmingen (San Diego, CA). Tricolour (PE/Cy5 and APC/Alexa750) conjugated antibodies and fluorophore-streptavidin conjugates were from Caltag (Burlingame, CA). Thymocytes were stained for cell surface markers and analysed on a FACS Calibur (Becton Dickinson Immunocytometry Systems Franklin Lakes CA). Data were analysed using CellQuest BDIS or FlowJo (Treestar Inc, Ashland OR) software. CD4CD8 double negative thymocyte subsets were analysed for CD44 and CD25 expression following lineage exclusion of mature DP and SP cells as well as non-T cell lineage cells using a cocktail of biotinylated antibodies (CD4, CD8, CD3 B220, Mac-1, NK, Gr-1, and γδ) revealed with streptavidin-tricolour and costained with CD25-FITC, CD44-PE and Thy1.2-APC. Intracellular phospho-S6 and TCR-β intracellular staining was carried out on thymocytes pre stained to identify DN3 and DN4 subpopulations. Cells were subsequently fixed in 1% paraformaldehyde for 10 min at room temperature, washed in PBS and permeabilized with saponin buffer (0.5% saponin, 5% FBS, 10 mM HEPES [pH 7.4] in PBS) for 10 min at room temperature. Permeabilized cells were incubated with PE-conjugated TCRβ antibody and phospho S6 antibody for 45 min at room temperature, washed in saponin buffer and subsequently stained with FITC conjugated donkey anti rabbit IgG (Jackson Immunoresearch, West Grove PA). Cells were analysed on an LSR Flow Cytometer (BDIS).
Thymocytes were isolated from 4- to 6-week-old RAG2, CD2–C3 and C57BL/6 control mice. DN3 pre-T cells were purified firstly by removing CD4, CD8 DP cells using MACS CD4+ and -MACS CD8+ T cell isolation kits (Miltenyi Biotec, Ltd.) followed by magnetic autoMACS separation. CD4/CD8 depleted thymocytes were then labelled with Thy1.2-APC, PE conjugated CD4, CD8, CD44 and TCRγδ antibodies, FITC conjugated CD25 antibodies and the DN3 subpopulation (Thy1CD4/CD8/CD44/CD25) was purified by cell sorting on a FACS Vantage (BDIS) using Cell Quest software (BDIS). All sorted cells were analysed by flow cytometry and only used for experiments if the purity was 95–98%. Total RNA was extracted from sorted cells using the Qiagen RNeasy kit according to manufacturers instructions. Total RNA was quantified using the RNA 6000 Nano LabChip kit and analyzed on an Agilent 2100 Bioanalyzer (Agilent Technologies). Total RNA from up to 10 thymi were pooled for microarray analysis and quantitative PCR.
Microarray analysis was carried out by the Finnish DNA Microarray Centre at the Centre for Biotechnology, Turku, Finland. For Affymetrix sample preparation, 100–1000 ng of total RNA was used as the starting material to synthesise target cRNA using the GeneChip Eukaryotic Small Sample Target Labeling Assay Version II according to manufacturers’ instructions (Affymetrix, Santa Clara, CA). The cRNA target sample was hybridised to the Affymetrix Mouse Genome 430 2.0 Array (45,101 probes sets and expressed sequence tag (EST) clusters). Data was analysed with Affymetrix Gene Chip Operating Software (GCOS) version 1.1 and filtered according to recommendations of the manufacturer. Comparison analysis was used to compare the expression profiles from two arrays – results were obtained by analyzing the experimental array in comparison to the baseline array. The global method of scaling was used for normalizing hybridization intensity between arrays. The statistical algorithm used by GCOS defined the expression status of each gene-specific probe set as not changed, increased, marginally increased, decreased or marginally decreased. Genes that were absent or unchanged between comparisons were excluded from the results. Comparison analysis defines a gene as up-regulated if the signal log ratio between the baseline sample and the experimental sample is larger than 1 (2-fold) and the experimental sample is present. Similarly changed genes were further analyzed/classified using the gene ontology tool () and represented by biological function. Gene lists were complied and sorted by genes that were either commonly or uniquely regulated in CD2-C3 and RAG2 thymocytes.
Purified total RNA (200 ng) was reverse transcribed using the iScript™ cDNA synthesis kit (BioRad). Real time RT-PCR was performed in a 96-well plate using iQ™ SYBR Green based detection (BioRad) on a BioRad iCycler in 20 μl reaction volume containing 1 μl cDNA (20 ng), 0.8 μl 10μM sense and antisense primers, 10 μl iQ™ SYBR Green supermix, and 4 μl nuclease free water. Each reaction was performed in duplicate and each experiment repeated in triplicate. 18S rRNA levels were used to normalise RNA concentrations between samples and the relative mRNA levels were calculated using the equation:where is the efficiency of PCR, ct is the threshold cycle, u is the mRNA of interest, r is the reference gene (18S rRNA), s is the sample and c is the control sample. Primers used for RT-PCR were designed using Beacon Designer 2 software. Primer sequences are: Egr1 forward 5′-ACAGAAGGACAAGAAAGCAGAC-3, reverse 5-CCAGGAGAGGAGTAGGAAGTG-3′, Fos forward 5′-CTACTGTGTTCCTGGCAATAGC-3′, reverse 5′-AACATTGACGCTGAAGGACTAC-3′, Egr3 forward 5′-TGACCAACGAGAAGCCCAATC-3′, reverse 5′-GCTAATGATGTTGTCCTGGCAC-3′, Jun forward 5′-CGCCTCGTTCCTCCAGTC-3′ reverse 5′-ACGTGAGAAGGTCCGAGTTC-3′, JunB forward 5′-CTTCTACGACGATGCCCTCAAC-3′, reverse 5′-GTTCAAGGTCATGCTCTGTTTTAGG-3′, Nur77 forward 5′-CCTGTTGCTAGAGTCTGCCTTC-3′, reverse 5′-CAATCCAACACCAAAGCCACG-3′, 18S RNA forward 5′-GTAACCCGTTGAACCCCATT-3′, reverse 5′-CCATCCAATCGTAGTA-3′.
Mice expressing the RhoA inhibitor Clostridium Botulinum C3 transferase under the control of the CD2 locus control region (LCR) have been described previously . Loss of RhoA function in CD2-C3 mice results in loss of CD4/CD8 double positive thymocytes (A) with a block in thymocyte development at the DN3 stage (B). Rag2 mice also block thymus development at the DN3 stage (A and B) due to failed expression of the pre-TCR complex. Despite this superficial similarity, loss of Rho function does not prevent TCR beta locus rearrangements although there is a reduced frequency of DN3 cells expressing intracellular TCR-β subunits (C). Ectopic expression of transgenic TCR complexes cannot reverse the developmental block in CD2–C3 thymocytes demonstrating that failed expression of TCR complexes does not explain the developmental defects caused by loss of Rho function . Rho thus appears necessary for pre-TCR function although these results do not discriminate between a direct role for RhoA in pre-TCR signalling or a role for this GTPase in the cytokine/stromal signalling pathways that synergise with pre-TCR signals to control pre-T cell differentiation .
CD2–C3 transferase mice provide a good model system to probe the immediate transcriptional consequences of losing RhoA function in pre-T cells since the CD2-LCR initiates expression of transgenes in T cell progenitors as they transit from the DN2 to DN3 stage of thymocyte development . Accordingly, it is possible to isolate DN3 thymocytes ex vivo that have only just switched on expression of C3 transferase and prepare RNA for DNA microarray analysis. The Affymetrix Mouse Genome 430 2.0 array, representing 39 000 murine gene transcripts, was used to transcriptionally profile DN3 pre-T cells purified from the thymi of CD3–C3 transgenic mice. In these experiments DN3 thymocytes from wild type control mice were transcriptionally profiled and used for comparisons.
The microarray data were analyzed with the Affymetrix Gene Chip Operating Software (GCOS, version 1.1) by comparing the gene expression profiles of CD2–C3 pre-T cells to the profile of normal cells to reveal changes above or below wild type levels. DN3 pre-T cells expressed approximately 18 000 genes and there were only small changes in the transcriptional profile of CD2–C3 DN3 thymocytes compared to control wild type DN3s. We focused our analysis on ⩾2-fold changes, which were further analyzed using the Affymetrix gene ontology tool. A full list of the genes regulated 2-fold or greater by loss of RhoA function in DN3 thymocytes is shown in . There were approximately 18 000 genes expressed in wild type DN3 thymocytes: loss of RhoA function caused a ⩾2-fold decrease in expression of 383 genes and a ⩾2-fold increase in expression of 190 genes. RhoA regulated genes in DN3 thymocytes encode proteins with diverse biological functions including serine/threonine kinases, protein phosphatases, enzymes that regulate protein biosynthesis.
This point is illustrated in A, which shows the genes whose expression is regulated more than 10-fold. The largest decrease was in expression of the gene encoding FosL2, a Fos family transcription factor (630-fold decrease) but other large changes included a subunit of eukaryotic translation initiation factor 2, Protein Kinase C beta and the membrane receptor CD69. There was also a loss of expression of genes encoding proteins that regulate nuclear structure and function such as genes encoding histones that are involved in nucleosome assembly. In a similar category there was reduced expression of mRNA encoding special AT-rich sequence binding protein 1 (SATB1) in CD2-C3 DN3s (). SATB1 acts as a scaffold for chromatin modifiers and is known to regulate higher order chromatin structure and to regulate gene expression by acting as a “docking site” for several chromatin remodeling enzymes .
Of 55 genes whose expression was decreased ⩾5-fold, 23 encoded transcription factors; of 384 genes decreased ⩾2-fold, 63 were transcription factors (B). Transcription factors downregulated in CD2-C3 DN3 thymocytes included Kruppel family transcription factors KLF4, KLF6 and KLF9. There was also loss of expression of members of the immediate early gene family Egr1 and Egr3, AP-1 family members Fos, FosL2, FosB, Jun, JunB and downregulated expression of genes encoding Ets1 and Nurr77. To validate the microarray analysis, quantitative real-time PCR was used to compare expression of a selection of these different transcription factors in wild type versus CD2-C3 thymocytes (C). The experiments confirmed the reduction in Fos, Jun, JunB, Egr1 and Egr3 and NURR77 in 3 DN3s compared to wild type controls.
A number of gene targets for Egr transcription factors have been described and in this context, RhoA inactivation caused loss of expression of multiple members of the MAP kinase phosphatase family DUSP 1, 2, 4, 6, 16. The latter are induced in negative feedback mechanisms that modulate MAP kinase activity and have been described as targets for Egr family transcription factors . The loss of DUSP expression is thus almost certainly a secondary consequence of failed expression of Egr molecules. Egr1 and Egr3 are well characterised pre-TCR induced genes . To verify this we compared the transcriptional profile of CD2-C3 DN3s with Recombinase gene 2 (RAG2) DN3s which lack expression of the pre-TCR complex because they fail to rearrange their TCR beta locus. Of the 23 transcription factors downregulated ⩾5-fold in CD2–C3 thymocytes 18 were also downregulated in Rag2 DN3s (D) and these included Egr1, Egr3, Fos, Fosl2, FosB, Ets1, Nurr77. These microarray analyses were confirmed by quantitative real-time PCR analysis of wild type versus Rag2 thymocytes (data not shown). The common loss of expression of members of the Egr family in CD2–C3 and Rag2 pre-T cells is thus consistent with a RhoA requirement for the pre-TCR induced gene transcription. However, comparisons of the total number of genes regulated uniquely in CD2–C3 and Rag2 DN3s () (VENN diagrams) indicate that there are a number of gene changes in CD2–C3 cells that are not seen in Rag2 DN3s. The full list of genes that were commonly and uniquely regulated 2-fold and greater in CD2 C3 versus Rag2 pre-T cells is shown in . In terms of ⩾ 2-fold, gene repression, CD2-C3 and Rag2 DN3s had 180 gene targets in common with 202 genes being downregulated uniquely in CD2–C3 pre-T cells and 120 genes downregulated uniquely in Rag2 cells. In terms of ⩾2-fold gene increases, CD2–C3 and Rag2 DN3s had 65 gene targets in common, 129 genes that were increased uniquely in CD2-C3 DN3s and 182 increased uniquely in Rag2 cells ().
The common transcriptional changes caused by loss of the pre-TCR in Rag2 DN3s or loss of Rho function in CD2–C3 mice indicates that Rho function is necessary for some pre-TCR induced transcriptional responses. However, there are genes downregulated in DN3s lacking RhoA function that are unchanged in Rag2 DN3s () indicating that RhoA is not just required to mediate the preTCR induced gene program but may also participate in cytokine/stromal cell-initiated signalling pathways that control pre-T cell differentiation. The genes uniquely lost in CD3–C3 DN3 thymocytes include serine kinases (Protein Kinase C beta and Map3k14); transcription factors (SpiB and Dlx1) and regulators of protein biosynthesis (Eif2s3y) (). One gene of interest lost in CD2–C3 DN3s but not Rag2 null DN3s was Hes-1 which is induced by Notch receptor-ligand interactions. Notch signalling is required throughout DN to DP stage of thymocyte development to support T cell metabolism and survival . A RhoA requirement for Notch signalling has not been described but it is known that RhoA is necessary for integrin mediated cell adhesion and thymocyte cell migration . Accordingly, decreased Hes-1 expression could reflect that pre-T cells lacking Rho function cannot make normal contacts with stromal cells that express the Notch ligands .
There were also unique transcriptional changes seen in Rag2 cells that were not found in the CD2–C3 DN3s () which indicates that the pre-TCR can regulate a program of gene transcription via RhoA independent pathways The unique changes in Rag2 cells not surprisingly included loss of expression of rearranged TCR-β subunits. One other gene downregulated in Rag2 DN3s but not CD2–C3 DN3s was , which encodes the ribosomal S6 subunit. In this respect, we recently described high basal levels of S6 phosphorylation in ex vivo β selected DN3s and no detectable S6 phosphorylation in ex vivo Rag2 DN3s . The data in quantify S6 phosphorylation in CD2–C3 DN3s compared to wild type cells. Normal DN3s are heterogeneous for phosphoS6 with the majority of cells being phosphoS6 but a significant percentage of cells are phosphoS6. The DN3 thymocyte subpopulation can be subdivided into cells that have not yet completed TCR-β locus rearrangements and those that express a functional TCR-β subunit that allows surface expression and signalling of the pre-TCR complex. Analysis of TCRβ expression by intracellular staining revealed that DN3s that express icTCR-β chains are generally phosphoS6 whereas DN3s that are icTCR-β null are uniformly phosphoS6. In CD2–C3 DN3s there were also high levels of phosphoS6 in TCR-β high cells. The presence of normal S6 phosphorylation in β selected DN3s lacking Rho function is a further indication that Rho is necessary for a subset of responses in DN3 thymocytes but does not globally block signalling.
One important question is whether the transcriptional changes in CD2–C3 DN3s explain why loss of RhoA function results in failed thymocyte differentiation? In this respect, microarray analysis of Rag2 pre-T cells revealed decreased expression of TCR-β subunits as a major defect () and failed expression of the TCR-β subunit is indeed responsible for the DN3 developmental block in Rag2 mice. In CD2–C3 DN3s there were several gene defects that would explain why RhoA function is essential for pre-T cell development. For example, RhoA is required for expression of Egr3 and Ets1 and loss of either of these transcription factors inhibits pre-T cell proliferation and is suboptimal for transition through early pre-TCR-dependent stages of thymocyte development . Loss of individual Egr family or AP-1 transcription factors does not completely abrogate thymocyte differentiation but this reflects that there is considerable redundancy in AP-1 complexes due to the ability of different family members to pair as dimers . The simultaneous elimination of Egr1, Egr3, Ets1 and multiple Fos/Jun family members in CD2–C3 DN3s would circumvent the possibility of redundancy . This comprehensive loss of AP-1 activity in pre-T cell lacking RhoA function plus the decreased expression of other transcription factors would rapidly cause global changes in the transcriptional program of a cell. Moreover, the transcription factor defects would be exacerbated by the decreased expression of chromatin modifying enzymes such as SATB1.
In summary, loss of Rho function or failed expression of the pre-TCR complex in Rag2 null mice block T cell development at a common stage. The present data show that the genetic consequences of loss of Rho function versus loss of pre-TCR expression for T cell progenitors are not identical. The present results also show that loss of Rho function in pre-T cells results in downregulation of genes encoding members of the Fos/Jun and Early growth response (Egr) family of transcription factors. The collective loss of these transcription factors and the resultant secondary genetic changes explains why loss of RhoA function prevents pre-T cell development. |
Genetically programming cells require sensors to receive information, circuits to process the inputs, and actuators to link the circuit output to a cellular response (; ; ; ). In this paradigm, sensing, signal integration, and actuation are encoded by distinct ‘devices' comprised of genes and regulatory elements (; ). These devices communicate with one another through changes in gene expression and activity. For example, when a sensor is stimulated, this may lead to the activation of a promoter, which then acts as the input to a circuit. There has been a large effort to create and characterize different classes of devices and to make this information publicly available in the Registry of Standard Biological Parts (parts.mit.edu).
We have constructed a device that functions as an AND gate that can integrate two input signals and control a cellular response. An AND gate is a logical operation that integrates multiple input signals. The output of an AND gate is only ON when all of the inputs are ON. If any of the outputs are OFF, then the output is OFF. An AND gate forms the core of electronic computing and is a critical device to create different genetic programs. It is particularly useful to integrate signals from multiple sensors to identify an environment with high specificity.
Bacteria use a variety of mechanisms to sense their environment, including two-component systems, transcription factors, and small RNA molecules (). In some cases, an environment is defined by a single signal, such as the presence or absence of a small molecule (e.g., the , , and operons; ). In other cases, it is a complex array of signals that are integrated by the bacterium to identify an environment. Integrating multiple signals can increase the sensing specificity. Even signals that are too general to identify a specific environment (e.g., pH, temperature, and osmolarity) can achieve higher specificity together. A similar problem arises when programming cells to identify an environment that is not naturally encountered, and for which there is not a single dominant signal. In this case, genetic logic gates are required to integrate multiple signals to achieve sensing specificity.
Different logic gates have been built using biological components such as transcription factor genes and regulatory elements (; ; ) or protein–protein interactions (). To date, the architecture of these circuits relies on the specific identity of a particular set of inputs and a particular output. In these examples, changing the identities of the inputs and outputs is not simple. In contrast, a circuit is modular if it can be rapidly connected to different inputs and used to drive different outputs. Modularity facilitates the incorporation of a circuit into different genetic programs.
We have designed and constructed a modular genetic AND gate whose inputs and outputs are promoters (). Only when both of the input promoters are active is the output promoter turned on. The architecture of the AND gate involves two parts, both of which are required to express T7 RNA polymerase. The first part is the T7 RNA polymerase gene, which has been modified to contain two amber stop codons that block translation. The second part is the nonsense suppressor tRNA , which enables the translation of polymerase. When both of these parts are transcribed from the input promoters, polymerase is expressed and this activates an output T7 promoter.
Using two inducible systems as inputs (promoters that respond to arabinose and salicylate) and connecting the output to the expression of green fluorescent protein (gfp), we demonstrate that this circuit behaves as a near-digital AND gate. These data are used to parameterize a simple model of the steady-state input–output response (transfer function) of the circuit. This formalization will facilitate the integration of this circuit into larger genetic systems. To demonstrate the circuit is modular, two constructs are made that switch the input promoters and connect the output to a different response. First, new inputs are added that respond to the quorum signal AI-1 () and magnesium limitation (). Second, the output is switched to the invasin gene, which enables the bacteria to invade mammalian cells. In both cases, the circuit behaves as an AND gate.
A two-input AND gate activates an output only when both inputs are on. If either or both of the inputs are off, then the output is off. Ideally, the circuit should be designed to be modular, such that the inputs and outputs can be rapidly rewired. In transcription-based systems, it is convenient that the connections between devices be promoters (, ; ). For example, if a two-component system turns a promoter on, then this promoter can be used as the input into the next device. Similarly, if the output is a promoter, then this can either express a gene that produces a cellular phenotype or act as an input into the next device.
Our AND gate design uses two promoters as inputs and turns on an output promoter. Transcription occurs from the output promoter only when both input promoters are active. The circuit integrates the inputs via a translational interaction (). The first input promoter drives the transcription of mRNA encoding T7 RNA polymerase that by itself cannot be translated due to the presence of amber stop codons. The second input promoter drives the transcription of an amber suppressor, which allows the activator to be expressed and activate an output promoter.
The suppression is based on the TAG amber stop codon and the SupD amber suppressor tRNA derived from tRNA (). In wild-type bacteria, the TAG codon is decoded by release factor 1 resulting in translation termination. In the presence of SupD, TAG codons are decoded as serine, and translation resumes to generate the full-length protein. Because there are only 326 TAG codons in (), suppressed expression can be as efficient as 50% with no loss of viability (). We verified that the T7 RNA polymerase and parts do not affect the growth rate or morphology when expressed in ().
T7 RNA polymerase was chosen to be the activator in the circuit, although in principle any transcriptional regulator could be used in this design. The T7 gene was modified to contain amber codons at positions 8 and 14 (). This causes premature translational termination, resulting in a non-functional polypeptide. This combination of mutations afforded the lowest basal T7 RNA polymerase activity and the highest gain in activity when coexpressed with an amber suppressor (). When both the polymerase and genes are expressed, full-length polymerase is synthesized, and the output T7 promoter becomes activated.
To characterize the circuit dynamics, two promoters that can be induced with small molecules were used as inputs (, plasmid details in ). The gene was placed under the control of a salicylate-activated promoter () (input 1). The gene was placed under the control of an arabinose-inducible promoter () (input 2). To monitor activation, a fast folding green fluorescent protein containing a degradation tag (), was placed under the control of the T7 promoter.
The first construct contained a strong ribosome binding site (rbs) () and did not function as an AND gate. The circuit was always inducible by arabinose, independent of the concentration of salicylate (not shown). Intuitively, this could result if the basal expression of was high, where a sufficient amount of activator was produced even in the absence of arabinose. In other words, the range of the activity of the input promoter did not match the range required for the proper behavior of the circuit. This problem has been observed before in genetic circuit design (). A successful approach to matching the range of the input to a downstream circuit has been to mutagenize the rbs, either using rational substitutions or random mutagenesis (; ; ).
To tune the range of the input, we designed a saturation mutagenic library of three positions in the rbs and the first base of the start codon (). This library of 128 theoretical variants was plated on media containing arabinose and salicylate, and 50% of the colonies were visibly fluorescent green. Of these, 48 green colonies were subsequently grown in liquid media with no inducer, only salicylate, or both inducers and assayed by fluorimetry. Of the 48 assayed variants 44 showed at least five-fold gain in fluorescence when both inducers were added compared to values obtained when only one or no inducer was added (). Therefore, most variants displayed AND-gate behavior. Two variants, B9 and F11, were chosen for further characterization. The B9 clone has a weaker rbs and behaves as a functional AND gate. The F11 clone has a weaker rbs than the initial sequence, but it produces a similar salicylate-independent response.
The B9 clone was used to further characterize the function of the AND gate circuit. The output of the circuit was measured by growing cells to mid-log phase in different combinations of the two inducers (Materials and methods). The output of the circuit was measured using fluorimetry (). The transitions between the on and off states were very steep, thus producing a near-digital AND gate. Flow cytometry was used to measure the population heterogeneity ( and ). There was no detectable expression in the absence of either inducer and a 1000-fold induction when both inducers are present.
The transfer function of a genetic circuit describes the steady-state response as a function of the activity of the input promoters. For a logic gate, this is a two-dimensional function, where two inputs are being integrated into a single output. An analytical form for the transfer function was derived on the basis of biochemical interaction underlying the circuit architecture and a simple model of translation control (). The model relates the normalized output of the AND gate (/) to the individual transfer functions of the two input promoters, and .
The transfer function was derived in order to understand how the range of the input promoters affects the function of the circuits. To characterize the circuit, two promoters are used that can be induced by small molecules (salicylate and arabinose). However, to generalize the model, and should be the activity of the and promoters and not dependent on the concentrations of the small molecules. The activity of these promoters can be measured independently by fusing to the promoter and measuring the output in response to the small-molecule inducer, thus producing a one-dimensional function (). The individual responses of the two promoters are then used to parameterize the transfer function.
The full derivation of the transfer function is described in the . The form of transfer function relating fluorescence measurements is as follows:
where is the maximum fluorescence observed for the output. Once and are calculated, these parameters could be used in conjunction with any two input promoters, provided that their one-dimensional transfer function was determined under the same standard growth conditions and plasmids.
Equation was parameterized using the full set of experimental data for the B9 clone when both inducers were systematically varied (). To calculate the one-dimensional transfer functions for and , the gene for a fast-degrading green fluorescent protein was fused to the salicylate and arabinose promoters (). For the salicylate promoter, a strong rbs was used as a standard measure of promoter activity (). For the arabinose reporters, the original, B9, and F11 rbss were inserted upstream of GFP. The fusions were cloned into a plasmid and the fluorescence was measured as a function of inducer concentration (). For each pair of salicylate and arabinose concentrations, the one-dimensional fluorescence data were used to obtain and , respectively. The two-dimensional data were used to obtain the value of /. Fitting these data to equation yielded the following: =50±20 and =3000±1000 (; ).
The parameterized transfer function captures the behavior of the circuit when input promoters with different transfer functions are connected to the AND gate (). When the model is extrapolated outside of the B9 data used for parameterization, it can be seen that different ranges of lead to different responses. In particular, a stronger rbs (as for the original and F11 clones) extends the upper limit of the range. This leads to a loss of the AND gate function, as now the circuit is inducible independent of . Note that for any range of , it is possible to generate a functional gate by altering the strength of the rbs of . This suggests that it is better that leaky promoters (such as ) be used as , as this can be compensated for by adjusting the rbs of . Here, we have relied on random mutagenesis and screening to identify functional rbss. Our ultimate goal is to use the transfer function and lists of standardized functional rbss (e.g., Registry of Standard Biology Parts entries BBa_J61100-39) to predict the rbs required to make the circuit functional for two input promoters.
To demonstrate the modularity of the circuit, two constructs containing different inputs and outputs were designed and analyzed. First, the inputs are swapped for two natural promoters (quorum sensing and Mg responsive). Next, the output is swapped from to the invasin gene, which enables the bacteria to invade mammalian cells. Both of these circuits demonstrated AND-gate behavior, thus demonstrating the modularity of the circuit and the capability to integrate natural inputs and control cellular behavior as an output.
To replace the inputs, and were replaced with and , which respond to magnesium limitation and the AI-1 quorum signal, respectively (). Two-component systems are a ubiquitous sensing motif in bacteria, consisting of a membrane-bound sensor and a cytoplasmic response regulator (). When stimulated by an environmental signal, the sensor phosphorylates the response regulator, which can then modulate gene expression. The PhoPQ two-component system responds to the external magnesium concentration. The PhoQ regulator activates the promoter in the absence of exogenous magnesium (; ). Quorum sensing systems are used by bacteria to communicate (). These systems have been used extensively as communication devices in synthetic genetic systems to program cells to form patterns (), regulate their density (), and kill malignant cells in response to bacterial density (). The promoter and gene, derived from the quorum sensing circuit, is induced in response to exogenous -3-oxohexanoyl--homoserine lactone (AI-1) ().
The inputs were connected to the circuit one at a time. First, the promoter was placed in front of the gene with the B9 rbs. The circuit function was then assayed using the salicylate input to drive . This construct did not generate an AND-gate as cells showed no fluorescence when induced with salicylate. Presumably, this is because insufficient transcript was produced from . As before, this problem was overcome by retuning the rbs. A clone was identified with a stronger rbs () that yielded a functional AND-gate. Next, and the gene were used to transcribe the gene. The circuit shows 15-fold fluorescence over background when fully induced but undetectable fluorescence in the presence of Mg or the absence of AI-1 ().
We next examined whether the output of the AND gate could control a cellular behavior. The expression of the invasin gene () of in confers the ability to invade mammalian cells expressing β1-integrin (). We have previously shown that singular environmentally-responsive promoters controlling confer environment-dependent invasion (). Here, the gene is substituted for and the AND gate is tested using the salicylate and Mg inputs.
In a previous study, we created an rbs variant of that conferred arabinose-dependent invasion under a promoter (). This construct was placed under the control of the T7 promoter and combined with the AND gate plasmids that have and as inputs. After being grown in different combinations of the inputs, the bacteria were assayed for invasion of HeLa cells (Materials and methods). Bacteria grown in the presence of Mg or the absence of salicylate show no detectable invasion (). The bacteria only invade when both of the input promoters are on. These experiments demonstrate the ability of a modular AND gate to integrate multiple environmental signals and respond with cellular behavior.
We have constructed a modular AND gate based on the amber suppression of T7 RNA polymerase. Only when two input promoters are active is an output turned on. Because the inputs and outputs of this gate are transcriptional signals, they can be easily replaced. This modularity is demonstrated by swapping the inputs and outputs of the circuit while preserving the AND-gate behavior. In this paradigm, changes in transcription become a common currency allowing the modular integration of individual devices (; ).
The modular nature of this type of transcriptional logic gate will facilitate its use in a variety of engineering applications. In particular, AND logic would be useful in obtaining gene expression in a specific microenvironment. A set of promoters—identified rationally or with a microarray—could be used as inputs to the AND gate. Rather than directly detecting the presence of a new environment with a single engineered promoter or sensing system, independent promoters sense different aspects of the environment. The AND gate activates only when all conditions are present to induce cellular responses. Often microenvironments are defined by multiple nonspecific signals such as oxygen, pH, cell density, lactate, and glucose. It is only when several of these inputs are integrated that specificity can be achieved.
A two-dimensional transfer function of a simple mathematical form captures the input–output behavior of the AND gate. This model could be used to predict the genetic changes required to connect two inputs to the circuit to produce a functional AND gate. To be able to use the model, the inputs have to be characterized using the plasmid and fluorescent reporter system used in this study. In this sense, this work represents a step toward standardization, where circuits are characterized quantitatively to understand their collective function when connected in series. This will be a critical approach in the design of large integrated systems consisting of multiple genetic circuits.
The transfer function we derived relies on a steady-state assumption and is based on a deterministic model. However, the response of the circuit may have dynamic or stochastic aspects that are not predicted from the model. For some applications where the induction of inputs is transient, the dynamics of the circuit could be critical to the successful implementation of the output process. A further obstacle to standardization is that the circuit may show different response characteristics in different environments or stages of growth. For example, this AND gate produces a lower gain at low cell densities (not shown). This change in the output range could impact its connection to a downstream circuit.
The ultimate goal in genetic circuit design is to incorporate them into more complex systems consisting of multiple circuits, sensors, and actuators. Unlike electronic circuits, where the spatial wiring of a circuit determines the flow of information through the circuit, intracellular circuits prevent cross-communication through specificity of biochemical interactions. Once a part—such as T7 or SupD—is used, it cannot be used in any of the other devices. Therefore, the use of the T7 RNA polymerase-based gate makes this valuable gene unavailable for protein overexpression within the same system. An advantage of our design is that any particular transcriptional activator, including engineered sequence-specific transcription factors (), could be used in place of the polymerase gene. Similarly, the use of amber suppression precludes its use in other systems. For example, translational recoding with unnatural amino acids using amber suppression could not coexist with this logic gate (). Other translational regulators that could be used in this gate include nonsense, missense, and frameshift suppressors, or riboregulators (; ).
Pushing the boundary of genetic engineering will require a toolbox of genetic circuits that perform prescribed functions and are designed to be incorporated into larger systems. Toward this end, we have described the construction and analysis of a genetic AND gate. We have demonstrated that this gate is modular, so that it can be connected to different promoter-based inputs and used to drive different outputs. Further, in a step toward standardization, we developed a model that could be empirically parameterized, and used to predict how new promoters will connect to the gate. Circuits like this will find broad application in genetic engineering of systems in which multiple transcriptional signals must be combined to produce a specific cellular response.
All manipulations were performed in strain MC1061, DH10B, or EC100D™ pir-116 (Epicentre, Madison, WI) growing in 2YT liquid media or LB agar plates supplemented with antibiotics at 25 μg/ml at 37°C. HeLa cells were obtained from the UCSF Cell Culture Facility (San Francisco, CA) and grown in DMEM media supplemented with 10% FCS and 1% streptomycin/penicillin solution. DNA-modifying enzymes were purchased from New England Biolabs. Oligonucleotides were synthesized by Sigma-Genosys (The Woodlands, TX) and used unpurified. PCR was performed with the Roche High-Fidelity PCR kit. Plasmid pLC113 containing the salicylate promoter was a gift from Sandy Parkinson (). Plasmid pAC-SupD and the double amber mutant of T7 RNA polymerase were described previously (; ). -3-oxohexanoyl--homoserine lactone, -arabinose, and sodium salicylate were from Sigma.
Sequences of the plasmids constructed for this study are available through the Registry of Standard Biological Parts (). Reporter plasmid pBR939B is a pBR322 derivative containing the pMB1 origin of replication, an ampicillin resistance gene, and a gene under the control of a T7 promoter and a terminator. Plasmid pAC-SalSer914 is a pACYC184 derivative containing the p15A origin of replication, a chloramphenicol resistance gene, and the gene under the control of a salicylate operon and an terminator. To construct pAC-SalSer914, the salicylate operon was PCR-amplified from plasmid pLC113 with oligonucleotides ca899F and ca899R and inserted into the I and HI sites of plasmid pAC581 () to obtain plasmid pAC899. Subsequently, the gene was PCR-amplified from plasmid pAC-SupD with oligonucleotides ca914F and ca914R and inserted into the HI and RI sites of plasmid pAC899. T7 RNA polymerase-expressing plasmids were constructed in plasmid pBAC872s () containing the F plasmid origin of replication and genes, an R6K origin of replication, a kanamycin resistance gene, and a HI/RI cassette flanked by an arabinose promoter and a terminator. In strain EC100D™ pir-116, the R6K origin confers high copy number. In strain MC1061, pBAC872s derivatives are single-copy plasmids. Arabinose-inducible GFP reporter plasmids pBAC987 and pBAC978 containing the B9 and F11-derived rbss were constructed from plasmid pBAC872s. The GFP cassette present in pBAC872s was PCR-amplified with oligonucleotides ca978F or ca987F and ca606R and inserted into the HI and RI sites of pBAC872s. The GFP sequence contains a degradation tag that confers a half-life of 40 min ().
To construct Plasmid pSupDLuxR, a variant of pAC581 (pAC-SupDb) was first constructed with HI and RI sites upstream of the gene. The entire operon was PCR-amplified with oligonucleotides ca742F and ca721R from plasmid pAC-LuxGfp () and inserted into the HI and RI sites of pAC-SupDb to obtain plasmid pSupDLux. The gene in pSupDLux was then excised by inverse PCR with oligonucleotides ca752F and ca747R and recircularization with II to obtain pSupDLuxR.
Plasmid pBACr-Mgr940 was constructed from plasmid pBAC874t, a variant of pBAC872s with a promoter (). The promoter was PCR-amplified from strain MG1655 genomic DNA with oligonucleotides ca901F and ca901R and inserted into the I and HI sites of pBAC874t to obtain plasmid pBACr-Mgr901. Subsequently, variants were inserted into the HI and RI sites of pBACr-Mgr901.
Plasmid pBACr-Mgr951 was constructed from plasmids pBACr-Mgr940 and pBACr-Inv939. Plasmid pBACr-Inv939 is a pBAC874t derivative containing a T7-GFP cassette derived from pBR939b within the I and HI sites upstream of the HI/RI cassette from plasmid pBACr-AraInv () containing the rbs and the invasin gene. The T7-GFP-Invasin fragment of pBACr-Inv939 was PCR-amplified with oligonucleotides ca279 and ca951R and inserted into the I site of pBACr-Mgr940 to yield plasmid pBACr-Mgr951.
To construct the T7 RNA polymerase rbs library, the T7 RNA polymerase gene with two amber stop codons and a short 5′ linker sequence was PCR-amplified with oligonucleotides ca940F and ca564R from plasmid pREP2-HLAA02 () and inserted into the HI and RI sites of pBAC872s or pBACr-Mgr901 under the control of the arabinose promoter. In this manner, the 5′ flanking bases of the rbs and the first base of the start codon were replaced with random sequence. The library of 5′UTR variants was inserted into MC1061 cells harboring pBR939b and pAC-SalSer914 and plated on LB media supplemented with the appropriate antibiotics, 100 μg/ml salicylate and arabinose. Individual green-fluorescing colonies were grown in 500 μl aliquots in a 96-well block in the presence of 100 μg/ml salicylate, 100 μg/ml arabinose, 10 μM N-3-oxohexanoyl--homoserine lactone and/or 50 mM MgCl, or no additive, and then assayed for fluorescence in a Tecan Safire fluorescence plate reader (Tecan). From this screen, individual variants B9, F11, and pBACr-Mgr940 were characterized further. A third variant was sequenced and found to introduce a hairpin occluding the rbs, which may result in aberrant translation (). Thus, this variant is not considered further.
To observe AND gate induction by cytometry, 50 ml cultures of MC1061 cells (salicylate and arabinose gate) or DH10B cells (AHL and no Mg gate) were grown in 2YT media in the presence or absence of inducers and the appropriate antibiotics for 3 h at 37°C to OD=1 in a baffled flask with aeration. Cytometry analysis was performed with a Becton Dickinson FACSCalibur™ on bacteria diluted in PBS buffer. Counts were gated by side and forward scatter. Data for 30 000 cells were collected for each experiment. Fluorescence was tuned relative to bacteria without a GFP gene centered within the first decade of fluorescence. For fluorimetry measurements, bacteria were grown in 400 μl cultures in 96-well blocks with shaking for 8 h at 37°C to OD=1. Aliquots of 100 μl each were transferred to 96-well plates and assayed for fluorescence in a Tecan.
Bacterial cultures were diluted 100-fold in 2YT media, and 50 μl was added to 1 ml of DMEM media in 24-well plates containing a confluent culture of HeLa cells (MOI=5). After 1 h incubation at 37°C, the wells were washed once in DMEM media and then incubated for 1 h with 1 ml of DMEM supplemented with 100 μg/ml gentamicin. Subsequently the wells were washed three times with DMEM, lysed with 1 ml of 1% Triton X-100, and then spread on LB agar plates and grown 24 h. Percent invasion was determined as the ratio of recovered bacteria divided by the number of CFU present in the original diluted culture as determined by titer. Values were averaged over four repetitions of the experiment |
The experimental identification of essential genes has been carried out in some bacteria and yeast (; ; ; ; ; ; ). The methods often used to disrupt genes to determine whether they are essential or not included transposon mutagenesis and targeted disruption by homologous recombination. Using transposon mutagenesis, whole regions of chromosomes can be examined; however, the results are inconclusive, because not all regions are inactivated by random insertion. Targeted disruption, which can identify essential genes expressed in the diploid stage or expressed conditionally, is a suitable method to show whether annotated genes are essential. But intergenic regions have not been investigated. Although most essential -acting genes have been identified through gene disruption studies, the necessity of the intergenic regions has not been sufficiently clarified. For example, some genetic information in the intergenic regions is transcribed, whereas other genetic information is not. The former sites are -acting, and the latter sites are -acting.
In bacteria, chromosomes are generally uni-replicons; therefore, the origin of replication () is -acting and essential. In , other -acting sites have also been reported. The is a -acting site, which is important for cell proliferation (). This site was identified because a mutant with a very large deletion around the replication terminus () grew slowly, and the chromosomal region responsible for this growth defect was identified and termed deletion-induced filamentation or (). This site was eventually shown to be a site for recombination catalyzed by the XerC-XerD recombinase, which resolves chromosome multimers resulting from homologous recombination between replicated sister chromosomes. Although this site affects cell growth, its deletion leads to a relatively minor growth defect (). Another -acting site, , was identified as being responsible for the polar movement of but this site was not essential for cell growth ().
To understand the essential genetic information of prokaryotic chromosomes, a genomic survey of -acting essential regions is necessary. An efficient way to identify essential factors, particularly -acting chromosome regions, is thought to be the systematic construction of large-scale chromosomal deletions. If unique and essential -acting regions are on a chromosome, the deletion mutants of these regions are no longer viable even in the presence of complementing plasmids. Previously, we constructed long-range deletions of the chromosome, which led to the reduction of the genome size (; ). First, we constructed 75 deletions (medium-scale deletions (MD)) in regions lacking the essential genes, which were identified through a survey of the published literature, using the homologous recombination system. We then constructed a second series of deletions (large-scale deletions (LD)) and combined them to construct an engineered strain lacking 29.7% of the parental chromosome. In this study, we constructed deletion mutants for other chromosomal regions, particularly those containing essential genes, to identify additional essential -acting chromosome regions, while maintaining the viability of the mutants with complementing plasmids expressing the deleted genes.
First, for the chromosomal regions that were not deleted during the construction of the first MD series, the identified essential genes were cloned and the chromosomal regions containing these genes were deleted. We cloned the essential genes into mini-F plasmid vectors ( and ) either using restriction digestion and ligation methods or using the recombination system ( and ). Seven and 34 MDs were obtained in the absence and presence of the complementing plasmids, respectively. We also constructed one new LD. However, we did not succeed in constructing MD and LD deletions for the entire chromosome. Second, we investigated whether or not there were any essential -acting chromosome regions in the regions not deleted in the MD and LD. Therefore, we developed a system of moving chromosome regions into mini-F plasmids using the yeast FLP-FRT site-specific recombination system 1 (FLP-FRT1) (see , for details). shows the improved system 2 (FLP-FRT2), which is essentially the same as FLP-FRT1. Using this system, 30 additional chromosome deletions were constructed, indicating that these regions have no essential -acting chromosome regions. Third, for these regions and the other regions that were not deleted, small-scale deletions (SD) were constructed using lambda (λ)-phage and the recombination system (see , for details.). Two hundred thirty-eight and 116 small-scale deletions (SD) were obtained using these methods in the absence and presence of complementing plasmids, respectively. Fourth, we tried to construct the deletions again for the other undeleted regions using improved systems. We obtained 5 and 15 deletions with system 2 (FLP-FRT2) and system 3 (FLP-FRT3), respectively. In total, 551 chromosome deletions mutations covering all of the genome, except for and were constructed ( and and ; see the profiling of chromosome (PEC) database () for details).
The region does not contain a site that affects cell growth, other than (); therefore, the results of this work indicate that there are no unique, -acting, and essential regions other than . Eukaryotic chromosomes are multireplicons, and thus each origin of replication is not necessarily essential. Apart from the origin of replication, other -acting chromosome regions in eukaryotic cells include telomeric sequences, which are necessary for chromosome maintenance and centromeric regions, which are required for stable segregation of eukaryotic chromosomes. The centromere is a unique region in each chromosome: in theory, two centromeres on one chromosome can pull apart the chromosomal DNA between two daughter cells during mitosis (). In prokaryotic cells, the mechanism underlying bacterial chromosome segregation is not understood. So far, a prokaryotic centromere has not been identified and it is not known if one exists. Low-copy number bacterial plasmids have their own partition systems, in which a -acting DNA region plays an essential role (). But unlike eukaryotic centromeres, a plasmid carrying two copies of the -acting sequence is structurally stable. It is not known whether a eukaryote-like centromere functions in chromosome segregation in prokaryotes. Here, we did not identify any -acting and essential sites other than . Because and regions are not thought to contain a site for chromosome stability (; ; ), our results suggest that a potential -acting site for chromosome segregation may be dispensable or redundant. Alternatively, prokaryotic cell sequences equivalent to the eukaryotic centromere may not exist. Global reorganization of chromosomes triggered by a loss of this cohesion resembles eukaryotic prometaphase (; ). It is suggested that the bacterial mechanism of chromosome segregation is a primordial one to which microtubule-based processes were added later.
Our results also show that all of the -acting essential genes were cloned on the complementing plasmids; however, the cloned genes (501 genes) are not necessarily essential. reported 303 -acting essential genes by targeted disruption, but 35 of them were not cloned on our complementing plasmids, indicating that these genes are nonessential (‘Class A' in the ). The discrepancy between the two studies may be due to a difference in the strains and culture conditions used. For example, the culture media Antibiotic medium 3 (this work) contains a glucose, but LB () does not. Alternatively, it may be ascribed to the difference between a single gene knockout () and a large-scale deletion of genes (this work). For example, the anti-toxin genes and , which were identified as essential genes of the toxin-antitoxin system (), were deleted with the toxin genes and , respectively, in our study. In addition, the gene encoding a repressor of a cell division inhibitor was deleted in our study with the , the inhibitor gene, whereas the gene disruptant was not obtained in a previous study (). Furthermore, we identified 25 genes (‘Class B') and 15 genes encoding small RNA (‘Class C') as essential genes and 2 genes as nonessential genes, which were determined from the results of our gene disruption experiments (J Kato, unpublished data, 2006) and other reports (We listed the relevant PMID number in the ). In total, we identified 303 essential genes (). Sequence comparison of these essential genes with those of (271 genes) revealed that 177 were conserved between these two genera (; ). When functionally classified, genes involved in translation, protein translocation, and lipid synthesis were well conserved, whereas those involved in cell wall and membrane synthesis were not (), which may reflect structural differences in the cell wall and membrane.
In our study, 50 chromosome regions were moved to a mini-F plasmid using the FLP-FRT systems. Forty-six of them were found to contain essential gene(s), whereas 2 regions had no essential genes and were deleted without complementing plasmids. The other two regions (OCL30 and OCL34) did not contain any essential genes, but these regions were essential and therefore were not deleted without the complementing plasmids. In these regions, there may be the functionally redundant genes; one of which may be at least essential. Thirty chromosome regions were first moved to a mini-F plasmid using system 1 (FLP-FRT1). When we tried to move the other 20 regions to the plasmids using the improved system 2, (FLP-FRT2), 5 regions were moved, but the other 15 regions were not. For the regions affected by these deletions, the chromosomes and plasmids of the strains obtained at each step of system 2 (FLP-FRT2) were analyzed. The results indicated that the chromosomal regions flanked by two FRT sites had been excised after induction of the FLP recombinase, but the resultant plasmids were not stably maintained even in the wild-type strain using the mini-F temperature-sensitive replicon. Analyses of some of the excised plasmids suggested a rearrangement of the plasmid structure (data not shown). Unexpected recombination between the cloned chromosomal regions and a mini-F plasmid may cause instability of the excised plasmids. Therefore, we developed an improved system, termed system 3 (FLP-FRT3), in which the excised chromosome region was maintained by an R6K replicon. This finally allowed us to construct the remaining 15 deletion mutations.
Developments in synthetic biology have made it possible to reduce the size of the genome of K-12 (; ; ; ), and recent work indicates that genome reduction can have unanticipated benefits (). To further engineer and to make useful improvements for industry and therapeutics, such as facilitating the production of metabolites and proteins, it is important to understand both the - and -acting essential genetic information. Further analyses are necessary to experimentally clarify the minimal set of genetic information necessary and sufficient to sustain a functioning cell.
All strains used are derivatives of MG1655. The MD series was constructed in MG1655 . The LD and SD deletion series were built using MG1655 rsh3 (: (Δ()∷) :Ap), which was constructed using KM22 (). MG1655 was used to combine LD deletion units. Antibiotic Medium 3 (Becton Dickinson, USA) was used for all experiments except for those involving selection, for which LB (-N) Suc was used (LB broth with 10% sucrose and lacking NaCl). The approximate formula per liter of the Antibiotic Medium 3 is beef extract 1.5 g, yeast extract 1.5 g, peptone 5.0 g, dextrose 1.0 g, sodium chloride 3.5 g, dipotassium phosphate 3.68 g, and monopotassium phosphate 1.32 g.
Complementing plasmids were constructed or . chromosome regions were amplified by PCR using primers flanked by restriction sites, digested with restriction enzymes, and ligated into mini-F vectors ( and ) with T4 DNA ligase. (), DNA fragments to be cloned were prepared and flanked by two DNA fragments, ‘KmN' and ‘mF', by two successive PCR reactions and introduced into the strain with the gene of λ-phage and a mini-F vector, mFCm4-2. Introduced fragments were cloned into the mini-F vector by recombination, resulting in kanamycin-resistant (Km) and chloramphenicol-sensitive (Cm) complementing plasmids.
The MD series was constructed with the homologous recombination system using ColE1-related plasmids and the mutant (; ). The vector 664BSCK2-4 has two positive selection markers (Cm and Km), two negative selection markers ( streptomycin-sensitive (Sm) and ), and multicloning sites flanking the Km marker. Both chromosomal regions flanking the targeted region were cloned into 664BSCK2-4, and the resulting plasmid introduced into MG1655 . A Cm transformant, in which the plasmid was integrated by homologous recombination between the cloned region and the same region on the chromosome, was selected at 42°C. After incubation at 30°C, a Sm Km Cm colony, in which the plasmid was excised by another homologous recombination between the other cloned chromosomal region and the same region on the chromosome, was isolated and deletions were confirmed by PCR.
The SD system has been described previously (). Briefly, a linear DNA fragment encoding the Cm gene was generated by PCR using oligonucleotide primers with a 40-base-pair region of homology to regions flanking the targeted deletion. The frequency of recombination was low using primers containing a 40-base-pair region of homology, but improved upon attachment of an approximately 1-kb region of homology to either end of the Cm gene. Fragments were introduced into the strain MG1655 by electroporation and Cm recombinants were isolated. Deletions were confirmed by PCR analysis.
The LD series was constructed by the ‘CRS cassette method' using the gene-mediated λ-phage homologous recombination system and linear DNA fragments (; ). The CRS cassette is approximately 5 kb and bears one positive selection marker, Cm, and two negative selection markers, (Sm) and . A DNA fragment in which chromosomal regions flanking the region to be deleted were joined to the ends of the CRS cassette was introduced into MG1655 rsh3. Cm colonies were selected and deletions confirmed by PCR. To remove the CRS cassette, a DNA fragment in which the same flanking chromosomal regions were directly joined to each other was introduced into Cm colonies. Sm and sucrose-resistant colonies were selected and deletions confirmed by PCR.
The FLP-FRT2 system is shown in (for details, see ). In our FLP-FRT 1 prototype system, A-Km and B-Km DNA fragments, which contain the Km gene joined to two (A and B) chromosomal regions flanking the region to be deleted, were prepared by PCR using 664BSCK2-4 derivative plasmids used to construct MD deletions described above. The A-Km DNA fragment was inserted into a mini-F plasmid, mini-FtsFA (suicide plasmid A with a Cm marker), which is replication-defective at 42°C, and the B-Km DNA fragment was inserted into a R6K-related plasmid, pSG76SA (suicide plasmid B with an Ap marker), which lacks necessary for replication (; ). First, pSG76SA carrying B-Km was introduced into the wild-type strain MG1655 and the Km recombinants in which the plasmid was integrated by homologous recombination were isolated. Second, the plasmid mini-FtsFA carrying A-Km was introduced into Km colonies obtained as described, and Cm colonies were obtained. Third, was disrupted by P1 transduction to inhibit homologous recombination beyond this stage. The FLP-containing plasmid (recombinase plasmid) was introduced and recombinase expression was induced, resulting in plasmid excision and chromosome deletion. To obtain a strain lacking the FLP-plasmid, cells were incubated at 35°C, at which the FLP-plasmid does not replicate but the miniF ts replicon remains functional.
The FRT2 system introduces improvements to the FRT1 system (). Briefly, the two chromosomal regions (A and B) flanking the targeted region were joined to the Cm gene to create A-Cm and B-Cm by PCR. A-Cm was inserted into mini-FtsFAK, and B-Cm into pSG76SA. pSG76SA carrying B-Cm was introduced into the wild-type strain MG1655. Cm colonies in which the plasmid was integrated were isolated. Next, mini-FtsFAK carrying A-Cm was introduced into the Cm recombinants, and the ampicillin-resistant (Ap) recombinants were isolated at 42°C. To inhibit homologous recombination beyond this stage, was disrupted by P1 transduction. The FLP-plasmid was introduced into Cm Ap recombinants and plasmid excision and chromosome deletion was induced. In the FRT3 system, the plasmid 184 Km , encoding a functional copy of , was co-introduced with the FLP-plasmid, and the excised plasmid was maintained with the R6K replicon in addition to the miniF ts replicon (). In all other aspects, the FRT3 system was the same as the FRT2 system. |
In evolutionary protein engineering, selection using a cell-free translation system carries advantages of large library size and applicability to cytotoxic proteins. Although various protein selection techniques such as virus , mRNA display and ribosome display have been developed, most of them are the techniques in which 3'-terminus of the genotype molecule is linked to the decoded peptide via its C-terminus. Thus, they are not adequate to the evolution of the protein, of which C-terminus is essential for its function (such as various peptide ligands, protein-protein interaction domain, membrane proteins, peptide hormones, etc). Therefore, it is worthwhile to devise the method in which the genotype molecule is linked to the decoded protein via its N-terminus, i.e. C-terminus of the protein is displayed freely.
Also in phage display method, the peptide is fused usually at the N-terminus of coat protein. However, the display method in which the peptide library was fused at the C-terminus of M13 phage major coat protein , and T7 phage D-coat protein has been also reported. The carboxyl terminal end of PDZ domain-ligands, which is essential to be recognized by PDZ domain, was screened , and mapping of the C-terminal epitope of the Alzheimer's disease specific antibody was carried out using such carboxyl-terminal phage display.
Several methods that link the protein to its mRNA (or DNA) via N-terminus have been reported. Sawata and Taira reported the binding procedure between MS2 coat protein dimer fused at N-terminus of the protein and the specific recognition sequence in its mRNA . Yanagawa and coworkers developed the method named “STABLE” in which the streptavidin fused at N-terminus of a peptide was bound to biotin labeled gene DNA in emulsion ,. These methods are simple and effective; however, screening condition is limited because its linkage between the peptide and its genotype molecule is non-covalent. Baskerville and Bartel made the ribozyme that catalyzes a phosphoamide bond formation between the 5'-terminus of the encoding RNA and the N-terminus of the decoded polypeptide . Nester and coworkers developed the method that employed the property of VirD2 protein, which covalently binds to the specific DNA sequence . These methods employ a covalent bond to linkage the peptide and its genotype molecule; however, these have not been applied yet to selection.
To date, many studies on the non-natural amino acid incorporation into a protein using nonsense , or frame-shift suppression have been reported. And recently, these have been applied to protein display methods to introduce the mRNA-peptide fusion molecule library containing non-natural amino acid ,.
In this article we report the novel method to link the N-terminus of the peptide to its mRNA. In our method, the mRNA is linked via ssDNA and spacer to phenylalanine derivative, which is acylated to amber sup tRNA. When this modified mRNA is subjected to a cell-free translation system, this phenylalanine derivative would be incorporated into the N-terminus region of the nascent peptide at the amber stop codon inserted at the beginning of the encoded region. Consequently, C-terminus of the peptide is displayed on its mRNA (Fig.).
The DNA-linker having a hydrazide group at 5'-terminus (hydrazide-linker) was synthesized as follows. 5'-NH-(CH)-(CHCHO)-(dA)-(CHCHO)-GGG-(dT-FITC)-CGGGGGGCAAAA-3', where (dA) is 30 nucleotides of deoxy adenosine, (dT-FITC) is fluorescein-modified deoxy thymidine, was purchased from Tsukuba Oligo Service Co., Ltd (Ibaraki, Japan). Single riboadenosine was added to 3'-terminus of the oligomer (1.0 nmol) using terminal deoxynucleotidyl transferase (24 U, Takara Bio) with 20 µmol ATP at 37 °C for overnight in 50 µl of supplied TdT buffer. The product was purified on P6 column (Bio Rad) to remove excess ATP, followed by phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation. The precipitate was dissolved in 65 µl of 0.5 M phosphate buffer (pH 9.0) and then mixed with 10 µl of 0.025 M succinic anhydride in DMSO for overnight at room temperature and purified on P6 column (Bio Rad) pre-equilibrated with HO. Purified sample was mixed with 2 mg adipodihydrazide and 5 mg EDC (1-ethyl-3- (3-dimethylaminopropyl) carbodiimide) in 70 µl of 0.1 M phosphate buffer (pH 7.2) for 4 hour at room temperature and purified on P6 column pre-equilibrated with HO.
where ν is the distance-through-bond in nucleotides .
Probability distribution of local concentration j (r) is described as :
The value of r for our case is 14 nm that is the distance-through-space between the exit position of mRNA on ribosome and the entrance of A-site of ribosome . On the assumption that mRNA and the linker were composed of ssRNA, the probability distribution of local concentration of linked aminoacyl sup tRNA on the entrance of A-site of ribosome is given as:
The His-tag encoding DNA (178 bp, Fig.A) was constructed by three steps of PCR amplification. The first step PCR was performed using three oligomers; NL-temp (5'-ATGGGCTAGGGTGTGAGCAAGGGCGAGGAGCTGTTCACCGGGGCAGCGCATCAC-3'), prNL+ (5'-CGGTTTCCCTCTAGAAATAATTTTGTTTAACTTTAAGAAGGAGATATATCCATGGGCTAGGGTGTGAGCA-3') and prNL- (5'-ACTCCTTAATGATGGTGATGGTGATGCGCTGCCCCGGT-3') and the resulting dsDNA was further amplified using two primers; prNLφ2.5GC+ (5'- CAGTAATACGACTCACTATTAGGGCCCCCCGACCCCGGTTTCCCTCTAGAAATAATTTTGT–3') and prNLHis- (5'-GCATCGACTCCTTAATGATGGTGATGGTGATGCGCTGCCCCGGTG-3'). The final step PCR was performed using the second step product dsDNA and two primers; prRT+ (5'-CAGTAATACGACTCACTATTAGGGCCCCCCGAC-3') and prRT- (5'-GAATTCGCCCTTGCATCGACTC-3'). The FLAG encoding DNA (190 bp, Fig.A) was constructed by two steps of PCR amplification. The first step PCR was performed using the His-tag encoding DNA and two primers; prRT+ and prNLFLAG- (5'-GCATCGACTCCTTATTACTTGTCATCGTCATCCTTGTAATCCGCCGCTGCCCCGGTGAACAG-3') and the resulting dsDNA was further amplified using two primers; prRT+ and prRT-. These PCR products were purified using QIAquick PCR purification kit (Qiagen) and transcribed in vitro using RiboMAX Large Scale RNA Production System - T7 (Promega) in the presence of 36 mM AMP and 3.6 mM each NTPs. The RNA transcript was purified on P6 column to remove excess AMP, followed by phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation.
The mRNA (2 µM) was hybridized to the DNA moiety of the hydrazide-linker (4 µM) and prRT- (4 µM) by heating at 95 °C and cooling to 25 °C in 50 µl of T4 RNA ligase buffer (Takara Bio) and ligation reaction was started by adding T4 RNA ligase (40 U, Takara Bio) and ribonuclease inhibitor (40 U, Takara Bio). Ligation reaction was performed at 25 °C for overnight and the ligated product was analyzed by 8 M urea 8 % PAGE using TBE running buffer at 65 °C and were visualized with fluorescence of FITC and then visualized again after staining by SYBR Green II (Cambrex) using a fluorescence imager (Pharos FX; Bio-Rad). Ligated product was purified using RNeasy Mini Kit (Qiagen).
4-acetyl-L-phenylalanine cyanomethyl ester was synthesized by as follows; 10 mg of Boc-4-acetyl-L-phenylalanine (Chem Impex) was dissolved in 1 ml of acetonitrile and mixed with 350 µl of triethylamine and 100 µl of chloroacetonitrile. This mixture was stirred at room temperature for over night and the reaction mixture was diluted with 10 ml of ethyl acetate and extracted with 20 ml of 0.5 N HCl. The organic layer was then washed with 20 ml of 4 % NaHCOand 10 ml of saturated aqueous sodium chloride and dried using a rotary evaporator. The resulting residue was mixed with 250 µl of TFA/anisole mixture (9:1) and stirred at room temperature for 30 min. The solvent was removed using a rotary evaporator and 1 ml of 4 M HCl in dioxane was added to form the hydrochloride salt. The solution was concentrated using a rotary evaporator and 2 ml of ether was added to precipitate the hydrochloride salt. These reactions were traced by Thin Layer Chromatography.
Aminoacyl-sup tRNA was prepared by the same procedure previously reported with minor modification using 4-acetyl-L-phenylalanine cyanomethyl ester as the amino acid substrate.
The template DNA encoding the aminoacyl ribozyme was constructed by two steps of PCR amplification. The first PCR was performed using three oligomers; Fx (5'-ACCTAACGCCAATACCCTTTCGGGCCTGCGGAAATCTTTCGATCC-3'), P5-1 (5'-ACGCATATGTAATACGACTCACTATAGGATCGAAAGATTTCCGC-3') and P5-2 (5'-GGTAACACGCATATGTAATACGACTC-3'). The resulting dsDNA was further amplified using two primers; p5-2 and p3-2 (5'-T-ACCTAACGCC-3'). The template DNA encoding the engineered tRNA was constructed by two steps of PCR amplification. The first PCR was performed using three oligomers; tR (5'-TGGTGCCTCTGACTGGACTCGAACCAGTGACATACGGATTTAGAGTCCGCCGTTCTACCGACTGAACTACAGAGGC-3'), p5-3 (5'-ACGCATATGTAATACGACTCACTATAGCCTCTGTAGTTCAGTCGGT-3') and p3-3 (5'-TGGTGCCTCTGACTGGACTC-3'). The resulting dsDNA was further amplified using two primers; p5-2 and p3-3. Amplified these two dsDNAs, encoding aminoacyl ribozyme or tRNA, were transcribed in vitro using RiboMAX Large Scale RNA transcription System – T7 in the presence of 7.5 mM GMP and 3.75 mM each NTPs. The RNA transcript was purified on P6 column to remove excess GMP, followed by phenol/chloroform/isoamyl alcohol extraction and ethanol precipitation.
400 µl of 0.1 M NaIO was added to 1 ml of 30 µM ribozyme, and the mixture was incubated on ice for 20 min. The oxidized RNA was then precipitated with 14 ml of 2 % LiClO in acetone and washed with 1 ml acetone. The pellet was dissolved in 1.4 ml of 0.1 M sodium acetate (pH 5.0) and added to 0.7 ml of adipic acid dihydrazide-agarose (Sigma) that was prewashed with RNase-free water. The mixture was stirred at room temperature for three hour and then 300 µl of 1 M NaBHCN was added and the mixture was incubated for 30 min to stabilize the imine bond. The resin-immobilized ribozyme was washed with 1.4 ml of W1 (0.1 M sodium acetate, 300 mM NaCl, 7.5 M urea, 0.1% SDS, pH 5.0) and suspended in 2.1 ml of W1.
30 µl of 50 µM tRNA was folded by heating at 95 °C and cooling to room temperature in a buffer (50 mM EPPS, 12.5 mM KCl, 12 mM MgCl, pH 7.0) and then diluted in 15 µl of a buffer (50 mM EPPS, 12.5 mM KCl, 3.6 M MgCl, pH 7.0) and 20 µl of a buffer (50 mM EPPS, 12.5 mM KCl, 1.2 M MgCl, pH 7.0). The folded tRNA solution was mixed with 5 nmol of resin-immobilized ribozyme that was prewashed with RNase-free water, and then 12.5 µl of 100 mM 4-acetyl-L-phenylalanine cyanomethyl ester in DMSO was added to this mixture followed by adding 1.6 µl of 0.25 M KOH (to adjust pH to 7.0-7.2). After 2 hour incubation under rotation at 4 °C, EK buffer (50 mM EPPS, 12.5 mM KCl, 10 mM EDTA, pH 7.0) was added and the supernatant, containing aminoacyl-tRNA product, was recovered and ethanol precipitated. Confirmation of aminoacylation was performed as follows. 50 pmol of aminoacylated tRNA was dissolved in 5 µl of EPPS (0.1 M, pH 5.9) containing 20 mM 6-(Biotinylamino)hexanoic acid N-hydroxy-sulfosuccinimide ester (Dojindo), and 1.0 µl of EPPS-KOH (0.3 M, pH 9.1) was added to adjust pH to 8.0. The mixture was incubated for one hour on ice and ethanol precipitated. The pellet of biotinylated aminoacyl-tRNA was dissolved in 10 µl of the loading buffer (167 pmol streptavidin, 0.1 M sodium acetate, 8 M urea, 0.05 % bromophenol blue, 0.05 % xylene cyanol, pH 5.0) and analyzed by 8 M urea 15 % PAGE using 50 mM sodium acetate (pH 5.0) as a running buffer at 30 °C and visualized after staining by SYBR Green II.
10 pmol of the mRNA-linker ligation product encoding His-tag sequence was mixed with 100 pmol of 4-acetyl-phenylalanyl tRNA in 10 µl of 1 mM sodium acetate (pH 5.2) for overnight at room temperature and translated in 16 µl volume of S30 Extract System for Linear Templates (Promega) for 5 min at 37 °C with SUPERase-In RNase inhibitor (20 U, Ambion). The translated sample was incubated in 100 µl of Ni-NTA agarose (Qiagen) suspended in wash buffer (50 mM NaHPO, 300 mM NaCl, 20 mM imidazole, pH 8.0) for 1 hour at 4 °C using a rotary mixer. The Ni-NTA agarose was washed two times with 200 µl of wash buffer and eluted with 50 µl of elution buffer (50 mM NaHPO, 300 mM NaCl, 250 mM imidazole, pH 8.0). The eluted sample was desalted with MicroSpin G-25 Column (GE Healthcare) and amplified by RT-PCR using . Polymerase One-Step RT-PCR System (Roche) using prRT+ and prRT- as primers. As negative controls, the translation product without methionine and the translation product without modification of 4-acethyl-phenylalanine tRNA to the linker before translation were screened with Ni-NTA agarose and amplified by RT-PCR. The amplified products were analyzed by 8 M urea 8 % PAGE using TBE running buffer at 65 °C and visualized after staining by SYBR Green I (Cambrex).
In selective enrichment analysis of the DNA encoding the His-tag sequence, Mixture of the mRNA-linker ligation products (1 pmol of the His-tag encoding molecule and 10 pmol of the FLAG-tag encoding molecule) was reacted with 100 pmol of 4-acetyl-phenylalanyl tRNA in 10 µl of 1 mM sodium acetate (pH 5.2) for overnight at room temperature and translated in 16 µl volume of S30 Extract System for Linear Templates for 5 min at 37 °C with SUPERase-In RNase inhibitor (20U). The translated sample was incubated in 100 µl of Ni-NTA agarose suspended in wash buffer (50 mM NaHPO, 300 mM NaCl, 20 mM imidazole, pH 8.0) for 1 hour at 4 °C using a rotary mixer. The Ni-NTA agarose was washed two times with 200 µl of wash buffer and eluted with 50 µl of elution buffer (50 mM NaHPO, 300 mM NaCl, 250 mM imidazole, pH 8.0). The eluted sample was desalted with MicroSpin G-25 Column and amplified by RT-PCR using . Polymerase One-Step RT-PCR System using prRT+ and prRT- as primers. The amplified product was transcribed and ligated with hydrazide-linker for the next round of screening. The amplified product of each screening round was analyzed by 8 M urea 8 % PAGE using TBE running buffer at 65 °C and visualized after staining by SYBR Green I.
In selective enrichment analysis of the DNA encoding the FLAG-tag sequence, Mixture of the mRNA-linker ligation products (1 pmol of FLAG-tag encoding molecule and 10 pmol of His-tag encoding molecule) was reacted with 100 pmol of 4-acetyl-phenylalanyl tRNA in 10 µl of 1 mM sodium acetate (pH 5.2) for overnight at room temperature and translated in 16 µl reaction volume of S30 Extract System for Linear Templates (Promega) for 5 min at 37 °C with SUPERase-In RNase inhibitor (20U, Ambion). The translated sample was diluted in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 % Triton X-100, pH 8.0) and added to 40 µl of anti-FLAG M2 agarose (Sigma), followed by incubation for 1 hour at 4 °C using a rotary mixer. The anti-FLAG agarose was washed three times with 500 µl lysis buffer and incubated with 100 µl of TBS containing 0.1 mg / ml of FLAG peptide for 1 hour at 4 °C. The supernatant was desalted with MicroSpin G-25 Column and amplified by RT-PCR using . Polymerase One-Step RT-PCR System using prRT+ and prRT- as primers. The amplified product was transcribed and ligated with hydrazide-linker for the next round of screening. The amplified product of each screening round was analyzed by 8 M urea 8 % PAGE using TBE running buffer at 65 °C and visualized after staining by SYBR Green I.
Selective enrichment of the DNA encoding the His-tag sequence using PURESYSTEM was performed as follows. The mixture of the mRNA-linker ligation products (0.5 pmol of His-tag encoding molecule and 5 pmol of FLAG-tag encoding molecule) was reacted with 50 pmol of 4-acetyl-phenylalanyl tRNA in 5 µl of 1 mM sodium acetate (pH 5.2) for overnight at room temperature and the reaction product was ethanol precipitated. The precipitated sample was translated in 3.13 µl reaction volume of PURESYSTEM classic II (Post Genome Institute Co., Ltd.) for 15 min at 37 °C with SUPERase-In RNase inhibitor. The translated sample was incubated in 50 µl of Ni-NTA agarose suspended in wash buffer for 4 hour at 4 °C using a rotary mixer. The Ni-NTA agarose was washed five times with 400 µl of wash buffer and eluted with 100 µl of elution buffer. The eluted sample was desalted with MicroSpin G-25 Column and amplified by RT-PCR using . Polymerase Two-Step RT-PCR System (Roche) and KOD dash DNA polymerase (Toyobo) using prRT+ and prRT-FITC (fluorescein was modified on the 5'-terminus of prRT-) as primers. The amplified product was analyzed by 8 M urea 8 % PAGE using TBE running buffer at 65 °C and visualized with fluorescence of FITC using a fluorescence imager.
Selective enrichment of the DNA encoding the FLAG-tag sequence using PURESYSTEM was performed as follows. The mixture of the mRNA-linker ligation products (0.5 pmol of FLAG-tag encoding molecule and 5 pmol of His-tag encoding molecule) was reacted with 50 pmol of 4-acetyl-phenylalanyl tRNA in 5 µl of 1 mM sodium acetate (pH 5.2) for overnight at room temperature and the reaction product was ethanol precipitated. The precipitated sample was translated in 3.13 µl reaction volume of PURESYSTEM classic II for 15 min at 37 °C with SUPERase-In RNase inhibitor. The translated sample was diluted in TBS buffer and added to 40 µl of anti-FLAG M2 agarose, followed by incubation for 4 hour at 4 °C using a rotary mixer. The anti-FLAG M2 agarose was washed five times with 400 µl of TBS buffer and incubated with 100 µl of Elution buffer (0.1 M glycine-HCl, pH3.5) for 10 min at room temperature. The supernatant was desalted with MicroSpin G-25 Column and amplified by RT-PCR using . Polymerase Two-Step RT-PCR System and KOD dash DNA polymerase using prRT+ and prRT-FITC as primers. The amplified product was analyzed by 8 M urea 8 % PAGE using TBE running buffer at 65 °C and visualized with fluorescence of FITC using a fluorescence imager.
5 pmol of mRNA-linker ligation product was reacted with 50 pmol of 4-acetyl-phenylalanyl tRNA in 0.1 M sodium acetate (pH 5.2) for overnight at room temperature and the reaction product was ethanol precipitated. The precipitated pellet was translated in 6.25 µl reaction volume of PURESYSTEM classic II for 30 min at 37 °C with SUPERase-In RNase inhibitor. The translation product was blotted to PVDF-membrane after 15 % SDS-PAGE and labeled with anti-FLAG M2 monoclonal antibody (Sigma) as primary antibody and Cy3-linked anti-mouse IgG (GE Healthcare) as secondary antibody. Fluorescence of Cy3 was detected using a fluorescence imager.
The connecting efficiency between the mRNA and the linker-molecule affects the diversity of the library. Thus, various sophisticated procedures have been devised -. In this report, we applied the Y-ligation method , which is the method of two single-strands ligation in the presence of a double-stranded stem to make a stem-loop structure in high efficiency. Because the transcripts were used as a phosphate donor substrate and the DNA moiety of hydrazide-linker were used as an acceptor substrate for T4 RNA ligase (Fig.B), several modifications were required. First, transcription must be performed with excessive AMP (for T7φ2.5 promoter, GMP for normal T7 promoter and SP6 promoter) to generate the monophosphate at the 5'-terminus of the transcript and a riboadenosine must be attached to the 3'-terminus of the DNA moiety of hydrazide-linker by terminal deoxyribonucleotidyl transferase to be recognized by T4 RNA ligase (Fig.B) . Unless these modifications were performed, the ligated product was hardly observed (data not shown). Second, since 3'-OH of the transcript can be ligated with 5'-monophosphate of itself, self-ligation of the transcript was occurred (Fig., lane 1-2). For blocking of the 3'-OH of the transcript, the DNA oligomer prRT- was hybridized to 3'-terminus region of the transcript before ligation reaction (Fig.B). Self-ligation was effectively prevented (Fig., lane 4-5) and yield of desired product was increased from 10 % to 90 % (Fig.B, lane 1 and 4). This blocking method is simple and effective compared with other methods (e.g., making 2', 3'-cyclic phosphate group by self-cleavage of ribozyme placed at 3'-terminus , or cleavage with deoxyribozyme ,.). Third, it is reported that T4 RNA ligase does not prefer guanine residue as a phosphate donor ,, but common promoters for transcription start the transcription with guanine residue, which is located at 5'-terminus of transcripts and act as phosphate donor in ligation reaction. Therefore, it is suspected that the transcripts generated by guanine-initiation type promoter are not suitable for donor substrate of T4 RNA ligase. Hence, We applied T7φ2.5 promoter, which initiates the transcription with adenine residue , because adenine is better than guanine as a donor substrate of T4 RNA ligase ,. We compared the ligation efficiency of transcripts generated by three kind of promoters; T7, SP6 and T7φ2.5. These promoters generate the fixed triplet sequence at 5'-terminus of the transcripts; GGG, GAA and AGG respectively. The transcript by T7φ2.5 promoter gave the best efficiency of over 80 %. For SP6 and T7, the efficiency saturated at about 55 % and 20 %, respectively (Fig.).
tRNAwas acylated with 4-acetyl-L-phenylalanine using resin-immobilized ribozyme and aminoacylation efficiency was analyzed using a streptavidin-dependent gel-shift assay after biotinylation of the α-amino group of 4-acetyl-L-phenylalanine. The efficiency was approximately 15 % in our procedure (Fig.A).
The yield of conjugation between mRNA-linker-hydrazide and 4-acetyl-L-phenylalanyl tRNA was estimated as follows. 7.5 pmol of hydrazide-linker and 75 pmol of 4-acetyl-L-phenylalanyl tRNA were incubated in 5 µl of 0.1 M sodium acetate (pH 5.2) for overnight at room temperature. The mixture was mixed with equal volume of loading buffer (0.1 M sodium acetate, 8 M urea, 0.05 % bromophenol blue, 0.05 % xylene cyanol, pH 5.0) and analyzed by 8 M urea 15 % PAGE using 50 mM sodium acetate (pH 5.0) as a running buffer at 30 °C and visualized with fluorescence of FITC and then visualized again after staining by SYBR Green II. The yield of the conjugation was approximately 10 % because of 90 % of hydrazide-linker was left without conjugation (Fig. B). This yield may be including the event of hydrolysis of aminoacyl-tRNA that has been conjugated to hydrazide-linker.
We demonstrated a model selection experiment using a histidine-tag having high affinity to Ni. The translation product displaying the His-tag peptide was screened using a Ni-NTA agarose column. The bound sample was eluted and used as a template for RT-PCR. As negative controls, the translation product without methionine that is necessary to initiate translation and the translation product without modification of 4-acethyl-phenylalanine sup tRNA to the linker before translation were screened with Ni-NTA agarose. The RT-PCR products were analyzed by electrophoresis and the desired product with higher amount was confirmed than one of negative controls generated by non-specific binding (Fig.A).
Next, we performed enrichment experiments for the DNA mixture. It was achieved to enrich the DNA encoding the His-tag sequence from a mixture containing an order of magnitude excess of the other DNA encoding FLAG-tag sequence (Fig.A) in three rounds screening with Ni-NTA (Fig. B). Furthermore, enrichment of the DNA encoding the FLAG-tag sequence from a mixture containing an order of magnitude excess of the other DNA encoding His-tag sequence was also accomplished in single round screening with anti-FLAG M2 antibody (Fig. B). Additionally, we also performed enrichment experiments using another translation system (PURESYSTEM) and achieved the enrichment of desired sequence (Fig. C). These results indicated that the enrichment was not an artefact by nonspecific binding of displayed peptide but a fact by specific binding of each peptide sequence.
The mRNA-linker ligation product encoding the FLAG peptide was translated and the mRNA-peptide fusion product was analyzed by SDS-PAGE and western blotting. Although the migration of the band caused by fusion formation was not detected on SDS-PAGE (Fig. A), the fusion molecule was detected by the western blotting (Fig. B). This low yield of the fusion formation was corresponding to relatively low efficiency of the selective enrichment (Fig. B, C). The low yield (~10 %) of the linker-aminoacyl-tRNA conjugate (Fig. B) affected directly the yield of the final product. In general, non-natural amino acid possessing large side group was incorporated into protein with relatively low efficiency. However, it is also reported that incorporation efficiency was not dependent on size of side group but on its shape, and the linear shape was appropriate . In addition, non-natural amino acids possessing relatively long side chain (BODIPY FL-X-aminophenylalanine or ε-N-biotinyl-L-lysine ,,) were also successfully incorporated into the polypeptide. The side chain of the amino acid used in this study was linear and did not contain any large aromatic ring. However, since the side chain of the non-natural amino acid used in this study was extraordinarily long, this amino acid may be difficult to be incorporated into polypeptide or premature termination may be occur during peptide chain elongation if this amino acid was incorporated. The linker length also must affect the formation efficiency of the mRNA-peptide fusion. Therefore, experimental optimization of the linker length will give better efficiency although the length of the linker used in this study was optimized theoretically.
We have developed a new method to prepare an mRNA-peptide fusion molecule covalently linked via N-terminus during translation. Because this fusion molecule displays the C-terminus of the peptide, it makes it possible to selection of the peptides for which function the C-terminus is essential. And also because this fused molecule has the free 3'-terminus of mRNA, it would make it possible to select an RNA replicase of de novo initiation-type.
And because a stop codon is inserted near downstream of the initiation codon, the translation reaction terminates just after initiation unless the sup tRNA is incorporated. Therefore unfused peptides that must compete with the fused peptides are never generated. Furthermore, this system is able to work even if the mRNA contains a termination codon. This method may be extended to the fusion formation linked at any position of protein by changing the position of amber codon that corresponds to the incorporation site of the linked aminoacyl sup tRNA. Thus, it may be possible that displaying the protein which having free terminus at both end and that arbitrary selection of the linking point on the protein according to its structure. |
Reactive oxygen species (ROS) are produced endogenously, during normal aerobic metabolism and under various pathological conditions, and exogenously, such as upon exposure to UV light, ionizing radiation, environmental mutagens and carcinogens. DNA is susceptible to damage by ROS, and the accumulation of oxidative DNA lesions is associated with aging and a variety of human diseases, including cancer and neurodegeneration (,).
A large number of single-nucleobase lesions induced by ROS have been widely studied for their formation, mutagenesis and repair (,). Most common point mutations induced by ROS are C→T transitions, suggesting that modified cytosine derivatives are among the most abundant and mutagenic oxidative DNA lesions (). In mammalian cells, cytosines at CpG sites are frequently methylated at the C5 carbon to form 5-methylcytosine (mC), which accounts for ∼4% of all dC residues in humans (,). The methylated CpGs are mutational hot spots in human tumor suppressor gene, and the most common mutation is mC→T transition ().
Pfeifer and coworkers () recently observed unusual mCG→TT tandem mutations when CpG-methylated pSP189 plasmid was treated with Cu(II)/HO/ascorbate and replicated in nucleotide excision repair (NER)-deficient human XPA cells, suggesting that vicinal base damages or intrastrand cross-link lesions formed at methylated CpG dinucleotide sites might be involved. In this context, we previously identified the mC[5m-8]G and G[8-5m]mC cross-link lesions (structures shown in ), where the methyl carbon of mC and the C8 of its adjacent guanine are covalently bonded, in oligodeoxyribonucleotides (ODNs) upon exposure to γ rays under aerobic and anaerobic conditions (,). Along this line, CG and GC () cross-link lesions, where the C5 carbon of cytosine is coupled with the C8 carbon of its neighboring guanine, can also be induced in aqueous solutions of synthetic ODNs exposed to γ- or X-rays (,).
Others () and we (,,) demonstrated that intrastrand cross-link lesions could be initiated from a single radical of pyrimidine bases. However, the same radical is also able to couple readily with molecular oxygen and the resulting peroxyl radical can be transformed to give single-nucleobase lesions. For instance, 5-formyl-2′-deoxycytidine (5-FmdC) and 5-hydroxymethyl-2′-deoxycytidine (5-HmdC) can form from the methyl radical of 5-methyl-2′-deoxycytidine (). Likewise, 5-hydroxy-2′-deoxycytidine (5-OHdC) can be induced from the 6-hydroxy-5-yl radical of 2′-deoxycytidine (). The quantification of intrastrand cross-link and single-nucleobase lesions initiated from the aforementioned radials under aerobic conditions can allow for the assessment of the relative contributions of these two pathways, i.e. coupling with O or with its vicinal guanine base. Furthermore, it is important to compare the profiles of oxidative lesions formed in methylated and unmethylated sequences, which may offer insights into understanding the ubiquitous C→T and the unusual mCG→TT mutations found at CpG sites (,).
Hydrogen peroxide (HO) is produced by endogenous metabolic processes, and it can lead to the formation of hydroxyl radical in the presence of transition metal ions in their reduced states, e.g. Fe(II) or Cu(I). In this context, copper binds to DNA with high affinity to G:C base pairs (), and it plays a pivotal role in maintaining the structure and integrity of chromatin (). Cu(II) and HO, often with the addition of ascorbic acid, can result in the formation of highly reactive species and produce extensive strand breaks () as well as many types of single-base lesions in DNA (), which was found to be enhanced while DNA was packed into nucleosomes (). It was proposed that free copper ion primarily mediates the formation of frank strand breaks, whereas DNA-bound copper induces mainly nucleobase modifications through the formation of DNA–Cu(I)–HO complexes (,).
Iron is another biologically relevant transition metal. studies showed that Fe(III)-dependent DNA fragmentation is much less extensive than that produced by equivalent amount of Cu(II) ions under otherwise comparable reaction conditions (), and Fe(II) induces significantly less nucleobase damage in the presence of HO than does Cu(II) (,). On the other hand, studies with Jurkat cells revealed that intracellular iron, but not copper, plays an important role in HO-mediated formation of strand breaks in DNA ().
Although the structures of various single-nucleobase lesions induced by Fenton reactions have been well established, there were few reports on the formation of cross-linked nucleobase lesions from treatment with Fenton-type reagents. In the latter respect, Randerath and others () detected, by P-postlabeling assay, several bulky DNA adducts that are induced endogenously in animal tissues or from the Fenton reaction mixture of synthetic ODNs. On the grounds that some of the bulky adducts were commonly formed in ODNs housing-specific dinucleotide sequences, these authors proposed that these adducts might be intrastrand cross-link lesions (). These bulky lesions were also found in tissues from animals exposed to pro-oxidant chemicals, which firmly linked the formation of the bulky adducts to ROS (,). More recently, Randerath . () demonstrated that some of the bulky DNA adducts actually contained the 8,5′-cyclo-2′-deoxyadenosine (cyclo-dA). The cyclo-dA bears a covalent bond between the C8 of adenine and the C carbon in the same nucleoside. There were also indications that intrastrand cross-links might be induced in salmon sperm DNA upon treatment with Fenton-type reagents; the structures of these putative cross-link lesions, however, remain elusive (,). Recently, we first reported the formation of a structurally defined G[8-5m]T intrastrand cross-link lesion, which bears a covalent bond between the methyl carbon of thymine and the C8 of its adjacent 5′ guanine, in calf thymus DNA upon treatment with Cu(II)/HO/ascorbate (). It remains to be established whether intrastrand cross-link lesions can also form between guanine and its neighboring cytosine or mC in DNA upon exposure to Fenton-type reagents.
In the present study, we synthesized several isotope-labeled oxidative single-base and cross-link lesions originated from 2′-deoxycytidine (dC) and 5-methyl-2′-deoxycytidine (5-mdC) (), and employed LC-MS/MS with the standard isotope dilution method to identify and quantify the single-nucleobase and intrastrand cross-link lesions formed in synthetic duplex ODNs upon incubation with Cu(II)/HO/ascorbate or Fe(II)/HO/ascorbate. The results allowed us to compare the reactivities of methylated and unmethylated cytosine residues toward Fenton-type reagents mediated by two different transition metal ions, i.e. iron and copper. Moreover, we demonstrated that GC, mC[5m-8]G and G[8-5m]mC cross-link lesions, which were previously shown to be generated upon exposure to γ rays (), could also be induced by Fenton-type reagents.
CuCl, (NH)Fe(SO)·6HO, -methionine, -ascorbic acid and alkaline phosphatase were from Sigma–Aldrich (St Louis, MO, USA). Hydrogen peroxide (30%) and nuclease P1 were purchased from Fisher Scientific (Fair Lawn, NJ, USA) and MP Biomedicals (Aurora, OH, USA), respectively. Snake venom phosphodiesterase and calf spleen phosphodiesterase were obtained from US Biological (Swampscott, MA, USA). Common reagents for solid-phase DNA synthesis were obtained from Glen Research Co. (Sterling, VA, USA). Unmodified ODNs () used in this study were purchased from Integrated DNA Technologies (Coralville, IA, USA) and purified by HPLC. [2-amino-1,3,7,9–]-8-oxo-2′-deoxyguanosine was synthesized as described previously ().
depicts the structures of the lesions that we quantified in this study. The sites of isotope incorporation are indicated by bold italic font, in which the nitrogen, carbon and hydrogen atoms were replaced with N, C and deuterium, respectively. The labeled and unlabeled single-nucleobase lesions were prepared according to previously described procedures ().
The concentrations of stock solutions of 5-FmdC, 5-HmdC, 5-OHdC, 8-oxodG, d(G[8-5m]mC) and d(GC) were determined by UV absorbance measurements. The molar extinction coefficients (in l/mol/cm) used to quantify the standard solutions were: 5-FmdC, ε = 13 700; 5-HmdC, ε = 9070; 5-OHdC, ε = 6025 (); 8-oxodG, ε = 9700 (); d(G[8-5m]mC), ε = 23 800 and d(GC), ε = 22 800 (). The values for 5-FmdC, 5-HmdC and d(G[8-5m]mC) were determined by H-NMR (Figures S4–S6) following the previously described method ().
The ODNs were annealed in a buffer containing 50 mM NaCl and 20 mM phosphate (pH 6.9) by heating the solution to 90°C and cooling slowly to room temperature. Aliquots of ODNs (5 nmol) were incubated with CuCl or (NH)Fe(SO) (6.25–100 µM), HO (50–800 µM) and ascorbate (0.5–8 mM) in a 100-µl solution containing 25 mM NaCl and 50 mM phosphate (pH 7.0) at room temperature for 60 min. All chemicals were freshly dissolved in doubly distilled water and the reactions were carried out under aerobic conditions. The detailed concentrations of individual Fenton-type reagents used for the reactions were shown in . The reactions were terminated by adding an excess amount of -methionine, and the ODN samples were desalted by ethanol precipitation.
Control experiments were also carried out to examine the effect of individual components in Fenton-type system on the formation of DNA lesions. First, we incubated ODNs with 200 µM HO and 2 mM ascorbate in the absence of Cu(II) or Fe(II). Second, ODNs were incubated with 25 µM CuCl or (NH)Fe(SO) and 200 µM HO, and no ascorbate was added. In addition, experiments were carried out in the presence of 4 mM dimethyl sulfoxide (DMSO), a hydroxyl radical scavenger, for Fenton reaction condition C ().
Four units of nuclease P1, 0.005 unit of calf spleen phosphodiesterase, and 1.5 µl solution containing 300 mM sodium acetate (pH 5.0) and 10 mM zinc acetate were added to 15 µl ODN samples treated with Fenton-type reagent, and the digestion was carried out at 37°C for 6 h. To the digestion mixture were then added 10 U of alkaline phosphatase, 0.05 U of snake venom phosphodiesterase and 5 µl of 0.5 M Tris–HCl buffer (pH 8.9). The digestion was continued at 37°C for 6 h, and the enzymes were removed by passing through a 10-kDa cutoff Centricon membrane (Millipore, Billerica, MA). The amount of nucleosides in the mixture was quantified by UV absorbance measurements, and to the mixture were then added isotope-labeled standard lesions. The resulting aliquots were subsequently subjected to HPLC enrichment and LC-MS/MS analysis.
The enrichment of DNA lesions from the digestion mixture of Fenton-type reagent-treated ODNs was performed on a Surveyor HPLC system (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a 4.6 × 150 mm Zorbax SB-AQ reverse phase C18 column (5 µm in particle size, Agilent Technologies). A gradient of 40 min 0–25% acetonitrile in 10 mM ammonium formate (pH 6.3) was employed, and the flow rate was 0.60 ml/min. We collected fractions in a wide retention time range (3–4 min) to ensure that the lesions were completely collected while the unmodified nucleosides were excluded as much as possible. The collected fractions were dried in a Speed-vac, and re-dissolved in 10 µl of HO for LC-MS/MS analysis.
On-line HPLC–MS/MS measurements were carried out using an Agilent 1100 capillary HPLC pump (Agilent Technologies) and an LTQ linear ion-trap mass spectrometer (Thermo Fisher Scientific), which was set up for monitoring the fragmentation of the [M + H] ions of the labeled and unlabeled single-base or cross-link lesions. A 0.5 × 150 mm Zorbax SB-C18 column (particle size 5 µm, Agilent Technologies) was used for the separation of the DNA hydrolysis samples, and the flow rate was 6.0 µl/min. A 60-min gradient of 0–25% acetonitrile in 20 mM ammonium acetate was employed for the analysis of HPLC-enriched oxidative lesions.
We annealed separately the self-complementary ODNs 1 and 2 () to form duplexes and treated them with different concentrations of Cu(II)/HO/ascorbate (), digested the resulting DNA samples with enzymes, and analyzed the digestion mixtures by LC-MS/MS with stable isotope-labeled lesions as internal standards.
The selected-ion chromatogram (SIC) for the 569→275 transition, which monitors the loss of 2-deoxyribose/phosphate backbone, showed a fraction eluting at the same time as the labeled d(G[8-5m]mC) internal standard (a and b; the sample was from ODN2 treated under condition C listed in ). Similarly, the SIC for 569→275 transition, which showed a peak at the same retention time as the labeled d(mC[5m-8]G) at around 13 min, supported the formation of d(mC[5m-8]G) (c and d). The product-ion spectrum of the ion of 569 further revealed the presence of d(G[8-5m]mC) and d(mC[5m-8]G) in the enzymatic digestion mixture of the treated ODNs (). Moreover, the amounts of d(G[8-5m]mC) and d(mC[5m-8]G) are significantly lower in the control samples without hydrogen peroxide treatment, showing that the G[8-5m]mC and mC[5m-8]G intrastrand cross-link lesions can indeed be induced by Fenton-type reagents. In this regard, our previous studies revealed that G[8-5m]mC and mC[5m-8]G could be initiated from the 5-(2′-deoxycytidinyl)methyl radical (,), which can result from the ·OH-mediated hydrogen abstraction from the methyl group of mC ( and ) ().
We also observed the formation of d(GC) cross-link based on the peak at 20.8 min found in the SIC for the 555→457 transition (). This component co-elutes with the isotope-labeled internal standard of d(GC). Furthermore, the relative abundances of the two major fragment ions formed from the cleavage of the ion of 555 match those formed from the fragmentation of the [M + H] ion of the internal standard, further supporting the presence of d(GC) in the enzymatic digestion mixture of ODN1 after treatment with Fenton-type reagents (). It is worth noting that we attempted, but failed to detect the d(CG) in the enzymatic digestion mixture. The formation of G[8-5m]mC, mC[5m-8]G and GC intrastrand cross-link lesions exhibited linear dose-dependent increase when the concentrations of HO were up to 800 µM; however, the yield of G[8-5m]mC is ∼13 and 16 times greater than those of GC and mC[5m-8]G, respectively, upon treatment with the same dose of Cu(II)/HO/ascorbate (, and calibration curves for LC-MS/MS quantification are shown in Figure S12).
Previously we demonstrated that the yields of mC[5m-8]G and G[8-5m]mC are much higher under anaerobic than aerobic conditions, suggesting the competition between coupling with molecular oxygen and conjugating with its neighboring base (). Given that ·OH can also result in the formation of single-base lesions under the same oxidation conditions, it is important to examine the formation of these lesions. Not surprisingly, LC-MS/MS analysis also confirmed the formation of 5-FmdC and 5-HmdC in ODN2 (Figures S7 and S8) and 5-OHdC in ODN1 upon treatment with Fenton-type reagents (Figure S9). The quantification results showed that, at doses up to 800 µM HO and 100 µM Cu(II) (), the yields of the lesions increased proportionally with the rise in the concentrations of the Fenton-type reagents. Moreover, the yields of single-base lesions were 2–3 orders of magnitude higher than those of the cross-link lesions ().
Iron is another biologically important transition metal which can participate in Fenton-type reactions. To assess the different effects of iron and copper on inducing oxidative DNA lesions, we also examined DNA lesions produced in ODNs 1 and 2 upon treatment with Fe(II)/HO/ascorbate. It turned out that the yields of cross-link and single-base lesions were markedly lower when Cu(II) was replaced with Fe(II) under otherwise identical experimental conditions ( and ). In addition, the amounts of d(mC[5m-8]G), d(CG) and d(GC) induced in ODNs 1 and 2 were below the detection limits of our LC-MS/MS method. Furthermore, the yields of 5-HmdC, 5-FmdC and d(G[8-5m]mC) rise proportionally with the increase of the concentration of Fenton-type reagents; however, the amount of 5-OHdC exhibited a slightly different trend. The yields of 5-OHdC at low-dose range increased linearly with the rise in the concentrations of Fenton-type reagents. On the other hand, when the DNA was treated with the highest dose of Fe(II)/HO/ascorbate ([HO] = 800 µM), the yield of 5-OHdC increased considerably, which was 10-fold higher than the yield found at the second highest dose ([HO] = 400 µM, C). The exact reason for the disproportionally high yield found at the highest dose is not clear, though we suspect that the duplex DNA may undergo a conformational change at this high concentration of Fe(II).
It is known that the methylation in poly(GC) sequence can facilitate the conformational change of DNA, most notably the transition from normal B to Z-form DNA in the presence of divalent metal ions (). Therefore, we also treated two mixed-sequence ODNs (ODNs 3 and 4 in ), which contain both the GX and XG sites (X = C or mC), with Cu(II)/HO/ascorbate at the two highest doses. It turned out that the yields of oxidative lesions are comparable with what we found for ODN1 and ODN2 (). In addition, we quantified the yield of 8-oxo-7,8-dihydro-2′-deoxyguanosine (8-oxodG, Figures S10 and S11), an abundant and well-studied oxidative DNA lesion. Our results showed that the amount of 5-HmdC, 5-FmdC or 5-OHdC was 4–6-fold lower than that of 8-oxodG (a and b).
We also carried out several control experiments to gain insights into the mechanism for the formation of the oxidative lesions under Fenton-type reaction conditions. In the absence of transition metal ions, the yields of lesions were sharply reduced (). Leaving out ascorbate resulted in pronounced decrease in the yields of both single-base and intrastrand cross-link lesions (). However, treatment of duplex DNA with Fe(II)/HO results in the formation of higher levels of lesions than the corresponding treatment with Cu(II)/HO (). This observation is in keeping with the fact that Fe(II) can react directly with HO to generate ·OH, whereas Cu(II) has to be reduced by ascorbate to give Cu(I) before it can participate in Fenton-type reaction to afford ·OH.
Here we demonstrated, by using LC-MS/MS with the standard isotope dilution method, that the treatment of duplex ODNs with Cu(II)/HO/ascorbate can lead to the formation of GC, mC[5m-8]G and G[8-5m]mC. Quantification results revealed that the yields of GC and mC[5m-8]G cross-links (a and b) are comparable with what we reported for another intrastrand cross-link, G[8-5m]T, in calf thymus DNA treated with the similar doses of Cu(II)/HO/ascorbate (). In contrast, the CG lesion was below the detection limit of our LC-MS/MS method. The yield of G[8-5m]mC formed in methylated duplex ODNs is, however, 13-fold greater than that of GC produced in unmethylated ODNs under the same oxidation conditions (a and b). Together, these data support that, upon treatment with Fenton-type reagents, the formation of intrastrand cross-link between guanine and its neighboring mC occurs at a much greater efficiency than that between guanine and its neighboring unmethylated cytosine. Our results also revealed that Cu(II)/HO/ascorbate system is ∼10 times as effective as Fe(II)/HO/ascorbate system in inducing these cross-link lesions.
Previous studies demonstrated that GC, G[8-5m]T and G[8-5m]mC cross-link lesions reduced the stability of duplex DNA (,), and the former two lesions blocked DNA synthesis by replicative DNA polymerases (,). Moreover, GC could be bypassed by yeast pol , a translesion synthesis DNA polymerase, with reduced efficiency and fidelity of nucleotide incorporation at the site opposite the 5′ guanine moiety of the lesion (). Recent repair studies revealed that both GC and G[8-5m]mC could be recognized and incised by UvrABC nucleases (), suggesting that oxidative intrastrand lesions might be substrates for NER enzymes . Furthermore, we found that G[8-5m]mC-bearing substrates could be incised by UvrABC nuclease with lower efficiency than the corresponding GC-containing duplex (). The poorer repair and more efficient formation of intrastrand cross-link lesion at GmC than at GC site may lead to the accumulation of such lesions and contribute to increased mutation rate at methylated CpG sites. The formation of the intrastrand cross-link lesions, proposed to be substrates for NER pathway (,,), may pose a risk for people with diseases associated with defects in NER, e.g. (XP) (). Future studies on intrastrand cross-link lesions reported here may contribute to a better understanding of pathological symptoms associated with NER-deficient diseases.
Pfeifer . () recently found an unusually high frequency of mCG→TT tandem double mutation when Cu(II)/HO/ascorbate-treated, CpG-methylated pSP189 shuttle vector was replicated in NER-deficient XPA cells. The corresponding CG→TT mutation occurred at a much lower frequency when the corresponding unmethylated vector was treated with Cu(II)/HO/ascorbate and propagated in the same cell line. These authors proposed that vicinal or cross-linked base damage originated from mC and its neighboring guanine might be involved (). Quantification data presented in this study support that intrastrand cross-link lesion formed between mC and its adjacent guanine might contribute to such tandem base substitutions since the mC[5m-8]G cross-link lesion was observed, whereas the CG cross-link was not detectable. Along this line, if a guanine is situated on the 5′ side of mCpG sites, the yield of G[8-5m]mC cross-link lesion is dramatically enhanced, by more than 10-fold, compared with corresponding GC induced at the unmethylated CpG sites.
For comparison, we also quantified the single-base lesions induced by the Fenton-type reagents under the same experimental conditions. It turned out that the yields for these single-base lesions were 2–3 orders of magnitude greater than those of intrastrand cross-link lesions formed at GC, mCG or GmC site (). These results indicate that the addition of molecular oxygen to pyrimidine radicals is much more facile than the coupling of the pyrimidine radicals to their adjacent purine bases.
Quantification results also showed that 5-OHdC is induced at a similar efficiency as 5-HmdC and 5-FmdC regardless of the sequences of ODNs and the types of transition metal ions being Cu(II) or Fe(II), whereas 5-hydroxy-2′-deoxyuridine (5-OHdU) is below the detection limit of our LC-MS/MS method, particularly at lower doses. The hydroxyl radical couples preferentially to the C5=C6 double bond in cytosine, which can lead to the formation of cytosine glycol. Cytosine glycol can undergo facile dehydration to yield 5-OHdC or undergo both deamination and dehydration to give 5-OHdU () ().
Similar addition of hydroxyl radical to the C5=C6 double bond of mC can lead to the formation of mC glycol, which can deaminate to give thymine glycol (,). Because of the lack of isotope-labeled internal standards of thymidine glycol, the formation of this lesion from the mC-bearing strand was not quantified in the present study. In addition, the hydroxyl radical can abstract a hydrogen atom from the methyl group of mC to give the methyl radical of the pyrimidine base, which can further transform to give 5-HmdC and 5-FmdC under aerobic conditions (). With the above quantification results, it is reasonable to speculate that the total amount of hydroxyl radical-induced major single-base lesions from mC can be several fold higher than those formed from unmethylated cytosine. The more facile formation of single-base lesion at mC than at unmethylated cytosine may account for the prevalent C→T transition mutation found at CpG site in human gene, and mutation in is a hallmark for many types of human tumors (). In this context, both 5-OHdC and 5-FmdC were found to be mutagenic ; a mutation frequency of 0.05% was reported for the former after 5-OHdC-bearing single-stranded M13 genome was propagated in cells (), and a mutation frequency of 0.03–0.28% was found for the latter after the 5-FmdC-carrying double-stranded shuttle vectors were replicated in simian COS-7 cells (). To our knowledge, the mutagenic properties of 5-HmdC have not yet been examined.
The methylation pattern of cytosine residues at CpG sites and post-translational modifications of histones are crucial for maintaining the epigenetic code in human cells (). The modification at CpG sequence can result in disturbed gene regulation and heritable epigenetic changes in chromatin. Recent studies revealed that the presence of 8-oxodG or 5-HmdC at CpG site reduced significantly the binding of DNA to methyl-CpG-binding proteins by at least 10-fold (). These lesions also alter the site selectivity of cytosine methylation at CpG site induced by human maintenance DNA methyltransferase DNMT1 (,). High yield of 8-oxodG, 5-HmdC and 5-FmdC observed in our experiment () may indicate that DNA damage-mediated alteration in methylation pattern and subsequent binding by methyl-CpG-binding proteins can be significant .
The intracellular concentrations of iron and copper ions can be dramatically increased under oxidative stress conditions through their release from the iron- or copper-storage proteins (,). In humans, genetic hemochromatosis and Wilson's disease cause abnormal accumulation of iron and copper, respectively, in various organs. An accumulation of highly mutagenic oxidative lesions in genomic DNA has been considered to be relevant to iron-induced carcinogenesis in iron-overload diseases (,). Bulky DNA lesions were found in the liver of patients with Wilson's disease and primary hemochromatosis (). Our current results support the argument that transition metal ions, especially copper, can induce significant amount of DNA damage, including bulky intrastrand cross-link lesions, in the presence of antioxidant such as ascorbate, which reduces the high-valence Cu/Fe so as to generate hydroxyl radical via the Fenton-type reaction.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
In the past decade, increased availability of large amounts of life sciences data, including low-throughput data accumulated by generations of scientists over a half-century, and high-throughput data acquired through newly developed biotechnologies, has coincided with great advances in data analysis and modeling techniques, most notably in the machine-learning area, leading to an increase in computational prediction programs in various important domains in life sciences research. In more and more problem domains, multiple prediction programs have emerged from independent efforts by different groups. These programs differ by what data features they use, or by what the methods or algorithms they apply in the classification tasks, or by both. These prediction programs may be complementary; i.e. one program performs better for one type of data under one set of circumstances, but another prediction program performs better for another type of data or under other circumstances. By proper exploitation of the combined strengths of these prediction programs, it may be possible to construct whose performance surpasses that of all existing prediction programs.
The meta-prediction problem is one that seeks to construct a prediction program (termed a ), which makes predictions by organizing and processing the prediction results of a number of other prediction problems (termed ). The meta-predictor takes the output of element predictors as its sole input. No explicit attention is paid to the feature definition or the underlying classification algorithms of the individual element predictors. Rather, the strengths and weaknesses of each element predictor, and the similarities and differences between different element predictors are visible to the meta-predictor only through the prediction results they make. The hope of meta-prediction is to develop a meta-predictor, which can combine the strengths of the element predictors and produce more accurate predictions than any of the element predictors. In this study, we focus on the meta-prediction of subcellular localization of proteins.
Subcellular localization is a key functional characteristic of eukaryotic proteins. Most proteins must be localized to the correct subcellular compartment or organelle in order to properly execute their biological function(s). Cooperating proteins must be present in the same location in order for them to interact. Since Nakai and Kanehisa's pioneering work (), a large number of computational prediction programs have been developed in this field [see recent reviews, e.g. (,)]. These programs use many different data features, such as N-terminal signal sequence information [TargetP (), PSORT () and iPSORT ()]; amino acid composition [NNPSL (), PLOC (), FUZZY_LOC () and SubLoc ()]; evolutionary information obtained by multiple sequence alignment or PSI-BLAST, and/or calculated physicochemical properties of the proteins [Proteome Analyst (), LOCSVMPSI (), ESLpred () and LOCtree ()]; 3D structural data [LOC3d ()]; or even gene expression data (). They also use many different classification methods, such as expert systems (PSORT, iPSORT); artificial neural networks (ANN) (LOCnet, LOC3d, TargetP); -nearest neighbor (-NN) [PSORT II ()]; Naive Bayes (NB) classifier (Proteome Analyst); fuzzy -NN (FUZZY_LOC); or support vector machines (SVM) (SubLoc, LOCSVMPSI, PLOC, ELSpred and LOCtree).
These different data features and classification methods may give these prediction programs different, complementary strengths. In this study, we develop meta-predictors that harness the combined strengths of these individual element predictors. We first compiled an unbiased subcellular localization dataset that does not overlap with any data used in the development of these predictors; we then examined the performance of these predictors using this unbiased dataset; and explored several voting-based strategies for constructing meta-predictors. We show that, using a simple reduced voting strategy, an excellent meta-predictor can be developed, with a predicting performance substantially exceeding that of all element predictors, and that this meta-predictor's excellent performance persists with data not used in its development.
In this study, we focus on the subcellular localization predictions of animal proteins. The unbiased protein subcellular localization dataset MetaSCL06 was compiled from Swiss-Prot Release 50.2 (12 July 2006). The compilation procedure consisted of four steps: assembling an unbiased set of proteins, assigning class labels to the proteins based on gene ontology (GO) annotations in Swiss-Prot entries, assigning class labels to the proteins based on the comment field in Swiss-Prot entries, and manual reconciliation of protein sets from steps 2 and 3 ().
The MetaSCL07 dataset is a validation set that was not used in meta-predictor development. This dataset was compiled from Swiss-Prot Release 51.6 (6 February 2007), with the same procedure used in the compilation of MetaSCL06. All entries of proteins bearing an initial entry date on or before 12 July 2006 (date of Release 50.2) were removed. This dataset includes 579 proteins (145 nuclear, 50 cytoplasmic, 144 mitochondrial and 240 extracellular). This dataset is available as Supplementary Table 4.
The final list of prediction programs and the chosen element predictors are shown in . A total of 12 element predictors were chosen, derived from eight prediction programs. Each program is discussed below:
(). ELSpred uses the one-versus-the-rest SVM as the underlying classification method. It makes predictions into the four common subcellular localization compartments: nuclear, cytoplasmic, mitochondrial and extracellular. ELSpred provides five prediction options, each corresponding to a different feature formulation scheme: ELSpred_comp uses the compositions of the 20 amino acids as its features. ELSpred_physicochemical uses 33 physicochemical properties as its features. ELSpred_dipeptide defines features using dipeptide compositions. The features used in ELSpred_EuPSI are constructed following three iterations of EuPSI-BLAST through which the similarity between the protein and 2427 eukaryotic proteins is obtained. ELSpred_hybrid, uses a feature scheme that combines all the above four feature schemes. These five prediction options are considered as different element predictors in this study.
(). LOCtree uses amino acid compositions of the proteins as its features. It goes through a three-level binary tree-structured process with a binary SVM model working at each node in the tree. In addition to the four common subcellular localization compartments, LOCtree also makes predictions about whether a protein is an organelle protein, and about whether a nuclear protein is a DNA-binding protein.
(). PLOC uses five different types of compositions (amino acids, amino acid pairs, one gapped amino acid pairs, two gapped amino acid pairs and three gapped amino acid pairs) as its features. The predictions are made by one-versus-the-rest SVMs followed by voting. In addition to the four common subcellular localization compartments, PLOC also makes predictions into six other subcellular localization compartments: cytoskeleton, endoplasmic reticulum, the Golgi apparatus, lysosome, peroxisome and plasma membrane.
(). Proteome Analyst adopts features calculated with PSI-BLAST against the Swiss-Prot database, and employs a Naïve Bayes (NB) algorithm for making predictions. Besides the four common subcellular compartments, Proteome Analyst also makes predictions into five additional subcellular compartments: endoplasmic reticulum, Golgi apparatus, lysosome, peroxisome and plasma membrane.
and (). PSORT and PSORT II utilize a large number of features, including the presence of N-terminal sorting signals, the presence of RNA/DNA-binding motifs, amino acid compositions and some calculated structural information. PSORT is a knowledge-based system with a set of ‘if-then’ rules. PSORT II employs a -NN learning algorithm. Besides the four common subcellular compartments, PSORT also makes predictions into endoplasmic reticulum, the Golgi apparatus, lysosome, microbody, plasma membrane, and PSORT II makes predictions into cytoskeleton, endoplasmic reticulum, the Golgi apparatus, plasma membrane, peroxisome and secretary vesicles.
(). SubLoc uses amino acid compositions as its features. Four one-versus-the-rest SVMs are trained to make predictions on a given protein into one of the four common subcellular localizations.
(). WoLF PSORT defines features using amino acid compositions and N-terminal signals that are encoded by AAindex, and also adopts some PSORT features. A -NN classifier is trained following the WoLF feature selection and weighting procedure. WoLF PSORT makes predictions into cytoskeleton, endoplasmic reticulum, the Golgi apparatus, lysosome, peroxisome and plasma membrane in addition to the four common subcellular compartments.
Prediction jobs were submitted to each of the element prediction programs with the protein sequences in the MetaSCL06 and MetaSCL07 datasets. Some of the element prediction servers, ELSpred, PSORT, PSORT II and PLOC, do not provide a batch-processing option. For these prediction servers, simple Java programs were developed and used to handle the job submission and result retrieval. Other Java programs were developed to parse and analyze the prediction results returned from the element prediction servers. Some prediction programs, including PLOC, ELSpred_EuPSI and ELSpred_hybrid, provide in their output the most likely subcellular compartment, but other prediction programs generate numerical scores in their output, for example, the ‘reliability indices’ produced by LOCtree and SubLoc, the ‘certainty score’ produced by ELSpred_comp, ELSpred_physicochemical and ELSpred_dipeptide, and the percentage scores produced by Proteome Analyst and PSORT II. When multiple compartments appeared in the output with numerical scores, the one with the highest value was picked as the predicted compartment. Two of the prediction servers (PLOC and WoLF PSORT) can predict two compartments for a single protein both with the highest scores (e.g. ‘nuclear’ and ‘cytoplasmic’). In these cases, both predictions were considered valid and they were given equal weights when the predicting performance of the prediction program was evaluated.
For a two-class classification problem, commonly used performance measures include sensitivity, specificity, accuracy and Matthew's correlation coefficient (MCC) ().
For a multi-class classification problem, the definitions of sensitivity, specificity and MCC are no longer valid, but that of accuracy continues to be useful. In addition, Gorodkin () defined a correlation coefficient formula, which we will call Gorodkin correlation coefficient (GCC), a measure of predicting performance for a multi-class classifier. GCC calculates the correlation between two × matrices, the observation matrix and the prediction matrix, where is the number of samples, and is the number of classes. In the protein subcellular localization prediction problem, is equal to four for meta-predictors and the element predictors making predictions into the four common subcellular compartments only (e.g. SubLoc). For element predictors that make predictions into the ‘other compartment’ class (e.g. LOCtree and PLOC), is equal to 5. An element in the observation matrix, , is set to be 1 if the th sample is known to belong to class , and it is set to be 0 if otherwise. An element in the prediction matrix, , is set to be 1 if the th sample is predicted to belong to class by the predictor, and it is set to be 0 if otherwise. For PLOC and WoLF PSORT, if a sample is predicted into two equally probable compartments, the corresponding elements in the prediction matrix are set to be 1/2 for both compartments. If no prediction is made for a sample by an element predictor, all corresponding elements in the prediction matrix are set to be 0.
Given the observation matrix and the prediction matrix GCC is defined as follows:
The element predictors are not completely compatible with one another in the types of predictions they make. First, 6 of the 12 element predictors make predictions for other subcellular compartments than the four we choose to use (). For simplicity, we lump all other subcellular compartments together for each of these programs and call them ‘other compartments’. Since the gold standard dataset (MetaSCL06) contains data for the four chosen compartments only, any predictions made into the ‘other compartments’ class by any element predictors are classified as wrong predictions. Second, some element predictors do not make predictions for all proteins. For instance, Proteome Analyst made predictions on 1523 of the 1693 proteins in the MetaSCL06 dataset, and ELSpred_EuPSI made predictions on only 624 of the 1693 proteins in the dataset. For the sake of making a fair performance comparison, we classed these ‘no prediction’ cases as wrong predictions when calculating the accuracy of these element predictors. However, we adopt an additional performance measure to assess the performance of an element predictor for the proportion of proteins in the dataset where predictions are actually made. We term this accuracy measure the ‘relative accuracy’, and calculate it as the ratio of the number of corrected predicted proteins and the number of proteins for which predictions are made. Finally, for some proteins, some element predictors (e.g. PLOC and WoLF PSORT) output two predicted subcellular compartments, which are considered equally probable by the predictors. For these samples, the predictions are considered ‘half correct’ if one of the two compartment output matches the true compartment label for the protein in the dataset.
(,) describes the prediction made by the th ( = 1, … 12) element predictor: (,) = 1 if the predictor made by the th element predictor is , and (,) = 0 if otherwise. If, for a given protein, two equally probable compartments are output by an element predictor (PLOC or WoLF PSORT), the score is split so that (,) = 1/2 for both compartments. The notation ‘arg max’ stands for the ‘argument of the maximum’, and it returns the value of that leads to the highest value of the formula that follows (in this case, ). That is, for each input protein, the unweighted voting meta-predictor sums the number of element predictors that make positive predictions for each of the four subcellular localization compartments, then picks the compartment with the largest number. When there are two or more compartments with the highest score, one compartment is picked at random.
The weighted voting strategy differs from the unweighted voting strategy in that the predictions made by element predictors are multiplied by a weight, which varies among predictors, before being summed up to produce the prediction of the meta-predictors.
Although the prediction results of all element predictors are available to the meta-predictors, it is not necessary for all of them to be used. Indeed, if we exclude from consideration some of the element predictors that do not perform well, it may be possible to obtain meta-predictors with further improved performance. Thus, we applied the so-called ‘reduced voting strategy’: starting from a full (or ‘unreduced’) meta-predictor, we iteratively reduce the number of element predictors included in the construction of meta-predictor, by picking the next element predictor with the lowest performance, and setting its weight to 0. This process continues until only one element predictor remains in consideration. There are three performance measures used in evaluating the element predictors—accuracy, reduced accuracy and GCC. Therefore, for each (unreduced) meta-predictor, there are three different ways in which the reduction can be done. They are named accuracy-guided reduction (or AG), relative accuracy-guided reduction (or RAG), and GCC-guided reduction (GG), respectively. In each of these reduction methods, the lowest scoring element predictors are excluded one by one, producing a series of reduced voting meta-predictors.
summarizes the predicting performance of the 12 element predictors assessed on the MetaSCL06 dataset. The predictions made by the element predictors vary considerably with one another. Proteome Analyst (accuracy: 0.821, GCC: 0.811) offers the best performance among all predictors, followed by LOCtree (accuracy: 0.746, GCC: 0.663) and WoLF PSORT (accuracy: 0.733, GCC: 0.635). Some predictors from the ELSpred program (ELSpred_hybrid and ELSpred_dipeptide) are ranked among the lowest in predicting performance. Proteome Analyst is also the element predictor that offers the highest relative accuracy (0.913), followed by ELSpred_EuPSI (0.880) and LOCtree (0.766).
With the performance of every element predictor assessed using the unbiased MetaSCL06 dataset, we set out to explore strategies to construct meta-predictors on top of these element predictors. First, we attempted a simple unweighted voting strategy. The meta-predictor constructed using the unweighted voting strategy (accuracy: 0.754, GCC: 0.651) offers better predicting performance than the average performance of the 12 element predictors (accuracy: 0.578, GCC: 0.459), but it does not reach the performance of the most accurate element predictor (Proteome Analyst, accuracy: 0.821, GCC: 0.811) ().
As is shown in , improved performance is achieved in these weighted voting meta-predictors. All three weighted voting meta-predictors show accuracy values that approach or slightly exceed that of the most accurate element predictor, Proteome Analyst. However, using GCC, none of these meta-predictors have reached the level of Proteome Analyst (GCC: 0.811).
The four voting schemes (unweighted voting and three weighted voting schemes) are combined with the three reduction methods [accuracy-guided reduction (or AG), relative accuracy-guided reduction (or RAG) and GCC-guided reduction (GG)], giving rise to a total of 12 series of reduced voting meta-predictors. In each of these predictor series, the predicting performance (measured in accuracy or in GCC) shows a biphasic relationship with the number of excluded element predictors (). When the number of excluded element predictors is small, the predicting performance increases with the number of excluded element predictors, agreeing well with our conjecture that excluding badly performed element predictors may lead to improved predicting performance of the meta-predictors. The predicting performance reaches a peak when about 6–9 element predictors are expelled, then declines as more element predictors are excluded. Apparently, following this critical point, further removing of the more accurate element predictors is detrimental to the predicting performance of the resultant meta-predictor.
As is shown in , the best predictor in each of the reduced voting meta-predictor series demonstrates better predicting performance than the best performed element predictor (Proteome Analyst) in both accuracy and GCC. Most of these best reduced meta-predictors show significantly higher GCC than that of Proteome Analyst in Fisher's Z-transformation test. The meta-predictor with the best performance was found to be the relative accuracy weighted, reduced by relative accuracy guiding, with six element predictors excluded (denoted as RAW-RAG-6). This meta-predictor makes predictions based on the predictions made by six element predictors: ELSpred_PhysicoChemical, ELSpred_EuPSI, LOCtree, Proteome Analyst, PSORT II and WoLF PSORT (). RAW-RAG-6 reaches a remarkable accuracy of 0.902, a nearly 8% improvement over Proteome Analyst (A: 0.821); and a GCC of 0.856, significantly higher than that of Proteome Analyst (GCC: 0.811), the best element predictor examined ( = 8.2 × 10, Fisher's Z-transformation test).
Element predictor performance was evaluated on data not used in their development. To impose this same limit on RAW-RAG-6, the element predictors and RAW-RAG-6 were evaluated using the MetaSCL07 dataset, containing data not used in RAW-RAG-6 development (). Proteome Analyst remains the element predictor with the best predicting performance based on GCC (0.783), though LOCtree offers better accuracy (0.829) than Proteome Analyst (0.775) with the MetaSCL07 dataset. The superior performance offered by RAW-RAG-6 persists with this dataset, with an accuracy of 0.888 and GCC of 0.840, significantly better than those of any element predictors.
RAW-RAG-6 outperforms each of the 12 element predictors in accuracy and MCC for all four two-class classification problems (). Comparing with the element predictor with the best performance (Proteome Analyst), the biggest improvement was achieved for the extracellular compartment, with 2.7% increase in accuracy (from 0.956 to 0.983) and 5.7% increase in MCC (from 0.909 to 0.966, < 2 × 10, Fisher's Z-transformation test). It is followed by the nuclear compartment, for which a 2.1% increase in accuracy (from 0.908 to 0.929) and 4.6% increase in MCC (from 0.801 to 0.847, = 1.4 × 10, Fisher's Z-transformation test) are achieved. The smallest improvement is found for the cytoplasmic compartment, where 0.7% increase in accuracy (from 0.922 to 0.929), and 1.3% improvement in MCC (from 0.617 to 0.630, = 0.27, Fisher's Z-transformation test) are observed. Overall, the RAW-RAG-6 meta-predictor achieves remarkable performance in these two-class classification problems, and consistently outperforms every element predictor in identifying proteins localized in each of the four subcellular compartments.
In many life science domains, several prediction programs have emerged that often have different strengths due to different types of data (or different aspects of the same data) used, and/or different classification methods adopted in their development. When more than one of these is used on the same data, they may produce conflicting predictions. Users are often confused and frustrated by such conflicting results, because they may lack the knowledge to make a sensible choice among them. If a meta-predictor can be developed with predicting performance exceeding that of any individual element predictors, it may resolve this quandary.
Meta-predictors cannot replace element predictors. Rather, they are enhancements. Meta-predictors are constructed from element predictors, and their performance depends on accurate predictions made by element predictors. Without good element predictors, it is not possible for good meta-predictors to be developed. In addition, meta-predictors (in particular, voting-based meta-predictors) are effective only within the scope of the prediction problem that is common to multiple element predictors. Often, element predictors make unique predictions. For example, among the prediction programs discussed in this study, only PSORT II makes predictions about protein localization to secretory vesicles. For unique predictions, one has to rely on an element predictor.
The linear voting strategies explored in this study are related to several well-known online learning algorithms, including Littlestone and Warmuth's weighted majority (WM) algorithm () and Freund and Schapire's Hedge algorithm (). Those algorithms are applied to situations where one person is trying to make predictions based on the opinions of several ‘experts’ from whom he seeks advice. If the weights are properly chosen, there are theoretical bounds of the maximal number of wrong predictions made by the ‘master predictor’, i.e. the performance of the ‘master predictor’ will not be ‘too much worse’ than that of the predictions made by the best ‘expert’. The meta-prediction problem discussed in this study differs from those previous studies in that ‘batch learning’, rather than ‘online learning’, applies. In other words, the training samples are assumed to be provided together, instead of one at a time. In the same paper in which the Hedge algorithm was discussed (), Freund and Schapire introduced the well-known Adaboost algorithm, which applies to batch learning. The major difference between meta-prediction problem discussed in this study and Adaboost and other ensemble learning algorithms [e.g. Logitboost () and Bagging ()] is that in ensemble learning, the ‘element predictors’ are results of an identical training algorithm applied to different samplings of the training data. In meta-prediction, the ‘element predictors’ are assumed to be known and unchanged, and all training data is used for all element predictors.
e
s
u
c
c
e
s
s
f
u
l
d
e
v
e
l
o
p
m
e
n
t
o
f
R
A
W
-
R
A
G
-
6
d
e
m
o
n
s
t
r
a
t
e
s
t
h
e
e
f
f
e
c
t
i
v
e
n
e
s
s
o
f
v
o
t
i
n
g
-
b
a
s
e
s
t
r
a
t
e
g
i
e
s
i
n
t
h
e
m
e
t
a
-
p
r
e
d
i
c
t
i
o
n
p
r
o
b
l
e
m
s
.
P
r
o
p
e
r
e
m
p
l
o
y
m
e
n
t
o
f
v
o
t
i
n
g
-
b
a
s
e
d
s
t
r
a
t
e
g
i
e
s
i
s
l
i
k
e
l
y
t
o
l
e
a
d
t
o
g
o
o
d
m
e
t
a
-
p
r
e
d
i
c
t
o
r
s
i
n
o
t
h
e
r
l
i
f
e
s
c
i
e
n
c
e
s
p
r
o
b
l
e
m
d
o
m
a
i
n
s
.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
The knowledge of the distribution of excess electron sites for DNA single strands is attracting increasing attention. (,) The formation of anions in DNA fragments has been found to be related to important biochemical processes such as DNA damage and repair (), charge transfer along DNA (), and the initiation of reactions leading to mutation (,,).
Experimentally based investigations suggest that the nucleobases have small electron affinities (EAs), ∼0.1 eV for thymine (T), cytosine (C) and uracil (U) (). Negative EA values have been determined for adenine (A) and cytosine in gas-phase experiments (,). The identification of the existence of two types of anions (dipole bound and covalent) () partly explains the differences among the different experimental EA values (). Recent experiments on the electron-capturing efficiencies of short DNA oligomers provide the only estimate of the relative order of the vertical attachment energies (VAEs) for DNA single strands (). However, the direct experimental determination of the EAs of nucleosides and nucleotides has proven difficult. Large DNA fragments, such as nucleotides, are non-volatile and the requirement for the vaporization of the species without thermal degradation makes it difficult to carry out reliable experimental studies in the gas phase.
Theoretical investigations at various levels of sophistication have complemented the experimental studies. While second-order Møller–Plesset perturbation theory (MP2) underestimates the adiabatic electron affinities (AEAs) for the five bases (,), density functional theory (DFT) approaches yield experimentally consistent AEA values for the individual bases (). Theoretical studies of excess charge in DNA have been extended to the prediction of the EAs of subunits such as nucleosides and nucleotides. Semi-empirical methods have been applied to evaluate the EAs of nucleosides and nucleotides. (,) With the reliably calibrated B3LYP/DZP++ approach (), meaningful predictions of the EAs of the 2′-deoxyribonucleosides have been completed (). Recently, the EAs of the pyrimidine nucleotides (3′-dCMP, 3′-dTMP, 5′-dCMP, 5′-dTMP) have also been predicted at the B3LYP/DZP++ level of theory (,).
To understand the effects of excess electrons on single-strand DNA, the backbone of DNA should be realistically simulated in the model study. Previous studies reveal that the phosphate group at either the 3′ or 5′ position increases the electron acquisition ability of the pyrimidines (,,). Thus, properly modeled systems representing single-strand DNA should include the phosphate group at both the 3′ and 5′ positions of the nucleotides. Moreover, the knowledge of the distribution of the excess electron sites and the availability of reliable EAs for the purine nucleotides are of equal importance. The vertical attachment of electrons to guanine-rich DNA single strands appears to predominate over entrapment by the cytosine-rich strands, as reported in the important 2005 experiment by Ray (). Although previous studies of the pyrimidine nucleosides and nucleotides suggest that the excess electron is mainly located on the bases in the electron-attached radical anions, the small AEA values and the large dipole moments of the purine bases (,) suggest that the situation for the purine nucleotides might be different. Here, we report a theoretical investigation of electron attachment to reasonable models of DNA single strands. The 2′-deoxyguanosine-3′,5′-diphosphate (3′,5′-dGDP), 2′-deoxyadenosine-3′,5′-diphosphate (3′,5′-dADP), 2′-deoxycytidine-3′,5′-diphosphate (3′,5′-dCDP) and 2′-deoxythymidine-3′,5′-diphosphate (3′,5′-dTDP) systems in their protonated forms have been selected as models (see ). For a better description of the influence of the 3′–5′ phosphodiester linkage in DNA, the -OPOH moiety at the 5′ position was terminated with a methyl group. These systems represent the most complete descriptions to date of the minimal DNA subunits and are expected to provide reliable information concerning electron attachment to single-strand DNA.
The B3LYP functional approach (,) with basis sets of double-ζ quality plus polarization and diffuse functions (denoted DZP++) was used to obtain optimized geometries, vibrationally zero-point corrected energies and natural charges for the model molecules in both neutral and anionic forms. The DZP++ basis sets were constructed by augmenting the Huzinaga–Dunning (,) set of contracted double-ζ Gaussian functions with one set of -type polarization functions for each H atom and one set of five -type polarization functions for each C, N, O and P atom (α(H) = 0.75, α(C) = 0.75, α(N) = 0.80, α(O) = 0.85, α(P) = 0.60). To complete the DZP++ basis, one even-tempered diffuse function was added to each H atom, while sets of even-tempered diffuse and functions were centered on each heavy atom. The even-tempered orbital exponents were determined according to the prescription of Lee and Schaefer (). Each adiabatic electron affinity was computed as the difference between the absolute energies of the appropriate neutral and anion species at their respective optimized geometries AEA = − .
To evaluate the electron-capture abilities of DNA single strands in aqueous solution, the polarizable continuum model (PCM) () with a dielectric constant of water (ε = 78.39) was used to simulate the solvated environment of an aqueous solution. To analyze the distributions of each unpaired electron, molecular orbital and the spin-density plots were constructed from the corresponding B3LYP/DZP++ densities. Natural population analysis (NPA) was determined using the B3LYP functional and the DZP++ basis set with the Natural bond orbital (NBO) analysis of Reed and Weinhold (,). The GAUSSIAN 98 system of DFT programs () was used for all computations.
The positive EAs () of the model systems suggest that the DNA single strands have a tendency to capture low-energy electrons and to form electronically stable radical anions.
To understand the electron-capturing ability of DNA single strands at the nascent stage of electron attachment, it is necessary to estimate the vertical electron attachment energies (VEAs). Relatively large VEAs are predicted for 3′,5′-dTDP (0.17 eV) and 3′,5′-dGDP (0.14 eV) in this investigation. This indicates that both the thymine-rich and the guanine-rich DNA single strands are prepared to capture low-energy electrons. Conversely, the near zero VEA values for 3′,5′-dCDP (0.03 eV) and 3′,5′-dADP (0.02 eV) suggest less effective electron capturing abilities for cytosine and adenine-derived DNA single strands. It must be noted that the high electron-capturing ability of guanine-rich DNA single strands and the low electron-capturing ability for cytosine-rich and adenine-rich DNA single strands have been observed in the experiments of Ray (). Although the VEA value is not the sole factor related to the electron-capture ability, our predictions for the VEAs of the nucleoside-3′,5′-diphosphate systems allow us to understand these experimental observations from the viewpoint of energy.
The AEAs of the nucleotides follow the order 3′,5′-dTDP > 3′,5′-dCDP > 3′,5′-dGDP > 3′,5′-dADP, which is consistent with that for the nucleobases and the nucleosides. However, substantial increases in the AEAs are predicted compared to those for the corresponding nucleic acid bases (0.52 eV versus 0.20 eV for T, 0.44 eV versus 0.03 eV for C, 0.36 eV versus −0.07 eV for G and 0.22 eV versus −0.28 eV for A, respectively; see ). In addition, the increase in the AEAs from the nucleosides to the corresponding nucleotides amounts to ∼0.08–0.27 eV, signifying the importance of the phosphate groups in the stabilization of the radical anions of the DNA components. It is interesting to note that the AEA of 3′,5′-dTDP is almost the same as that of 3′-dTMP (2′-deoxythymidine-3′-monophosphate). The influence of the OH group at the 5′ position on the AEA of 3′-dTMP is equivalent to that of the 5′-phosphate group in 3′,5′-dTDP. The similar AEA value of 3′,5′-dCDP and that of 3′-dCMP (2′-deoxycytidine-3′-monophosphate) suggest the same equivalency.
To evaluate the electronic stability of the radical anions, the vertical detachment energies of the anions have been predicted. The VDEs of the radical anions of the purines are found to be about 0.3 eV, significantly smaller than those of the radical anions of the pyrimidines (∼0.7 eV). Consequently, reactions with activation barriers higher than 0.3 eV (∼7 kcal/mol), such as that for N-glycosidic bond breakage (), are not expected for the radical anions of 3′,5′-dGDP and 3′,5′-dADP. As will be shown below, the low VDE values for 3′,5′-dGDP and 3′,5′-dADP may be traced to the nature of the phosphate group anion, while the high VDE values for 3′,5′-dTDP and 3′,5′-dCDP are concordant with the characteristics of the base-centered anions.
Examination of the molecular orbital that the ‘last’ electron occupies provides an electronic structure-based rationale for the electron-attracting capabilities of the nucleotides. Plots of the singly occupied molecular orbitals (SOMOs) for the radical anions are shown in . The most striking feature revealed by the SOMOs of these nucleoside-3′,5′-diphosphates molecules is that the excess electron density resides in part in the vicinity of the phosphate moiety, at the 3′ position. This phenomenon has not been found for either the pyrimidine nucleoside-3′-monophophate radical anions (,) or the corresponding 5′-monophophate radical anions (). Previous investigations have shown that the excess electron density is not located around the -POH moiety of the stable radical anions of the pyrimidine nucleoside monophosphate (,,). The present results seem to suggest that introducing a phosphate group at the 3′ or 5′ position of the ribose improves the electron-capturing ability of the 3′-phosphate fragment. The partial electron distribution on the -POH moiety might have its origin in the relatively low VDEs of the nucleoside–diphosphate complexes. However, guanine is different, in that the excess electron is partly covalent bonded to the 3′-phosphate in 3′,5′-dGDP while it appears to be dipole-bound in the nucleoside (dG) (). The VDE of 3′,5′-dGDP is thus significantly higher than that of dG (0.32 eV versus 0.05 eV). Examination of the SOMOs depicted in enables one to conclude that the excess electron largely resides in the vicinity of the phosphate moiety in the purine radical anions, while it is mainly located near the nucleobase unit in the pyrimidines. Recall that the AEA of the sugar-phosphate-sugar model is close to zero (∼0.00–0.03 eV) (). In this light, the difference in electron distributions between the purines and the pyrimidines is expected.
It should be noted that the excess electron at the base fragment in 3′,5′-dADP, 3′,5′-dCDP and 3′,5′-dTDP is valence-bound (the π* anti-bonding orbital of the base is partly occupied, as shown in the SOMOs), while it is typically dipole-bound in the guanine moiety in 3′,5′-dGDP. This feature can also be seen in electron attachment to the nucleosides (). Compared to the nucleosides, the net result of the introduction of phosphate groups at the 3′ and 5′ positions is to reduce the population near the base moiety by increasing the electron density around the phosphate groups.
The location of negative charge on the constituent parts of the nucleoside pair also provides some insight into the overall electronic effect of the charge. summarizes the charge distributions among the bases, ribose and phosphates for both the neutral and anionic complexes. The analysis of the NPA charge differences between the neutral and anionic nucleotides supports the conclusion that the excess electron mainly resides in the vicinity of the nucleobase moiety in the pyrimidine radical anions, while it is largely located at the 3′-phosphate group in the purine. The NPA charge differences suggest that there is 0.63 a.u. of negative charge located on the thymine and 0.29 a.u. on the 3′-phosphate for 3′,5′-dTDP; with 0.63 a.u. of negative charge located on the cytosine and 0.21 a.u. on the 3′-phosphate for 3′,5′-dCDP. Conversely, the charge distribution differences of the purine nucleotides show significant increases in the negative charge resident on the 3′-phosphate group (0.46 a.u. for both 3′,5′-dADP and 3′,5′-dGDP) and remarkable decreases in the negative charge populations on the base moieties (0.42 a.u. for 3′,5′-dADP and 0.46 a.u. for 3′,5′-dGDP). The NPA analyses for the pyrimidine nucleoside-5′-monophosphate () and nucleoside-3′-monophosphate () indicate that the excess electron does not qualitatively reside on the phosphate group for the nucleoside-monophosphate. Approximately 20% of the negative charge distribution on the 3′-phosphate group for the pyrimidine nucleoside-3′,5′-diphosphates thus originates from cooperative effects owing to the coexistence of these two phosphate groups. The distribution of the ‘last’ electron amongst the constituent parts seems directly correlated to the VDEs discussed above.
The fully optimized geometries of the neutral and anionic nucleoside-3′,5′-diphosphates are depicted in . The most striking finding is that pyramidization of atom C8 in the radical anion dA and atom C6 of dC and dT (also in the pyrimidine nucleoside-monophosphates) (,,,) is barely detected in the anionic nucleoside-3′,5′-diphosphates. The dihedral angle D is less than 3° in 3′,5′-dCDP and 3′,5′-dTDP [it is ∼10–20° in the corresponding nucleosides and their monophosphate complexes (,,)], and the dihedral D in 3′,5′-dADP is less than 1° (in contrast, it is about 30° in dA) (). Compared to the nucleosides and the nucleoside-monophosphates, the geometric variations due to electron attachment are less obvious for the bases of the nucleoside-3′,5′-diphosphates. Typically, the C5–C6 and C6–N1 bond distances in 3′,5′-dCDP are 0.03 and 0.04 Å longer than those for the neutral species. These differences are about 0.02 Å less than the corresponding bond-length elongations found for the formation of 5′-dCMP (the C5–C6 and C6–N1 bond increases amount to 0.05 and 0.06 Å for in the radical anion of 5′-dCMP, and 0.05 and 0.06 Å in the radical anion of dC, respectively) (,). Similarly, the C5–C6 and C6–N1 bond distance increases are 0.03 and 0.03 Å in 3′,5′-dTDP, while they are 0.05 and 0.06 Å in dT and 5′-dTMP, respectively. (,) Moreover, the bond elongations in N7–C8 and C8–N9 due to electron attachment to 3′,5′-dADP are significantly less than those () for the related nucleoside (0.01 Å for 3′,5′-dADP versus 0.07–0.09 Å for dA). The reduction of the glycosidic bond distance in the radical anions of the nucleoside-3′,5′-diphosphates is about 0.01 Å less than that for the corresponding nucleosides and nucleoside-monophosphates. The less significant geometric alterations of the base moiety of the radical anions of 3′,5′-dADP, 3′,5′-dCDP and 3′,5′-dTDP might originate from the less negative charge distribution near the base moiety, consistent with the molecular orbital analysis and the charge distribution analysis discussed above. Due to the nature of the dipole-bound anion, it is not unexpected that the geometry of the guanine base of 3′,5′-dGDP is little changed by electron attachment.
Analogous to the pyrimidine-monophosphates, interaction with water greatly improves the electron-capture ability of DNA single strands. Note that when we speak of the ‘electron affinity’ of a solvated molecule M, the physical situation described is a microsolvated M(HO) system, in which the water uniformly enclose M and becomes arbitrarily large. In this sense, the ‘AEA’ values are 1.59 eV for 3′,5′-dADP, 0.95 eV for 3′,5′-dGDP, 1.99 eV for 3′,5′-dCDP and 1.98 eV for 3′,5′-dTDP in aqueous solution. The increases in the AEAs with solvation range from 0.6 eV (3′,5′-dGDP) to 1.5 eV (3′,5′-dCDP), similar to those for the pyrimidine-monophosphates (,).
It is even more important to notice that the polar solvent surroundings associated with aqueous solution alter the distribution of the excess electron in each DNA single strand dramatically (see ). Due to the solvation effects, the excess electron resides almost completely near the base for 3′,5′-dCDP, 3′,5′-dTDP and even for 3′,5′-dADP, for which case the base fragment excess electron density is low in the gas phase. Conversely, the excess electron distribution in aqueous solution is largely limited to the vicinity of the 3′-phosphate group in 3′,5′-dGDP, resulting in a phosphate-located radical anion. Although there is a hypothesis () that the excess electron might reside on the phosphate group of DNA single strands, the present study provides the sound theoretical evidence for only the guanine-related nucleotides. However, the present study also reveals that this hypothesis might be applicable for the purine nucleotides in the gas phase and only valid for the guanine nucleotide in aqueous solution.
It is well known that the phosphate groups of the nucleotides are deprotonated in aqueous solution. Recent studies demonstrate that the EAs of the pyrimidine nucleotides are nearly independent of the existence of counterions in aqueous solution (,). Accordingly, the deprotonation of the phosphate groups in aqueous solution should not affect the EAs of 3′,5′-dCDP and 3′,5′-dTDP. Note that the excess electron density in 3′,5′-dADP is distributed mainly on the base moiety in aqueous solution. Thus, one might expect that deprotonation of the phosphate in dADP would influence its electron affinity to a minor degree. However, since the unpaired electron is primarily located on the phosphate moiety of 3′,5′-dGDP in aqueous solution, the deprotonation of the phosphates might reduce the electron affinity of dGDP considerably.
The 2′-deoxyguanosine-3′,5′-diphosphate, 2′-deoxyadenosine-3′,5′-diphosphate, 2′-deoxycytidine-3′,5′-diphosphate and 2′-deoxythymidine-3′,5′-diphosphate systems are the simplest fragments of DNA single strands that might be considered representative. Exploring electron attachment to these archetypal units of DNA single strands enables one to approach reliable predictions of the EAs of DNA single strands.
DNA single strands have a strong tendency to capture low-energy electrons and to form electronically stable radical anions. The relatively large VEAs predicted for 3′,5′-dTDP (0.17 eV) and 3′,5′-dGDP (0.14 eV) in this investigation indicate that both the thymine-rich and the guanine-rich DNA single strands have the ability to capture low-energy electrons. Conversely, the small VEA values of 3′,5′-dCDP (0.03 eV) and 3′,5′-dADP (0.02 eV) imply that the electron-capture efficiency of the cytosine and adenine-rich DNA single strands is lower than those for guanine and thymine.
The AEAs of the nucleotides range from 0.22 to 0.52 eV and follow the order 3′,5′-dTDP > 3′,5′-dCDP > 3′,5′-dGDP > 3′,5′-dADP. A substantial increase in the AEA is predicted compared to that of the corresponding nucleic acid bases and the corresponding nucleosides. The effect of the phosphate groups in stabilizing the radical anions of the DNA components is crucial.
The coexistence of the phosphate groups at the 3′ and 5′ positions of the ribose results in a more delocalized electron distribution around both the base and the phosphate moieties. About 20–30% of the negative charge is located at the phosphate group for the pyrimidine nucleotides, while the analogous distribution is about 50% for the purine nucleotides.
Aqueous solution dramatically increases the electron-capturing ability of the DNA single strands, by up to 1–2 eV. Moreover, the solvent effect localize the excess electron on either the base (for 3′,5′-dADP, 3′,5′-dCDP and 3′,5′-dTDP) or the phosphate (for 3′,5′-dGDP) subunit.
It is worthwhile to point out that base stacking seems unlikely to affect the EAs of single strands of DNA, as shown in the computational study by Simons’ group (). The primary results of our own study of the solvated dGdC base-stacking model suggest that this is also true for the stacked bases in aqueous solution.
Low energy electron (LEE) attachment induced DNA SSB has attracted great interest from both experiment and theory (,,,,). Electron attachment to DNA single strands may lead to either N–C glycosidic bond rupture or C3′–O3′ and C5′–O5′ σ-bond breakage. Distinct electron attachment positions have been proposed based on different model studies (,). The present investigation demonstrates that in the gas phase, the excess electron is located both on the nucleobase and the phosphate moiety for DNA single strands. However, the distribution of the negative charge is not even. The attached electron favors the base moiety for the pyrimidine DNA oligomers, and it prefers the 3′-phosphate subunit for the purine DNA single strands. In contrast, the distribution of the extra negative charge is more localized in aqueous solution. The attached electron is tightly bound to the base fragment for the cytidine, thymidine and adenosine nucleotides; for the guanosine nucleotides, this ‘last electron’ resides primarily in the vicinity of the 3′-phosphate group due to the solvation effects.
The comparatively low VDE values predicted for 3′,5′-dADP and 3′,5′-dGDP indicate that both are electronically less favorable, and electron detachment might compete with reactions involving relatively high activation barriers such as glycosidic bond breakages. However, the radical anions of the pyrimidine nucleotides with high VDE values are expected to be electronically stable. Thus the base-centered radical anions of the pyrimidine nucleotides might be the possible intermediates for DNA SSB. |
The high fidelity displayed by replicases from families A, B and C outcomes from 10 to 10-fold polymerase preference for inserting correct rather than incorrect nucleotides (). Such selectivity has traditionally been considered to rely on the base-pair shape and size, as the result of the geometric restraints imposed by the polymerase active site to tolerate equivalent Watson–Crick base pairs, ruling out those differing from this geometry (). Additionally, recent studies performed with non-natural nucleotides also suggest a role for the π–π stacking interactions between the aromatic rings of the incoming dNTP and amino acid residues for an efficient polymerization at least in family B DNA polymerases (). Polymerases contact DNA at positions in which both, topology and chemistry are identical among the four canonical base pairs. Thus, interactions occur mainly with the base, sugar and triphosphate moieties of the incoming nucleotide, with the nucleotide placed immediately 5′ with respect to the templating nucleotide and, most importantly, with the minor groove of the nascent base pair (,). In this sense, structural studies of DNA polymerase complexes have revealed the presence of conserved residues at their catalytic sites that make contacts with the purine N and pyrimidine O atoms that act as hydrogen-bond acceptors and are positioned at similar locations in the minor groove (,). Incorrect insertion of a nucleotide will move such H-acceptors out of position, breaking interactions with the polymerase and diminishing the DNA-binding stability at the polymerization site. This will result in an increased chance of the primer DNA to be switched to the 3′-5′ exonuclease site of the replicase to have the incorrect nucleotide removed ().
Lesions in the genomes arise by their continuous exposure to the action of toxic environmental agents such as ultraviolet and ionizing radiation, genotoxic chemicals, by-products of the normal metabolism like the reactive oxygen species (ROS), in addition to spontaneous breaks of chemical bonds (). Most organisms code for a number of enzymes to repair such damages before replication fork meeting them since replicative DNA polymerases stall when they come across one of these lesions as they cannot form a proper and catalytically competent ternary complex with the damaged nucleotide at their catalytic sites. One exception is the specially deleterious lesion 8-oxo-7,8,-dihydro-2′-deoxyguanosine (8oxodG), produced by ROS inside the cell () as, if it escapes the repair machinery, it can be used by the replicase as template or as incoming nucleotide. The harmfulness of this lesion resides in its dual coding potential as it can pair with both cytosine and adenine during DNA synthesis, in the latter case leading to G to T transversions (,). Structural studies of both 8oxodG:dC and 8oxodG:dA base pairs showed that 8oxodG adopts the classical conformation opposite dC, pairing in a Watson–Crick fashion, whereas its glycosidic bond adopts the orientation to form a Hoogsteen base pair with dA ().
Structural modelling of catalytically competent complexes of RB69 and T7 DNA polymerases have suggested that the preferential insertion of dC opposite such a lesion is accomplished by a handicapped dA incorporation since the O atom of the 8oxodG():dA mispair sterically clashes with specific residues at the corresponding active sites (,). However, notwithstanding their high insertion fidelity and preferential dC insertion during the bypass of 8oxodG-containing templates, replicases can misincorporate dA with a moderately high efficiency () due to the fact that such a base pair establishes appropriate hydrogen-bond interactions with the minor groove sensing residues at the catalytic site ().
Bacteriophage ϕ29 DNA polymerase is a protein-primed DNA-dependent replicase belonging to the eukaryotic-type family of DNA polymerases (family B). Like many other replicases, it possesses, within a single polypeptide chain, both 5′–3′ polymerization and 3′–5′ exonuclease activities. It displays a high intrinsic nucleotide insertion discrimination [10 to 10 ()], which is further improved 100-fold through proofreading by the exonuclease domain (). ϕ29 DNA polymerase displays two unique characteristics compared with most replicases. First, a DNA polymerase molecule replicates the entire genome processively without the assistance of processivity factors (), in contrast to most replicases that require accessory proteins to clamp the enzyme to the DNA (). Second, ϕ29 DNA polymerase couples processive DNA polymerization to strand displacement. This ability allows the enzyme to replicate the ϕ29 double-strand genome without the need for a helicase ().
In this article, we study the ability of ϕ29 DNA polymerase to perform translesional synthesis past 8oxodG by assaying the nucleotide insertion opposite the lesion and further extension steps, as well as its capacity to proofread the formed pairs, this issue being of importance, as this enzyme is currently used for isothermal rolling circle amplification and whole genome amplification (,). Structural models mentioned above, together with multiple sequence alignments of DNA polymerases (), as well as the availability of the crystallographic structure of ϕ29 DNA polymerase (), have led us to analyse the role in translesional synthesis past 8oxodG of the invariant Tyr residue of the highly conserved B motif of family B DNA polymerases by means of substitutions at the corresponding ϕ29 DNA polymerase residue Tyr390. The results obtained allow us to propose a principal and dual role for this residue as one of the key determinants that dictate nucleotide insertion and extension preferences during translesion synthesis past 8oxodG by family B replicases.
Unlabelled nucleotides were purchased from Amersham Pharmacia Biochemicals. [γ-P]ATP (3000 Ci/mmol) was obtained from Amersham Pharmacia. Oligonucleotides Pber (5′CTGCAGCTGATGCGC), Pber-2 (5′CTGCAGCTGATGC), PberA (5′CTGCAGCTGATGCGC), PberC (5′CTGCAGCTGATGCGC) and sp1 (5′GATCACAGTGAGTAC) were 5′-labelled with [γ-P]ATP and phage T4 polynucleotide kinase and purified electrophoretically on 8 M urea–20% polyacrylamide gels. Labelled Pber, Pber-2, PberA and PberC were hybridized to the 34-mer 8oxodG containing oligonucleotide 5′GTACCCGGGGATCCGTACGCGCATCAGCTGCAG, and labelled sp1 to sp1cC+5 (5′TCTATGTACTCACTGTGATC) and to sp1cA+5 (5′TCTATGTACTCACTGTGATC). Hybridizations were performed in the presence of 0.2 M NaCl and 50 mM Tris-HCl (pH 7.5), resulting in primer/template structures. Oligonucleotides were obtained from Invitrogen.
Phage T4 polynucleotide kinase was obtained from New England Biolabs. ϕ29 DNA polymerase variants at Tyr390 residue Y390F and Y390S were constructed by J. Saturno in an exonuclease deficient background (D12A/D66A) (hereafter PolExo and PolExo, respectively (unpublished data) and further purified from BL21(DE3) cells harbouring the corresponding recombinant plasmid ().
The hybrid molecules Pber-2/T4 and Pber-2/8oxodG, containing the unmodified dG and an 8oxodG lesion at the +3 position of the template, respectively, can be used both as substrate for the 3′–5′ exonuclease activity and for DNA-dependent DNA polymerization. The incubation mixture contained, in a final volume of 12.5 μl, 50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml bovine serum albumin (BSA), 10 mM MgCl, 1.2 nM of either 5′-labelled Pber-2/T4 or Pber-2/8oxodG, 24 nM of either wild-type or mutant D12A/D66A (Exo) DNA polymerase, and the indicated concentration of the four dNTPs. After incubation for 5 min at 25°C, the reaction was stopped by adding EDTA up to 10 mM. Samples were analysed by 8 M urea–20% PAGE and autoradiography. Polymerization or 3′–5′ exonuclease activity are detected as an increase or decrease, respectively, in the size (13-mer) of the 5′-labelled Pber-2 primer.
The incubation mixture contained, in 12.5 µl, 50 mM Tris-HCI (pH 7.5), 10 mM MgCl, 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml BSA, 6 nM of wild-type ϕ29 DNA polymerase and 1.2 nM of either 5′-labelled PberC/T4, PberA/8oxodG or PberC/8oxodG double-stranded DNA (dsDNA) substrate. Samples were incubated at 25°C for the indicated times and quenched by adding EDTA up to 10 mM. Reactions were analysed by 8 M urea–20% PAGE and autoradiography.
The incubation mixture contained, in a final volume of 12.5 μl, 50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml BSA and 10 mM MgCl. As substrate, 1.2 nM of the 5′-labelled hybrid molecule Pber/8oxodG dsDNA was used. The amount of DNA polymerase added (12, 30 and 18 nM of PolExo, PolExo and PolExo mutants, respectively) was adjusted to obtain linear conditions. Samples were incubated for 15 s at 25°C, in the presence of the indicated concentrations of either dCTP or dATP, and quenched by adding 3 μl of gel loading buffer. Reactions were analysed by electrophoresis in 8 M urea–20% PAGE and quantified using a Molecular Dynamics PhosphorImager. Formation of the extended product was plotted against dNTP concentration. Apparent values for Michaelis–Menten constant and for incorporation opposite 8oxodG were obtained by least-squares nonlinear regression to a rectangular hyperbola using Kaleidagraph 3.6.4 software. was calculated by dividing the by the enzyme concentration. Catalytic efficiency was obtained by dividing by . Discrimination factor () for the nucleotide insertion opposite 8oxodG was obtained by dividing the catalytic efficiency for the incorporation of dCTP by the corresponding one for the insertion of dATP. Assays were performed at least three times to guarantee reproducibility.
As substrate, 1.2 nM of the 5′-labelled hybrid molecules PberA/8oxodG and PberC/8oxodG were used to analyse extension of 8oxodG:dA and 8oxodG:dC base pairs, respectively. Assay was carried out under the same experimental conditions as described above, but in the presence of the indicated concentrations of dGTP, the next correct nucleotide to be inserted. Extension parameters were obtained as mentioned. Discrimination factor () for the extension of DNA synthesis past the 8oxodG lesion was obtained by dividing the catalytic efficiency for the extension of 8oxodG:dA by the corresponding one for the extension of 8oxodG:dC. Assays were performed at least three times to guarantee reproducibility.
Reactions were carried out essentially as described (,).The incubation mixture contained, in a final volume of 12.5 μl, 50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml BSA, 10 mM MgCl, 1.2 nM of the 5′-labelled hybrid molecule PberA/8oxodG or PberC/8oxodG and 25 nM of either PolExo or 75 nM of PolExo mutant. After 5 min at 25°C to reach the binding equilibrium, samples were mixed with a 1000-fold excess of unlabelled substrate to trap dissociated DNA polymerase molecules. After the delay times indicated, 100 μM dGTP was added. Reactions were quenched by adding 3 μl of gel loading buffer 15 s after addition of dGTP. For 0 min delay time point, dGTP was included in the solution with the trapping DNA. Control reaction of the effectiveness of the trap was performed by incubating the DNA polymerase with the labelled and trapping DNA simultaneously. Reactions were analysed by electrophoresis in 8 M urea–20% PAGE and quantified using a Molecular Dynamics PhosphorImager. The fraction of primer molecules extended () by the saturating concentration of dGTP was plotted against the delay time (). These data were fit to the single exponential = + to obtain the rate constant for dissociation (). Assays were performed at least three times to guarantee reproducibility.
The incubation mixture contained, in a final volume of 12.5 μl, 50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml BSA, 10 mM MgCl and 12 nM of PolExo. As substrate, 1.2 nM of the 5′-labelled hybrid molecule sp1/sp1cC+5 or sp1/sp1cA+5 was used. Samples were incubated at 25°C for 15 s in the presence of the indicated concentrations of 8oxodGTP and quenched by adding 3 μl of gel loading buffer. Reactions were analysed and quantified as described above.
The incubation mixture contained, in a final volume of 12.5 μl, 50 mM Tris-HCl (pH 7.5), 1 mM dithiothreitol, 4% glycerol, 0.1 mg/ml BSA, 10 mM MgCl and 12 nM of PolExo. As substrate, 1.2 nM of the 5′-labelled hybrid molecule sp1/sp1cC+5 or sp1/sp1cA+5 was used. Samples were incubated at 25°C, in the presence of either 100 nM dGTP or 64 μM 8oxodGTP to obtain the dC:dG, dC:8oxodG and dA:8oxodG base pairs poised for extension by the indicated dATP concentrations. Samples were incubated at 25°C for 15 s and quenched by adding 3 μl of gel loading buffer. Reactions were analysed by electrophoresis in 8 M urea–20% PAGE.
ϕ29 DNA polymerase is the most processive replicase known, being able to incorporate, without the assistance of processivity factors, more than 70 000 nt during a single encounter with an undamaged DNA template (). Considering the presence of ROS derived from normal cellular metabolism that react with and modify DNA, is predictable that ϕ29 DNA polymerase has to deal with such DNA lesions during replication. Here, we have analysed the efficiency of translesion synthesis past one of the most abundant DNA lesions caused by ROS, the 8oxodG. To test ϕ29 DNA polymerase ability to carry out nucleotide insertion and further extension opposite such a lesion, primer extension reactions were conducted in the presence of increasing concentrations of the four dNTPs (see the Materials and Methods section). The 5′-labelled primer molecule was annealed to templates containing at the +3 position with respect to the 3′-end of primer terminus either the natural dG (X=G) or 8oxodG (X=8oxodG). As shown in , in the absence of nucleotides, the wild-type enzyme digests the primer strand by means of its 3′–5′ exonuclease activity and requires 50–100 nM dNTPs to get a net elongation of the primer strand with the unmodified template, yielding full-length (34-mer) products. In contrast, at those high dNTPs concentrations (100 nM), the polymerase stalls at the +1 and +2 positions, that is, prior to the 8oxodG lesion, showing nucleotide insertion opposite damage as a limiting step. As dNTP concentration increases, a band at the +3 position appears just opposite 8oxodG, indicating that extension of the primer terminus paired to 8oxodG is also impaired, needing an 800 nM dNTP concentration to come across the damage. In parallel, we tested the behaviour of the exonuclease deficient ϕ29 DNA polymerase mutant D12A/D66A on these molecules to study the effect of the 3′–5′ exonuclease activity during translesion synthesis past 8oxodG (). The absence of competition by the 3′–5′ exonuclease activity allows mutant polymerase to insert and further elongate nucleotides at dNTPs concentrations lower than those of the wild-type enzyme with the unmodified template. Interestingly, the mutant polymerase was able to insert nucleotides opposite 8oxodG at 100 nM dNTPs, as manifested by the presence of a band at the +3 position. To get full-length products with the damaged DNA, both mutant and wild-type DNA polymerases required a dNTP concentration 16-fold higher than that needed with the undamaged template. From this, it could be inferred that the 3′–5′ exonuclease activity of the wild-type enzyme exerts a similar pressure with both substrates, suggesting that during translesional synthesis DNA polymerase could proofread the primer terminus paired with 8oxodG.
To carry out a comparative analysis of the capacity of the wild-type DNA polymerase to excise 3′ terminal dA or dC paired to 8oxodG versus dC opposite the unmodified dG, we performed a 3′–5′ exonuclease assay using as substrates the hybrid molecules depicted on top of (see also Materials and Methods section). As shown in , the wild-type DNA polymerase was able to degrade the three substrates with the same efficiency, irrespective of their terminus base pair. The inherent potential deleteriousness of the 8oxodG():dA mispair is a consequence of not being identified as aberrant by the polymerase since, in such orientation, the O group mimics the O group of a thymidine, establishing a proper hydrogen-bond contact with the minor groove sensing residues at the catalytic site (). This fact allows stable mispair formation at the polymerization site, avoiding its exonucleolytic correction. In addition, DNA polymerases contain residues that make interactions through the minor groove of the DNA at post-replication positions that contribute to binding affinity and to the ability to sense base-pair geometry (). The misinsertion and further extension of nucleotides provoke changes in the geometry of the DNA with the subsequent lost of several of these contacts, causing DNA polymerase to stall and proofread the misinserted nucleotides (). Protrusion of the Hoogsteen base pair 8oxodG():dA into the major groove does not alter the overall structure of the DNA, contributing to avoid its exonucleolytic proofread.
As most DNA polymerases assayed in their capacity to insert nucleotides opposite 8oxodG [see () and references therein], ϕ29 DNA polymerase is able to catalyse mainly the insertion of both dC and dA (data not shown). Crystallization of bacteriophage T7 (family A) () and RB69 (family B) () DNA polymerases complexed with a nascent 8oxodG():dCTP base pair have given light about the structural rationale for the efficient bypass of 8oxodG, in contrast to other lesions as abasic sites and thymine dimers that completely stall the replication machinery until the lesion is either repaired or bypassed. In both structures, the unfavourable steric interaction between the O group and O4 oxygen of the 8oxodG in its non-mutagenic () conformation is alleviated by a kinking of the phophodiester backbone of the template strand at the catalytic site. In T7 DNA polymerase, there is an additional contact between the ε amino group of residue Lys536 and the O group of 8oxodG steadying the conformation for base pairing with dCTP (,). On the other hand, modelling of a nascent 8oxodG():dATP base pair in the active site of both DNA polymerases allowed Brieba . () and Freisinger . () to predict steric and electrostatic clashes between the O group and the side chain of residues located at the fingers subdomain, as responsible for the discrimination against the 8oxodG in its (mutagenic) conformation. In spite of the presence of polymerase-dependent specific residues, in both DNA polymerases the Tyr residue of the highly conserved motif B (Tyr567 and Tyr530 in RB69 and T7 DNA polymerases, respectively) () is proposed to play a role in discrimination during nucleotide insertion opposite an 8oxodG site.
To study the role of such Tyr residue in 8oxodG translesion synthesis, we have analysed the insertion and elongation capacities opposite 8oxodG-containing templates of ϕ29 DNA polymerase variants at the homologous Tyr390 residue, in an exonuclease deficient background (D12A/D66A). PolExo and PolExo mutants, allowed us to evaluate the importance of the hydroxyl and aromatic groups, respectively, in both translesional phases by comparing their activities with the wild-type enzyme (PolExo).
To analyse the ability of ϕ29 DNA polymerase mutants to carry out nucleotide insertion opposite 8oxodG, primer extension reactions were conducted in the presence of increasing concentrations of the corresponding dNTP (see Materials and Methods section). A 5′-labelled primer molecule (15-mer) was annealed to a template (34-mer) containing 8oxodG at the nascent position (see Materials and Methods section). As it can be observed in and , mutant PolExo displayed a clear preference to incorporate dC with respect to dA, mainly due to the lower for dC, showing a discrimination factor ( in ) of about 5, a moderately high insertion fidelity in comparison with other polymerases belonging to families A, B, X, Y and reverse transcriptase (RT) (), in which the dC:dA insertion ratio ranges from 0.075 for HIV-1 RT () to 91 for Dpo4 DNA polymerase () [compiled in ()]. The incorporation of dA cannot be attributed to a decreased insertion fidelity of PolExo, as it discriminates 3 × 10-fold against dA insertion opposite the undamaged dG (data not shown). Under those conditions, and consistent with their reduced efficiency in incorporating nucleotides opposite undamaged templates due to a reduced dNTP-binding capacity (,), PolExo and PolExo mutants were also less active than PolExo in inserting both dA and dC at 8oxodG ( and ). Thus, correct dC insertion by PolExo and PolExo mutants is reduced to two and one orders of magnitude relative to the wild-type polymerase, respectively. These data are in good agreement with the comparative efficiencies displayed by these two mutants during incorporation of dNTPs on undamaged substrates (), precluding any additional incorporation defect due to the presence of 8oxodG at the templating position. However, their discrimination factors clearly changed. Whereas PolExo mutant discriminates against dA insertion in a similar fashion as PolExo (∼ 4), PolExo mutant displays nearly identical insertion efficiency for both nucleotides (∼ 1). Such behaviour is not a direct consequence of the diminished insertion fidelity showed by this mutant, as it has a discrimination factor against erroneous nucleotide insertion opposite a non-damaged nucleotide of 3 × 10 (,). This result clearly reflects the importance of the aromatic moiety of ϕ29 DNA polymerase residue Tyr390 in discriminating against the orientation (error-prone) of the purine base of the 8oxodG. Freisinger . modelled a nascent 8oxodG:dATP base pair at the catalytic active site of RB69 DNA polymerase (). Such a model led them to propose that the O group of 8oxodG() would clash with the molecular surface formed by residues Tyr567 and Gly568, allowing the canonical and non-mutagenic conformation to be preferentially placed (). The DNA and incoming nucleotide from the structure of the RB69 ternary complex () can be homology modelled onto the structure of the apo ϕ29 DNA polymerase by aligning the palm subdomains of both ϕ29 and RB69 DNA polymerases, the position of the modelled DNA and incoming nucleotide being consistent with the placement of the ϕ29 DNA polymerase catalytic carboxylates, steric gate residues that distinguish ribo- from deoxyribonucleotides, and residues involved in binding DNA (). This fact allowed us also to model a nascent 8oxodG:dATP in the corresponding active site of ϕ29 DNA polymerase () in which ϕ29 DNA polymerase residues Tyr390 and Gly391 are placed at the same location and orientation that the homologous RB69 DNA polymerase residues Tyr567 and Gly568, mentioned above. Based on that, the absence of discrimination showed by PolExo mutant could reflect that the lack of the aromatic group avoids the steric contacts with the O of the 8oxodG() either by removing a direct interaction with Tyr390 or by tolerating a subtle shift of Gly391 backbone to accommodate the O atom, or both. Interestingly, recent studies have involved the homologous Tyr766 residue of DNA polymerase I in preventing incorrect nucleotide insertion across from the dG:-2-aminofluorene adduct, as mutant Y766S was significantly less selective for correct nucleotide incorporation than the wild-type enzyme ().
To study the next step after insertion opposite 8oxodG, 16/34-mer hybrid molecules, with the primer 3′-end containing either dC or dA paired with 8oxodG at the template (see Materials and Methods section), were used as substrate to analyse single nucleotide incorporation of dG, the next correct nucleotide following the 8oxodG:dC/8oxodG:dA base-pair formation. As seen in (see also Supplementary Figure S1), PolExo mutant displayed a 14-fold favoured extension of the 8oxodG:dA base pair (see in discrimination factor, ). Interestingly, PolExo mutant lost discrimination ability to elongate the base pairs under study, as it showed a 5-fold preference to extend the 8oxodG:dA pair. In contrast, PolExo mutant recovered a nearly wild-type phenotype, exhibiting a 12-fold preference in the extension of such a base pair (see ). These results were pointing to the importance of the hydroxyl group of ϕ29 DNA polymerase residue Tyr390 in the establishment of contacts with the conformation of 8oxodG that allow the pair 8oxodG:dA to be preferentially extended.
The structural basis for the poor extension efficiency of the 8oxodG:dC base pair have been highlighted by modelling this pair, poised for extension, at the polymerization site of T7 and RB69 DNA polymerases (,). Thus, the conformation of 8oxodG would promote local perturbations of the DNA backbone to alleviate steric clashes between the O group of 8oxodG and its 5′ phosphate group, leading to a subtle shift in the sugar moiety of 8oxodG which would decrease the catalytic efficiency of the elongation of this base pair (). In addition, avoidance of those steric clashes would be accounted by a slight movement of the newly replicated duplex towards the DNA major groove with the direct consequence of a diminished binding stability at the polymerase active site (). To elucidate differences in the proper stabilization at the polymerase active site of both DNA substrates, assays with ϕ29 PolExo were carried out by measuring the DNA-binding stability of the enzyme to DNA hybrid molecules containing either a dA- or dC- 3′-end primer terminus paired to 8oxodG (see Materials and Methods section and Supplementary Figure S2). Interestingly, the 8oxodG:dA base pair ( = 0.23 ± 0.01 min) was bound 3-fold more stably to the PolExo polymerization site than the 8oxodG:dC pair ( = 0.70 ± 0.02 min). Thus, our results support the original proposal of a diminished binding stability of the template 8oxodG() when is forming a base pair poised for extension ().
RB69 DNA polymerase Tyr567 forms part of the minor groove mismatch sensing mechanism () by means of making, through a water molecule, hydrogen bonds to the sugar moiety and the acceptors located at the N of a purine or O of a pyrimidine of the ultimate template base (). Modelling of 8oxodG():dA pair ready for extension predicted that the O group could substitute this water molecule and make a hydrogen bond with the hydroxyl group of Tyr567, in an analogous manner as with the N of a non-modified dG () (), not detecting this mispair as being aberrant. In fact, the minor groove surface of this mispair is identical to the normal dT:dA base pair (). On the contrary, the steric clashes between the O of 8oxodG() and sugar moiety could difficult the proper formation of the water-mediated hydrogen bond between the Tyr residue and the N of 8oxodG in the 8oxodG:dC pair. This elegant model proposed by Freisinger . () to explain the structural determinants for insertion and further extension opposite 8oxodG is substantiated by the biochemical results presented here. Thus, the absence of the hydroxyl group in PolExo mutant would preclude stabilization of the 8oxodG() through such hydrogen bond, diminishing the discrimination for extending both base pairs. In this sense, relative stabilization of the 8oxodG:dA base pair ( = 0.32 ± 0.02 min) with respect to 8oxodG:dC ( = 0.58 ± 0.03 min) at the polymerization site of PolExo mutant decreased in comparison with PolExo (Supplementary Figure S2), while PolExo variant displayed a wild-type-like phenotype (data not shown). The smaller size of serine could allow one of its rotameric conformations to locate its hydroxyl group in a position suitable to contact the O group of 8oxodG(). Additionally, the absence of the bulky aromatic ring in mutant PolExo could favour a better accommodation of the nitrogen base of 8oxodG() that would permit the establishment of the above-mentioned interaction. Thus, allowing this mutant to discriminate against 8oxodG:dC elongation at a wild-type level. In this extension step, the aromatic group of Tyr would be dispensable, their role being restricted to confer specificity during the insertion step.
Our biochemical results, together with the structural models, let to predict the dual role of the conserved Tyr residue of motif B of family B DNA polymerases during the bypass of the 8oxodG lesions of the DNA by means of its aromatic and hydroxyl groups. The former will dictate the preferential 8oxodG() conformation during the nucleotide insertion step, and as a consequence will favour the non-mutagenic dC incorporation, while the latter will do it during further base-pair extension, stabilizing the 8oxodG() form promoting extension of the mispair 8oxodG:dA. Although the role of this Tyr as a key residue for controlling nucleotide misincorporation on natural DNA has been extensively described in family B DNA polymerases (), this is the first time that the specific contribution of its chemical substituents during the two steps of 8oxodG replication is biochemically demonstrated. On the other hand, contrarily to thermophilic A-family DNA polymerases, currently used for PCR, as polymerase from that incorporates indistinguishably dC and dA opposite 8oxodG (), preferential incorporation of dC opposite 8oxodG makes ϕ29 DNA polymerase one of the most suitable enzymes to amplify faithfully DNA substrates that, as ancient DNAs, could be widely oxidized.
In addition to promote lesions in the DNA, the presence inside cells of ROS also generates oxidatively altered purines and pyrimidines, the most abundant being 8oxodGTP (). Although organisms synthesize 8oxodGTPases to sanitize the dNTP pool, the incorporation of such modified nucleotide could have adverse consequences as pairing opposite adenine of the template strand would result in A:T to C:G transversions ().
Primer extension analyses similar to those described above were performed to analyse the templating nucleotide preference (dA versus dC) during 8oxodGMP insertion by ϕ29 DNA polymerase, using the 15/21-mer hybrid molecules sp1/sp1cC+5 and sp1/sp1cA+5 described in the Materials and Methods section (see also Supplementary Figure S3). As it can be seen in , ϕ29 DNA polymerase inserts 8oxodG preferentially opposite dC, although 2000-fold less efficiently than insertion of the unmodified dGMP. Additionally, qualitative analysis of the extension of both dA:8oxodG and dC:8oxodG base pairs obtained after insertion of 8oxodG in these molecules (see Materials and Methods section), show the preferential extension of the correct dC:8oxodG base pair (), with an efficiency similar to that of the normal dC:dG base pair (the appearance of the 18- and 20-mer bands is due to the pairing of 8oxodG with the templating dA and dC at these positions, respectively). These are very unusual results as most of DNA polymerases assayed displayed a favoured 8oxodG insertion opposite dA, as T7 and γ DNA polymerases from family A (,), β and λ DNA polymerases from family X (,,) and Dpo4 and Dbh DNA polymerases belonging to family Y (). Solely HIV-RT, Pol Ik (family A), DNA pol II (family B) and DNA polymerase from African Swine Fever Virus (family X) showed an 8oxodG insertion pattern similar to that described for ϕ29 DNA polymerase (,). Conversely, PolIk and ϕ29 DNA polymerases are the only enzymes studied so far exhibiting a prone extension of the correct dC:8oxodG base pair () (this study).
At present, DNA polymerase ternary complexes with 8oxodGTP as incoming nucleotide have not been solved. This fact precludes the analysis of the structural rationale for 8oxodG insertion preferences. In , we have modelled dC:8oxodGTP() and dA:8oxodGTP() base pairs into the RB69 DNA polymerase active site. As it happens with a non-modified incoming nucleotide, Lys560 residue, placed at the fingers subdomain, will make a hydrogen bond to one of the two negatively charged equatorial oxygens of the α phosphate of the incoming 8oxodGTP(), enhancing catalysis by allowing the proper stabilization of the catalytic state (). Nevertheless, protrusion into the major groove of N, N and O atoms of the incoming 8oxodGTP() could led to a sterical and/or electrostatic clash between the N atom and the edge of the side chain of Lys560 (red patch in ), affecting the proper interaction between this residue and the α phosphate of the 8oxodGTP() and, as a consequence, its insertion. Structural alignments of both RB69 and ϕ29 DNA polymerases have allowed us to identify ϕ29 DNA polymerase Lys383 as the homologous lysine residue (). By the contrary, we have not found a structural rationale for the preferential extension of the dC:8oxodG base pair, as once inserted, the 8oxodGMP() should be translocated backwards to the primer terminus site. At this location, the only contacts between the protein and the 3′-end nucleotide are through the phosphodiester bond. The impaired elongation of the incorrect base pair could suggest that the conformation of the 3′ terminus either promotes a subtle distortion in the proper orientation of the attacking 3′-OH group to form the phosphodiester bond with the next incoming nucleotide, or a slightly variation in the angle of the glycosidic bond that could make difficult the establishment of the proper stacking interactions between the nitrogen bases of the incoming nucleotide and primer terminus. Difficulties in the translocation step due to the presence of 8oxodGMP() at the 3′ primer terminus cannot be ruled out.
As mentioned above, several DNA polymerases from eukaryotes and prokaryotes can use 8oxodGTP as incoming nucleotide, leading to transversions. Surveillance mechanisms have evolved to maintain a low frequency of these types of mutations, owing to the action of enzymes able to degrade such mutagenic substrates (). Thus, synthesizes MutT protein that hydrolyses 8oxodGTP to 8oxodGMP (), this latter form being unable to be rephosphorylated, preventing insertion of 8oxodGTP during DNA synthesis. Analogously, mammalian cells code for MTH1, a protein showing a similar enzymatic activity (). However, proteins as MutY, which excises a dA mispaired with 8-oxodG as part of the process to restore the original G:C base () could have a deleterious effect if 8oxodG had been the erroneously incorporated nucleotide. , the host of bacteriophage ϕ29, encodes two potential 8oxodGTPases that could prevent the mutagenic effects of 8oxodGTP, as the lack of both proteins sensitized growing cells to oxidizing agents (). Thus, if bacteriophage ϕ29 had to develop under conditions that could increase the pool of 8oxodGTP inside the cell, as severe oxidative stress or dysfunctions in the activity of any 8oxodGTPase, the discrimination against 8oxodGMP insertion as well as the great fidelity exhibited by ϕ29 DNA polymerase during the unlikely insertion and further extension of 8oxodGMP would guarantee the maintenance of the genetic information of the bacteriophage.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
The regulation of eukaryotic promoters is complex and involves the combinatorial action of transcription factors (). These transcription factors often form distinct modules which coregulate the expression of groups of genes with related function. One example of such a module, is the complex formed by the association of the Forkhead transcription factor Fkh2p and the MADS-box transcription factor Mcm1p in . This complex controls the cyclical activation of a cohort of genes expressed during the late G2 and M phases of the cell cycle (). Mcm1p itself functions in a number of alternative complexes with other coregulators to target genes involved in different biological processes (,). Indeed, combinatorial interactions with other transcription factors are typical of members of the MADS-box transcription factor family, including the mammalian protein SRF. Moreover, some of these combinatorial interactions appear to be evolutionarily conserved as, for example, both SRF and Mcm1p show interactions with homeodomain transcription factors ().
The yeast forkhead protein Fkh2p contains two defined domains, the Forkhead DNA-binding domain and the phospho-peptide binding FHA domain. In Fkh2p, the FHA domain acts as a transcriptional activation domain that functions through phosphorylation-dependent recruitment of the coactivator Ndd1p (,). There are over 40 human Forkhead transcription factors (). However, only two of these are known to possess FHA domains in addition to their Forkhead DNA-binding domain, FOXK1 and FOXK2. The mouse version of FOXK1, Foxk1/MNF exists as two isoforms, MNFα and MNFβ which differ through alternative splicing leading to the production of the C-terminally truncated MNFβ isoform (). Foxk1/MNF has been implicated in the correct functioning of myogenic stem cells (), which is in part due to proliferative defects caused by Foxk1/MNF loss (). Molecularly, MNFβ has been shown to act as a transcriptional repressor protein, but the role of MNFα is unclear (). Other forkhead proteins such as FOXM1 have been linked to controlling cell cycle-dependent expression of genes expressed at the G2-M boundary during the cell cycle (), suggesting that Forkhead proteins play a similar role in cell cycle control in yeast and mammalian systems.
Mcm1p and SRF are highly related proteins, especially within their DNA-binding domains, and this is reflected by their ability to bind similar DNA motifs known as CArG boxes (). Moreover, many of the protein–protein interfaces are conserved, exemplified by the observation that the yeast Forkhead transcription factor Fkh2p can form ternary DNA bound complexes with both Mcm1p and SRF (). Thus, we hypothesized that in an analogous manner to the yeast Mcm1p-Fkh2p complex, human SRF might also interact with a Forkhead transcription factor to control gene expression. Here we demonstrate that SRF interacts physically and functionally with FOXK1, demonstrating the evolutionary conservation of this transcription factor module.
The following plasmids were used in mammalian cell transfections. The L8G5E1a-Luc and LexA-VP16 constructs were provided by C. Lemercier (). pCH110 (Pharmacia), pEF1/myc-His/LacZ (Invitrogen), p6xFOX-Luc (kindly provided by R. Costa; 21), p5xCArG-Luc (Stratagene), pGL3-CDC25C-luc (pAS2608) (kindly provided by K. Engeland), pGL3-SM α-actin-Luc/pAS2268 (containing the rat promoter sequences −713 to +51 bp; kindly provided by S. Pham; 22), pAS2511, pAS2512 and pAS2513 (containing the rat promoter sequences −713 to +51 bp and mutations in either CArG box A, CArG box B or both CArG boxes A and B, respectively), were made by Quikchange mutagenesis using the primer-template combinations ADS1617/ADS1618-pAS2268, ADS1619/ADS1620-pAS2268, ADS1619/ADS1620-pAS2511, respectively. pCGNSRF (encoding HA-tagged full-length SRF, kindly provided by R. Prywes) was described previously.
pAS2256 [encoding CMV-driven full-length His-Flag tagged FOXK1(1–733)] was constructed by a two-step procedure. First, the HindIII/XbaI fragment from pAS1169 was cloned into the same sites in pCMV5 (encoding amino acids 97–733) to create pAS1173. Next, a HindIII/AscI-cleaved PCR fragment (primer pair ADS1315/ADS1316 and Image clone 30345138/pAS1186 as a template) was ligated into the same sites in pAS1173 to create pAS2256. pAS2265 [encoding CMV-driven full-length His-Flag tagged FOXK1(1–733)(H355A)] was constructed by ligating a KpnI/XbaI fragment from pAS2259 into the same sites in pCMV5. pAS1175 [encoding CMV-driven full-length His-Flag tagged FOXM1b(1–763)] was constructed by ligating the HindIII/XbaI fragment from pAS1171 into the same sites in pCMV5. pAS2257 [encoding CMV-driven FOXK1(1–262) fused to the GAL4 DNA-binding domain] was constructed by ligating a BamHI/XbaI-cleaved PCR fragment (primer pair ADS1315/ADS1297 and pAS2256 as a template) was ligated into the same sites in pAS2063 ().
For bacterial expression, pGEX-KG and pAS58 (encoding GST-core[amino acids 132–222]; 24) have been described previously.
For transcription/translation, pAS1242 [encoding full-length Fkh2p(1–862)] has been described previously (). pAS2255 [encoding T3-driven full-length His-Flag tagged FOXK1(1–733)] was constructed by a two-step procedure. First, HindIII/XhoI-cleaved PCR fragment (encoding amino acids 97–733; primer pair ADS1168/ADS1169 and Image clone 5168241/pAS1184 as a template) was cloned into the same sites in pAS728 () to create pAS1169. Next, a HindIII/AscI-cleaved PCR fragment (primer pair ADS1315/ADS1316 and Image clone 30345138/pAS1186 as a template) was ligated into the same sites in pAS1169 to create pAS2255. pAS1168 [encoding T3-driven full-length His-Flag tagged FOXK1(216–418)] was constructed by ligating a NcoI/XhoI-cleaved PCR fragment (primer pair ADS1166/ADS1167 and Image clone 5168241/pAS1184 as a template) into the same sites in pAS728. pAS2259 and pAS2258 [encoding T3-driven full-length His-Flag tagged FOXK1(1–733)(H355A) and FOXK1(216–488)(H355A), respectively] were created by QuikChange mutagenesis using the primer pair ADS1342/ADS1343 on the templates pAS2255 and pAS1168, respectively. pAS1171 [encoding T3-driven full-length His-Flag-tagged FOXM1b(1–748)] was constructed by ligating a HindIII/XhoI-cleaved PCR fragment (primer pair ADS1177/ADS1178 and Image clone 3834244/pAS1181 as a template) into the same sites in pAS728.
For reporter gene assays, typically 0.25 μg of reporter plasmid and 50 ng of pEF1/myc-His/LacZ or pCH110 were co-transfected with 0.005–2 μg of expression plasmids. Cell extracts were prepared and equal amounts of protein were used in luciferase and β-galactosidase assays as described previously ().
Real time RT-PCR was carried out as described previously (). The following primer-pairs were used for RT-PCR experiments. : ADS1372 (5′-CGAGTTCGAGTTCCTCATGC-3′) and ADS1373 (5′-GGGAGATCTGGGGGTACAGT-3′), : ADS1358 (5′-GCGTGGCTATTCCTTCGTTA-3′) and ADS1359 (5′-ATGAAGGATGGCTGGAACAG-3′), , ADS1348 (5′-AGCTGCTTCCACCTACCTCA-3′) and ADS1349 (5′-CTTCTGGTTGAGGGAATCCA-3′), SRF ADS1027 and ADS1028 (), , ADS4007 (5′-ACAGTCAGCCGCATCTTCTT-3′) and ADS4008 (5′-TTGATTTTGGAGGGATCTCG-3′) and 18S internal control, ADS4005 (5′-TCAAGAACGAAAGTCGGAGGTT-3′) and ADS4006, (5′-GGACATCTAAGGGCATCACAG-3′).
siRNA against , and matched control, were constructed by the Silencer™ siRNA construction kit (Ambion). Human target sequences were: FOXK1-1 5′-TTGTGATAGAGCGACGTG-3′ (ADS1403/4) and FOXK1-2 5′-TTCTGCACAAAGCGGTAAC-3′ (ADS1405/6) (italicized residues represent sequences used in the transcription process in producing the constructs). The siRNAs against and matched control siRNA (Santa Cruz) were made synthetically. To carry out RNA interference (RNAi), a two-step transfection protocol was carried out in 12-well plates as described previously ().
Western blotting was carried out with the primary antibodies; Flag (Sigma), GAPDH (Abcam), Erk2 (sc-154; Santa Cruz) and SRF (Santa Cruz) essentially as described previously ().
Coimmunoprecipitation analysis of overexpressed proteins was performed using protein A sepharose beads (Sigma) as described previously (). GST pulldown analysis was performed essentially as described previously () with purified bacterially expressed recombinant GST-core and translated FOXK1 derivatives.
Gel retardation assays were performed as described previously using a fragment from the promoter (), and core purified from bacteria () and translated FOXK1 and Fkh2p derivatives.
The biotinylated DNA-binding assays were performed essentially as described previously (). Total 2 × 10 HEK293 cells were cotransfected with 8 μg Flag tagged FOXK1 plasmids and 200 pmol control siRNA or siSRF using lipofectamine 2000. Twenty-four hours after transfection, the whole-cell extracts were incubated with 1 μg biotinylated Rat promoter DNA or 500 pmol of CArG-A element, which were immobilized on streptavidin–agarose beads (Dynabead M-280 streptavidin), in binding buffer [50 mM HEPES·KOH, pH 7.9/150 mM NaCl/0.5% Triton X-100/2 mM EDTA/20 mM NaF/1 mM NaVO/20 μg/ml poly(dI-dC) and protease inhibitors] at 4°C for 60 min. The beads were washed four times with the binding buffer, and precipitated proteins were analysed by western blotting. The biotin-linked Rat promoter was generated by PCR using the template pAS2268 (for the wild-type promoter) or pAS2513 (for the mutant promoter with both CArG boxes disrupted and the primers ADS1692 (5′-AAGGGTCAGCGATAAACCAA-3′) and ADS1693 (Biotin-CTTACCCTGATGGCGACTG-3′). The biotin-linked site containing CArG element A was created by annealing the oligonucleotides ADS1694 (biotin-CCTGTCTTTGCTCCTTGTTTGGGAAGCGAGTGGG) and ADS1695 (CCCACTCGCTTCCCAAACAAGGAGCAAAGACAGG).
Chromatin immunoprecipitation (ChIP) assays using control IgG (Upstate) or antisera specific to SRF (Santa Cruz), E2A (Santa Cruz: E2A, E12 (V-18) SC-349) and FOXK1 (Abcam) were performed as described previously () except that cross-linking was performed for 10 min. Bound promoters were detected by PCR using primers for the human promoter (ADS1650, 5′-CTCTGGGCATTTCTGCAGTT-3′ and ADS1651, 5′-TTCTGCTCTCCTCCCACTTG-3′), the human promoter (ADS1696, 5′-ACTTAGCCGTCCACAACAGG-3′ and ADS1697 5′-GGGGACTGGAAGTCATGTGT-3′) or for the promoter or intron 3 as described previously ().
The Forkhead transcription factor Fkh2p interacts with the MADS-box transcription factor Mcm1p in . To establish whether a similar interaction could be detected between the closest human homologue of Mcm1p, SRF and a forkhead transcription factor, we first cloned the full-length human homologue of Foxk1/MNF, by combining the sequences from two IMAGE clones. The sequence we cloned was identical to that recently identified by approaches (). Human FOXK1 shares a conserved domain structure with mouse Foxk1/MNF and yeast Fkh2p, with an FHA domain preceding the Forkhead DNA-binding domain (A). Overall, there is 90% sequence identity between human and mouse FOXK1/Foxk1. The sequence conservation between human FOXK1 and yeast Fkh2p is substantially lower (20% identity) and is concentrated mainly in the FHA and Forkhead domains.
Next, we co-expressed epitope-tagged versions of FOXK1 and SRF in human 293 cells and carried out co-immunoprecipitation analysis. Co-precipitation of SRF and FOXK1 was observed which was dependent on the expression of both transcription factors (B). To establish whether endogenous SRF could interact with FOXK1, we expressed Flag epitope-tagged FOXK1 and attempted to co-precipitate SRF. Endogenous SRF was co-precipitated with FOXK1 (C). We were however unable to detect interactions between SRF and FOXK1 expressed at endogenous levels, most likely due to the low levels of expression of these proteins coupled with the likely substoichiometric association of these transcription factors (data not shown). Interactions between yeast Fkh2p and Mcm1p occur with the DNA-binding domain of Mcm1p. We therefore tested whether we could detect analogous interactions between FOXK1 and SRF . Indeed, the DNA-binding domain of SRF (core) was sufficient for interaction with FOXK1 in a GST pulldown assay (D). This interaction was specific as different forkhead proteins FOXM1b and FOXN3 were unable to form specific complexes with core (data not shown). Moreover, both FOXK1 and Fkh2p could form complexes with SRF on a fragment derived from the promoter (E), a known binding site for the Fkh2p-Mcm1p complex, which contains a composite binding motif for both SRF and the forkhead transcription factor (). In contrast, FOXM1b was unable to form a similar complex (data not shown).
Collectively, these data demonstrate that in common with their yeast counterparts the human Forkhead FOXK1 and MADS-box SRF transcription factors can interact and .
A short splice form of the mouse homologue of FOXK1, MNFβ has previously been shown to act as a transcriptional repressor (). Full-length FOXK1 acts as a potent repressor of a reporter gene controlled by multimerized forkhead binding sites (A). In contrast, FOXM1b acts as a transcriptional activator in this assay (A). The N-terminal part of FOXK1 which includes the FHA domain is sufficient to repress transcription, as revealed by the ability of a GAL-FOXK1(1–262) fusion protein to repress transcription (B). This effect is specific, as the GAL4 DNA-binding domain alone or other control GAL4 fusions do not cause such dose-dependent decreases in reporter activity (data not shown).
To establish potential functional interactions with SRF, we examined the activity of FOXK1 on a reporter driven by multimerized SRF binding sites (CArG-luc). FOXK1 functioned as repressor of this reporter gene while the alternative forkhead transcription factor FOXM1b had little effect on the activity of this reporter (C). Importantly, upon depletion of SRF by siRNA, the repressive effect of FOXK1 on the CArG-luc reporter was lost (D). Thus, FOXK1 can repress SRF-dependent promoter activation.
A previous microarray study has identified a number of potential FOXK1 target genes through the analysis of muscle side population cells derived from knockout mice (). Similarly, a number of SRF target genes were identified through microarray analysis following the expression of SRF-VP16 fusion proteins in ES cells derived from knockout mice (). Two of the deregulated genes in both cases were smooth muscle () and (encoding a lysosomal protective protein) which were induced by either loss of Foxk1/MNF or expression of SRF-VP16, making this a likely candidate for detecting functional interactions between these two transcription factors.
First, we confirmed the involvement of human FOXK1 and SRF in controlling and expression. Two different siRNAs were derived which caused the depletion of FOXK1 expression at both the protein and RNA levels (A and B). Upon depletion of in 293 cells, we observed a large increase in and levels with both siRNA constructs, consistent with a role for FOXK1 in repressing the activity of these genes (B). Similar results were obtained in muscle-derived RD18 rhabdomyosarcoma cells (data not shown). Conversely, the depletion of SRF caused a substantial decrease in the expression of and expression (C), confirming the ability of SRF to activate the expression of these genes. Indeed, other studies have previously also implicated SRF in controlling the activity of (). Importantly, ChIP analysis demonstrated that both FOXK1 and SRF could be found associated with the endogenous and genes in 293 (E) and RD18 cells (data not shown). In contrast, only SRF was found to associate with the promoter and neither protein associated with an intron found in the gene (E, bottom panels). Thus, both FOXK1 and SRF can directly associate with the promoter regions and control the expression of and .
To establish whether FOXK1 could affect the activity of the promoter, we carried out reporter gene assays. Overexpression of FOXK1 caused the dose-dependent repression of a promoter–reporter construct (A). In contrast, at the same levels of expression, little effect was observed on the activity of a promoter–reporter construct (B). Conversely, knockdown of expression caused an enhancement in the activity of the promoter (C). This effect was specific, as no effect was seen on the promoter (D). Thus, FOXK1 acts to repress the activity of the promoter.
FOXK1 might regulate the promoter through either direct promoter binding or indirectly through association with SRF. We therefore created a mutant form of FOXK1 containing a point mutation in a region that is predicted to disrupt its DNA-binding surface [FOXK1(H355A)]. This point mutation did not affect interactions between FOXK1 and SRF (A), but severely compromised the ability of FOXK1 to bind to DNA (B, lanes 7–9). Indeed, FOXK1(H355A) exhibited much reduced repressive ability on a reporter driven by six forkhead binding sites (6xFox-luc; C).
Next, we demonstrated that increasing amounts of SRF were able to activate the promoter in 293 cells (D). Conversely, when we co-expressed increasing amounts of wild-type FOXK1 with SRF, we were able to demonstrate that FOXK1 was able to repress the activity of this SRF-activated reporter in this cell type (E). Moreover, the DNA-binding defective FOXK1(H355A) also repressed the activity of the promoter to a similar extent as the wild-type protein (E) suggesting that FOXK1 does not need to bind to DNA directly. Indeed, in agreement with this, we were unable to detect significant binding of wild-type FOXK1 to fragments of the promoter (data not shown). Similarly, both wild-type FOXK1 and FOXK1(H355A) were able to repress the promoter in the presence of endogenous levels of SRF in RD18 cells (F).
Together these data demonstrate that FOXK1 can regulate the promoter in a DNA-binding independent manner.
A likely mechanism through which FOXK1 regulates the promoter is via direct interactions with SRF. To establish whether this is the case, we first carried out reporter assays using a series of promoter–reporter constructs in which either one or both of the two SRF binding sites were mutated (A). FOXK1 was able to repress the activity of the wild-type promoter, but this activity was reduced upon mutation of either of the two SRF binding sites. Moreover, the repressive activity of FOXK1 was virtually abolished upon mutation of both SRF binding sites (CArG-A/Bmut; A). Next, we depleted endogenous SRF levels by siRNA treatment (B). Upon depletion of SRF, the ability of FOXK1 to repress the promoter was blunted.
Finally, we asked whether SRF was required for FOXK1 recruitment to the promoter. We were unable to detect FOXK1 binding to the promoter by gel retardation analysis (data not shown), therefore, we used an immobilized biotinylated-DNA-binding assay to detect FOXK1-DNA interactions . Either the wild type or a mutant version (with both CArG boxes disrupted) of the promoter were coupled to streptavidin-linked beads. As a further control, we created duplexes which spanned just the proximal CArG (CArG-A) from this promoter. Extracts from 293 cells were used in the binding assays from cells that had been transfected with Flag-tagged FOXK1 in the presence of control siRNA or siRNA constructs against SRF. Binding of FOXK1 to the wild-type promoter could be detected and this binding was reduced upon depletion of SRF (A, lanes 5 and 6). Similar effects were seen on the isolated CArG-A element (A, lanes 9 and 10). These reductions in binding mirrored the reductions in SRF binding. In comparison to the wild-type promoter, the mutant promoter showed reduced levels of both SRF and FOXK1 binding (A, lanes 7 and 8). Thus, SRF is required for efficient recruitment of FOXK1 to the promoter.
To establish whether the same mechanism is operative , ChIP analysis was used to monitor the occupancy of the promoter upon SRF depletion. The reduction of SRF levels caused the expected reduction of SRF binding to the promoter. However, the binding of FOXK1 was also compromised (B, top two panels). This reduction was not due to a general effect on promoter occupancy as the binding of an alternative transcription factor E2A () was maintained upon SRF depletion. Similar results were obtained on the promoter, where SRF depletion led to reductions in both SRF and FOXK1 binding (B, panels 3 and 4).
Together these results therefore demonstrate that SRF is required for the FOXK1 to regulate and be recruited to the promoter. Thus, FOXK1 and SRF interact both physically and functionally to control the activity of target genes such as and .
Complexes between Forkhead and MADS-box transcription factors have previously been shown to be an important common combination involved in controlling the cyclical expression of cell cycle genes in . Here, we demonstrate that this combination of transcription factors is also functionally important in human cells, adding to the repertoire of transcription factor modules that function in metazoan systems. Specifically, we demonstrate that the human Forkhead transcription factor FOXK1 functionally interacts with the MADS-box protein SRF.
FOXK1 can form complexes with SRF in the absence of DNA through binding to the minimal core DNA-binding domain of SRF which includes the MADS-box (). Indeed, a mutant form of FOXK1 that cannot bind to DNA efficiently can still control SRF-dependent promoter activity (). However, SRF is required for the efficient recruitment of FOXK1 to target promoters and ( and ). This suggests a model whereby SRF acts as a platform to recruit FOXK1 (C). FOXK1 can then repress promoter activity. These observations are fully consistent with the known roles of SRF and other MADS-box proteins in acting as a platform for the assembly of many different types of transcriptional regulatory complexes, some of which like MRTFs make minimal DNA interactions (,).
The paradigm for interactions between Forkhead and MADS-box transcription factors is the yeast Fkh2p-Mcm1p complex (). However, while there are important overall similarities between the Fkh2p-Mcm1p and human FOXK1-SRF complexes, their modes of interaction and regulation are not identical. Both Fkh2p and FOXK1 share a similar domain structure, with both possessing an N-terminally located FHA domain in addition to the Forkhead DNA-binding domain. FOXK1 is a transcriptional repressor protein, and Fkh2p can also repress transcription of its target genes during the early part of the cell cycle (). The mouse homologue of FOXK1, Foxk1/MNFβ represses transcription through the recruitment of the Sin3 corepressor () and Fkh2p can also bind to Sin3 (our unpublished data). In contrast, to date, no transcriptional activation capacity has been identified for mammalian FOXK1 proteins, and the region encompassing the FHA domain has repressive activity rather than the transactivation ability exhibited by the same region in Fkh2p (; 12,13,38). Moreover, to date, we have been unable to establish a role for the FOXK1-SRF complex in cell cycle control (our unpublished data), and instead, an alternative forkhead protein FOXM1 appears to perform the major role in controlling G2-M phase transcription in mammalian cells (). Our data indicate that FOXM1 does not function through binding and changing the activity of SRF (, data not shown).
There also seem to be important differences between human and mouse FOXK1 proteins. In mice, there are two isoforms of Foxk1/MNF, MNFα and the shorter splice form MNFβ. However, only the latter apparently shows DNA-binding and transcriptional regulatory capacity (). In contrast, full-length human FOXK1 (equivalent to MNFα) can bind to DNA and regulate transcription ( and ). Secondly, mutations designed to disrupt the phosphopeptide binding activity of the FHA domain of MNFβ partially diminished the repressive activity of MNFβ () but were without effect in FOXK1 (data not shown). It is currently unclear why these proteins apparently function differently but these observations might reflect important evolutionary differences.
To establish the importance of the FOXK1-SRF interaction, we demonstrated that this complex functions on the and genes, and that FOXK1 has a repressive role in the complex. However, FOXK1 does not seem to be an obligate partner for SRF as we could not detect FOXK1 binding to the different target gene, (). Thus, FOXK1 is likely to be a substoichimetric partner for SRF, as suggested by our inability to co-immunoprecipitate endogenous FOXK1 and SRF. Biologically, FOXK1 is likely to restrict the expression of expression in non-smooth muscle cell types such as the stem cell-like myogenic side population cells (). Recently, another forkhead transcription factor Foxo4 was shown to repress expression in proliferating smooth muscle cells (). In common with FOXK1, Foxo4 repressed transcription in a DNA-binding independent manner and achieved this through interacting with and inhibiting the SRF-myocardin activator complex. Direct interactions between Foxo4 with SRF were not however shown. A different SRF partner protein Elk-1 was also shown to inhibit the expression of a number of smooth muscle-specific genes (). However, Elk-1 is ineffective against . Thus, several different ways might have been devised to reduce the expression of different smooth muscle genes in different cell types through impacting on the activity of SRF.
In summary, we have identified a novel combination of functionally interacting human transcription factors, the Forkhead protein FOXK1 and MADS-box protein, SRF. Future studies will focus on how common this mode of SRF target gene regulation is in human cells. |
The large subunit of ribosomes comprises two RNA species, 23S and 5S rRNA, and 33 proteins (). 5S rRNA is the smallest RNA component of the ribosome, 120 nucleotides long. Its secondary structure has been extensively investigated using phylogenetic data and a suite of biophysical and biochemical techniques, including NMR spectroscopy and microarrays (,). It consists of five helices (I–V), two hairpin loops (C and D), two internal loops (B and E) and a hinge loop (A) organized in a three-helix junction.
The tertiary structure of 5S rRNA in solution is not known yet. The most detailed picture of the tertiary structure of 5S rRNA has emerged from models reconstructed on the basis of crystallographic and cryo-EM analysis of the large ribosomal subunit or the whole 70S ribosome from bacteria and archaea (). Beyond this contribution, these studies revealed that 5S rRNA, in conjunction with proteins L5, L18 and L25, forms the central protuberance of the large ribosomal subunit. Extending from there it connects the upper part of the large subunit with the peptidyl transferase (PTase) center and the binding site of elongation factor EF-G (). Consistent with this 5S rRNA topography, cross-linking and foot printing analysis (), site-directed mutagenesis studies () and screening of RNA motifs interacting with 5S rRNA () have indicated that 5S rRNA interacts with multiple functional regions of 23S rRNA in ribosomes, including helices 38 and 39, the terminal loop of helix 89, and sites between nucleosides 2272–2345 in domain V. In parallel, mutagenesis studies in 5S rRNA from (,) and functional substitution of 5S rRNA within ribosomes by antibiotics interacting simultaneously with domains II and V of 23S rRNA () demonstrated that 5S rRNA is actively involved in different ribosomal functions. Combined with early studies on the reconstitution of active 50S subunits (, ), this extensive investigation on the structural and functional role of 5S rRNA led to the hypothesis that 5S rRNA assists in stabilizing the PTase center and in facilitating communication between different functional centers of the ribosome. Nevertheless, despite the tendency in this series of experiments to attribute all of the observed effects to 5S rRNA, it is in certain cases difficult to distinguish the role of 5S rRNA from that of the associated ribosomal proteins. An alternative experimental approach to investigate the functional role of 5S rRNA is to specifically alter its folding and to correlate the resulting conformational changes with alterations in the function of the 50S ribosomal subunit.
Polyamines, like other monovalent and divalent ions, cause on 5S rRNA a conformational switch to a more compact form (). For instance, spermine has been found to stabilize the secondary and tertiary structure of 5S rRNA quite strongly, more so and in a different way than Mg (). This was attributed to the hydrogen bonds afforded by all the primary and secondary amines of spermine. Obviously, to unveil the molecular basis of spermine effect on the structure and function of 5S rRNA, the localization of polyamine binding sites on 5S rRNA is a prerequisite.
In the present study, mapping of spermine binding sites in 5S rRNA was achieved by photo-affinity labeling of 5S rRNA, either free in solution or as part of 50S subunits or whole ribosomes. In each case, the cross-linking sites were identified by primer extension analysis. As expected, binding of spermine to 5S rRNA caused changes in the conformation of 5S rRNA. The influence of these changes on several ribosomal properties was investigated.
[γ-P] ATP and [α-P] GTP were obtained from Izotop (Budapest, Hungary). L-[2, 6 H] Phenylalanine was from Moravek Biochemicals Inc. (Brea, CA, USA). [C] Spermine tetrahydrochloride was from Amerscham Biosciences Inc. (Piscataway, NJ, USA). RNase H was purchased from Promega (Madison, WI, USA), and AMV reverse transcriptase from Roche Diagnostics (Mannheim, Germany). The dNTPs and ddNTPs were from Boehringer (Mannheim, Germany). Spermine tetrahydrochloride, dimethyl sulfate (DMS), DMS stop solution, puromycin dihydrochloride, thiostrepton and heterogenous tRNA from were from Sigma (St Louis, MO, USA). Cellulose nitrate filters (type HA; 0.45 μM pore size) were from Millipore Corporation (Bedford, MA, USA), while thin plastic sheet of PEI-cellulose F were from Merck KGaA (Darmstadt, Germany). 17-deoxynucleotides, used as primers in the primer extension analysis, were synthesized by Invitrogen (Paisley, UK). Elongation factor G (EF-G) from was kindly provided by Prof. K. H. Nierhaus (Max-Planck Institute, Berlin). ABA-spermine was synthesized from methyl-4-azidobenzoimidate and spermine, and purified according to Clark . ().
Polyamine-depleted 70S ribosomes, native 50S and 30S subunits, and partially purified translation factors were prepared from K12 cells, as reported previously (). 23S rRNA, 5S rRNA and total proteins (TP50) were isolated from 50S ribosomal subunits and activated prior to their use by incubation for 20 min at 42°C in a buffer containing 40 mM HEPES-KOH, pH 7.2, 20 mM magnesium acetate and 100 mM NHCl (). Ac[H]Phe-tRNA was prepared from heterogeneous tRNA (Sigma), charged to ∼80%, 100% being 28 pmol of [H]Phe per A unit (). Post-translocation complex of poly(U)-programmed 70S ribosomes (POST-complex) bearing tRNA at the E-site and Ac[H]Phe-tRNA at the P-site, pre-translocation complex (PRE-complex) containing tRNA at the P-site and Ac[H]Phe-tRNA at the A-site and a ribosomal complex (P-complex) carrying a single tRNA in the P-site were prepared as described by Dinos . (). The state of tRNA binding (hybrid versus classical) in the prepared ribosomal complexes was established using radioactivity and puromycin-reactivity measurements, as well as chemical protection assays (). POST-complex prepared according to this experimental protocol was reactive against puromycin to ∼90%. Reconstituted 50S ribosomal subunits from TP50, 23S and 5S rRNA were prepared by a two-step incubation procedure (). When required, 5S rRNA labeled at the 3′-end with [5′-P] pCp and T4 RNA ligase () or photolabeled by ABA-[C]spermine was used in the reconstitution experiments. After reconstitution, each one of the samples was loaded on a 10–30% linear sucrose gradient in buffer A containing 100 mM Tris–HCl, pH 7.2, 6 mM MgCl, 100 mM NHCl and 6 mM 2-mercaptoethanol. The gradients were centrifuged for 6 h at 85 000, at 4°C. The percentage of total reconstitution was then estimated by processing the gradients for optical scanning at 260 nm and radioactivity counting.
Naked 5S rRNA, 50S ribosomal subunits or POST-complex were photolabeled with ABA-spermine and purified as shown previously (). The ABA-spermine cross-linking sites in 5S rRNA were determined by primer-extension analysis (), making use of the fact that by linking the photoprobe to a nucleoside, it acts as a barrier for reverse transcriptase. The DNA primer used was complementary to the 5S rRNA region encompassing nucleosides 104–120. The entire 5S rRNA, except its extreme 3′-end (nucleosides 95–120), was analyzed in this way. To check the extreme 3′-end of 5S rRNA, modified 5S rRNA with ABA-SPM was radioactively labeled at its 3′-end with [5′-P] pCp and T4 RNA ligase and then extended by a 38-oligonucleotide [5′-(A)(G)(A)-3′]. The product was purified by gel electrophoresis on an 8% (w/v) polyacrylamide/7M urea gel, excised and ethanol precipitated. Such extended 5S rRNA was analyzed by primer extension using a deoxy-nucleotide complementary to the extended arm, 5′-(T)(C)-3′, as primer. The stops of reverse transcriptase reaction were visualized on a gel autoradiogram (). Controls with untreated 5S rRNA or samples photolabeled in the simultaneous presence of a 250-fold excess of spermine were run in parallel in order to distinguish nicks in the RNA, non-specific photoincorporation or autonomous pauses of reverse transcriptase. Only bands reproduced at least three times were taken into account. To confirm the results of primer extension analysis and to search for possible photoincorporation into positions non-recognizable by reverse transcriptase, samples were first hybridized with selected pairs of 11-deoxynucleotides complementary to 5S rRNA at positions 40 nts apart, digested with RNase H, and applied to a 8% (w/v) polyacrylamide/7M urea gel for analysis (). 5S rRNA, untreated or photolabeled, was modified with DMS (), and then subjected to primer extension analysis.
The number of cross-linking sites, the dissociation constant () of the reversible complex between 5S rRNA (R) and the photoprobe (L) and the Hill coefficient were estimated as shown previously ().
To evaluate the ability of the reconstituted 50S subunits to associate with native 30S subunits, 5 A units of each subunit were incubated in buffer A for 30 min at 37°C. The mixture was then applied to a linear 10–30% sucrose gradient in buffer A, centrifuged for 6 h at 85 000 and then analyzed by optical scanning at 260 nm.
Binding of Ac[H]Phe-tRNA to the ribosomal P- and A-sites was performed in buffer B (40 mM HEPES-KOH, pH 7.2, 6 mM magnesium acetate, 100 mM NHCl and 6 mM 2-mercaptoethanol) containing or not containing 50 μM spermine, by incubating Ac[H]Phe-tRNA with poly(U)-programmed 70S ribosomes pre-filled (A-site binding) or not pre-filled (total binding) at their P-sites by tRNA (). The Ac[H]Phe-tRNA binding was measured by nitrocellulose filtration. The P-site bound Ac[H]Phe-tRNA was estimated from the total binding by titrating the resultant Ac[H]Phe-tRNA·poly(U)·70S ribosome complex with puromycin (2 mM, 2 min at 25°C).
PTase activity of the POST-complex was assessed by the puromycin reaction carried out at 25°C in buffer B. When required, 50 μM spermine was also included in the incubation mixture. The catalytic rate constant () of PTase and the affinity constant () of puromycin were estimated as described previously ().
For translocation assays, aliquots of PRE-complex (1.6 pmol) were added in 12 μl of buffer B containing 0.015 μM EF-G and 0.12 mM GTP. When desired, spermine at 50 μM was also included in buffer B. The mixtures were incubated at 25°C for specified time intervals. Translocation was monitored by titrating the resultant POST-complex with a solution of puromycin and thiostrepton (). Spontaneous translocation was measured in the absence of EF-G. In another series of experiments, increasing concentrations of EF-G were added in buffer B containing 0.12 mM GTP and PRE-complex at a fixed concentration. The reaction was carried out for 5 min at 25°C, after which translocation was again estimated by reaction with puromycin. Possible interference in the PRE-complex samples by POST-complex was measured at the start of the translocation time-course and subtracted.
To estimate the capacity of ribosomes for EF-G binding, PRE-complex was prepared either from native subunits (control samples) or from native 30S subunits and reconstituted large ribosomal subunits (). Aliquots of these complexes (1.6 pmol) were incubated at 25°C for 30 min in 12 μl of buffer B containing 1 μM EF-G, 10 μM GTP, 5 μCi of [α-P]GTP (400 Ci/mmol), and 0.5 mM fusidic acid. When desired, spermine at 50 μM was also included in the incubation mixture. Six microliters of each incubation mixture were filtered through nitro-cellulose filters, washed twice with 2 ml of ice-cold buffer B, and the radioactivity retained on the filters was determined by scintillation counting. Control experiments were performed in which ribosomes were incubated in the absence of EF-G, and the radioactivity measured was subtracted. P-complex and empty ribosomes are also capable of binding EF-G and triggering the GTPase activity of EF-G (,). Therefore, the participation of such complexes in our experimental system was quantified and used to correct the raw data, applying ,
Ribosome-dependent GTP hydrolysis catalyzed by EF-G was determined in 15 μl of buffer B containing 1.6 pmol of PRE-complex, 8.8 pmol of EF-G, 50 μM GTP and 5 μCi of [α-P]GTP (400 Ci/mmol). The mixture was incubated in the presence or absence of 50 μM spermine at 4°C for specified time intervals. An aliquot (3 μl) was withdrawn at each time-point, and the reaction in this aliquot stopped by adding 1 μl of 11 M formic acid. The samples were kept on ice for 5 min, centrifuged at 10 000 for 10 min, and then analyzed by thin-layer chromatography on PEI-cellulose F in 1 M formic acid and 1 M lithium chloride. The extent of hydrolysis was quantified using a FUJIFILM phosphoimager. [α-P]GTP hydrolysis by ribosomes solely were measured separately and subtracted. In addition, the capacity of P-complex and empty ribosomes to trigger the GTPase activity of EF-G was separately measured and taken into account in data processing.
One-way ANOVA was used to estimate the mean values, data variability and significant differences between means.
The arylazido group, ABA, is positioned 9 Å from the -amino group of spermine. In the dark, the photoprobe interacts in a reversible fashion with 5S rRNA. By irradiation at 300 nm, however, the arylazido group is converted to nitrene, which reacts rapidly and covalently with a variety of adjacent groups of the target molecule. Precise identification of the cross-linking sites was achieved by primer extension analysis, taking advantage of the fact that reverse transcriptase pauses or stops one position before an ABA-spermine-modified nucleoside. The modified nucleosides were then determined by electrophoretic and autoradiographic analysis of the reverse transciptase products. To make the above analysis possible in the extreme 3′-end region of 5S rRNA, the target molecule was extended after its labeling with ABA-spermine by ligating the 5′-(A)(G)(A)-3′ oligonucleotide at its 3′-end and using a primer complementary to this extension. A representative auto-radiogram obtained by primer-extension analysis is shown in A. The cross-linking sites of ABA-spermine in 5S rRNA, either free in solution or assembled into 50S subunit and POST-complex, are summarized in a secondary structure model of 5S rRNA (B). Detailed presentation of the 5S rRNA nucleosides labeled under various conditions is given in Supplementary Table 1. The authentic character of the cross-linking sites was confirmed by RNase H cleavage of 5S rRNA modified with ABA-[C] spermine, in the presence of selected pairs of 11-deoxynucleotide complementary to sequences located 40 nts apart in the primary structure of 5S rRNA (Supplementary Table 2).
Except for helix I, ABA-spermine cross-linking was observed throughout the 5S rRNA molecule in both helices and loops of the secondary structure, but it was detected more frequently in helix III and loop C (B). The primarily labeled nucleosides in single-stranded regions were A and C (84%), while in double-stranded regions the label was more equally distributed among the four nucleosides. In most cases, reverse transcriptase stopped at a single nucleoside. Only three exceptions were recorded, in which two adjacent nucleosides were modified; the one closer to the 3′-terminus corresponded to a band with stronger intensity. Such a profile is probably due to a ‘stuttering’ of reverse transcriptase, observed also previously (,). Evidence for site-specific photo-incorporation was sought by running in parallel control experiments, in which spermine was added in excess in the incubation mixture during photolabeling. As shown in A, the authentic stops of reverse transciptase are essentially abolished under such conditions. As detected by additional experiments, spermine exhibited the best competition potency, compared with spermidine and putrescine. Monovalent ions, such as Na or NH did not compete, while Mg decreased the labeling of 5S rRNA in a specific and dose-dependent manner. At 10 mM Mg, a 10% reduction in photolabeling was recorded; nucleosides U32, A73, A52, A53, C71, A73 and U80 were the cross-linking residues most affected by Mg. In contrast, the cross-linking pattern for the remaining positions was not significantly altered.
An important feature of B is that the cross-linking pattern depends on the photoprobe concentration and the assembly status of the target molecule. With regards to the former factor, an increase of ABA-spermine concentration from 50 to 300 μM enhanced the susceptibility of A52, A53 and G56 to the photoprobe, while it reduced the cross-linking at nucleosides G67 and A73. In addition, the increase of photoprobe concentration resulted in enrichment of cross-linking by additional sites dispersed throughout the 5S rRNA molecule. Assembly of 50S subunit resulted in protection of several positions from cross-linking by ABA-spermine, especially of nucleosides belonging to helices III and V and to loops C, D and E. In contrast, it favored cross-linking to helix II and loop B. Finally, when 50S subunit was incorporated into the ribosomal POST-complex, a conspicuous change in the susceptibility of several nucleosides against ABA-spermine was recorded. Namely, nucleosides A46, C47, A52, A59, C62, G69 and A73 were freshly protected, while C42 and A108 exhibited increased protection.
Chemical probing experiments using DMS indicated that the protection pattern of 5S rRNA changes, when it is incorporated into 50 subunits. In agreement with previous observations (), nucleosides C11, A15, C35, A57, A58, A59, A66, A73 and A99 became protected, while nucleoside A78 became more accessible. Formation of the ribosomal POST-complex reduced the susceptibility of nucleosides C38 and A45 against DMS, but increased the reactivity of A15 and A52. On the other hand, cross-linking of ABA-spermine induced further alterations in the reactivity of 5S rRNA towards DMS by increasing the accessibility of some positions and decreasing the accessibility of others. The results of the latter experiments are summarized in .
Photolabeling data at various concentrations of ABA-spermine established a sigmoidal hyperbola (). Therefore, they were fitted to ,
In , [] is the concentration of the free photoprobe, is the amount of the covalently bound photoprobe to naked 5S rRNA, is the value of at saturation conditions, is the Hill coefficient, and is the overall dissociation constant of the encounter complex between the photoprobe and 5S rRNA. From the value of and the amount of 5S rRNA used in the assay, we estimated a value for the number of cross-linking sites equal to 6.35.
The crude product derived from reconstitution experiments using photolabeled 5S rRNA, TP-50 and wild-type 23S rRNA was found to exhibit decreased capability to form 70S ribosomes upon association with native 30S ribosomal subunits, compared with that obtained by utilizing wild-type 5S rRNA. Two factors may cause such an effect. First, photolabeling of 5S rRNA molecule may directly influence the ability of the reconstituted particles to associate with 30S subunits. Second, reduced ability could be the result of an impaired reconstitution, and thus of a decrease in the net fraction of the 50S subunits. In order to discriminate between these two possibilities, the crude product was subjected to sucrose gradient centrifugation and the yield of reconstitution was estimated. As indicated in , ribosomal particles deprived of 5S rRNA sedimented at about 47S. Such particles, upon association with native 30S subunits, led to the formation of 62S ribosomal complexes. When 5S rRNA photolabeled with 50 μM ABA-spermine was used as a component of the reconstitution mixture, both 50S and 47S particles were obtained, thus resulting in a decreased yield of 50S subunits. However, isolated 50S subunits from this mixture were fully active at associating with native 30S subunits and forming 70S ribosomes. In contrast, photolabeling of 5S rRNA with 300 μM ABA-spermine prior to its use in the reconstitution experiments, disturbed more dramatically both the yield of 50S subunits and their capability to associate with native 30S subunits.
When 70S ribosomes obtained from reconstituted 50S subunits and native 30S subunits were poly(U)-programmed and incubated with equimolar amount of AcPhe-tRNA, their capacity for binding to the P- and A-site was found 0.171 and 0.092, respectively. AcPhe-tRNA binding to both sites was remarkably improved in the presence of 50 μM spermine (). At higher molar ratio of AcPhe-tRNA to ribosomes (2:1) or at higher concentrations of spermine (300 μM), the binding was further improved. In this series of experiments, wild-type 5S rRNA was used in the reconstitution of 50S subunits. Next, 70S ribosomes containing 5S rRNA photolabeled with 50 μM ABA-spermine were prepared and tested under identical conditions, but in the absence of free spermine. As shown in , cross-linking of ABA-spermine to 5S rRNA caused a slight, but statistically significant improvement in the binding properties of the modified ribosomes. Moreover, cross-linking of ABA-spermine to 5S rRNA stimulated by 30% the catalytic rate constant of PTase, without affecting essentially the affinity of ribosomes towards puromycin. In agreement with previous studies (,), 70S ribosomes deprived of 5S rRNA displayed extremely low capability to bind AcPhe-tRNA, particularly at the A-site. These ribosomes also exhibited impaired PTase activity ().
Two species of PRE-complex were tested for their efficiency to translocate AcPhe-tRNA from the A- to the P-site: one reconstituted from wild-type 5S rRNA, and the other reconstituted from 5S rRNA photolabeled with 50 μM ABA-spermine. For the construction of PRE-complexes, poly(U)-programmed 70S ribosomes were incubated for 20 min at 37°C with tRNA (molar ratio to ribosomes 2 : 1) in order to pre-fill the P-site. Chemical protection assays verified that tRNA sampled the classical P/P binding state. Using [P]tRNA, we found that occupation of the P-site in this complex (P-complex) ranged from 40% to 75%, depending on the 5S rRNA species (photolabeled or wild-type) and the ionic conditions used. Subsequently, Ac[H]Phe-tRNA was added (molar ratio 2:1) and incubated for an additional 20 min at 37°C to allow non-enzymatic A-site binding (PRE-complex). Ribosomal complexes prepared in this way were reactive against puromycin to ∼20%. In addition, chemical protection assays indicated very low, <10%, but measurable protection of nucleoside in 23S rRNA (E-site) against DMS, while radioactivity measurements detected no release of the bound [P] isotope from the ribosome. This means that more than 80% of the bound AcPhe-tRNA sampled the classical A/A binding state. Our results are in agreement with previous observations that -protected aminoacyl-tRNAs bound in a pre-translocation state on the ribosome generally occupy the classical A/A site (). The small amount of AcPhe-tRNA which was reactive toward puromycin, probably translocating spontaneously from the A/A to the P/P site during preparation of the PRE-complexes, was measured in each case and taken into account in data processing. Translocation was studied in buffer B containing 0.12 mM GTP, in the absence or in the presence of 50 μM spermine. Spermine 50 μM was chosen on the basis of previous findings that such a concentration is optimum for translocation (). EF-G at 15 nM promoted relatively fast translocation that was almost complete within 2 min, independently of whether free spermine was present or absent. In the absence of EF-G, however, translocation was much slower and proceeded in a linear fashion at least up to 30 min. Spermine either free or covalently bound to 5S rRNA improved spontaneous translocation (A). To examine the effect of spermine on EF-G requirements for efficient translocation, increasing amounts of EF-G were added in buffer B containing PRE-complex at fixed concentration, and translocation was allowed to proceed for 5 min at 25°C. As shown in B, the extent of translocation in each case followed a hyperbolic curve reaching the same plateau at high concentrations of EF-G. Nevertheless, the concentration of EF-G at which 50% translocation was achieved, differed. Namely, PRE-complexes exposed to free spermine or possessing covalently bound spermine at their 5S rRNA required lower concentrations of EF-G for efficient translocation.
The sparing effect of spermine on EF-G-requirements could be explained either by a beneficial effect of spermine on the binding of EF-G to ribosomes, and/or by an influence of spermine on the ribosome-dependent GTP hydrolysis catalyzed by EF-G. To clear this point, we applied the following two approaches: First, we used fusidic acid, an antibiotic that binds to the EF-G·GDP-ribosome complex and prevents dissociation of EF-G·GDP from the ribosome. Under these conditions, the EF-G·GDP-ribosome complex is stalled in a post-translocation state (). Employment of [α-P] GTP in the assay allows the indirect estimation of EF-G binding, by measuring the trapped radioactivity on ribosomes relative to controls incubated in the absence of EF-G. The results from these experiments are summarized in . It is evident that exposure of whole PRE-complexes to free spermine or covalent attachment of spermine to their 5S rRNA promotes EF-G binding. Similar treatment of P-complex or empty ribosomes results in analogous improvement of EF-G binding. Nevertheless, it is apparent from the data shown in that the interaction of EF-G with the ribosome is generally stimulated when deacylated tRNA occupies the P-site. In agreement with previous observations (), ribosomes deprived of 5S rRNA were almost inactive to bind EF-G. Second, the capacity of PRE-complex to activate hydrolysis of GTP by EF-G was determined. As shown in , the initial velocity of the reaction was found to be almost identical for either ribosomes reconstituted with wild-type or ribosomes containing photolabeled 5S rRNA, and similar to that possessed by native ribosomes, regardless of whether free spermine was present or not in the reaction mixture [compare panels (A) and (B) in ]. Similar behavior was exhibited by P-complex and empty ribosomes (C). In agreement with previous studies (,), empty ribosomes or PRE-complex were less active in stimulating the GTPase activity of EF-G, compared with P-complex. Interestingly, ribosomal complexes deprived of 5S rRNA exhibited considerably reduced efficiency to activate the GTP hydrolysis by EF-G, much less than that reported by Dohme and Nierhaus ().
In the present study, mapping of spermine binding sites in 5S rRNA was achieved by a photoaffinity labeling approach combined with primer extension analysis. The rationale is that attachment of ABA-spermine to a nucleoside acts as a barrier for reverse transcriptase. This technique bypasses several disadvantages associated with other cross-linking methods utilizing homobifunctional reagents (,). It has been successfully applied so far for mapping spermine binding sites in AcPhe-tRNA free or bound to the P-site of ribosomes (), as well as in ribosomal proteins () and 16S and 23S rRNA naked or incorporated into ribosomes (,). It should be mentioned that, apart from monovalent and divalent cations, polyamines too are important components of the ionic environment of ribosomes, playing an essential role in the structural integrity of ribosomes and hence, in translational accuracy and efficiency (). Although the existence of spermine in cells is questionable (,), accumulated evidence supports the notion that almost all of the cellular functions of the naturally occurring polyamines can be fulfilled by spermine (). On the other hand, due to its four positive charges and the hydrogen bonds afforded by its primary and secondary amines, spermine is the most effective polyamine in stabilizing the RNA folding. Therefore, it can be experimentally used in micromolar concentrations, a very advantageous fact in bypassing artifacts associated with non-specific binding, frequently encountered in the application of photolabeling techniques.
The number of ABA-spermine cross-linking sites in 5S rRNA, determined in the present study by primer extension analysis, is higher than that calculated by Hill-plot analysis. This contradiction can be explained by the fact that apparently different cross-links localized to adjacent positions in the structure of 5S rRNA may represent the same binding site. To some degree, non-specific binding may also account for it. In fact, as revealed by competition experiments using natural polyamines () or Mg ions as competitors, most of the cross-linking sites appear to be specific. The remaining sites are placed on the cytosolic surface of the large ribosomal subunit (C71), at the top edge of the central protuberance (U32, A52, and A53) or are located in defined electro-negative pockets (A73, U80). Nevertheless, independent of the specificity, polyamine binding to certain regions may be beneficial for the stabilization of the 5S rRNA tertiary structure and its communication with functional regions of 23S rRNA. Noteworthy is the binding of spermine within and around loop E, a region implicated in the binding of ribosomal protein L25. Crystal structure analysis of 5S rRNA fragments encompassing the loop E sequence, alone () or in complex with protein L25 (), has demonstrated that loop E contains seven non-Watson–Crick base-pairs stabilized by several Mg ions. Interestingly, molecular dynamics simulations revealed extensive binding of monovalent ions in the 5S rRNA loop E motif, even in the presence of 4–5 mM Mg (,). Most of the monovalent ion binding sites coincide (G98, U80) or are in close proximity (G105, G72, U77) to cross-linking sites of ABA-spermine. Another issue of importance is related to the occupancy of distant sites from the symmetry center (the G75-A101 pair) by freely diffusing ligands, like charged chemical groups belonging to proteins or drugs. Indeed, in a complex of L25 with a 5S rRNA fragment studied by Lu and Steitz (), an amino group of L25 binds close to U80. Therefore, polyamine binding to this site may modulate directly the interaction of 5S rRNA with protein L25. In contrast, sites located close to the symmetry center display high electronegative potential. This implies that ions associated with these regions may be more difficult to displace. In line, we observed that binding of ABA-spermine to A73 is highly resistant to competition by monovalent cations. Other examples of cross-linking, which may interfere in protein binding, are related to helices II and III, and loops B and C. Crystallographic studies () as well as foot printing and cross-linking data [reviewed in ()] have indicated that these regions constitute the primary binding sites for proteins L5 and L18. Detailed analysis of ribosomal protein L5 in complex with a 34-nt fragment comprising helix III and loop C of 5S rRNA revealed that nucleosides C42 and C43 are involved in both hydrophobic interactions and hydrogen bonding with no polar and side chain atoms of L5, respectively (). It is noteworthy that these residues of loop C are susceptible to ABA-spermine cross-linking, regardless of the 5S rRNA assembly status or the photoprobe concentration.
Photolabeling of 5S rRNA with 50 μM ABA-spermine causes profound changes in the susceptibility of several nucleosides in 5S rRNA against DMS (). Therefore, it is tempting to suggest that binding of spermine to 5S rRNA promotes changes in its folding. Namely, loop A adopts an apparent ‘loosening’ of its structure, while loops C, D, E and helix III achieve a more tight structure. These findings are consistent with previous observations obtained by means of calorimetric studies (,). Photolabeling of 5S rRNA with 300 μM ABA-spermine results to a conformer, which does not meet the operational definition of the ‘A’ form (). 50S subunits reconstituted from 5S rRNA modified in this way associate with native small ribosomal subunits at low yield () and are functionally inactive.
The yield of the 50S rRNA subunit reconstitution obtained by 5S rRNA photolabeled with 50 μM ABA-spermine was found to be approximately half that achieved with wild-type 5S rRNA. Kakegawa . () indicated that the reconstitution of 50S particles from 23S rRNA, 5S rRNA and TP50 is not influenced greatly by polyamines. On the other hand, there is evidence that 5S rRNA undergoes conformational changes during its assembly into the central protuberance of the 50S subunit (). In agreement, we observed that some of the ABA-spermine crosslinks in 5S rRNA get more pronounced in the 50S subunit than in the isolated state, clearly indicating that 5S rRNA undergoes conformational changes upon incorporation into ribosomes. Therefore, it is tempting to suggest that ABA-spermine cross-linking, although not directly affecting the reconstitution process, may hinder the required conformational changes by stabilizing a rigid conformation in 5S rRNA. In support of this hypothesis, labeling of 5S rRNA with 300 μM ABA-spermine reduces further its incorporation into 50S subunits. In contrast to 50S subunits containing 5S rRNA photolabeled with ABA-spermine at 300 μM, those assembled from 5S rRNA photolabeled with ABA-spermine at 50 μM were fully active in associating with native 30S subunits.
The loop B →loop C arm in 5S rRNA is the most affected region by ABA-spermine photolabeling. A great body of experimental data has previously demonstrated that this region participates in several signal transmission chains connecting 5S rRNA with P-site-bound tRNA. For instance, this region binds protein L5 which interacts with the T-loop of P-site bound tRNA (,,). Our data show that photolabeling of 5S rRNA with 50 μM ABA-spermine results in a subtle but visible stimulation of the AcPhe-tRNA binding to the P-site. Binding to the A-site is not affected, albeit loop B →loop C arm is near the tip of H38 of 23S rRNA () which communicates via protein S13 of the small subunit (bridge B1a) with the anticodon region of A-site-bound tRNA. In fact, several studies revealed that loss of contacts between protein S13 and H38 has little consequences on tRNA binding (). Whole ribosomes containing either wild-type or photolabeled 5S rRNA, exhibit greater capacity for AcPhe-tRNA binding to the A-site in the presence of free spermine than in its absence. In this case, however, 30S subunit and AcPhe-tRNA also interact with spermine, a fact that favors tRNA binding (). In agreement with previous findings (), we observed that omission of 5S rRNA from the large subunit suppresses AcPhe-tRNA binding, in particular to the A-site. Such ribosomes, even though provided with free spermine, cannot recover from the lack of 5S rRNA.
Ribosomes containing photolabeled 5S rRNA exhibit 30% higher PTase activity than those containing wild-type 5S rRNA. Catalysis of peptide bond formation is mediated almost entirely through precise alignment of the A- and P-substrates within the active center, coupled to substrate-assisted catalysis (). Therefore, it is tempting to suggest that besides the effect of the 5S rRNA photolabeling on the affinity of P-site, the signals transmitted from 5S rRNA via protein L5 to the P-site may also affect the functional positioning of AcPhe-tRNA within the catalytic center. Since the affinity of puromycin for the A-site is not altered upon 5S rRNA photolabeleling, analogous effects on the A-site cannot be postulated. Ribosomes deprived of 5S rRNA retain 10% of their PTase activity, in agreement with previous observations according to which 5S rRNA enhances ribosomal activity (,) but is not absolutely essential for it ().
Spontaneous translocation is an inherent property of the ribosome itself, even though extremely slow (,). This is verified by the present work, also showing that reversible interaction of spermine with whole ribosomes or covalent binding of ABA-spermine to 5S rRNA stimulate EF-G-independent translocation. This prompted us to suggest that certain changes in the folding of ribosomal RNA may be beneficial to the translocation of tRNAs. Supporting evidence is provided by DMS-protection results, indicating that the region around the G1338-U1341 ridge in the small-subunit head and A790 terminal loop in the platform achieves a more ‘open'structure upon ABA-spermine photo-incorporation. Crystallographic studies in ribosomes have postulated that the gap left between the G1338-U1341 ridge and A790 operates as a ‘lock’, controlling the passage of the P-site tRNA anticodon stem-loop (,). Steady-state data derived in the present work show that translocation in the presence of high concentrations of EF-G and GTP is fast, even in the absence of polyamines. Therefore, monitoring the ABA-spermine cross-linking effect on EF-G-dependent translocation is impossible, without using rapid kinetic techniques. Following an alternative approach, we found that interaction of spermine with the whole ribosome or with its 5S rRNA has a sparing effect on EF-G requirements. Further experiments revealed that photoincorporation of ABA-spermine into 5S rRNA does not affect the initial rate of EF-G catalyzed GTP hydrolysis by 70S ribosomes, but enhances the binding of EF-G to ribosomes. In agreement with previous observations (,), we found that ribosomes deprived of 5S rRNA retain some EF-G-dependent GTPase activity but are unable to bind efficiently EF-G, even in the presence of spermine. The EF-G binding site in ribosome consists of the GTPase-associated center (GAC) located in domain II of 23S rRNA (helices H42-H44) and the sarcin–ricin loop (SRL) located in domain VI of 23S rRNA (helix H95) (). Direct contacts of 5S rRNA with GAC, SRL or EF-G have not been identified so far in ribosomes. Nevertheless, accumulated evidence suggests that several allosteric signal transmission pathways set 5S rRNA off coordinating the PTase reaction with subsequent EF-G·GTP binding and GTP hydrolysis (,).
In conclusion, a comprehensive view of the 5S rRNA nucleotide residues involved in spermine binding is obtained by the present study. The results suggest that binding of spermine to 5S rRNA causes conformational changes in certain regions of the molecule. Consequently, these changes have beneficial effects on the ribosome functioning. The general improvement observed in various functional tests is an important argument that the identified sites in 5S rRNA may have an important functional relevance to the overall active structure of the ribosome.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Telomerase is a ribonucleoprotein (RNP) that is responsible for maintaining the terminal repeats of telomeres in most organisms (). The catalytic core of telomerase consists minimally of two components: an RNA in which the template is embedded (named TER in general and TLC1 in the budding yeast ), and a reverse transcriptase (RT)-like protein that mediates catalysis (named TERT in general and Est2p in yeast). In addition, telomerase from different organisms have been shown to possess a number of accessory or regulatory subunits that promote telomerase RNP assembly, recruitment and activity.
Though the template region of telomerase RNA typically contains no more than two copies of the telomere repeat, telomerase is known to add long tracts of telomeric DNA to a given primer following a single binding event (,). The enzyme is also capable of extending telomeres by multiple repeats in a single cell cycle (,). The ability of telomerase to mediate repetitive copying of the template RNA requires two types of processivity: nucleotide addition processivity (NAP), defined as propensity of the enzyme to add nucleotides successively as the template RNA is moved through the active site; and repeat addition processivity (RAP), defined as propensity of the enzyme to initiate another round of copying when the active site reaches the 5′ boundary of the template (,). RAP has been postulated to require an interaction between telomerase and the more 5′ region of DNA primer, one that is distinct from the interaction at the catalytic site (). This ‘anchor site’ interaction is thought to allow telomerase to remain bound to DNA in between rounds of template copying, when the DNA product presumably unpairs from the RNA template. A substantial body of evidence now implicates an N-terminal domain of TERT (known variously as GQ, RID or TEN) as the anchor site. For example, mutations in this domain of yeast and human TERT selectively impair RAP and alter the ability of telomerase to utilize primers in a length- and sequence-dependent manner (). In addition, this region of yeast and TERT can bind DNA with low affinity when recombinantly expressed and purified (,). Finally, the same domain of yeast and TERT can be crosslinked to the 5′ region of the DNA primer in the context of a telomerase-primer complex (,). However, the detailed molecular mechanisms by which anchor site promote DNA binding and continuous polymerization is not understood.
Recently, Jacobs crystallized and determined the structure of the N-terminal domain of TERT (). The structure revealed a single globular domain with a novel fold consisting of 4 β strands and 7 α helices. Well-conserved residues were found to be scattered on two surfaces. Further biochemical analysis demonstrated the importance of an invariant Gln residue in promoting telomerase activity and DNA binding. Somewhat surprisingly (in light of the ‘anchor site’ notion), mutating this Gln residue (and several other residues) severely reduced overall activity of telomerase , but did not impair its processivity. Thus, the study not only provides the first detailed structural framework for analyzing a critical telomerase protein domain, but also raises interesting questions concerning the precise mechanism of this domain. In the current report, we build on the observations of Jacobs and constructed a model of the homologous domain in yeast TERT. Electrostatic calculation revealed two positively charged surface patches, both of which were found to be functionally important through site-specific mutagenesis analysis. Importantly, mutations of residues lining one of the patches (including the invariant Gln and several non-conserved residues) specifically impaired the RAP of yeast telomerase. Even a mild reduction in RAP is correlated with significant telomere loss. Moreover, we found that a previously identified point mutation in the anchor site domain of yeast TERT simultaneously enhanced RAP of telomerase and provoked telomere elongation (). Our data reveal specific molecular features of the yeast TERT N-terminal domain that are required for function, and suggest that telomerase anchor site has evolved species-specific residues to interact with species-specific telomere repeats. The results also reinforce the importance of telomerase processivity in regulating telomere length.
PSI-BLAST searches were carried out using yeast TERT as the query on the non-redundant protein sequence database at NCBI (). Five iterations were run using an -value cut off of 0.0001. The sequence hits were compiled into a multiple-sequence alignment using ClustalW v1.83, from which very remote homologs were removed and only the known telomerase reverse transcriptase sequences were selected (). This purged alignment was then used to create a sequence-based profile to which the template (PDB Id: 2B2A) was aligned, creating a structure-to-profile alignment. Pairwise sequence alignment of the template with the query sequence was retrieved from this structure-to-profile alignment for use in homology modeling.
The homology modeling program NEST was used to construct the yeast model. In particular, the template has two gaps of 10 and 6 residues, respectively, with respect to the query sequence. The equivalent regions of the yeast model were filled with the loop building function in NEST. The final model was evaluated by the programs Verify 3D and Prosall, which score structures according to how well each residue fits into its structural environment based on criteria derived from statistical analysis of high-resolution structures in the PDB (). The yeast model was judged to be of good quality and found to superimpose well to the template with an RMS deviation for the backbone atoms of 0.8 Å (Supplementary Figure 1C). The electrostatic properties of the TERT GQ domains were calculated with a modified version of the program Delphi and visualized in the program GRASP, as previously described (,).
Plasmid pRS426-EST2 was made by cloning a PCR fragment containing EST2 and 400 bp of flanking sequences in between the Bam HI site and Sac I site of pRS426. Plasmid pSE-EST2-C874, containing a protein A-tagged gene, has been described (). This fully functional Est2p is designated wild type telomerase throughout the text. All substitution mutations in the GQ domain of were generated by using the QuikChange protocol (Stratagene), appropriate primer oligonucleotides and pSE-EST2-C874 as template. All point mutations were confirmed by sequencing. A ‘plasmid shuffle’ strategy was used to assess the functionality of mutants. First, pRS426-EST2 was introduced into an -Δ strain to restore telomere length. Subsequently, the pSE-EST2 series of plasmids were transformed into the strain. The strains bearing both plasmids were grown briefly in YPD and then streaked on 5-FOA-containing plates to select for clones that have lost pRS426-EST2. The colonies were then re-streaked multiple times and monitored for growth defects and telomere length alterations.
Chromosomal DNAs were isolated from successive streaks of yeast clones bearing wild type and mutant genes using the ‘Smash and Grab’ protocol, digested with PstI, and electrophoretically separated on a 0.9% agarose gel. Following capillary transfer to nylon membranes, telomere-containing fragments were detected by hybridization with a P-labeled poly(dG-dT) probe ().
Whole cell extracts and IgG-Sepharose purified telomerase were prepared as previously described (,). Each primer extension assay was carried out using 15 μl of IgG-Sepharose pretreated with 4 mg of protein extract, and was initiated by the addition of a 15 μl cocktail containing 100 mM Tris–HCl, pH 8.0, 4 mM magnesium chloride, 2 mM DTT, 2 mM spermidine, primer oligodeoxynucleotides and varying combinations of labeled and unlabeled dGTP and dTTP. Primer extension products were processed and analyzed by gel electrophoresis as previously described. The oligodeoxynucleotide primers used for telomerase assays were purchased from Sigma-Genosis and purified by denaturing gel electrophoresis prior to use. The primers have the following sequences: TEL15, TGTGTGGTGTGTGGG; OXYT1, GTTTTGGGGTTTTGGG; TEL15(m4,5), TGTGTGGTGTCAGGG.
For determination of the processivity of substitution mutants, assays were performed using OXYT1 or TEL15(m4,5) as primer. The signal for each product was determined by PhosphorImager (Molecular Dynamics) and normalized to the amount of transcript by dividing against the number of labeled residues. The OXYT1 primer ends in 3 Gs, and can align to only one site along the yeast RNA template, thus supporting the addition of a 7-nucleotide sequence (TGTGGTG) up to the 5′ boundary of the template. For nucleotides beyond the +7 position, we assume a sequence of TGTG … (Calculations were made assuming other compatible sequences, and the conclusions were not altered.)
RAP is defined as P, i.e. the processivity at the point of translocation.
The levels of protein A-tagged yeast TERT in cell extracts were determined by enrichment on IgG-Sepharose and subsequent western blotting as previously described (). The levels of TERT-associated TLC1 RNA were determined by semi-quantitative RT-PCR using 2X RT-PCR Master Mix (USB Corp.) and primers designed to amplify an ∼300 bp product (TLC509: GCAAAGTTTGCACGAGTT and TLC793R: CTTTTTGTAGTGGGATTTATTC).
To engage in a structure-based analysis of the yeast telomerase GQ domain, we first constructed a homology model of this domain using the structure of domain as the template. The alignment of the template and target sequence is a critical step in generating a high-quality model, the accuracy of the modeled structure being greatly dependent on the quality of the alignment (). Because the level of sequence homology between these domains of yeast and TERT is only 14%, we used a structure-to-profile alignment approach (which incorporates sequence information from many family members) to obtain the optimal sequence alignment (see Materials and Methods Section and Supplementary Figure 1). The pairwise sequence alignment between the yeast TERT GQ domain and the template was then extracted from the structure-to-profile alignment and used in the modeling program (Supplementary Figure 2A). This alignment approach has demonstrated success in providing the basis for generating high-quality homology models for proteins with limited sequence similarity to the template (,). For model building, we utilized the NEST program, which employs rigid-body assembly coupled with loop modeling (). A recent comparative evaluation of commonly used modeling programs indicates that NEST performs better than or similarly to other programs (). The final yeast model of the GQ domain was evaluated with Verify 3D and ProsaII and judged to be of reasonable quality (data not shown). The model also superimposed well onto the template with an RMS deviation for the backbone atoms of 0.8 Å (Supplementary Figure 2B). Notably, the yeast model can readily account for some of the prior mutagenesis results. For example, both W115 and G123, which were previously demonstrated to be essential for telomerase function, are likely to be important for the structural integrity of this domain because the former packs into the interior and the latter is located within a tight turn, as glycines often are (B).
We then used electrostatic calculation to identify potential DNA-binding sites on the surface of the yeast domain (). This analysis revealed two positively charged patches separated by a ridge on one face of the domain (). Comparable patches can be observed on the surface of the domain (data not shown). The top patch in the yeast domain (named patch 1) is lined by several basic residues including K111, K116 and H119. The bottom patch (named patch 2) contains the invariant Gln (Q146, which is equivalent to Q168 in TERT) as well as R151, N153 and H156. Notably, none of these residues with the exception of Q146 are conserved in .
Based on modeling and electrostatic calculation, we chose nine residues in the GQ domain of yeast TERT for detailed mutagenesis analysis. Seven of these are located on either patch 1 (K111, K116 and H119) or patch 2 (Q146, R151, N153 and H156). G149 and C152 are also included in the analysis for comparative purposes. G149 is universally conserved and likely to be functionally important. C152 is located next to patch 2 but points away from it, and may therefore be unimportant. A ‘plasmid shuffle’ strategy was used to generate a set of isogenic yeast strains containing plasmid-borne wild type or mutant genes (see Experimental Procedures Section). To facilitate detection and telomerase isolation, the genes were all fused to a protein A tag. The strains were propagated by repeatedly restreaking for single colonies on plates, and then assessed for growth and telomere defects.
As shown in , of the 10 mutants, only Q146A and G149A (Ala substitution of invariant residues) exhibited severe growth defects by the third re-streak (∼80–100 generations after loss of wild type EST2). The Q146E mutant failed to show significant growth defects, suggesting that this substitution is less disruptive. Not surprisingly, all three mutants also manifested severe telomere loss (B). Interestingly, despite their lack of growth defects, 4 of the other 7 mutants (K111A, H119A, R151A and N153A) also manifested moderate to severe telomere attrition, indicating that these residues are also functionally important. The remaining three mutants (K116A, C152A and H156A) appear to be functionally intact in their ability to maintain wild type telomere lengths.
To investigate the basis for the observed telomere maintenance defects, we first measured the levels of Est2 protein in the mutant strains. Cell extracts were prepared from the strains and the protein A-tagged Est2p isolated on IgG-Sepharose. The tagged Est2p were then detected using antibodies directed against protein A. As shown in A (top panel), different mutations in the GQ domain had different impacts on Est2p protein levels in cell extracts, with H119A and G149A causing the most severe reduction and K111A, Q146A and Q146E causing moderate protein loss. We next assessed the levels of Est2p-associated TLC1 RNA using RT-PCR, and found a correlation between the levels of Est2p and TLC1 RNA in IgG-Sepharose precipitates (A, bottom panel). For example, both the H119A and G149A samples showed very low levels of TLC1 RNA, whereas the K111A, Q146A and Q146E samples contained slightly higher levels (but still significantly reduced in comparison with the wild type sample). These findings suggest that mutations in the GQ domain can have a destabilizing effect on either the TERT protein or the telomerase complex.
To further examine the effect of GQ domain mutations, we assayed the levels of telomerase primer extension activity in IgG-Sepharose precipitates using a 15-nt primer with a canonical yeast telomere repeat sequence (B). As demonstrated previously by us and others, yeast telomerase is relatively non-processive on this primer and can complete mostly one round of repeat addition, adding up to 7 nt to the starting primer. Again, the amounts of primer extension activity correlated with the levels of Est2p and TLC1 RNA in the IgG-Sepharose precipitates. Thus, none of the mutations seem to impair greatly the catalytic competence of an assembled telomerase complex.
In several previous studies, we demonstrated that yeast telomerase becomes more processive on primers that contain non-yeast telomere sequences (,,). The use of such primers allowed us to identify residues in Est2p that specifically impair RAP. We therefore subjected the newly constructed GQ domain mutants to telomerase assays using one such primer, named OXYT1 (A). As expected, wild type telomerase became more processive on this primer and extended it by as many as 18 nt. Interestingly, mutants that exhibited reduced RAP (Q146A, Q146E, R151A, N153A) were all located on patch 2. The alterations for all of the affected mutants were qualitatively similar and can be visualized by comparing the product distribution of the wild type, R151A and N153A enzymes (B and C). During the first round of synthesis, the R151A and N153A mutant appear to be slightly more proficient at adding nucleotides successively such that a greater proportion of the extension products reached the +7 length. However, the fraction of products that were further extended by the mutant was significantly reduced, indicative of a clear defect in repeat addition. More specifically, while the WT, R151A and N153A telomerase generated similar amount of +7 product (pre-translocation), the mutant synthesized 50–70% less longer products (post translocation, shaded region). Quantitative analysis revealed statistically significant reductions (∼2-fold) in RAP for all of the affected mutants (D). In contrast, the K111A, K116A, C152A and H156A mutant were found to have normal RAP. The processivities of the H119A and G149A telomerase were not assessed quantitatively due to low product intensities. To examine the generality of the RAP defect, we performed an additional series of telomerase assays using primer TEL15(m4,5), which also contained non-yeast telomeric sequence. Again, the R151A and N153A mutation caused significant RAP loss, whereas the C152A and H156A mutation had no effect (Supplementary Figure 3). Together, these findings indicate that residues in the positively charged patch 2 are important for the RAP of telomerase.
In addition to loss of RAP, deletion of the GQ domain was shown previously to result in telomerase with reduced ability to utilize primers [such as OXYT1 and TEL15(m4,5)] that form short hybrids with telomerase RNA (). To determine if the substitution mutations caused the same defect, we calculated for each mutant the ratio of its activity on the OXYT1 primer to its activity on the TEL15 primer (which can form a long hybrid with telomerase RNA) and normalized the ratio to the wild type enzyme (E). Interestingly, we found that all of the patch 2 mutants manifested relative reductions in their activities on the OXYT1 primer. The magnitude of decrease ranged approximately from 2 to 4-fold. A relative reduction in activity was observed as well for the R151A and N153A mutant when the assays were performed using the TEL15(m4,5) primer (Supplementary Figure 3). Thus, the patch 2 mutants evidently exhibited both the primer utilization and processivity defects associated with deletion of the entire GQ domain.
Ji and colleagues recently described a mutation (E76K) in the anchor site domain of EST2 that caused telomere elongation. Based on their studies, they suggest that the mutant is not altered with respect to enzyme activity, but rather abnormal in -mediated regulation (). However, in their analysis, Ji and colleagues used primers that did not support RAP, making it difficult to determine if the mutation affected this aspect of telomerase property. We therefore decided to re-examine the and functions of this mutant using our system. Consistent with earlier findings, we observed substantial telomere elongation in yeast strains contain the E76K allele (A). More interestingly, we found that the E76K telomerase exhibited an ∼20% increase in RAP relative to WT telomerase (B–D). The difference is relatively modest and somewhat difficult to visualize on the PhosphorImager scan (B). However, quantitative comparison of the gel lanes showed a significant increase in the relative abundance of the second round products for the E76K enzyme (C, compare the blue and red trace in the shaded region). Moreover, the difference is reproducible and statistically significant in different reaction conditions (D). For example, the probability of initiating a second round of repeat addition on the OXYT1 primer was 0.55 for WT telomerase and 0.64 for the E76K enzyme when dGTP and dTTP were included at 50 and 0.3 µM, respectively. This difference in RAP was also observed in reactions using 0.3 µM dGTP and 50 µM dTTP (D). Notably, this kind of increase was not observed in any other mutants characterized in this study, including the ones that did not induce any telomere length alteration (e.g. H119A, C152A and H156A). We examined a second mutant reported to induce telomere elongation (N95D). However, in our system, this mutant caused a very mild telomere lengthening phenotype, and we did not observe a statistically significant increase in RAP (Supplementary and data not shown). We conclude that the E76K mutation has a previously undetected effect on the intrinsic enzymatic properties of yeast telomerase, in a manner that is consistent with its ability to induce telomere elongation.
The present analyses of the GQ domain of yeast TERT have revealed functionally important residues. Here, we divide the residues into three classes based on phenotypes of the mutants and discuss their mechanisms (see B for a visual summary of the results). The first class, including K111 and H119 and located in patch 1, appears to have a ‘stabilizing’ effect on the telomerase complex; substitution at these positions reduced the levels of RNP and telomerase activity. However, the enzymatic property of telomerase showed no detectable change. The second class, including R151 and N153 and located in patch 2, appears not to be required for telomerase stability, but rather the primer interaction function of GQ domain. Substitution at these positions resulted in loss of activity on primers that form short hybrids with telomerase RNA and loss of RAP. These defects are qualitatively similar to defects exhibited by deletion of the entire GQ domain (), suggesting that the key function of the GQ domain requires patch 2. While the level of telomerase RNP were reduced slightly in both the R151A and N153A mutant (A), it is unlikely that this reduction can account for the telomere shortening phenotype of the mutants; the C152A and H156A mutants suffered slight RNP reduction, yet exhibited no telomere phenotypes. The third class, comprising only of Q146, is apparently required for complex stability as well as primer interaction, because substitution at this position results in the combined phenotype of the first two groups of residues. Thus, despite its surface location, Q146 must mediate important structural as well as enzymatic functions, consistent with its absolute conservation through evolution. Like the R151A and N153A mutant, the senescent and telomere shortening phenotype of the Q146A mutant cannot be attributed to the reduction in RNP level alone because the K111A suffered a comparable degree of RNP loss, yet exhibited only mild telomere shortening. Because all of the functionally important residues with the exception of Q146 are not well conserved, our data suggest that telomerase has evolved species-specific residues to mediate important GQ domain functions. This notion is consistent with the high degree of telomere sequence divergence in different organisms. Interestingly, the mutation that confers increased RAP (E76K) does not map near patch 2. Instead, the residue packs against two alpha helices (α5 and α6 in the designation scheme by Jacob ) that are positioned away from the basic patch (B). Thus either other regions of the domain also regulate DNA binding, or the effect of this mutation may be indirect.
A particularly interesting question raised by previous and current analyses of telomerase in different organisms is whether the contribution of GQ domain to RAP is universally conserved. Such a contribution has been unequivocally demonstrated for the yeast and human domain (,,). However, several analyses of TERT mutants have failed to disclose such a contribution (,,). Indeed, the same mutation in yeast and can apparently have different effects on telomerase enzymology. For example, Ala substitution of the invariant Gln caused a significant reduction in RAP in yeast, but not in (). The reason for this disparity is not understood, but several potential explanations may be considered. One possibility is that the DNA-binding function of TERT has evolved to serve another aspect of telomerase enzymology, e.g. primer binding during initiation. Alternatively, the presence of other DNA-binding domains or proteins in telomerase may obscure the contribution of the GQ domain to RAP. Finally, it should be noted that all of the studies have been performed using reconstituted telomerase, which differs significantly with respect to processivity from the endogenous enzyme (). In this regard, we note that a protocol for reconstituting mutant telomerase was recently reported (). It may be interesting to re-examine the contribution of GQ domain to RAP in this presumably more physiologically relevant system.
In earlier studies, we demonstrated that yeast telomerase mutants with reduced and enhanced NAP supported the maintenance of shorter and longer than wild type telomeres, respectively (). We have also identified mutations that simultaneously impaired RAP and telomere maintenance (). In this report, we further show that a mutation in the GQ domain of yeast TERT that resulted in increased RAP also induced telomere elongation. Collectively, these data provide strong argument for the notion that the intrinsic processivity of telomerase is an important determinant of telomere length. Furthermore, they support a recently proposed model of telomere length homeostasis (,). This model has two key premises. First, there is a feedback loop that regulates the activity of telomerase as a function of telomere length, such that longer telomeres are more refractory to extension. Second, regulation of telomerase activity occurs at the step of initiation rather than processivity.
Note that as stated before, Pi is a function of telomere length () and is negatively correlated with . Thus for example, when processivity is enhanced by a telomerase mutation (e.g. the E76K mutation), equilibrium can only be achieved by reducing Pi. A reduction in Pi, in turn, can be achieved by increasing telomere length. Thus, the model predicts an increase in equilibrium telomere length in the presence of a processivity-enhancing mutation, precisely as observed.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
RNA structures play important roles in the HIV-1 life cycle and several structured motifs are clustered in the untranslated regions (UTR) that are present at both ends of the viral RNA. The 5′UTR contains the repeat region (R), U5 region (unique at the 5′ terminus) and the leader sequence. The R region folds the TAR and the polyA hairpins (), and TAR binds the viral Tat protein to activate transcription (). Among many other RNA signals, the dimer initiation site (DIS) is located in the HIV-1 leader RNA and forms a stem-loop structure with a palindromic sequence. Direct base pairing between the loop-exposed palindromes of two DIS hairpins results in RNA dimerization. The 3′UTR is subdivided into the U3 (unique at the 3′ terminus) and the R region. The 3′R also folds the TAR hairpin, but the polyA hairpin is shortened due to polyadenylation ().
HIV-1 full-length RNA serves both as mRNA for the Gag-Pol proteins and as genomic RNA that is packaged in virion particles. It therefore contains characteristics for both functions. HIV-1 genomic RNA is 5′-capped and 3′-polyadenylated like cellular mRNAs, but it performs some unique virus-specific functions, such as dimerization, packaging in virion particles and reverse transcription. Reverse transcription is initiated within the leader RNA and the initial copy DNA of the 5′R is transferred to the 3′ to allow copying of the full-length genome (,). It has been suggested that the 5′/3′ ends should be close together in virions to facilitate this obligatory strand transfer step ().
Cellular mRNAs and the RNA genomes of several viruses were demonstrated to have a 5′/3′ interaction (). Specifically, genome circularization plays a role in RNA viruses that lack the cap structure (the Picornaviridae family), the polyA tail (the Reoviridae family) or both elements (the Flaviridae family). In some cases, a protein bridge is involved in RNA circularization, e.g. for members of the Picornaviridae and the Reoviridae family (,,,). Sequences in the 5′ and 3′ end of the RNA genome base pair in members of the Flaviridae family and the LTR-retrotransposon Ty1 (,,,,). The retrotransposon Ty1 shares several features with HIV-1, including its genome organization and the process of reverse transcription. Ty1 RNA circularization was shown to be involved in reverse transcription initiation (). It has been suggested that RNA circularization is a common feature of all RNA viruses (). We therefore set out to test this for HIV-1 RNA using assays.
The full-length HIV-1 molecular clone pLAI was used as a template for PCR amplification of certain genome regions (). Mutant GC1 contains a GCGC deletion in the palindromic sequence of the DIS hairpin (). The DNA templates for transcription were obtained by PCR amplification with sense primers with 5′-flanking T7 RNA polymerase promoter sequence and antisense primers (). The PCR fragments were ethanol-precipitated and dissolved in water.
transcription was performed with the Ambion MegaShortscript T7 transcription kit according to the manufacturer's instructions. The transcripts were separated on a 6% denaturing polyacrylamide gel containing urea, subsequently excised after visualization by UV shadowing and eluted from the gel fragments by overnight incubation at 30°C in 1× TBE. The RNA was subsequently ethanol-precipitated and dissolved in water. The transcripts were purified by Nucaway columns (Ambion). The quality of the unlabeled RNA was analyzed on a 2% agarose gel containing ethidium bromide, and the concentration of the RNA was determined by UV spectrometry (Smartspec™ 3000, Bio-Rad).
For 5′ end-labeling, the transcripts were dephosphorylated and end-labeled with [γ-P]-ATP (Amersham) according to the KinaseMax (Ambion) protocol. Transcript 3′ end-labeling was performed with T4 RNA ligase (Ambion) and [P]pCp (Amersham) according to the 3′ end-labeling protocol available at the Ambion website. End-labeled transcripts were purified on NucAway Spin Columns (Ambion).
Electrophoretic mobility shift assays were performed in 20 μl reactions, containing a physiological buffer (final concentration: 125 mM K acetate, 2.5 mM Mg acetate, 25 mM HEPES, pH 7.0), ∼200 counts/s of P-labeled transcript and different concentrations of unlabeled transcripts (end concentrations of 0, 5, 20, 80 and 320 nM). This mixture was heated at 85°C for 2 min to denature the transcripts and slowly cooled to room temperature for refolding. After addition of 4 μl non-denaturing loading buffer (25% glycerol with bromophenol blue dye), 10 μl was loaded on a 4% 0.25× TBM non-denaturing polyacrylamide gel (Boehringer Mannheim). Electrophoresis was performed at 150 V for at least 3 h for short transcripts or 100 V overnight for large transcripts. Subsequently, gels were dried and exposed overnight to a PhosphorImager screen. Quantification was performed with the ImageQuant software.
RNA complexes were formed in 200 μl physiological buffer (final concentration: 125 mM K acetate, 2.5 mM Mg acetate, 25 mM HEPES, pH 7.0), ∼2000 counts/s of P-labeled 3′ transcript with unlabeled transcript (final concentration 320 nM). This mixture was heated at 85°C for 2 min to denature the transcripts and slowly cooled to room temperature for refolding. The mixture was split in 20 μl samples and kept on ice. Non-radiolabeled 3′ transcript was added (320 nM) to prevent regeneration of the labeled gag–U3R interaction upon melting. To determine the melting temperature, the samples were incubated at different temperatures for 5 min and subsequently frozen in dry ice. After addition of 4 μl loading buffer (25% glycerol with bromophenol blue dye), 10 μl was loaded on a 4% 0.25× TBM non-denaturing polyacrylamide gel (Boehringer Mannheim). Electrophoresis was performed at 150 V for 3 h. Subsequently, gels were dried and exposed overnight to a PhosphorImager screen. Quantification was performed with the ImageQuant software.
Radiolabeled 5′ or 3′ transcripts (1000 count/s) were either incubated alone or mixed with non-radiolabeled 5′ or 3′ transcript (320 nM) in 90 μl physiological salt buffer (final concentration: 125 mM K acetate, 2.5 mM Mg acetate, 25 mM HEPES, pH 7.0) at 85°C for 2 min and slowly cooled to room temperature. The sample was split into 4 × 20 μl and kept at room temperature. Lead acetate (5 mM final concentration) was added for 2.5, 5 and 10 min. Cleavage by lead ions was stopped by addition of EDTA (100 mM final concentration) and one volume of formamide containing loading buffer. A hydrolysis ladder was made by incubating the radiolabeled transcripts in hydrolysis buffer (50 mM NaCO/NaHCO, pH 9.2) at 95°C for 5, 10 and 15 min, followed by addition of loading buffer. All samples were heated at 95°C for 2 min and separated on a 6% denaturing acrylamide gel at 50 W for 3 h. The gel was dried and exposed overnight to a PhosphorImager screens. Gels were analyzed by ImageQuant.
To determine if RNA circularization occurs in the HIV-1 genome, we screened for an interaction between two transcripts encompassing the 5′- and the 3′-terminal 1 kb of the HIV-1 genome, fragments 1–1036 and 8206–9229 (). To prevent homodimerization of the 5′ transcript, we used the GC1 mutant with a deletion in the loop of the DIS hairpin. This mutation has previously been demonstrated to effectively block HIV-1 RNA dimerization (). We first checked for the lack of homodimerization. Incubation of the radiolabeled 5′ transcript with increasing amounts of non-radiolabeled 5′ transcript in physiological salt buffer and analysis on a non-denaturing gel did not show homodimers (, first 5 lanes), demonstrating the effectiveness of the DIS mutation. The 3′ transcript is not expected to form homodimers, although a small amount of slow-migrating complexes are observed (20%) at the highest RNA concentration (290 nM, lane 10). We next mixed the 5′ and 3′ transcripts. The mixture was heated in physiological salt buffer to 75°C and slowly cooled to room temperature and analyzed on gel. The radiolabeled 5′ transcript was incubated with increasing amounts of non-labeled 3′ transcript, and the radiolabeled 3′ transcript was incubated with an increasing amount of non-labeled 5′ transcript. Both experiments show an apparent band shift signal (, marked by arrow). The radiolabeled RNA was efficiently shifted in this complex at relatively low concentration of the non-labeled RNA. The formation of homodimers and heterodimers was quantified as shown in B. These results indicate that the 5′ and 3′ ends of HIV-1 genomic RNA can interact.
To map the sequences involved in the 5′/3′ interaction, we made a 3′-truncated set of both transcripts with steps of 100 nt (). The radiolabeled 5′ set is shown upon gel analysis in the left panel of A. These transcripts were subsequently incubated with non-labeled 3′ 1-kb transcript (320 nM) (A, right panel). Only the four largest 5′ transcripts are able to interact with the 3′ transcript, yielding a prominent band shift (marked with arrow). Transcripts of 600 nt or shorter do not interact, indicating that the interaction domain requires sequences between position 600 and 700 of the gag coding sequence. The 3′-truncated set of 3′ transcripts is shown in the left panel of B. Upon incubation with non-labeled 5′ RNA (320 nM), only the largest transcript that includes the 3′ terminal 123 nt of HIV-1 genomic RNA is able to shift (B, right panel). These interaction results are summarized in with +/− signs. The analysis implicates gag sequences (position 600–700) and 3′UTR sequences (position 9127–9229) in the 5′/3′ interaction. This 3′ domain encodes the U3 and R sequences, the latter includes the 3′ TAR and polyA hairpins ().
To confirm the involvement of nt 600–700 and 9127–9229 in heterodimer formation, we performed band shift assays with short transcripts encompassing these gag and U3R sequences. These radiolabeled transcripts do not form significant homodimer levels (A), although the U3R transcript forms ∼10% slowly migrating complexes at the highest RNA concentration. This result is consistent with that of the band shift assay performed with the 1-kb transcripts. Most importantly, a prominent heterodimer complex is formed upon mixing of the transcripts (A, marked with arrow). The signals were quantified and the percentage bandshift was calculated (B). When the radiolabeled gag RNA is incubated with increasing amounts of non-radiolabeled 3′U3R RNA or , the transcripts shift efficiently (80 and 100%, respectively). These results demonstrate the minimal sequence requirements for the 5′/3′ interaction in gag (600–700) and U3R (9127–9229).
To further analyze whether the correct 5′/3′ interaction partners were mapped in the truncation series, we compared the melting temperature of the complexes formed with the 1-kb 5′/3′ transcripts versus the minimal gag/U3R domains. A large batch of the respective complexes was made in physiological salt buffer by heating to 75°C and slowly cooling to room temperature and samples were aliquoted. The samples were incubated at different temperatures for 5 min, snap-cooled and analyzed on a native acrylamide gel, of which the appropriate sections are shown (). To prevent renewed complex formation after heat-induced dissociation, a large amount of non-radiolabeled RNA corresponding to the radiolabeled transcript was present. The is ∼57°C for the complex with 1-kb transcripts and 55°C for the gag–U3R transcripts, indicating that essentially the same base-pairing interaction is formed. This result confirms that the minimal gag and U3R domains are sufficient for the 5′/3′ interaction.
We next performed RNA structure probing of the minimal gag and U3R transcripts before and after complex formation. We used lead (Pb) acetate to specifically probe the single-stranded, unpaired nucleotides. The samples were subsequently denatured and separated on a sequencing gel. The Pb-induced cleavage pattern of the individual gag transcript shows reactivity along the entire transcript, indicating the absence of prominent secondary structure (A). In the presence of unlabeled U3R transcript, a profound footprint is induced in the gag RNA. This footprint domain shows reduced Pb reactivity in two regions (marked 1 and 3 in A), with a hyperreactive region in between (marked 2).
The Pb-induced cleavage pattern of the individual U3R transcript shows an extended domain that lacks sensitivity with a hyperreactive segment in the middle. This pattern correlates with the stable secondary structure of the 3′ TAR hairpin (). The hyperreactivity of the TAR UUU-bulge is caused by the magnesium-binding pocket that strongly induces cleavage in the presence of Pb ions (). Upon mixing with the gag transcript, the TAR footprint remains unaltered, but several flanking sequences become protected (marked 4 and 6).
To analyze the Pb-induced cleavage pattern within the downstream half of the transcripts, we repeated the experiment with 3′ labeled transcripts (A, right panel). This experiment revealed clear signals in the gag transcript that become protected upon complex formation (marked 1, 2, 7 and 9). Sequences that become protected were also observed in the U3R transcript (marked 4, 6 and 10).
Analysis of the cleavage patterns, combined with Mfold analysis (), resulted in a prediction of the gag–U3R interaction (B). The base-pairing model shows that the two segments flanking the 3′ TAR hairpin in U3R interact with partially complementary sequences in gag. The TAR hairpin itself remains unaffected by formation of the 5′/3′ interaction. The gag–U3R interaction protects certain nucleotides against Pb-induced cleavage (indicated by red circles). The nucleotides of the gag–U3R heterodimer that are cleaved by Pb ions are mainly located in internal loops (indicated by blue arrows).
To provide phylogenetic support for the proposed 5′/3′ interaction, we analyzed the RNA genome of different HIV-1 subtypes. A representative sequence was selected for subtype A1, A2, B, C, D, AE, F, G, H and J. We fused fragments 600–700 nt and 9079–9229 nt (subtype B LAI coordinates) and analyzed the RNA secondary structure with Mfold (). The proposed 5′/3′ interaction can be formed by all HIV-1 subtypes, despite significant sequence variation (). All subtypes form a similar long distance base-pairing interaction between gag and U3. Within the gag–U3 interaction, a typical 4-nucleotide pyrimidine bulge that is flanked by two stretches of interacting nucleotides is apparent in several subtypes. One of these stretches, which is boxed in subtype A1, is analyzed in more detail in the insert of . This stretch can be formed by all subtypes and is supported by some interesting base-pairing co-variations. In particular, the marked C-G base pair of subtype A1, B, D and F (insert of ) is changed to U-G in subtype C, E and J, and true co-variations are also seen: A-U for subtype A2 and U-A for subtype H. The observation that the 5′/3′ interaction is conserved among the HIV-1 subtypes strongly suggests a role for this interaction in HIV-1 replication.
We identified an RNA–RNA interaction between the 5′ and the 3′ ends of the HIV-1 RNA genome. Mapping of the minimal domains indicated that the interaction is formed by a 5′ sequence in the gag open reading frame and the 3′-terminal U3R region. Transcripts comprising these short sequences form a complex with a similar as the complex formed by the 1-kb transcripts, confirming that we identified the minimal sequences that constitute the 5′/3′ interaction. Detailed RNA structure probing indicates that sequences flanking the 3′ TAR hairpin base pair with gag sequences. This gag–U3R interaction can be formed by all HIV-1 subtypes despite considerable sequence variation, thus providing phylogenetic support for the proposed interaction. Further experimentation is needed to elucidate the exact role of RNA circularization in the HIV-1 life cycle.
The proposed gag–U3R interaction is supported by the result of two previous studies. Kanevsky . () analyzed the role of secondary structure in the 5′ and 3′R regions on the strand transfer process of reverse transcription. This study used RNA structure probing to show that the TAR hairpin is present at both ends, but the polyA hairpin is presented differently, either in extended form (5′) or shortened form (3′) due to 3′-polyadenylation (). Interestingly, this study also placed the R region within the gag coding sequence. In this context, the shorter 3′ polyA hairpin is no longer observed, possibly due to pairing with gag sequences. Because the RNA structure probing data reveal a surprisingly strong correlation with our probing data, the same interaction between gag and polyA sequences may be formed. In another study, Gee . () used computational folding predictions of the 5′ and the 3′ polyA hairpin on transcripts encompassing the 5′-terminal 662 nt fused to the 3′ region of subtypes AE, AG, B and D. The 5′ polyA sequence predominantly folds the polyA hairpin, but the 3′ polyA sequence preferentially interacts with a sequence in gag. This structure, which was predicted for all four subtypes, is the same as the gag–U3R interaction described in this study. Since only the 5′ and 3′ 1000 nt were examined in this study, we cannot exclude the possibility that other sequences might influence the gag–U3R interaction. Computational RNA structure predictions of complete genomic sequences revealed that such interactions are not expected ().
Circularization of HIV-1 RNA seems exclusive for unspliced genomic RNA because all spliced subgenomic mRNAs lack the gag interaction domain. The gag–U3R interaction may distinguish genomic RNA from spliced transcripts and could thus function in genomic RNA-specific processes, e.g. packaging and reverse transcription. Especially for the latter process, the gag–U3R distance interaction may bring both ends of the genomic RNA close together. During reverse transcription, the (–) strong-stop DNA is transferred to the 3′R sequence to allow reverse transcription to proceed. Bringing the 5′ and the 3′ ends of the genomic RNA together could facilitate this strand transfer step (,). Furthermore, a long distance RNA interaction between the 5′ polyA region and a sequence in gag was described recently (). Surprisingly, this gag sequence is located only 170 nt upstream of the gag domain that interacts with U3R. Combining the two long-distance RNA interactions would bring the 5′R and 3′R sequences in very close proximity, which may facilitate strand transfer. The gag–U3R interaction may also positively influence polyadenylation at the 3′ polyA site, which is inhibited by the polyA hairpin as present in 5′R (,).
RNA circularization has been described for several positive-stranded RNA viruses (). Circularization plays an important role in essential steps of the viral replication cycle such as translation, transcription and reverse transcription (). The 5′/3′ interaction in HIV-1 is similar to the long-distance interaction described for the retrotransposon Ty1. For Ty1, gag sequences base pair with the U3 domain, and circularization of the retrotransposon RNA genome is required for efficient reverse transcription initiation (). For HIV-1 RNA, there are multiple sequence motifs near the 5′ end that regulate initiation of reverse transcription (), but there is no evidence that the 3′ end is involved. Further research will be needed to address this issue.
The HIV-1 genome contains many inhibitory sequences (INS) that block RNA export from the nucleus to the cytoplasm (,). One INS element is located within the gag sequence that is involved in the gag–U3R interaction (). The gag sequence has been reported to bind polyA-binding protein (PABP1), which usually interacts with the 3′-polyA tail of mRNAs (). Gag mutations that prevent binding of PABP1 increase gag translation significantly (). It has been speculated that PABP1 binds to the 3′ polyA tail at low concentrations, but also to the INS in gag at higher concentrations. Since PABP1 can multimerize, the proteins bound to gag and the 3′ polyA tail could also interact, thus forming a 5′/3′-protein bridge. The possibility that PABP1 modulates the gag–U3R interaction is an interesting option. |
Selenium (Se) is an essential micronutrient due to its requirement for biosynthesis and function of the 21st amino acid, selenocysteine (Sec). This amino acid is typically found in the active sites of a small number of selenoproteins in all three domains of life: archaea, bacteria and eukaryotes (). Biosynthesis of Sec and its cotranslational insertion into polypeptides require a complex molecular machinery that recodes in-frame UGA codons, which normally function as stop signals, to serve as Sec codons (). Although the occurrence of selenoprotein genes is limited, the Sec UGA codon has become the first addition to the universal genetic code since the code was deciphered 40 years ago ().
The mechanism of Sec insertion differs in the three domains of life. In bacteria, this process has been most thoroughly elucidated in (,,). Translation of bacterial selenoprotein mRNA requires both a selenocysteine insertion sequence (SECIS) element, which is a stem-loop structure immediately downstream of Sec-encoding UGA codon (,,), and -acting factors dedicated to Sec incorporation (). In archaea and eukaryotes, SECIS elements are located in 3′-UTRs and some factors involved in Sec biosynthesis and insertion are different. Recent identification of Sec synthase, SecS, in eukaryotes, which is different from the bacterial Sec synthase, SelA, provided important insights into Sec biosynthesis in these organisms ().
Recently, an additional rare amino acid pyrrolysine (Pyl), was identified, which expanded the canonical genetic code to 22 amino acids (,). Pyl is inserted in response to UAG codon in several methanogenic archaea (). Although the mechanism of Pyl biosynthesis and incorporation into protein is not fully understood, the presence of a tRNA gene () with the CUA anticodon and of class II aminoacyl-tRNA synthetase gene () argued for cotranslational incorporation of Pyl (). In , a single bacterium, in which a Pyl-containing protein was found, PylS consists of two proteins: PylSn and PylSc ().
In recent years, large-scale genome sequencing projects, including both organism-specific and environmental metagenomic projects, provided a large volume of gene and protein sequence information. However, selenoprotein genes are almost universally misannotated in these datasets because UGA has the dual function of encoding Sec and terminating translation, and only the latter function is recognized by current annotation programs. Several bioinformatics tools have been developed to address this problem and can be used to identify selenoprotein genes (). These programs have successfully identified many new selenoproteins in both prokaryotic and eukaryotic genomes, as well as in the Sargasso Sea environmental samples ().
Complex symbiotic relationships between bacteria and multicellular eukaryotes have evolved in several environments, but science has traditionally focused on interactions that are pathogenic (). Recently, there has been increased recognition of symbiotic interactions that benefit both the microorganism and the host (). A recent metagenomic analysis of the symbiotic microbial consortium of the marine oligochaete , a worm lacking a mouth, gut and nephridia, revealed four major co-occurring symbionts, which belong to (δ1 and δ4) and (γ1 and γ3), as well as one minor species. Since some are selenoprotein-rich organisms (), we analyzed the selenoproteomes of these symbionts to examine a possible relationship between selenium and symbiosis.
To characterize selenoproteome in these symbionts, we adopted a Sec/cysteine(Cys) homology-based search approach, which has been successfully used to characterize the selenoproteomes of both prokaryotes () and one of the largest prokaryotic sequencing projects, the Sargasso Sea microbial sequencing project (). We detected known selenoproteins present in this metagenomic dataset and identified several novel selenoproteins. Interestingly, one deltaproteobacterium, δ1 symbiont, contains at least 57 selenoproteins, which is the largest number of selenoproteins reported to date in any organism. In addition, several Pyl-containing proteins were identified and most were also found in the same δ1 symbiont. Our results provide new insights into understanding evolution and function of these rare amino acids.
Assembled sequences of the symbionts’ metagenome were obtained from NCBI with the project accession number AASZ00000000 (). The database contained 5597 genomic sequences, which corresponded to a total of 23.7 million nucleotides. Non-redundant (NR) protein database was downloaded from NCBI ftp server. This dataset contained a total of 4 644 764 protein sequences (1 603 127 260 amino acids). BLAST () was also obtained from NCBI.
Each Cys-containing protein sequence in the NR database was initially searched against the symbionts’ metagenomic database for possible TGA/TAG/TAA-containing homologs using TBLASTN with default parameters. Only local alignments, in which Cys in the query protein was aligned with TGA codon in the nucleotide sequence from the symbionts’ metagenomic database, were selected for further analysis. For each TGA-containing nucleotide sequence identified in the metagenomic database, regions upstream and downstream of the putative in-frame TGA codon were analyzed to identify a minimal ORF. If a stop codon was found between the in-frame TGA codon and an initiation codon (ATG or GTG), such a TGA-containing sequence was discarded.
We used BL2SEQ to cluster remaining protein sequences into different groups. If a local alignment of two proteins had an E-value below 10 and was at least 20 amino acid long, as well as the predicted Sec residues were located at the same position or very close (no more than three residues apart) in the alignment, the two proteins were assigned to the same cluster.
The remaining clusters were analyzed for occurrence of bacterial SECIS elements, located immediately downstream of the in-frame TGA codons, using bSECISearch program (). The final clusters were manually analyzed and divided into three groups: known selenoproteins, new selenoproteins (clusters containing at least two different sequences with conserved in-frame TGA codons) and selenoprotein candidates (clusters containing only one sequence). It should be noted that sequencing errors that generate in-frame UGA codons could not be excluded for selenoprotein candidates.
PylT and PylS sequences from (accession number AY064401) were used to search for possible homologs in the metagenomic dataset. Candidate tRNA was further analyzed to identify structural features associated with known tRNA, including a six base-pair acceptor stem and a base between the D and acceptor stems (). Other genes in the Pyl operon (, , ) were also analyzed by comparative sequence analyses.
The TBLASTN program with default parameters was used to search for known Pyl-containing methylamine methyltransferases. Open reading frames (ORFs) and conservation of UAG-flanking regions were examined manually. Multiple alignments were generated with ClustalW ().
xref
fig
#text
Whole-genome shotgun and metagenomic sequencing projects have provided a new and powerful tool in the study of community organization and metabolism in natural microbial communities (). Recently, such methods have been extended to analyze symbiotic relationships. One project involved an analysis of microbes from a marine oligochaete , which lacks a mouth, gut, anus or nephridial excretory system, and contains several bacterial endosymbionts that are located just below the worm cuticle (). These endosymbionts include two sulfur-oxidizing gammaproteobacteria (γ1 and γ3) and two sulfate-reducing deltaproteobacteria (δ1 and δ4). Identification of selenoprotein genes in such an unusual symbiotic system may help understand the role of selenium and other micronutrients in the intricate interactions that form such a complex, adaptive consortium.
In the present study, we employed a procedure that analyzes Sec/Cys pairs in homologous sequences to characterize the selenoproteomes of symbiotic microorganisms in the gutless worm. A total of 82 genes that belonged to 24 previously described prokaryotic selenoprotein families and 17 sequences that belonged to six new selenoprotein families were identified. Most selenoproteins were found to occur in δ1 symbiont, which contained 44 known selenoproteins (21 families) and 13 new selenoproteins (6 families). Although the genome size of δ1 symbiont is ∼13.5 Mb, which is larger than most other deltaproteobacteria, its reconstruction revealed a single species (). If this is the case, then our study identified an organism, which has the largest selenoproteome reported to date (57 selenoproteins) of any organism, including eukaryotes and archaea.
Most detected selenoproteins were homologs of thiol-based redox enzymes and contained conserved redox motifs. In contrast, such known redox motifs were largely absent in new selenoproteins identified in the metagenomic dataset. In addition, analysis of secondary structures revealed that these new selenoproteins did not contain thioredoxin-like fold, which is a dominant fold in selenoproteins identified in several marine environmental sequencing projects (,). Perhaps, additional redox reactions that are carried out by new selenoproteins occur in these symbionts.
Besides the unusually high number of selenoproteins, 10 Pyl-containing proteins were identified in the metagenomic dataset. δ1 contained eight of these sequences that belonged to MtbB and MttB families. Thus, the δ1 symbiont is also the organism, which has the largest number of Pyl-containing proteins in bacteria. Previously, only one bacterial protein, from , was known to possess Pyl. Therefore, identifying so many pyrroproteins in the same bacterium is truly remarkable.
We previously proposed that UAG may be an ambiguous codon in some archaea, wherein it could serve as either Pyl codon or a stop signal. However, in , UAG is frequently used as a stop signal, suggesting an unknown mechanism that allows ribosomes to recognize function of specific UAG codons. By analogy to Sec, which is inserted with the help of SECIS elements, PYLIS elements may be present in bacterial pyrroprotein genes. However, our analysis of genes coding for Pyl-containing proteins revealed no common RNA structures. Additional RNA structure searches should be carried out in the future. The current set of Pyl-containing proteins provides an excellent dataset for further interrogation.
Given that most symbiotic and host-associated bacteria have lost the ability to utilize Sec or only possess a limited number of selenoproteins, the dramatic abundance of selenoproteins in the two endosymbiotic deltaproteobacteria, especially δ1 that also contains many Pyl-containing proteins, is remarkable, raising a series of questions regarding evolution and function of these proteins, as well as their roles in symbiosis. It has been suggested that most selenoproteins evolved from their Cys-containing homologs and anaerobic environments could support the use of Sec (). Compared to most other symbionts and host-associated organisms, which seem to live under aerobic or microaerobic conditions, the obligate anaerobic environment of the two symbionts may be one reason for evolution of new selenoproteins. In addition, compared to the environments where other hosts live, seawater could provide a constant supply of selenium for Sec biosynthesis in these symbionts. An alternative hypothesis is that the host worm needs more efficient metabolism and waste management, which are provided by its symbionts because of the lack of digestive and excretory systems. These special needs might have led to selective advantage of harboring multiple symbionts that utilize amino acids that provide catalytic advantages to various metabolic systems, such as Sec in many redox proteins and Pyl in methylamine methyltransferases.
Symbiotic deltaproteobacteria in the gutless worm evolved as organisms that support the broadest use of the genetic code, utilizing 63 of 64 codons to code for 22 amino acids. It would be interesting to examine if this and other symbiotic systems provide selective advantage to further expand the genetic code, either utilizing a third stop signal, UAA, or using some codons to insert multiple non-canonical or common amino acids.
#text
S
u
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
o
n
l
i
n
e
. |
Double-strand breaks (DSBs) can be produced exogenously by ionizing radiation and chemical DNA damaging agents or by endogenous free radicals generated during cellular metabolism. Eukaryotic cells have evolved two main pathways to repair DSBs: homologous recombination relies on extensive sequence homology between the damaged DNA and donor template, whereas illegitimate recombination, or nonhomologous end joining (NHEJ), involves end joining in the absence of DNA sequence homology (). Some illegitimate recombination events are characterized by a few basepairs (bp) of homology shared at the ends of the two recombination junctions, so called microhomology-mediated recombination (MHMR) (). This class of events is independent of the end-binding Yku70/Yku80 heterodimer, and thus is distinct from the classical NHEJ pathway (). As well, MHMR is a mutagenic pathway that is often associated with deletion of the sequences spanning one of the microhomology regions and the intervening sequence.
MHMR has been implicated in generation of large deletions and other genomic rearrangements in mammalian cells. Microhomology is often found at the recombination junctions of radiation-induced genomic rearrangements (,), implying that radiation-induced DSBs are repaired preferentially by MHMR. Furthermore, 50% of large deletions mediated by MHMR are associated with human disorders ().
In yeast, DSBs are mainly repaired by homologous recombination, but different mechanisms of illegitimate recombination have been evident in the absence of a homologous donor or an essential gene in the homologous recombination pathway (,,). By transformation of a -containing fragment into haploid yeast cells lacking the gene and subsequent selection for Ura transformants (,), three classes of spontaneous illegitimate recombination events were identified. The first class is due to MHMR, characterized by 2–4 bp of microhomologies at the junctions (). The second class is mediated by topoisomerase I where the transforming DNA integrates next to the preferred cleavage sites of topoisomerase I, CTT or GTT (). The final class involves ligation of transforming DNA to mitochondrial DNA fragments that seem to be capable of autonomous replication ().
Co-transformation of a recombination substrate with restriction enzymes HI, II and I increases the efficiency of DNA integration by several fold and the majority of events integrated into the restriction sites introduced by the co-transformed restriction enzyme (,), this process being termed restriction enzyme-mediated integration (REMI). The simple interpretation is that the co-transformed restriction enzymes create DSBs at the genomic restriction sites that attract the transforming DNA ends. Restriction enzymes were subsequently used to facilitate integration in for random mutagenesis and simultaneous tagging of the mutant gene ().
Ionizing radiation enhances the frequency of nonhomologous integration (NHI) in yeast (). In addition, other DNA damaging agents including UV, bleomycin, camptothecin, VP-16 and hydrogen peroxide also enhance gene integration in mammalian cells (). These DNA damaging agents may eventually lead to the formation of DSBs and suggest that DSBs could be the inducing agent for the enhancement of integration.
The DNA sequences of recombining junctions have been determined for very few DNA damage-enhanced illegitimate recombination events and the possibility that DNA damage induces illegitimate recombination events at undamaged sites has not previously been investigated. In this manuscript, we investigate the effects of ionizing radiation and restriction enzymes on NHI by sequence analysis of the integration target sites. We report here that both ionizing radiation and restriction enzymes increase the frequency of microhomology-mediated integration. In addition, a site-specific I-I-mediated DSB induces microhomology-mediated integration randomly throughout the genome, suggesting that DSBs induce MHMR probably at non-targeted sites.
Experiments were performed in the haploid strain RSY12 (
), in which the entire open reading frame and promoter sequence was replaced by the gene (). Strain YWY200 was constructed in the RSY12 background by integration of plasmid pWY203 containing I-I gene under the promoter into locus (Yap,W. and Schiestl, R. H., unpublished results). The construct containing the 24-bp I-I recognition site and a marker flanked by 50 bp of upstream and downstream sequence of was generated by PCR and transformed into YWY200 to generate strain YWY200-I-SceI. strain DH5α was used for the maintenance and amplification of plasmid DNA.
YEp195 contains the marker for selection and the two-micron origin of replication and was used to control for transformation efficiency. Plasmids pM150 and pM151 were described previously () and contain a 1.1 kb III fragment.
For yeast transformation the ‘Lithium Acetate/Single-Stranded DNA/PEG’ transformation method was used (). In previous studies, ∼10% of spontaneous NHI events were ligated to mitochondrial DNA fragments (). The occurrence of these events is probably related to the fact that some mitochondrial DNA fragments can function as an origin of replication in the nucleus (). We excluded these mitochondrial events in the current study by checking the frequency of loss of the Ura3 phenotype on medium containing 5-fluoro-orotic acid (FOA) because these events are not due to integrations into genomic DNA.
Standard methods were followed except as noted below. pM151 was digested with II and pM150 with I and prepared as described (). The spontaneous and radiation-induced integration target sites were determined by direct genomic sequencing. Yeast genomic DNA was purified using the genomic purification tips (100/T) (Qiagen, Valencia, CA, USA). Fifteen microgram of genomic DNA was used for a double big dye sequencing reaction () with the primer pM151-3559 5′-GTGCACCATATG CGGTGTGAAATACC-3 to sequence the forward direction and the primer pM151-71 5′-AATCTAAGTCTGTGCTCCTTCCTTCG-3′ to sequence the reverse direction. Sequencing products were purified by Centri-Sep columns (Princeton Separations, Adelphia, NJ, USA) and loaded on an ABI 371 sequencer.
Plasmid DNA was digested with restriction enzyme, precipitated with ethanol and resuspended in 200 µl of 0.01 M Tris-HCl (pH 7.5), 0.05 M EDTA, 1% sodium dodecyl sulfate and 100 µg of proteinase K/ml. After 30 min incubation at 37°C, the sample was extracted with phenol–chloroform–isoamyl alcohol, precipitated with ethanol, washed with 70% ethanol and vacuum dried. The pellet was dissolved in sterile water, and the yeast cells were then transformed with this solution. Also, 200U of restriction enzyme and 1/10 volume of the restriction enzyme buffer were added to the transformation mixtures. The restriction enzyme-induced integration target sites were cloned by plasmid rescue as described previously (). 718 was purchased from Roche (Basel, Switzerland). All other enzymes were purchased from New England Biolabs (Ipswich, MA, USA). Thirty-eight junctions from 19 integration events were sequenced in which both junctions resulted from NHI ( and B). Also included in our analysis are three junctions from NHI events that were previously published [eventBH10 in A, events SB3 and SB13 in B in ref. ()].
After transformation of plasmid DNA, yeast cells were plated onto synthetic complete medium lacking uracil and exposed to 50 Gy of γ-ray using a Mark-I irradiator (Shepherd and Associates, San Fernando, CA, USA), with a Cs-137 source and a dose rate of 529 cGy/min.
The frequencies of microhomology-mediated recombination with and without exposure to γ-ray, restriction enzymes and I-I endonuclease were analyzed by the χ test.
To examine the effect of γ-irradiation on NHI, a II-linearized plasmid pM151 containing the gene was transformed into yeast strain RSY12, which lacks any homology to the gene. Cells were then exposed to 50 Gy of γ-rays which causes an average of two DSBs per cell (). Cells expressing the Ura phenotype are the results of the containing plasmid pM151 that has either integrated into the genome or has captured an origin of replication. In parallel, a self-replicating plasmid YEp195 that contains a two-micron origin and the gene was used to transform RSY12 to normalize changes in the transformation efficiency caused by radiation. The frequency of spontaneous NHI events was about 0.75 per μg DNA per 10 YEp195 transformants. Exposure of yeast cells to γ-irradiation in three different experiments caused an average of 4.7-fold increase in NHI frequency, which is similar to previous results (). In addition, γ-irradiation caused an average of 6.3-fold increase in NHI frequency using a I-linearized plasmid pM151 as integrating DNA, indicating that radiation-induced integration is not specific to the II-linearized substrate (data not shown).
To investigate the mechanism of radiation-enhanced integration, we isolated genomic DNA from yeast colonies that underwent NHI events and sequenced the regions spanning the junctions between genomic DNA and the integrated plasmid. Obtained sequences were aligned with yeast chromosomal DNA sequences in the Genome Database (SGD) to identify integration target sites. We randomly selected and analyzed 13 stable transformants of spontaneous integration events and 11 transformants derived after irradiation. Among the spontaneous events, 7 out of 26 junctions (27%) contained 2–3 bp of microhomology between the protruding single-stranded (PSS) ends of the integrating DNA and the genomic target sites (A). Remarkably, among integration events in irradiated cells, 17 out of 22 junctions (77%) displayed 2–7 bp of microhomology (B). This 2.9-fold increase in the fraction of MHMR events is significantly different ( < 0.001) in comparison to spontaneous events. Forty-five percent of events in irradiated yeast cells utilized 4 or more bp of microhomology (27% with 4 bp, 9% with 5 bp, 4.5% with both 6 and 7 bp) while no events with 4 or more bp were observed in unirradiated cells () indicating that microhomology length is increased in irradiated cells in comparison to spontaneous events ( < 0.0005).
As radiation-induced DSBs are modified and required to be processed prior to religation (), deletions are often found at the breakpoints (,). However, the majority of spontaneous and radiation-induced integration events had the genomic sequences flanking the integrating substrate maintained without any deletion, implying that spontaneous and radiation-induced integrations are likely to occur at sites other than the sites of DSBs. All spontaneous and radiation-induced illegitimate integration events were randomly distributed throughout the yeast genome without bias for any particular chromosome position, expressed gene or ORF.
Previous experiments have shown co-transformation with restriction enzymes increase the efficiency of DNA integration into genomic sites of the restriction enzymes (,). When 718-linearized plasmid was transformed into yeast cells in the presence of I, the integration efficiency increased 2- to 3-fold yet less than 1/3 integrated at the I sites (). Furthermore, when II was co-transformed with a I-linearized plasmid, no integrations were observed at the II sites (). These results suggest that co-transformation of linearized DNA substrate with restriction enzymes may lead to increased integration efficiency into genomic sites other than the ones recognized by the restriction enzymes. In this work, we continued our previous studies and analyzed the events obtained after co-transformation with restriction enzymes by sequence analysis. We examined the following enzyme combinations: HI-HI, I-II, RI-II, I-II, 718-I and 718 (filled PSS ends)-I, in which the former enzyme was used to linearize the integrating plasmid pM150 and the latter enzyme was added during the transformation process (). We classified all events after restriction enzyme addition into different integration classes (restriction enzyme-mediated integration (REMI) and NHI) and compared this distribution with the control experiments lacking co-transformation with restriction enzymes.
Among events from co-transformation with restriction enzymes, 12 out of 41 junctions (29%) displayed 4 or more bp of microhomologies between the PSS ends of the integrating DNA and the genomic sequences whereas only 1 out of 36 junctions (2.7%) was observed in the control ( < 0.005) (, and B). We compared 14 junctions from spontaneous integrations of a 3′ PSS substrate linearized by I (A) with 14 junctions obtained from co-transformation with II, producing a 5′PSS (B). The distances to the closest II sites for each of the 14 integration junctions were in agreement with a random distribution in relation to those II sites in the genome (data not shown). This suggests that a specific pathway utilizing microhomology was induced by restriction enzymes increasing integration efficiency .
Restriction enzymes sometimes show ‘star activity’ in the presence of detergents in the buffer or with inappropriate buffers () (). It is possible that the enzymes show such star activity under the conditions that occur in the nuclei of cells. The most common types of star activity are single base-pair substitutions or truncations of outer bases (). However, none of the integration sites could be accounted for by the restriction enzyme cutting at such common star activity sites.
To determine if integration occurs close to the site of breakage or at random locations throughout the genome, we overexpressed the I-I endonuclease to generate a single DSB at a defined site in the genome. I-I is a meganuclease that cleaves at a 24-bp recognition site not present in the yeast genome (). Cells overexpressing the endonuclease, and having a recombinant I-I site, undergo competition between cleavage and religation resulting in a constitutive DSB. Strains YWY200 and YWY200-I-I express the I-I enzyme under the promoter, inserted at the locus, while the YWY200-I-I strain in addition contains an I-I recognition site at the locus. Upon 6 h of galactose induction in the YWY200-I-I strain, cleavage of the recognition site in the genome by the overexpressed I-I was verified by PCR analysis (data not shown). Strains YWY200 and YWY200-I-I were grown in complete medium with 2% glucose or 2% galactose for 6 h and transformed with II-linearized pM151 to determine the NHI frequency. After induction on galactose-containing medium, the relative frequency of NHI (transformants per μg DNA per 10 YEp195 transformants) in the strain containing the I-I site increased by 4.8-fold, from 0.11 ± 0.05 to 0.50 ± 0.14 whereas in control cells on galactose, not containing the I-I site, there was no increase in the NHI frequency; 0.11 ± 0.01 and 0.13 ± 0.04, respectively.
After induction in galactose media, we sequenced 7 and 11 transformants derived from YWY200 and YWY200-I-I cells respectively (). In the YWY200 cells, 2 out of 14 (14%) junctions showed 2–3 bp of microhomologies whereas 15 out of 22 (68%) junctions from the YWY200-I-I cells displayed 2–6 bp of microhomologies, resulting in a significant difference in the occurrence of microhomology usage ( < 0.005) (). In seven integration events from YWY200 cells combined with 13 spontaneous events from the isogenic wild-type RSY12 strain (A), 9 out of 40 (23%) junctions showed 2–3 bp of microhomologies versus 16 out of 22 (73%) junctions from the YWY200-I-I cells utilized microhomologies from 2 to 6 bp, which is also significantly different ( < 0.0005). Strikingly, the integrations from the YWY200-I-I cells were distributed randomly over the genome, as opposed to being targeted in proximity to the I-I-mediated DSB (), indicating that DSBs induce MHMR that is probably at random, non-targeted sites.
Our results show that a MHMR pathway is induced after exposure to ionizing radiation, restriction enzymes and I-I endonuclease. Radiation caused a 3-fold increase in microhomology-mediated integration. Radiation-induced integration showed lack of deletions at genomic insertion sites, implying that such events are likely to occur at sites that are not damaged. Similarly, restriction enzymes and I-I endonuclease increase the frequency of microhomology-mediated integration. Restriction enzymes induce microhomology-mediated integration at non-restriction sites that are undamaged. Furthermore, the I-I induced microhomology-mediated integration events were distributed randomly throughout the genome, as opposed to being targeted in proximity to the I-I-mediated DSB, suggesting that radiation-induced MHMR events may also occur at undamaged sites. Taken together, these results demonstrate that DSBs induce MHMR that is probably at non-targeted sites.
Ionizing radiation and other DNA damaging agents have previously been shown to increase the frequency of NHI in mammalian cells (), and there is evidence for DNA damage or a DSB-induced pathway of MHMR (,). Moreover, the recombined products of MHMR show deletions from the site of breakage to the closest region of microhomology (). However, in these studies only the rejoining of the DSB at the target site was examined, and illegitimate recombination events at non-targeted sites were not detectable because of their design. Consequently, the author's interpretation in these studies was that during the repair of a DSB, degradation of DNA ends ensues until microhomology is encountered. Few studies have reported ionizing radiation-induced rearrangements not targeted at the DSB. These include radiation-enhanced formation of the transducing phage during prophage induction (,). Noteworthy, radiation-enhanced excision by recombination was not observed between random sequences but between hotspot sites of microhomology (). The author's interpretation was that it is unlikely that these events were DSB mediated and therefore another radiation-induced DNA damage at these sites of microhomology might have enhanced the efficiency (). This is an unlikely but theoretically possible explanation of our radiation-induced integration events into the genome. However, this possibility does not account for restriction enzyme and I-I-induced integration events that are most likely to the DNA damage.
We utilized a system that overexpresses I-I endonuclease to constitutively generate a DSB at a specific locus, which enables us to examine whether the enhanced integration events were targeted to the site of DSB or occurred at non-targeted sites throughout the genome. Similar to ionizing radiation and restriction enzymes, I-I endonuclease induces microhomology-mediated recombination. Strikingly, all of the I-I-mediated DSB-induced integration events were distributed randomly at non-targeted sites throughout the genome, as opposed to being targeted in proximity to the I-I-mediated DSB. These results demonstrate that DSBs induce a genome-wide MHMR pathway that facilitates integration probably at non-targeted sites.
Ionizing radiation generates single-stranded breaks, DSBs, base damage and DNA crosslinks. It is possible that radiation-induced integration events occur at these damaged sites. Based on the observations that deletions are frequently observed at the sites of DSBs in yeast (,,) and mammalian cells (,); if the integrating substrate was targeted to the damaged site, deletions would have been expected to occur at the site of damage until microhomology was encountered, similar to microhomology-driven deletions in plasmid end-joining assays (,). However, this possibility is excluded by our results, which show lack of deletions at the genomic insertion sites in 9 out of 11 independent radiation-induced MHMR events. Furthermore, the probability of radiation-induced damage coinciding with the sites of microhomologies is extremely low. It has been shown that radiation-induced damage occur rather randomly (). The expected random occurrence for 4, 5, 6 and 7 nt of microhomology is ∼1, 0.3, 0.09 and 0.03%, respectively (), whereas our frequencies of usage of these microhomologies after irradiation were 27, 9, 4.5 and 4.5%, respectively ( < 0.005). In addition, it is very unlikely that radiation-induced damage occurs at the specific sequences GATC or TCGA that is homologous to the ends of the integrating substrate linearized by II or I. Even though unlikely it is still possible that radiation-induced integration occurs at the subset of damaged sites that coincide with the microhomology for MHMR.
Among the radiation-induced integration events, the total loss of nucleotides at the PSS ends is 9 out of 88 compared to 23 out of 104 in spontaneous events ( < 0.05) indicating that radiation protected the ends of the integrating DNA substrate. Of the restriction enzyme induced NHI events, 14 out of 56 nt were lost from spontaneous integrations of a I-linearized substrate compared to only 3 out of 56 after co-transformation with II ( < 0.005). Thus co-transformation with restriction enzymes, like radiation, seems to protect DNA ends from degradation. This may reflect a faster or more efficient microhomology search after introduction of DSBs somewhere else in the genome. Alternatively, a hypothetical activity that preserves the PSS ends induced by DSBs might have a similar effect and increase the availability of plasmid PSS ends for microhomology search.
A time course study showed that exposure of yeast cells to radiation 6 h before transformation of the integrating substrate increases the frequency of NHI by 2-fold (data not shown), indicating that an increase in microhomology usage after irradiation is associated with an inducible process since the majority of radiation-induced damage is repaired within 2 h in yeast (). Based on our data, it is likely that a factor is induced that facilitates microhomology search and mediates MHMR.
MHMR has shown to be independent of that is distinct from the prototypical NHEJ pathway (). The possibility that induced MHMR could be mediated by a type of homologous recombination with low fidelity is much less likely since homologous integration requires at least 30 bp of homology on each end of the DNA fragment (,), as opposed to only a few base pairs of homology in the induced MHMR events. Alternatively, the induced MHMR may occur via a pathway similar to single-strand annealing that takes place between two direct repeats; however, this is also not likely since homologous sequences as small as 29 bp was observed in only 0.2% of the time in single-strand annealing ().
Under environmental stress, a process called adaptive or stationary-phase mutations arise, which are different from spontaneous growth-dependent mutations in (,). Harris . showed that the adaptive mutation is dependent on the RecA-RecBCD homologous recombination pathway (). The adaptive mutations occur at unselected genes throughout the bacterial chromosome (,). The transient global mutagenesis increases the rate of evolution enormously and produces a small population of cells with a hypermutable state that can adapt to a new environment, which is advantageous when genetic variation is a limiting factor for evolution (). Similarly, cancer cells assume a mutator phenotype to produce high levels of genomic instability during cancer development. Similar mechanisms may be involved in evolution and carcinogenesis. According to our data, it is plausible that a DSB-induced genome-wide MHMR pathway could lead to large-scale genomic rearrangements after a single DSB end invades another genomic location. Such a phenomenon may provide benefits to evolve genetic variants that have growth advantages under genotoxic stress. We propose that this inducible MHMR pathway could be a potential mechanism of adaptive evolution and carcinogenesis in eukaryotes. |
xref
italic
#text
xref
sub
fig
#text
Buck and Zechiedrich () proposed that if type II DNA topoisomerases were able to recognize specifically inter-hooked juxtapositions as sites of strand passages, then the steady-state level of knotting maintained by type II topoisomerases could be diminished as compared to completely random strand passages. Liu () defined the criteria of inter-hooking for portions of random polygons in the cubic lattice, where due to the lattice constraint each subchain involved in an inter-hooked juxtaposition needs to have at least four segments. For polygons in free space, inter-hooked juxtaposition can be already formed by two subchains composed of two segments each. illustrates what geometric characteristics of juxtapositions of two subchains composed of two segments each were analysed by us. It is somewhat arbitrary to decide what geometry of inter-segmental juxtaposition should be considered already as inter-hooked and where the limit is. To illustrate the possible effect of inter-hooking, the data in show a topological difference between passages at random juxtapositions and these that fulfiled the following three conditions: (i) each subchain intersects with a sector of the plane extending from angle defined by the other subchain; (ii) the facing angle > 105°; (iii) the mean opening angle ( + )/2 < 22.5° (). The mean opening angle of 22.5° (i.e. the deflection angle of 157.5°g) might seem to be very high, however recent experimental studies of DNA cyclization demonstrated that DNA fragments as short as 94 bp efficiently circularize and therefore show a cumulative deflection angle of 360° (). Those results have lead to a conclusion that sharp kinks form spontaneously in the DNA (). While the presence of spontaneous DNA kinks in DNA is presently a matter of debate (,) our modelling results can be equally well applied to strong bends resulting from localized kinks or strong bends resulting from the cumulative bending angle of . 150° that is redistributed over the length of 40 bp or so.
A shows the effect of the selection of inter-hooked juxtapositions (green bars), as compared to random juxtaposition (red bars) as sites of strand passage within random configurations of trefoil knots (the analysed configurations were limited to left-handed versions of trefoil knot and are therefore denoted as 3 knots). It is visible that a passage at inter-hooked juxtaposition has 3.6 times higher probability to result in unknot as compared to a passage at a random juxtaposition. B shows how the topological consequences of strand-passage reactions within unknots depend on whether the selected juxtaposition was inter-hooked or just random. The observed inhibition of the formation of trefoil knots for passages at inter-hooked juxtapositions as compared to random juxtapositions was of the order of 3.9. Combining the data from A and B, one can conclude that the selection of inter-hooked configurations as sites of inter-segmental passages is expected to decrease the steady-state ratio of trefoils to unknots by about 14 times (3.9 × 3.6) as compared to random passages, provided that the frequency ratios between random and inter-hooked configurations is the same in trefoil knots and unknots. However, our analysis revealed that the ratio of inter-hooked to random configurations is 1.65 times higher in trefoil knots than in unknots. The rate of the strand-passage reaction in unknotted and trefoil-forming molecules can be assumed to be proportional to the frequency of appearance of recognized juxtapositions within unknots or trefoils, respectively (). Therefore, the selection of inter-hooked juxtapositions (with the cutoff parameters described above) as sites of strand passage would be expected to diminish the level of knotting by more than 23 times (14 × 1.65) below the topological equilibrium in this system. This is 2-fold less than the effect observed in biochemical experiments performed by Rybenkov () who revealed a 50-fold reduction of the knotting level in DNA molecules with a size closely corresponding to 32 statistical segments. To reach this level of knotting reduction in simulation studies, one would need a more strict selection of inter-hooked juxtapositions, requiring a smaller mean opening angle (α + α)/2 or a bigger facing angle γ, for example. Alternatively, the knotting reduction level could be further increased by introducing a modification of a kinetic proofreading mechanism proposed by Yan (). This mechanism assumes that type II topoisomerases have a strong binding preference for inter-hooked juxtapositions, but after first binding to DNA, the energy of ATP is used to liberate the bound juxtaposition without performing the passage reaction but instead activating the enzyme for a passage during a short time period. If during this short activation time a juxtaposition with the preferred geometry is bound by the enzyme it is used for the inter-segmental passage. If no juxtaposition with preferred geometry is bound within the short time after the first binding, the activation state decays and the entire cycle has to be repeated. The mechanism of kinetic proofreading is based on the fact that inter-hooked juxtapositions are more frequent in knotted than in unknotted molecules and it essentially squares the probability that a passage happens in a knotted DNA molecule rather than in unknotted one (). After introducing a kinetic proofreading factor, we observed that the same selection criteria for inter-hooked juxtapositions (as described above) caused a 40-fold knotting reduction.
Type II DNA topoisomerases were shown not only to be able to maintain the knotting level below the topological equilibrium but also to choose a specific pathway of relaxation of more complex knots (). The chosen pathway maximized the efficiency of unknotting by selecting the passages that lead to direct unknotting instead of progressive relaxation through knots with intermediate level of knotting (). To investigate the effect of selection of inter-hooked juxtapositions on the topological outcomes of strand passages in more complex knots we studied topological transitions in left-handed version of 5 knots since left-handed twist knots arise naturally as result of intermolecular passages in negatively supercoiled DNA molecules (,) and these DNA knots were used in experiments that demonstrated selective pathway of knots’ relaxation (). C shows that random passages within 5 knots result in roughly equal proportions of unknots and 3 knots (. 10% each), and in a high proportion of strand passages that maintain the original topology (>50%). Selection of inter-hooked juxtapositions, however, greatly stimulates knots simplification and preferentially leads to events that cause unknotting (∼40% of events) rather than to forming trefoil knots (<20% of events). Mann () observed that human topo II acting on 5 knots showed high selectivity for passages that directly lead to formation of unknots, while passages to 3 knots were hardly observed. In order to understand the possible basis of this specificity we investigated how the selection of specific geometry of the juxtaposition affects the topological consequences of the strand passages occurring within 5 knots. A shows how the mean opening angle [(α + α)/2], calculated for both two-segment subchains forming juxtapositions (), influences the efficiency of unknotting and of other topological transitions resulting from a strand passage at these juxtapositions. It is visible that the narrower the mean opening angle the more efficient the unknotting. The unknotting approaches 40% frequency for the juxtapositions with the mean opening angle between 0° and 22.5° and it remains below 7% for the mean opening angle between 157.5° and 180°. The probability of not changing the knot type after a strand passage has the opposing tendency. For juxtapositions with the mean opening angle <22.5° only 30% of passages maintained the topology of 5 knot, while for the mean opening angles between 157.5 and 180° ∼80% of passages did not change the knot type. The observations that strongly hooked juxtapositions (with small opening angle) are likely to give rise to simpler knots upon strand passage while weakly hooked juxtapositions (with large opening angle) result in a passage that does not change the topology are consistent with the proposal by Buck and Zechiedrich () and with the simulations studies by Liu (). In general, inter-hooked juxtapositions are more likely to be encountered in the knotted portion of a random chain. Therefore, passages occurring at such juxtapositions are likely to result in a simplification of the knot. Unhooked juxtapositions are more likely to occur between subchains that are not prompted to bend over each other as those in a trivial loop. Therefore, strand passages at such juxtapositions are more likely to result in no change of the topology. More difficult to explain is the observation that the ratio of passages to unknots versus those to 3 knots strongly decreases (from three to one) as the mean opening angle increases. Apparently, the portion of the knot (topological domain) where passages between opposing segments lead to formation of unknots is more bent and has the opposing segments more inter-hooked with each other than the topological domain within which inter-segmental passage leads to creation of trefoil knots. More detailed description of topological domains of various knots can be found in Ref. ().
B shows the effect of the geometric chirality of the juxtaposition on the topological outcomes of passages occurring at these juxtapositions. The concept of geometric chirality is explained in the inset to . To characterize the geometric chirality of juxtapositions we measure it for the two unoriented tangential lines a and a passing through the vertices denoted as A and A on . In contrast to A, where unknotting frequency of inter-segmental passages is progressively decreasing with the increase of the mean opening angle, the unknotting dependence on the geometric chirality of juxtapositions is more complex. Formation of unknots is strongly favoured by inter-segmental passages at juxtapositions with strong right-handed geometric chirality while it is least efficient for juxtapositions that are achiral or show weak left-handed chirality. Interestingly, the formation of unknots increases again for juxtapositions with strong left-handed chirality. Formation of trefoil knots is more efficient than that of unknots for juxtapositions that are achiral or show a weak left-handed chirality. Apparently, the juxtapositions that after a passage lead preferentially to creation of 3 knots frequently show weak geometric chirality, while the juxtapositions leading to the preferential formation of unknots are usually strongly chiral. C shows the dependence of the topological outcomes on the distance between the two central vertices A and A of each subchain forming the juxtapositions (). Interestingly, the distance of approach does not influence significantly the topological consequences of strand passages occurring at these juxtapositions. Notice though that all juxtapositions analysed in were pre-selected to be weakly hooked, i.e., have the γ angle >90° and satisfy the criteria that one subchain pierces the sector of plane defined by the other subchain ().
We have shown above that the selection using just one criterion, like the mean opening angle (A) or the geometric chirality angle, can result in up to 3-fold higher conversion rate to unknots than that to 3 knots. However, the steric selectivity of DNA topoisomerases for the geometry of juxtapositions is most likely a combination of several criteria. A and B show how the ratio of passages from 5 knots to unknots versus those from 5 knots to 3 knots depends on various pair-wise combinations of selection criteria. A uses a colour code to show how the combination of opening angles (α) at both subchains forming the selected juxtaposition () affects the topological consequences of the reaction. It is visible that the 0/3 ratio can reach a value as high as six depending on the combination of opening angles α. The majority of the configuration space where the transitions to unknots are highly favoured over those to 3 knots, are in the region where both opening angles α are small (<25°). However, the 0/3 ratio also reaches a moderate value in a region of the configurations space where one subchain has a very small opening angle (<10°) and the other has a large opening angle of (130–170°). This would correspond to a situation where a strongly bent subchain is hooked over a rather straight subchain. B illustrates how the combination of the mean opening angle and the geometric chirality angle affects the 0/3 ratio resulting from passages at the corresponding juxtapositions. The ratio between passages to unknots and those to 3 knots exceeds six for juxtapositions with strong negative geometric chirality and having in addition small average opening angle. The examples shown in demonstrate that steric selectivity can very well lead to a situation where the great majority of topoisomerase-mediated passages in 5 knot could result in unknots (). Specific combinations of more than two criteria are likely to define a portion of the configurations space where juxtapositions would result in a still higher bias toward passages to unknots than those shown in A and B. Although we do not know yet what the rules of steric selection are in the case of human topoisomerase II we should be able to find out what are the steric properties of 5 knots that could be used to guide the topoisomerase to convert it into unknots rather than into trefoils.
Looking at standard diagrams of 5 knot () one sees no reasons why low geometrical chirality of juxtapositions should favour passages to 3 knots. However, standard minimal diagrams of knots may not reflect the average 3D structure of a random chain forming a given knot. It was shown earlier that ideal geometric configurations of a given knot reflect certain global statistical properties of corresponding random knots of a given type (). Thus, for example, the writhe of ideal knots had practically the same value as the average writhe of randomly fluctuating knotted chains forming the same knot type (,) or even random knots of the corresponding type formed on a cubic lattice (). We decided therefore to have a closer look at the ideal configuration of the 5 knot hoping to see some hints why the selection of inter-hooked juxtapositions favours specifically transitions to unknots. C shows the ideal geometric representation of the knot 5. Inter-segmental passages between opposing segments in a region presented on a yellow background would lead to formation of trefoil knots and it is visible that the two big loops forming this region are not inter-hooked but rather point in the same direction (if both loops were reduced to two segment subchains, the bisectrices would point in both cases in the same direction i.e. toward right of the figure). Inter-segmental passages between opposing segments in a portion of the knot that is drawn on dark blue background (C) would lead to formation of unknots. Interestingly, the opposing regions there are inter-hooked. Therefore, if random configurations of knots retain some characteristics of ideal knots then selection of inter-hooked juxtapositions in 5 knots is more likely to cause unknotting than a change to 3 knot. Can ideal configurations also explain the data presented in A and B? From the data presented in A, we can conclude that in the portion of 5 knot where passage between opposing segments leads to the creation of unknots, the juxtapositions of highly bent subchains are much more likely to be present. Indeed, our measurements of local radius of curvature in the ideal geometric configuration of 5 knot revealed that regions where passages between opposing segments leads to formation of unknots are significantly more bent than the regions where inter-segmental passages lead to formation of trefoil knots (data not shown). Similar observations can also be made for chirality. The opposing segments in the domain in which a strand passage leads to formation of 3 knots (yellow background) are practically perpendicular to each other and thus show low chirality. Strong negative chirality is however observed within a domain in which passage between opposing subchains leads to formation of unknots (shown on dark blue background). Our analysis suggests, therefore, that ideal geometric configurations of knots reflect not only global statistical properties of random knots of a given type such as time-averaged writhe (,) but also such local characteristics like the curvature and chirality () within corresponding topological domains of a given knot.
The presented simulation study has dealt only with the unknotting activity of type II DNA topoisomerases, however the same selection principle is also expected to greatly favour the decatenation activity of type II DNA topoisomerases (). Simulation studies using polygons confined to a cubic lattice have in fact demonstrated that the selection of hooked juxtapositions strongly stimulate both unknotting and decatenation of modelled DNA rings (,). Until recently, a much higher biological importance was attributed to the decatenation activity of type II DNA topoisomerases than to their unknotting activity, since catenated and pre-catenated rings are normal intermediates of DNA replication process (,) and it is essential to remove all interlinking between daughter DNA molecules before the completion of the cell division (,). A very recent study showed however that unknotting activity is also of very high biological importance as the presence of knots in DNA was shown to promote severe replicon dysfunctions and induced high mutation rate ().
For the description of the selection model to be more complete, one needs to consider at which point of the reaction the energy gained from ATP hydrolysis is used by type II DNA topoisomerases. It is important to realize that topoisomerase-mediated strand cleavage and re-sealing are isoenergetic, reversible reactions, as the energy of the hydrolysed phosphodiester bond in the DNA is maintained in a phosphodiester bond between the active tyrosine residue of the protein and the phosphate group of the DNA. Therefore, type I topoisomerases do not depend on the energy gained from ATP hydrolysis to perform efficient relaxation of supercoiled DNA. However, relaxation reactions mediated by type I DNA topoisomerases can only diminish the free energy of DNA substrates and therefore can be considered as driven by the DNA relaxation. For thermodynamic reasons the catalytic action of topoisomerases that is not coupled to ATP hydrolysis cannot move the system out of topological equilibrium (). To understand better the energetic requirements of the reaction let us consider what could happen if type II DNA topoisomerases with high affinity for hooked juxtapositions were not able to use the energy gained from ATP hydrolysis. Without ATP hydrolysis the enzyme can bind very efficiently to hooked juxtapositions as this step of the reaction is driven by the binding energy. Subsequently, the enzyme can perform the inter-segmental passage from inside to the outside of the inter-hooked juxtaposition, provided that the affinity of the enzyme for the DNA arrangement after the passage is higher than to the DNA arrangement before the passage. In fact, all these steps of the reaction can be performed by type II topoisomerases when ATP is replaced by non-hydrolysable analogues (). For all these steps of the reaction to proceed in a correct order without the need of ATP hydrolysis, it is necessary that the energy of the system progressively diminishes and this means that the complex between the topoisomerase and the DNA after the passages should be very stable. As a consequence, at the end of the reaction one would need the energy to dissociate this very stable complex so that the DNA and topoisomerase could function again. In biochemical experiments that established that type II DNA topoisomerases can perform one cycle of the reaction without ATP hydrolysis, the ‘dead-end’ complexes between the enzyme and the DNA were disrupted by SDS and proteinase K () and this was equivalent to providing the energy to dissociate those stable complexes. In reactions performed in the presence of ATP, the energy gained from ATP hydrolysis is used by the enzyme for its resetting for next rounds of the reaction. In principle, the energy of ATP hydrolysis occurring at just one point of the catalytic cycle can be stored in form of a mechanical stress that can be then used stepwise to induce a cascade of conformational changes required during several distinct steps of a complex enzymatic reaction. Observations showing that, in the presence of non-hydrolysable analogues of ATP, type II topoisomerases can perform one round of the reaction but are unable to perform another round () indeed suggested that ATP hydrolysis only happens after the strand passage is complete. However, more recent studies suggested that two ATP molecules bound in quasi-symmetrically located regions of the dimeric enzyme are hydrolysed sequentially at two specific points during each catalytic cycle of type II topoisomerases (,). The first hydrolysis precedes the phase of inter-segmental transport and therefore could be directly coupled to the active transfer () or to the kinetic proofreading mechanism (). Since the complete strand passage happens also without the ATP hydrolysis it is likely that the energy gained from the hydrolysis at this point just speeds up of the reaction or is used for quality control. The second hydrolysis happens at the end of the reaction (), which is consistent with the earlier proposal that ATP hydrolysis serves to dissociate the enzyme from the DNA after inter-segmental passage and is used for resetting the enzyme for another binding and cleavage cycle of the reaction (,). The principle of using the energy of ATP hydrolysis to dissociate stable complexes formed between the enzyme and DNA products of the reaction is frequently encountered in biological systems and RecA protein, for example, uses its ATPase activity to release itself from the stable complex formed with the DNA after the strand-exchange reaction is completed ().
#text |
RNAs are capable of carrying out a multitude of diverse biological functions. Many biologically active RNAs have to adopt intricate 3D structures that rival protein structures in their complexity to be functional in a cellular environment.
Folding of RNA chains into compact globular 3D structures represents a problem. Due to the polyanionic character of the RNA phosphodiester backbone, it inevitably brings negative charges in close proximity to each other. This results in highly unfavorable electrostatic interactions that are anisotropically distributed in the folded form of the RNA. Therefore, divalent metal ions in general and Mg with its favorable charge/size ratio in particular () play an important role in RNA folding. Mg-ions not only stabilize the final structure through either direct coordination (inner sphere contact) with negatively charged groups of the RNA or in a water-mediated interaction (outer sphere contact) with the hexahydrated ion (Mg(HO)). They also influence the rate of folding, stabilize folding intermediates or destabilize alternative conformations ().
Riboswitches control the expression of a significant number of bacterial genes by binding with high affinity and specificity to small metabolite molecules. To achieve the binding specificity and affinity required for their function, the ligand binding or aptamer domains of these RNA elements must fold into intricate tertiary structures. This is illustrated in a number of X-ray structures of aptamer domain/ligand complexes of different riboswitches (). In many cases, the X-ray structures revealed the presence of defined Mg-binding sites.
For instance for the thiamine pyrophosphate (TPP)-sensing riboswitch it was shown that Mg binding is absolutely required for TPP binding () and cannot be replaced by monovalent ions or Ca (). The X-ray structure of the TPP-sensing riboswitch from bound to TPP reveals that one Mg-ion chelates oxygen atoms in the pyrophosphate moiety of TPP and links them to the RNA whereas there are two bridging Mg-ions in the X-ray structure of the TPP-sensing riboswitch from bound to TPP. The presence of bridging Mg-ions in the X-ray structures is consistent with the strict Mg requirements for TPP binding to the TPP-sensing riboswitch (). In contrast, the GlmS riboswitch—a ribozyme that undergoes a self-cleavage reaction in the presence of glucosamine-6-phosphate (GlcN6P)—does not strictly require Mg for the self-cleavage reaction. Here, Mg can be substituted by either other divalent metal ions or Co(NH) and to some extend by high concentrations of monovalent ions () indicating that the metal ion is not involved in catalysis but only in electrostatic stabilization of the structure consistent with the X-ray structure of this complex ().
The purine-sensing riboswitches are among the smallest riboswitches found so far. All purine riboswitches fold into a three-way junction (A) where central structural elements such as the ligand-binding core region and the loops L2 and L3 that cap helices II and III, respectively, show a very high degree of sequence conservation. However, despite the sequence conservation and similarity in structure, the adenine-sensing riboswitch binds adenine with high specificity while the guanine-sensing riboswitch binds guanine. This specificity is mediated by a single nucleotide in the ligand-binding core region which is a cytidin in the guanine-sensing riboswitch and a uridine in the adenine-sensing riboswitch (,). It was shown that the purine ligand is bound to this specific nucleotide in the core region by forming an intermolecular Watson–Crick base pair ().
The importance of Mg-ions for ligand binding to these riboswitches is less clear. Thermodynamic studies of ligand binding supported a requirement for the presence of Mg in the case of hypoxanthine binding to the guanine-sensing riboswitch () and FRET studies of the adenine-sensing riboswitch from suggested an important role for Mg in ligand binding (). Specifically, Mg was required for the formation of the long-range base-pairing interactions between nucleotides in loop 2 and 3. Furthermore, the X-ray structure of the adenine-sensing riboswitch bound to adenine revealed five well-defined Mg-binding sites () whereas 11 bound Co(NH)-ions were found in the X-ray structure of the guanine-sensing riboswitch in complex with hypoxanthine (). In contrast, NMR studies found ligand binding to be independent of the presence of Mg for both the adenine and the guanine-sensing riboswitch (,). In addition, in the guanine-sensing riboswitch from the loop–loop interaction between loops 2 and 3 is preformed in the free form of the riboswitch and becomes strengthened in the presence of Mg but is stable enough to exist even outside the context of the riboswitch (). Given the high degree of sequence identity between the guanine- and adenine-sensing riboswitch (A) the requirement for Mg for the formation of the loop–loop interactions in the adenine-sensing riboswitch as reported in the FRET studies () is surprising.
Specifically, with respect to the loop sequences the guanine- and the adenine–sensing riboswitch differ only at the position 32 (A) in loop 2. At position 32, the guanine-sensing riboswitch bears a guanosine whereas it is an adenosine in the adenine-sensing riboswitch. The X-ray structures of both riboswitches show that the purine at position 32 stacks between the apical closing base pair of helix II and A33 without making any stabilizing hydrogen bonds and that the residue at position 62 loops out into the solvent (,). In addition, there are differences in the closing base pairs of helices II and III in the two riboswitches. The base pair adjacent to loop 2 is an asymmetric U:U base pair in the adenine-sensing riboswitch and a Watson–Crick G:C base pair in the guanine-sensing riboswitch. The closing base pair of helix III is a non-canonical A:A base pair in the adenine-sensing riboswitch but a Watson–Crick A:U base pair in the guanine-sensing riboswitch.
The reported discrepancies in the behavior of the closely related adenine- and guanine-sensing riboswitches prompted us to investigate the binding of divalent metal ions to the adenine-sensing riboswitch and their influence on ligand-induced RNA folding in solution by high-resolution NMR methods in this study.
N,C-labeled adenine was synthesized as described in (). The ligand concentration was determined by UV absorbance at 261 nm using an extinction coefficient ɛ = 13 400 mol cm ().
Labeled RNAs were prepared by transcription with T7 RNA polymerase from linearized plasmid DNA templates using N- and N,C-labeled nucleotides purchased from Silantes (Munich, Germany). The labeled RNA was purified as described () and concentrations were determined by UV absorbance at 260 nm. The RNA was folded into a homogenous monomeric conformation through denaturation at 363 K for 5 min and rapid cooling to 273 K by 10-fold dilution with ice-cold HO. Folding into a homogenous monomeric form was verified by native PAGE (data not shown). The RNA was subsequently exchanged into NMR buffer (25 mM KPO, pH 6.2, 50 mM KCl) using Centricon-10 microconcentrators.
All NMR experiments were recorded on Bruker NMR spectrometers operating at 600, 700, 800 and 900 MHz equipped with a 5 mm HCN cryogenic probes and -axis gradients. The spectra were recorded and processed using Bruker X-WIN NMR3.5 and TopSpin2.0 and were analyzed using XEASY (). All NMR spectra were recorded at a temperature of 283 K in 90% HO/10%DO in NMR buffer (25 mM KPO, pH 6.2, 50 mM KCl). Water suppression was achieved using the WATERGATE water suppression scheme () including water flip back pulses (). H,N-HSQC, J-H,N-HSQC () and 2D- or N-edited 3D-NOESY experiments were performed using standard pulse sequences (). The J-HNN-COSY experiments () were performed as described previously ().
Titrations of the RNA with Mg, Mn and Co(NH) were performed directly in the NMR tube by adding aliquots from 125 mM or 500 mM MgCl, 1 mM MnCl or 100 mM Co(NH)Cl stock solutions. The equilibrium binding constant for Mg to the RNA was measured by the chemical shift perturbation (CSP) of the imino group of G37. CSPs were calculated using the formula and were correlated with the ratio of [Mg]/[RNA]. Non-linear regression using SigmaPlot9.0 was performed using the following fitting function:
and are the fitted parameters where is the ratio of the dissociation constant () to [RNA] and is the CSP for infinite Mg concentrations ().
The imino region of the H,N-HSQC of the aptamer domain of the adenine-sensing riboswitch in the absence of Mg shows well-resolved and sharp signals in the presence of 1.2 equivalents adenine (B). The presence of a signal for the N9H9 imino group of the adenine ligand, which is undetectable in its free form indicates the formation of a high-affinity RNA–ligand complex in the absence of Mg. By analyzing 2D-H,H-NOESY and 3D-H,H,N-NOESY-HSQC experiments (data not shown) the imino resonances of the RNA–adenine complex could be assigned. The sequence of the aptamer domain of the riboswitch RNA contains 16 guanosine and 19 uridine nucleotides (A). For 12 of the guanosine and 17 of the uridine nucleotides imino resonances are detectable in the H,N-HSQC spectrum at 10°C. Imino resonances that could not be assigned correspond to residues G13, G14, U36, G48, G62 and U63. Imino resonances of terminal nucleotides such as G13 and G14 in RNA helices are often not detectable in NMR spectra due to their reduced stability and the resulting fast exchange of the imino proton with the bulk water solvent. The imino groups of residues U36, G48, G62 and U63 are solvent exposed according to the X-ray structure of the adenine–RNA complex () which also renders them undetectable due to fast exchange with the bulk water solvent.
The binding mode of adenine to the adenine-sensing riboswitch RNA and tertiary structure characteristics of the adenine–RNA complex were further elucidated by an HNN-COSY experiment () which detects hydrogen bonds of the type N-H…N. In the HNN-COSY spectrum of the adenine-sensing riboswitch in complex with adenine (C), the imino groups of U74 and of U51 show correlations with the N1 and the N3 nitrogen of the bound adenine, respectively. U74, U51 and adenine form an intermolecular base triple (D) as observed previously for a mutant of the adenine-sensing riboswitch aptamer domain () and in the X-ray structure of the adenine-sensing riboswitch RNA in complex with adenine crystallized in the presence of 200 mM Mg (). Furthermore, several non-canonical base pairings and tertiary interactions can be identified in the HNN-COSY or the NOESY spectra. In the HNN-COSY spectrum, the imino group of U49 is correlated to an adenosine N3 as identified by its chemical shift which is consistent with the hydrogen bond between the U49 imino group and the N3 of A76 in the X-ray structure (E). The imino group of U34 shows a correlation to the N7 nitrogen of an adenosine as expected for the reversed Hoogsteen A65:U34 base pair () (F) observed in the X-ray structure. Nuclear Overhauser effect (NOE) data and the chemical shifts of the C2 and C4 carbonyl groups of U31 and U39 indicate that these nucleotides form an asymmetric U:U base pair (data not shown) () also in agreement with the X-ray structure.
In summary, the number of observed imino proton signals, the obtained assignments, the observed intra- and intermolecular NOEs, the binding mode of adenine to the RNA as deduced from the HNN-COSY experiment as well as the observed tertiary non-canonical base-pairing interactions strongly indicate that the teriary structure of the adenine-sensing riboswitch aptamer domain bound to adenine in solution in the absence of Mg-ions is virtually identical to the corresponding X-ray structure obtained in the presence of high concentrations of Mg. Therefore, the presence of Mg-ions is not a prerequisite for either ligand binding or for the correct folding of this RNA–ligand complex.
We applied three different but complementary techniques to characterize divalent metal binding to the adenine–RNA complex and to analyze possible structural effects in solution by NMR spectroscopy. First, we observed CSPs in Mg-titration experiments caused by the interaction of the metal ion with the RNA–ligand complex. Second, we used the Mn-ion as a paramagnetic Mg-analog to observe paramagnetic line broadening for NMR signals in close proximity to putative divalent metal-binding sites. Third, we utilized the [Co(NH)]-ion which is considered to be an analog of an hexahydrated Mg-ion in CSP and NOE experiments to identify imino groups spatially close to putative divalent metal-binding sites.
We titrated the adenine-sensing riboswitch RNA–adenine complex with Mg and compared the imino group region of the H,N-HSQC spectrum (A). The number of the imino resonances remains unchanged and no large chemical shift changes are observed. This indicates that the overall structure of the complex is not perturbed upon Mg binding and that no new structural elements are formed. However, while some imino resonances show no or only small CSP other resonances showed significant CSP upon addition of 5 mM Mg to the adenine–RNA complex. Such significant CSPs were observed for U22, G37, G38, U39, U47, U71 and the adenine N9H9 imino group. When these imino groups are mapped on the X-ray structure of the complex, four well-defined areas are visible which are affected by Mg binding (A). G37 and G38 are located in loop L2 and are involved in the loop–loop interaction forming long-range Watson–Crick base pairs with C61 and C60 in L3, respectively. U39 is located at the apical tip of helix II, U71 is situated in the lower half of helix III pointing to the core region whereas U22, U47 and the adenine N9H9 imino group are all at the interface of the RNA and the bound ligand.
Due to enhanced spin relaxation binding of the paramagnetic Mn-ion leads to line broadening of resonances in close proximity to the ion with a distance dependence of 1/r (). Only micromolar concentrations of Mn are required to induce the effect (). In the RNA–adenine complex in the presence of 5 mM Mg imino resonances that show significant line broadening upon titration with 15 µM Mn are those of U22, G37, G38, G44, G46, U47, G56, U71 and G72 (B). Except for U39 and the N9H9 group of the bound ligand all the resonances that show significant CSP in the Mg -titrations also show paramagnetic line broadening induced by the presence of Mn. Of the four resonances that show line broadening but no CSP G56 and G72 are close in space to U71 that is affected by both ions. Similarly, G46 where only line broadening is observed is adjacent to U47 sensitive to both ions. The residues showing Mn-induced line broadening are clustered in four distinct regions when mapped onto the structure (B). Three of those regions were previously identified in the Mg -titrations. These regions are close to G37 and G38 involved in the loop–loop interaction, the center of helix III (U71, G56 and G72) and in the ligand-binding core (U22, G46 and U47), respectively. G44 located in the lower half of helix II does not show CSP upon titration with Mg and is not close in space to other residues that are affected by the presence of either Mg or Mn.
Co(NH3) has a geometry similar to the hexahydrated Mg(HO)-ion. However, the ligand shell of amino groups is inert to exchange and in contrast to Mg cannot form inner sphere contacts (). In order to complement our magnesium-binding studies we used the adenine–RNA complex and performed Co(NH) -titration experiments. As for the Mg -titration of the complex, the overall conformation of the RNA–ligand complex remained unperturbed and no additional structural elements are formed as indicated by the constant number of observable imino resonances. In a H,N-HSQC spectrum the imino resonances of U17, U18, U22, G37, G38, G44, G46, U47, G56, G57, U71 and G80 showed significant CSP upon addition of 3 mM Co(NH) (C).
The binding site of Co(NH) to the RNA–adenine complex cannot only be studied by CSP but also by intermolecular NOEs between the protons of the NH-groups of Co(NH) and RNA protons in spatial proximity. Co(NH) in contrast to Mg(HO) is suitable for such NMR experiments since the hydrogens of the ammonia ligands of Co(NH) exchange slowly with the bulk solvent water and are therefore detectable by NMR. We recorded a 2D-H,H-NOESY spectrum for the adenine–RNA complex in the presence of 5 mM Co(NH) in which we detected intermolecular NOEs from the protons of Co(NH) to imino group protons of U22, G38, G46, G47, G56, G57, U71, G72 and the N9H9 imino group of the bound adenine (D). The residues showing either significant CSP upon the addition of 3 mM Co(NH) or having an intermolecular NOE in the presence of 5 mM Co(NH) include all the residues affected by paramagnetic line broadening in the presence of Mn-ions. G57 which only shows NOEs and CSP to Co(NH) but no Mn-induced line broadening is located in the center of helix III close to G56, U71 and G72 that are affected by both the presence of Co(NH) and Mn. Similarly, the N9H9 group of the bound adenine which displays an NOE to Co(NH) and CSP in the presence of Mg but no line broadening in the presence of Mn is in the ligand-binding core in close proximity to U22, G46 and G47. The imino resonances of U17, U18 and G80 in helix I only showed CSP in the presence of Co(NH) but not in the presence of Mg and no line broadening in the presence of Mn-ions.
Mapping of the imino groups experiencing CSP and showing NOEs in the presence Co(NH) on the X-ray structure reveals five distinct binding sites for Co(NH) to the adenine-sensing riboswitch RNA in complex with adenine (C).
Three of these binding sites have been identified before in the titration experiments with Mg and in the line-broadening experiments with Mn. These are located in the vicinity of G37 and G38 in the loop–loop interaction, the center of helix III (U71, G56 and G72) and in the ligand-binding core (U22, G46 and U47), respectively. This is further illustrated by mapping only those imino groups with resonances perturbed by the presence of all three Mg-, Mn- and Co(NH)-ions on the X-ray structure (D). One binding site located in the center of helix II is only identified in the line-broadening experiment with Mn and by CSP in the presence of Co(NH)-ions.
The X-ray structure shows five hexahydrated Mg-ions [Mg1 to Mg5, nomenclature of Mg atoms as in (), D] bound to the adenine-sensing riboswitch in complex with adenine (). One of these Mg-binding sites (Mg3) is involved in a crystal packing interaction (). Mg1 in the X-ray structure binds to nucleotides in J2-3 and at the lower end of helix II close to U22 and U47. U22 and U47 have been identified in all four experiments as proximal to a divalent metal ion-binding site in solution. Mg2 in the X-ray structure is located close to U71 in helix II that has been identified as a consensus-binding site in solution. Divalent metal ion binding to the lower end of helix II in solution that would correspond to Mg4 in the X-ray structure was only evident in the line-broadening experiments with Mn and in the titrations with Co(NH) (G44). Therefore, this Mg-ion might bind only weakly.
No divalent cation binding in solution could be detected in the vicinity of the positions for Mg3 where the binding site is formed by crystal packing interactions and Mg5 in the X-ray structure. On the other hand, no bound Mg is observed in the X-ray structure in the vicinity of G37 and G38 in the loop–loop interaction where all the solution experiments unambiguously indicate the presence of a divalent cation-binding site. Interestingly, Mg binding at this site appears to be important for the formation of the loop–loop interaction in the adenine-sensing riboswitch.
The H,N-HSQC spectrum of the free form of the adenine-sensing riboswitch RNA in the absence of metal ions shows imino resonances that are broad and that vary largely in intensity (A). In comparison with the adenine bound form, the number of signals is significantly smaller in the free state of the RNA. The signals for U20, U22, G45, G46, U47, U49, U51, U71, G72, U74, U75 located in the ligand-binding core and at the ends of the helices facing the core region are completely absent. This indicates that there are no stabilizing hydrogen-bond interactions between nucleotides in the core region and the core region apparently is largely disordered. In addition, the signal for the imino group of G38 is absent and those of G37 and U34 are barely detectable. This indicates that the loop–loop interaction between loop 2 and loop 3 is strongly destabilized in the free form of the RNA. Only a minor part of the RNA apparently samples a conformation where the two loops interact in exchange with totally open conformations.
However, besides the signals of imino groups belonging to nucleotides in stable Watson–Crick base pairs in the central regions of helices I, II and III, respectively, there are five uridine imino resonances and one guanosine imino resonance with chemical shifts indicating involvement in non-canonical base-pairing interactions. NOEs observed between these imino resonances (B) in conjunction with the H,N-HSQC-spectrum (C) suggest the presence of two asymmetrical U:U base pairs and of one G:U wobble base pair in the free RNA. However, there is only one asymmetrical U:U base pair (U31:U39) observed in the adenine bound form of the RNA. In addition, there is no evidence for the presence of a stable G:U wobble base pair in the adenine bound RNA. The second U:U base pair and the G:U wobble base pair therefore represent an alternative base-pairing pattern unique to the free form of the RNA and have to dissociate in the process of ligand binding. The location of these alternative base pairs either in the loops 2 and 3 or the ligand-binding core cannot unambiguously be established due to the lack of NOEs to other imino protons. In addition, the presence of the second U:U base pair could be either due to an alternative conformation for the U:U base pair between U31 and U39 (D) or due to an additional long-range U:U base-pairing interaction. In the first case, exchange cross peaks should be observable between the uridine imino resonances involved in the U:U base-pairing interactions. Unfortunately, both the proton and the N chemical shifts of the uridine imino resonances are very similar and such cross peaks would be obscured by the strong diagonal peaks in either H,H-NOESY spectra and N-ZZ-exchange experiments (). However, Mg-titration experiments of the free RNA suggest that the alternative G:U and U:U base pairs are located in the region of the loop–loop interaction and compete with the proper formation of the long-range base pairs between loops 2 and 3.
We titrated the free form of the adenine-sensing riboswitch RNA with Mg and observed changes in the signals of the imino groups in a H,N-HSQC spectrum. The overall number of imino residues decreases and the resonances reach equal intensities in the presence of 5 mM Mg (A). During the stepwise addition of Mg-ions a signal for the imino group of G38 becomes observable and signals for G37 and U34 strongly increase in intensity. This indicates that the presence of Mg promotes the formation of the long-range base-pairing interactions between loops 2 and 3. The signals for G37 and G38 continuously shift upon increasing the Mg concentration. This suggests that Mg shifts the fast equilibrium between open conformations without a loop–loop interaction and closed conformations strongly toward a closed conformation with a stable interaction between loop 2 and loop 3.
At the same time the stepwise addition of Mg causes the imino proton signals of the non-canonical G:U wobble base pair and one of the U:U base pairs to disappear whereas the imino protons of the other U:U base pair shift strongly towards chemical shifts similar to those of the adenine bound RNA in the presence of Mg. Therefore, the formation of the ‘native’ stable loop–loop interaction upon addition of Mg directly leads to the dissociation of the alternative base pairs observed in the free RNA in the absence of Mg. Taken together, these observations indicate that the ensemble of rapidly interconverting conformations observed for the free form of the RNA in the absence of Mg converges to a more homogeneous conformation where the loop–loop interaction is present and stable in the presence of Mg.
The stepwise CSP of the G37 imino group during the Mg -titration was used to estimate the dissociation constant for Mg binding to the area of the loop–loop interaction (B). G37's imino resonance strongly increases in intensity and shifts subsequently by a total of 0.15 p.p.m. in H and 0.35 p.p.m. in N during the titration in agreement with a rapidly associating and dissociating complex of Mg and RNA. Correlating the CSP for each titration point with the ratio [Mg]/[RNA] yielded a saturation curve revealing an equilibrium dissociation constant of 1.5 mM ± 0.4 mM for Mg to the loop L2 region after non-linear regression (C).
Interestingly, titrations of the free RNA with Co(NH)-ions lead to virtually identical results with regard to the formation of the native loop–loop interaction and the destabilization of the alternative base-pairing patterns. Therefore, Mg apparently induces the native folding of this RNA purely by using outer-shell coordination.
In order to probe the effect of the stability in helix II on the loop–loop formation properties in the free form of the adenine-sensing riboswitch RNA, we introduced a stabilizing mutation in helix II. In this A30G/U40C double mutant the A30:U40 base pair adjacent to the U31:U39 base pair is replaced by a more stable G:C base pair. In fact, the guanine-sensing riboswitch displays a U:A base pair at position 30–40 like the adenine-sensing riboswitch but in contrast the adjacent base pair facing the loop–loop interaction is a stable canonical G:C base pair in the case of the guanine-sensing riboswitch and not a weaker non-canonical U:U base pair as the case in the adenine-sensing riboswitch (A and inset). The A30G/U40C double mutant binds to adenine in a manner identical to the wild-type RNA (). The imino group H,N-HSQC spectrum of the free form of the mutant shows significant differences compared to the one of the wild-type. In the mutant imino group resonances for G37 and U34 which are located in the loop L2 and form long-range base pairs with nucleotides in loop L3 are readily detectable () with chemical shifts similar to those in the adenine bound RNA. In addition, only one U:U base pair is detected which can be assigned as the U31:U39 base pair and no signals corresponding to a G:U wobble base pair can be found. These data show that the mutant RNA in contrast to the wild-type RNA can already form a stable native loop–loop interaction in the free form in the absence of Mg.
Divalent cation binding plays a prominent role for RNA folding and stabilization. Mg-ions in particular are often required for the proper folding of complex RNA structures giving rise to the notion of an Mg-ion-core () in highly structured globular RNAs in analogy to the hydrophobic core of protein structures. The X-ray structures of the aptamer domains of the closely related guanine- and adenine-sensing riboswitches are good examples for such intricate globular RNA structures. Not surprisingly, those structures revealed a number of well-defined binding sites for either hexahydrated Mg-ions or its structural analog the Co(NH)-ion. Furthermore, an important role for Mg-ions for the folding and ligand binding of the adenine-sensing riboswitch was reported. On the other hand, our own NMR investigations of the guanine-sensing riboswitch despite revealing a number of well-defined divalent cation-binding sites in solution showed that Mg binding was not required for ligand binding and that in the free state of the RNA important tertiary interactions were already pre-organized in a Mg-independent manner.
Here, we investigated the binding of divalent cations to the aptamer domain of the adenine-sensing riboswitch in solution and its importance for ligand binding and folding of this RNA. We found that although adenine binding to the aptamer domain is independent of the presence of divalent cations there are a number of well-defined binding sites for divalent cations on the adenine–RNA complex in solution. In contrast to the related guanine-sensing riboswitch, the free state of the adenine-sensing riboswitch is conformationally heterogeneous and displays alternative base-pairing patterns detrimental to ligand binding. The addition of Mg-ions induces the formation of a crucial tertiary interaction, the formation of base-pairing interactions between nucleotides in loops 2 and 3, and in turn destabilizes the alternative base pairs thereby pre-organizing the structure of the RNA for ligand binding.
Possible divalent cation-binding sites in solution were identified by using four different techniques: Mg-induced CSP, paramagnetic line broadening induced by Mn-ions, CSP induced by Co(NH)-ions and intermolecular NOEs between Co(NH)-ions and RNA protons. Despite differences in their inherent sensitivity these experiments yielded consistent results. Three distinct Mg-binding sites were identified in all four experiments. Apparently, in all three binding sites the ion binds through ‘outer shell’ coordination since Mg titrations and Co(NH)-binding experiments show similar effects. Two of those binding sites are also seen in the X-ray structure and were also found in solution for the structurally very similar guanine-sensing riboswitch. The third binding site identified in all experiments was not seen in the X-ray structure. It is located close to nucleotides involved in tertiary base-pairing interactions between the loops 2 and 3. This is consistent with FRET results showing that Mg binding promotes the formation of this loop–loop interaction already in the absence of ligand (,) and with our own results that show Mg-dependent formation of the tertiary base-pairing interactions between nucleotides in these two loops. A divalent cation-binding site at this position was also identified in solution for the structurally highly similar guanine-sensing riboswitch (). A fourth possible binding site was only identified in the line-broadening experiments with Mn and the Co(NH)-binding experiments. Interestingly, this binding site also corresponds to the position of a bound Mg-ion in the X-ray structure. Probably the affinity of this site is lower and the magnitude of the Mg-induced CSP is too small to identify this site in the Mg experiments. Two Mg-binding sites found in the X-ray structure (Mg3 and Mg5) were not found in solution. One of those is created by crystal packing interactions (Mg3). In principle, the failure to detect the Mg-binding site corresponding to Mg5 () of the X-ray structure might be due to our exclusive use of guanosine and uridine imino groups as probes for metal-induced chemical shift changes. This would prevent the detection of metal ion binding in A, C-rich regions of the structure due to the lack of suitable probes. However, Mg5 is close in space to the imino groups of U74 and U75. Thus, our results suggest that this binding site is apparently not or only rarely occupied by a divalent cation in solution.
On the other hand, Co(NH)-ions in solution seem to bind to a site in helix I in an area with a high density of uridine carbonyl groups where neither the X-ray structure nor the Mg- and Mn-titration experiments showed a binding site. Due to their higher charge density Co(NH)-ions have an ∼10-fold higher affinity () than divalent ions and induce larger CSPs than divalent ions so it seems likely that in this experiment a very weak binding site was detected. The value of the Co(NH)-ion as a structural mimic of hexahydrated Mg-ions has been questioned recently (). Our results indicate that it is a useful probe as long as ion binding occurs through outer shell coordination but also that exclusive reliance on Co(NH)-ions as a probe might lead to an overestimation of the amount of divalent cation binding.
Whereas Mg binding does not influence the structure of the RNA–ligand complex it has a significant influence on the conformation of the free state of the adenine-sensing riboswitch. In contrast to our findings for the aptamer domain of the closely related guanine-sensing riboswitch in the absence of Mg the free aptamer domain of the adenine-sensing riboswitch studied here does not show a pre-organized stable tertiary interaction between the loops 2 and 3. Instead, an alternative base-pairing pattern is observed that most likely involves loop nucleotides and competes with the formation of this loop–loop interaction and ultimately ligand binding. Mg promotes the formation of this crucial tertiary interaction and directly induces the formation of base-pairing interactions between loop 2 and 3 in agreement with the FRET results reported by Lafontaine and co-workers and Micura and coworkers (,). Therefore, the small sequence differences in the closing base pairs of helix II and III and loop 2 and 3, respectively, between the guanine and the adenine-sensing riboswitch lead to significant differences in the conformation of the free state of the RNA and a different role for Mg in the folding pathway (). The stabilization of the loop–loop interaction through either Mg binding or more stable base-pairing interactions in helix II as found in the majority of the guanine-sensing riboswitches (,) and the resulting pre-organization of the RNA fold might actually have important consequences for the kinetics of ligand binding () which would most likely be faster in the presence of Mg or the stabilizing base-pairing interactions. In turn, the ‘on’ rate of ligand binding is important for the proper function of a ‘kinetically’ controlled riboswitch such as the riboswitch (). On the other hand, the base pairing is the same in helix II of the riboswitch from which appears to be thermodynamically controlled (). In fact, all the adenine-sensing riboswitches described so far () contain either at least one non-canonical base pair or two A:U base pairs as closing base pairs of loop 2. In contrast, two Watson–Crick base pairs with either one or even both being a G:C base pair are found at these positions in most of the guanine-sensing riboswitches. Therefore, it is tempting to speculate that the Mg concentration might modulate the effectivity of riboswitch-mediated gene regulation in the case of the adenine-sensing riboswitches.
The detrimental effect for ligand binding of an alternative long-range base-pairing interaction in the core region of the adenine-sensing riboswitch (G48A) has been shown before () and the sequence conservation patterns the different variants suggest that they have evolved in a way to minimize this possibility. Here, we demonstrate that the presence of Mg-ions apparently helps to avoid an alternative base-pairing pattern in the loop regions. Finally, we demonstrate that small sequence variations can have a profound effect on the folding pathway of this RNA since a slight variation of an apical base pair in helix II is as effective as Mg in stabilizing the loop–loop interaction. Similar effects on RNA-folding pathways have been reported for instance for point mutations in the P5abc domain of the group I intron (,).
The large influence of small sequence variations on the structure of the free state of the purine-binding riboswitches and their folding pathways might have interesting implications for the design of drugs that are targeted against purine-sensing riboswitches (). The high similarity of the structures of their aptamer domain ligand complexes will render it difficult to find drugs that bind selectively to only a subset of these riboswitches such as those of a given bacterium. However, targeting the free form of these riboswitches where small sequence differences result in different conformations will allow selectivity. For some of these riboswitches it might be promising to develop inhibitors for the formation of the loop–loop interaction which is essential for ligand binding () while those with a stable preformed loop–loop interaction will not be affected in their function. |
Retroviruses occasionally integrate into the germ line and may then be transmitted vertically to new generations as ‘endogenous’ retroviral sequences (ERVs) (). A substantial part of extant eukaryotic genomes consists of ERVs (). ERVs are one of many kinds of transposable genetic elements (). Transposons far outnumber conventional genes in higher eukaryotic genomes (). ERVs are often severely mutated, which makes them difficult to recognize. Detection, classification and pathophysiological studies of ERVs are accelerating.
ERV detection has mostly been conducted by BLAST algorithms using the non-redundant (nr) sequence database at NCBI (), or using the BLAT search at the UCSC genome browser interface (). This requires a preconceived notion of the query sequence and, although computer-aided, is largely a time-consuming manual process. A further difficulty is that current ERV classification is nonsystematic, which complicates evaluation of recognition techniques. The primary classification principle for human ERVs (HERVs) has been tRNA complementary sequences in the primer binding site (PBS) (). The RepBase nomenclature () is based on nucleotide identity to machine-generated consensus sequences () of repetitive elements. Although efficient and pervasive, this approach does not in itself identify the repetitive element as retroviral. This is completed by manual inspection, a slow and sometimes error-prone process. RepeatMasker (,), is a system for genome wide screening for repetitive sequences, based on RepBase. It gradually developed from simple detection to a degree of characterization of the repeats. The characterization is however still limited. HERVd (,), in its turn, is a derivative of RepeatMasker. These sequence collections are the main references. They are further described below.
A number of algorithms have been developed for sequence searching, see e.g. (,). In general, they are not suitable for the task of large-scale identification of ERVs in genomic material. This is because the conserved features of ERVs are short and sparse, while the intervening sequences are highly variable, even before degradation by mutations sets in. Published methods for retrieval of retroviral sequences from genomic databases either center on detection of long terminal repeat (LTR) pairs, specific conserved sequences, or general repeat detection. To our knowledge, there is however no comprehensive attempt to both detect ERVs and to characterize their internal structure.
More and more vertebrate genomes have been sequenced. A generic tool for detection of a broad range of retroviral sequences, which is not limited to primate genomes, is needed. We have therefore developed a procedure, which concentrates on the conserved features (motifs). The intervening regions come in only as rough measures of distances between motifs, though they are the subject of follow-up analysis in regions focussed by the primary search. The search procedure for each individual motif may be chosen according to its characteristics, though so far straightforward codon-by-codon comparison with a consensus amino acid sequence dominates. Modules for search, analysis and result presentation have been united as a package, RetroTector, ReTe, with large scope for modification to meet particular needs, even possibly for other tasks than ERV searching.
ReTe is an expert system, which strives to embody and generalize present knowledge of retroviral genomic structures. It uses a combination of several novel heuristic algorithms. The primary algorithm is based on the principle of ‘fragment threading’. It first detects candidates for the LTRs, then different conserved retroviral motifs. Having reduced the search space, more time-consuming and exhaustive algorithms come into play. The LTRs and motifs are then connected into chains, indicating more or less complete ERVs. Finally, it attempts to reconstruct the four major retroviral proteins Gag, Pro, Pol and Env. The findings are collected into a database for convenient retrieval. Data are presented in an interactive graphical format, akin to the format used in textbooks ().
Reference retroviral sequences were collected from GenBank. Whole genomic sequences (human genome versions hg15, hg16, hg17 and hg18, chimpanzee genome versions panTro1 and panTro2, wild red jungle fowl genome versions galGal1 and galGal3, dog genome canFam2, Rhesus macaque rheMac2, and Mouse genome mm8) were downloaded via the UCSC Genome Browser (). Results from the analyses of the various assemblies will be published separately. Most of the results in this methodological paper are based on early versions (hg15 and hg16) of the human genome assembly. However, more recent data, from hg18, panTro2, rheMac2, canFam2, musMus8 and galGal3, and the corresponding RepeatMasker output files, are also included. Reference retroviral sequences RSV (J20342 and NC_001407), ALV (NC_001408), MMTV (NC_001503), MPMV (NC_001550), JSRV (NC_001494), FLV (NC_001940), MLV (J02255 and NC_001501), HTLV1 (NC_001436), HTLV2 (M10060 and NC_001488), WDSV (NC_001867), Snakehead retrovirus (NC_001724), Xen1 (AJ506107), HIV (K03455 and NC_001802) and HFV (NC_001736), were analyzed with ReTe. The errantiviruses ZAM (AJ00387) and CER1 (U15406) were also analyzed.
The genomic analyses were performed on (i) seven Dell Optiplex 260 office computers with 2–2.5 GHz Pentium processors, and 40 GB hard disks, and (ii) at the Uppmax () computer cluster of AMD Opteron 250, 850 and 875 CPUs running Scientific Linux 4.2. The latter configuration yielded 2–5 times shorter execution times.
We coined this term to describe the central procedure in ReTe. It depends on a database of conserved motifs, constraints on the distances between motif ‘hits’ and a matrix of similarities between amino acids. From a programmer's point of view, ‘motifs’ are procedures for detection of conserved ERV traits in the face of mutations. The bulk of the motifs operate through simple comparison (using the acid similarity matrix) against a conserved amino acid sequence, but there are several other types (, which also includes motifs for other purposes). Each motif is connected to one or more retrovirus genera (at present alpha-, beta-, gamma-, delta-, epsilon-, spuma- and lentiretroviruses, and the related viruses Gypsy and Copia). The constraints on the distances between motif ‘hits’ are based on the position distances in known retroviruses, extended with a ‘safety margin’. At present the motifs and constraints are adapted primarily to vertebrate, especially primate, sequences, but are also flexible to change in order to accommodate other sequence analyses.
The principle of ‘fragment threading’ is illustrated in . The most likely of the often many possible combinations of motif hits is chosen according to a heuristic procedure. Motif hits are combined into ‘chains’ satisfying distance constraints, corresponding to potential ERVs, though ‘broken’ chains violating one or two constraints are also possible, to account for ERVs containing insertions or deletions (indels). To evaluate the chain, it is assigned a score and a retroviral genus (or more than one in ambiguous cases) through a vector procedure: Each motif hit is assigned a vector. Its direction is dependent on its genus and its length depends on a weight factor for the motif (see Supplementary Data S3), and how well the hit fits the motif. The motif-hit vectors are summed (with some modifications) into a vector for the whole chain. The length of this vector determines the chain score and its direction determines the retroviral genus assigned.
‘Fragment threading’ is simple in principle, but in order not to miss mutilated or previously unknown ERVs, the motif hit and distance constraints must be so lax that an exhaustive search of all possible combinations is not practical. This ‘combinatorial explosion’ has been countered in several ways: (i) The search is hierarchical. The motifs are grouped into 14 ‘subgenes’ (5′LTR, PBS, MA, CA, NC, DU, PR, RT (incl. RNH), DL, IN, SU, TM, PPT and 3′LTR) according to established retrovirus terminology (DL is here used to denote a dUTPase sequence integrated in the integrase region). Exhaustive ‘fragment threading’ of motif hits is applied within each subgene (except the first two and last two, which contain only one motif each) to generate subgene hits. The subgene hits are then threaded to form chains, with a limit on the number of hits tried for each subgene. (ii) Another limit is set on the length of gaps in the subgene sequence. (iii) A subset of the motifs (notably the PBS and PPT subgenes) is normally not used in the primary search, but only in refining already found chains. (iv) Long sequences are split into chunks (typically 115-kb long with 15-kb overlap) before processing. The 15-kb overlap is sufficient to minimize loss of ERVs, normally up to 10-kb long (), in the sequence chunk border region.
#text
list
#text
xref
#text
ReTe is written in Java and should run on any computer with Java runtime 1.4.3 or later. For full functionality an SQL database manager, preferably MySQL, should be available. ReTe has been extensively used and tested under the Windows (with Sun Java runtime), MacOS X 10.4 (i.e. UNIX), LINUX (Red Hat 9, Shrike) and Scientific Linux operating systems. In the programming, the shareware source (.eps file module) has been utilized.
ReTe is designed for searching entire genomes, i.e. the algorithms were chosen for speed rather than refinement. Also, information outside the ERV proper, such as integration repeats, also referred to as ‘target site duplications’, is utilized by ReTe. The design is flexible, with numerous variable parameters and facilities for plugin extensions (in the form of Java classes). Further information and documentation about ReTe is available at the URL: .
The variable parameters allow the user to adjust the unavoidable tradeoff between speed, sensitivity and selectivity. The standard settings (see the URL above) provide for the processing of a genome in about one month processor time (2 GHz Pentium) with sensitivity prioritized over selectivity. The processing time can be drastically reduced by running the program on a cluster of faster processors. The selectivity may be increased in retrospect by disregarding low-scoring results.
ReTe contains modules for various operations that can be executed from a menu. However, for many of them execution is normally initiated by a script file, generated by another module, containing necessary information. The script files thus serve to connect the different modules. One of the modules (SweepScripts) handles automatic execution of scripts, so that an entire chromosome can be processed automatically.
bold
italic
fig
#text
‘LTRID’ is a module that identifies potential LTRs (). LTRID first aims to find the polyadenylation signal, always present in an LTR, either as the characteristic sequence (AATAAA, ATTAAA or AGTAAA) or as a high score by a neural network trained for this purpose. It detects the R-U5 portion of LTRs of many Alpha-, Beta- and Gammaretroviruslike sequences. After detection of the polyadenylation signal, LTRID calculates a score for the LTR candidate by searching for other LTR characteristics (GT accumulation, a further neural network, TATA box, characteristic nucleotide sequences, binding sites of selected transcription factors, CpG-rich regions) within realistic distances from the polyadenylation signal. All of these LTR-specific features were found using the database of annotated reference retroviral genomes (Blomberg, J., unpublished data).
If the LTRID score exceeds a specified threshold, the single LTR candidate is accepted and included in the script for RetroVID (see below). So also are pairs of similar (by algorithm A2) LTR candidates separated by a realistic distance, irrespective of LTR scores. In both cases, start and end points of the LTRs are suggested based on similarities to the characteristic direct and short inverted repeats formed during the retroviral integration and flanking the provirus ().
LTRID recognizes LTR pairs adequately, though with many false positives. Pairs of ALUs, LINEs or other transposons not found and masked by SweepDNA may be reported as LTR pairs. Identification of solitary LTRs is at present not satisfactory. However, inclusion of HMMs, see e.g. (), will probably improve this.
‘RetroVID’ is normally initiated by a script generated by LTRID and containing its findings of LTR candidates. These are included as motif hits in the subsequent procedure. The other motifs are given a score threshold by sampling each of them in 1000 positions along the target sequence, the score threshold for motif hits then being determined by the statistics of these scores, typically at mean +5.5 SD, thus adjusting score thresholds to local genetic noise. A subset of the motifs is then scored in all positions and hits recorded where the score threshold is exceeded. The hits are combined into chains through ‘fragment threading’, and a subset of non-overlapping, high-scoring chains is subsequently selected. The selected chains are further refined, mainly by including the full set of motifs and by making a renewed attempt to include LTRs, by searching for pairs with algorithm A1. The resulting chains are output to one file, and if appropriate, scripts for ORFID, EnvTracer and XonID are also generated.
‘ORFID’ is normally started by a script generated by RetroVID, containing information about detected motif hits and ranges for the start and end of the protein. There is one script for each gene in which motif hits were found. Moreover, if the genus of the chain was ambiguous, a set of ORFID scripts for each genus are generated. ORFID then constructs a putative protein, or ‘putein’, passing through most of the motif hits. Very weak or otherwise doubtful motif hits are ignored. The frequent post-integrational mutations in ERVs makes it especially important to have multiple criteria for continuously selecting the most likely reading frame. Essentially, ORFID strives for an optimal pattern of frame shifts using algorithm A4.
All codons between frame shifts are included in the putein, with no attempt to identify indels. A4 is applied between each pair of consecutive motif hits, where it is in general reasonably stable. It is also applied to the end portions, but is then combined with a procedure that evaluates the fitness of each position within the range as starting- (or end-) point for a protein. The details of this procedure are different for each gene and retroviral genus and built on similarity to known protein ends, the relations of known viral proteins to stop codons, Kozak start consensus (), protease cleavage sites, slippery sequences, pseudoknots, splice sites and von Heijne signal sequence (). If puteins are made for adjacent genes, they are also adjusted to each other. This selection of putein ends is not always satisfactory and will be improved. ORFID also identifies the longest ORF coinciding with the putein, as a possible present-day coding sequence.
‘XonID’ identifies other possible exons than the four fundamental retroviral ( and ) genes, since ORFID only constructs rather obvious puteins. XonID may be applied to search for more vague traces of exons, combining criterion (ii) in algorithm A4 with data about canonical splice sites and start/stop codons.
‘EnvTracer’ is based on similar principles as XonID, but is specialized to finding likely reading frames. It is focussed on long open reading frames not recognized by ORFID, rich in predicted N-glycosylation sites, which occur between the predicted end of and a 3′LTR.
Within ReTe, there are about 20 other command modules, some for visualization and storage of results, others mainly for maintenance and debugging. Of particular interest are the modules:
‘PseuGID’ is applied if a chain has none or very short LTRs, suggesting that the ERV may be a processed pseudogene (). RetroVID generates a script for this module, which searches for the structures characteristic of processed pseudogenes. This function has not yet been fully tested.
‘Chainview’ displays one chain at a time in detailed graphics (; Supplementary Data S2) and in detailed text. Apart from the motif hits and the course of the predicted chain, it may also display an analysis of LTR structure, puteins and EnvTracer and XonID output related to the chain, start and stop codons, splice donors/acceptors and several other features.
‘Puteinview’ shows details, such as the full amino acid sequence of a putein or an exon suggested by XonID or EnvTracer.
‘CollectGenome’: The mass of text files containing all the results from a ReTe analysis may be forbidding. This module extracts selected results and collects them into an SQL database, with separate tables for LTR candidates, chains and puteins. Thereby it also compares each chain to a set of RepBase consensus sequences and a set of known annotated retrovirus sequences, using algorithm A1, and Pol puteins to a set of known Pol proteins (using algorithm A3) for classification purposes. Judgment of the content compared to nonretroviral repetitive sequences is conducted as an internal control.
‘Genomeview’ may be used to inspect the database generated by CollectGenome. It shows the distribution of LTR candidates and chains within the chromosomes. Chains and their relation to the RepBase reference sequences and known retroviruses may also be viewed graphically (with less detail than Chainview).
‘RetroTectorShell’: This is a Windows-specific program separate from ReTe, written in Visual FoxPro (VFP). It was developed in parallel with the Java ReTe kernel for user interaction and data handling. It performs similar functions as CollectGenome and Genomeview, but has a somewhat different profile. Features so far found only in RetroTectorShell are: (i) A mechanism for collecting ReTe output into a VFP table. Each genome's retroviral content is thus contained in a single table. It disregards single LTRs and alternative ORFs. (ii) A BLAST-like algorithm for searching chains in this table according to protein or nucleic acid similarity.
A 3 × 10 nt stretch of random nucleotide sequence with equal A, T, G and C frequencies was run. Two chains with a score above the minimum, 258 and 273, resulted. Thus, none were above 300. The experience from many genomic analyses, primarily from the human genome versions hg15–hg18, has shown that a score cutoff of 300 almost totally eliminates spurious chains results from motif-hit combinations occurring by chance (; ROC curves in Supplementary Data S10–S15).
In order to test the detection limits of ReTe with regard to degraded retroviral sequences (i.e. very old insertions), a test set with artificially degraded sequences was created. These sequences were based on the complete genome of HIV-1, isolate MNCG (Genbank accession no M17449), which was degraded according to four different mutational models, with decay ranging from 1 to 60% mutation (see Supplemental Data S1 for further details). The resulting test set was analyzed with ReTe and also with BLAST (see below), for comparison. Four different mutational models were selected: (i) Random substitutions with equal probabilities (Jukes–Cantor); (ii) Higher probability of transition over transversion (Kimura 2-parameter model); (iii) The Kimura 2-parameter model with insertions and deletions, with frequencies of indels applying to human pseudogenes () (‘indel model’); (iv) Simulation of an active retrovirus, where both endogenous and exogenous phases are subjected to purifying selection during each round of infection and replication (‘Exogenous model’). Mutations harming features important for the retroviral function will not persist in the viral population. HIV is a highly replicative retrovirus. The Los Alamos National Laboratory () presents a large database with sequence data for many subtypes, including a set of full-length HIV genomes aligned with respect to nucleotide codon triplets. The ‘exogenous’ model uses this aligned data set to test which mutations are allowed.
The ReTe analysis utilized default settings. BLAST (version 2.2.6, obtained from NCBI, ), run locally under Linux Red Hat 9, with default settings except that word length was 7. The sequences were matched against the reference sequence HIVMNCG, as well as against the entire non-redundant nucleotide database (see Supplementary Data S1 for further details).
In the simulated evolution model based on HIV, ReTe chain detection was more resistant to mutational decay than BLAST sequence detection. ReTe also attempts to reconstruct the retroviral proteins. Sequence similarity in the form of percent identity between the Pol puteins and the original HIVMNCG Pol protein, shows that ReTe can detect even extensively mutated and evolutionarily distant Pol sequences (, and ; Supplementary Data S1).
As seen in and , chain scores in the human genome version hg18 ranged from the cutoff of 300–4400. High scoring (>2000) chains were from exogenous retroviruses, and from structurally intact endogenous proviruses like the betaretroviruslike HERV-K(HML2) (). Incomplete proviruses, with LTR--LTR etc. scored 250–400 points. sequences like scored 250–950. sequences like scored 200–350, or not at all. chains scored 350–1050.
As shown in and , and , ReTe can detect retroviral sequences in a wide variety of genomes, also from less complete assemblies (e.g. panTro1). The four major genes were detected in many of the chains. However, sequences were less common than the other three. Detection of the gene, the least conserved of the four major retroviral genes, poses special difficulties. There are few conserved motifs in its SU portion, and the conserved motifs in the transmembrane protein often do not provide enough basis for a putein reconstruction. The inclusion of the EnvTracer module was intended to diminish false negativity in gene detection. The lower frequency of in the retroviral chains () is probably due to mutational decay, occasional misses by EnvTracer or to -less proviruses which may transpose without leaving the cell, e.g. the betaretroviruslike IAP elements in mice (), and possibly the bulk of the HERVH elements in humans ().
As discussed below, a basic problem for sensitivity and specificity determination for an ERV detection algorithm is the absence of a generally recognized and curated HERV database. An established general mechanism for repeat detection, and a limited level of repeat characterization, is provided by RepeatMasker (,), based on RepBase (). HERVd (,) represented an attempt to reduce the fragmentation of Repeatmasker output, and to amend its retroviral nomenclature. Unfortunately, it could not be maintained. Nevertheless, these sources are the best established references.
ReTe-derived sequences were evaluated versus RepeatMasker output and HERVd annotations on the human genome hg15, see also supplementary Data S6 and S7, as well as (). Using ≥300 as a chain score cutoff, 3373 retroviral chains were detected in hg15. Repeatmasker reported 457 600 ‘LTR’ elements, generally as fragments. A total of 2625 ReTe chains were colocalized with a Repeatmasker entry, using a criterion of overlap within 12 000 nt of the start point of the ReTe chain. The Supplementary Data S2 and S4 shows a chain missed in hg15. The remainder occurred in both data sets. A detailed comparison of these discrepancies requires a detailed ERV classification, which is out of scope for this article. The Supplementary Data (S6–S17) does however give a survey. Repeat-based recognition does not in itself identify retroviral sequences. RepeatMasker relies on the man-made assignments in RepBase to characterize elements as ‘LTR’ elements or not.
On the other hand, ReTe is not dependent upon repetition for detection, and therefore could detect single or low-copy-number retroviral elements. Examples are ERV-FRD on chromosome 6p24.2 (), and HERV-Fc1 () on chromosome Xq21.33, which are single or low-copy-number elements with unusually open reading frames (Supplementary Data S2 and S4). Both elements are however now fully or partially covered in an April 2007 version of the RM output of the human genome version hg18.
Reference retroviral genomes give chains with scores of; 3341 (RSV), 3792 (MMTV), 3439 (MPMV), 3022 (MLV), 2934 (FLV), 2814 (HIV-1), 2345 (HTLV2) and 879 (WDSV). The errantivirus elements ZAM and CER1 give 935 and 687, respectively. They are elements from and , respectively. They are related to the Orthoretroviruses (; and Supplementary Data S2), and have the same genome organization. The four major genes are predicted in all of the 10 viruses, except for the gene in RSV. This may be due to deranged env distances due to the presence of , and/or insufficient coverage of alpharetroviral SU and TM motifs. The start and stop positions of the respective ORFs were correct within 10% of the annotated position, see e.g. (). Larger deviations are occasionally observed in complex retroviruses (lenti, delta and epsilon retroviruses), which have one or several additional regulatory protein genes. They can occur before and around . Similar problems are caused by the sarcoma viruses, where oncogenes disrupt the retroviral structure (Supplementary Data, S2 and S5).
As mentioned in the discussion, ReTe has already been used in several studies on human ERVs (,,).
ReTe was applied to the the human genome versions hg15, hg16, hg17 and hg18, the chimpanzee genome versions panTro1 and panTro2, the chicken genome versions galGal1 and galGal3, the dog genome version canFam2, the mouse genome version mm8 and the opossum genome monDom4. Depending on settings, a full analysis of the human and chimpanzee genomes takes 5–6 days, the chicken genome 3 days, using the computer set described in Systems and Methods. On the Uppmax Opteron Linux-based cluster, 1–2 days suffices.
Solitary LTR detection is one of the most demanding aspects of retroviral sequence recognition. The principle of LTR selective split octamers used in ReTe is similar to the one used in the program Matinspector® (Genomatix Gmbh, Germany) (). Despite a rather high LTR selectivity of the presented LTR recognition algorithm (LTRID), the number of false-positive hits is overwhelming when entire genomes are processed. Initially, we considered using hidden Markov models (HMMs) () for the detection of retroviral structures. This is computationally intensive, and we instead chose the faster algorithm ‘fragment threading’. We are currently attempting a limited introduction of HMMs for improved detection of solitary LTRs, and a few other motifs. The principle of ORF-selective hexamers used in ORFID is also used in gene-finding algorithms like GenScan (). Clearly, certain binary triplet combinations are more likely to occur in retroviral ORFs compared to the two alternative reading frames. The functional basis for this selectivity is obscure.
We had the practising retrovirologist and geneticist in mind in the design. Although the modular design of ReTe leaves ample scope for future improvement, it has already proved useful in its present form (,,). ReTe gives a rich basis for assessment of the functionality and taxonomy of a retroviral element. The ability to export protein and nucleic acid sequences of the four major retroviral genes ( and ) from ERVs of an entire genome in FASTA format will aid phylogenetic studies, and promote the understanding of the ‘retroproteomes’ of these organisms. The usefulness of the ready availability of nucleic acid frequency, LTR divergence, Pol-based classification versus reference retroviral elements, and degree of nucleotide identity to RepBase elements for all detected ERVs in a genome has been demonstrated in studies on HERV-H (,), ERV3 (), ERV9/HERV-W () and a comparison of ERVs unique to humans and chimpanzees (). In addition, features such as splice prediction, prediction of additional ORFs besides Gag, Pro, Pol and Env, readthrough mechanisms and LTR structure aid further functional studies on ERVs.
Repeat-based recognition of ERVs, using RepBase () and its corollaries RepeatMasker (,), Hubley,R. and Green,P., unpublished data () and HERVd (,), has been conducted for over a decade. An elaborate classification (RepBase) based on nucleic acid identity to machine-generated consensus sequences, through the Censor program (), exists for many repeated elements. It is gradually being supplemented by user contributions. This classification and detection procedure should be scrutinized against other alternatives. ReTe provides an independent route to ERV detection and classification. Only by using several approaches can a rational classification of ERVs be achieved. ReTe favors elements >1000-bp long due to its dependence on the presence of several retroviral fragments in the right order at approximate distances typical of retroviruses. RepeatMasker, however, can detect considerably shorter sequences but is limited by the need for a minimum number of repeats for recognition. It is also limited by a lack of internal retroviral structure interpretation. An exact appraisal is not possible due to the often fragmented nature of both HERVd and Repeatmasker outputs. Published methods for retrieval of retroviral sequences either center around detection of LTR pairs (,), specific conserved sequences, like TM (,), or RT, combined with an ORF search (,), or general repeat detection, collected in RepBase (,) and used in RepeatMasker (,) and HERVd (,). HESAS (HERVs Expression and Structure Analysis System) () merges dbEST information with Repeatmasker-based output. It yields information about the expression and structure of HERVs. However, none of them has the broad scope of ReTe.
Speed of analysis is essential as the sequencing and assembly of genomes becomes faster. Sequence lengths from 10 to 10 nt can realistically be analyzed. As demonstrated, ReTe faithfully reconstructed many features both of the simple retrovirus MoMLV, and the complex retrovirus HIV (; Supplementary Data S2 and S5). Many of the ReTe functions may be further optimized, e.g. splice site prediction based on canonical consensus sequences and LTR detection.
The limitation to the four different nucleotides in DNA imposes restrictions on the possibility for sequence recognition of mutated sequences. This is observed in multiple nucleic acid alignments, where identities <50% must be regarded with caution. The typical mutation frequency (here discussed as substitutions) for sequences without selection pressure is ∼0.2% per million years (). Thus, 1% substitution corresponds to ∼5 Mya, and 50% corresponds to ∼250 Mya. Consequently, 200–300 Mya is a detection limit for selection neutral retroviral sequences, which probably are the majority of the ERVs. Selection for a functional protein, leading to persistence of conserved retroviral amino acid motifs, can push back the limits for recognition considerably. This is demonstrated by the ability of ReTe to recognize widely divergent retrovirus-related retrotransposons like Errantiviruses of invertebrates (). Thus, using a collection of motifs (Supplementary Data S3) largely (but not exclusively) derived from retroviral sequences of higher vertebrates, ReTe can detect retroviruslike sequences in amphibians, insects and worms. In this situation, the model-based approach of ReTe surpasses the unbiassed recognition via the BLAST algorithm ( and ). This attests to the structural antiquity of retroviruses. However, the demonstrated ability to detect highly mutated ERVs requires a relatively intact structural backbone. Secondary integrations of a few large (e.g. LINEs) or many smaller (e.g. SINEs) elements into an ERV can derange structure beyond repair by the ‘broken chain’ function of ReTe, and masking of nonretroviral repeats. This problem is inherent to the systematic structural approach of ReTe. On the other hand, ReTe often provides an interpreted proviral structure, which can be used in further studies.
The structural model of ReTe thus allows recognition of many retroviral sequences. There are both minor and major obstacles to widening the scope of detection. Adjustments of the distance constraints and inclusion of more motifs, are simple measures which may lead to an enhanced recognition of retroviral sequences like HERV-L. However, the gene of has the gene order IN..RT instead of the usual RT..IN, which would require the use of alternative models for these elements. The MalR retrotransposons () are incomplete and very divergent from orthoretroviral gene structure, with very few recognizable conserved motifs. The latter two are major challenges for ReTe.
ReTe analysis of five vertebrate genomes () demonstrated that vertebrate lineages have a variable number and type of ERVs. The mouse had a high number of high-scoring chains, and the dog and chicken genome had a low number. The human, chimpanzee and rhesus genomes were intermediate. Especially complete proviruses were betaretroviruslike in mouse and gammaretroviruslike in humans (). The findings extend and confirm previous observations (,). Several of these species differences must have arisen relatively late in evolution, probably due to more or less successful modifications of antiretroviral restrictions or changes in habitat and habits (). Further work with the extensive retroviral sequence data set provided by ReTe will undoubtedly shed light both on retroviral and vertebrate evolution.
#text
email
ext-link
#text |
Cockayne Syndrome (CS) is a hereditary human disease characterized by profound deficiency of post-natal growth with heterogeneity of clinical symptoms and neural degeneration resembling premature aging. CS is associated with defective transcription-coupled nucleotide excision repair (). At the cellular/molecular level, the phenotype of CS includes sensitivity to ultraviolet (UV)-irradiation and delayed recovery of RNA and DNA synthesis after exposure to UV () There are two CS complementation groups, A and B (), corresponding to genetic defects in the and genes, respectively. The gene encodes a 168-kDa protein which is a member of the SWI/SNF family of DNA-dependent ATPases (). Previous studies show that primary cells from CS patients have a defect in initiating (i.e. incision) repair of oxidative DNA lesions (). In addition, cells that are deficient in CSB-dependent ATPase (motif VI mutants) are sensitive to γ-irradiation and defective in general genome repair of oxidative DNA damage (,). These data suggest that CSB may play a role in repair of oxidative DNA damage via the base excision repair (BER) pathway.
c-Abl is a non-receptor protein tyrosine kinase whose activities are tightly regulated in the cell. For example, both nuclear and cytoplasmic c-Abl is activated in response to genotoxic or oxidative stress (), DNA damage (), or by a protein kinase Cδ (PKCδ)-dependent mechanism (). Activated c-Abl plays a role in inducing apoptosis and/or necrosis (,) after exposure to oxidative and other types of cellular stress. c-Abl is phosphorylated by DNA-PK (,) and interacts with DNA repair associated proteins () including p53, p73 and Rad9 (). Oxidative stress activates c-Abl (,).
Structure-function analysis of c-Abl indicates that it has a protein–protein interaction SH3 domain, which is likely to participate in interactions between c-Abl and its kinase substrates. SH3 domains bind preferentially to proline-rich motifs such as P-X-X-P (,) (B). CSB is a phosphoprotein with an N-terminal proline-rich region (E). No prior reports have indicated that CSB undergoes tyrosine phosphorylation.
Tyrosine phosphorylation, particularly by c-Abl tyrosine kinase, has been reported to play a role in the regulation of some DNA repair proteins. c-Abl-mediated phosphorylation of DNA-topoisomerase I (topo I) at Tyr268 and in cells conferred activation of the topo I function. Moreover, activation of c-Abl by treatment of cells with ionizing radiation was associated with c-Abl-dependent phosphorylation of topo I and induction of topo I activity (). Rad51 is a key element of recombinational DNA repair and its activity is regulated by phosphorylation of the tyrosine residue at position 315 by c-Abl tyrosine kinase (). This study explores the interactions between c-Abl and CSB and the role of this interaction in the response to oxidative DNA damage. Evidence is presented that CSB binds to c-Abl and is a substrate for c-Abl tyrosine kinase. We speculate that tyrosine phosphorylation of CSB plays an important role in regulating and/or coordinating the cellular response to oxidative stress.
CS1AN.S3.G2 (CS1AN) cells, a SV40- transformed human CS-B fibroblast, either with empty vector or complemented with wild-type CSB, E646Q, Q942E point mutants, ECFP-CSB were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 50 U of penicillin/ml, 50 µg of streptomycin/ml and 50 µg of genetectin/ml or neomycin (ECFP-CSB). K562 lymphoblasts were cultured in RPMI medium 1640 containing 10% fetal bovine serum and antibiotics. CS1AN cells, mouse wild-type and c-Abl-/- embryonic fibroblats were cultured in Dulbecco's modified Eagle medium containing 10% fetal bovine serum, 50 U of penicillin/ml and 50 µg of streptomycin/ml. GM00969, AG08470, AG01428 and AG05012 primary fibroblasts were grown to confluence in minimum essential medium containing 15% FBS, 2X MEM non-essential amino acids, 2X MEM vitamins and 1X MEM essential amino acids. The human cells were from the American Type Culture Collection (Manassas, VA) or Coriell Cell Repositories (Camden, NJ). In some experiments, CS1AN and CSB-WT complemented cells were incubated with hydrogen peroxide (Sigma, St. Louis, MO) in serum-free media at a concentration of 250 µM for 1 h, either in the presence or absence of STI-571 (5 µM).
The purified HA/His-tagged CSB protein, GST–c-Abl–Src homology domain 2 (SH2) and SH3 fragments and the purified histidine-tagged c-Abl from a baculovirus insect expression system were prepared as previously described (,). The c-Abl fragment (45 kDa, containing the SH2 and kinase domains, SH2-PTK) was purchased from New England BioLabs (Beverly, MA) and a human active almost full-length c-Abl was purchased from Upstate technologies (Beverly, MA).
Procedures for preparing cell lysates and conducting immunoprecipitation and immunoblot analysis were described previously (). The pre-cleared lysates were immunoprecipitated with rabbit polyclonal anti-CSB (H-300; Santa Cruz biotechnology, Santa Cruz, CA) or mouse monoclonal anti-Abl (8E9; BD Pharmingen) antibodies. The protein G Sepharose-precipitated proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS–PAGE), transferred to polyvinylidene difluoride membranes, and analyzed by immunoblotting with anti-CSB, anti-Abl (8E9; BD Pharmingen), or anti-phosphotyrosine (Upstate Technology, Beverly, MA) monoclonal antibodies followed by the chemiluminescent method for detection with SuperSignal substrates (Pierce, Rockford, IL). Membranes were stripped by using the Restore western blot stripping buffer (Pierce, Rockford, IL) when necessary. Anti-CSB immunoprecipitates were used either to study the interactions or to analyze tyrosine phosphorylation. Anti-Abl immunoprecipitates were used either to study the interactions or to analyze tyrosine phosphorylation or in kinase assays using GST–Crk (aa 120–225) as c-Abl substrate.
The CSB and GST–c-Abl fragments and GST was incubated and the adsorbents were analyzed by SDS–PAGE. The membrane was analyzed by immunoblotting with the monoclonal anti-HA or anti-c-Abl antibody. Also, GST–c-Abl and GST was incubated with HeLa whole cell extracts (1 mg/ml) and the adsorbents were analyzed by SDS–PAGE. The membrane was analyzed by immunoblotting with the polyclonal anti-CSB or stained with amido black.
CS1AN cells were transfected with cDNA (6 µg) expressing various S-tagged fragments of CSB protein (C) for 30 h with the PolyFect reagent (Qiagen, Valencia, CA). Whole cell extracts were prepared as described above. Equal amount of proteins were subjected to S-protein pull down using S-protein conjugated agarose beads (Novagen). The pull-down products were centrifuged at 500 × for 2 min to collect the beads and washed five times with ice-cold lysis buffer. The final bead pellets were either used to incubate with full-length human c-Abl for binding studies or for kinase assay.
Cells were fractionated to remove the cytoskeleton according to Mirzoeva and Petrini (), and then were processed for indirect immunofluorescence and analyzed as described in Partridge . (). For quantification, images were analyzed with co-localization software (Zeiss) using the exact same set of criteria for all cells.
For kinase assays, the purified CSB was incubated with the c-Abl fragment, mouse full-length c-Abl or human full-length GST-c-Abl or GST-c-Abl (K-R) in a kinase buffer (50 mM Tris–HCl, 10 mM MgCl, 1 mM EGTA, 2 mM dithiothreitol) for 15 min at 28°C. The c-Abl kinase inhibitor STI-571 (Novartis Pharma AG) () was used as a control. Phosphorylated proteins were separated by SDS–PAGE and analyzed by immunoblotting.
For phosphorylation, CSB-WT or CS1AN cells (2.4 × 10 cells/10 cm culture dish) were transfected with 6 µg of wild-type c-Abl or the dominant-negative c-Abl (K-R) mutant vectors () for 30 h with the PolyFect reagent (Qiagen, Valencia, CA). The cells were lysed as described above. Equal amount of protein from the whole cell extracts was used to immunoprecipitates CSB followed by phosphotyrosine analysis. Whole cell extracts were also immunoblotted to detect c-Abl, and CSB.
MALDI-MS analysis was used to identify the tyrosine-phosphorylation site of CSB. In brief, kinase assay was done as described above. The phosphorylated proteins were separated on 4–20% Tris–HCl gel by SDS–PAGE. The gel was then stained with Commassie blue stain. The bands were then excised and subjected to in-gel tryptic digestion (Pierce, Rickford, IL) according to the manufacturer's instructions.
The digests were then subjected to phosphopeptide isolation using a phosphopeptide isolation kit (Pierce, Rockford, IL). Elutes were subjected to MALDI-MS analysis. Mass spectra were collected on a Voyager MALDI-TOF mass spectrometer using delayed extraction parameter. The eluted phosphopeptide mixtures were analyzed in reflector mode. 70–200 shots from a nitrogen laser (337 nm) were averaged to yield each recorded mass spectrum. Spectra were externally calibrated with angiotensin I (MH = 1296.6853) and adrenocorticotropic hormone fragment (7–38; MH = 3657.93). The matrix solution was prepared immediately before use by mixing 20 mg of α-cyano-4-hydroxycinnamic acid and 4 mg of 2,5-dihydroxybenzoic acid in 500 µl of 0.1% trifluoroacetic acid in 50% CHCN/dHO, vortexing for 1 min, centrifuging for 1 min, and subsequently diluting the clear supernatant 1:4 with 0.1% TFA in 50% CHCN/dHO. Five microliters of the diluted matrix solution was mixed with the eluted tyrosine-phosphorylated peptides, and a 1 µl suspension was spotted on the target and allowed to air-dry completely before mass analysis.
For interpretation of the mass spectra, a list of predicted molecular weights was generated by theoretical cleavage with trypsin as a specific endoproteinases using the MS-Digest program available at the University of California, San Francisco (). The masses of the peaks recorded in the mass spectra were matched to the calculated masses within ∼0.1% or better.
Physical interaction between purified CSB and c-Abl was examined using GST-pull-down or co-immunoprecipitation assays (A, 1B and 1C). GST-pull-down products revealed that CSB and c-Abl interact with each other (A). A GST-pull-down assay using GST-c-Abl and HeLa whole cell extracts (1 mg/ml) also revealed a potential interaction between c-Abl and endogenous CSB (B). assays were also performed with extracts of CSB-deficient CS1AN cells or CS1AN cells complemented with wild-type CSB. Immunoprecipitation was carried out with anti-c-Abl antibody and the immunoprecipitates were blotted with anti-CSB. The results suggest that CSB interacts with c-Abl in CS1AN cells expressing wild-type CSB (C).
The region of c-Abl that interacts with CSB was identified by incubating CSB with GST fusion proteins carrying the SH2 or the SH3 signaling domain of c-Abl. The results show that CSB binds to GST-c-Abl-SH3 (A). Although we were not able to see any interaction of SH2 domain of c-Abl with CSB, there is a possibility of transient interaction of the SH2 domain with CSB. This interaction mapping was analyzed in more detail using S-protein tagged fragments of CSB. The results demonstrate that the N-terminal region of CSB, which includes proline-rich residues 301–304, interacts with the SH3 domain of c-Abl (C, D and E). S-protein tagged CSB fragments (C) were used in an alternate pull-down assay, which also indicated that the N-terminal region of CSB interacts with the SH3 domain of c-Abl (C, D and E).
To determine whether CSB is a substrate for c-Abl, an kinase assay was performed by incubating CSB with a truncated form of c-Abl including its SH2 and protein tyrosine kinase domains (SH2-PTK). The results show that phosphotyrosine (p-Tyr) accumulates in CSB in the presence of active c-Abl kinase (SH2-PTK) but not in the presence of heat-inactivated c-Abl (SH2-PTK), a full-length kinase inactive c-Abl mutant (GST-c-Abl K-R), or after pre-incubation of either SH2-PTK fragment or full-length c-Abl with the kinase inhibitor STI-571 (A and C). (Note that the amount of CSB protein was similar after phosphotyrosinylation; A.) The kinase activities of full-length human c-Abl (B) and human GST-c-Abl also phosphorylate CSB (C). However, GST-c-Abl (K-R) failed to tyrosine phosphorylate CSB (C). The location(s) of phosphotyrosine residues in SH2-PTK phosphorylated CSB were identified using an kinase assay with immunocomplexed S-protein fragments of CSB. The results indicate that phosphotyrosine is located only in CSB fragment aa 465–1056 (D). The conserved ATPase and helicase motifs of CSB are also located in this region of CSB.
The ability of c-Abl to phosphorylate CSB was examined in CSB-deficient CS1AN cells expressing ectopic wild-type CSB. These cells were transfected with a plasmid expressing c-Abl or kinase-inactive c-Abl (K-R). CSB immunoprecipitates had significantly more phosphotyrosine in cells expressing c-Abl than in cells expressing c-Abl (K-R) or in cells that were not overexpressing c-Abl (E). A low level of tyrosine phosphorylation was evident in cells expressing c-Abl (K-R), which could be due to endogenous c-Abl phosphorylation of CSB (E). These results indicate that c-Abl tyrosine-phosphorylates CSB and and are consistent with the possibility that CSB is a physiologically important substrate of c-Abl.
CSB tyrosine-phosphorylation was also examined in chronic myeloid leukemia (CML) K562 lymphoblasts, which have a constitutive high level of Abl tyrosine kinase activity. These cells have been shown to express both cytoplasmic and nuclear c-Abl (). CSB immunoprecipitates from whole cell extracts of these cells are significantly tyrosine-phosphorylated and the amount of phosphorylation is reduced if cells are pre-treated with STI-571 (F). Similarly, CSB is tyrosine-phosphorylated in wild-type mouse embryonic fibroblasts (MEFs) but not in Abl-/- MEFs (G). The amount of phosphotyrosine incorporated into CSB increased in MEFs exposed to oxidative stress and this increase was blocked by pre-treatment with STI-571 (G). These data confirm the specificity of c-Abl for CSB and suggest that CSB may be a preferred substrate of c-Abl in cells exposed to oxidative stress.
The tyrosine phosphorylation site of CSB was mapped at the amino acid level by MALDI-MS analysis of tryptic fragments of tyrosine-phosphorylated CSB. This experiment was carried out using the product of an kinase reaction containing c-Abl and CSB. The assay product was resolved by SDS–PAGE and digested by trypsin . Phosphopeptides were eluted from the gel, concentrated using phosphopetide columns and analyzed by MALDI-MS (). The peptides obtained from full-length CSB had a 95–99% match frequency with 59% coverage of full-length CSB. Mass analysis of each peptide indicated that a phosphate group was added to a 2262 kDa peptide VVIDPDWNPSTDTQARER (929–947); this peptide contains Tyr932 (B). Treatment with alkaline phosphatase released the phosphate moiety from the putative phosphotyrosine 932, suggesting correct identification of the peptide structure and correct mapping of tyrosine phosphorylation in CSB (C). Tyr932 is located in motif VI of the putative ATPase/helicase domain of CSB.
Indirect immunofluorescence of CSIAN cells expressing ECFP-CSB indicate that c-Abl (red, ) and CSB (green, ) are present throughout the nucleus in a punctate pattern in non-treated cells (Control). After exposure to hydrogen peroxide (HO), expression of ECFP-CSB and c-Abl increase and they re-distribute and co-localize in the nucleus and nucleolus (yellow in Merge, ). These effects can be inhibited with STI-571 (HO + STI, ). These data raise the possibility that tyrosine phosphorylation of CSB by c-Abl promotes redistribution of CSB in the nucleus and enrichment of CSB in the nucleolus in response to oxidative stress; furthermore, these events may require c-Abl tyrosine kinase activity. Interestingly, when similar experiments were performed in CSB-deficient CS1AN cells (Bottom Panel, ), c-Abl redistribution to the nucleolus was not observed in response to oxidative stress. This result suggests that c-Abl and CSB might play complementary roles in a stress-response pathway that is activated by oxidative damage. presents the quantitation of the data observed in .
As stated above, we postulate that the interaction between CSB and c-Abl and the phosphorylation of CSB by c-Abl may play a role in the response to oxidative stress . This idea was tested by examining the efficiency of oxidative stress-induced c-Abl phosphorylation and autophosphorylation in the presence of wild-type and mutant forms of CSB. CS1AN cells expressing wild-type CSB, motif II E646Q CSB (ATPase-deficient) or motif VI Q942E CSB (BER-deficient) or no CSB were exposed to HO and CSB phosphorylation and c-Abl autophosphorylation were examined. Autophosphorylation of c-Abl was significantly induced in cells expressing wild-type (WT) CSB and motif VI mutant CSB but not in cells expressing the CSB motif II mutant (A). Tyrosine-phosphorylation increased significantly for WT CSB, but less for the motif VI mutant and not at all for the motif II mutant of CSB. Pre-treatment with STI-571 (5 µM) completely blocked c-Abl auto-phosphorylation and c-Abl mediated tyrosine phosphorylation of CSB (A). c-Abl and CSB protein levels were similar in all cell extracts (B). c-Abl auto-phosphorylation and tyrosine-phosphorylation of CSB was observed in normal human primary fibroblasts exposed to HO, but not in primary fibroblasts from CS patients (C). Oxidative damage did not result in Abl auto-phosphorylation or tyrosine-phosphorylation of CSB in AG05012 or AG01428 primary fibroblasts obtained from CS patients. However, a significant auto-phosphorylation of c-Abl followed by a resultant tyrosine phosphorylation of CSB was observed in GM00969 and AG08470 fibroblasts derived from matched normal human subjects, after oxidative damage (C). Pre-treatment with STI-571 inhibited auto-phosphorylation of c-Abl and tyrosine-phosphorylation of CSB (C). These data suggest that CSB stimulates auto-phosphorylation of c-Abl and its activation in response to oxidative stress; this in turn results in tyrosine phosphorylation of CSB itself.
It is possible that auto-phosphorylation of c-Abl in response to oxidative damage influences c-Abl activity in a CSB-dependent manner. This idea was tested by measuring oxidative stress-induced c-Abl kinase activity towards a 15 amino acid fragment of Crk (aa 120–225) conjugated to GST in extracts of CS1AN cells with an ectopic plasmid expressing wild-type CSB. The results show that autophosphorylation of c-Abl increased in response to oxidative stress in extracts of CS1AN cells expressing wild type CSB but not in extracts of control cells carrying empty vector. Furthermore, activation/autophosphorylation of c-Abl stimulated tyrosine-phosphorylation of GST–Crk (aa 120–225) (D). Pre-treatment with STI-571 inhibited c-Abl autophosphorylation and c-Abl-mediated tyrosine-phosphorylation of GST–Crk (aa 120–225) (D). In order to decrease the possibility of phosphorylation by c-Abl immunoprecipitates from resting cells, low amounts of protein was used in these studies for immunoprecipitating c-Abl for kinase activity.
This study provides the first evidence of a physical interaction between CSB, a protein involved in the transcription-coupled DNA nucleotide excision DNA repair pathway, and the non-receptor tyrosine kinase c-Abl. CSB interacts with and is tyrosine phosphorylated by c-Abl and . Serine and threonine phosphorylation of CSB were reported previously (), but this is the first report of tyrosine phosphorylation of CSB. Tyrosine-phosphorylated CSB is constitutive in CML K562 lymphoblasts. Also, wild-type MEFs exposed to hydrogen peroxide have increased CSB tyrosine-phosphorylation while it is lacking in Abl-/- MEFs. Tyrosine 932, which lies in motif VI of the ATPase domain of CSB, is the site of CSB phosphorylation by c-Abl. Oxidative stress stimulates c-Abl-mediated phosphorylation of CSB and this event is blocked by STI-571, a specific inhibitor of c-Abl kinase. Oxidative stress-induced c-Abl auto-phosphorylation and tyrosine phosphorylation of CSB appear to require CSB ATPase, because it was not observed in cells expressing an ATPase-deficient mutant of CSB. These data suggest that CSB and c-Abl may participate in and possibly regulate a common oxidative stress-response pathway.
c-Abl is a tightly regulated non-receptor tyrosine kinase that contains a Src homology domain 3 (SH3). The SH3 domain is a protein–protein interaction region which usually binds to proline-rich motifs (P-X-X-P) (,). This study shows that the interaction between c-Abl and CSB is mediated by the c-Abl SH3 domain and the N-terminal 355 amino acids of CSB (aa 1–355). This region of CSB includes a proline-rich motif at aa 301–304 (PVTP).
CSB cells and cell extracts are deficient in incision at 8-oxoguanine and 8-oxoadenine lesions (,) suggesting that CSB may play a role in the repair of these lesions. These lesions are believed to be removed by BER, and there is additional evidence for a role of CSB in the general genome BER pathway (). Additional evidence comes from studies in mouse models () where increased accumulation of 8-oxoG is observed when mice are deficient in OGG1 in addition to lacking CSB. Also CSB is in protein complex with OGG1 () although 8-oxo-G appears to be removed by general genome BER, and there is no indication of transcription coupled repair of this lesion (). This deficiency in 8-oxoguanine incision activity is complemented by ectopic expression of WT CSB; the deficiency is partially complemented by a motif II CSB mutant (ATPase deficient, E646Q) but not complemented by motif VI (Q942E) CSB mutants. (,,). This result suggests that domain VI of CSB is required for repair of oxidative DNA lesions such as 8-oxoguanine. It is interesting to note that Tyr932, the site of c-Abl mediated tyrosine phosphorylation of CSB, also resides in domain VI of CSB protein.
In cells exposed to oxidative stress, c-Abl auto-phosphorylation and its phosphorylation of CSB increases. However, these effects are reduced and a lower extent of phosphorylation occurs in cells expressing the motif VI (Q942E) CSB mutant. CSB motif II is also required for these phosphorylation events, because they do not increase in response to oxidative damage in cells expressing a motif II CSB mutant or in cells lacking CSB. CSB is also required for oxidative stress-induced phosphorylation of Crk, a previously characterized substrate of c-Abl. This finding of the requirement of CSB in the activation of c-Abl during oxidative damage might have a greater role in the understanding of mechanisms of leukemia in CML patients. A constitutively active Abl and Bcr-abl feature the physiology of CML (). Furthermore, a disruption of the gene product has been reported as a possible anti-cancer target (,). It is possible that the disruption of might help in the inactivation of Abl in case of CML. In addition, the lack of skin cancer in CSB patients might be due to the fact that because of lack of CSB in these patients, activation of c-Abl is hindered inhibiting the process of onset of carcinogenesis. These findings suggest that CSB may regulate auto-phosphorylation and activation of c-Abl in response to oxidative stress and it might play a role in the development of cancer. In addition, we also speculate that tyrosine phosphorylated CSB may play a specific role in regulating the response to oxidative DNA damage.
This study also shows that CSB and c-Abl co-localize in the nucleus and re-distribute in the nucleolus in cells treated with HO. This co-localization is inhibited by pre-treatment with STI-571. The redistribution of c-Abl and CSB may facilitate interaction between the two proteins and tyrosine phosphorylation of CSB. Alternatively, phosphorylation of CSB might promote its redistribution to sites of oxidative damage; this is consistent with the observation that redistribution of CSB and c-Abl is inhibited by STI-571. Furthermore, c-Abl and CSB do not co-localize to the nucleolus in unstressed cells, suggesting that activation of c-Abl may play a role in initiating redistribution. In future experiments, CSB tyrosine 932 mutants could be used to determine the role of c-Abl mediated tyrosine phosphorylation of CSB in the process of CSB redistribution and in the response to oxidative stress.
In conclusion, our results suggest that c-Abl interacts with and tyrosine phosphorylates CSB. This interaction may play an important role in the response to oxidative stress, resulting in activation of c-Abl, tyrosine phosphorylation of CSB and more efficient BER of oxidative DNA damage. Tyrosine-phosphorylated CSB may serve as a signal for repair proteins to localize to DNA damage and may help maintain active transcription in the nucleolus. These data also provide a first insight into a novel possible role of CSB as a signaling molecule in response to oxidative damage. This role can further be exploited to understand the phenotype of CS as well as in a general understanding of the role of CSB in response to various oxidative and DNA damage. Further studies are needed to understand the impact of tyrosine phosphorylation on the physiologically important functions of CSB . |
Estrogen receptor alpha (ERα) is a ligand-activated transcription factor that alters the expression of a wide variety of estrogen-responsive genes in target cells (,). It is essential for development of the reproductive tract and maintenance of reproductive function (,).
ERα is comprised of six functional domains (A–F) that have been evolutionarily conserved (,). The most highly conserved region is domain C, the DNA-binding domain (DBD), which is comprised of two zinc finger domains. The DBD is necessary and sufficient for specific interaction of the receptor with its DNA recognition sequence, the estrogen response element (ERE). Domain E, the ligand-binding domain (LBD), is also highly conserved and directs the specific interaction of the receptor with hormone. In addition to these two highly conserved domains are regions with considerable variation in amino acid sequence, including the amino terminal A/B domain, the carboxy terminal F domain, and the centrally located hinge region, domain D. Sequence analysis of ERα from different species in combination with functional studies of mutant receptors have identified two regions of the receptor that are important in enhancing estrogen-responsive gene expression (,). The ligand-independent activation function 1, AF-1, is localized in the amino terminal A/B domain of the receptor and the hormone-inducible activation function 2, AF-2, is present in the LBD (,).
Upon binding hormone, ERα undergoes a conformational change, binds to EREs residing in estrogen-responsive genes, and recruits co-regulatory proteins to initiate changes in gene expression (,). These co-regulatory proteins include chromatin remodelers, modifiers of post-translational acetylation and phosphorylation, and an increasing number of cell-cycle and DNA repair-related factors (). This extensive array of co-regulatory proteins, which possess a wide variety of functional activities, helps to ensure fine-tuned control of estrogen-responsive gene expression.
In order to identify novel co-regulatory proteins involved in ERα-mediated gene expression, we utilized a modified gel mobility shift assay to isolate proteins associated with the DNA-bound receptor and then identified the isolated proteins by mass spectrometry analysis (,). One protein of particular interest was proliferating cell nuclear antigen (PCNA), which is required for DNA replication and repair. Interestingly, PCNA interacts directly with the DNA repair protein flap endonuclease-1 [FEN-1 ()], which we recently identified as a modulator of ERα-mediated transcription (). In addition, PCNA has been used as an independent marker of breast, renal and skin cancer ().
We have characterized the association of PCNA with ERα and find that PCNA interacts with ERα, enhances the receptor–DNA interaction , and associates with endogenous, estrogen-responsive genes. Rather than influencing estrogen responsiveness in MCF-7 breast cancer cells, PCNA helps to maintain the basal expression of estrogen-responsive genes.
Nuclear extracts (20 μg) from HeLa cervical cancer cells were incubated with annealed, P-labeled oligos containing the vitellogenin A2 ERE (5′-GAT TAA CTG TCC AAA GTC A CAG T GAT CAA AGT TAA TGT AA-3′ and 5′-TTA CAT TAA CTT TGA TCA C TGTG ACT TTG GAC AGT TAA TC-3′) in the absence or presence of 400 fmol of purified, baculovirus-expressed ERα. Incubations were performed in agarose-binding buffer (15 mM Tris pH 7.9, 56 mM KCl, 0.2 mM EDTA, 4 mM DTT, 5 mM MgOAc, 0.05 mM ZnCl) with 10% v/v glycerol, 100 ng of poly dI/dC, 1 μg salmon sperm DNA and 10 nM 17β-estradiol (E) in a final volume of 12.5 μl for 10 min on ice. Proteins associated with the ERE-bound ERα were separated on a 1.75% low melt agarose gel with modified TBE buffer (4.5 mM Tris pH 7.9, 44.3 mM boric acid, 5.2 mM MgOAc and 1 mM EDTA).
For large-scale isolation of protein complexes, reactions were increased 10-fold and proteins were identified using mass spectrometry analysis essentially as previously described (). Nine discrete peptide fragments with amino acid sequence identical to that found in PCNA (LVQGSILKK, NLAMGVNLTSMSK, FSASGELGNGNIK, LMDLDVEQLGIPEQEYSCVVK, YLNFFTK, ATPLSSTVTLSMSADVPLVVEYK, DLSHIGDAVVISCAK, FSASGELGNGNIKLSQTSNVDKEEEAVTIEMNEPVQLTFALR, AEDNADTLALVFEAPNQEK) were identified in two independent experiments. These peptides comprised 57% of the total PCNA amino acid sequence. Control lanes lacking ERα were run on the agarose gels in parallel to ensure that PCNA was associated with the DNA-bound ERα and did not simply co-migrate with the receptor–DNA complex.
A bacterial expression vector encoding his-tagged PCNA (pHKEp-PCNA) was graciously provided by Zvi Kelman [University of Maryland Biotechnology Institute, Rockville, MD, USA ()]. Expression and purification of his-tagged PCNA was performed as previously described (). Protein purity was assessed on Coomassie stained gels and protein concentration was determined using the BioRad protein assay (BioRad, Hercules, CA, USA) with BSA as a standard.
Pull-down assays using transcribed and translated S-labeled full-length ERα or truncated ERα proteins ABC, AB, CD and DEF () were performed essentially as described (). Expression vectors for ▵CD1 (amino acids 180–292), ▵CD2 (amino acids 180–272) and C (amino acids 180–262) were provided by Kendall Nettles (The Scripps Institute, Jupiter, FL, USA) and synthesized as described in Stols . (). Full length and truncated ERα proteins were synthesized using the TNT T7 Quick Coupled Transcription/Translation system (Promega, Madison, WI, USA) and incubated with immobilized, his-tagged PCNA. For E domain interaction studies, pET15b-ERα (304–554) [kindly provided by Benita Katzenellenbogen, University of Illinois, Urbana, IL ()] which encoded the his-tagged E domain of ERα, was transformed into , expressed and immobilized on Ni-NTA beads as previously described (), followed by incubation with full length, untagged PCNA (a gift from John Bruning and Kendall Nettles, The Scripps Institute, Jupiter, FL, USA). Incubations were done at 4°C for 45 min in binding buffer (15 mM Tris pH 7.9, 20 mM KCl, 0.2 mM EDTA, 4 mM DTT) with or without 10 μM E.
Bound proteins were washed once with binding buffer, once with wash buffer (15 mM Tris pH 7.9, 100 mM KCl, 0.2 mM EDTA, 4 mM DTT), and then eluted with 2 × loading buffer (125 mM Tris pH 6.8, 4% v/v SDS, 20% v/v glycerol, 1.44 M β-mercaptoethanol). Eluted proteins were separated by SDS–PAGE and subjected to autoradiography (for S-labeled proteins). For pulldowns using purified PCNA, eluted proteins were separated by SDS–PAGE and subjected to western blot analysis with antibodies specific for PCNA, his-tag and ERα (sc-7907, sc-803 and sc-8002, respectively, Santa Cruz Biotechnology, Santa Cruz, CA, USA). The blots were probed with a horseradish peroxidase-conjugated secondary antibody and developed using a chemiluminescent detection system as previously described ().
Flag-tagged full-length ERα was expressed in Sf9 cells as previously described (,), immobilized on M2-agarose (Sigma, St Louis, MO, USA), and washed with purification buffer (20 mM Tris pH 7.5, 300 mM NaCl, 1.5 mM MgCl, 0.2 mM EDTA and 10% v/v glycerol). Twenty micrograms of MCF-7 nuclear extracts or purified, untagged PCNA (a gracious gift from John Bruning and Kendall Nettles, The Scripps Institute, Jupiter, FL, USA) were incubated with immobilized ERα in the absence or presence of DNA oligos containing an ERE (described above for PCNA isolation) or non-specific DNA sequence (5′-CTA GAT TAC TTC TCA TGT TAG ACA TAC TCA GAT CTA GAC ATA CTC AGA TC-3′ and 5′-GAT CTG AGT ATG TCT AGA TCT GAG TAT GTC TAA CAT GAG AAG TAA TCT AG-3′) in binding buffer, followed by washing and elution as described. Eluted proteins were separated by SDS–PAGE and subjected to western blot analysis with PCNA (sc-56 or sc-7907) or ERα (sc-8002) specific antibodies (Santa Cruz Biotechnology, Santa Cruz, CA, USA). The blots were probed with a horseradish peroxidase-conjugated secondary antibody and developed using a chemiluminescent detection system as previously described ().
ERα was expressed and purified to near homogeneity as we have previously described (). Purified ERα measuring 10–50 fmol were incubated without or with 0.5–2.5 μg of purified his-PCNA in binding buffer with 10% v/v glycerol, 100 ng of poly dI/dC and 10 nM E in a final volume of 20 μl for 10 min on ice. BSA and His elution buffer (50 mM NaHPO, 300 mM NaCl, 250 mM imidazole) were included as needed to maintain constant protein and salt concentrations. For antibody supershift experiments, an ERα- or PCNA-specific antibody (sc-8002 or sc-56, respectively, Santa Cruz Biotechnology, Santa Cruz, CA, USA) was added to the binding reaction and incubated for 10 min on ice prior to addition of DNA. Radiolabeled ERE-containing oligos were added to the binding reactions and incubated for 10 min at room temperature prior to fractionation on low ionic strength polyacrylamide gels () at 4°C with buffer re-circulation. Radioactive bands were visualized by autoradiography or were quantitated by phosphoimager analysis with Image Quant software (GE Healthcare, Piscataway, NJ, USA).
MCF-7 cells were maintained in phenol red-containing MEM supplemented with 5% v/v calf serum, placed on phenol red-free MEM with 5% v/v CDCS for at least 72 h and exposed to ethanol vehicle or 10 nM E for 2 h. Chromatin immunoprecipitation assays were carried out essentially as recommended by Millipore (Charlottesville, VA, USA) except that pelleted cells were washed three times in lysis buffer (10 mM Tris, pH 7.5, 10 mM NaCl, 3 mM MgCl) with 0.5% v/v NP-40, re-suspended in lysis buffer with 10 mM CaCl and 4% v/v NP-40, and treated with 75 U micrococcal nuclease (USB, Cleveland, OH, USA) for 10 min prior to sonication. The ERα- and PCNA-specific antibodies sc-8002 and sc-56, respectively (Santa Cruz Biotechnology, Santa Cruz, CA, USA), were used for immunoprecipitation of protein–DNA complexes. PCR primers () flanking the pS2 ERE or a region 2.8-kb upstream of this site, which contains no known ERα-binding sequence, were used for quantitative PCR analysis with iQ SYBR Green Supermix and the iCycler PCR thermocycler according to manufacturer's directions for 40 amplification cycles (BioRad, Hercules, CA, USA). One thousand, 5000, 10 000 and 25 000 genomic copies were run in parallel with each primer set during each experiment to derive a standard curve. The relative copy number for each sample was determined from the standard curve. Each sample was run in triplicate and data from four independent experiments is reported as the relative number of copies of specific DNA sequence. Significant changes in induction were calculated using the student's -test.
For agarose gel analysis, MCF-7 cells were maintained as above and treated with ethanol vehicle or 10 nM E for 15, 45 or 120 min before chromatin was isolated as described above using antibodies specific to ERα, PCNA or non-specific mouse IgG (sc-8002, sc-56 and sc-2025, respectively, Santa Cruz Biotechnology, Santa Cruz, CA, USA). Purified ChIP DNA or 10% of input was subjected to 30 cycles of PCR with 0.5–1.0 mM MgCl and i DNA polymerase (BioRad, Hercules, CA, USA) in a 20 μl reaction according to manufacturer's directions using primers flanking the oxytocin ERE-containing gene region, or a control region of the non-estrogen-responsive 36B4 gene which lacks an ERα-binding sequence (). The entire PCR reaction was run on a 1.5% agarose gel and visualized using SYBR Safe DNA gel stain (Molecular Probes, Eugene, OR, USA) under UV light on a Gel Doc with Quantity One software (BioRad, Hercules, CA, USA).
For siRNA experiments, MCF-7 cells were maintained as stated above and seeded in 12 well plates 24 h prior to transfection. Cells were transfected with 50 pmol of control (renilla luciferase, 4630, Ambion, Austin, TX, USA) or PCNA-specific siRNA oligos (Silencer validated siRNA ID 42853, Ambion, Austin, TX, USA) in the absence of antibiotics using siLentFect (BioRad, Hercules, CA, USA) for 48 h. Medium was replaced with phenol red-free MEM containing 5% v/v CDCS for an additional 24 h, followed by treatment with 10 nM E or ethanol vehicle for 24 h. Preliminary time course experiments were performed with 0–72 h of siRNA exposure to determine the amount of time required to maintain reduced PCNA protein levels. Protein knockdown was monitored by western blot analysis of whole cell lysates as described above using antibodies to PCNA and Sp1 (sc-56 and sc-59, respectively, Santa Cruz Biotechnologies, Santa Cruz, CA, USA). RNA was harvested using Trizol (Invitrogen, Carlsbad, CA, USA) and processed according to manufacturer's directions. cDNA was synthesized using the Reverse Transcription System (Promega, Madison, WI, USA). Real-time PCR was performed using iQ SYBR Green Supermix and the iCycler PCR thermocycler (BioRad, Hercules, CA, USA) according to manufacturer's directions with primer sequences specific for the PCNA, progesterone receptor (PR), pS2 and ERα mRNA (). The 36B4 gene, which is not regulated by E, was used as a control. Samples were run in duplicate for each primer set during each experiment and standard curves were derived using serial dilutions of cDNA equivalent to 0.02, 0.2, 2 and 20 ng of input RNA. The relative nanogram of RNA was determined from the standard curve. Significant changes in RNA levels due to specific siRNA or hormone exposure were calculated by analysis of variance (ANOVA) with ezANOVA (C. Rorden, , Columbia, SC, USA).
To isolate novel proteins associated with the DNA-bound ERα, we performed agarose-based gel mobility shift assays with P-labeled ERE-containing oligos, purified ERα and HeLa nuclear extracts. Although ERα slightly altered the migration of the ERE-containing oligos (A, compare lanes 1 and 2), HeLa nuclear extracts alone did not alter migration of the radiolabeled probe (compare lanes 1 and 3). However, when both ERα and HeLa nuclear extracts were utilized, a higher order protein–DNA complex was formed (lane 4). We have shown previously that these higher-order complexes are supershifted by an ERα-specific antibody (). This large protein–DNA complex was excised from the gel, proteins were subjected to trypsin digestion and the peptides were subjected to mass spectrometry analysis. Nine discrete peptides, which had amino acid sequence identical to that found in PCNA, were identified. As seen in B, these peptides, which were identified in two independent experiments, represented a major portion (57%) of the PCNA protein. Since we had previously identified two DNA repair proteins, -methylpurine DNA glycosylase (MPG) and FEN-1, which alter the ability of ERα to activate gene expression (,), we were intrigued by the identification of another DNA repair protein associated with the DNA-bound receptor.
Because PCNA was isolated as a component of a large, multi-protein complex associated with the DNA-bound ERα, it was possible that other proteins were required for the ERα–PCNA interaction. To determine whether purified PCNA and ERα could interact, flag-tagged ERα was immobilized and incubated with purified, untagged PCNA. The purified PCNA was able to interact with the immobilized ERα in the absence and in the presence of E (A, lanes 3 and 4), but did not interact with the resin alone (lane 2). Thus, PCNA interacts directly with ERα.
To define which regions of the receptor are required for interaction with PCNA, bacterially expressed, his-tagged PCNA was immobilized and incubated with transcribed and translated full-length or truncated ERα. As shown in B, full-length ERα and the amino-terminal ABC interacted with PCNA in the absence and in the presence of E. Deletion of the C domain abolished the interaction with PCNA (AB). The failure of AB alone to interact with PCNA may reflect the relatively unstructured nature of the AB domains in the absence of the C domain ().
Interestingly, a small portion of ERα containing only the DBD and hinge region interacted strongly with PCNA (CD). To further identify the central region of ERα required for PCNA interaction, sequential truncations of the hinge region were made. When 15 (▵CD1) or 35 (▵CD2) amino acids of the D domain were deleted, the interaction of ERα with PCNA was maintained. However, deletion of the entire D domain to amino acid 262 (C) eliminated the interaction of the two proteins. Thus, the inclusion of amino acids 262–272, which were previously defined as the C-terminal extension (CTE) of the DBD (,), is required for interaction with PCNA. Interestingly, the CTE not only stabilizes the receptor–DNA interaction, but also serves as a site for ERα acetylation, which in turn influences ERα-mediated transcription ().
In addition to interacting with the central portion of ERα, PCNA also interacted with the DEF domains. Unfortunately, because the ERα LBD (E) interacted with the resin used for PCNA immobilization (data not shown), we were unable to determine whether PCNA interacted with the E domain using this method. However, his-tagged E domain was expressed, immobilized on nickel-NTA resin and incubated with untagged, purified PCNA. The his-tagged E domain bound to the resin (C, lower panel), but unlike the full-length ERα (A, upper panel), failed to interact with PCNA (C, upper panel) suggesting that additional receptor domains must be required for PCNA interaction. Thus, while neither AB, C, nor E alone was able to interact with PCNA, combining these individual domains with additional, adjacent amino acid sequence induced formation of specific structural features required for PCNA interaction. These findings are in agreement with previous studies carried out with the progesterone and glucocorticoid receptors, which highlighted the interdependence of the receptor domains in maintaining the structural and functional integrity of these proteins ().
Since DNA can induce allosteric changes in ERα conformation (), it seemed possible that binding of the receptor to DNA might influence the ERα–PCNA interaction. Baculovirus-expressed, flag-tagged ERα was immobilized on anti-flag resin and incubated with nuclear extracts from MCF-7 breast cancer cells, which express endogenous PCNA (, lanes 1 and 2). ERα and its associated proteins were isolated, run on a denaturing polyacrylamide gel, and subjected to western blot analysis. PCNA interacted with ERα (lanes 5 and 6), but not with the resin alone (lanes 3 and 4). The interaction of PCNA with ERα was not influenced by addition of oligos containing a non-specific DNA sequence (lanes 7 and 8) or an ERE (lanes 9 and 10). Thus, in contrast to another DNA repair protein, MPG, which interacts more efficiently with the ERE-bound receptor than with free ERα (), PCNA interacted with ERα in the absence and in the presence of DNA.
Although the addition of DNA did not influence the ERα–PCNA interaction, it was possible that the interaction of PCNA with ERα might alter the ability of the receptor to bind efficiently to DNA. To determine whether PCNA influenced ERα–ERE complex formation, gel mobility shift assays were performed with purified ERα in the absence and in the presence of increasing amounts of purified, his-tagged PCNA. When 10 fmol of purified ERα was combined with radiolabeled ERE-containing oligos, the receptor–DNA complex was barely detectable (, lane 1). However, a dose-dependent increase in complex formation was seen when increasing amounts of PCNA were included in the binding reactions (lanes 2–4). Inclusion of an ERα-specific antibody (lane 5), but not a PCNA-specific antibody (lane 6), supershifted the receptor–DNA complex. Thus, although PCNA dramatically increased the ERα–ERE interaction, it was not present in the receptor–DNA complex. The failure of PCNA to form a stable ternary complex with the DNA-bound ERα could result from the absence of other nuclear proteins required for stabilization of the PCNA–receptor–DNA interaction and/or the extended period of electrophoresis required for gel mobility shift assays, which could prohibit the formation of a stable ternary complex containing these two purified proteins. The ability of purified co-regulatory proteins to enhance the receptor–DNA complex, but not form a stable ternary complex in gel mobility shift assays, has been reported with a number of ERα-associated proteins by our laboratory and others (,), including the DNA repair proteins MPG and FEN-1 (,). Furthermore, these findings are consistent with the ability of PCNA to enhance the interaction of FEN-1 with DNA but not form a ternary complex with the DNA-bound FEN-1 ().
Although we were unable to isolate a ternary complex containing purified ERα and PCNA with DNA in our gel mobility shift experiments, it seemed possible that PCNA might be able to interact with an ERE-containing region of an endogenous estrogen-responsive gene in its native chromatin environment. Thus, chromatin immunoprecipitation assays were performed to examine the interaction of endogenously expressed ERα and PCNA with the native pS2 gene in MCF-7 cells using real-time PCR analysis. Consistent with previous studies (,,,), more ERα was associated with the ERE-containing region of the pS2 gene in the presence than in the absence of E (A). In contrast, no change was observed in the association of ERα with a region 2.8-kb upstream of the pS2 ERE (Control), which lacked an ERα-binding site. Likewise, when a PCNA-specific antibody was utilized, there was no change in the association of PCNA with the upstream pS2 region in the absence or presence of E. However, a modest, statistically significant increase was observed in the association of PCNA with the ERE-containing region of the pS2 gene in the presence of E.
We also examined another ERE-containing gene region using agarose gels to visualize the PCR products. While both ERα and PCNA were present at the ERE-containing region of the oxytocin gene in the absence of hormone, more ERα and PCNA were associated with this gene region when cells had been treated with E for 45 min or 2 h. The protection of the pS2 ERE and the association of ERα with the ERE-containing region of the pS2 gene in the absence of hormone have been reported previously (,). In contrast, neither ERα nor PCNA was associated with the 36B4 gene, which contains no ERα-binding site and is unaffected by hormone treatment. When a non-specific IgG control antibody was utilized, the amount of amplicon produced was far less than observed when an ERα- or PCNA-specific antibody was used. Thus, PCNA was present at the ERE-containing region of the estrogen-responsive oxytocin gene in the absence and in the presence of hormone. Combined with our gel shift assays, which demonstrated that PCNA enhances the ERα–ERE interaction, these findings suggest that PCNA may help to stabilize the interaction of ERα with endogenous, ERE-containing DNA regions in the absence and in the presence of E.
The ability of PCNA to enhance the ERα–ERE interaction and associate with estrogen-responsive genes suggested that it might be able to influence ERα-mediated transcription. To examine the potential effect of PCNA on transcription of endogenous, estrogen-responsive genes in their native chromatin environment, siRNA was employed to knock down PCNA expression in MCF-7 cells. In addition, siRNA directed against renilla luciferase was used as a control. When PCNA-specific siRNA was used, PCNA protein levels were decreased and remained low for 72 h after siRNA treatment (A). In contrast, PCNA levels were not affected by control siRNA. The level of Sp1 protein was also monitored to help ensure that neither of the siRNAs utilized had an overall effect on protein expression. Sp1 is a transcription factor that plays an important role in regulating expression of a number of estrogen-responsive genes () including the progesterone receptor (PR) gene (). The level of PCNA mRNA was also examined using quantitative real-time PCR. Exposure of MCF-7 cells to E increased PCNA mRNA levels when control siRNA was used (B). When PCNA-specific siRNA was used, significant decreases in PCNA mRNA were observed in the absence and in the presence of E.
To determine whether decreasing endogenous PCNA expression would alter the expression of endogenous estrogen-responsive genes, we examined the expression of the well-studied pS2 gene. Exposure of MCF-7 cells to E increased pS2 mRNA in the presence of control siRNA as has been reported previously (,). Interestingly, the PCNA-specific siRNA decreased basal pS2 mRNA levels, but did not affect estrogen-induced pS2 mRNA levels. We also examined expression of the human PR gene, which lacks a palindromic ERE sequence, but instead derives at least part of its estrogen responsiveness from multiple AP-1 and Sp1 sites (). In agreement with previous studies (,), PR mRNA levels increased when MCF-7 cells were treated with E and control siRNA. Interestingly, as we had observed with the pS2 gene, basal PR mRNA levels were significantly decreased in the presence of PCNA-specific siRNA, but no change was observed in the level of PR mRNA when MCF-7 cells were treated with hormone.
Since ERα regulates estrogen responsiveness, we also examined the ERα mRNA levels. When control siRNA was used, an E-induced decrease in ERα mRNA level was observed as has been reported previously (,). However, when PCNA-specific siRNA was utilized, ERα mRNA levels were decreased in the absence, but not in the presence of E as we had observed with the pS2 and PR genes. In contrast, no change was observed in the level of 36B4 mRNA, which is constitutively expressed, in the absence or in the presence of hormone when control siRNA was used. A slight increase in the level of 36B4 mRNA was detected in the presence of E when the PCNA-specific siRNA was used, but the magnitude of this increase was less than the decreases in basal mRNA levels observed for the estrogen-responsive genes.
We have identified a novel interaction between ERα and PCNA, a protein required for DNA replication and repair. We have shown that PCNA interacts with ERα, enhances the ERα–ERE interaction and helps to maintain basal expression of estrogen-responsive genes.
PCNA is present in cells as a homotrimer comprised of three PCNA monomers that encircle the DNA helix in a head-to-tail orientation (). While associated with DNA, PCNA serves as a loading dock for proteins involved in DNA replication and repair. It binds to DNA polymerase δ and FEN-1 and increases their catalytic activities thereby enhancing DNA replication and processing of Okazaki fragments (,).PCNA participates in numerous DNA repair pathways including nucleotide excision repair, base excision repair, mismatch repair and double-strand break repair (). The extraordinary versatility of PCNA is evident in its ability to interact with more than 20 polymerases, ligases, endonucleases and helicases involved in DNA replication and repair ().
PCNA functions as a sliding clamp that advances along template DNA while recruiting DNA replication and repair factors and tethering them to DNA (). Although the association of many replication factors with DNA is transient, the interaction of PCNA with DNA is more sustained (). It has been suggested that the continued association of PCNA with DNA and the transient association of its binding partners enables PCNA to simultaneously coordinate DNA replication, DNA repair and cell-cycle progression (,). This persistent association of PCNA with DNA might explain the presence of PCNA at the endogenous ERE-containing gene regions in the absence of E and the modest increases we observed in the presence of E.
In addition to its interaction with ERα, previous studies have documented the interaction of PCNA with other transcription factors including the retinoic acid receptor (RAR) and the nuclear receptor coactivator p300 (,). Since the E-occupied ERα interacts with p300 (,,,) and p300 interacts with PCNA, PCNA may help to form an interconnected network of regulatory proteins associated with nuclear receptors to modulate gene expression in the presence of hormone. Our ChIP assays () suggest that PCNA may also be essential in stabilizing the receptor–DNA interaction and maintaining basal expression of estrogen-responsive genes in the absence of hormone, when fewer co-regulatory proteins are present to stabilize the ERα–ERE interaction. Thus, in addition to serving as a platform for the recruitment of proteins involved in DNA replication and repair, PCNA may serve as a platform for nuclear receptors and co-regulatory proteins involved in modulating transcription.
Recently, Ivanov . () used computer modeling to examine PCNA–DNA interaction and showed that PCNA contacts the minor groove of the DNA helix and causes the DNA to tilt. Previous work from our laboratory demonstrated that ERα CTE induces conformational changes in DNA structure resulting in compression of the major groove and expansion of the minor groove (,). Since the ERα CTE is required for interaction with PCNA and stabilizes the ERα–ERE interaction (), it is possible that PCNA may foster the ERα–ERE interaction by altering DNA structure.
Although increased PCNA expression has been reported in proliferating normal cells (,,) and at the site of uterine implantation (,), increased expression of PCNA has also been linked to decreased survival of breast cancer patients (,). Interestingly, a post-translational modification of PCNA that lowers its DNA replication fidelity has been found in MCF-7 cells and in malignant breast and ovarian cancers, but not in normal tissues (,). It has been suggested that this decreased fidelity of the modified PCNA in cancer cells may play a role in tumor progression and decreased survival, whereas the unmodified PCNA in normal cells may help to protect the integrity of the genome.
While PCNA is required for DNA replication and repair and has been used as a marker of proliferation, our studies expand the functional repertoire of PCNA from the realm of DNA replication and repair and highlight its involvement in altering gene expression. |
Since its introduction, the benefits of 454 sequencing () have been exploited for an increasing number of applications, including genomic sequencing (), cDNA sequencing () and ultra-deep amplicon sequencing (). Despite its power in generating a great number of sequences from few samples, 454 sequencing is at present unsuitable for studies requiring targeted sequence data from many different samples. This limitation is particularly pertinent to population and medical genetics studies. The 454 sequencing plate can be physically divided into a maximum of sixteen regions, each of which yields an average of 0.63 and 2.88 Mb per run for the GS20 and the new GS FLX sequencing platforms, respectively. If shotgun libraries from 16 human mtDNA genomes are sequenced in a single run, each genome will on average be covered 70- and 350-fold for GS20 and GS FLX, respectively. Despite the low cost per sequenced nucleotide, the high coverage and cost per run make this approach impractical. Furthermore, the physical separation of the plate reduces the total number of sequences retrieved from one run to roughly half and consumes more time and material, as each sample must be processed separately. Thus, even further physical separation of the plate would not solve these inherent problems.
One approach to overcoming these limitations is to barcode samples with sample-specific sequence tags. Samples are pooled prior to 454 sequencing and are identified after sequencing by their unique sequence tags. This approach has been used to barcode cDNA libraries with tagged primers for reverse transcription (). Recently, Binladen () used 5′-tagged PCR primers to distinguish amplicon sequences derived from different sources. However, this method requires the synthesis of sample-specific primers for each target under study, which is time-consuming and cost-prohibitive when dealing with large sample sizes. Currently, no method exists for barcoding small genomic libraries or amplicons derived from untagged PCR primers. The resources of 454 sequencing therefore cannot be fully exploited for applications that require low coverage sequencing of many different samples.
Despite the potential benefits a barcoding method for 454 sequencing offers, any proposed technique must ensure efficient use of sequencing resources and high data reliability. Incomplete reactions and sequencing errors can result in background sequences without a sequence tag, heterogeneous sequence representation among samples and false-assignment of sequences to their sample origin. We have developed a method called parallel tagged sequencing (PTS) that largely alleviates these problems.
PTS is based on a ligation strategy analogous to the one utilized in the standard 454 library preparation procedure (), the use of barcoding adapters and a restriction system that excludes background sequences. An outline of the method is displayed in A. In separate reactions DNA molecules from different samples are blunt end repaired and phosphorylated. Subsequently, sample-specific barcoding adapters are ligated to both ends of the molecules, and the resulting nicks are removed by a strand-displacing polymerase. Each barcoding adapter is comprised of a single self-hybridized palindromic oligonucleotide containing an SrfI restriction site flanked by complementary sequence tags (B). This setup requires only a single oligonucleotide to be synthesized in order to barcode each sample. The barcoded samples are then quantified, pooled in a ratio reflecting the proportion of sequences desired from each sample and treated with phosphatase to remove residual 5′ phosphates of unligated ends. Such unligated ends may arise during the adapter fill-in step in molecules containing single-strand nicks caused by nebulization. Half of the adapter is then cut off by SrfI, which is a rare cutter in mammalian genomes (). It leaves blunt ends with 5′ phosphates and allows the pooled samples to be directly processed using the standard 454 library preparation. Phosphatase treatment in conjunction with SrfI digestion effectively reduces background sequences when starting from nebulized DNA; untagged molecule ends are prevented from being ligated to the universal 454 adapters during library preparation.
The sequence output from barcoded libraries begins with the sequence key TCAG, which originates from the universal 454 adapters, followed by four non-informative nucleotides GGGC from the SrfI site. Since the 454 technique uses sequencing-by-synthesis, all G's are incorporated in the same nucleotide flow as the last key base, and therefore do not decrease the read length. The origin of the sequence is identified through the adjacent sequence tag. We developed a tag design that is particularly robust to sequencing errors associated with homopolymers, a well-known problem in 454 sequencing (,,). At a length of 6 bp, it allows the pooling of a maximum of 72 samples into a single sequencing library (B). Since the 454 sequencing plate is always split into at least two regions, this configuration allows for the parallel processing of up to 144 samples in a single run.
We demonstrate the power of parallel tagged 454 sequencing by shotgun sequencing six human mtDNA genomes on two 16th plate regions of the GS20 platform. Published Sanger sequences for these samples allow a comparison of the two sequencing approaches. Additionally, we verified the reproducibility of the method by independently sequencing a second set of six human mtDNA genomes.
Six human genomic DNA samples and their respective mtDNA genome sequences were obtained from a previous study and the CEPH panel (,). For the replicate experiment, six additional samples were kindly provided by Mark Stoneking. The mtDNA genomes were amplified in two overlapping fragments using the Expand Long Range dNTPack kit (Roche) and equimolar mixtures of 5′ blocked and unmodified PCR primers in order to avoid sequence overrepresentation within the two amplicon overlaps (oligonucleotide sequences listed in Supplementary Table 2). Amplicons were purified using the PCR purification kit (Qiagen) and quantified using an ND-1000 spectrophotometer (Nanodrop Technologies). Amplicons from each sample were combined in equal molar ratios to obtain roughly one microgram of DNA per sample. Shotgun DNA libraries with a mean fragment size of ∼500 bp were prepared using nebulizers and chemicals from the GS20 library preparation kit (Roche) according to the manual.
Raw 454 reads from the standard flow files (SFF) begin with the artificial sequence TCAGGGGC followed by the 6 bp barcode. To compensate for homopolymer miscalls and carry forward effects, two to six Gs, followed by one or two Cs and an additional G were accepted while scanning for barcode sequences. If the following 6 bp had the general pattern of a barcode, starting with A or T, ending with C or G with adjacent bases differing, the ‘clip_adapter_left’ entry in the SFF file was overwritten to the point after the barcode, causing the 454 software tools to ignore this part of the sequence. The 3′ ends of the reads were then scanned for the reverse-complement of the barcode sequence and, if found, the ‘clip_adapter_right’ entry was set to also remove this part. The modified SFF files were then separated according to the barcode sequences using ‘sfffile’ and directly submitted into runMapper v1.0.53.17 for assembly. The consensus sequences are available as Supplementary Data.
Nebulized DNA fragments from six human mtDNA genomes were barcoded, pooled into a single sequencing library and sequenced on a small region of the 454 picotitre plate. An independent replicate experiment was carried out using six different samples. Only a limited amount of variation was observed for the sequence representation among samples: between 1995 and 4154 sequences were obtained from each sample in the first experiment (A); the second experiment yielded between 3267 and 5731 sequences from each sample (B). The variation was observed both within and between runs, indicating that there is no underlying ligation bias leading to differential efficiencies in the tagging reactions. In fact, we could not detect incomplete tagging with any of the barcoding adapters when optimizing the protocol using a PCR product as template. Therefore, we think that a more accurate quantification of the number of molecules contained in each barcoded sample is likely to further increase the homogeneity of sequence representation among samples. This could be achieved by DNA quantification methods more sensitive than UV absorption, such as fluorescence-based picogreen quantification assays (). In addition, for each barcoded sample the mean fragment size could be determined by capillary electrophoresis (), which allows for inferring the actual molar concentration rather than the simple mass concentration, thereby accounting for differences in fragment size distributions after nebulization.
As little as ∼3% of the sequences obtained from both runs could not be assigned to a correct sequence tag. Whereas most of these sequences did not begin with a tag, 260 of the 44 016 sequences generated in both runs carried tag sequences that differed by at least one substitution from the tags that were used. In 154 of these cases, the tag sequences differed by at least two substitutions and could therefore possibly be assigned to a false sample. Depending on the number of tags used, this represents a maximum false assignment rate of 0.35%. High coverage makes it unlikely that single misidentified reads will affect consensus sequence accuracy, but if necessary, the false assignment rate could be reduced by extending the tag length. The average coverage for each genome ranged from 10- to 22-fold and from 19- to 32-fold in the first and second experiment, respectively (see Supplementary Table 1). The coverage distribution along the mtDNA genome shows only moderate variation, with the exception of under-representation of reads close to the amplicon ends ( and Supplementary Figure 1). This arises due to reduced breakage of molecules at their ends. Thus, apart from the regions adjacent to the shorter amplicon overlap, all positions of the mtDNA genomes were covered by at least one sequence read, except for 3 bp from one of the 12 sequences.
To assess the resulting sequence quality, the consensus sequences of six mtDNA genomes were compared to their previously published Sanger sequences. In 99 kb of sequence four substitutions were detected between the Sanger and the 454 consensus sequences, corresponding to more than 99.99% agreement. However, we observed single base-pair insertions or deletions at a frequency of 0.27% on average (see Supplementary Table 1). Indel rate estimates derived from previous 454 genome sequencing projects are lower (,,), but in-depth comparisons are difficult due to the different natures of the sequenced genomes and the limited amount of data available in this study. The frequency of indels decreases as coverage increases (see Supplementary Figure 2), but higher coverage sequencing would be required to determine whether the indel rate decreases to previously determined rates at >30-fold coverage.
PTS allows the parallel generation of DNA sequences from a large number of samples using 454 high-throughput sequencing. It provides several advantages over simple physical separation of the sequencing plate or previously published tagging techniques. First, it allows highly parallel sequencing of small shotgun libraries derived, for example, from long-range amplicons or plasmids. Second, PTS can be applied to the sequencing of single or pooled PCR products. It thereby facilitates a simple switch from classical Sanger sequencing to 454 sequencing without the need to change the experimental design of existing PCR applications (e.g. by ordering new primers). This is particularly useful for applications that require microbial subcloning, since 454 sequences derive from single molecules. The oligonucleotide tags can be reused for any PCR system and on any type of double-stranded DNA. Third, the number of samples that can be sequenced in parallel using PTS is increased by an order of magnitude compared to the use of a plate's 16 subunits. Moreover, since physical separation decreases the overall number of sequences obtained, PTS enables targeted DNA sequencing at a lower price per nucleotide. In fact, optimal use of this technique can drastically reduce sequencing costs compared to traditional Sanger sequencing. Using PTS, a single run of the GS20 platform, which usually yields at least 20 Mb, can currently provide sequences for more than 50 mtDNA genomes at ∼20-fold coverage. With an output of 100 Mb per run it should be possible to obtain more than 250 mtDNA genome sequences at this coverage on the higher throughput GS FLX platform. Our data indicate that aiming for 20-fold average coverage is sufficient to obtain complete consensus sequences for all or the vast majority of the samples, as we expect each genome to be covered between 10- and 30-fold. Even at 10-fold average coverage, the resulting sequence quality is sufficient to accurately identify substitutional differences. However, the frequency of single base-pair insertion and deletion errors in and around homopolymers increases with lower coverage. These errors can be easily identified and eliminated when closely related sequences are available for comparison, but may represent a more serious problem when this is not the case. Therefore, the optimal coverage range should be determined according to each study's particular requirements. For nuclear sequences, higher average coverage will be necessary if polymorphic positions must be reliably detected.
For sequencing continuous stretches of DNA, the shotgun sequencing of long-range PCR fragments using PTS provides several advantages over traditional Sanger sequencing. First, it is much less time-consuming, as it reduces the work load for setting up PCRs and sequencing reactions. Second, it decreases the risk of allelic dropout due to primer template mismatches, which is especially important when working with species for which no genomic sequence information exists. Third, apart from the primer sequences, no prior information is needed for sequence assembly. This feature is especially useful for reliably obtaining unknown exon–intron structures of genes when primers are designed to anneal to conserved exonic sequences. The use of SrfI, a restriction enzyme with restrictions sites approximately every 150 kb in the human genome, guarantees a minimal number of dropouts due to restriction occurring within a sequence fragment. Even if a sequence fragment contains an SrfI site, due to the random fragmentation of DNA by nebulization, this would only prevent coverage immediately adjacent to the SrfI site itself.
In summary, we have developed a system that allows for high-throughput sequencing of targeted DNA sequences using the 454 system or any other parallel sequencing system (). Our technique substantially increases the throughput for targeted sequencing compared to traditional Sanger sequencing, while at the same time decreasing the cost per nucleotide. PTS can be readily adapted to any existing experimental design and enables sequencing any type of double-stranded DNA from multiple samples.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Today's DNA microarray devices contain upwards of five million features, each containing a unique probe sequence. Technological advances have continually pushed this feature density higher, ultimately allowing the construction of genomic tiling microarrays wherein large stretches of genomic sequence are represented by probes targeting it at regular intervals (). These intervals are typically 100 nt or finer and allow the unbiased monitoring of genomic functions such as DNA transcription (,) and replication (), among many other uses.
From a technological standpoint the tiling microarray's greatest achievement is in moving the DNA microarray technology from an application-specific (gene expression or genotyping) one that relies heavily on genomic annotation to a more general purpose tool. For instance, a single tiling microarray design can be used for transcript mapping, transcription factor localization and DNA replication timing, as evidenced by the recent ENCODE consortium's series of genomic experiments ().
In this respect, it may be argued that the goals of DNA microarray technology are coming full circle—a general application tool for detecting nucleic acids. With this aim, an initial vision for the DNA microarray was a matrix of oligonucleotide containing features, each containing unique -mer probes (). This matrix could, in theory, be used to query a biological sample for the presence of any nucleic acid sequence. A hindrance to the -mer construction is that such an array requires synthesizing 4 features. Naturally, larger values of infuse greater specificity into the arrayed probes, but as increases, the number of required features grows rapidly. Despite this limitation, generic -mer microarrays were initially conceptualized as a means to generate primary sequence data for the human genome sequencing effort and although this ‘sequencing by hybridization’ (,) approach has been demonstrated in a number of test cases (,), it has not enjoyed widespread use because of somewhat unrealistic thermodynamic assumptions about microarray hybridization. For the arguably simpler application of measuring gene expression, theoretical studies have suggested that universal arrays containing all possible 10-mers would be adequate () but this claim is yet to be substantiated in a working system.
Although the -mer approach has largely been abandoned with at least one notable exception (), we hypothesize that generic microarrays may be unintentionally re-emerging with the development of tiling arrays. Several contributing factors have led us to contemplate this hypothesis. First, oligonucleotide fabrication technology has improved microarray feature density upwards to five million features per array. This allows for the vast sequence coverage needed in a universal array system. Second, in many tiling array applications (e.g. transcript mapping and ChIP-chip applications) only a very small fraction of the genome is expected to be ‘active’. This would leave most of the array's features with very little target-specific activity, if any. Third, it is well known that the short oligonucleotides used in many tiling microarrays may be prone to bind weakly with off-targets (). These points, when taken together, suggest that biologically active regions of a genome not represented on a tiling microarray may still leave weak signatures of their activity in the ‘inactive’ regions targeted by tiling array probes.
If this hypothesis were true, a consequence would be that tiling microarrays targeting the human genome (or random oligonucleotides, for that matter) could be used to bridge the gap towards making DNA microarrays generally applicable to any organism and/or application. One would simply hybridize labeled nucleic acids to the tiling (or random) array and then read off intensities corresponding to probes that cross-hybridize to the targets that they are interested in. The target specificity for a single cross-hybridizing probe would, of course, be much less than that of a perfectly complementary probe but one could theoretically pool data from the features that might cross-hybridize to the subsequences present within the target sequence. In this way, the loss in specificity may be made up for by greater coverage of the region.
Should such data prove useful, this approach would certainly be attractive to researchers studying organisms poorly supported by array manufacturers. A similar method suggested for coping with this reality is to perform so-called cross-species hybridizations (). As the name implies, this procedure calls for the hybridization of RNA (or reverse-transcribed cDNA) obtained from one species to a microarray designed to target another species’ genetic material. Cross-species strategies have yielded many meaningful results (), indicating that useful information can be measured from cross-hybridization signals alone.
To investigate whether the concept of a species-non-specific universal array may be re-emerging with tiling arrays, we have simulated the scenario of using nearest-neighbor features to measure transcript abundances by using tiling microarray data that targets one part of the human genome to predict expression levels genome-wide. We have adopted an intensity prediction strategy for gene expression profiling and while we believe that this approach would not replace existing microarray strategies currently in use for studying human and other model organisms’ gene expression patterns, we do offer the technique as a theoretically viable option for someone studying RNA expression in a species for which no commercial arrays exist or for someone wishing to assay non-genic regions in an organism for which no tiling arrays are available. While our results do not indicate perfect concordance with signals that might be obtained via traditional means, we do demonstrate very significant trends that can certainly be useful in a hypothesis-generating setting, where DNA microarrays are typically employed ().
The data set we studied uses 98 unique microarray designs to tile ten human chromosomes at a five base pair resolution (). Each array probes approximately 760 000 unique genomic tiles with one perfectly matching 25 nt oligo and one 25 nt oligo identical to the perfect match, save the 13th nucleotide; this nucleotide is replaced by the complement nucleotide of the perfect match probe's 13th nucleotide. With these arrays, eleven different RNA populations isolated from nine different cell lines were probed. Samples were probed an average of three times. Nine samples contained polyA-selected RNA and two contained total RNA. Nine of the eleven samples contained cytosolic RNA while two contained nuclear RNA.
Microarray data were normalized as follows. First, the minimum feature intensity was computed for each array and decremented by one intensity unit. This value was then subtracted from every measurement in every array such that each array subsequently had a minimum signal intensity of one. This subtraction approximates the removal of optical background noise (). Each array's signals were then log transformed and the entire data set was subsequently quantile normalized () to remove any array-specific effects such as differences in cDNA concentration hybridized to the arrays.
To find features that are close in sequence to a desired nucleic acid target, we first divided the target's nucleotide string into all of its length 25 substrings. Then each of these substrings was used as a query to the database of probe sequences that exist in a given microarray design (e.g. chip01). The nearest-neighbor feature for a substring was then defined such that its probe sequence had fewer mismatches to the query substring than any other probe sequence present on the array. If multiple features had probes with an identical maximal number of matches to the query substring, then one is chosen at random to be the nearest-neighbor feature. This procedure is schematized in . Unless otherwise noted, we ignored features whose probes were an exact match to a query substring.
The Refseq database of well-curated human genes, based on the March 2006 build of the human genome, was downloaded on 6 February from the UCSC table browser (). As of this download date, Refseq contained 25 319 nucleotide sequences for which our nearest-neighbor queries were conducted.
In determining whether or not a gene is transcribed, we first found all of its nearest-neighbor features as described above. For each of these identified features, we sampled an additional feature from the same array whose probe sequence had identical GC content. We then counted the number of times the nearest-neighbor feature had signal greater than the GC-content matched feature's signal and added to this quantity half the number of times these two quantities were equal. Dividing this value by gave us the observed proportion, , of nearest-neighbor features exhibiting signal greater than their GC-content matched control features. The significance of this proportion under the null hypothesis of = 0.5 can be computed directly via summing the tail of the binomial probability distribution function. Since we had a very large number of nearest-neighbor features per gene, we simplified this computation by converting to a standardized -score:
Specifically, the 0.25 in the denominator follows from the formulation of a Bernouli random variable's variance as its expected value multiplied by one minus its expected value. Since our expected value under the null hypothesis is 0.5, our variable's variance is 0.5(1−0.5) = 0.25. The -score was then converted to a -value using the standard normal curve.
s
o
u
g
h
t
t
o
d
e
t
e
r
m
i
n
e
w
h
e
t
h
e
r
e
x
i
s
t
i
n
g
t
i
l
i
n
g
m
i
c
r
o
a
r
r
a
y
p
l
a
t
f
o
r
m
s
m
i
g
h
t
b
e
c
o
n
v
e
r
g
i
n
g
t
o
w
a
r
d
s
g
e
n
e
r
a
l
-
p
u
r
p
o
s
e
n
u
c
l
e
i
c
a
c
i
d
d
e
t
e
c
t
i
n
g
d
e
v
i
c
e
s
w
i
t
h
t
h
e
i
r
h
i
g
h
e
r
f
e
a
t
u
r
e
d
e
n
s
i
t
i
e
s
.
T
o
a
s
s
e
s
s
t
h
i
s
h
y
p
o
t
h
e
s
i
s
,
w
e
s
i
m
u
l
a
t
e
d
t
h
e
s
c
e
n
a
r
i
o
b
y
‘
m
e
a
s
u
r
i
n
g
’
R
e
f
s
e
q
t
r
a
n
s
c
r
i
p
t
a
b
u
n
d
a
n
c
e
s
w
i
t
h
e
x
i
s
t
i
n
g
t
i
l
i
n
g
m
i
c
r
o
a
r
r
a
y
d
a
t
a
f
o
r
w
h
i
c
h
t
h
e
s
e
t
r
a
n
s
c
r
i
p
t
s
w
e
r
e
n
o
t
t
h
e
i
n
t
e
n
d
e
d
t
a
r
g
e
t
.
Tiling microarrays allow for unbiased analysis of genome function. This is achieved by allocating a microarray's features to probes that target genomic sequence at regularly spaced intervals. These intervals of genomic DNA are largely inactive in transcript mapping experiments. We sought to exploit these voids and the fact that short oligonucleotides can cross-hybridize to unintended sequences to measure gene expression solely with off-target nearest-neighbor features. Specifically, we have shown that these potentially cross-hybridizing features can detect the transcription of a large number of known genes. We complemented this analysis by showing that nearest-neighbor-derived summaries of exon expression correlate within genes, that nearest-neighbor-derived gene summaries correlate with perfect-match-derived gene summaries, and that nearest-neighbor summaries derived from different array designs agree with one another. Together, these findings provide evidence that a tiling microarray can function as a ‘universal’ array that could be applied to the study of any query nucleic acid sequence. This approach differs from the complete -mer complement of oligos that traditionally define universal microarrays and is potentially useful for multi-species studies such as those being carried out by the ENCODE consortium.
In addition to our proof-of-principle work, we have quantified the main limitation of using our technique. This limitation manifests as a fair amount of gene-to-gene variation in how our nearest-neighbor strategy performs with respect to correlation with signals from traditional microarray measurements (A). Therefore, we would urge that results obtained with our method be taken as suggestive and warranting of follow-up study with traditional, lower-throughput experiments. This suggestion is generally true for almost any microarray technology and is why these platforms are usually deemed hypothesis-generating ones. In our current work, we have extended this ability for hypothesis generation to a wider spectrum of applications and for a more inclusive list of species. While a greater fraction of the generated hypotheses are probably false, the technique still whittles down the large space of putative hypotheses to a more manageable list suitable for further experimentation.
The statistically significant trends present in our analyses further suggest that our approach could enable genomic-scale hypotheses to be investigated in non-model systems, where higher error rates are easily accommodated by large sample sizes (e.g. 20 000 genes). Such hypotheses might involve biological network prediction, sample clustering and classification or ontological analyses. Many valuable conclusions have been made by pursuing these questions in model organisms with traditional DNA microarrays, even when they were in their infancy and contained very high levels of gene-to-gene variability in their performance.
Again, the work that we have presented is largely a proof-of-principle. There are several extensions that could broaden the approach's usefulness. In the current work, we have used a very simple function for assessing tile:probe similarity, namely the number of mismatches between the two short oligonucleotide sequences. Functions based on their dinucleotide mismatch distance, Gibbs free energy, or length of longest common substring could be explored. Beyond changing the similarity function for finding a single nearest-neighbor probe, one can imagine using several nearest-neighbor probes’ expression profiles to predict the query's in a weighted fashion. For example, we have explored using different values for in our -nearest-neighbor lookups but found increasing beyond = 4 steadily decreased performance. We have not only concentrated on using = 1 for simplicity in our analyses and discussion, but also because the increase in correlations with perfect match probes proved to be quite small (). There are a plethora of further directions research in this area could go, especially when one considers various weighting functions. Here, we have limited ourselves to just the simplest of models for probe: target similarity to demonstrate feasibility.
Another area that might benefit from further research is the algorithm's runtime. As we have implemented our strategy, we compare a query sequence to each probe sequence within the database. Since there are upwards of five million probe sequences in the database, and query transcripts consist of thousands of queries, finding expression summaries for all of Refseq can be a time-consuming task (several days to lookup all of Refseq). Currently, we have used a brute-force parallelization to perform our lookups, but more elegant strategies may be applicable. One obvious approach would be to use short sub-sequence hashing to accelerate lookups. This could be achieved by splitting up a query into all of its component 8-mers, for example, and using these as keys into a data structure (such as a hash table or suffix tree) that records the identities of probes having all possible 8-mer sub-sequences. Such an approach would enable fast lookups and would find similar probe sequences but would not guarantee the identification of the nearest-neighbor probe. This is analogous to the ability of BLAST to identify very similar sequences to a query despite not guaranteeing the identification of the closest sequence within a nucleic acid database ().
Our work has similar aims as those where one species’ genetic material is hybridized to arrays targeting that of another, closely related species. Both this strategy, and that described here, seek to obtain functional genomic data for unintended nucleic acid targets. Data obtained in this fashion can be used in comparative genomics and other evolution-based studies of gene expression, a currently very exciting field of study (,). It is likely that gene expression summaries derived from arrays targeting phylogenetic neighbors will yield better estimates of gene expression since the arrays’ probes would have few mismatches with their cross-species targets. However, the approach outlined in this article might be better suited for studies probing material from a number of different species since a random array would contain probes equally dissimilar to any of the target species sequences being studied. No biases can arise from platform selection with a random array.
Finally, the main conclusion we wish to make is that short oligonucleotide cross-hybridization is not necessarily a bad thing. In this work, we have exploited its presence to use a microarray for an unintended purpose. In doing so, we have demonstrated that microarrays need not consist completely of probe sequences that are perfect complements to the target nucleic acid. We believe that moving forward, the design of species-specific microarrays may want to take advantage of this fact as well. |
An ideal vector for functional genetics or for gene therapy applications should allow long-term expression of the delivered transgene at close to physiological levels. Many features that regulate gene expression have been characterized in detail but other contributory factors are poorly understood. It is known, for example, that expression of a gene can be influenced by several factors such as its replication timing during the cell cycle () and its localization within the nucleus (,). Complexity of gene expression also includes alternative splicing mechanisms, alternative promoter usage and effects of DNA polymorphisms, which could explain the production of a large number of proteins from a relatively small number of human genes (around 22 000 known genes, see ). To achieve physiological conditions of expression and retention, an efficient vector system for gene therapy or expression studies should be based on sequences derived from the human genome.
The delivery of a complete genomic locus has proven to be an excellent means to express a transgene at physiological levels, in contrast to cDNA-based expression cassettes (). cDNA-based constructs have shown many limitations due to the silencing of the heterologous promoter or transgene overexpression, which can be toxic for the cell. The use of an entire genomic DNA locus, in which the native promoter of a gene and all regulatory sequences are included in the vector, has demonstrated to be effective in rescuing deficiency phenotypes by providing physiological levels of expression and correct alternative splicing and promoter usage mechanisms ().
Long-term transgene retention can be achieved through the vector integration into the host genome or through the replication and persistence of the vector in the nucleus as an extrachromosomal unit (episome). Episomal vectors are capable of long-term persistence in mammalian cells requiring the two main features of replication and segregation into daughter cells. Episomal maintenance systems offer many advantages over integrating vectors as they avoid unpredictable integration into the host genome and the risk of cellular transformation.
Many episomal systems described in the past have been derived from viral genomes (). To be maintained extrachromosomally, most episomal vector systems rely on the expression of viral products, which may possibly confer immunogenic or oncogenic properties. Recently, a new episomal system has been described, named pEPI-1, which possesses replication and episomal retention features through the function of a scaffold/matrix attachment region () isolated from the human β-interferon gene (,).
In this project, we describe the development of a high-capacity episomal vector system based on the above-described human sequence. are 70% AT-rich sequences, which are believed to play many important roles in chromatin function. The human genome is organized in a complex structure within the nucleus where DNA interacts with histones and chromatin proteins to form a 30-nm fibre, which appears to be organized into loops by interaction of with the nuclear matrix (). Besides their structural function, also play an important role in temporal and spatial organization of gene expression (,).
The episomal features of an -based self-replicating vector can be explained by its interaction with the nuclear matrix (), since cellular processes such as DNA replication and transcription are closely associated with this proteinaceous structure (). Close involvement in DNA replication has also been confirmed by the finding of sequences close to replication sites (,). The inclusion of an sequence in an expression vector can help increase the level of expression and avoid silencing of the transgene (,).
Here we have developed an episomal vector system based on an element able to deliver, maintain and express a genomic DNA transgene at physiological levels. To investigate regulated expression from a genomic locus, we used a previously characterized BAC clone carrying the 45-kb human low density lipoprotein receptor locus () within a 135-kb genomic DNA insert () since this gene is an excellent example of finely regulated expression (,). Delivery of such large genomic DNA loci by non-viral means can be technically challenging and exhibit poor efficiency. Therefore, to achieve a high level of delivery in cells we used the herpes simplex type 1 (HSV-1) amplicon vector exploiting its high transgene capacity, unique amongst viral vectors, in a system we have called the infectious bacterial artificial chromosome, or iBAC.
Here we describe the construction and analysis of the 156-kb iBAC-- vector and demonstrate intact delivery to cells by HSV-1 amplicons. We generated stable clonal cell lines carrying the iBAC-- vector as a low copy, stable episome. Functional expression studies on two independent clones demonstrated a full functional recovery of LDLR activity in a deficient cell line. transgene expression retained physiological regulation and was repressed by high sterol levels. Finally, we showed the high mitotic stability of iBAC-- by demonstrating long-term episomal retention in both clones grown in the absence of selection.
Overall, these data describe the development of a new episomal vector system lacking viral coding sequences and able to provide efficient delivery and long-term regulated expression of a large genomic DNA transgene.
p
sec
To construct pEPHZ-, we first modified the -containing plasmid pEPI to create pEPHZ, by inserting a site and the sequences necessary for vector packaging into HSV-1 amplicons (A). To incorporate the human locus into this vector we used a previously identified BAC clone containing the complete 45 kb genomic DNA locus within a 135 kb insert (CIT-B 164O19, GenBank AC011485) (). This BAC clone was retrofitted with pEPHZ using an reaction mediated by purified Cre enzyme, able to recognize the sites present on both plasmids (A).
Analysis of the resulting vector by NotI restriction digestion followed by pulsed-field gel electrophoresis showed the integrity of the 135 kb genomic insert and the presence of only one copy of pEPHZ in pEPHZ-. The presence of the expected bands confirmed the successful construction of pEPHZ- through a Cre--based recombination (B).
We packaged pEPHZ- amplicons in Vero 2–2 cells using an improved HSV-1 amplicon helper-virus-free system (). Typically amplicon titers of ∼1–2 × 10 transducing units (t.u.)/ml, measured on G16.9 cells by using X-gal staining, were achieved after concentration by ultracentrifugation.
To demonstrate intact delivery of the 135-kb genomic DNA insert, we infected CHO a7 cells with pEPHZ- amplicons at a multiplicity of infection (MOI) of 1. In all experiments, we used CHO a7 cells expressing the herpesvirus entry protein C (HveC) to enhance their transduction (). Forty-eight hours post-infection we prepared episomal DNA and performed a plasmid rescue assay. DNA inside HSV-1 amplicons is linear but it re-circularizes after entry into the nucleus. The plasmid rescue assay will shuttle re-circularized pEPHZ- molecules, from infected CHO a7 cells back into bacteria where the vector can be analysed. Plasmid DNA was prepared from bacterial colonies grown on kanamycin/chloramphenicol plates, digested with NotI and analysed by pulsed-field gel electrophoresis. pEPHZ- was shown to be delivered intact by HSV-1 amplicons and re-circularize upon infection unrearranged, with an efficiency of 82% ().
This experiment confirmed the high capacity of HSV-1 amplicons demonstrating the high efficiency of packaging and delivery of the intact 156-kb pEPHZ- vector.
In order to assess whether pEPHZ- can persist in cells as an episome, we generated CHO a7 stable clonal cell lines carrying this vector. CHO a7 cells were seeded and then infected 24 h later with pEPHZ- amplicons at an MOI of 0.1–0.5. Infected cells were diluted 48 h later in medium supplemented with G418 (450 μg/ml) and left to grow for about 10–14 days. We isolated a total number of 108 early stage single clones. In most cases the clones did not survive prolonged culture, which we interpret as being due to the episome only providing transient antibiotic resistance and not successfully becoming permanently established. We obtained ten established growing clonal lines and screened them for episomal status by plasmid rescue, which we consider the optimal test for the presence of established episome. Plasmid rescue demonstrated the vector to be present as an episome in three of the ten (30%) established clones. A shows NotI restriction analysis followed by pulsed-field gel electrophoresis of DNA rescued from two of the established episomal clones. These data demonstrate the replication/retention features of pEPHZ- since episomal plasmid DNA extracted from clones after several weeks of growth under antibiotic selection (∼70 cell generations) showed a restriction digest pattern identical to the one of the original starting DNA preparation used for packaging (A). These two clones named pEPHZ- clones 1 and 2 (abbreviated to cl 1 and cl 2) were used for subsequent work. Plasmid rescue analysis was repeated every two weeks throughout the duration of the experiment to confirm continuing episomal retention.
These results clearly show that pEPHZ- is replicated and retained episomally when delivered by HSV-1 amplicon infection and confirm the presence of an intact genomic locus as vector rearrangements were not observed.
To assess the number of copies of pEPHZ-, total genomic DNA was prepared from pEPHZ- clones 1 and 2, continuously grown in the presence of G418 selection. Southern blotting hybridization with a radiolabelled gene probe revealed the expected 8.5 kb band in the pEPHZ controls and in the pEPHZ- clones 1 and 2 (B). Band intensity was quantified using two independent systems (C) and pEPHZ- was estimated to be present in clones 1 and 2 at ∼1 and ∼4 copies/cell, respectively.
To assess gene expression from the genomic insert carried by pEPHZ-, episomal CHO a7 clones 1 and 2 were analysed for the presence of specific LDLR activity. Fluorescently labelled low density lipoproteins (DiI-LDL), routinely used for qualitative microscopy studies on LDLR, can also be used to quantify expression when the samples are processed and analysed on a spectrofluorometer ().
The two pEPHZ- clones were incubated in lipoprotein-deficient serum (LPDS) for 48 h to enhance expression. After incubation with LPDS, cell monolayers were incubated with DiI-LDL at a concentration of 10 μg/ml for 5 h after which time cells were washed and pictures were taken using an inverted fluorescence microscope. From these pictures expression similar to the wild-type cell line can be observed for both clones (A). The same samples were then lysed and analysed on a spectrofluorometer using excitation and emission wavelengths of 520 and 580 nm, respectively. The data described in B confirm the complete restoration of expression in pEPHZ- clones 1 and 2 and correction of the CHO a7 deficiency phenotype.
Expression from the endogenous locus is regulated to control the intracellular cholesterol concentration through a negative feedback mechanism. The pEPHZ- genomic DNA transgene contains the sterol response elements (), necessary for transcriptional regulation in response to sterols. To investigate regulation of expression from the 135-kb genomic DNA insert delivered by pEPHZ-, CHO a7 pEPHZ- stable clones were first incubated in LPDS for 24 h, after which time cholesterol and 25-hydroxycholesterol at concentrations of 12 and 0.6 μg/ml, respectively, were added. After a further 24 h in the presence of cholesterol and 25-hydroxycholesterol, cell monolayers were washed and the DiI-LDL assay was performed. Average reductions of expression in the presence of sterols by 66% in pEPHZ- clone 1, and 70% in clone 2, demonstrate that the classical mechanism of sterol inhibition acting on the endogenous site also applies to the iBAC-- transgene (A). Demonstration that only gene expression is affected by the presence of high sterol levels comes from fluorescence microscopy pictures of the same cell samples. No difference in reporter EGFP expression is found in control samples or samples treated with sterols (B).
Episomal vectors are slowly lost in rapidly dividing cells when selective pressure is removed and this occurs with a rate depending on the episomal retention efficiency of the vector. We used two different methods to calculate the efficiency of pEPHZ- retention. First, pEPHZ- clones 1 and 2 were grown in the absence of G418 selection and maintained in actively growing state. At each passage, episomal DNA was extracted and plasmid rescue assay was performed. DNA extracted from kanamycin/chloramphenicol-resistant bacterial colonies obtained by plasmid rescue was digested with NotI and compared with the original DNA preparation used for pEPHZ- packaging. It was possible to rescue the vector until 11 weeks in the absence of selection, demonstrating the very high retention efficiency of this episomal vector (data not shown).
Second, from the same two pEPHZ- clones grown without G418, genomic DNA was also extracted every two weeks. After 3 months (at the end of the experiment) all samples were analysed by Southern blot hybridization for copy number assessment and band intensity calculated by Phosphorimager quantitative analysis. The data shown in demonstrate a rate of loss of 0.2 and 2.4% per cell generation for pEPHZ- clone 1 and clone 2, respectively, similar to that reported for pEPI, the basic vector (,).
In this work we have developed an episomal vector for the efficient delivery and maintenance of genomic DNA loci. Many recent studies have demonstrated the advantages in gene expression that come from the delivery of a genomic DNA locus over expression cassettes based on cDNA driven by strong constitutive promoters (). These advantages include physiologically regulated levels of expression, long-term vector retention and transgene expression, and correct alternative splicing and promoter usage (,,,). The importance of physiological regulation of transgene expression can be exemplified by the gene which is mutated in familial hypercholesterolemia (FH), a disorder characterized by high plasma levels of low density lipoproteins, which form atherosclerotic plaques leading to premature cardiovascular disease (,). Previous gene therapy vectors developed for the treatment of FH have focused on the use of cassettes carrying the cDNA driven by strong promoters (). These small constructs were characterized, and , by induction of non-physiological overexpression of , which led to the formation of intracellular deposits of cholesterol followed by death of the transduced cells and, finally, loss of transgene expression (,). This does not occur in physiological conditions where levels of expression are controlled by regulatory elements present throughout the locus.
Long-term gene expression is dependent on efficient vector retention, which can be achieved by using a replicating episomal vector. Such vectors offer several advantages over integrating vectors by maintaining the transgene in an extrachromosomal state (). The transgene of interest will not be disrupted and subject to regulatory constraints (silencing), a phenomenon referred to as ‘position effect’, which can often occur in the event of integration into cellular DNA. Moreover an episomal transgene will not lead to the interruption of important cellular genes or to cell transformation by insertional mutagenesis, the latter of which can occur if the integration site is near to growth-promoting genes. It has been demonstrated for integrating vectors that these vectors tend to integrate in coding or regulatory regions of expressed genes, making the integration event even more likely to induce oncogenesis (,). In 2003, three years into a clinical trial using an MLV vector for the therapy of human severe combined immunodeficiency X-linked (SCID-X1), after an initial success in developing a functional immune system, three of the eleven patients developed leukaemia caused by vector integration and activation of the proto-oncogene (,). The important side effect observed in this clinical trial has raised important questions of safety, regarding the use of integrating vectors, which can be overcome by the use of episomal systems.
Delivery of vectors containing large fragments of DNA by non-viral means has proven to be inefficient. The obvious alternative to non-viral gene transfer is represented by the use of viral vectors, however the majority of them have a limited capacity. The HSV-1 amplicon system combines high capacity and efficient delivery with ease of construction and offers a capacity of up to ∼150 kb which can allow the delivery of ∼95% of human genomic DNA loci (see ). The recent development of an improved packaging system provides vector packaging into ‘gutless’ HSV-1 amplicons without helper virus contamination ().
In this work we describe the construction of a vector that achieves all the above-mentioned features that an efficient vector for gene therapy or expression studies should possess. We developed an iBAC- hybrid system for delivery and expression of a genomic DNA locus. In order to assess regulated levels of expression we chose to deliver the genomic locus since expression is tightly regulated by intracellular levels of cholesterol through a complex negative feedback mechanism, mediated by sterol response elements present in the promoter region (,). Therefore, we constructed the 156-kb vector pEPHZ- by recombining a modified pEPI vector, pEPHZ, with a BAC carrying the genomic locus. Infection of CHO a7 cells with HSV-1 amplicons carrying pEPHZ- and rescue of episomal DNA after 48 h showed that this vector can be delivered and re-circularizes upon entry into cells with high efficiency, demonstrating the feasibility of this approach. In order to investigate the episomal stability of the iBAC-- system, we infected CHO a7 cells and isolated stable clonal cell lines carrying pEPHZ-. The plasmid rescue assay performed on established dividing clones revealed the vector to be present in an episomal state in three out of ten pEPHZ- clones. Since no episomal vector could be rescued from the remaining seven clones we assume the vector to be present in an integrated form in these clones. We therefore observe a final percentage of 30% (3/10) of episomal clones out of the total number of isolated dividing clones. We note, first, that we did not observe a long-term clone to contain a re-arranged episome suggesting that once established the episome retains structural integrity, and second, that an efficiency of 30% for obtaining episomal retention is similar to that observed for an earlier HSV-1/EBV system when 3/9 episomal colonies of mammalian cells were obtained following infection with a similar large >100 kb episomal construct (). Plasmid rescue analysis carried out on pEPHZ- clones proved that this vector can replicate and be retained stably in an episomal state, and no rearrangements were observed. Moreover, Southern blot studies revealed the vector to be present at a low copy number, which allows pEPHZ- expression properties to be at physiological levels. Studies of vector retention demonstrated a very high retention efficiency for iBAC-- as the vector could be rescued after more than 100 cell generations in the absence of selection. Southern blot analyses confirmed these results by showing a pEPHZ- mitotic stability of ∼97.6–99.8%, in line with observations made on pEPI in other research studies (). To assess if pEPHZ- is able to express functional low density lipoprotein receptors on cell surface, we analysed pEPHZ- episomal stable clones for their ability to bind and internalize fluorescent-labelled LDLs. Qualitative and quantitative studies demonstrated a complete restoration of expression, at levels comparable to the wild-type cell line. Moreover incubation of pEPHZ- clonal cell lines with high sterol levels shows a typical reduction of ∼65–70% of expression in clones incubated with cholesterol and 25-hydroxycholesterol compared to the same clones in the absence of sterols, exhibiting an essentially identical response to that shown by endogenous expression from wild-type CHO cells.
The vector system described in this work represents a broadly applicable expression system for human genomic DNA loci. iBAC- is a novel episomal vector which is based on human sequences, producing efficient episomal retention in the absence of virally encoded proteins. Moreover, the presence of the whole genomic DNA locus allows regulated levels of transgene expression. In this study, we have shown the iBAC- system to be an efficient tool for long-term vector retention and regulated transgene expression, which are major targets of vector development. |
Circulating renin is of central importance for long term blood pressure regulation as well as for electrolyte balance in mammals (). Renin production is regulated at different levels including transcriptional regulation. Many aspects of the transcriptional regulation of renin are not unravelled, although a number of functional -regulatory regions of the renin promoter have been identified (). Non-coding DNA regions that regulate gene expression are often evolutionary conserved (). Thus genomic sequence comparisons between related species may guide the discovery of -regulatory sequences (). A study performed by Loots . showed that conserved elements can act over distances up to 120 kb in coordinating gene expression (). By means of a phylogenetic foot-printing techniques, regulatory regions which confer muscle-specific expression have been identified ().
The degree of conservation of non-coding sequences reflects evolutionary constraints. For example, if one compares experimentally validated murine enhancer sequences to those in the zebrafish genome it turns out that -elements important for developmental function are foremost conserved (), indicating substantial constraints during development and their effect on conservation.
Tools to identify conserved regions comprise numerical algorithms analysing multispecies DNA blocks, such as BLASTZ, AVID, GLASS, LAGAN and others (,).
Having identified conserved non-coding DNA in proximity to a gene of interest, the next step in the analysis is to reveal the detailed functional relevance of an identified -regulatory region. A classic way to investigate a possible role of a DNA region is to perform reporter gene assays with different single restriction pieces of the DNA region under investigation. Mathematical modelling can further help to understand how elements in the regulatory regions orchestrate transcription. For example, a -regulatory input function has been studied in the bacterium and it has been shown that the results of that study compare well with a mathematical model of the binding of the regulatory proteins cAMP receptor protein (CRP) and to the regulatory region (). Also eukaryotic transcription can be investigated using mathematical modelling. In an elaborate study dissecting the promoter of the sea urchin, it has been shown that multiple operations may be performed in the promoter complex ().
However, little is known about the complex operations of distant conserved -elements. We investigate a conserved region named CNSmd (conserved non-coding equence in mouse and dog) of ∼775 bp in size, 14 kb upstream of the REN gene, which shows conservation in mouse and dog (A). The genomic structure of CNSmd is shown in C. We extend the classic technique of reporter gene assay analysis by designing combinatorial assays, i.e. reporter plasmids containing multiple combinations of the DNA region. This design made it possible to investigate the regulatory role of this CNSmd region in combinatorial reporter gene assays. For this purpose, we investigated the -regulatory role of four ∼200 bp parts of the CNSmd in multiple combinations. We explore in 132 assays how these -elements of a human conserved non-coding sequences (CNS) act on promoter activity under different conditions and how this action may be modelled. To extract the underlying patterns of the reporter gene activity we scored multiple mathematical models. This analysis reveals that a switch model describes the biological function of this evolutionary conserved region most appropriate. The significance of the results was tested against a random background model. The biological implication of the switch is underscored by a comparison of the model parameters and a multi-species conservation map which shows that the most important region in the model shows the highest conservation scores.
For many years, research in renin transcriptional regulation was hindered by the fact that an appropriate cell model was missing (). In this work, we utilize one of the very few renin producing cell lines available nowadays, the human Calu-6 cell line ().
Approximately 14 kb upstream of the human renin gene, a conserved region of ∼775 bp was identified. This region is named CNSmd. In comparison to the renin's upstream region of the dog and mouse genome, this region shows blocks that are highly conserved and show more than 75% identity as seen in A and B. The genomic distance relationships of the conserved region are shown in C for several mammalian species.
To elucidate the regulatory impact of the different parts of the CNSmd region on renin expression, we constructed 11 reporter constructs that contain the renin promoter and varying combinations of four equally sized parts of the identified conserved non-coding region upstream as shown in D. The different promoter constructs exhibit different promoter activity when measured in the luciferase reporter gene assay. In addition, we tested the promoter activity under 12 different cellular conditions as obtained by the addition of substances which have an impact on cellular signalling cascades involved in blood pressure regulation. These different conditions provoke altered reporter activity. The detailed results of the 132 combinatorial reporter gene assays are given in .
In order to elucidate whether all four regions of CNSmd have an influence on promoter activity and to understand how the four regions of the CNSmd act together to modulate the overall promoter activity, we constructed six mathematical models and tested which one explains the data best. We start with a minimal model that assumes that the effects of the four regions modulate the expression independently. We modified this model in three directions. First, we allowed for interactions between neighbouring regions (See Interaction Model Section). Second, to include specific transcriptional modulators reacting on a certain substance in the medium, we constructed a condition-specific model where each region modulates the promoter activity depending on the substance added. Third, we constructed a switch model (A) where the most proximal region to the promoter is dominant. That is, if this region is in the construct, the other regions do not modify the promoter activity. The reason to construct such a model was first that the most proximal region had the highest influence on promoter activity. Second, we found that non-linearities that we introduced (data not shown) improved the goodness-of-fit drastically, and suggested that the most proximal region was dominant. Furthermore, we constructed a model where each region modulates expression in a multiplicative fashion, i.e. causes fold-changes on promoter activity. Finally, a combination of the switch and the multiplicative model was created, where CNS1-3 modulate the promoter activity multiplicatively and CNS4 is dominant over CNS1-3 such that if CNS4 is present in the construct, CNS1-3 have no influence. The models have different complexity, i.e. have different numbers of parameters. As more complex models usually yield better fits as they have more degrees of freedom, we tested the different complex models using two methods of model selection, namely the Akaike Information Criterion and the Likelihood Ratio Test (LRT).
We started with a simple linear model to fit the experimental results. The interaction model as well as the condition-specific model did not significantly increase the goodness-of-fit as compared with the minimal model according to the selection criteria. When a switch model was tested, it gave superior selection scores; hence it predicts the experimental data best (A–C). The multiplicative model had significant advantage over the minimal model, and the combination of the multiplicative model and the switch model gave best scores and fitted significantly better than all other models ().
Having found a model that explains the data best, we asked the question whether all four regions have a significant influence on promoter activity. To investigate this, we set the influence of each of the regions to a fold change of one and fitted the model. In line with the model selection procedure described above, we tested whether these reduced models fit the data significantly worse. We find that all four parameters are essential to explain the data. The AICs of the models without influence of CNS1, 2, 3 and 4 are 112, 116, 112 and 242.7, respectively, corresponding to -values below 0.005. Interestingly, the dominating impact of CNS4 and the fold-changes caused by the regions correspond well with the multi species conservation score (), i.e. the region with the highest influence shows the highest evolutionary conservation.
The best-scoring model is the model that has only free parameters for treatments 1 (adenosine), 5 (phorbolester), 8 (ethanol), 9 (retinoic acid) and 11 (low FCS) with an AIC 95.3. The parameters can be found in of the Supplementary Data.
To obtain an estimate of the confidence intervals of our calculated parameters, we estimated the distribution of the model parameters in a bootstrapping procedure. Assuming that the experimental error is close to a Gaussian distribution (see Supplementary Figure 2), we added Gaussian noise to the predicted data with an SD based on an error model comprising the SD of the actual measurements and the absolute value of the measurement. The results of the parameters and their 5 and 95% confidence intervals are given in of the Supplementary Data.
In order to exclude the possibility that our results may be explained simply by a random configuration, we generated a random background model. We shuffled the CNS composition of the constructs (by generating sets of construct where each construct has the same number of parts but randomly assigned to CNS1–4). For each shuffled sets, we have fitted the switch model and have estimated the prediction error. It turned out that the prediction error of the real situation is less than the 0.001 quantile of the distribution of prediction errors of the shuffled instances. Thus, we can reject the null hypothesis that our results may be explained by a random construct composition and thus provides evidence for the biological significance of our results. (For details, see Supplementary ).
Far distant regulatory regions may affect promoter activity to a large degree. Conservation in non-coding regions is an accepted guide to identify regulatory elements. We describe a conserved region of ∼800 bp in size, 14 kb upstream of the renin gene which is located upstream of a known 11 kb upstream regulatory region of the renin gene (,), and have investigated its regulatory role in newly developed combinatorial reporter gene assays. This conserved region exists in several mammalian species, such as human, dog, cow and mouse. We extend classical reporter gene analysis by a combinatorial cloning strategy with subsequent mathematical analysis of reporter gene activity. For this purpose, we have generated 11 reporter constructs with combinations of four subregions CNS1–4 with ∼200 bp in size. We measured reporter gene activity in 132 assays under 12 different cellular conditions. We aimed to investigate how the four regions act on the promoter activity and how this interaction may be described. Following the principle of maximum parsimony, we first considered a minimal model. Taking this model as an ancestor, we added three types of complexity to this minimal model. First, we included possible interactions between the four regions, second we allowed condition specific impact of each of the four regions and third, we implemented a non-linear transfer function. According to two widely used methods of model selections, namely the Akaike Information Criterion and the Likelihood Ratio Test, we have chosen the switch model, where the most proximal region to the promoter in the minimal model, CNS4, is dominant in regulation, i.e. the other regions do only contribute to regulation if there is no outside influence on CNS4. While varying our stimuli, we did not find a stimulus where the repression by CNS4 is inactive. Therefore, the regulatory function of this region remains to be further elucidated. The estimation of the CNS1- to CNS4- specific scaling parameters reveals that the influence on the gene activity is largest for region CNS4, for both, the minimal and the switch model. To corroborate this further, we have related this to a multi-species conservation score which we have obtained from University of California Santa Cruz Genome Bioinformatics server (UCSC, ). Most interestingly, CNS4 which has the highest regulatory impact on gene activity shows the highest overall conservation scores for the regions under investigation. This fits well with the notion that regulatory important regions are under a higher evolutionary pressure.
Our modelling strategy identifies five tested cellular conditions to be important for modulating our reporter gene system, namely adenosine, phorbolester, ethanol, retinoic acid and low FCS. Adenosine did slightly stimulate the reporter gene activity. Adenosine is released from macula densa cells under high salt conditions. The effect of high salt load is a reduction in renin production (). Thus, our finding of a slight increase suggests, that other regions than CNSmd modulate dominantly the adenosine effect in an situation. This is supported by the fact that the condition-specific model did not reveal an adenosine-specific impact on a subregion of CNSmd. Retinoic acid has been shown to stimulate renin transcription through the so-called renin enhancer () which is located ∼3 kb downstream of the CMSmd region. Retinoic acid showed in our system a repressing effect on the promoter activity. This finding is interesting, because it suggests that there are different competing influences from -regulatory regions acting on promoter activity. The strongest inducing effect on promoter activity was seen in our system by the condition ‘low serum’ (FCS). This condition constitutes a stress situation for cells, which has led to a strong induction of reporter gene activity. Several other conditions which are known to influence renin transcription such as angiotensin II () and vitamin D () did not show a modulating effect. One reason might be that the angiotensin II and vitamin D effects seen in other models are not mediated via the CNSmd conserved region. The modelling approach shows that the cellular conditions tested in our study do not affect the promoter activity in a subregion-specific manner for CNSmd. The reason for that might be that we did not apply a cellular condition which conveys a subregion-specific effect for subregions of CNSmd or that our cell line model lacks the required signal transducing elements.
The results of our investigation show that a multiplicative switch model fits well the experimental results obtained in reporter gene assays. The results of the tests with random noisy data underline the biological impact of the fitted parameters and the model predictions.
We were able to exclude the possibility, that the results might be explained by a random construct composition. This was done by a rigorous statistical analysis of the distribution of prediction errors of a model with shuffled CNS composition. The results of this analysis underscore the biological relevance of our findings, in particular the dominating role of CNS4.
We have subjected the CNS4 DNA sequence to a transcription factor binding analysis using the Transfac database. The complete result list of this analysis can be found in the Supplementary Table 2. Among the factors is, for instance, NF-Y () which has been shown to play a role in renin transcription in the region of the renal enhancer ∼3 kb downstream of CNS4 by competing with other factors. Further examples are multiple putative PAX family binding sites. Members of the PAX-family are important during foetal development (). This might suggest that this region conveys signals during development. Further, in this list we find AREB6 as well as p300, both playing a role in cell type transitions (). It is clear that all the individual putative binding sites and their corresponding -factors suggested by the theoretical analysis have to be addressed by further experimental approaches in order to confirm their potential role. An alternative explanation for activity modulation of is the nucleosome occupancy that may regulate availability of binding sites on the conserved regions and hence regulate promoter activity () which has to be addressed experimentally.
Renin production is complex and relies on a number of transcriptional and posttranscriptional (,) mechanisms modulated by cAMP (,) which is known to important for a number of genes via CREB -regulatory sites (). We wish to point out that our analysis did not address the role other known -regulatory regions such as the CREB binding sites (), regulatory regions in intron 1 (), the LXR region (), the chorionic enhancer () and the renal enhancer (). Further, renin transcription depends on a number of factors which were not addressed in our study such as cAMP () and TNFalpha (,). Finally, the effects were only tested in one cell line and not in other cell lines, such as the Y-1 cell line ().
Partitioning the 775 bp region into subregions might appear somewhat arbitrary but allows us to study the regions in an unbiased manner. The possibility of destroying putative regulatory sites has been excluded by design; since the analysed isolated subregions overlap each other (). Therefore, the second most prominent peak is fully contained within CNS3, clearly seen from the fine mapped positions seen in .
To obtain a good prediction in our approach, it was necessary to introduce inter-dependence in terms of a multiplicative model. This hints to the effect of possible competition in protein–DNA interactions. Our data indicates that at the chosen scale of 200 bp interaction plays a role and has to be considered in promoter studies. This finding is important in the context of biological studies of isolated -regulatory regions and demonstrates that the combinatorial cloning approach is able to extract additional information. Our findings and the statistical analysis underscore the usefulness of our combinatorial approach.
In summary, we have shown that a multiplicative switch model explains well the data obtained in our study of four conserved regions and their impact on promoter activity. The parameters of the model are specific for the genomic region CNS1–4. The experimental condition influenced gene activity in a modulating way. We were able to identify experimental conditions with a high impact on promoter activity. The -region of highest impact on promoter activity is CNS4. This corresponds with highest conservation in a multi species comparison. This finding is compatible with the notion that there is an evolutionary pressure on regulatory sites which leads to evolutionary conservation. The biological impact of the predictions was further verified by testing the prediction errors against a randomized model, and it was clearly shown that the error of the model is at the lower tail of the distribution.
Our combinatorial cloning strategy in combination with high-throughput assay systems together with modern methods of system analysis offer the possibility to study the -regulatory role of DNA subregions in detail and are capable to discover important regulatory regions and to characterize their action on promoter activity.
A CNS of 775 bp was identified on the human chromosome 1 at position (200881908–200881134, HG17 assembly, May 2004) which shows conservation when compared to mouse and dog. This region is located ∼14 kb upstream of the REN gene. This CNS was named CNSmd (). This identification was done by visual inspection of the upstream region of the human renin gene using the web interface software ECR Browser at with the standard default parameters.
We constructed Firefly-Luciferase reporter gene constructs using the pGL3-basic (Promega, E1751) as a backbone. First, we cloned the canonical promoter (−1 to −218) of the REN gene into the pGL3 basic at the multiple cloning site (MCS) and obtained the pGL3mp. During this cloning step, the MCS was extended by an adapter to allow convenient further cloning steps. In the second step, we generated 10 further constructs (in total 1–11) by inserting combinations of four approximately equally sized CNS1 to CNS4 which are part of the 775 bp CNSmd segment in the pGL3mp. Parts overlapped in order not to destroy regulatory sequences on the boundaries. The exact positions of the segments are given in Supplementary Table 3. The combinatorial summary of the constructs 1–11 is shown in D. Cloning was achieved by PCR amplification of CNS1 to CNS4 and combinations thereof with specific primer overhangs for restriction digestion, restriction and ligation of the appropriate fragments. We used a genomic human BAC AL592114 (obtained from RZPD, Germany) as template for PCR reactions. The sequence of each plasmid was checked by sequencing.
Human Calu-6 cells (,) from the American Type Culture Collection (ATCC #HTB-56) were cultured in MEM with Earle's salt supplemented with 2 mmol/l -glutamine, 7.5% sodium bicarbonate, 1x non-essential amino acids, 1 mM sodium pyruvate, 100 U/ml penicillin, 100 μg/ml streptomycin and 10% FCS (Biochrom KG, Germany). Cells were grown at 37°C in humidified air with 5% CO.
Experiments were performed in 96 well cell culture plates (Greiner bio-one, Germany, µclear white #655098). On 80% confluence Calu-6 cells were transfected with the Firefly-Luciferase constructs (100 ng) using 0.8 µl FuGENE 6 Transfection agent (Roche, Category Number 1815075) and 75 µl medium (DMEM-High Glucose, HEPES, P/S/G). The cells were co-transfected in the same reaction with (100 ng) Renilla-Luciferase Plasmid phRL-TK (Promega, E6241) as internal control for normalization.
We modified cell culture medium and generated 12 different cellular environments 24 h after transfection. (i) adenosine 10M, (ii) aldosteron 10M, (iii) angiotensin I 10M, (iv) angiotensin II 10M, (v) phorbolester 10M, (vi) atrial natriuretic peptide (ANP) 10M, (vii) dopamin 10M, (viii) ethanol 0.1%, (ix) retinoic acid 10M, (x) vitamin D 10, (xi) low FCS (0.5%) and (xii) normal medium denoting condition 1 to 12, respectively. All pharmacological compounds were obtained from Sigma-Aldrich.
Cells were lysed 48 h after transfection using 30 µl passive lysis buffer (Promega, E1941) after medium removal and gentle washing with PBS. The assays were performed on the Luminoskan RS (Sweden) plate-luminometer using the injector system. The firefly luminescence was measured by injecting 100 µl of buffer 1 (470 µM -luciferin, 27 µM Coencym A, 33.3 mM DTT, 530 µM ATP, 2.67 mM MgSO, 20 mM Tricine, 0.1 mM EDTA) and the renilla luminescence was measured after injecting 100 µl of buffer 2 [1.1 M NaCl, 2.2 mM NaEDTA, 0.22 M KPO (pH 5.1), 0.44 mg/ml BSA, 1.3 mM NaN, 1.43 µM Coelenterazin, adjusted finally to pH 5.0, all compounds were obtained from PJK, Germany]. The device Luminoskan RS was automatically controlled using customized software (in-house development by RM).
The relative light units of firefly luminescence were divided by the relative light units of renilla luminescence of each well to obtain normalization with respect to cell number and transfection efficacy. Each experiment was performed four times and the mean value was calculated.
We developed four models to describe the expression of construct under cellular condition : a minimal model assuming condition-independent action of the four regulatory regions, an interaction model, where the regulatory regions show interactions with each other, a model where the regulatory region act condition specific, and a switch model, where the regulatory regions influence the promoter activity in a non-linear way. The latter three models are extensions of the minimal model.
The models are described below. The structure of the constructs is reflected in matrix where is 1 if region is present in construct ( = 1,2, …,11), and is 0 otherwise (D). For example the raw = 6 in D reads 0,1,1,0,1 since the sixth construct contains CNS2, CNS3 and the promoter.
We used a maximum-likelihood method () to find optimal parameters for the model. Utilizing the matlab-function , we searched for parameters minimizing the χ-distance between the data and the model (this method is also referred to as weighted least square fit):
A critical point in this fitting procedure is to estimate the variance correctly. For each data point, we had four measurements from which we estimated each data point's variance. As underestimated variances will give higher weight to the corresponding data points, we smoothed the values by applying a mixture model, estimating by the mean of the variances from the measurements and of a variance given by a linear error model:
The values and were obtained by linear regression. The assumption of this error model is that the measurement errors and residuals are normally distributed, which is appropriate here (compare Supplementary Figure 2).
We started the fitting procedure from the minimal model and extended the model in several directions: by allowing interactions, by including condition-specific regulation and by including regulation by a dominant element at CNS4. As the extended models are richer in behaviour and possess a higher number of parameters, their fits will always yield higher likelihood values. To assess, whether an extension allows for better description of the data or is just over-fitting the data, we used two methods: Likelihood Ratio Test (LRT) () and Akaike Information Criterion (AIC) (). The LRT calculates the -value under which the fit of the extended model can be obtained under the null hypothesis that the true model is the minimal model. In contrast, AIC comes from information theory and scores the models penalizing more parameters. According to the AIC framework, the model having the smallest AIC is to be chosen. AIC is calculated by:
For the entire analysis, it is important that the error is close to a Gaussian distribution which we checked by inspecting qq-plots and histograms for the experimental error and the residuals (see Supplementary Figure 2).
To estimate the confidence interval of the fitted parameters, a bootstrapping procedure was applied: we used the fitted model to generate a data set, and subsequently added noise from the linear error model. The model was then fitted to these generated data sets and the distribution of the fitted parameters was taken as an estimate of the distribution of the real parameter. The 5 and 95% quantiles were used to define a confidence interval for the parameters.
Data for multi species conservation was obtained from in April 2005. The conservation score relates to human chromosome position and refers to a joined comparison with Chimp (panTro1, November 2003)–Dog (canFam1, July 2004)–Mouse (mm5, May 2004)–Rat (rn3, June 2003)–Chicken (galGal2, February 2004)–Zebrafish (danRer1, November 2003)–Fugu (fr1, August 2002).
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Steric-blocking antisense oligonucleotides (AOs) are considered potential therapeutics for genetic diseases such as Duchenne muscular dystrophy (DMD) and β-thalassemia. For their potential to be realized, however, the AOs must be effectively delivered to cell nuclei. Cationic lipoplex- or PEI-based transfection methods used to deliver charged AOs are not suitable for the delivery of uncharged AOs such as phosphorodiamidate morpholino oligomers (PMO, ) () and peptide nucleic acids (PNAs) (). Conjugation of PMO to short CPPs is a good method to enhance the cytoplasmic and nuclear delivery of PMO because the conjugates are simple to use and because the short peptides and their AO conjugates can be easily manufactured and characterized in a quality-controlled manner. Examples of well-studied CPP−PMO conjugates include those with Tat and oligoarginine peptides (,)
Important considerations in the design of effective CPPs include the ability to deliver AO efficiently, stability in living systems and toxicity. We have reported that Tat and oligoarginine peptides are not stable in human serum (), and are therefore ill-suited for applications. Oligoarginine peptides incorporating non-α amino acids have been proven superior to oligoarginine alone. CPPs containing 6-aminohexanoic acid (X) and β-alanine (B) were more stable in human serum than Tat or oligoarginine peptides (). A CPP−PMO conjugate, (RXR)−PMO, has been shown to be more efficient in the correction of pre-mRNA mis-splicing () and in inhibition of the replication of mouse hepatitis virus () than an oligoarginine peptide. In addition, (RXR)−PMO conjugates have been shown to cause effective exon skipping in muscle cells from DMD dogs (), in human muscle explants () and in mice, as well as inhibiting the replication of various viruses in cell cultures (,) and in mice (,).
The above studies have helped make it clear that unnatural amino acids can confer enhanced stability and activity, and therefore improve the potential of CPPs to deliver therapeutic PMO. In pursuit of CPPs with improved characteristics, we have carried out a structure–activity relationship study to investigate the effects of unnatural amino acid insertions in oligoarginine peptides on cellular delivery, nuclear antisense activity, toxicity and serum-binding characteristics of the resulting CPP−PMO conjugates. The unnatural amino acids studied here are X, B and -arginine (r). We chose to study the X amino acid based on the successes of the (RXR) CPP in several studies as shown in the previous paragraph. B and r amino acids were chosen because they have good enzymatic stability (). The CPPs are (i) the oligoarginine sequences, R and R, (ii) sequences with RXR, RX and RB repeats, as well as various combinations thereof, and (iii) sequences containing -arginine, r, (rX) (rXR), (rXr) and (rB). The CPP−PMO conjugates were evaluated for their relative (a) cellular uptake, as determined by flow cytometry, (b) antisense activity, as determined by a splice correction assay () and (c) cellular toxicity, as determined by MTT cell viability, propidium iodide membrane integrity and hemolysis assays, as well as by microscopic imaging.
CPP nomenclature and sequences are listed in . Chemical structures of PMO and (RX)−PMO are shown in . The antisense PMO (CCT CTT ACC TCA GTT ACA) is designed to target a β-thalassemic mutant splice site present in the human β-globin intron 2 of a positive-readout antisense activity assay system () as described in the Results section. Synthesis of PMO, described previously (,), and the CPPs, using standard Fmoc chemistry (), were performed at AVI BioPharma, achieving purities of >90% as determined by HPLC and mass spectrometry analysis. Conjugation of a CPP to a PMO through an amide linker, described previously (), was followed with an additional purification step to remove nonconjugated peptide. Samples were loaded on source 30S resin (Amersham Biosciences, Pittsburgh, PA) in a 2 ml Biorad (Hercules, CA) MT2 column at 2 ml/min with running buffer A (20 mM NaHPO, 25% acetonitrile, pH 7.0) and purified into 45-s fractions with 0–35% buffer gradient (buffer B: 1.5M NaCl, 20 mM NaHPO, 25% acetonitrile, pH 7.0) over 60 min, using a Biorad BioLogic low pressure chromatography system. The desired faction was desalted by a method described previously (). HPLC and MS analyses revealed that the final product contained >90% CPP conjugated to full-length PMO, with the balance composed of CPP conjugated to incomplete PMO sequence, nonconjugated full-length or incomplete PMO.
The HeLa pLuc705 (pLuc705) () cell line was obtained from Gene Tools, LLC (Philomath, OR). Human liver cell line HepG2 was from American Type Culture Collection (ATCC, Manassas, VA). Cells were cultured in RPMI 1640 medium supplemented with 2 mM -Glutamine, 100 U/ml penicillin and 10% fetal bovine serum (FBS) (HyClone, Ogden, UT) at 37°C in a humidified atmosphere containing 5% CO. All treatments were carried out in OptiMEM medium (Gibco, Inc., Carlsbad, CA.) with or without FBS.
pLuc705 cells were seeded 20 h prior to treatment in 12-well plates at 100 000 cells/well. Cells were treated with 2 μM fluorescein-tagged CPP−PMO conjugates for 24 h. After treatment, cells were washed with 500 μl of cold PBS, incubated with 400 μl of trypsin at 37°C for 10 min and combined with 500 μl medium containing 10% FBS. Cells were spun down at 1000 for 3 min, washed with 500 μl cold PBS twice, resuspended in 200 μl of PBS containing 1% FBS and analyzed by a FC-500 Beckman Coulter cytometer (Fullerton, CA). Data was processed using FCS Express 2 (De Novo Software, Thornhill, Ontario, Canada).
pLuc705 cells were seeded 20 h before treatment in 48-well plates at 30 000 cells/well. Cells were washed with 200 μl medium and treated with CPP−PMO conjugates for 24 h. Cells were washed twice with 200 μl PBS (HyClone, Ogden, UT), and incubated with 100 μl of cell lysis buffer (Promega, Madison, WI) at 4°C for 30 min. Cell lysate was separated from the cell debris by centrifugation at 1000 for 10 min at 4°C. Luciferase levels were determined by mixing 30 μl of cell lysate and 50 μl of luciferase assay reagent (Promega) and measuring subsequent light production using a Flx 800 microplate fluorescence/luminescence reader (Bio-tek, Winooski, VM). The relative light units (RLU) per sample were normalized to microgram of sample protein as determined by the bicinchoninic acid method, following the manufacturer's procedure (Pierce, Rockford, IL).
pLuc705 cells were seeded 20 h before the treatment in 96-well plates at 9000 cell/well and then treated with the conjugates. The microscopic phase images of treated cells were visualized by a Nikon Diaphot inverted microscope (Melville, NY), captured by an Olympus digital camera and processed by the Magnafire software (Optronics, Goleta, CA). After imaging the cells, the cell viability was determined by the methylthiazoletetrazolium assay (MTT, Sigma, St. Louis, MO) assay. MTT solution (5 mg/ml) was added to the treatment medium to a final concentration of 0.5 mg/ml and incubated for 4 h at 37°C. 85% of the media of each well was then replaced with DMSO containing 0.01M HCl and further incubated for 10 min at 37°C and the absorbance measured at 540 nm. Percent cell viability was determined by normalizing the absorbance of each treated sample to the mean of untreated samples.
pLuc705 cells were seeded 20 h before treatment in 12-well plates at 100 000 cells/well. Cells were treated by removing the medium, washing with 500 μl PBS and incubating with medium containing CPP−PMO conjugates. Treatment medium was collected in tubes and cells were washed with PBS once, and then treated with 400 μl of 10% trypsin for 10 min at 37°C. Trypsin was neutralized with 500 μl of the serum containing medium. Cells were transferred to the tubes containing previously collected treatment medium, pelleted by centrifugation at 1000 for 5 min, washed with PBS once, and re-suspended in 200 μl of 0.05 μg/ml propidium iodide (PI) in PBS. Cells were further incubated at 37°C for 15 min and analyzed by the Beckman Coulter cytometer (30 000 events/sample collected).
The hemolytic activities of the conjugates were determined in fresh rat blood according to a method described elsewhere ().
Cellular uptake of CPP−PMO conjugates was investigated using the 3′-carboxyfluorescein-tagged PMO (PMOF) and flow cytometry. We chose 2 μM as the treatment concentration because none of the conjugates caused any detectable cytotoxicity at this concentration, as demonstrated by the MTT and PI uptake assays. After treating with the conjugates, cells were treated with trypsin () to remove membrane-bound conjugates. We found that a heparin sulfate washing step prior to trypsin treatment did not remove additional membrane-bound conjugates but caused some cellular toxicity (data not shown); therefore, only the trypsin treatment step was used in this study.
To determine the effect of serum on cellular uptake of the various conjugates, uptake evaluation assays were carried out in the medium containing various concentrations of, or in the absence of, serum. Cellular uptake of CPP−PMOF conjugates increased with the number of arginines and decreased with the X and/or B residue insertion (A and B). The oligoarginine R−PMOF had a mean fluorescence (MF) of 662, nearly 3-fold higher than the 234 produced by R−PMOF, indicating that a difference of a single arginine can make a substantial difference in the biological properties of a CPP. Insertion of an X or B residue in the R sequence reduced the MF from 234 of R−PMO to 42, 70 and 60 of (RX)−, (RXR)− and (RB)−PMOF, respectively (A). The number of RX or RB repeats affected cellular uptake, with conjugates having fewer RX or RB repeats generating lower MF (B).
While the addition of 10% serum to the medium caused a decrease in the uptake of the R− or R−PMOF conjugates, it increased the uptake of conjugates containing RX, RB or RXR motifs (A and C). Serum reduced the MF of R− and R−PMOF from 662 and 234 to 354 and 158, respectively, and increased the MF of (RX)−, (RXR)− and (RB)−PMOF from 41, 70 and 60 to 92, 92 and 111, respectively. These differences were statistically significant (A). However, higher serum concentrations (30 and 60%) decreased the uptake of (RXR)−PMOF and oligoarginine−PMOF (C).
Arginine stereochemistry ( versus ) had little effect on the uptake of CPP−PMOF conjugates. We compared the MF of R−, (RB)− and (RX)−PMOF with their respective -isomer conjugates, r−, (rB)− and (rX)−PMOF and found that there was no significant difference between each pair, as shown in D for the (RX)− and (rX)−PMOF pair.
The effectiveness of each CPP−PMO conjugate was determined in a previously described splicing correction assay (), considered a reliable method to assess nuclear antisense activity of a steric-blocking AO. This assay utilizes the ability of steric-blocking AOs to block a splice site created by a mutation in order to restore normal splicing. The luciferase coding sequence was interrupted by the human β-globin thalassemic intron 2 which carried a mutated splice site at nucleotide 705. HeLa cells were stably transfected with the plasmid therefore named as pLuc705 cell. In the pLuc705 system, steric-blocking AOs must be present in the cell nucleus for splicing correction to occur. Advantages of this system include the positive readout and high signal-to-noise ratio. With this system the relative efficiencies of various CPPs to deliver an AO with sequence appropriate for splice-correction to cell nuclei can be easily compared.
The cellular toxicity of the various CPP−PMO conjugates was determined by MTT-survival, propidium iodine (PI) exclusion and hemolysis assays and microscopic imaging. The MTT and PI exclusion assays measure metabolic activity and membrane integrity of cells, respectively. The hemolysis assay determines compatibility with blood. Microscopic images were used to verify the MTT results and observe the general health of the cells.
The naturally occurring CPPs such as Tat peptide are not stable in blood and neither are oligolysine/oligoarginine (), rendering these CPPs unfavorable as transporters for therapeutic AOs. We reasoned that one approach to improve stability would be to use non-α amino acids or -amino acids. In this study, we investigated whether incorporation of 6-aminohexanoic acid (X), β-alanine (B) and -arginine (r) amino acids into the CPP would affect cellular delivery, antisense activity, toxicity and serum binding of the resulting CPP−PMO conjugates.
We found that CPP−PMOF conjugates containing X/B residues did not enter cells as efficiently as R− and R−PMO conjugates. This is consistent with our previous finding for the (RXR) conjugate (). We have found that cell surface proteoglycans were involved with binding of the Tat−, RF− and (RXR)−PMO conjugates with the (RXR) conjugate having the lowest binding affinity. Insertion of X into an oligoarginine CPP reduces the charge density and may lead to decreased binding affinity for proteoglycans. Despite the lower cellular uptake of X/B-containing CPP–PMO, they generated higher antisense activities in the cell nucleus than oligoarginine−PMO. We have found that endocytosis was the internalization mechanism (at least primarily) for oligoarginine- and (RXR)−PMO conjugates. Indication of different uptake mechanisms was not found among these conjugates (). Therefore we hypothesize that X/B-containing conjugates have a greater ability to escape from endosomes/lysosomes than oligoarginine conjugates by a mechanism as yet to be studied.
The number of X residues affects both the nuclear antisense activity and the toxicity of conjugates. The CPP−PMO conjugate with 8 X residues [(RX)−PMO] had the highest activity followed by one with 5 Xs [(RXR)−PMO] ( and ). However, these conjugates were toxic to cells at higher concentrations, which may be a concern when considering potential applications for delivery of PMO. Replacement of all 8 Xs with Bs decreased both toxicity and antisense activity. The combination of 3–4 Xs with several B residues yielded CPPs with no detectable toxicity, and at some concentrations several of them had similar antisense activity as (RX)−PMO. We think this type of CPP, having Bs and fewer than 5 Xs, will offer balanced activity and low toxicity as well as the stability, and have considerable potential for delivery of therapeutic AOs. Further investigation into the toxicity and activity versus dosing levels of these CPPs is warranted.
Surprisingly, the replacement of -arginine with -arginine enhanced neither uptake nor antisense activity for oligoarginine, or X- and B-containing conjugates. In the case of (RX)−PMO, the replacement actually caused a small but statistically significant decrease in activity. Our observation is different from the results reported by others (,) who found that -CPPs had higher cellular uptake than -CPPs, although no biological functional cargo was used in their study. The difference between results may be due to the type and size of cargos and the cell lines used for the assays. Whether the use of -arginine-containing peptides results in superior CPP−PMO functional activity remains to be tested.
We attempted to understand the nature of (RX)− and (RXR)−PMO toxicity. It is apparent that these two conjugates caused little immediate membrane damage with 0.5 or 5 h treatment at concentrations as high as 60 μM (). However, these two conjugates had dose-dependent toxicity with 24 hr treatment as shown by the leaky cell membranes and fewer cells compared to controls (Figure 6C&D, Figure 8). Interestingly, the replacement of Xs with Bs in (RX)−PMO abolished the toxicity, and the replacement of -arginine with -arginine reduced the toxicity of (RX)−PMO (). We have found that (rX)−PMO was completely stable and the peptide portion of (RB)−PMO was only partially degraded, whereas the peptide portion of (RX)−PMO was completely degraded in cells (). We wondered whether the difference in toxicity among (RX), (RB) and (rX)−PMO conjugates was caused by differences in intracellular stability, resulting in the metabolized products of (RX)−PMO producing toxicity. The identifiable metabolized products of (RX)−PMO were XRXB−PMO and XB−PMO () but neither product had any detectable toxicity as measured by MTT assay (data not shown). It is possible that the CPP portion was degraded into free amino acids and/or smaller peptide fragments which were toxic. However, our investigation revealed that neither free R nor X, alone or in combination, caused cellular toxicity. Another possibility is that because of the high hydrophobicity of X compared to B, X in combination with positively charged arginine residues leads to toxicity not generated by B residue combinations. However, this explanation does not account for the difference in toxicity observed between (RX)−PMO and (rX)−PMO, which have the same hydrophobicity. Perhaps the toxicity of (RXR)−PMO and (RX)−PMO was caused by the peptide fragments that we could not identify by mass spectrometry.
Unlike the toxicity difference between (RX)− and (rX)−PMO, the → replacement did not change the toxicity of (RXR)−PMO (). Substitution of either one R (rXR) or two R (rXr) from the RXR repeat neither reduced nor increased the toxicity profile of (RXR)−PMO. At this point, we do not fully understand the mechanisms of (RXR)− and (RX)−PMO conjugate toxicity, but look forward to studying this topic further.
Serum effect on the activity of a CPP−AO conjugates is an important issue when considering potential applications. X/B-containing conjugates were still active in 60% serum while oligoarginine conjugates were not. The greater stability of the X/B-containing conjugates to serum enzymes is likely a factor contributing to their high activity. The loss of activity in high serum concentrations makes oligoarginine CPPs undesirable as potential therapeutic AO carriers.
In summary, we have found that the X/B-containing CPP−PMO conjugates are superior to oligoarginine−PMO conjugates for the following reasons: they display higher activity in cell nuclei, are less affected by serum and are more stable in blood (). The toxicity of the X/B-containing CPPs can be reduced by keeping the number of X residues below 5 while still maintaining a reasonable delivery efficacy and stability. This study provides a basis for further optimization of CPP sequence using R, r, X and B residues in the interest of further reducing toxicity and increasing antisense activity, which will likely lead to more effective AO transporters for potential therapeutic applications. |
Guanine quartet (G-quartet) structures with four Hoogsteen-paired, coplanar guanines were first observed more than 40 years ago () and later demonstrated to be a unique structural motif of guanine-rich oligonucleotides () (A). G-quartets are found in nature and also in sequences identified by screening techniques such as () (). They have also aroused interest as therapeutic agents to inhibit human thrombin (,), HIV infection (), and as targets themselves to inhibit telomerase activity in anticancer drug design ().
G-quartets display extraordinary structural polymorphism (). The 15-nt DNA sequence d(GTGTGTGTG) () binds and inactivate thrombin, a key enzyme in the blood clotting cascade. NMR spectroscopy reveals that it adopts a uniquely folded structure with two stacked G-quartets connected through edge-loops (one TGT and two TT loops) involving antiparallel alignment of adjacent strands (B) (). A seemly different crystal structure of this aptamer–thrombin complex (,) was later analyzed to be in agreement with the previous NMR structure (,). Potassium ions stabilize the G-quadruplex by coordination to the G residues (). The G-quartet displays an alternating and conformation of guanine bases (- and -Gs) on the same plane and 5′--G/3′--G along each strand of the quadruplex (i.e. 5′---3′ connectivity where -G includes G1, G5, G10 and G14 and -G includes G2, G6, G11 and G15 in B) and all the thymidines are in the orientation ().
The telomeric sequence d(GTG) adopts symmetrical dimeric quadruplexes comprising four G-quartets linked through a diagonal loop as analyzed by NMR in sodium (Na) environment (C) (,) or an edge-loop structure as revealed by X-crystallography in potassium (K) environment (). More recent studies re-examined the crystal structure of this sequence and suggested the crystal structure also adopts a diagonal conformation, consistent with the NMR solution structure (). The diagonal loop configuration (C) consists of ′--G3′--G along each G-strand, conformations in a G-quartet and all sugar residues puckered in the (C2′- conformation (,).
Another interesting example is the phosphorothioate oligonucleotide PS-d(TGT) identified as an inhibitor of HIV-1 infection by combinatorial screening of a library of DNA strands. This aptamer inhibits HIV envelope-mediated cell fusion and its structure consists of a parallel-stranded tetramer with all guanines in the -conformation (D) (,).
Based on the above, the glycosidic conformations of guanines (i.e. - and -Gs) have a strong correlation with strand alignments. Parallel-stranded quadruplexes (D) support only -G residues, while antiparallel-stranded quadruplexes favor alternating Gs engaged in unimolecular (B) or intermolecular complexes (C) (,).
G-quadruplexes are sensitive to chemical modifications. Several studies aimed at modifying the thrombin-binding aptamer d(GTGTGTGTG) have been reported (), but very few, if any, have led to an improvement over the original molecule. For example, Seela and coworkers recently reported the insertion of a hairpin-forming sequence GCGAAG into the position of the central loop (TGT) of the thrombin-binding aptamer. This construct formed a G-quadruplex fused to a mini-hairpin structure. According to the data, the mini-hairpin induces a structural change in the aptamer section, leading to less stable G-quadruplex. Binding to thrombin was not investigated (). Saccà . studied the effect of backbone charge and atom size, base substitutions as well as the effect of modification at the sugar 2′-position as analyzed by spectroscopy. Fully modified aptamers with sugar modifications (ribose, 2′--methylribose) and phosphate backbone modifications (methylphosphonate, phosphorothioate) led to a reduction in the thermal stability (). In fact, the 2′--methylribose modification led not only to a destabilization of the structure but to a complete transformation of the G-tetrad conformation, as shown by spectroscopy in potassium buffer (). 2′--methylribose was also shown to causes structural changes in RNA aptamers and often resulted in a loss of activity (). Accordingly, there is a need for new chemical modifications to improve the nuclease stability of this and other aptamers. Ideally, these modifications will not alter the subtle binding interactions of the selected native aptamers and the thermal stability of G-quadruplexes.
2′-Deoxy-2′-fluoro--arabinonucleic acids (2′F-ANA) confer DNA-like ( conformations () to oligonucleotides while rendering them more nuclease resistant (). The incorporation of 2′F-araN units in oligonucleotides also raises the of different systems, i.e. duplexes (∼+1°C/nt) (), triplexes (∼+0.8°C/nt) () and C-rich quadruplexes (∼+1°C/nt, pH <4.0) (). In light of this and other advantageous characteristics of 2F-ANA, such as synthetic accessibility through conventional solid-phase phosphoramidite chemistry, and promising antisense/siRNA properties (,), we have undertaken the first study concerning the ability of 2′F-ANA to form various G-quadruplex structures as analyzed by and CD experiments. 2′F-ANA modified thrombin-binding aptamers were further evaluated by their nuclease resistance and binding affinity to thrombin. The outcome of these studies has opened new perspectives for the application of 2′F-ANA as aptamer oligonucleotides.
The sequence and composition of the oligomers prepared in this study are shown in and . Arabinose modified aptamer syntheses were carried out at a 1 μmol scale on an Applied Biosystems (ABI) 3400A synthesizer using standard β-cyanoethylphosphoramidite chemistry according to published protocols (). Deoxyribonucleoside phosphoramidites were purchased from ChemGenes (Waltham, MA) and 2′F-arabinonucleoside 3′--phosphoramidites were provided by Topigen Pharmaceuticals Inc. (Montreal, Canada). The final concentrations of the monomers were 0.10 M for 2′-deoxyribonucleoside phosphoramidites and 0.125 M for the 2′F-arabinose phosphoramidites. The coupling time was extended to 150 s for the 2′-deoxyribonucleoside phosphoramidites dC and dG, and 15 min for the 2′F-araG and T phosphoramidites. These conditions gave about 99% average stepwise coupling yields. With the exception of PG17 through PG24, oligonucleotides were purified by anion-exchange HPLC (Waters Protein Pak DEAE-5PW column; 7.5 mm × 7.5 cm), desalted by size-exclusion chromatography on Sephadex G-25 resin, and characterized by MALDI-TOF mass spectrometry (Kratos Kompact-III Instrument; Kratos Analytical Inc., New York). Purity of the isolated oligonucleotides was >95% by HPLC. PG17–24 were used as obtained following deprotection and desalting.
UV thermal dissociation data was obtained on a Varian CARY 1 spectrophotometer equipped with a Peltier temperature controller. Thrombin-binding aptamers (PG1–14) were dissolved in buffer (10 mM Tris, pH 6.8, with and without 25 mM KCl) at a final concentration of 8 μM (). Thrombin-binding aptamers were annealed in buffer at 80°C for 10 min, allowed to cool to room temperature and refrigerated (4°C) overnight before measurements. dTGT and related sequences (PG17–20) were dissolved in phosphate-buffered saline (PBS buffer, pH 7.2) composed of 137 mM NaCl, 2.7 mM KCl, 1.5 mM KHPO, 8 mM NaHPO at a final concentration of 20 µM (). The telomeric DNA dGTG and related sequences (PG21–24) were dissolved in 10 mM sodium phosphate buffer (pH 7, 0.1 mM EDTA and 200 mM NaCl) at a final concentration of 100 µM (). All samples (PG17–24) were annealed at 98°C for 5 min, naturally cooled down to room temperature and refrigerated (4°C) overnight before measurements. The annealed samples were transferred to pre-chilled Hellma QS-1.000 (Cat #114) quartz cell, sealed with a Teflon-wrapped stopper and degassed by placing them in an ultrasonic bath for 1 min. Extinction coefficients were obtained from the following internet site () based on the nearest-neighbor approach () and modified aptamers (phosphorothioates and 2′F-ANA) were assumed to have the same extinction coefficient as the natural DNA aptamer. Denaturation/cooling curves were acquired at either at 295 nm for d(GTGTGTGTG) and related sequences (PG1–14), dGTG and related sequences (PG21–24), or at 260 nm for dTGT and related sequences (PG17–20), at a heating/cooling rate of 0.5°C/min between 10 and 80°C (for PG1–14), 20 and 90°C (for PG17–20) or 40 and 98°C (for PG21–24). The data were analyzed with the software provided by Varian and converted to Microsoft Excel ( and ). The decreases in UV absorbance (hypochromicity) with increasing temperature were normalized between 1 and 0 by the formula: = ( − )/( − ), where is the absorbance at any given temperature (), is the minimum absorbance reading at high temperature and is the maximum absorbance reading at low temperature. concentration dependence studies were also conducted in the same way at 295 nm using thrombin-binding aptamers (PG1–14) with different concentrations ranging from 4 to 76 μM. Starna quartz cells (Starna Cells, Inc., Cat. # 1-Q-1) with 1-mm path length were used at high concentrations to reduce the amount of aptamers required and to avoid exceeding the Absorbance range of the instrument.
CD spectra (200–320 nm) were collected on a Jasco J-710 spectropolarimeter at a rate of 100 nm/min using fused quartz cells (Hellma, 165-QS). Measurements were carried out either in 10 mM Tris, pH 6.8 (with and without 25 mM KCl) at a concentration of 8 μM for thrombin-binding aptamers (PG1–14) (); in PBS buffer (pH 7.2, 137 mM NaCl, 2.7 mM KCl, 1.5 mM KHPO, 8 mM NaHPO) for dTGT and related sequences (PG17–20) at a final concentration of 20 µM (), or in sodium phosphate buffer (10 mM sodium phosphate buffer, pH 7, 0.1 mM EDTA and 200 mM NaCl) for dGTG and related sequences (PG21–24) at a final concentration of 100 µM (). Temperature was controlled by an internal circulating bath (VWR Scientific) at constant temperature. The data was processed using J-700 Windows software supplied by the manufacturer (JASCO, Inc.). To facilitate comparisons, the CD spectra were background subtracted, smoothed and corrected for concentration so that molar ellipticities could be obtained. Temperature-dependent CD spectra were also conducted for dTGT and related sequences (PG17–20). A 10-min equilibration time was allowed at each temperature before CD scanning. The profile was obtained by plotting the maximum molar ellipticities versus temperature and normalizing.
Nuclease stability of anti-thrombin aptamers was conducted in 10% fetal bovine serum (FBS, Wisent Inc., Cat. #080150) diluted with multicell Dulbecco's Modified Eagle's Medium (DMEM, Wisent Inc., Cat. #319005-CL) at 37°C. A single-strand DNA (ssDNA) 23mer (P-8), which is unable to form G-quadruplexes, was used as a control. Approximately 8 μmol of stock solution of aptamers and ssDNA control (∼1.2 O.D.U) was evaporated to dryness under reduced pressure and then incubated with 300 μl 10% FBS at 37°C. At 0, 0.25, 0.5, 1, 2, 6 and 24 h, 50 μl of samples were collected and stored at –20°C for at least 20 min. The samples were evaporated to dryness and then 10 μl of gel loading buffer and 10 μl of autoclaved water was added. 10 μl of the mixture was used for polyacrylamide gel electrophoresis (PAGE), which was carried out at room temperature using 20% polyacrylamide gel in 0.5× TBE buffer (Tris-borate-EDTA). The degradation patterns on the gels were visualized using Stains-All (Bio-Rad) according to the manufacturer's protocol.
Aptamers (PG1–14) were radiolabeled at the 5′-hydroxyl terminus with a radioactive P probe using a T4 polynucleotide kinase (T4 PNK) according to the manufacture's specifications (MBI Fermentas Life Sciences, Burlington, ON). Incorporation of the P label was accomplished in reaction mixtures consisting of DNA aptamers substrate (100 pmol), 2 μl 10 × reaction buffer (500 mM Tris–HCl, pH 7.6 at 25°C, 100 mM MgCl, 50 mM DTT, 1 mM spermidine and 1 mM EDTA), 1 μl T4 PNK enzyme solution (10 U/1μl in a solution of 20 mM Tris–HCl, pH 7.5, 25 mM KCl, 0.1 mM EDTA, 2 mM DTT and 50% glycerol), 6 μl [γ-P]-ATP solution (6000 Ci/mmol, 10 mCi/ml; Amersham Biosciences, Inc.) and autoclaved sterile water to a final volume of 20 μl. The reaction mixture was incubated for 30–45 min at 37°C, followed by a second incubation for 5 min at 95°C to thermally denature and deactivate the kinase enzyme. The solution was purified according to a standard protocol () and the isolated yield of P-5′-DNA following gel extraction averaged 50%. The pure labeled samples were kept at –20°C for future use.
Labeled aptamers (1.25 pmol) were heated to 95°C for 5 min in the binding buffer (Tris–Ac, pH 7.4, 140 mM NaCl, 5 mM KCl, 1 mM CaCl, 1 mM MgCl) () and immediately placed on ice for 5 min before binding to increasing concentrations of thrombin protease (Amersham Biosciences, Inc.) ranging from 10 to 1000 nM in the binding buffer at 37°C in a final volume of 20 μl for 30 min. Mixtures were filtered through a nitrocellulose filter (13 mm Millipore, HAWP, 0.45 μm) pre-wetted with binding buffer in a Millipore filter binding apparatus, and immediately rinsed with 600 μl ice-cold washing buffer [Tris–Ac, pH 7.4, 140 mM NaCl, 5 mM KCl, 1 mM CaCl, 1 mM MgCl, 1% sodium pyrophosphate (w/v)], then the filter was air-dried and the bound aptamer quantified by scintillation counting. The binding percentage (%) was calculated by the subtraction of the background from the counts in the microtube. was roughly estimated from the concentration (nM) where 50% of the maximum binding percentage was observed with a certain aptamer during the thrombin concentrations studied.
italic
sub
#text
sub
xref
italic
#text
S
u
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
DNA is modified by many mutagens, including reactive oxygen species (ROS) (). When ROS react with DNA, various kinds of modified base and/or sugar moieties are produced. The most common lesion is 7,8-dihydro-8-oxoguanine (8-oxo-G), which is highly mutagenic in bacterial and mammalian cells, due to its well-known miscoding potential leading to frequent G->T transversions (). The 1,2-dihydro-2-oxoadenine (2-OH-A) is another common DNA lesion produced by ROS. 2-OH-A possesses significant mutagenic potential in living cells (,). Replication in bacteria or mammalian cells of shuttle vectors containing a single 2-OH-A produces a broad spectrum of mutations. Mutation analysis indicated that a significant fraction of the oxidation-related mutations occur at A:T base pairs. The mismatch repair (MMR) enzyme MYH, the MutY homolog which excises A incorporated opposite DNA 8-oxoG, also removes 2-OH-A from 2-OH-A:G base pairs. It has been previously shown that overexpression of hMTH1 in MMR-defective mouse and human cells reduces the level of DNA 8-oxoG and significantly attenuates their characteristic mutator phenotype (). Mutation and microsatellite instability analysis indicated that a significant fraction of the oxidation-related mutations that were subject to correction by MMR occurred at A:T base pairs (). In particular, AT->TA, AT->GC mutations and frameshifts in runs of As were all affected. Since hMTH1 acts on both 2-OH-dATP and 8-oxodGTP (), its expression could influence mutation by either of the oxidized purines, suggesting that DNA 2-OH-A might make a significant contribution to the mutational burden.
No structural information for incorrect base pairs involving 2-OH-A are available, however thermodynamic analysis showed that 2-OH-A forms stable base pairs with T, C and G, and, to a lesser extent with A (,,). Moreover, the presence of the 2-hydroxy and 1,2-dihydro-2-oxo tautomers, and the possible presence of the and conformers, may lead to various types of base pairs opposite 2-OH-A. Accordingly, when challenged with a 2-OH-A lesion on the template, DNA polymerases (DNA pols), beside correctly incorporating T, often misinsert G and C nucleotides, with various efficiencies depending upon the sequence context. So far only few DNA pols have been studied in detail with 2-OH-A (,). It has been shown that the replicative enzyme DNA pol α has greatly reduced incorporation efficiency opposite a 2-OH-A, suggesting that 2-OH-A, contrary to 8-oxo-G, might constitute a block for DNA replication, requiring some specialized DNA pol to be bypassed (). Among translesion synthesis (TLS) DNA pols of the Y family, data are available only for human DNA pol η and for the archaeal enzyme Dpo4, a homologue to human DNA pol κ. Bypass of 2-OH-A by both enzymes was mutagenic, leading to AT->GC transitions and to AT->TA transversions, respectively (). Finally, the major DNA repair enzyme DNA pol β of family X, was shown to catalyse significant error-prone TLS in the presence of 2-OH-A ().
DNA pol λ belongs to the DNA pol family X, together with DNA pol β, μ and TdT (). We have recently shown that DNA pol λ is very efficient in performing error-free TLS past an 8-oxo-G, and its fidelity and efficiency is enhanced several-fold by the auxiliary proteins PCNA and RP-A, both for normal and translesion synthesis (). These results led us to hypothesize that DNA pol λ might be the principal enzyme involved in error-free bypass of oxidized bases. In this work we have analysed the 2-OH-A bypass in either a random sequence or in a A-run, in the presence of DNA pol λ. Polypurine tracts have been shown to reduce the intrinsic fidelity of DNA pol λ () as well as to constitute hotspots for oxidative lesions and genomic instability (). We therefore, assessed the fidelity of the 2-OH-A bypass by DNA pol λ in this highly mutagenic context. Our results suggested that DNA pol λ can perform error-free bypass of 2-OH-A. A specific role of the DNA pol λ active site residue Tyr 505 in determining nucleotide selectivity opposite the lesion was found. Finally, PCNA and RP-A specifically enhanced the fidelity of TLS by DNA pol λ even on ‘difficult’ sequence contexts such as A-runs.
A*-31mer 5′GCAAAGAACTTATAG3′
6A*-36mer 5′GCAAAGAACTTATAGAA3′
Recombinant human DNA pol β was from Trevigen; recombinant human his-tagged DNA pol λ wild type and the Tyr505Ala mutant were cloned, expressed and purified as described (). Recombinant human PCNA and human RP-A were expressed and purified as described (). After purification, all the proteins were >90% homogenous, as judged by SDS–PAGE and Coomassie staining.
For denaturing gel analysis of DNA synthesis products, the reaction mixtures contained 50 mM Tris–HCl (pH 7.0), 0.25 mg/ml BSA, 1 mM DTT and 20 nM (0.2 pmol of 3′ OH ends) of the 6-FAM-labelled primer/template (unless otherwise stated). Concentrations of DNA pol λ, PCNA, RP-A, dNTPs and Mg were as indicated in the Figure Legends. Reactions were incubated 10 min at 37°C and then stopped by addition of standard denaturing gel loading buffer (95% formamide, 10 mM EDTA, xylene cyanol and bromophenol blue), heated at 95°C for 3 min and loaded on a 7 M urea/14% polyacrylamide gel.
Experiments were performed under the conditions described earlier using the following nucleotide concentrations (dATP or dGTP or dTTP): 0.1, 0.25, 0.5, 2, 5, 10, 25, 100 µM, respectively. Data points were derived from the analysis of the intensities of the products bands. The values of integrated gel band intensities in dependence of the nucleotide substrate concentrations ([dNTP]) were fitted to the Equation ():
Before being inserted in Equation (), the intensities of the single bands of interest were first normalized by dividing for the total intensity of the lane. This reduced the variability due to manual gel loading. An empty portion of the gel was scanned and the resulting value was subtracted as background.
values were derived from the relationship: = [E], where E is the input enzyme concentration.
Nucleotide incorporation efficiencies were defined as the / ratio. Under single nucleotide incorporation conditions = /( + ) and = /( + ), where is the true polymerization rate, is the dissociation rate of the enzyme–primer complex and is the true Michelis constant for nucleotide binding. Thus, / values are equal to /.
The goodness of fit of the interpolated curve was assessed by computer-aided calculation of the sum of squares of errors SSE and the correlation coefficient 2. Interpolation, SSE, 2 and standard errors determination were done with the computer program GraphPadPrism.
DNA pol λ was tested on the A*31-mer template containing a single 2-OH-A lesion in the presence of 1 mM Mg++. As shown in A, DNA pol λ exclusively incorporated dTTP opposite the lesion. Titration of dTTP in the presence of Mg++, indicated that DNA pol λ was less efficient in incorporation in front of the 2-OH-A lesion than opposite a normal A (B, compare lanes 3–6 with lanes 7–10). In order to further investigate this effect, dNTPs titrations were performed in the presence of a higher enzyme concentration. As shown in C, there was a clear reduction in the elongation efficiency by DNA pol λ with the damaged template (lanes 6 - 9) with respect to the control (lanes 2–5). The major DNA repair enzyme, DNA pol β, which is closely related to DNA pol λ, catalysed high dGTP misincorporation in front of the lesion (D, lane 8). Interestingly, the lesion did not affect the elongation at all by DNA pol β in the presence of all four dNTPs (D, compare lane 2 with lane 7). Also the prototypic Y-family enzyme DNA pol η, showed significant misincorporation of dGTP opposite the lesion (E, lane 3). In contrast, however, under the same conditions, DNA pol λ did not show misincorporation (A). Thus, DNA pol λ can catalysed the error-free dTTP incorporation opposite 2-OH-A, albeit with reduced overall efficiency. On the opposite, DNA pol β catalytic efficiency was not affected by the lesion, but the bypass reaction was error-prone due to the significant dGTP misincorporation.
Next, the 2-OH-A lesion was placed in the middle of a six A run, a sequence context known to be extremely mutagenic () and (). On this 6A*36-mer template, in the presence of 1 mM Mg++, DNA pol λ misincorporated dGTP both in front of the lesion and in front of an undamaged template (A, lanes 3 and 8). In addition, slippage products of +4 and +5 were observed 6A*36-mer templates (lanes 6 and 11). dGTP titration experiments, however, indicated that the efficiency of dGTP misincorporation was comparable on both the control (B, lanes 2–5) and damaged (B, lanes 7–10) templates. Thus, this sequence context could induce dGTP misincorporation by DNA pol λ also in the presence of the lesion. Moreover, a reduced elongation efficiency of damaged versus undamaged template was observed on this template (B, compare lane 1 with lane 6). The relative incorporation efficiencies for DNA pol λ in the different sequence contexts are summarized in .
The data so far indicated that DNA pol λ was intrinsically able to perform error-free dTTP incorporation opposite 2-OH-A, but at the expense of a reduced catalytic efficiency. Moreover, when the lesion was placed in an A-run, the overall fidelity of DNA pol λ decreased, leading to significant error-prone translesion synthesis. We have previously shown that the auxiliary proteins PCNA and RP-A are able to increase DNA pol λ catalytic efficiency, fidelity and translesion synthesis (). Thus, PCNA and RP-A effects on the 2-OH-A bypass by DNA pol λ were investigated in the highly mutagenic sequence context of the 6A*36-mer template. As shown in A, RP-A alone was able to inhibit dGTP incorporation opposite 2-OH-A by DNA pol λ, albeit only partially (compare lanes 1, 2 with lanes 5, 6), whereas dTTP incorporation was not affected (lanes 7–12). When tested alone, PCNA showed a similar effect (B), with a slight inhibition of dGTP incorporation (compare lane 1 with lanes 2–4). Most interestingly, however, when PCNA was tested together with RP-A, a strong reduction of error-prone dGTP incorporation (B, lanes 5 and 6), but not of the faithful dTTP incorporation (C, lanes 2–6) opposite 2-OH-A was observed. Titration experiments confirmed the high selectivity for dTTP (D, lanes 1–4) versus dGTP (D, lanes 5–8) incorporation opposite the lesion in the presence of PCNA and RP-A. As summarized in , in the presence of PCNA and RP-A, dTTP incorporation efficiency (/) opposite 2-OH-A was increased 2.2-fold, whereas dGTP incorporation was reduced 3.2-fold. As a consequence the bias for dTTP versus dGTP incorporation by DNA pol λ on the 6A*36-mer template raised from 23.1 in the absence, to 166 in the presence of PCNA and RP-A (). Elongation past the lesion by DNA pol λ in the presence of all four dNTPs was also enhanced by PCNA and RP-A (D, compare lanes 1, 2 with lanes 6, 7), resulting in a 3.3-fold increase in the corresponding catalytic efficiency (/, see ).
We have previously shown that the DNA pol λ residue Tyr505 is important for nucleotide discrimination (), being involved in interactions with the incoming dNTP. We therefore tested the mutant Tyr505Ala in the presence of the 2-OH-A lesion. As shown in A, the Tyr505Ala mutant with the A*31-mer template showed both dGTP and dATP misincorporation opposite the lesion (lanes 8 and 10). When the same enzyme was tested with the damaged 6A*36-mer template (B), dGTP and dATP misincorporations were detected, on both the control and damaged template (B, lanes 3, 5 and 8, 10). In addition, on this template slippage products were also observed both in the absence and in the presence of the lesion (B, lanes 6 and 11). As shown in C, dATP misincorporation could be detected exclusively in front of the lesion with the A*31-mer template (lanes 9 and 10). Similar experiments on the same template were carried out for dGTP misincorporation. As shown in D, again dGTP misincorporation could be only observed in front of the lesion by the Tyr505Ala (lanes 10 and 11). These data indicated that misincorporation was not due to a reduction in the overall fidelity of DNA pol λ by the Tyr505Ala mutation, since error-free incorporation was observed on the undamaged templates in both cases. Under these conditions, wild-type DNA pol λ did not misincorporate either dGTP or dATP (A). Thus, the Tyr505Ala mutant showed significantly reduced fidelity specifically for 2-OH-A bypass on the A*31-mer template, indicating a key role for this residue in determining nucleotide selection for incorporation opposite this lesion, suggesting an important role of this active site residue in 2-OH-A translesion synthesis.
The presence of the oxidized base 8-oxo-G in the replicating strand has been shown to easily misdirect nucleotide incorporation by replicative DNA pols, and frequent misincorporation of A opposite 8-oxo-G is likely to occur (). Another oxidized base, 2-OH-A, is also a potential source of errors by different DNA pols, with A and G being the most frequently misincorporated nucleotides (,). A major difference between the two oxidized purines is however that 8-oxo-G lesion does not constitute a block for replicative (or repair) DNA pols, while 2-OH-A causes a serious reduction in the catalytic efficiency of several DNA pols, including the replicative enzyme DNA pol α (). Thus, this lesion might sensibly slow down replication fork progression and this might constitute the signal to recruit a specialized TLS DNA pol (). Recent data, also from our laboratories, suggested that the most accurate DNA pols in dealing with 8-oxo-G are DNA pol λ and DNA pol η, in combination with the auxiliary proteins PCNA and RP-A (,). Here, our results strongly suggested that DNA pol λ might also be a prime candidate in correctly coping the 2-OH-A lesion. DNA pol λ is intrinsically able to perform error-free TLS past a 2-OH-A lesion, although with a slight reduction in catalytic efficiency. This ability is not shared by other DNA pols and 2-OH-A produced high error rates when replicated by DNA pols belonging to the B or Y-family.
The structural information on 2-OH-A-containing duplexes indicates that 2-OH-A can assume or tautomeric forms in variable proportion depending on solvent polarity and neighbouring bases (). Thus the polarity of the microenvironment within the active site of various polymerases might influence the tautomeric equilibrium of 2-OH-A (). In particular the fraction of 2-OH-A enol tautomers, which closely resembles A, is favoured by a decrease in solvent polarity. The active site of DNA pol λ is known to assume a closed conformation, even prior to dNTP binding (,), and this less polar microenvironment might increase the enol fraction of 2-OH-A. In addition DNA pol λ shows limited interactions with the template strand and strict geometric requirements confined to the terminal and nascent base pairs. These factors together with the preferred enolic form of 2-OH-A might be responsible for an equivalent use of the oxidized and unmodified A in the pairing with dTTP in a random sequence.
In an attempt to clarify the molecular basis for the high selectivity shown by DNA pol λ for the T:2-OH-A base pair we have analysed the contribution of the critical residue Tyr 505, present in the DNA pol λ nucleotide-binding pocket. Kinetic and structural studies have shown that this residue Tyr 505 plays an important role in dNTP binding (). In the DNA-DNA pol λ binary complex, Tyr 505 obstructs the nucleotide-binding pocket through an hydrogen bond with the N1 of a templating adenine (,). In the precatalytic ternary complex, dNTP binding causes a rotation of the Tyr505 side chain, which repositions itself in order to take interactions with the base at the primer terminus. Such an interaction plays a crucial role in nucleotide selectivity, allowing the enzyme to assess proper base pair geometry. We have also shown that Tyr505 acts as a ‘steric gate’ checking the correct size of the nascent base pair (,). In this study, we showed that mutant DNA pol λ carrying an Ala instead of a Tyr at position 505, was able to misincorporate both dATP and dGTP opposite a 2-OH-A lesion, at variance with the wild type enzyme, which exclusively incorporated dTTP. These results suggest that the Tyr505 residue in DNA pol λ is the major determinant preventing 2-OH-A-mediated misincorporation. Thus it is possible that the shorter side chain of Ala does not block 2-OH-A in the suitable position for correct base pairing and is unable to sense the altered geometry of the nascent A:2-OH-A or G:2-OH-A mispairs.
On particular sequence contexts such as A-runs a reduction in DNA pol λ fidelity has been already reported (). The presence of 2-OH-A in a repeated sequence does not further worsen the error rate of DNA pol λ. However, due to the highly mutagenic properties of the A-run, significant dGTP misincorporation could be observed by DNA pol λ alone opposite the lesion. The auxiliary proteins PCNA and RP-A efficiently remove these drawbacks, ensuring efficient error-free TLS of a 2-OH-A lesion even in such a ‘difficult’ sequence context. In summary, our findings indicated that DNA pol λ couples flexible substrate recognition properties (being able to bypass a 2-OH-A lesion) to high intrinsic fidelity in TLS, and PCNA and RP-A auxiliary proteins enhance these properties. Both these proteins play essential roles in DNA replication and DNA repair. Moreover, while RP-A is an important ‘sensor’ protein for DNA replication fork stalling (,), PCNA has been proposed to coordinate the switching from replicative DNA pols to specialized TLS enzymes (). DNA pol λ interacts both functionally and physically with PCNA and its activity is modulated in various way by RP-A. DNA pol λ has been implicated in various pathways such as abasic site bypass, base excision repair () and non-homologous end joining (,). The present data show, for the first time, that the 2-OH-A lesion can be efficiently and faithfully bypassed by human DNA pol λ in combination with PCNA and RP-A, suggesting a role of this enzyme in multiple pathways, including the translesion bypass of 2-OH-A. |
One-third of humans worldwide are infected with the latent form of tuberculosis (TB), and almost two million people die each year from the deadly disease. To be such a successful pathogen, (MTB) adapts to myriad stresses at each stage of infection. Challenges to MTB's survival include reactive oxygen and nitrogen species of activated macrophages, low pH, hypoxia, anti-microbial peptides and starvation for essential nutrients. In addition, bacteria expelled from the host are often challenged with exposure to UV light, dehydration, starvation and low temperature. Because MTB survives most or all of these challenges by transcriptional regulation (), understanding transcription extends our ability to disrupt MTB's life cycle.
Studies of MTB to date suggest that transcription is as complex and varied as it is in other prokaryotes. For example, the genome shows approximately 190 putative transcriptional regulators. Even in the best studied class, MTB's 13 different sigma factors, complexity and unanswered questions are common (,) to date only five (SigA, SigC, SigE, SigF and SigH) have a defined putative promoter consensus sequence (). In addition, the genome contains at least five anti-sigma factors, each of which conduct post-translational regulation of one or more sigma factors, and seven genes encoding anti-anti-sigma factors. Furthermore, the examples of transcripts studied to date demonstrate that even within a single well-characterized operon, transcriptional regulation can be complex in MTB: alternative internal promoters and competing promoters on the opposite strand have been identified (); single genes may be regulated by multiple promoters (); and, as in other prokaryotes, supercoiling plays a role in gene expression (). Finally, little is known about transcription termination in MTB; a recent study found that transcriptional terminators could be found with only 15% sensitivity between opposite-strand genes, one of the worst rates for 96 species examined ().
Other than genes themselves, operons are the most basic unit of organization in bacterial genomes, and they provide the basis for understanding transcriptional regulation and the entire regulatory network of an organism. A search of the MTB literature, however, reveals relatively few well-defined operons (). The prediction of a complete operon map of the MTB genome would be a major milestone in understanding this important pathogen.
Continual development of computational methods for operon prediction in bacteria has been underway in recent years, primarily in . For example, Ermolaeva . () examined conserved gene groupings and proximity over a large set of complete prokaryotic genomes. Although fruitful, such predictions can only be made on the fraction of gene pairs that are conserved across clusters. Other researchers focused on knowledge of within-species gene organizations and characteristics. One widely used approach relies on the well-established notion that the likelihood that adjacent genes are transcribed in the same operon increases as the number of base pairs separating the two genes decreases. For example, Salgado . () used log-likelihoods based on distance to compare adjacent genes in an operon to those not in an operon, later enhancing their predictions with information about functional classes. Romero and Karp () used the methods of Salgado . as a starting point, improving their predictions by using information contained in pathway-genome databases at BioCyc. These authors attempt to enhance distance-based methods with information about metabolic pathways, protein complexes and transporters, an approach grounded in the observation that genes in the same operon often work together in pathways, processes or multimeric protein(s).
Sabatti . () added information from potentially large sets of microarray experiments to inter-gene spacing and directionality. The incorporation of microarray expression data in operon prediction, when available, is natural, since we would expect genes in an operon to be similarly expressed across a variety of conditions. In the absence of measurement error, internal secondary promoters, differential RNA stability or variation in RNA polymerase processivity (that is to say, in an imaginary, idealized biology), genes in an operon would always be perfectly coexpressed. A second group, Bockhorst . (), also successfully used intergenic distance and expression data to predict operons in ; they followed a Bayesian network approach. The synergistic power of intergenic distance and microarray coexpression to predict operons was further confirmed when De Hoon . () used these predictors to accurately build an operon map for .
Two of the methods described above, after being developed and evaluated in , have been used to predict operons in MTB. Using cross-species conservation of gene proximity, Ermolaeva . () offer predicted operon pairs for MTB at The Institute for Genome Research (TIGR) (). As of this writing, this site explicitly reports data for only a third of the available genome (1389 of 3999 potential operon pairs). When we use the 55 known, laboratory-verified operon pairs that we have identified (see ), we find that the TIGR cross-species comparative method correctly predicts that 26 of the 55 pairs would be cotranscribed. Of these 26, 16 are explicitly listed as being cotranscribed, the remaining can be found to be associated as subsets of larger predicted operons. No information regarding cotranscription is reported at TIGR for the remaining 29 of the 55 known operon pairs in MTB.
Romero and Karp () use data in their BioCyc database related to each gene's pathway, complexes and functional class (along with intergenic distance) to predict 2509 transcriptional units in MTB (). The accuracy of Romero and Karp's predictions in MTB itself was not verified. As of this writing, we found that, among the 55 operon pairs known to us, BioCyc correctly predicts 84% (36 of 43) of published operon pairs, but only 33% (4 of 12) of those not yet published (but which were confirmed in our laboratory; ). No assessment of the false positive rate in the BioCyc MTB data set is available. Reduced performance outside of well-annotated areas of the genome was predicted by Romero and Karp (), and likely results from the reduced pathway, complex and functional class data in these regions. Overall, the method's diminished performance outside encouraged Romero and Karp to advocate for building each model with species-specific known transcriptional data. They also point out that supplementing their model with expression data (like Sabatti .) might further enhance operon prediction ().
The predicted MTB operon maps on TIGR and BioCyc each offer valuable predictions, yet both have limitations in scope, power and performance. In this article, we show that methods which use gene expression data from microarray experiments substantially improve operon predictions in MTB, just as they did in (,), and (). Unlike the BioCyc and TIGR operon maps, our method uses data available for all areas of the genome (distance and coexpression) and works equally well in uncharacterized operons, where prediction is most important. We describe our work to organize substantial expression data (474 microarray experiments), confirm new operon pairs in the laboratory, build a predictive model for operon pairs based on intergenic distance and coexpression data, and evaluate model performance. Our best model achieves a true positive rate of over 90% at a false positive rate of 9.1%.
All microarray experiments were performed using standard protocols as previously described (). Previously published details of sample preparation for those microarray experiments can be found in referenced studies listed in . The expression values were downloaded from a local version of Michael Eisen's AMAD database () (). These experiments came in nine general experimental categories (see ), and were performed using either amplicon or oligonucleotide technology.
In order to calculate the desired gene expression correlations from our data set, we employed several data cleaning techniques. After extracting the background-adjusted intensities for each channel from the AMAD database for the 474 currently held microarray experiments, we used the R () package (version 1.0-5) to fill in values for missing data (). For any missing values for a gene, the function . finds the -nearest neighbor genes, referencing only experiments for which the original gene has data. Once these -nearest neighbors are found, the average expression of these neighbors for the experiment which the original gene has a missing value is used to fill in a reasonable approximation for that value. We settled on a -value of 30 (higher than the default -value of 10), since -values below 30 produced too many missing values and forced the imputed value to be based on the experiment average (rather than the more desirable average from the -nearest neighbors).
The expression correlation between gene pairs was based on the natural logarithm of the ratio of the normalized two-channel intensities across a set of experiments. Unique correlations for each gene pair were calculated across all experiments and across each of twelve subsets of experiments determined by experimental condition and microarray technology (amplicon or oligonucleotide; see ).
RT-PCR was used to test whether adjacent genes that represent potential operon pairs (POPs) in fact co-occur on a single cellular RNA molecule. MTB RNA samples were extracted, from logarithmic stage grown cells, and purified by Trizol (Invitrogen) extraction, DNAase treatment and further purified by using an RNA-Easy kit (Qiagen) as described (). To further purify the RNA and reduce the risk of contaminating genomic DNA, RNA samples were then purified with a second round of DNAase (Qiagen #79254) treatment followed by two consecutive RNA cleanup treatments using ‘RNeasy MinElute’ silica-membrane columns (Qiagen #74204). This RNA was then incubated with random primers (Promega #C1181) and reverse transcriptase (Promega #M5101) to create a cDNA pool. To control for the possibility of accidental PCR amplification from genomic DNA not eliminated by the multiple rounds of purification, an RT-control was prepared in parallel, which differed only by the absence of reverse transcriptase. To maintain buffer continuity between RT-PCR reactions, reverse-transcription 10× buffer (Promega #A3561) was used in both reverse transcription and PCR reactions. For each POP to be tested, MIT's Primer3 software was used to design primers anchored in the open reading frames of adjacent genes. Typically such primers were separated by 200–400 bp, and had annealing temperatures of 55–58°C. (See Supplementary Data for primer details). For each primer pair, the RT+ and RT− samples described above were used as templates in PCR reactions via standard methods. Primer pairs which produced a correctly sized product from the RT + PCR reactions but not the RT− control PCR reactions were interpreted as evidence that an RNA bridging the gene pair was present in the MTB mRNA pool. Such results were then confirmed from an independently prepared sample of MTB RNA.
The development of a predictive model for operon pairs in MTB requires the definition of a training set of known operon (OP) and non-operon (NOP) gene pairs. Evidence of 43 operon pairs has been confirmed in laboratories and subsequently published (). To build this set further, we have confirmed 12 new operon pairs () using RT-PCR.
The piloting work using microarrays for operon prediction in () and () used thousands of well-characterized, same-strand non-operon pairs to build predictive models. Unfortunately, most or all other organisms that could currently benefit from operon prediction (including MTB) have much more limited experimental verification; this is especially true for non-operon pairs. To overcome this potential roadblock to prediction, we use the 1340 pairs of consecutive genes on opposite strands of DNA as our NOP set. This substitution is supported by recent work (,), which finds comparable microarray expression between two types (same-strand versus opposite-strand) of non-operon gene pairs. However, as described in Price . (), the distribution of intergenic distance in same-strand and opposite-strand NOPs should differ; distances for same-strand NOPs should be greater in general. One consequence is that our predictive model using expression and distance will be somewhat conservative and underpowered—it is more difficult to distinguish OPs from NOPs using opposite-strand NOPs since their distance distribution is less distinguishable from the distance distribution of OPs. Thus, if our predictive model performs well using our definition of NOP to create a training set, we have reason to believe the model will perform even better when predicting operon status of same-strand gene pairs. Any loss of power from using opposite-strand gene pairs as NOPs in our training set will be more than offset by the additional power from the large number of available NOPs under this definition, especially when compared to the small number of laboratory-validated same-strand NOPs in MTB. In addition, investigations of conserved ‘known’ NOPs in () and () found evidence that many of them are indeed cotranscribed.
All sequence information on the MTB genome was obtained from the online database TubercuList, (), based on the H37Rv strain. Intergenic distance between two genes was found from the database by subtracting the ending location on the genome of the first gene from the initial location on the genome of the second gene.
We constructed a statistical model for predicting operon status of each potential operon pair (POP) based on intergenic distance and expression correlation. We based our predictive models on logistic regression with the logit link function.
and the model equation is
cyanide, hypoxia and nitric oxide) and microarray technology (oligo or amplicon). Using the known OPs and the NOPs as our training set, estimates of the coefficients of the model are found through the iteratively reweighted least squares technique.
To place these predictive probabilities on a more intuitive scale, we use the values to assign a ‘cotranscription rank percentile (CRP)’ for each gene pair by simply sorting all operon pairs from highest predicted probability of being an operon pair (CRP = 100) to lowest (CRP = 0) based on the model. Then, we select an appropriate CRP threshold to achieve a specified modeling goal (e.g. achieving a certain sensitivity, achieving a certain specificity, classifying a certain percentage of gene pairs as operon pairs). Based on the chosen CRP threshold, each POP in the entire genome can be classified as being either an operon pair or a non-operon pair.
Model performance is assessed and compared using several metrics. The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) both balance goodness-of-fit of a model with model complexity, where the BIC imposes a more severe penalty on extra parameters than the AIC (). Kendall's tau-a assesses the difference between concordant and discordant sets out of all possible sets of gene pairs, where a concordant set of gene pairs is one in which the pair that is an operon has a higher fitted probability of being an operon pair than the pair that is not an operon (). The c index provides the area under the receiver operating characteristic (ROC) curve, where the ROC curve provides a graphical representation of the trade off between the false negative and false positive rates for every possible cutoff in predictive probabilities (). In particular, the c index is then the percentage of all possible pairs of cases in which the model assigns a higher predictive probability to a correct case than an incorrect case ().
Finally, we compared models using cross-validated measures of overall model accuracy—the proportion of all gene pairs of known operon status which were correctly classified. We chose our classification threshold to provide a true positive rate in our final model of at least 90% while minimizing the false positive rate. Estimates of overall model accuracy were determined with 10-fold cross-validation using the function in R. This function was also used to obtain cross-validated estimates of specificity and sensitivity for our final model.
Preliminary data exploration verified a pattern of small intergenic distance and high expression correlation being associated with operon pairs in our data. A shows that the distribution of intergenic distances for operon pairs (OPs) is centered tightly near zero, whereas the distribution of intergenic distances for non-operon pairs (NOPs) is spread out with a median at 83 bp. This supports the hypothesis that intergenic distance can be used to separate OPs from NOPs in MTB as it has been used in other organisms.
Evidence that expression correlation distinguishes OPs from NOPs was also found. B presents densities of gene expression correlations across all experiments. The mean and median expression correlations for OPs are 0.60 and 0.66, respectively, and 0.16 and 0.14 for NOPs.
To examine the relationship between expression correlation and intergenic distance, we generated a scatterplot (C) with different colors distinguishing known operon pairs from potential operon pairs. We see that operon pairs are generally characterized by short intergenic distance and high correlation of expression. In fact, among POPs, most examples of strong coexpression have relatively short intergenic distance.
The logistic regression predictive model for a particular potential operon pair (POP) is based on available data for that POP. For this work, we have chosen to use the most recent annotation of MTB (), but some microarray experiments predate this reannotation. Because of this, amplicon microarray experiment data is unavailable for 76 genes, and data from more recent oligo experiments was missing for 37 of those 76 genes. This missing data translated into a slightly larger number of missing gene expression correlations since each individual gene is part of two gene pairs. To use as much information as possible for each gene pair, we constructed three different predictive models based on available data—predictions for 2572 POPs use intergenic distance and expression correlations from both oligo and amplicon experiments, 42 use intergenic distance and expression correlations from oligo experiments and the remaining 45 POPs use intergenic distance only.
Since we had expression data from a variety of experiments, both in terms of treatments (cyanide, hypoxia, nitric oxide, etc.) and microarray technology (oligo or amplicon), we explored the possibility that separating the expression data by experiment type would provide more predictive power of the true operons. In addition to the varying experimental conditions, not all the experiments were done at the same time or with the same microarray platform and thus might vary in quality. Instead of a single correlation of coexpression across all experiments, we calculated separate coexpression correlations within each of the 12 experimental types () and looked for evidence that model performance was improved.
Thus, for the 2572 gene pairs for which distance, oligo and amplicon data were available, final predictors in our logistic regression model were chosen from among intergenic distance and correlations of coexpression within each of the twelve experimental types. Insignificant predictors were removed through backward elimination methods. (Model A) shows the estimated coefficients, SEs and significance tests for our final logistic regression model for the POPs with complete data. Five of the twelve experimental types proved significantly helpful in distinguishing OPs from NOPs.
In , we compare the model in A (Model A) with several other models, including a model with gene expression correlation for all experiments as the only correlation predictor (Model G); a model with two correlation predictors, one for oligo microarrays and the other for amplicon microarrays (Model F); and the full model with gene expression correlations for all 12 experiment types (Model H). Several statistics summarizing model performance are presented, including c index (the area under the ROC curve), Kendall's tau-a, AIC and BIC. Note that lower values of the AIC and BIC indicate superior model performance. For each measure except BIC, Model A from was preferable to every other model except the full model. With BIC, Model A was the best performing model of all, which reflects our model building efforts to optimize model fit while minimizing unnecessary complexity. A few results in are especially pertinent. For instance, we have quantitative evidence that (i) adding coexpression data to intergenic distance clearly improves model performance (Model C versus all other models); (ii) adding data from amplicon microarray experiments improves upon models with intergenic distance and data from oligo experiments (Model A versus Model B; Model D versus Model F; Model E versus Model H) and (iii) using separate coexpression correlations from each experiment type leads to improved predictive performance when compared with models using a single term for coexpression across all experiments (Model A versus Model F and Model G).
In order to accurately predict operons in gene pairs for which amplicon data was not available (42 POPs with distance and oligo data), a new logistic regression model was developed (Model B). As in Model A, backward elimination methods were used to choose a final set of predictors from among intergenic distance and correlations of coexpression within the four experimental subgroups using oligo microarrays. The same training set of operons used in Model A was used to build Model B. B shows the estimated coefficients, SEs and significance tests for our final logistic regression model which should be used for prediction for the POPs with distance and oligo data only. It is not surprising that those oligo experiments which were found significant in the primary model (A) were also found significant in this model.
For the 45 POPs with no expression data at all, we developed a logistic regression model on the same training set of operons using only intergenic distance as a predictor. This model (Model C) appears in C. As expected, intergenic distance remains a significant predictor of operon status.
To illustrate these model comparisons, we created ROC curves, plotting false positive rates versus true positive rates. The strongest predictive models maximize the true positive rate while minimizing the false positive rate; in other words, they have maximum proximity to the upper left-hand corner of the graph. For instance, the ROC curves in show that the distance only model (Model C; dotted line) is inferior to the models with distance and expression data (solid and dashed). In addition, the model with expression data from both amplicon and oligo experiments (Model A; solid line) performed significantly better than the model with expression data from only oligo experiments (Model B; dashed line).
demonstrates that expression data can be used as a powerful tool for operon prediction in MTB; our current model (, Models A–C) classifies 39.3% of POPs as operon pairs, and achieves a true positive rate of 90.8% at a false positive rate of 9.1%. These rates are comparable with some of the best published to date, even in . For example, Sabatti ., who piloted the use of expression data and distance to predict operons in , report a true positive rate of 88% at a false positive rate of 12% (). Romero and Karp, whose method does not use expression data for but leverages the exhaustive knowledge of pathways and complexes in this best characterized of prokaryotes, report a true positive rate of 91% with a false positive rate of 13% (). De Hoon and colleagues, who (like us) extend Sabatti's methods to another genome, achieve a true positive rate of 88.8% with a false positive rate of 12.1% in (). Our model's performance compares even more favorably with the first generation of MTB operon predictions available at TIGR (which finds 47% of known operon gene pairs) and BioCyc (which finds 84% of published pairs but only 33% of unpublished ones). Unlike the BioCyc and TIGR operon maps, our method uses data available for all areas of the genome (distance and coexpression) and should work equally well in uncharacterized operons, where prediction is most important. In addition, since we used opposite-strand non-operon pairs in our training data set, the model performance we observed should improve as laboratory validated, same-strand non-operon pairs (whose intergenic distance distributions are more distinguishable from that of operon pairs) are used for training.
Finally, our ongoing work to validate model predictions and expand our training data with additional laboratory-identified pairs, though preliminary, also supports the value of our model ().
We have predicted the operon structure of MTB using intergenic distance, microarray expression data and information about the conditions of the microarray experiments. A predictive logistic regression model based on these inputs outperformed alternative models without expression data or even with all expression data condensed into a single correlation term.
Our predictive model produced a predictive probability that each potential operon pair in the MTB genome is truly an operon pair, and we transformed those predictive probabilities into cotranscription rank percentiles (CRPs). In order to simplify presentation and interpretation, we converted these continuous outputs into a discrete set of classifications by selecting a threshold for determining whether or not a potential operon pair should be classified as an operon. Although this binary classification knowingly oversimplifies the underlying biology (e.g. ignoring growing evidence of internal promoters, alternative transcriptional start sites and internal readthrough terminators), it is useful because it makes assembly of a genome-wide map of operon structure straightforward. In our case, we defined a CRP threshold to produce a true positive rate of at least 90% while minimizing the false positive rate. Our model performance—cross-validated sensitivity of 90.8% with specificity of 90.9%—compares favorably with models developed in the much better characterized ‘model organisms’ in () and (). By applying this 90% true positive threshold to the data, it is straightforward to generate a complete list of predicted operons (see Supplementary Data).
We have also made our full data set available to researchers who may want to further explore model building and performance (see Supplementary Data). The data contains all model inputs and outputs which we considered—intergenic distance, coexpression correlations by experiment type, predictive probabilities and cotranscription rank percentiles—for all 3999 gene pairs in MTB. While we believe the models and predicted cotranscription map for the entire MTB genome presented in this article represent a rich and complete view of the currently existing data, we also believe that understanding of MTB can be further enhanced using this data as a base. For example, researchers might wish to examine the impact of different thresholds for classification as an operon pair. In addition, researchers can update this data as new operons or non-operons are confirmed, or they can add additional potential predictors as more is learned about the biology of MTB. The accelerating pace of molecular research in this important pathogen is certain to provide additional data with which to refine the predictions described in this work, forming a solid empirical foundation for our future understanding of MTB transcription.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Restriction-modification (R-M) systems are ubiquitous in bacteria and are involved in protection of the bacterial cell from incoming foreign DNA. R-M systems are comprised of two enzymes: a methyltransferase and a restriction endonuclease. The methyltransferase catalyses the methylation of a specific DNA recognition sequence distinguishing ‘self’ DNA and protecting it from cleavage (,). The restriction endonuclease catalyses the double-stranded cleavage of unmethylated ‘non-self’ DNA. R-M systems are classified into three major groups: types I, II and III (). In type II systems, the most common type of R-M system (), the methyltransferase activity and restriction activity are performed by two independently acting enzymes (). In the more complex type I systems, methyltransferase activity and restriction activity are performed by the same holoenzyme, which consists of three types of subunits: S, M and R encoded by , and , respectively. The S and M subunits are required for methyltransferase activity, and all three subunits are necessary for restriction activity (). Type III systems consist of only two subunits, the methyltransferase (modification, Mod) subunit which can independently function as a methyltransferase (,) and the restriction (Res) subunit which must form a complex with Mod to recognize and cleave DNA ().
Phase variation is the reversible, high frequency switching of gene expression. In and several other mucosal pathogens, phase variation is commonly mediated by mutations in simple tandem DNA repeats in the open reading frame or promoter region of phase variable genes (,). Phase variation is usually associated with genes encoding surface expressed virulence determinants, where switching of expression allows the generation of a diverse population of phenotypically distinct cells—some of which will be better adapted to survival. There are several examples of type III R-M systems in a variety of pathogenic bacteria that have been proven to undergo phase variation [ () and ()] or from sequence analysis would be predicted to undergo phase variation [ (), , () and ()]. Several proposals have been made as to the functional significance of phase variable type III R-M systems but these have not been tested experimentally. Classically, R-M systems are thought to provide resistance to invading bacteriophages or foreign DNA acquired via natural transformation (). In this context phase variation may permit temporary removal of this restriction barrier, allowing the acquisition of foreign, potentially beneficial, DNA molecules (). It has also been suggested that phase variation of methyltransferases may lead to autolytic self-DNA degradation by the cognate restriction enzyme and that such systems may be suicidal (,). This would lead to release of DNA into the environment for uptake by other cells, potentially to the benefit of the population (,). An alternative function for phase variable methyltransferases may be gene regulation, mediated by differential methylation of the genome (). It has previously been established that DNA methylation can affect gene expression in several systems (e.g. Dam methylation) (), however, there were no examples where the methyltransferase affecting gene expression is itself phase variably expressed. Recent work from our laboratory has shown that in strain Rd, a phase variable methyltransferase (Mod) of a type III R-M system coordinates the random switching of expression of multiple genes, and constitutes a phase variable regulon—‘phasevarion’ (). is an important human pathogen causing invasive diseases, such as meningitis and respiratory tract infections. Functional phasevarions have also recently been confirmed in the human pathogens, . and , where a gene also coordinates expression of a regulon of many genes, several of which have a potential role in pathogenesis (Srikhanta , submitted for publication). Since a role in gene regulation is now established in several systems, the question arises of whether during evolution the role of these type III R-M systems has become exclusively gene regulation, or whether their role as DNA restriction systems has been retained.
Fifty-nine isolates were used in this study, as described in of Supplementary Data. Strains included a set of 24 Finnish otitis media isolates that have been used in several previous studies (). The other NTHi isolates have also been described previously (). Encapsulated strains included type strains from the American Type Culture Collection (ATCC) as well as clinical isolates from the A. Smith laboratory and strains from the E. R. Moxon laboratory. For strains that had not previously been typed, multilocus sequence typing (MLST) was carried out as previously described (). Strain details and sequence types were submitted to the public MLST website (haemophilus.mlst.net). An UPGMA (unweighted pair group method using arithmetic mean) dendrogram was constructed using the START2 collection of MLST-related software available at ().
was grown at 37°C in brain heart infusion (BHI) broth supplemented with hemin (10 μg/ml) and NAD (2 μg/ml). BHI plates were prepared with 1% (v/v) agar and supplemented with 10% (v/v) Levinthal base () and when appropriate kanamycin (10 μg/ml) or tetracycline (5 μg/ml).
All enzymes were sourced from New England Biolabs. PCR was performed using primers purchased from Sigma Proligo (Table 2 of Supplementary Data). Primers him6A and him11 were used to amplify the variable region of the gene, primers him1 and him3 or him4 and him5 were used to amplify the repeat tract, and primers HI1059for and HI1052rev were used to amplify across the / region. The region of the gene containing known frameshift mutations was amplified using primers HI1054for and HI1054rev or HI1054for2 and HI1054rev2. Sequencing reactions were prepared using PCR products as template and Big-Dye sequencing kit (Perkin Elmer). Samples were analysed using a 3130xl Capillary Electrophoresis Genetic Analyser (Applied Biosystems International). Data were analysed using MacVector (version 9.0) and DNA Sequencher. To analyse DNA fragment sizes, PCR products, amplified using a primer set in which the forward primer was labelled with 6-carboxyfluorescein (6-FAM), were analysed using the GeneScan system (Applied Biosystems International). Southern hybridization analysis was carried out as described by Sambrook () using a DIG-labelled (Roche) PCR product as a probe. To confirm the presence/absence of the gene, primers him7 and him2 (Table 2, Supplementary Data) were used to PCR amplify the 3′ conserved region of from strain Rd DNA and this PCR product was used as a probe.
Sheared genomic DNA from the Rd:: mutant strain previously described () was used to transform non-typeable isolates R2866 and 162 by the MIV method (). :: transformants were selected on BHI plates containing kanamycin and confirmed by PCR and Southern analysis. The gene was PCR amplified using primers ResF and ResR. The PCR product was cloned into pGEM-Teasy vector (Promega), digested with HindIII and blunted using Klenow polymerase (New England Biolabs). The Tn kanamycin resistance cassette from the pUC4K vector (Pharmacia) was excised using HincII and inserted into the blunt HindIII site. The resulting plasmid, pGEM::, was linearized by digestion with NcoI and used to transform non-typeable isolates 162 by the MIV method. :: transformants were selected on BHI plates containing kanamycin and confirmed by PCR analysis.
Bacteria harvested from five plates of confluent growth on BHI agar supplemented with Levinthal base were resuspended in 20 ml of ice-cold sterile water containing 15% v/v glycerol and 272 mM sucrose (pH 7.4). Bacteria were centrifuged for 2 min at 13 000 r.p.m. and resuspended in 1 ml sterile water (containing sucrose and glycerol). This step was repeated 4–5 times, keeping the cells on ice between spins. The OD of the cell suspension was measured and normalized to an OD 600 of 10. One microgram of DNA was added to the cells that were then incubated on ice for 2 min. Cells were electroporated (Bio-Rad micropulser electroporator, 2.5 kV, 0.2 cm cuvettes) and BHI broth was added immediately. After 90 min incubation at 37°C with shaking, cells were plated on BHI agar containing tetracycline. After overnight growth, single colonies were selected, grown in broth and plasmid prepared using the Qiagen Plasmid Midi Kit (Qiagen, Doncaster, Vic, Au). For the quantitative transformation experiments, numbers of transformants were recorded. Twelve colonies were picked for sequencing of the repeat tract and a sample of the cell population was subject to fragment size analysis using the GeneScan system, as previously described (). The transformation efficiency was calculated as number of colonies/μg DNA. The ratio of ON to OFF cells in the ON recipient cell population was checked pre- and post-transformation by fragment size analysis and confirmed to be unchanged.
Plasmid pHStet was extracted from strain Rd ON and cells, or R2866 ON and cells using the Qiagen Plasmid Midi Kit (Qiagen, Doncaster, Vic, Au). One microgram of each plasmid was digested overnight with ApoI according to manufacturer's instructions and the resulting fragments were separated on a 2% high resolution agarose gel (Nusieve 3:1 Agarose,Cambrex BioScience, Rockland, ME, USA) with TBE at 70 V for 2 h and visualized under UV illumination. Similarly, digests were carried out using TaqI.
Plasmid pHStet was extracted from strain R2866 ON and cells using the Qiagen Plasmid Midi Kit (Qiagen, Doncaster, Vic, Au). Mod ON cells were verified by sequencing the repeat tract. Five micrograms of each plasmid was digested overnight with DpnI according to manufacturer's instructions and the resulting fragments were separated on a 1.5% high resolution agarose gel (Nusieve 3:1 Agarose,Cambrex BioScience, Rockland, ME, USA) with TBE at 70 V for 2 h. The separated DNA fragments were transferred to nitrocellulose membrane (GeneScreen, Perkin Elmer, Rowville, Vic, Au) using overnight capillary transfer with 10× SSC. The DNA was cross-linked to the membrane by exposure to UV light, and then the membrane was washed three times in TBST (100 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5% Tween 20) for 5 min with gentle agitation. The membrane was blocked for 1 h in 3% BSA in TBST with gentle agitation after which it was incubated for 1 h in 10 ml of a 1:1000 dilution of a rabbit anti-N6-methyl-adenine monoclonal antibody (Megabase Research Products, Lincoln, NE, USA) in the above blocking solution. The membrane was washed three times for 5 min in TBST prior to being incubated in a 1:20 000 dilution of a goat anti-rabbit IgG alkaline phosphatase conjugate antibody (Sigma-Adrich, Castle Hill, NSW, Au) in blocking solution for another hour. After another three washes, the membrane was immersed in 10 ml of Sigma FAST BCIP/NBT (Sigma-Aldrich) substrate.
Early work on strain Rf DNA restriction systems characterized the recognition sequence for a type III R-M system (HinfIII) as 5′-CGAAT-3′ (). The strain Rd genome sequence contains a single R-M system with homology to type III systems, encoded by the and genes (). The gene is subject to phase variable expression due to a 5′-AGTC-3′ tetranucleotide repeat tract within the open reading frame (,). To investigate whether the recognition sequence for Mod in strain Rd is the same as that for HinfIII, plasmid pHStet () was grown in strain Rd ON (40 repeats) and strain Rd cells. The resulting methylated or non-methylated plasmids were then digested with an enzyme whose recognition sequence overlaps the proposed methylation sequence. Digestion by ApoI (5′-RAATTY-3′) is known to be inhibited by methylation of either of the adenines of the HinfIII sequence (a). Only one of the multiple ApoI sites in pHStet overlaps with a HinfIII site and would potentially be inhibited by methylation. A 1.3-kb ApoI fragment containing this overlapping HinfIII/ApoI site was observed after digestion of the strain Rd ON derived (methylated) plasmid with ApoI. This band was absent from the strain Rd-derived (non-methylated) plasmid digest (b), indicating inhibition of ApoI digestion by methylation of the DNA by Mod. We conclude that Mod from strain Rd has the same site specificity as HinfIII (5′-CGAAT-3′). Here, we also demonstrate that it is the second adenine in the HinfIII sequence that is methylated. This is demonstrated by digestion of pHStet with TaqI. The TaqI recognition site (5′-TCGA-3′) overlaps with the first adenine of the HinfIII sequence. The TaqI cleavage pattern of pHStet, isolated from Rd ON and cells, is identical, indicating that this base is not methylated by Mod (c). This confirms the findings of a previous study, in which DNA methylated by the HinfIII enzyme, isolated from strain Rf, using -[-H] adenosyl methionine, was cleaved by TaqI ().
An alignment of four gene sequences, from the four strains of for which the genomes have been sequenced and made available, indicated division of the gene sequence into three domains, typical of genes of type III R-M systems (). The N- and C-terminal regions of the protein show between 90 and 96% similarity in sequence amongst the genome-sequenced strains, and are separated by a central domain that is completely dissimilar between strains, showing only 29–31% amino acid sequence similarity. This is consistent with previous findings that Mod proteins are highly conserved in the N- and C-terminal thirds of the protein with relatively low conservation in the central third (,). This central portion of the protein has been proposed to play a role in target sequence recognition and binding, and in binding of the methyl donor (). The conserved regions are proposed to play a role in protein–protein interactions between Mod and Res.
To test whether the variant alleles present amongst genome sequenced strains of encode proteins with the same or distinct DNA recognition sequences, the ApoI restriction inhibition assay described above was carried out on plasmids derived from non-typeable (NTHi) strain R2866 (strain R2866). This strain was chosen since it contains a repeat tract length permissive for expression of the gene and it has a distinct DNA site specificity domain to strain Rd. In this strain background, the ApoI restriction pattern of plasmid pHStet was identical for plasmid derived from strain R2866 ON cells and :: cells (b), indicating that the methylation site specified by strain R2866 Mod, and hence the restriction site, is distinct from that of strain Rd Mod (i.e. not HinfIII; 5′-CGAAT-3′). However to confirm this, it was necessary to demonstrate that Mod is actually active in this strain. Plasmid pHStet isolated from R2866 ON and :: cells was digested; separated DNA fragments were transferred to nitrocellulose membrane and probed with anti-N6-methyladenosine antisera. The antisera bound to all of the separated DNA fragments, indicating that endogenous methylases were methylating many sites within the plasmid. With one of the DNA fragments, differential binding of the antisera to plasmid isolated from R2866 ON and :: cells was observed, indicating that many Mod methylation sites occur within this fragment allowing for an observable difference in antisera binding between the isogenic strains. This difference in antisera binding between wild-type and mutant confirms that Mod is an active methyltransferase in this strain (d).
The sequence heterogeneity observed amongst the genes of the genome sequenced strains of (see above) and our experimental data supporting the correlation between sequence type and recognition site specificity, highlighted the need for a phylogenetic analysis of the gene from a genetically diverse set of strains. Encapsulated strains of are associated with invasive diseases such as meningitis and pneumonia, while acapsular or non-typeable (NTHi) strains are associated with otitis media and respiratory tract infections (). We examined the gene in 59 strains, including 43 NTHi and 16 encapsulated strains. The genetic relatedness of these strains had been determined by multilocus sequence typing (MLST). PCR amplification and DNA sequencing of the central variable region of the gene and the repeat tract indicated significant heterogeneity.
We identified 15 sequence groups in the survey, such that within a group the variable region sequence showed >95% amino acid similarity. Between groups, the percentage sequence similarity of the variable region of the protein ranged from 29 to 38%. During the course of this study, Bayliss () reported sequences of 22 NTHi isolates, constituting a subset of the NTHi collection described above, and proposed groupings. To avoid confusion, we have assigned the same numbers for common groups. The relationship of the 15 sequence groups to a MLST dendrogram is shown in . A correlation is evident between sequence type and capsular serotype. Each strain from a particular capsular serotype shares an identical sequence type to all other members of that serotype, with the exception of one type f strain, ATCC 9833. All capsular type d strains possess a unique sequence type (Group 1). Capsular type e and f strains that contain the gene (two type e strains lacked and ), share a common sequence type that is unique from all other capsular or non-typeable strains (Group 14). Capsular type b strains share a sequence type with a number of NTHi isolates (Group 2). The serotype a strain examined does not possess a gene and the sequence from the one serotype c strain is unlike all other sequences identified (Group 16). The NTHi isolates show significant diversity in their gene sequence, with the NTHi isolates being distributed across 12 sequence groups. This finding is generally consistent with previous reports of the overall genetic diversity of NTHi strains, relative to encapsulated strains (,).
In three capsular strains (ATCC 9006, type a; ATCC 8142, type e; R3368, type e) no and gene could be identified. PCR amplification with primers that bind to the genes flanking and gave a product size consistent with the absence of and . DNA sequencing of this PCR product indicated that only the flanking genes and a small amount of intergenic sequence was present (). No remnant of the or gene was found, indicating that either a clean deletion event had occurred or that these strains had never acquired this R-M system. Southern analysis using a probe against the 3′ conserved region of indicated that these genes were not present elsewhere within the genome (data not shown).
A correlation is evident between repeat tract sequence and capsular serotype. The repeat unit 5′-AGTC-3′ was found in all strains of capsular serotype d, while the repeat unit 5′-AGCC-3′ was found in all other strains containing repeats, including NTHi isolates and strains of capsular serotypes b, c and f. A correlation is also evident between sequence group and the length of the repeat tract. In three of the 15 sequence groups, there is no repeat tract within the gene. Two of these groups are made up of NTHi isolates only (Groups 6 and 7), the other group contains capsular type e and f strains (Group 14). In these strains the DNA sequence 5′-TCAGATAGTCAG-3′ is present in place of the repeat tract. The gene is not predicted to be phase variable in these strains. Of the other twelve sequence groups, some groups contain predominantly low numbers of repeats or no repeats (Groups 3, 4, 5 and 8) and some contain predominantly high numbers of repeats (Groups 1, 2, 9 and 10). It has previously been reported in strain Rd that the length of the repeat tract correlates to the rate of phase variation (). The phase variability of expression and the rate at which this occurs may reflect functional differences amongst different alleles. No correlation was detected in this study between sequence group or repeat tract length and disease phenotype of the corresponding strains ().
Noted also are pathogenic genes that define groups 11, 12 and 13. Previous work by Kroll () reported the genes of the type III R-M system as clear examples of horizontally transferred genes between the pathogens and . In group 13 we show an example of NTHi isolate R3023, which has the same DNA recognition domain sequence as strain FA1090. The four NTHi isolates in group 7 have the same DNA recognition domain sequence as the genome strain (Sanger Institute, UK). These genes clearly have a common origin. Due to the high sequence conservation of in and species, the common DNA repeats mediating phase variation and the common origin via horizontal gene transfer, we propose the following nomenclature. The genes referred to in are hence called , followed by a number corresponding to the grouping based on the DNA recognition site allele, recognizing the key phenotype of DNA recognition specificity (e.g. the nt-R3023 gene is ). This avoids confusion in nomenclature with organisms like pathogenic , which contain multiple genes that are distinct over the whole length of their sequence.
Unlike the significant diversity observed in the sequence amongst strains of , the res sequence is more highly conserved amongst the genome sequence strains of NTHi and strain Rd. The protein encoded by the ORF varies in length amongst these strains from 722 amino acids (aa) in strain 86-028NP to 930 aa in strain Rd. The 930 aa Res protein from strain Rd and the 929 aa Res protein from strains R2866 and R2846 are similar in size to the homologous Res protein from other organisms. Alignment of the nucleotide sequence of the genes from these four strains indicates homology to the full-length gene sequence in all cases although in strain 86-028NP, a single base pair deletion results in a frame-shift mutation truncating 207 aa from the C-terminus of the protein. The Res subunit of type III R-M systems contains several sequence motifs characteristic of DNA and RNA superfamily II helicases (). These conserved motifs include the ATP-binding motif (TGxGKT) (,) and motif II (DEAH or DEPH) (). In addition, a weakly conserved sequence ‘PD … (D/E)XK’ is found at the C-terminus of the Res protein which represents the endonuclease domain or active site, a signature found in several types of nucleases (). This region of the protein is involved in metal binding, a requirement for DNA cleavage by several restriction enzymes (). The truncation at the C-terminus of the strain 86-028NP protein results in loss of the endonuclease domain, and the protein is predicted to be inactive ().
To investigate whether the same 86-028NP single base pair frame shift mutation is found in the gene of other strains, a 1.3-kb region of the gene encompassing the 86-028NP mutation was sequenced in the 56 strains containing and . The single base pair deletion in the gene of 86-028NP was also found in strains nt-723 and nt-1247 ( group 2). Unexpectedly, a different, 32-bp deletion was found in the gene of four strains in the group 2. This mutation introduced a frame-shift resulting in a peptide of only 495 aa,—a more extensive carboxy terminal truncation than the 86-028NP mutation (see ). All serogroup b strains tested contain this mutation, as does the genome sequence serogroup b strain 10810 (Sanger Institute, UK). Conversely, in the gene of strains containing a non-phase variable gene, no obvious frame-shift mutations were observed.
To further investigate restriction activity of the type III R-M systems, we used plasmids specifically methylated or non-methylated by Mod (i.e. isolated from ON or :: cells, respectively) to transform strains Rd and 162. Prior to transformation, the methylation state of plasmid isolated from strain Rd was checked using the ApoI inhibition assay described above and by sequencing of the repeat tract, this was not necessary for strain 162 since is constitutively expressed in this strain. A statistically significant difference in transformation efficiency was shown between Mod methylated plasmids and non-methylated plasmids transformed into the homologous strain ( ON cells of strain Rd or 162) (). The Mod methylated plasmids transformed only two times more efficiently into these strains than non-methylated plasmids. This difference in transformation efficiency is attributable entirely to the activity of the type III R-M system and is independent of other R-M systems, which remain unchanged in the isogenic wild-type/:: strain pairs.
When Mod methylated plasmids were transformed into the heterologous strain (i.e. plasmid isolated from Rd ON cells transformed into 162, and vice versa), a significant difference in transformation efficiency was observed between plasmids methylated in the same strain and plasmids methylated in a heterologous strain (). This significant reduction in transformation efficiency is due to differences in other, non-type III methylation systems between these two strains.
We have previously shown that strain Rd Mod is a phase variably expressed, epigenetic regulator of multiple genes—a phasevarion (). Here we report that the recognition sequence for Mod is 5′-CGAAT-3′, the same as previously reported for HinfIII (). This is a key finding enabling our studies resolving how ON/OFF switching of methylation influences the promoters of genes controlled by the phasevarion of strain Rd. Identification of the strain Rd Mod target site also enabled design of strategies used in this study to address functional and evolutionary questions presented by phase variable type III R-M systems of . Comparisons of the and genes from the genome strains, and our large scale survey of capsulate and NT revealed a wide range of variation in the region of the gene thought to dictate sequence specificity (,). This variation was also observed in the recent NTHi sequence analysis conducted by Bayliss (). Here we used an ApoI inhibition assay to demonstrate that the gene of strain R2866 ( group 10) does not modify the same sequence as of strain Rd ( group 1), confirming these enzymes methylate distinct target sequences. These data provide experimental evidence that sequence variation observed in the putative DNA specificity domain, used to generate groupings in , reflect Mod proteins with distinct target sequences. Consistent with these findings, our recent work in pathogenic has confirmed that phase variation of the ModA11, 12 and 13 methyltransferases (see ) control expression of distinct genes (Srikhanta, submitted for publication).
These findings raise the question of whether the 15 groupings in represent 15 different type III R-M systems or 15 different methyltransferases controlling phasevarions, and whether these functions are mutually exclusive. Sequence analysis revealed three groupings (Groups 6, 7 and 14) in which none of the genes contain tetranucleotide repeat sequences. These genes are therefore not phase variably expressed, and by definition these are not phasevarions. These groups include strain 162, in which we have demonstrated a functional type III R-M system. We propose all these groups, and any other strain with a non-phase variable , are likely to be functional, dedicated type III R-M systems. In contrast, representatives of group all have high numbers of tetranucleotide repeats, consistent with phase variable expression. Analysis of the gene from strains within these groups revealed frame-shift mutations in in many cases that are inconsistent with expression of a functional Res protein. We propose that these strains, and any other strains with tetranucleotide repeats in , and obvious inactivating mutations in , function exclusively as phasevarions. Hence, of the 41 strains in the survey that contain a phase variable gene, seven have an obvious mutation in and appear to be dedicated phasevarions. This represents 17% of strains containing a phase variable gene. The remaining 15 strains that do not contain a phase variable gene are likely to be dedicated, functional, type III R-M systems. There were no strains in this group that had a corresponding inactivating mutation in the gene. In the strains above the distinct functions of regulation and restriction appear to be mutually exclusive.
In a related study, another type of inactivating frame-shift mutation has been identified in the genes associated with allele of () that results in premature truncation of the encoded protein and loss of the endonuclease domain, this mutation is present in 70% of strains with the allele (Srikhanta , submitted for publication). In , a 250-aa in-frame deletion has been observed in Res associated with the allele, potentially inactivating restriction function (Srikhanta , submitted for publication). These findings illustrate that inactivation of restriction function is common amongst the containing phase variable type III R-M systems.
Transformation experiments with strain Rd and strain 162 measuring the effectiveness of type III-specific restriction revealed only a two-fold increase in protection when the system was active. This appears to be a marginal level of functionality and raises the question of the selective advantage of restriction function at this level. More detailed analysis of restriction function is required to determine how many of the remaining strains contain a functional type III restriction system. Sequencing a region corresponding to one-third of the gene to look for 86-028NP point mutations, we found a different 32-bp deletion mutation that inactivated . The possibility remains that the obvious, inactivating frame-shift mutations found in genes associated predominantly with group, the largest group (representing 25% of strains surveyed), may represent the ‘tip of the iceberg’ with many other possible silencing mechanisms either unsurveyed or currently undetectable by simple sequence analysis of these poorly defined enzymes.
The observation of non-functional type III R-M systems in the strains containing phase variable genes suggests two possibilities, first, that the restriction function is redundant and has been lost, and second, that phase variable expression of may be inconsistent with a viable organism if an active Mod-Res holoenzyme can form. In non-phase variable type III R-M systems, sites are modified during replication, and any unmodified sites in newly replicated DNA are either in the same orientation or paired with methylated sites, thereby preventing suicidal restriction of cellular DNA (). In a strain containing an active type III R-M system with a phase variable gene, if the gene switches OFF for several rounds of replication, then phase varies back to ON, the condition may be lethal or detrimental, as none of the Mod sites in the genome would be methylated and protected from cleavage. Our original work in strain Rd (), and recent work in pathogenic (Srikhanta , submitted for publication), have established a role for phase variable genes in control of gene expression. This study has revealed that the strains associated with human disease may contain a series of distinct phasevarions. Knowledge of the changes in gene expression that occur as these phasevarions switch the organisms between two distinct cell types will have a major impact on our ability to develop an understanding of the general principles of host pathogen interactions for , and to assess antigens as vaccine candidates. Our current understanding is that the role of the locus in biology is in transition. In some strains of the population it retains its function as a type III R-M system, while in others it plays a key role as a dedicated, randomly switching, epigenetic mechanism for controlling gene expression.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Double-strand breaks (DSBs) can be created by different DNA-damaging agents or can occur spontaneously during cell growth. If not properly repaired, DSBs have the potential to affect cell viability, or to cause the loss or modification of genetic information (). In all organisms studied to date, two different mechanisms are used to repair DSBs: homologous recombination (HR) and non-homologous end joining (NHEJ) [reviewed in ()]. In HR a DNA break or gap is repaired by copying similar information present on a sister chromatid, a homologous chromosome, or at an ectopic location. In contrast, in NHEJ broken ends are ligated together without the need for extensive homology.
Among eukaryotes, these processes have been best characterized at the genetic and molecular level in the yeast . Homologous recombination is the major repair pathway in yeast. Yeast cells are able to repair broken chromosomes by carrying out a genome-wide search for homology, and using that information as a template to patch the broken chromosome. This usually results in the transfer of genetic information between the two loci, and may also lead to crossovers. If the information is present at different genomic locations, HR may result in genomic rearrangements, such as translocations, inversions and deletions [reviewed in ()]. HR usually requires a set of genes termed the epistasis group [reviewed in ()]. These proteins help search for homologous sequences and carry out a strand exchange reaction between regions sharing sequence similarities.
NHEJ is a repair mechanism that is conserved from bacteria to higher eukaryotes [reviewed in ()]. While NHEJ appears to be the major pathway for DSB repair in human cells (), it represents a relatively minor pathway in . A conserved set of proteins is required for NHEJ, including the Ku70/Ku80 heterodimer, DNA IV ligase and its associated factor Lif1/XRCC4. The MRX complex (MRN in humans) plays a role in promoting DNA joining both by HR and by NHEJ ().
The relative contribution of HR and NHEJ varies depending on both the organism and the context of the DSB. In organisms with genomes rich in repeats, recombination between non-allelic sequences can potentially lead to crossover and genome rearrangements. In such a case, NHEJ might prove safer (). However, NHEJ has also been linked to chromosomal rearrangements associated with the repair of a specific, induced DSB. These repair events consist of insertions, deletions, translocations and inversions ().
Previous studies have shown that unrelated DNA fragments, sometimes termed ‘filler DNA’ can be inserted into junctions during the joining reaction. For example, in mammalian lymphoid cells, extra nucleotides of filler DNA are usually found at VDJ joints; only part of this filler DNA is generated by terminal transferase activity (). DSB repair events in mammalian cells can also be associated with the capture of endogenous or exogenous DNA sequences up to several kilobases in length (). Filler DNA is also commonly observed in the repair of DSBs in plant cells (,). The insertion of exogenous DNA has also been observed in yeast cells. cDNA sequences of the natural Ty1 yeast retrotransposon could be found at repaired DSB sites (,,). Additionally, short mitochondrial DNA segments can be captured at break sites. This integration was not accompanied by modification of the junction sites, except for the loss or gain of 1–5 nucleotides (,).
A selectable assay system has been developed to study NHEJ-mediated chromosomal rearrangements associated with repair of a unique DSB in haploid yeast cells. Using this system Yu and Gabriel showed that extrachromosomal DNA fragments could be captured at the DSB site, in particular Ty1 cDNA and mitochondrial DNA sequences. In both wild type and cells, the inserted sequences were flanked by sequences with microhomology to the cut site, suggesting that NHEJ plays a central role in generating these events. Accordingly, the insertion process was shown to be -independent and dependent (,).
In order to study in more detail the mechanism that captures filler DNA, we created yeast strains that allow the generation of a linear DNA fragment in each cell undergoing a chromosomal DSB. The experimental setup allowed us to directly identify events in which insertions had occurred at the break site. As there are no sequences in the genome of these strains sharing extensive homology to the region undergoing the DSB, survival becomes dependent on the NHEJ repair pathway. Here we use this selectable assay system to show that annealing of complementary sequences at the termini of linear fragments play an important role in linear fragment capture. We also show that more than one linear fragment can be used during a single repair event. Finally, we present a model for DSB repair by capture of a linear fragment.
All yeast strains were originally derived from YFP17 (, , , , , , , , , ) (). Strains AGY670 and strain AGY673 carry an intronless copy of the gene ( and a allele into which the intron was inserted ( or , respectively). Strain AHY22 is a derivative (YFP17, , , ).
Strains AHY52 (AHY117, ) and AHY51 (AHY117, ) were made in several steps. Strain AHY117 was streaked onto 5-FOA plates to obtain a spontaneous Ura-clone (AHY156). The non-functional allele on chromosome was replaced with the -containing HI fragment from either plasmid pAH146 ( cassette) or plasmid pAH148 ( cassette). Correct integration of the cassette at the locus was confirmed by PCR and sequencing.
, and derivatives of AHY119 were obtained by one-step gene replacement.
All plasmids were transformed into DH5 competent bacteria. SURE competent bacteria (Stratagene) were used for the palindrome-containing inverted orientation.
pAH150 (‘Direct’ plasmid) and pAH129 (‘Inverted’ plasmid) were derived from pLAY98 () and pRS414, a -CEN plasmid (). Plasmid pAH174 (No-HOcs) was constructed by ligation of the II fragment into I and I-digested pRS414.
pAH146 (- cassette) and pAH148 (-) were constructed simultaneously. First, the 117-bp HOcs was amplified using primers pMH1, which contains an engineered I site upstream of the HOcs, and primer pMNH2, containing both I and I sites downstream of HOcs. The HOcs from plasmid pAH129 was used as a PCR template. The amplified fragment was then cut with I and ligated to I-digested pAGE1658 () (). Plasmids carrying both possible orientations were distinguished by digestion with I and I (a site unique to the plasmid). Candidates were subjected to sequencing to confirm the presence of the predicted configuration.
To construct plasmid pAH316 [referred to as ‘Inverted +2invHOcs’ plasmid, the double HOcs from plasmid pAH148 was amplified and ligated to a -digested pAH129 (‘Inverted’ plasmid)].
To construct plasmid pAH256 (‘Inverted-Fit’), pLAY98 was modified by replacing the upstream HOcs with an HOcs engineered to contain I and I/I ends, thus ensuring its orientation. The fragment cut with I and I was ligated into this modified pLAY98 plasmid that was cut with the same restriction enzymes. The resulting gene flanked with inverted HO cut sites was cut with II and ligated to pRS414 as described earlier.
Yeast cells were grown in yeast extract-peptone-dextrose (YP-dextrose) or synthetic complete media (SC) with appropriate amino acids missing (). Yeast extract-peptone-galactose (YP-galactose) and Yeast extract-peptone-raffinose (YP-raffinose) contain 2% galactose and 1% raffinose (w/v), respectively. 5-fluoroorotic acid (5-FOA) plates are SC-glucose plates supplemented with 1 mg/ml of 5-FOA ().
Twelve to 36 independent colonies of various yeast strains were inoculated into 3 ml liquid medium and grown at 30° to a final concentration of ∼3 × 10 cells/ml. Serial dilutions were then plated on YP-dextrose or YP-galactose plates and counted after 3–5 days. The survival frequency was calculated by determining the ratio of colonies growing on YP-galactose plates versus YP-dextrose plates (median values). Statistical analysis was carried out using a non-parametric Mann–Whitney rank test. Standard deviations were always lower than 20% of the median value.
YP-galactose plates were replica plated onto 5-FOA plates. The ratio of colonies growing on 5-FOA to those growing on galactose was used to calculate the frequency of 5-FOA resistance per survivor. The frequency of 5-FOA per plated cells was calculated by multiplying the previous two terms.
Colonies growing on 5-FOA plates were replica plated to SC-leu and SC-trp plates. The frequency of 5-FOA per plated cell with particular marker combinations was determined by multiplying the previous calculation by the proportion of the total 5-FOA colonies shown to have or lack those markers (i.e. Leu Trp–, Leu Trp and Leu– Trp–).
5-FOA resistant colonies were purified and grown in liquid YP-dextrose at 30° over night before total genomic DNA was isolated (). Genomic DNA samples were used as templates for PCR, using primers upstream and downstream of the HO cut site and the cs cassette. The PCR products were analyzed by sequencing.
The experimental system used () is based on the strains developed by Yu and Gabriel (). Haploid yeast cells contain a single HO recognition sequence (HO cut site or HOcs) placed into a non-essential portion of the intron, which has itself been engineered into the coding domain of the gene on chromosome . This allele is efficiently spliced, resulting in uracil prototrophy. In addition, this strain (AGY117) has been deleted of all -related sequences, and also contains an integrated, galactose-inducible HO endonuclease gene (). When transferred to galactose-containing media, a persistent and lethal DSB is formed on chromosome , unless the break is repaired in a way that eliminates the HO recognition sequence. Using this system Yu . captured extrachromosomal DNA fragments at the DSB site, by selecting for resistance to 5-fluoroorotic acid (5-FOA, i.e. loss of uracil prototrophy). In particular Ty1 cDNA (140 bp to 5.6 kb) and mitochondrial DNA sequences (33 to 219 bp) were identified.
To expand this work and study the mechanism by which filler DNA is incorporated into DSBs, we set up a related system to examine the fate of a defined extrachromosomal fragment in a cell with a defined DSB. We created a plasmid that upon induction of the HO endonuclease generates a linear DNA fragment carrying the marker gene. In our strains, a centromeric plasmid carries a selectable marker that is flanked by two HOcs. The two HOcs sequences are present either in the same (referred to as ‘Direct’) or in opposite (referred to as ‘Inverted’) orientation relative to each other (). These plasmids were independently transformed into AHY22, a strain carrying the cs allele in which all the relevant homologous sequences (the intron and the gene) had been deleted. Upon transfer of cells to galactose-containing medium, the HO endonuclease is transcriptionally activated and cleaves the three HO cut sites, causing the release of the fragment from the plasmid and creating at the same time a single DSB in the genome (). The cell's ability to survive depends on successful repair of the genomic DSB in a way that modifies the HO recognition sequence and forms HO endonuclease-resistant colonies. In previous studies, carried out in the absence of a linearized DNA molecule, the most common form of repair was an imprecise end joining event within the intron, that does not interfere with splicing, and which therefore does not make the cells resistant to FOA (,). We reasoned that if the linear fragment could participate in these repair events, its insertion at the DSB site should interfere with splicing and thus generate FOA resistant uracil auxotrophs. This experimental setup therefore allows screening for repair events in which the DSB has incorporated a defined linear fragment: these are detectable as colonies that, upon loss of the -marked plasmid, remain Ura and Leu. ‘Direct’ fragments can in principle be joined with the DSB either through annealing of the complementary 3′ overhanging four bases at both ends and precise re-ligation or by imprecise end joining of the non-complementary 3′ overhanging four bases (). However, if the break is repaired by precise re-ligation, it will be re-cut by the HO endonuclease. In order to stably insert a linear fragment at the DSB site, cells must use more complex mechanisms that modify the junction sequences, either by imprecisely joining complementary overhangs or by joining ends where the overhanging bases are non-complementary. For ‘Inverted’ fragments only one terminus can anneal, while the other must be joined in the absence of terminal complementarities. This can occur in either orientation of the fragment relative to the gene (). As a control we constructed a plasmid without any HO cut sites. In this case, there is no release of a linear fragment.
Wild-type cells carrying plasmids with the ‘Direct’, ‘Inverted’ or control ‘uncuttable’ fragment were plated on galactose-containing medium. Under these conditions, chromosome is cut by the HO endonuclease, and only cells able to repair the broken chromosome will survive to form colonies. Southern blot analysis showed that the efficiency of DSB formation by HO was very similar in all strains (data not shown). Survival of the three strains was also very similar (∼1% survival). These results suggest that the availability of linear fragments does not significantly increase a cell's ability to survive a DSB.
Most of the surviving colonies were 5-FOA sensitive, as expected from previous studies (). Our experimental system allows us to directly identify potential insertion events, which result in FOA resistance. The frequency of these events differed greatly among the various strains (). Except for cells carrying the ‘No-HOcs’ plasmids, all Ura colonies recovered were Trp, i.e. they had lost their plasmids. This result demonstrates that the HOcs in the plasmids were cut with high efficiency. The 5-FOA resistant colonies were then tested for leucine prototrophy, to identify fragment insertions, and individual colonies were analyzed by PCR amplification of the genomic region surrounding the chromosomal DSB at the locus.
In wild-type cells carrying a plasmid that released a linear ‘Direct’ fragment, the frequency of 5-FOA survivors per plated cell was 2.05 × 10 and 94% of the 5-FOA cells carried a insertion (). This is in marked contrast to cells carrying the linear fragment with inverted termini. In this configuration, the frequency of Leu colonies per plated cell was 0.01 × 10, a 200-fold difference (). These results demonstrate that a linear fragment present in the cells can be captured and used to repair a broken chromosome, but that the termini of the linear fragments affect the frequency of recoverable insertional repair. As expected, in cells carrying the plasmid lacking HO cut sites, no insertions were detected among the FOA resistant survivors (less than 10 Leu colonies per plated cell).
We carried out PCR analysis of the junctions of 52 insertion events. All ‘Direct’ linear fragments were inserted in the same orientation relative to the target locus ( gene). This is the orientation expected from events in which the complementary ssDNA ends can undergo annealing (A). This bias strongly suggests that annealing of the compatible ends plays an important role in directing the capturing of the linear fragment. We sequenced 10 of the insertions, and found that in 9 out of 10 survivors both ends of the inserted fragment exhibited an addition of 2 bp (+GT) (A). Addition of GT appears to be the predominant modification of the HOcs that prevents the site from being re-cut by the HO endonuclease. The GT addition was detected in previous studies (), and in more than 100 other sequences analyzed in this work (see further data later).
Sequence analysis of the junctions of three insertion events obtained with the ‘Inverted’ plasmid revealed simple end-joining events at one terminus with addition of 2 bp (+GT) while the other terminus exhibited a complex event (B). In one case the terminal region of the fragment was deleted, and in one case a genomic sequence (part of the gene) was inserted at the deleted end including the downstream gene. These results suggest that repair of a chromosomal end where simple end-joining cannot occur is associated with degradation of either the linear fragment, the chromosomal HOcs and associated sequences or both. As these results suggested that degradation of the ‘Inverted’ fragment could in many cases result in insertion of a partially deleted gene, which could produce a Leu- phenotype, we analyzed 20 independent 5-FOA resistant Ura– Leu– colonies obtained from the strain with the ‘Inverted’ plasmid, by PCR analysis with primers on either side of the HOcs. In these cases, the repaired chromosome exhibited either Ty or mitochondrial sequences insertions, or deletion of the HO cut site (). No truncated fragment was detected suggesting that mechanistically, end-joining precedes degradation, since the compatible ends of the same ‘Inverted’ fragment are not degraded.
To investigate whether the homologous recombination repair system affects the ability of cells to assimilate linear fragments, we carried out similar experiments in cells deleted either for the or for the gene. encodes a RecA homolog, and is known to be involved in strand exchange. In contrast, the function of Rad52 is less understood; it is, however, an essential component of the HR pathway and its inactivation affects most forms of recombination ().
cells carrying the linear fragment with direct or inverted termini exhibited a 2–4-fold decrease in viability and a frequency of FOA colonies per plated cell that was 10–20-fold lower than the wild type (). Furthermore, only 7% of the FOA resistant surviving cells with the ‘Direct’ fragment carried insertions (0.011 × 10 Leu colonies per plated cell, ). This represents a 183-fold reduction in the frequency of DNA capture events. None of the cells carrying the plasmid with the inverted configuration had a insertion (). These results suggest that Rad52p plays a central role in the capture of linear fragments, despite the lack of extensive homology between the chromosomal ends and the captured fragment.
We carried out an analysis of the junctions of 13 insertion events obtained from colonies carrying the ‘Direct’ plasmid. In 12 out of the 13 samples the ‘Direct’ linear fragment was inserted in the same orientation to the target locus ( gene).
We sequenced both junctions of 11 insertion events obtained from colonies carrying the ‘Direct’ plasmid. All survivors carried mutations in the upstream and downstream HO cut sites. However, in contrast to wild-type cells, only half of the junctions exhibited the common GT insertion. Additional junctions included insertion of TT, TGT or a simple G (B). In the only case in which the ‘Direct’ linear fragment was inserted in the opposite orientation to the target locus, sequencing revealed a complex event: at the upstream HOcs, the terminal 3′ T nucleotide of the chromosomal HOcs was deleted, and ligated to the HOcs of the fragment that lost its 6 terminal nucleotides. In addition, a 50-bp deletion of the HOcs of the fragment and 43-bp deletion of the chromosomal HOcs sequences were seen at the downstream HOcs (B).
cells carrying the ‘Direct’ plasmid exhibited a 5-fold lower frequency of FOA colonies per plated cell, compared to the wild type (). Furthermore, a 6-fold decrease in the frequency of cells that repaired the chromosomal break by capturing the ‘Direct’ fragment was observed (0.31 × 10 Leu colonies per plated cell). In contrast, the number of FOA resistant cells carrying the ‘Inverted’ fragment that captured the fragment was similar to that of the wild-type control (0.011 × 10 Leu colonies per plated cell).
Junction analysis revealed that in all samples the ‘Direct’ linear fragment was inserted in the same orientation to the target locus ( gene). These results, similar to those observed in wild type and cells, suggest again that the cohesive orientation is preferred, and that this preference is independent of the main HR proteins. Sequence analysis revealed a pattern very similar to that observed in wild-type cells: GT or TGT additions at both ends for the ‘Direct’ insertions, and GT addition on one end and rearrangement in the other for the ‘Inverted’ fragment (C).
These results show that, in contrast to the results obtained in the strain, lack of activity lowers the efficiency of capture in the ‘Direct’ orientation, without affecting the mechanism of capture in the ‘Inverted’ orientation.
Yku80 is part of the Ku heterodimer (composed of Yku70 and Yku80 in yeast), required for accurate NHEJ. To investigate whether Ku plays a role in linear fragment assimilation, we repeated our experiments in cells. Survival of cells on galactose was severely impaired (1000-fold lower than wild type). We were unable to recover any survivor carrying a fragment. These results indicate the importance of the Ku complex for survival after a DSB when homologous recombination is not an option, and further, its apparently absolute requirement for the incorporation of linear fragments into the broken chromosome.
In the vast majority of events examined, the ‘Direct’ linear fragment was inserted into the target locus in the same orientation. This suggests that simple annealing of the ends plays a dominant role in the capture of our linear fragments. To test this possibility, we changed the genomic target locus to an HO cut site that is incompatible with the fragment termini. Two new yeast strains were created. The first (AHY51, ) carries two HOcs in an inverted orientation relative to one another (hereafter named ‘’ strain). The second strain (AHY52, ) contains two HOcs in direct orientation (‘’ strain). Upon transfer of cells to galactose-containing medium, the HO endonuclease cleaves both HOcs simultaneously, leaving non-complementary 3′-overhanging termini on chromosome in the ‘’ strain and complementary 4-base 3′-overhanging termini in the ’ strain. The ’ strain exhibited a survival frequency similar to that observed in cells with a single HOcs. In contrast, the ’ strain exhibited a 24-fold lower survival frequency in the presence of either plasmid (). These results indicate that the presence of complementary ends at the DSB site on chromosome is the major determinant of survival, despite the additional requirement for a mutation that prevents re-digestion by the HO endonuclease.
We carried out PCR analysis of insertion events of ‘Direct’ fragments inserted at the locus, resulting in Leu, 5-FOA resistant colonies. These were of particular interest, since cleavage of the HO cut sites on chromosome should leave non-complementary ends, making the complementary overhang present at one end of the fragment as the only likely joining molecule. The fragments appeared in both orientations with respect to the target gene (12/20 insertions in the same orientation, and 8/20 insertions in the opposite one, A). Surprisingly, all survivors, irrespective of the orientation of insertion, still carried two inverted HO cut sites upstream of and a single HOcs downstream of the insert. All the double HOcs analyzed carried inactivating mutations resulting from insertion of GT dinucleotides in each HOcs (A). These results indicate that both HOcs at the chromosomal locus were cut and repaired, but that the 95 bp DNA fragment located between them was not lost or degraded. Our observations suggest a repair mechanism that keeps the broken ends together before the insertion of the linear fragment (see ‘Discussion’ section).
For cells with the locus carrying the linear ‘Inverted’ fragment, no complementary ends from the fragment should be present to be captured at the DSB site, so insertions should require a complex NHEJ event at each end. In all samples analyzed (10 independent colonies) the linear fragment was inserted in the same orientation as the target locus. Sequence analysis revealed that the two inverted HOcs were retained either upstream (4/10 insertions) or downstream (5/10 insertions) of the insert, mutated by GT/AC additions and in one case (1/10 insertion) the two inverted HOcs were retained in both upstream and downstream of the insert. In all cases examined, the single HOcs at the other end of the insert contained a partial deletion (B). Thus, as with the ‘Direct’ linear fragment (A) the chromosomal DSB is repaired by using two linear fragments, one released from the double-HOcs digestion, and a second one released from the cuts in the donor plasmid.
The results presented earlier suggest that the small fragments generated by the endonuclease are utilized for DNA capture at the nearby break. This could suggest that the small fragment remains somehow associated with the break. In order to test whereas this adaptor molecule acts only in or represent the capture of an additional linear molecule that can be supplied in , we created a new plasmid carrying the ‘Inverted’ linear donor, plus, at a second location in the plasmid, a copy of the construct. Upon expression of the HO endonuclease, five different DSBs are created: one at the locus on chromosome , and two pairs of breaks that release a linear fragment and a linear ‘adaptor’ fragment.
The presence of the construct on the plasmid increased the efficiency of fragment capture: the frequency of 5-FOA resistant colonies was 8 × 10 per plated cell, a 44-fold increase with respect to the ‘Inverted’ linear fragments. Moreover, the frequency of Leu colonies was similar to that observed with the ‘Direct’ linear fragment (1.44 × 10 per plated cell). An analysis of insertion events revealed that in all cases, DNA capture involved the use of the small HOcs-derived fragments. In 13 out of 20 samples analyzed the fragment was oriented as the target locus, whereas in the seven remaining samples it was oriented in the opposite direction. In all cases analyzed, all HOcs carried GT insertions. These results demonstrate that a broken chromosomal end can efficiently capture two independent linear fragments.
The highest frequency of capture was observed in strains carrying a single DSB at the :: locus and ‘Direct’ plasmids. We reasoned that if this is due to its ability to carry out simple annealing of both ends (provided there is an alteration of the terminal sequences during the joining process), then similar combinations of chromosomal breaks and linear ends should result in similar high levels of capture. We created an additional linear substrate, which we call ‘Inverted-Fit’ (C), which can be inserted into the DSB at the locus by simple annealing. A similar case is found in cells with the locus carrying ‘Direct’ fragment.
As predicted, in both cases the frequency of 5-FOA colonies was similar to that of cells with a single HOcs carrying the ‘Direct’ plasmid (). Similarly, nearly all (99 and 84%) of the cells able to grow on 5-FOA had an insertion of the fragment. An analysis of Leu colonies of these strains showed that the small excised inverted HOcs pair fragment was not used during repair. As before, both junctions carried single HOcs with GT insertions (C and data not shown).
italic
sup
#text |
Association of the matrix attachment regions (MARs) in the chromosomal DNA with the nuclear matrix organizes the higher order structure of the genome, forming looped structures that are likely to be equivalent to active chromatin domains in terms of transcription as well as replication (). The MAR sequences are generally AT-rich at 70% and possess potential of DNA bending (,). They have regions where base pairs tend to break under an unwinding stress (base-unpairing region: BUR), centered at a sequence ATATAT, which was termed as BUR nucleation sequence (). The tendency of base unpairing in the MAR DNA was shown to be essential in binding to the nuclear matrix and enhancing the promoter activity ().
Special AT-rich sequence binding protein 1 (SATB1) was originally isolated as a factor that specifically binds to the BUR sequence, where mutation of the ATATAT sequence impaired the binding (). SATB1, predominantly expressed in thymocytes, recruits histone deacetylase complex to the MAR site inside the interleukin-2 receptor α gene, in order to repress its expression (). SATB1-null mice exhibit irregular expression of the above gene and related genes in the premature CD4CD8 T-cells, which caused small thymi and spleens, and death at age of 3 weeks (). SATB1 also regulates the expression of fetal globin genes in the erythroid progenitor cells, by binding to MARs in the locus control region and the ε-globin promoter region in the β-globin cluster, and forming a complex with CREB-binding protein that possesses histone acetyltransferase activity (). Inside the cells, SATB1 is localized at nuclei and surrounds heterochromatin, forming a cage-like network structure (). Recently, it was shown that phosphorylation by protein kinase C alters preference for proteins to bind, i.e. histone deacetylase or histone acetyltransferase, and that acetylation by the latter impairs DNA-binding activity ().
SATB1 is ∼800 amino acids in length, possessing three DNA-binding motifs, i.e. two CUT domains and a homeodomain, in which a region of ∼150 amino acids that includes the N-terminal CUT domain (CUT repeat 1, CUTr1) is reported to be the region relevant to the recognition of the MAR DNA (MAR-binding domain; MBD) (). We have revealed by NMR that the CUTr1 region of ∼80 amino acids is folded whilst the remainder of MBD is largely unfolded (). The solution structure of CUTr1 contained five α-helices, in which the N-terminal four are arranged similarly to the four-helix structures of the CUT domain of hepatocyte nuclear factor 6α and the POU-specific domains of POU-homologous proteins (). Our NMR titration analysis and surface plasmon resonance (SPR) experiments using groove-specific binding drugs and methylated DNAs indicated that SATB1 possesses a DNA-binding mode similar to that of the POU-specific domain, in which third helix recognizes DNA bases from the major groove side (), although the mechanism of specific recognition of the MAR sequence has yet to be revealed.
In this study, we determined the crystal structure of the complex of SATB1–CUTr1 and a MAR DNA, by a molecular replacement method using the NMR solution structure. The domain indeed contacted DNA from the major groove side, similarly to the POU-specific domains. The contacting manners by the two types of domains show striking similarity to each other at least in part, revealing that they share a framework of the DNA binding.
A pET15b (Novagen) expression vector of a SATB1-CUTr1 mutant fragment where four basic residues Arg-Lys-Arg-Lys were attached to the C-terminus of fragment Asn368–Leu452 was produced by PCR from a previously made expression vector of fragment Val353–Asn490 () with primers including the mutational sequence (5′-CAACAGGTTAACACAGAGGTGTCTTCCGAAATC-3′ and 5′-CATATTACAAGCTCCTTTCCCTTTCGTCCTGG-3′; mutational sequence in the reverse primer is underlined and recognition sequences of I or HI are shown in italic.). Mutations of Gln402 and Gly403 to alanine (Q402A and G403A mutations, respectively) were introduced to fragment Val353-Asn490 by two-step PCR: the first step includes two independent reactions using a forward subcloning primer (5′-CCACCCTGTCAGTAGATCTATGAATAAGCCTTTG-3′; the I recognition sequence is shown in italic.) and a reverse mutational primer (5′-AAGCAAGCCCAGTTCTGTTAAAAGCCACACG-3′ for the Q402A mutation or 5′-TGAAAGCAAGCCTGAGTTCTGTTAAAAGCCA-3′ for G403A mutation; mutational bases are underlined) or a forward mutational primer (5′-AACAGAACTGGGCTTGCTTTCAGAAATCCT-3′ for the Q402A mutation or 5′-AGAACTCAGGCTTGCTTTCAGAAATCCTCCG-3′ for the G403A mutation; mutational bases are underlined) and a reverse subcloning primer (5′-CATTATATTAATTGTTCTCTGGTTTCCCATTCCTTTC-3′; the HI recognition sequence is shown in italic), and the second step includes a reaction with the two subcloning primers shown above, with mixture of the products of the first reactions as the template. The mutant proteins were expressed in cells and purified as described previously (). After the protein was mixed with a double-stranded 12-mer DNA (5′-GCTAATATATGC-3′/5′-GCATATATTAGC-3′) at a 1:1 molar ratio, the complex was purified by a Sephadex G-75 (Amersham) column chromatography. The buffer used for the chromatography was 50 mM sodium phosphate (pH5.5). Before crystallization, sample was dissolved in 10 mM Tris–HCl buffer (pH 8.0) by dialysis, and concentrated by ultracentrifugation with Amicon Ultra device (Millipore).
Experiments were carried out at 293 K using a Biacore X apparatus (BIAcore) essentially as described previously (). The running buffer was 50 mM sodium phosphate (pH 5.5) containing 0.005% Tween-20 for the mutants based on the Val353–Asn490 fragment, while for the mutant with four basic residues attached to the C-terminus of the Asn368–Leu452 fragment, 50 mM NaCl was supplemented to the buffer in order to minimize non-specific electrostatic attraction. A total of 766 resonance unit (RU) of a double-stranded 16-mer DNA (5′-bio-CGTTTCTAATATATGC-3′/5′-GCATATATTAGAAACG-3′) (‘bio’ indicates biotinylation at the 5′ end) were immobilized on the surfaces of Sensor Chip SAs (BIAcore) in one of the two flow cells. Experiments were repeated four times in order to estimate uncertainty of the binding constant. Data were analyzed as described previously ().
The protein–DNA complex of an initial concentration of 15 mg ml was subject to crystallization by a sitting-drop vapor diffusion method at 293 K on a 96-well Protein Crystallography Plate (Corning), with a reservoir solution of 50 mM Tris–HCl buffer (pH 8.0) containing 20% (w/v) polyethylene glycol 20000 (Wako, Japan), 10 mM MgCl or MgSO and 20% ethylene glycol. Rod-like crystals of 100–200 μm appear within a week.
The crystal produced in the buffer containing MgCl was flash-frozen in a nitrogen gas stream at 90 K and subjected to the measurement on a SMART6000 diffractometer (Bruker AXS) with Cu α radiation from a M06X rotation-anode generator (MAC Science) by two-axis crystal rotation. Diffraction data were indexed and scaled by programs SMART and SAINT (Bruker AXS), and merged and converted to structure-factor amplitudes by the SCALA () and TRUNCATE () programs from the CCP4 program suite ().
The crystal made in the buffer containing MgSO was flash-frozen in a nitrogen gas stream at 95 K and diffraction data were collected on beam line NW12A at PF-AR synchrotron (KEK, Tsukuba, Japan) by one-axis crystal rotation. HKL2000 (), SCALA and TRUNCATE programs were used for the data processing. The diffraction data statistics are listed in .
The protein–DNA complex structures for the laboratory diffraction data were calculated by a molecular replacement method, using the NMR solution structure of SATB1-MBD (PDB entry 1YSE). Initially, the structure of the protein moiety in the complex was obtained from the starting template of the NMR structure, by a cross-rotation method implemented in the CNS package (). From this protein structure and a standard B-DNA structure produced by Insight II program (Accelrys), the complex model was obtained by the cross-rotation and translational search methods, where the coordinate of the protein model was fixed. Iterative cycles of model building to fit to the electron density map and dynamical refinement were carried out by XtalView () and CNS, respectively. The water molecules were picked up at 1.5 σ level and within distance of 4.0 Å from the protein or DNA atoms by a program implemented in CNS, although some waters shifted slightly more distant during final refinements.
Root mean square deviation (RMSD) values from the idealized geometry with regard to bond lengths and angles were obtained by CNS. Ramachandran profile and secondary structure elements of the protein moiety are analyzed by Procheck (). The statistics of the structures are listed in .
Intermolecular contacts are analyzed by in-house FORTRAN programs, XtalView () and Insight II (Accelrys). Hydrogen bonds are defined by distance of hydrogen donor and acceptor <3.4 Å and predictable donor-hydrogen-acceptor angle >110°. Electrostatic contacts are defined by N-O distance <5.0 Å, unless the groups form a hydrogen bond(s). Among the van der Waals interactions, contacts between apolar atoms (‘apolar contacts’ in this study), are defined by C-C distance <4.5 Å, unless an adjacent N or O atom is closer to the counterpart C atom. Atom pairs not involved in the above three contacts, with distance <3.9 Å are classified into other van der Waals contacts (or simply ‘van der Waals contacts’ in this study), unless an adjacent atom is closer to and forms a hydrogen bond to the counterpart atom. Effects of amino acid or base substitutions are evaluated by the Biopolymer module in Insight II. RMSD values between the molecules are calculated by using MOLMOL ().
For crystallization of protein–DNA complex, we initially intended to use a CUTr1 fragment, Asn368–Ala455, which is essentially the folded region of MBD as revealed by the NMR analysis (), since including unfolded regions may interfere crystallization. However, this fragment is rather acidic, with a pI value of 5.8 and a net charge of −1 at neutral pH, which is unfavorable for interaction with the negatively charged DNA, causing electrostatic repulsion. Consequently we did not observe significant DNA binding for this fragment (). Therefore, we prepared a mutant containing four basic amino acids at the C-terminus of fragment Asn368–Leu452 (a), which possesses a pI value of 10.1 and a net charge of +3. This fragment showed DNA binding with a constant of 9.5(±1.2) × 10 M, which is slightly better than that of the basic MBD fragment with a net charge of +2 [4.8(±0.6) × 10 M; ()], even though the experiment for the former was carried out in a higher salt concentration (). We used this mutant CUTr1 fragment for crystallization of the complex with DNA, which will be treated as SATB1-CUTr1, hereafter, unless otherwise stated.
For DNA, we used a double-strand dodecamer including a sequence from the IgH enhancer region, CTAATATAT (b), which was included in DNAs for the gel retardation and missing nucleoside experiments () as well as for our SPR experiment (). In the sequence, the ATATAT sequence is termed as the BUR nucleation sequence that is critical in the base unpairing and at least partially in the binding of SATB1 (,).
The 1:1 mixture of SATB1-CUTr1 and dodecamer DNA was purified by a gel filtration column and subjected to crystallization. Crystals produced with different salts were subjected to the measurements with laboratory and synchrotron X-ray sources under freezing condition by nitrogen gas stream, resulting in diffraction data of resolution up to 2.00 and 1.75 Å, respectively. The structure was determined by a molecular replacement method using the NMR structure of SATB1-MBD () and a standard B-DNA structure. The statistics for the diffraction data, refinement results and structural properties are shown in .
A triclinic crystal lattice contained a single structure of the complex of SATB1-CUTr1 and a dodecamer DNA possessing overall dimensions of ∼40 Å × 40 Å × 40 Å. Crystal packing was promoted by DNA–DNA, protein–protein and protein–DNA interactions between the adjacent lattices (Supplementary Data and Figures S1–S3), which involves an artificially introduced Arg residue. Since the two structures determined from the two datasets are essentially the same (Supplementary Figure S4), we will describe on the structure from the laboratory data, which possesses better values (), unless otherwise stated.
The protein moiety consists of five α-helices (α1: Ile375–Ala386, α2: Gln390–Phe398, α3: Gln402–Lys411, α4: Gln420–Leu433 and α5: Glu437–Leu452; a short helical region of Pro415–Thr417 appears only in the structure from the laboratory data, which is likely to be related to the crystal packing; Supplementary Data), as identified by the program Procheck () (a and a). The structure of DNA is essentially in the standard B-form without significant deformation. The α3-helix, especially in its N-terminal half, deeply enters the major groove of DNA acting as the recognition helix, in a manner where the helical axes of the α-helix and DNA are perpendicular to each other. The backbone structure of the protein in the crystal is similar to that of the solution structure determined by NMR (the minimized average structure of the ensemble) (b). The RMSD value for the Cα atoms in the Val371-Gln445 region is 0.99 Å after superimposition. However, the two structures significantly differ in Arg400–Gln402 in the N-terminal cap region of α-helix 3, with RMSD of 2.59 Å (b). In this region, α-helix 3 starts from Gln402 in the crystal structure, even though Gly at position 403 generally acts as a helix breaker, although it starts from Leu404 in the solution structure. This is likely to be caused by the interaction with DNA, since the side-chains of the residues in this region directly contact DNA bases and phosphates, as described below. The difference in the C-terminal region is more noticeable, with RMSD of 6.12 Å for Asp446–Leu452, where the helix of the solution structure is truncated at Gln445 while that in the crystal structure elongates to Leu452. This is not likely to be the result of the protein–DNA interaction, since the previous NMR titration experiment did not show significant chemical-shift changes in this region (). Alternatively, we suggest that the solution structure of the domain also contains the long helix at least in some conformations, although NMR structural constraints based on nuclear Overhauser enhancements and amide hydrogen exchange protection were not obtained because of the significant flexibility in this region. The Asp446–Leu452 region shows relatively large temperature factors, with average of 39.2 ± 8.9 Å for backbone atoms, where the whole determined region (Glu370–Arg453) shows 22.4 ± 9.2 Å. In addition, the two crystal structures showed relatively large RMSD value of 1.29 Å in this region, which also indicates the high flexibility (Supplementary Figure S4).
The mode of the protein–DNA interaction in this crystal structure is consistent with our previous observation by the NMR titration analysis and SPR experiments indicating that α-helix 3 enters the major groove of DNA (). A computational model of the complex based on these results shown in the same report is indeed similar to the present crystal structure.
Residues in α-helix 3 and its N-terminal cap (Thr401–Glu407) contact eight bases in six consecutive base pairs, through direct or water-mediated hydrogen bonds, and contacts between apolar atoms (apolar contacts) and other van der Waals interactions ( and a) (see Methods section, for definition). All the contacts including those to the sugar-phosphate backbone cover the range of eight base pairs (a). Details of the contacts are as follows.
The O atom of Thr401 and the backbone amide of Gly403 form hydrogen bonds to a same water molecule that possesses hydrogen bonds to the N and N atoms of A4 (a). At the same time the O atom of Thr401 forms van der Waals contacts to C and C of T3 (data not shown). Also, the C and backbone O atoms of Gly403 form contacts to the C and/or O atoms of T5′ (a). In addition, Gly403 is important in that it does not contain side-chain, since presumable C and H atoms would have steric hindrance with NH of A4 and O of T5′ (data not shown). Consistently, a mutation of Gly403 to Ala resulted in a reduction of DNA-binding activity by ∼10-fold (b).
The O and N atoms of Gln402 form hydrogen bonds to N and N, respectively, of A6′ in a manner typical of the DNA base recognition by proteins (,) (b). At the same time, the C, C, O, and N atoms form contacts to the C, C and/or C atom of T7′. In addition, O of Gln402 and N of A5 form a water-mediated hydrogen bond (a). These interactions are very important, as confirmed by a mutation of Gln402 to Ala, which results in an impairment of DNA-binding activity by ∼50-fold (b).
The C atom of Leu404 forms an apolar contact to the C atom of the C2 base, with distance of 4.3 Å (data not shown). Although this contact is not likely to strongly exclude other bases, a pyrimidine base may be preferable, since C may expand the area of apolar contacts, whilst the presumable position of C of a purine base, when simply replaced, would be slightly more distant to Leu404 C, with a distance of 4.4 Å (data not shown). In addition, the C, C and backbone N atoms of Leu404 form contacts to C of T3.
The O atom of Ser406 form van der Waals contacts to C of T5′ and C of A6. The C, C and O atoms of Glu407 form contacts to the C methyl groups of T4′ and/or T5′ (c and a). It is noteworthy that two basic residues in α-helix 3, Arg410 and Lys411, form hydrogen bonding and/or electrostatic contacts with Glu407. Especially, Arg410 form two hydrogen bonds to Glu407 (c). These interactions are likely to constrain the conformation of Glu407 side-chain so that the C and C atom are closer to the C atoms of T4′ and T5′ than the O atoms, which is essential for the apolar contacts. Importance of the Glu407–Arg410 interaction is evident from our previous observation that mutation of Arg410 to Asn results in a decrease of DNA-binding affinity by ∼40-fold ().
DNA sugar-phosphate backbones are contacted through direct or water-mediated hydrogen bonds, electrostatic attractions and apolar and van der Waals contacts (a). Contacting residues are Ser389, Gln390 (b), Ala391, Arg395, Arg400, Thr401, Gln402, Leu404, Ser406, Glu407 (c), Arg410 (c), Ser421 and Asn425. In addition, O of Ser421 is in a position allowing a hydrogen bond to the backbone phosphate of G1, if the DNA is not artificially truncated. Among these residues, Gln390 is of special note in that, in addition to two hydrogen bonds to two phosphate groups (a), the O atom forms a hydrogen bond to the N atom of Gln402 that is the only residue forming direct hydrogen bonds to the base (b). This hydrogen-bonding network is likely to contribute primarily to fixing the relative position of protein and DNA, namely, defining the DNA-binding framework.
The contacts to the DNA backbone should be important in the protein–DNA binding, enhancing the affinity, although not directly contributing to the sequence-specific recognition. It should be noted that Thr401, Gln402, Leu404, Ser406 and Glu407 contact to both the DNA bases and backbones, simultaneously. It was shown for Ser406 that a mutation to Ala reduces the DNA-binding activity by more than 10-fold ().
It should be emphasized that mutations of the base-contacting residues conserved among CUT domains (a), i.e. Gln402 and Gly403, on the MBD fragment (Val353–Asn490), as well as those of Ser406 and Arg410, significantly reduced the DNA-binding activity [b; ()]. Therefore, the binding mode described above is unlikely to be induced by the basic residues introduced at the C-terminus of the present SATB1-CUTr1 fragment.
The base contacts described above are likely to define the sequence specificity of SATB1-CUTr1 to the MAR DNA, resulting in a recognition sequence of CTAATA/TATTAG, or (Y)TAATA/TATTA(R) considering the relatively weak contact and the presumable preference to pyrimidine (Y) rather than purine (R) at the C2 position. The recognition sequence only partially includes the BUR nucleation sequence (), ATATAT (b), to which SATB1 is initially expected to bind (). The mutation that impaired the binding of SATB1 changed the CTAATA sequence to CTACTG (), which should interfere the contacts by Gln402, Gly403, Ser406 and Glu407, at least partly, according to the present contacting profile (a). Consistently, a missing nucleotide experiment in the same report showed that SATB1 contacts the DNA in the region centered at a sequence CTAATA, but not at ATATAT. We have previously introduced methylation at the N atom of adenine base at position of A3′ or A6′ and observed that DNA binding was significantly interfered only when A6′ was methylated (), which is also consistent with the present observation that the O atom of Gln402 forms a hydrogen bond with the N atom of A6′, while the A3′ base is not contacted by the protein (b and a)
This study is the first report regarding the geometry of sequence-specific DNA recognition by a CUT domain. There are three subgroups of CUT-domain proteins, i.e. (i) SATB proteins including SATB1 and SATB2, possessing two CUT-domain repeats, (ii) ONECUT group including HNF6α and ONECUT proteins, possessing a single CUT domain and (iii) human CCAAT displacement protein (CDP) and its murine autholog, Cut homeobox (Cux), possessing three repeats (a). Sequence specificity and/or DNA-binding activities of these CUT domains are significantly different among one another. Namely, HNF6α recognizes the ATCAAT sequence (), and the N-terminal CUT domain, repeat 1, of CDP/Cux recognizes the CCAAT sequence, while the other two CUT domains, repeat 2 and repeat 3, recognize ATCGAT (), which are different from the present SATB1-binding sequence (c). In addition, Cut repeat 2 of SATB1 protein is dispensable for binding to the MAR DNA (). However, many DNA-contacting residues observed in this study are conserved among these CUT domains (a), indicating that the DNA-binding mode is essentially shared. Especially, the important base-contacting residues, Gln402 and Gly403, are perfectly conserved, which form hydrogen bonds to A6′ base, and apolar and van der Waals contacts to T5′, respectively. Also, Thr or Ser is conserved at position 401, which is, together with Gly at position 403, capable of forming the water-mediated hydrogen bonds to the N and N atoms of the A4 base. With regard to these hydrogen bonds, the A4 base can be replaced by a G base, since O and N of the G base would act as two hydrogen-bond acceptors. Therefore, a sequence of the three base pairs R4A5T6/A6′T5′Y4′ is likely to be recognized commonly by the CUT domains. Considering this, the recognition sequences were aligned as in c, resulting in the consensus core recognition sequence of YYRAT/ATYRR.
The recognition sequence of HNF6α, ATCAAT/ATTGAT, fits the YYRAT consensus in the both directions (c). It should be noted that HNF6α possesses Thr and Asp in the positions equivalent to Leu404 and Glu407, respectively, of SATB1-CUTr1 (a). When Leu404 is simply replaced by Thr in the present SATB1-CUTr1/DNA structure, it is still possible to form apolar contacts to the C atom in the T3 base, but not to atoms of the C2 base. In contrast, when Glu407 is replaced by Asp, apolar contacts to the C atoms of the T4′ and T5′ bases are likely to be disrupted. These observation suggest that HNF6α recognizes the T3/A3′ base pair, but not the A4/T4′ base pair, and, therefore, that the ATTGAT sequence (‘HNF6aR’ in c), but not the ATCAAT, is the direction equivalent to the YYRAT consensus core sequence.
The three CUT domains of CDP possess Ser at the position equivalent to Leu404 of SATB1-CUTr1 (a). When Leu404 is simply replaced by Ser in the SATB1-CUTr1/DNA structure, an apolar contact between the Ser C atom and the T3 C atom would still be possible depending on the conformation (data not shown), although it should be much weaker than those from the C and C atoms of the Leu residue and may allow a C base at the T3 position (c). It should be noted that CUT repeat 1 possesses Glu at the position equivalent to Glu407, although it is Asp for CUT repeats 2 or 3. For CUT repeat 1, the interactions between Glu and the T4′ and T5′ bases should be conserved, although for CUT repeats 2 and 3, those between Asp and the T bases are unlikely. Therefore, repeat 1, but not repeat 2 or 3, is likely to require an A base at position 4.
SATB1-CUTr2 is not likely to contribute to the MAR-DNA binding (), which may be explained as follows. Among the several DNA-contacting residues that differ between CUTr1 and CUTr2, those likely to largely influence the binding are Trp at position 404 and Cys at position 406 (a). The former possesses a bulky side-chain compared to Leu and, when simply replaced, shows a steric hindrance with DNA sugar-phosphate backbone or other amino acids, such as Arg400 and Asn425, probably disrupting contacts including hydrogen bonds to the phosphate. Also, the side-chain of Cys at position 406 is unable to form a hydrogen bond to a DNA phosphate.
We previously predicted that the framework of DNA binding by the five-helix SATB1-CUTr1 structure is similar to that by the basically four-helix POU-specific domains of POU-homologous proteins, in that the third helix acts as the recognition helix and deeply enters the major groove of DNA, judging from the structural similarity and the NMR titration and SPR experiments (), which was confirmed by the present crystal structure. In the crystal structures of the complex of DNA and POU-homologous proteins, Oct-1, Pit-1 and HNF1α (PDB entries 1OCT, 1AU7 and 1IC8, respectively), the third helix of the common four helices of the POU-specific domains acts as the recognition helix, achieving the sequence-specific DNA recognition () (c and d). We have re-analyzed contacting profiles in these crystal structures, by the same criteria adopted in this study.
Although the amino acid sequences in the recognition helices of POU-specific domains are not very similar to those of CUT domains (b), a Gln residue that is strictly conserved among the POU-specific domains forms a pair of hydrogen bonds to adenine N and N atoms, essentially in the same manner as Gln402 of SATB1 (b and d). In addition, another Gln residue is also conserved, which forms hydrogen bonds to the side-chain of the above Gln residue and at the same time to phosphate groups from both the side-chain and backbone N atoms, in the similar manner as Gln390 of SATB1 (a, c and d, and b and d). Since this conserved hydrogen-bonding network is most important in defining the DNA-binding framework, as described above, SATB1-CUTr1 and POU-specific domains are likely to share the DNA-binding framework.
Therefore, in order to compare the DNA-contacting profiles, amino acid sequences of the recognition helices are aligned so that the Gln residue equivalent to Gln402 of SATB1 (hereafter we use term ‘residue 402’, etc.) is conserved among the proteins (b). Also, sequences of the DNA regions contacted in the crystal structures are aligned in reference to the T6-A6′ base pair that is contacted by Gln402 (c). Based on the alignments, the contacting profiles i.e. locations and modes, of SATB1-CUTr1 and POU-specific domains are summarized in d. In addition to the conserved hydrogen bonds to the A base at position equivalent to A6′ of SATB1-CUTr1 recognition sequence (hereafter we use term ‘base 6′’, etc.), there are several points where the contacting profiles of the SATB1-CUTr1 and POU-specific domains are similar to each other, as indicated in red color in d. For example, hydrogen bonds between residue 403 and base 4 are observed for SATB1 and all the POU-specific domains. It should be noted, however, that the amino acids of residue 403 (Gly, Thr or Ser) and bases of base 4 (A or C or T) are different and therefore, manners of hydrogen bonding are different. Namely, for OCT1, O of Thr accepts a hydrogen bond from N of C base, while for Pit-1, O of Thr donates a hydrogen bond to O of T base, and for HNF1α, O of Ser accepts a hydrogen bond from N6 of A base (data not shown), which are also different from SATB1-CUTr1, where backbone N of Gly donates a hydrogen bond to a water that behaves as both hydrogen donor and acceptor to the A4 base (a).
There are four contacts that are observed only for SATB1-CUTr1, while 16 are observed only for either of the three POU-specific domains (d). Among the latter 16, only four are observed for all the POU-specific domains, i.e. hydrogen bonds between residue 403 and base 5′, hydrogen bonds between residue 407 and base 3, van der Waals contacts between residue 403 and base 4, and van der Waals contacts between residue 407 and base 3. Even for these four, manner of contacts are diverse, similarly to as described above. Thus, the difference between the base-contacting profiles of SATB1-CUTr1 and POU-specific domains are at a similar level to that among the POU-specific domains. In fact, common and different contacts between SATB1-CUTr1 and HNF1α are both 14 in d, which is better than the case between OCT-1 and HNF1α, showing 11 common and 15 different contacts, and the case between Pit-1 and HNF1α with 11 common and 18 different contacts.
Contacts to DNA backbone are also similar among these domains at least partially. Namely, contacts equivalent to the hydrogen bonds from Gln390 (Gln27 of OCT-1 and of Pit-1, and Gln130 of HNF1α; mentioned above; d), Thr401 (Ser43 of OCT-1 and Pit-1), Ser406 (Ser48 of OCT-1 and Ser145 of HNF1α), Ser421 (Ser56 of OCT-1 and Pit-1) and Asn425 (Asn59 of OCT-1 and Pit-1) of SATB1-CUTr1 to phosphate groups (a) are observed also for either of the POU-specific domains. In addition, apolar contacts similar to that from Ser389 of SATB1 are observed for Thr26 of OCT-1 and Pit-1. It should be noted that the contacted phosphates and sugars are equivalent to those of the nucleotides contacted by SATB1-CUTr1 on the basis of the alignment in c.
It is concluded, therefore, that the SATB1-CUTr1 shares its DNA-binding framework with POU-specific domains, where a significant number of contacts to DNA bases and backbones are conserved. Within the framework, contacts of bases by similar or different types of amino acids located at similar positions yield a partly similar recognition sequences. In the alignment in c, a preference to the ATA/TAT sequence at positions 5–7/7′–5′ is clear. It should be noted that one of the sequences contacted by HNF1α (TTAATA) matches the (Y)TAATA recognition sequence of SATB1 when the relatively weak contact and the presumable preference to pyrimidine (Y) rather than purine (R) at the C2 position was considered, as described above.
As described above, the CUT domains and POU-specific domains are two very similar subtypes of helix–turn–helix DNA-binding domains, which are highly likely to be evolutionarily related to each other. The conserved hydrogen-bonding network involving two Gln and an A base (a,c and d; b and d) is also observed for some other proteins with helix–turn–helix DNA-binding motif, such as 434 repressor (), 434 Cro () and λ repressor (), which all are from phages, but not for Trp repressor (), CAP (), lac repressor (,) or purine repressor (), which are of bacterial origin. Therefore, this hydrogen-bonding network can be a strong clue to classifying the subfamilies of the helix–turn–helix DNA-binding domains, and CUT and POU-specific domains are likely to be related to phage repressors, rather than bacterial ones.
The proteins containing CUT or POU-specific domains are also likely to be related to each other. They commonly possess homeodomain regions more C-terminal to CUT or POU-specific domain, except for minor entries possessing only CUT or POU-specific domain, which appear in the NCBI Conserved Domain Database (). Both the groups of the proteins are identified from group of animals that includes nematoda, insects and vertebrates, but not from more primitive animals, yeast or plants, while homeodomain proteins exist more widely in eukaryotes. Together with the structural relationship with phage repressors, we propose a hypothesis of a lateral transfer in which a phage repressor-like DNA-binding domain was incorporated into a homeodomain protein of a nematoda-like animal that is fed on bacteria occasionally transfected with phages. After the transfer, the proteins became divided into two groups along with the evolution of animals. During the evolution, these proteins are likely to keep functioning as transcriptional repressors or activators, although the target systems became quite diverse, e.g. nervous systems including pituitary gland, and blood and immune systems (,).
The most characteristic point in the interaction between SATB1–CUTr1 and a MAR–DNA is that only a single pair of direct hydrogen bonds are used to recognize bases (b and a), which is atypical of DNA recognition from the major groove side mostly driven by direct hydrogen bonds (,). Although water-mediated hydrogen bonds, apolar contacts and other van der Waals contacts compensate them to achieve the sequence specificity, the affinity is not expected to be very strong. In fact, this CUT domain itself did not show significant DNA-binding activity when isolated from the other regions of MBD (), and we attached four basic residues at C-terminus for the present crystallographic study. To gain stable DNA binding by the full-length protein, attachment of homeodomain at more C-terminal region () and dimerization driven by the N-terminal PDZ domain () are likely to be necessary. When the PDZ domain is removed by proteolysis by caspase 6, the DNA-binding activity of the protein as well as the activity in the transcription regulation is significantly reduced (). Recently it was reported that acetylation by histone acetyltransferase at the N-terminal PDZ-domain region impairs the DNA-binding activity, implying a possibility of loss of dimerization (). Therefore, affinity in the DNA-binding domain itself should not be too strong, in order to allow post-translational regulation of the DNA-binding activity of the protein in cells, and thereby effective transcriptional regulation.
The co-ordinates of the structures determined from the laboratory and synchrotron data sets as well as the relevant structural factors have been deposited to the Protein Data Bank under accession IDs 2O49 and 2O4A, respectively.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Telomeres are specialized DNA sequences that cap the ends of chromosomes and play an important role in cancer (), aging (,) and genetic stability (). Human telomeric DNA consists of tandem repeats of hexanucleotide d(TTAGGG) for 5–8 kb in length, terminating in a single-stranded 3′-overhang of 100–200 bases in length (). In normal somatic cells, each cell replication results in a 50–200 base loss of the telomere, and after reaching a critical shortening of the telomeric DNA, the cell undergoes apoptosis (). In contrast, telomeres of cancer cells do not shorten on replication, due to the activation of a reverse transcriptase telomerase that extends the telomeric sequence at the chromosome ends (). Telomerase has been shown to be activated in 80–85% of human cancer cells (), and has been suggested to play a key role in maintaining the malignant phenotype by stabilizing telomere length and integrity (). The G-rich telomeric sequence can form DNA G-quadruplex structures consisting of stacked G-tetrad planes with Hoogsteen-type hydrogen bonds and stabilized by monovalent cations such as Na and K. Formation of DNA G-quadruplex in the human telomeric sequence has been shown to inhibit the activity of telomerase, thus the intramolecular telomeric G-quadruplex has been considered to be an attractive target for cancer therapeutic intervention (,,,).
Structural information of the intramolecular human telomeric G-quadruplex formed under physiologically relevant conditions is necessary for structure-based rational drug design. Because the K structure is considered to be biologically more relevant due to the higher intracellular concentration of K, it has been the subject of intense investigation (), although the Na structure was reported more than a decade ago (). The minimal requirement for an intramolecular telomeric G-quadruplex is a four-G-tract human telomeric sequence (A and B). We have recently determined the folding and solution structure of a hybrid-type mixed parallel/antiparallel-stranded intramolecular human telomeric G-quadruplex (Hybrid-1) in K solution (C, left) using a 26-nt sequence that contains the wild-type 22-nt four-G-tract human telomeric core sequence with modified flanking sequences (Tel26, A) (,). An adenine triple capping structure was found to form in the Hybrid-1 telomeric G-quadruplex. The Hybrid-1 folding has also been independently reported by others (,).
Here we report the NMR solution structure of the major intramolecular G-quadruplex formed in a wild-type 26-nt four-G-tract human telomeric sequence wtTel26 in K solution. This is the first intramolecular G-quadruplex structure of a biologically native, unmodified human telomeric sequence. This structure is also a hybrid-type mixed parallel/antiparallel-stranded G-quadruplex (Hybrid-2) (C, right) but differs from the Hybrid-1 structure in its loop arrangements, strand orientations and capping structures. The Hybrid-2 folding topology has also been independently reported (). In our NMR structure of the Hybrid-2 telomeric G-quadruplex, a novel T:A:T triple capping structure is found to form, which covers one end of the Hybrid-2 telomeric G-quadruplex and appears to play an important role in the selective stabilization of the Hybrid-2 structure. The structural information obtained from this study and our previous study () allowed us to understand the favored formation of a specific hybrid-type telomeric G-quadruplex, which appears to be largely determined by the specific capping structures. The distinct capping structures of each hybrid-type intramolecular telomeric G-quadruplex may provide specific binding sites for drug targeting. To gain insight into the structure polymorphism and equilibrium of human telomeric DNA, we have systematically examined the four-G-tract human telomeric sequences. We found that the hybrid-type G-quadruplexes appear to be the major conformations formed in these human telomeric sequences in K solution and are always in equilibrium between Hybrid-1 and Hybrid-2 structures, which is largely determined by the 3′-flanking sequence. The energy barrier between the two hybrid forms appears to be rather small and can be readily shifted by minor changes. Intriguingly, both hybrid-type G-quadruplex structures suggest a straightforward means for multimer formation with effective packing in the human telomeric sequence (D) and provide important implications for specific drug targeting of G-quadruplexes in human telomeres.
The DNA oligonucleotides were synthesized using β-cyanoethylphosphoramidite solid-phase chemistry on an Expedite™ 8909 Nucleic Acid Synthesis System (Applied Biosystem, Inc.) in DMT-on mode, and were purified using C18 reverse-phase HPLC chromatography, as described previously (,,). Samples in DO were prepared by repeated lyophilization and final dissolution in 99.96% DO. Samples in water were prepared in 10%/90% DO/HO solution. The final NMR samples contained 0.1–2.5 mM DNA in 25 mM K-phosphate buffer (pH 7.0) and 70 mM KCl.
NMR experiments were performed on a Bruker DRX-600 spectrometer as described previously (,,,). Standard 2D NMR experiments, including NOESY, TOCSY and DQF-COSY, were collected at 1, 5, 10, 15, 20, 25 and 30°C to obtain the complete proton resonance assignment. Non-exchangeable protons were estimated based on the NOE cross-peak volumes at 50–300 ms mixing times, with the upper and lower boundaries assigned to ±20% of the estimated distances. Distances between exchangeable protons were assigned with looser boundaries of ±25%. The methyl base proton Me-H6 distance (2.99 Å) was used as a reference. The distances involving the unresolved protons, e.g. methyl protons, were assigned using pseudo-atom notation in X-PLOR.
Metric matrix distance geometry (MMDG) and simulated annealing calculations were carried out in X-PLOR () to embed and optimize 100 initial structures based on an arbitrary extended conformation for the single-stranded wtTel26 sequence, as described previously (,). The experimentally obtained distance restraints and G-tetrad hydrogen-bonding distance restraints were included during the calculation.
All of the 100 molecules obtained from the DGSA calculations were subjected to NOE-restrained simulated annealing refinement in XPLOR () with a distance-dependent dielectric constant, as described previously (,). The force constants were scaled at 10–30 and 80–100 kcal mol Å for NOE and hydrogen bond distance restraints, respectively. A total of 727 NOE distance restraints, of which 324 are from inter-residue NOEs, were incorporated into the NOE-restrained structure calculation.
Dihedral angle restraints were used to restrict the glycosidic torsion angle (χ) for the experimentally assigned and conformations. A dihedral angle restraint of 60(±35)° was applied to the G-tetrad guanines, and a dihedral angle restraint of 240(±40)° was applied to the G-tetrad guanines. The force constants of dihedral angle restraints were 10 kcal · mol · rad.
NOE-restrained simulated annealing refinement calculations were performed as described previously (,). The time steps for all processes of heating, cooling and equilibration were set to 1 fs. The 10 best molecules were selected based both on their minimal energy terms and number of NOE violations and have been deposited in the Protein Data Bank (PDB ID 2JPZ).
We have very recently reported a hybrid-type intramolecular telomeric G-quadruplex structure formed in potassium solution using a 26-nt sequence, which contains the wild-type 22-nt four-G-tract human telomeric sequence that was used for the previous NMR and X-ray structural studies (,) with a flanking AA at each end (Tel26, A). During this work, we found that the Tel26 sequence forms two stable G-quadruplex conformations, with well-defined guanine imino-peaks, when examined right after it was prepared in K solution (Figure S1). The second conformation (∼40%) slowly converts to the first or final conformation after overnight, while the complete conversion takes about 1 day (Figure S1). The CD spectrum of Tel 26 is very similar to that of wtTel26, indicating that the two conformations are similar in nature. We decided to go back and re-examine the wild-type 26-nt telomeric sequence wtTel26, (TTAGGG)TT (A).
The 1D H NMR spectrum of the wtTel26 sequence in K solution is shown in (top). It shows a major species with twelve resolved imino proton resonances between 10.5 and 12 p.p.m., characteristic of a G-quadruplex structure. The NMR resonances of this major species have sharp line widths (5–6 Hz at 25°C), suggesting a unimolecular G-quadruplex structure. Minor conformations are also present, as indicated by the presence of weak and broader resonances, which account for ∼20–25% of the total population. The melting temperature of this major conformation is around 53°C (Figure S2). The melting temperature is concentration independent as shown by both NMR and CD, indicating that the major G-quadruplex structure is unimolecular.
The presence of twelve imino peaks indicates that the major G-quadruplex structure contains three G-tetrads. Using N-filtered experiments as previously reported (,,,,), the imino (one-bond connection to N1) and base aromatic H8 (two-bond connection to N7) protons of tetrad guanines were unambiguously assigned by site-specific low-enrichment (6%) of 1, 2, 7-N-labeled guanine nucleoside at each guanine position one at a time. The assignment of each imino proton H1 of the twelve guanines involved in the three G-tetrads is shown in . The base H6 and methyl proton resonances of thymines have been unambiguously assigned by substituting deoxyuridine (dU) for each thymine one at a time (Figure S3). We used multiple 2D NMR experiments, including 2D-NOESY, TOCSY and COSY, at various temperatures, to assign the proton resonances of wtTel26 in K. All proton resonances, except some of the H5′/H5′′ protons, were unambiguously assigned ().
It needs to be noted that it is a challenging process to get complete resonance assignment for wtTel26, as there are clearly minor species. To get an unambiguous assignment, we have collected data at various temperatures, 1, 5, 10, 15, 20, 25 and 30°C, with multiple mixing times, in both DO and HO, as the chemical shifts are not only temperature dependent but also solvent dependent. The complete spectral assignment was achieved by examining all of these spectra at different conditions. We made sure that the same NOE assignment pathway for the major conformation can be followed for all experimental conditions, i.e., the connectivities with the adjacent residues can be followed and are consistent at different conditions. For weak NOE peaks, only those that could be consistently observed at all different conditions and that follow the same connectivities with the flanking residues were used for structure determination and calculation.
The assignment of the imino and base H8 protons of guanines leads to the direct determination of the folding topology of the major G-quadruplex structure formed in wtTel26 in K solution (C, right). In a G-tetrad plane with a Hoogsteen-type H-bond network, the imino proton NH1 of a guanine is in close spatial vicinity to the NH1s of the two adjacent guanines, and to the base H8 of one of the adjacent guanines (C). Three G-tetrad planes were determined based on the NOE connectivities of exchangeable protons (A and B). For example, the GH1/GH1 NOE interactions, including G6H1/G10H1, G10H1/G18H1, G18H1/G24H1 and G24H1/G6H1 (A); and the GH1/GH8 NOE interactions, including G6H1/G10H8, G10H1/G18H8, G18H1/G24H8 and G24H1/G6H8 (B), define a G-tetrad plane of G6-G10-G18-G24 (C, right). The overall G-quadruplex alignment is further defined based on the inter-tetrad NOE connections from residues that are positioned far apart in the DNA sequence. For example, the strong NOE interactions of G5H1/G12H1, G11H1/G16H1 and G17H1/G22H1 (A) connect the top two G-tetrads and define their reversed G-arrangements (C, right). The sequential inter-tetrad NOE interactions, including G5H1/G6H1, G17H1/G18H1 and G23H1/G24H1 (A) indicate the same G-conformations of the bottom two G-tetrads (C, right), while the inter-residue NOEs between the two G-tetrads, e.g. G6H1/G11H8, G10H1/G17H8, G18H1/G23H8 and G24H1/G5H8 (B), reflect the right-handed twist of the DNA backbone.
An expanded region for base and sugar H1′ protons of non-exchangeable proton NOESY is shown in C. Five guanine residues are in -conformation, including G4, G10, G11, G16 and G22, as indicated by the very strong H8–H1′ NOE intensities (C). The regular sequential NOE connectivities are either missing or very weak at the N()− G( + 1) steps, i.e. A3–G4, A9–G10, G10–G11, A15–G16 and A21–G22. The characteristic G()H8/G( + 1)H1′ NOEs are observed, including those of G4–G5, G10–G11, G11–G12, G16–G17 and G22–G23 (C). A characteristic downfield shift is observed for the H2′/H2′′ sugar protons of the -guanines (). For the N()− G( + 1) steps, the sequential NOE crosspeak connectivities of the base H8 protons to the 5′-flanking residue sugar H1′/H2′/H2′′ protons, typical for right-handed DNA twist, are clearly observed (C).
The major G-quadruplex in the wtTel26 also adopts a hybrid-type folding structure, but is different from that found in the Tel26 sequence, so we named this new folding the Hybrid-2 type and the first one the Hybrid-1 type (C). Hybrid-2 folding differs from Hybrid-1 in its arrangement of the three TTA loops, which is lateral-lateral-side versus side-lateral-lateral. Thus the first, third and fourth G-tracts of Hybrid-2 folding are parallel while the second G-tract is antiparallel, whereas in Hybrid-1 folding the third G-tract is the antiparallel one. The same number of guanines was observed in the Hybrid-1 folding in the Tel26 sequence (,), with the only difference being the -G11 in Hybrid-2 versus -G17 in Hybrid-1.
Many inter-residue NOE interactions are observed in 2D-NOESY of wtTel26 in K. Critical inter-residue NOE interactions are schematically summarized in , which immediately define the overall structure of the Hybrid-2 G-quadruplex formed in wtTel26 in K. Except for residues A21 and T26 whose H2′ and H2′′ protons are overlapping, all the other residues have resolved H2′ and H2′′ resonances, and the H1′–H2′′ NOE is stronger than the H1′–H2′ NOE while the H1′–H4′ is considerably weaker, indicating the C2′-endo sugar pucker conformations. All residues in the loop regions are in the conformation except T19, which shows strong H8–H1′ NOEs (C) even at a short mixing time. Solution structures of the Hybrid-2 type human telomeric G-quadruplex were calculated using a NOE-restrained distance geometry (DGSA) and molecular dynamics (RMD) approach (, PDB ID 2JPZ), starting from an arbitrary extended single-stranded DNA. A total of 727 NOE distance restraints, of which 324 are from inter-residue NOE interactions, were incorporated into the NOE-restrained structure calculation (). Very few NOEs associated with H5′ or H5′′ were used for structure calculation, due to possible ambiguities associated with H5′ or H5′′ protons. Dihedral angle restraints were used for the glycosidic torsion angle (χ) based on the experimentally determined and conformations. The structure statistics are listed in . Remarkably, like the Hybrid-1 human telomeric G-quadruplex structure (), the Hybrid-2 telomeric G-quadruplex structure is very well defined (A). For the 10 best NMR refined structures, the RMSD is 1.38 Å for all residues except the 5′-terminal T1 ().
A representative model of the Hybrid-2 type human telomeric G-quadruplex structure in K is shown in two different views in B. The guanine distribution of the three parallel G-strands (first, third and fourth) is (5′-), while that of the antiparallel G-strand (second) is (5′-). The widths of the four grooves are 12.98 Å (groove I), 8.62 Å (groove II), 13.56 Å (groove III) and 9.88 Å (groove IV), as measured by the closest P-P distance across groove, with groove III occupied by a double-chain-reversal loop.
A very well-defined T:A:T capping structure is formed at the bottom end of the Hybrid-2 telomeric G-quadruplex structure in K (A). It appears that the 3′-flanking-TT of wtTel26 (A) plays an important role in the favored formation of the Hybrid-2 structure (C, right). T8 and A9 of the first TTA lateral loop, as well as T25 of the 3′-flanking segment, appear to form a T:A:T triple capping structure, with potential H-bonds formed between T8 and T25, and between T25 and A9 (B). Both imino H3 protons of T8 and T25 were observed in NMR even at 15°C. Clear NOEs were observed between T8H3 and T26H1′ as well as T25H1′. The 3′-terminal T26 appears to stack above the T:A:T triple. It is interesting to note that, in the modified Tel26 sequence (A), the A25 of the 3′-flanking-segment was involved in the formation of a stable A:T base pair with T14 and appears to be important for the stabilization of the Hybrid-1 structure (C, left) (,) (see Discussion).
In accord with the T:A:T triple capping structure, T8, A9 and T25 all appear to stack very well with the bottom tetrad ( and A). A9 is positioned largely underneath the G18 base. A9H8 has strong NOEs with both G10H1 and G6H1, while A9H2 has strong NOEs with G10H1 and G18H1 (B). A9H2 shows clear NOEs with sugar protons of G18 (), including H1′ (medium-strong), H2′/′′ (strong) and H3′ (medium), indicating the H2 end of A9 is quite close to the sugar moiety of G18. An unusually large number of NOEs were observed between T8 and G6, including T8Me/G6H8, as well as NOEs between the base protons Me&H6 of T8 and G6 sugar protons, e.g. G6H1′, H2′/′′ and H3′ (). Such NOEs are indicative of two stacking bases. T8Me shows clear NOEs with T26H1′/H4′, as well as T25 H1′ (), consistent with the T:A:T triple capping conformation (A). Clear NOEs were observed between the sugar protons of T8 and G6H1 as well as G10H8 (), indicating the T8 sugar is stacked below G6 and G10 of the bottom tetrad. In fact, most base and sugar protons of T8 and A9 are significantly upfield shifted (), indicating the two residues are covered by the bottom G-tetrad, whose ring-current effect likely induces the resonance upfield shifting of T8 and A9. A9H8 showed unusual medium-strong NOEs with T8H1′/H4′, but not H2′/′′ and H3′, indicating a left-handed conformation at the T8-A9 step with the A9 positioned at the H1′/H4′ side of T8. T7, which is adjacent to G6 of the bottom G-tetrad, appears to fall into groove I at the H1′/H4′ side of the G6 sugar, as supported by clear NOEs between G6H4′/H1′ and T7Me/H6 combined with weak sequential NOEs between T7H6/Me and G6H2′/′′ and H3′ (). Potential H-bonds could be formed between T7O2 and G6NH21, and between T7O4 and G5NH21. The H3 imino proton of T7 was also observed in NMR at low temperatures. The 3′-terminal G24-T25-T26 segment is very well stacked and A), as supported by clear sequential NOE connectivities observed in this region (C), as well as the NOEs between base protons, such as G24H8 with T25H6/Me T25H6 with T26H6/Me, and T25Me with G18H1 (strong) and G24H1 (medium) (). Consistent with the stacking conformation, the base and H1′, H2′/′′ sugar protons of T25 and T26 are markedly upfield shifted ().
As observed in the Hybrid-1 structure C, left, A21 of the double-chain-reversal loop, formed by the third TTA linker in the Hybrid-2 structure (C right and C), appears to position partially above G16 and G22 of the top G-tetrad with its H2 end pointing toward the G-tetrad center, as indicated by NOEs between A21 and G16/G22 (), similar to A9 of the first TTA loop in the Hybrid-1 structure. The base and sugar protons of A21 are downfield shifted (), as also observed for A9 in the Hybrid-1 structure, likely due to their partially stacked position with the top G-tetrad. T20 and T19 appear to position in groove III (C), with T19 positioned closer to the core G-tetrads as indicated by a number of NOEs between T19 and the third G-tract, G16-G17-G18 ().
The loop and flanking residues above the top end of the Hybrid-2 human telomeric G-quadruplex are not as well structured as those covering the bottom end (). A3 of the 5′-flanking segment, T13 and A15 of the second TTA lateral loop, and A21 of the third TTA double-chain-reversal loop, all appear to position right above the top G-tetrad, forming the first layer of capping at the top end; however, no H-bond formations were detected for this layer. A15 appears to stack well with the top tetrad, as indicated by NOEs between A15H2 and G22H1/G16H1, and between A15H8 and G16H1/G12H8 (). T13 appears to stack over G12, as indicated by NOEs between T13H6/Me and the sugar protons of G12, and between T13Me and G12H8/G16H1 (). The methyl end of T13 appears to be close to A15, as a number of NOEs were observed between T13Me and A15H8 as well as A15 sugar protons (). A3 is positioned above the G4 and G12 of the top G-tetrad, with its H2 end pointing toward G12, as indicated by NOE interactions of A3H8/G4H1 and A3H2/G12H1 (). T2 of the 5′-flanking segment and T14 of the second TTA loop appear to position above the first layer, while the 5′-terminal T1 is not very well defined.
WtTel26 contains four TTA segments, including the 5′-flanking TTA and three TTA loops. Adenine residues have been shown to be largely involved in the formation of capping structures, which is an important factor for the stability and thus for the preferred formation of a particular G-quadruplex structure. We have carried out systematic mutational analysis of wtTel26 to determine the functional role of adenines in the stability of the Hybrid-2 telomeric G-quadruplex structure, using A-to-T mutation one at a time for each adenine, including A3, A9, A15 and A21. The 1D NMR spectra of each A-to-T mutant sample are shown in . In accord with the important role of A9 in the T:A:T capping structure (B), mutation of A9T clearly destabilizes the Hybrid-2 structure. Interestingly, it appears that the major imino peaks of wtTel26-A9T resemble those of Tel26 Figure S1, which forms a Hybrid-1 structure C, left, implying the A9T mutation may favor the Hybrid-1 structure. A possible explanation is that while A9T destabilizes the T:A:T triple capping structure in the Hybrid-2 structure, the thymine of the A9T mutant may form a stable H-bonded base pair with A21 and thus favor the Hybrid-1 structure C, left. In addition, while wtTel26 is a mixture of multiple conformations, with the major conformation (∼75%) being the Hybrid-2 type telomeric G-quadruplex structure, the Hybrid-2 type structure only accounts for ∼50% in the T25U sequence with a single T-to-U substitution at position 25 Figure S3. This result implies the importance of T25 for the favored formation of Hybrid-2 structure; and that T8:A9:U25 triple is not as stable as the T8:A9:T25 triple capping structure, as the methyl group of a thymine (T25 here) is likely to be important for the stacking interactions. The important role of T25 was also reflected by the observation that four-G-tract telomeric sequences lacking the 3′-flanking segment (wtTel24, wtTel23) B) appear to form a Hybrid-1 type major structure ().
In contrast, A-to-T mutations at position 3 or 15 markedly improve the NMR spectra (), indicating the formation of a more predominant G-quadruplex structure. The folding structures of both the mutant sequences, A15T and A3T, has been shown to be Hybrid-2 type by unambiguous assignment of the tetrad guanine imino protons using site-specific incorporation of N-labeled guanine Figure S4. It is likely that a stable base-paired capping structure can be formed at the top end by the two mutations, e.g. (A15T):A3 or A15:(A3T), respectively, which further stabilizes the Hybrid-2 structure. In addition, the A21T mutation appears to destabilize the Hybrid-2 conformation ().
As we found that human telomeric sequences can adopt both Hybrid-1 and Hybrid-2 types intramolecular G-quadruplexes, we have carried out systematic analysis to obtain insights into the major G-quadruplex conformations of different wild-type telomeric sequences in K solution. We found that the 3′-flanking segment is important for the formation of a specific G-quadruplex conformation, hence we systematically prepared the wild-type four-G-tract human telomeric sequences with various 3′-flanking segments of null, T, TT and TTA, respectively (B). The 1D H NMR spectra of these sequences in K solution are shown in . It appears that all sequences form a mixture of multiple conformations, with a major conformation existing in most cases. It is clear that the 3′-flanking segment has a determinant role in the major G-quadruplex structure formation in human telomeric sequences, as the four-G-tract sequences with the same 3′-flanking segment give rise to very similar 1D NMR spectra ().
Human telomeric sequences with the 3′-flanking segment of TT (wtTel26, wtTel25b, wtTel24b) (B) all appear to form the major Hybrid-2 structure as observed for wtTel26 (). This observation is in good agreement with the NMR structure of the Hybrid-2 type intramolecular telomeric G-quadruplex, in which the 3′-flanking TT is important for the formation of a stable T:A:T triple capping structure that selectively stabilizes the Hybrid-2 type structure. For the human telomeric sequence wtTel27 with a 3′-flanking TTA segment B), the Hybrid-2 structure still appears to be the major conformation and S5.
Interestingly, while it was shown in our structure that T25 is very important for the Hybrid-2 structure, human telomeric sequences with 3′-flanking-T (wtTel25a, wtTel24a, wtTel23a) gave rise to 1D NMR spectra not very similar to that of wtTel26 and other telomeric sequences with 3′-flanking TT. In fact, those spectra look more like that of wtTel23, which has been shown to form a major structure of Hybrid-1 type (). In order to understand the effect of the 3′-flanking sequences, we decided to determine the major conformation in these four-G-tract telomeric sequences, using site-specific incorporation of N-labeled guanines for unambiguous assignment of tetrad guanine imino protons for each sequence. For wtTel25a with a 3′-flanking T (T25) B), the Hybrid-2 type structure also appears to be the major conformation ( and S6). These results are in good agreement with the important role of the 3′-flanking segment, especially T25, in selective stabilization of Hybrid-2 structure due to the stable formation of a T:A:T triple capping structure. However, while wtTel24a has the same 3′-flanking T as wtTel25a, it appears that wtTel24a forms a mixture of two conformations with similar populations (∼50–50 distribution) ( and S6), as a second set of imino peaks was detected for each N-G labeling. This second set of guanine imino protons appears to be consistent with the Hybrid-1 type structure, and was also detected in the wtTel25a sequence, with a much smaller population (∼25%) ().
Here we have determined the NMR-refined molecular structure of the major intramolecular telomeric G-quadruplex formed in the native 26-nt human telomeric sequence wtTel26 (A) in K solution, which was a Hybrid-2 type telomeric G-quadruplex (C, right). We have very recently determined the NMR structure of the intramolecular G-quadruplex formed in a modified 26-nt telomeric sequence Tel26 (A) in K, which was a Hybrid-1 type telomeric G-quadruplex (C, left) (,). The two sequences contain the same 22-nt core with four G-tracts and differ only in the flanking segments, with Tel26 containing the modified flanking-AA at both ends instead of the native TT (A). It thus appears that while human telomeric sequences form predominantly hybrid-type G-quadruplex structures in K solution, the flanking sequences determine the specific structure, namely, Hybrid-1 or Hybrid-2, formed.
While both telomeric intramolecular structures are hybrid type with mixed parallel/antiparallel strands, they differ in loop arrangements, strand orientations and capping structures (C). The Hybrid-1 structure has sequential side-lateral-lateral loops with the first TTA loop adopting the double-chain-reversal conformation, whereas the Hybrid-2 structure has lateral-lateral-side loops with the last TTA loop adopting the double-chain-reversal conformation. Consequently, in the Hybrid-1 intramolecular telomeric G-quadruplex structure, the top end was covered by the 5′-flanking segment and the third TTA lateral loop, while the bottom end was covered by the second TTA lateral loop and the 3′-flanking segment. In contrast, the top end of the Hybrid-2 intramolecular telomeric G-quadruplex was covered by the 5′-flanking segment and the second TTA lateral loop, while the bottom end was covered by the first TTA lateral loop and the 3′-flanking segment (C).
Our study indicated that the specific hybrid-type intramolecular telomeric G-quadruplex structure was selectively stabilized by specific capping structures. In the Hybrid-2 structure of the native wtTel26 sequence, a novel stable T:A:T triple capping structure was formed at the bottom end by T8 and A9 of the first TTA lateral loop and T25 of the 3′-flanking segment ( and A and B). This T8:A9:T25 capping structure appears to play an important role for the stabilization of the Hybrid-2 structure, as demonstrated by mutational analysis and sequence screening. The T8:A9:T25 triple structure is specific to the Hybrid-2 folding structure (C, right) and thus selectively favors the Hybrid-2 structure. This T:A:T triple is not able to form in the modified Tel26 sequence, as T25 was mutated to A25 (A and C). In fact, the mutated A25 in the Tel26 sequence was shown to be involved in the formation of a different capping structure, the A25:T14 base pair, which is specific to the Hybrid-1 structure (C, left) (). It has been found in our previous study that the mutated A25 is important for the stable formation of the Hybrid-1 structure (). In addition, an adenine triple capping structure was found to form in the Hybrid-1 structure (), which provides additional stabilization specific to the Hybrid-1 structure.
The third TTA linker adopts the double-chain-reversal loop conformation in the Hybrid-2 telomeric G-quadruplex structure, while the first TTA linker in the Hybrid-1 structure adopts the double-chain-reversal loop conformation (C) (). Similar to the observation in our previous Hybrid-1 structure that A9 of the double-chain-reversal loop is positioned partially above the top tetrad (), A21 of the double-chain-reversal loop in the Hybrid-2 structure is also found to position partially above the top G-tetrad (C). The double-chain-reversal loop conformations have been shown to favor short loop sizes of 1 and 2 nt (,,), while a loop longer than 2 nt is in general not as favored for the formation of such a loop conformation, likely due to the lack of stacking interactions of loop residues in the groove region. The specific conformation observed in both Hybrid-1 and Hybrid-2 structures makes the TTA double-chain-reversal loop equivalent to a two-nt sized loop, as only the two thymine residues are positioned in the groove region. This may explain why a 3-nt double-chain-reversal loop can be stably present in the hybrid-type human telomeric G-quadruplexes in K.
A four-G-tract human telomeric sequence is required for the formation of an intramolecular telomeric G-quadruplex. To gain insight into structure polymorphism of human telomeric DNA, we have systematically examined the four-G-tract human telomeric sequences (A and B). We found that the G-quadruplexes formed in these human telomeric sequences in K solution are always in equilibrium between Hybrid-1 and Hybrid-2 conformations. The equilibrium of Hybrid-1 and Hybrid-2 structures appears to be largely determined by the 3′-flanking sequence.
In the wtTel26 sequence with a 3′-flanking-TT (B), the Hybrid-2 structure is the major form in K solution (∼75%) (, top). The same major conformation, the Hybrid-2 structure, appears to form in wtTel25b and wtTel24b (B) with the same 3′-flanking sequence but different 5′-flanking sequences (TA and A, respectively) (). The Hybrid-2 structure also appears to be the major conformation in wtTel27 (B) with a 3′-flanking-TTA segment that contains an additional 3′-terminal adenine ( and S5). In contrast, telomeric sequences wtTel23 and wtTel24 (B), which lack the 3′-flanking segment, appear to form the Hybrid-1 structure as the major conformation () (). This result indicates that the 3′-flanking-T (T25 in our structure), which is necessary for the formation of the T:A:T triple capping structure (A and B), is important for the selective formation of Hybrid-2 type intramolecular telomeric G-quadruplex structure. The telomeric sequences lacking this thymine are unable to form the T:A:T triple capping structure and thus a stable Hybrid-2 structure.
While the 1D NMR spectrum of wtTel25a with a 3′-flanking-T (B) appears to be different from that of wtTel26 (), we found that the Hybrid-2 form is still the major conformation in wtTel25a, although the Hybrid-1 structure can also be detected as a minor conformation ( and S6). The T:A:T triple capping needed for the Hybrid-2 structure can be formed in wtTel25a; however, it appears that the second thymine (T26) of the 3′-flanking-TT in wtTel26 (B) may provide additional stabilization for the T:A:T capping as well as the Hybrid-2 structure (). Interestingly, we found that wtTel24a (B) forms a mixture of Hybrid-1 and Hybrid-2 structures with comparable populations (), although the 1D spectrum of WtTel24a looks quite similar to that of wtTel25a (). WTel24a differs from wtTel25a only in its 5′-flanking segment, which is a TA instead of TTA (B). Whereas wtTel24a contains the 3′-flanking-T that can form the T:A:T triple capping to stabilize the Hybrid-2 structure, it has been shown that the 5′-terminal thymine of the 5′-flanking TA can fold back and form an base-paired capping structure with A21 to stabilize the Hybrid-1 conformation in a modified mutTel24 sequence (A) (). The observation of the Hybrid-2 structure being clearly favored over Hybrid-1 in wtTel25a which contains an additional 5′-T (B) suggests that this A:T base-paired capping structure involving the terminal thymine () is unlikely to form in an extended telomeric sequence. It also needs to be noted that the prediction of G-quadruplex folding topology based on 1D NMR spectral patterns should be made with caution.
It appears that Hybrid-1 and Hybrid-2 conformations coexist in four-G-tract human telomeric sequences in K solution. The energy difference between the two hybrid-type core G-quadruplex conformations appears to be rather small, as the equilibrium of Hybrid-1 and Hybrid-2 structures can be readily shifted by minor changes, such as the lengths or the minor modifications of the flanking sequences. However, the kinetics of the interconversion between the two conformations appears to be slow on the NMR timescale, as very few exchange peaks were observed in NOESY experiments. It may be noted that while the Hybrid-2 form appears to be the major conformation in the elongated sequences (such as wtTel26 and wtTel27), it accounts for only ∼70% in wtTel26 and less than 65% in wtTel27, and the Hybrid-1 form can be detected to some extent in both sequences. It is thus suggested that both forms can form and coexist in the extended human telomeric sequence. Due to the low energy barrier between the two forms, the equilibrium of the two forms may be affected and readily shifted by different factors, such as body temperatures, ion concentrations and protein bindings.
The structure polymorphism and dynamic equilibrium of the human telomeric sequence appear to be intrinsic to its sequence and may be important for the biology of human telomeres. Unlike the G-rich sequences in gene promoter regions (,,), the telomeric DNA sequences are unique in that they contain the same tandem repeats with the same linker segments. Thus in a telomeric sequence, any four-G-tract region has the same potential of forming an intramolecular G-quadruplex structure. The linker segment in the human (and vertebrates) telomeric sequence is TTA, whereas the linker segments in the telomeric sequences of most lower organisms only contain thymines. The presence of adenine in the TTA loops adds an asymmetry in human telomeric sequences, and as a consequence, provides a more selective basis for different capping structures and thus different G-quadruplex conformations. For example, the T:A:T triple capping with T8, A9 and T25 can only be formed in the Hybrid-2 structure, while the adenine triple capping with A3, A9 and A21 can only be formed in the Hybrid-1 structure (C). It is interesting to note that the hybrid-type G-quadruplex is an asymmetric structure; and it is this asymmetry that determines the possibility of forming the two very closely related but yet distinct intramolecular G-quadruplexes, the Hybrid-1 and Hybrid-2 structures.
Significantly, as discussed in our previous report (,), both hybrid-type G-quadruplex structures provide an efficient scaffold for a compact-stacking structure of multimers in human telomeric DNA. The 5′- and 3′-ends of the hybrid-type G-quadruplex structure point in opposite directions, allowing the hybrid-type G-quadruplex to be readily folded and stacked end to end in the elongated linear telomeric DNA strand (D). Intriguingly, the capping structures observed, such as the adenine triple in the Hybrid-1 structure and the T:A:T triple in Hybrid-2 structure, can provide not only stacking interactions between the two adjacent telomeric G-quadruplexes, but also specific binding sites for small molecule ligands. As the energy barrier between the two hybrid-type structures is rather small, it can be easily surpassed by a specific ligand binding, which can change the G-quadruplex conformation. Consequently, the different G-quadruplex structure may disrupt the existing protein interactions and introduce new protein recognitions. The human telomeric sequence has been shown to be highly conserved. Nature may have chosen this specific sequence with its asymmetry and the low energy barrier between different forms, which may provide a means to affect the protein recognition and to control the biology of human telomeres.
In summary, we have determined the molecular structure of the major intramolecular G-quadruplex formed in the native, non-modified 26-nt human telomeric sequence wtTel26 in physiological K solution. Our study revealed a hybrid-type intramolecular G-quadruplex structure (Hybrid-2), which is different from the hybrid-type (Hybrid-1) structure that has been recently determined in our lab (,), in loop arrangements, strand orientations and capping structures. The distinct capping structures appear to determine the favored formation of a specific hybrid-type telomeric G-quadruplex structure, and may provide specific binding sites for drug targeting. The hybrid-type G-quadruplex structures appear to be the major conformations formed in four-G-tract human telomeric sequences in K solution, with a dynamic equilibrium between Hybrid-1 and Hybrid-2 conformations that appears to be largely determined by the 3′-flanking sequence. Both hybrid-type G-quadruplex structures suggest a straightforward means for multimer formation with effective packing in the human telomeric sequence and provide important implications for drug targeting of G-quadruplexes in human telomeres.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
DNA and RNA molecules naturally occurring in living cells derive from nucleoside triphosphates, through an iterative catalytic process condensing the nucleoside 5′-phosphate moiety of these activated substrates to the 3′-hydroxyl moiety of the elongating strand and releasing pyrophosphate. A crucial aspect of this polymerization process is that the pyrophosphate leaving group undergoes hydrolysis under the action of pyrophosphatase, the essential enzyme in charge of the irreversibility of macromolecular biosynthesis (). The energetics of protein synthesis is also based on the release and subsequent destruction of pyrophosphate through the transient formation of amino acyladenylates by amino acyl-tRNA synthetases (). As an anhydride made up of two identical phosphoryl moieties, pyrophosphate is endowed with a compositional simplicity that provides the basis of straightforward recycling processes for reconstituting the pools of RNA and DNA precursors through P-O bond rearrangements catalyzed by nucleotide kinases and nucleoside diphosphate kinase (). Altogether, cells have evolved an efficient network of enzymatic phospho-transfers so as to reload phosphoanhydride bonds in nucleotides using the potential energy of activated intermediates in metabolism, e.g. phosphoenolpyruvate, or using the chemo-osmotic potential of the cell membrane proton gradient (). This coupling between energy storage and nucleic acid polymerization through phosphoanhydride formation and pyrophosphate release lies at the core of cells chemical, energetic and genetic design. If, for synthetic biology purposes, one attempted to diversify enzymatic polymerization of nucleic acids by condensing an additional category of activated nucleotides bearing no pyrophosphate leaving group, the design of such precursors would have to integrate features similar to those embodied in the economy and simplicity of phosphoanhydride metabolism. As a first step toward experiments along these lines, we set out to explore metabolic prototypes of activated DNA precursors such that: (i) they would serve as substrate for a DNA polymerase; (ii) their consumption in DNA biosynthesis would release as leaving group a metabolite common in cells; (iii) the leaving group could be actively degraded or recycled so as to enforce irreversibility of polymerization and (iv) the leaving group could facilitate uptake of activated DNA precursors through cell membranes.
Few successful attempts have been reported in the literature to substitute functionally the pyrophosphate leaving group of nucleoside triphosphates in nucleic acid polymerization. Phosphorimidazolides and their 2-methyl derivatives have been the topic of a systematic investigation of non-enzymatic polymerization of canonical and non-canonical nucleotides in prebiotic studies (). Systematic substitution at the beta and gamma positions of deoxynucleoside triphosphates has been studied by Krayevsky and collaborators (). Remarkably, dTTP analogues bearing bulky hydrophobic groups at the gamma position of dNTPs were shown to undergo polymerization catalyzed by HIV reverse transcriptase and other viral and cellular DNA polymerases (). More recently, deoxynucleoside triphosphate analogues with a P-C-P distal bridge replacing the P-O-P phosphoanhydride have also been demonstrated to undergo condensation by DNA polymerase beta (). Among conceivable P-O, P-S, P-C and P-N conjugates of deoxynucleoside monophosphates with physiological metabolites, phosphoramidate conjugates with amino acids (alias dNAP) seemed particularly worthy of interest because their structure lends itself to activation by the catalytic Mg of DNA polymerases, much as deoxynucleoside triphosphates (). Their condensation by polymerases is designed to release common amino acids, which can be recycled or destroyed in subsequent metabolic steps. In addition, they offer an easy synthetic access and a high enough stability toward spontaneous hydrolysis under physiological conditions (37°C, pH = 7.5).
The second aim of this research project is to bypass the kinase pathway for intracellular activation of modified nucleosides. Nucleoside reverse transcriptase inhibitors (NRTI) are designed to be recognized as substrates for RT and incorporated into a growing strand for further termination of chain elongation (). Inhibition of reverse transcriptase activity and chain termination by NRTI's is achieved by introduction of structural modifications in the sugar moiety. These RT inhibitors are usually administered as biologically inactive free nucleosides or nucleoside phosphonates, or as nucleoside monophosphate/phosphonate prodrugs where the phosphate moiety is masked with a lipophilic group (). In the case of nucleoside administration, three steps of kinase-mediated activation are needed to generate the triphosphate in the cell. In the prodrug concept, nucleoside 5′-monophosphate kinase and nucleoside 5′-diphosphate kinase activities are needed to provide the biological active congener (). The efficiency of these enzymatic phosphorylation reactions depends on the substrate specificity of the different kinases. A nucleotide analogue that would not depend on activation by nucleoside/nucleotide kinases whilst serving as a natural substrate mimic, would be of great interest.
Here, we present a series of amino acids phosphoramidate analogues in which a natural -amino acid moiety is linked through a P-N bond to 2′-deoxyadenosine 5′-O-monophosphate () and which might serve as potential leaving group in a nucleotidyl transfer reaction. Especially, the 5′-aspartyl-phosphoramidate and the 5′-histidyl-phosphoramidate can mimic a natural triphosphate moiety quite effectively for the incorporation of 2′-deoxyadenosine into a growing DNA strand by HIV RT and Therminator DNA polymerase. As demonstrated by modelling studies, the amino acid moiety of this deoxynucleotide analogue provides structural and electrostatic features essential for salt formation and/or metal coordination and assembly of the catalytic residues in the polymerase active site. In this respect, this study might contribute to a better understanding of the mechanism of biological polymerization reactions.
The synthesis of methyl ester amino acid phosphoramidate nucleotides analogues was accomplished according to the method described by Wagner and colleagues starting from nucleoside monophosphate (). The details of the synthesis of the compounds used in the study have been described in a preliminary communication of this work (). -amino acids were used for synthesis of the first series of phosphoramidate analogues. The deprotection of the amino acid moiety was carried out with 0.4 M sodium hydroxide in methanol-water solution. The series of phosphoramidate analogues coupled to a variety of natural -amino acids, synthesized for the study, is shown in .
HIV reverse transcriptase is involved in copying of the HIV genome and uses deoxynucleotides as substrates. HIV RT is an error-prone polymerase and has high mutation rate (). The essential role of the HIV RT in viral replication and its flexibility and tolerance toward modified nucleotides renders this enzyme a primary target in treatment of HIV infection. In the presented study, the ability of HIV RT to incorporate a series of amino acid phosphoramidate nucleotide analogues was investigated by the gel-based single nucleotide incorporation assay (,).
Among amino acid phosphoramidate nucleotides () remarkable results were observed with Asp-dAMP () ().
This phosphoramidate analogue was recognized by HIV RT and efficiently incorporated into a growing primer strand resulting in 90% conversion to an ( + 1) strand in 60 min (500 μM nucleotide concentration). At the same conditions, incorporation of His-dAMP(), Gly-dAMP () and Pro-dAMP () were 1.5-, 6.5-, and 3.7-fold less efficient, respectively. Efficient incorporation of Asp-dAMP (24.1%) was also observed when the substrate concentration was decreased 10-fold. However, significantly lesser incorporation of amino acid phosphoramidate was detected for nucleotides coupled to non-polar, hydrophobic amino acids. Ala-dAMP and Tyr-dAMP behaved as poor substrates leading to merely 7- and 10-fold reduction in primer extension, respectively ().
Interestingly, no incorporation occurred when respective methyl ester derivatives of (dimethylester for and as also the carboxylic acid function in the side chain is methylated) were used as substrates in the polymerase reaction (, Panel II). An unexpected result was observed with Glu-dAMP analogue () that also acted very poorly as an HIV RT substrate. These observations suggest that recognition and incorporation of amino acids (AA) dAMPs is a very specific process and is likely to be dictated by the chemical structure and electrostatics of the amino acid moiety.
Since it is known that polymerases have tendency to incorporate dAMP in a non-template manner, we investigated whether the former observations were due to a true base-pair extension. A control experiment with a mismatch sequence (A against A) was carried out (, Panel III). As expected, Asp-dAMP () was not incorporated at all into the growing primer strand. After 2 h of the polymerase reaction at 500 μM substrate concentration 0% conversion was observed. The same results were observed when the substrate concentration was increased to 1 mM.
Another polymerase enzyme that demonstrates similar trends in recognition and utilization of AA-dAMPs is the thermostable Therminator DNA polymerase, a variant of (9°N−7) DNA polymerase. This enzyme demonstrated effective recognition and incorporation of a number of nucleotides bearing unnatural nucleobase and sugar moieties (). Likewise, probing of AA-dAMP incorporation directed by Therminator DNA polymerase revealed property of analogues , and to act effectively as alternative substrates in the DNA polymerization reaction ( and ).
Yet again, the best results were obtained with Asp-dAMP, which led to 25.2% primer extension over 60 min at 500 μM nucleotide concentration. At the same conditions, similar results were obtained for Gly-dAMP and His-dAMP (26 and 25.4% primer extension, respectively). In the case of Glu-dAMP and methyl protected AA-dAMPs, Therminator DNA polymerase displays selectivity analogous to HIV reverse transcriptase and fails to direct incorporation of those phosphoramidate analogues ().
The remarkable property of Asp-dAMP encouraged further investigation and testing as a substrate for other DNA polymerases. However, in the case of Taq, Vent (exo), and KF (exo) DNA polymerases, recognition and incorporation efficiency were significantly less appealing. Incorporation and primer extension were observed only in the case of KF (exo) DNA pol demonstrating 32.5% conversion of the primer strand in 60 min. This is in contrast to Taq and Vent (exo) DNA polymerases that failed to insert into a growing primer strand. The diversity in incorporation selectivity that are observed among the polymerases [Therminator, Taq, Vent (exo), KF (exo) and HIV reverse transcriptase] could indicate the differences in the active site flexibility and tolerance to the triphosphate modifications (). Although no data from human polymerases are yet available, the selectivity for HIV-RT points to the potential of this approach for the design of direct reverse transcriptase inhibitors as potential anti-HIV agents.
Among this series of phosphoramidate nucleotide, the most encouraging results using T3 template were observed with Asp-dAMP and His-dAMP which were used by HIV RT to extend a primer with three adenine nucleobases ( + 3 product). The results of these gel electrophoresis experiments are described in the preliminary communication ().
However, after 60 min of the polymerase reaction the ( + 2) product predominates over the ( + 3) product (56.3% versus, 5.2% for Asp-dAMP and 67.1% versus 13.5%). Interestingly, efficiency of DNA synthesis with His-dAMP at 500 μM substrate concentration is similar or better to that when Asp-dAMP serves as the substrate (67.1% versus 56.2%, respectively, for the synthesis of the ( + 2) primer). This is in contrast to the single nucleotide incorporation results that indicate that His-dAMP is less good than Asp-dAMP as a substrate for HIV RT (). Efficiency of chain elongation and single nucleotide incorporation is dependent on the structure of the ‘leaving group’.
In the case of the T5 template with the overhang of seven thymidine nucleobases, HIV RT indeed generates ( + 6) and ( + 7) products at a very little extent while the ( + 2) and ( + 3) products are prevalent (). The obvious stalling of the HIV RT polymerase after incorporation of two adenine nucleobases might indicate substrate inhibition or a template sequence effect. The primer strand extension for 1 h with 500 μM of Gly-dAMP or Pro-dAMP takes place with low efficiency and does not result in the formation of the full-length extension products.
The Therminator DNA pol mediated addition of amino acid phosphoramidate nucleotides instead of natural dNTPs at the 3′ terminal end was investigated for several AA-dAMPs. Similarly, to the case of HIV RT, the best results were observed with Asp dAMP phosphoramidate, which was successfully incorporated across from a string of thymidine residues (T3 template) to provide an ( + 3) product ().
However, when His-dAMP and Gly-dAMP were used as substrates for Therminator DNA polymerase, the primer extension took place with significantly lesser efficiency (13.6 and 18.1% of primer extension, respectively) with the ( + 1) product being predominant and halted after addition of 2 nt phosphoramidate residues (). The primer extension with Pro-dAMP was very ineffective and resulted only in addition of 1 nt phosphoramidate residue at the primer's end.
It is interesting to note that in the case of the T7 template with the overhang of seven thymidine residues the predominant product of the primer extension was the ( + 2) oligonucleotide. Nonetheless, Therminator DNA polymerase was able to carry out the extension of the T7 primer with Asp-dAMP phosphoramidate and incorporate up to five adenine residues.
In order to evaluate the influence of chirality of the amino acid to function as a leaving group during the incorporation reaction, we have compared -Asp-dAMP and -Asp-dAMP as substrate for RT using template T1 and primer P1.
From the gel electrophoresis experiments it is clear that the natural -Asp-dAMP is a better substrate for RT than -Asp-dAMP (). The stereoselectivity of the reaction further support the idea that appropriate binding and coordinating of the amino acid in the active site of the enzyme is needed, in order to function as substrate.
From the point of view of applicability in the antiviral field, we have evaluated the potential of -Asp-PMEA (PhosphonoMethoxyEthylAdenine) to function as chain terminator in the RT assay. PMEA is a potent anti-HIV agent with a phosphonate moiety instead of a phosphate group and lacking a free hydroxyl in the nucleoside moiety, so that chain elongation is not possible (). The incorporation of PMEA using -Asp as leaving group is demonstrated in , although high concentration of the substrate is needed (100 µM). It can be concluded that -Asp-PMEA is a substrate for RT and incorporation of PMEA leads to chain termination. This experiment demonstrates the potential of the approach to use modified nucleosides having an alternative leaving group as potential anti-HIV agents.
With the -Asp-dAMP and -His-dAMP molecules bound to reverse transcriptase, stable molecular dynamics (MD) trajectories were obtained.
In the -Asp-dAMP complex (b), the 2 Mg ions are comparable in position to their situation in the original TTP complex (a). They are tightly bound to both COO groups of the -Asp-dAMP (mimicking the second and third phosphate groups), the phosphoramidite group and to three Asp groups in the enzyme: Asp110A and Asp185A which are widely conserved in the polymerases and Asp186A (,). In the original X-ray structure, Asp186A is not involved in an ionic bond with one of the Mg ions. In the -His-dAMP complex MD simulation reveals that the -His-dAMP binds in a different manner ().
The Mg ions shift during the simulation, i.e. Mg601A moves 3.44 Å to Asp186A, while Mg600A moves 2.51 Å in the direction of Asp185A. There is an ionic bond from one phosphate oxygen to Mg, and the coordination of the Mg ions to three Asp residues is still present. The amino acid COO of -His-dAMP however is involved in a salt bridge with Lys65A. The neutral His group of -His-dAMP is in an orientation facilitating a cationic-aromatic interaction with one of the Mg ions ().
Only a few dAMP residues are incorporated into the primer, when feeding the enzyme with -Asp-dAMP. An explanation could be that the leaving group is bound too tight to the enzyme so that it stays in the enzyme, becoming an obstruction for new entering residues. When mutating the -Asp-dAMP into -Glu-dAMP, this glutamine side chain is too far from the Mg ions to interact. Although the -Glu-dAMP molecule fits into the NTP binding pocket (model not shown), no conclusion on why this molecule is not incorporated in the chain can be drawn based on that model.
and are generated using Bobscript, Molscript and Raster3d ().
In order to evaluate whether the low incorporation level using some thermostable polymerases is not due to the instability of the compound at higher temperature, we have determined the chemical stability of dNAP in different conditions. The stability of -Asp-dAMP was investigated using 1D P NMR spectra. Within a pH range of 6–8 and at a temperature of 25°C no degradation could be observed after a period of 2 days. At pH 7 and 8.8, at a temperature of 70°C, the product degraded following first order reaction kinetics with a half life of 3.3 h = and 2.8 h, respectively.
A considerable amount of work has been done in the last two decades exploring DNA polymerase mechanisms and the factors involved in nucleotide recognition and polymerase fidelity (). Importantly, these studies have provided extremely valuable information for rational design of anti-viral, anti-cancer compounds as well as nucleoside probes for various biotechnological applications (). The presented study, hence, demonstrates the successful and efficient use of the amino acid moiety linked to the α phosphate via a (P-N) phosphoramidate bond as a diphosphate (pyrophosphate) group mimic.
This study revealed that the most remarkable recognition and nucleotide analogue incorporation was achieved with 2′-deoxyadenosine-5′-aspartyl phosphoramidate nucleotide analogue. Gel-based polymerase assay shows that this analogue is successfully inserted into a growing DNA strand by HIV Reverse Transcriptase and Therminator DNA polymerase and that this incorporation is selective. It was also demonstrated that HIV RT is capable of incorporation of 2–3 consecutive residues of in template-directed DNA synthesis. Likewise, Therminator DNA polymerase also efficiently extends a DNA primer by several nucleobases using the modified substrate. Notably, in the case of HIV RT, the observed stalling and termination of the DNA synthesis after incorporation of two Asp-dAMP residues could possibly indicate enzyme inhibition by either a competitive or a non-competitive mechanism. Incorporation of the modified substrate and the primer extension were observed with His-dAMP analogue. Although His-dAMP was inserted at a lesser extent than Asp-dAMP, primer extension was comparable to the extension with and also resulted in a noticeable stalling at the +2 position. Thus, it would be of a great interest to explore the causes of the stalling and possible modes of substrate inhibition.
The kinetic analysis of Asp-dAMP incorporation shows that the specificity for incorporation of this modified substrate by HIV reverse transcriptase is ∼1300-fold lower than that for the natural substrate (dATP) (). The significantly higher value for the amino acid phosphoramidate analogue than for the natural substrate, suggests that the phosphoramidate substrate dissociates from the active site more readily and faster. Thus, high value for Asp-dAMP implies weak binding to the polymerase active site. However, the measured is only 3-fold lower than this for the natural substrate suggesting fast and efficient nucleophilic displacement of amino acid moiety once the amino acid phosphoramidate substrate is bound at the active site and formation of a phosphodiester bond. Therefore, it might be feasible to design an amino acid phosphoramidate analogue with higher affinity for the polymerase active site while retaining structural features responsible for efficient recognition and nucleotidyl transfer.
To be useful in biological experiments, it is important that the amino acids-dAMP analogues are chemically and enzymatically stable. The enzymatic stability has not been evaluated, yet. However, the compounds have been proven to be stable in water between pH 6 and pH 8 for at least 2 days, their half life at pH 7 and pH 8.8 at 70°C is 3.3 and 2.8 h, respectively. These stability studies have been performed using NMR spectrometry to follow the potential degradation reactions (data not shown). It is expected that the phosphoramidate bond will become unstable at a pH lower than 5. Exploration of the substrate properties among a series of nucleotide phosphoramidates coupled to a wide spectrum of natural amino acids revealed several important structural and electrostatic features involved in triphosphate moiety binding, recognition and nucleotidyl transfer. Numerous structural studies indicate that the binding of the nucleoside triphosphate at the polymerase active site results in coordination of at least two catalytically essential metals (Mg) (). Furthermore, several catalytic, highly conserved amino acid residues are also involved in chelation of metal ions and proper positioning of the triphosphate moiety for an ‘in-line’ nucleophilic attack at the α phosphorus. It was suggested previously that initial recognition of an incoming dNTP occurs through the binding of the triphosphate moiety (). Structural and genetic analysis of a number of DNA polymerases and reverse transcriptase indicates that amino acid residues involved in the triphosphate binding are highly conserved (,,). In the case of HIV RT, binding of the incoming dNTP is coordinated by Arg72 and Lys65 that make interactions with the α- and γ-phosphates, respectively (a) (). This dNTP is also accompanied by 2 Mg ions, which are bound to the phosphates of the nucleotide and to the two residues Asp185 and Asp110 (a) (,).
Importantly, studies have shown that the binding of the incoming dNTP and catalytic metal ions is responsible for further rearrangements of the catalytic amino acid residues as well as the relocation of the 3′ primer terminus in a position for the effective nucleotidyl transfer (). Therefore, a proper geometric and spatial arrangement of all reacting residues and atoms are essential for the formation of the productive tertiary complex (,,). Efficient incorporation of might imply that the aspartyl amino acid effectively replaces the β and γ phosphate groups, likely. It can also be suggested that aspartate moiety acts as a leaving group and mimics of a pyrophosphate group. Another example of activated nucleotide and a use of a good leaving group is phosphorimidazolide nucleotides, in which the α-phosphorus atom is activated by imidazolide or methyl imidazolide moieties. Such phosphoimidazolides deoxy- and ribonucleotides were described as efficient substrates for non-enzymatic templated and non-templated oligomerization ().
Furthermore, the presented study clearly demonstrates the requirement for the presence of the negative charge and electrostatic interactions for efficient binding of an incoming nucleotide. This is evident from the study with the aspartyl phosphoramidate () and bis-methoxy aspartyl phosphoramidate () where the protection of a carboxylate groups brings drastic changes in ability of HIV RT or Therminator DNA pol to recognize and incorporate these two modified substrates. Although aspartyl phosphoramidate nucleotide behaves as a good nucleotide triphosphate analogue and substrate for HIV RT and Therminator DNA polymerase, its methyl protected derivative does not support DNA synthesis at all. Comparison among the series of synthesized amino acid phosphoramidates demonstrates that Asp-dAMP, which possesses an extra negative charge, displays superior properties as a polymerase substrate. The lower efficiency of DNA synthesis using -Asp-dAMP (when compared with dATP) may also be attributed to these electrostatic effects (three negative charges for -Asp-dAMP versus four negative charges for dATP, respectively), and the suggestion that the incoming dATP brings 1 Mg in the active site, while the pyrophosphate leaving group takes 1 Mg with it (). Unexpectedly, Glu-dAMP analogue that also possesses an extra negative charge failed to serve as a substrate for both HIV RT and Therminator DNA polymerase. A possible explanation for this outcome could be steric clashing (crowding) due to a longer amino acid side chain (an extra methylene group) as compare to the Asp-dAMP analogue. However, a modelling study (data not shown) demonstrates that Glu-dAMP can also be accommodated in the active site of the enzyme complex. Apparently, the binding of Glu-dAMP is not optimal to function as substrate in the polymerization reaction. It can also lead to misalignment of the negative charge in the polymerase active site and disruption of the binding of the catalytic metal ions. HIV RT and Therminator DNA pol were effective in incorporation of His-dAMP suggesting that imidazole moiety of the histidine side chain might likewise be involved in binding interactions of metal ions and the triphosphate group. Interestingly, a modelling study demonstrates that His-dAMP is bound in a different way as Asp-dAMP and that the α-carboxylate group is involved in interactions with Lys-65 and Arg-72, while the imidazole ring of His-dAMP is involved in cation-π interactions.
Moreover, very little or no nucleotide incorporation occurred when phosphoramidate nucleotides coupled to non-polar amino acids were used as substrates for HIV RT or DNA polymerase. These observations provide further support for the rationale that the electrostatic interactions are indispensable for polymerase recognition, the assembly of the catalytic complex in the active site and for chain elongation. These suggestions, however, need to be confirmed by co-crystallization experiments and molecular dynamics.
In conclusion, aspartyl phosphoramidate moiety serves as a mimic of a pyrophosphate group and behaves as a good leaving group in a nucleotidyl transfer reaction. Incorporation of nucleotides, although to a lesser extent, was likewise observed for histidyl and glycinyl phosphoramidates, respectively. The fact that different AA-dAMP function as substrate for the polymerase reaction, does not mean that they bind to the enzyme in the same way and that their mode of action is identical.
Therefore, is seems feasible to use chain terminating nucleotide analogues coupled to newly designed leaving groups through a phosphoramidite or phosphodiester linkage for a direct inhibition of HIV-RT or other viral polymerase. Another application may be the enzymatic synthesis of DNA containing natural and unnatural nucleobases, avoiding at times cumbersome nucleoside triphosphate synthesis and purification. As we emphasized elsewhere (), the propagation of certain genes and replicons that would require the exogenous supply of nucleic acid precursors absent from cells and natural food chains, such as deoxynucleotide-aspartate conjugates, stands as a promising option for preventing genetic pollution by nutritionally containing the dissemination of genetically engineered microbes. The logical next step toward implementing dNAPs, e.g. the set of deoxynucleotide-aspartate conjugates corresponding to the four bases A, C, G and T, as precursors of certain DNA plasmids in a bacterial cell will be to evolve a DNA polymerase variant with an enhanced efficiency and selectivity in condensing these substrates but having lost the capability to condense canonical dNTPs. Techniques of computational design () and of evolution () could be applied for accomplishing such a swap of substrate specificity in the active site of HIV reverse transcriptase and other DNA polymerases.
The synthesis and analysis of all -AA-dAMP compounds have been described in the Supplementary Data of the first communication of this research ().
The assays for single nucleotide incorporation by HIV reverse transcriptase and the steady state kinetics has been described ().
In the case of Therminator DNA polymerase, the enzyme was obtained from Westburg (NEB) (2 U/µl) and the reactions were carried out in a 10X Thermopol reaction buffer containing 20 mM TRIS-HCl, 10 mM KCl, 2 mM MgSO, 0.1% Triton X-100, pH 8.8. The final concentration of the Therminator DNA pol in the reaction mixture was 8.33 × 10 U/µl. The polymerase reaction involving Therminator DNA polymerase or any other thermostable polymerase [Vent (exo), Taq DNA polymerase] was carried out in a similar way as the HIV RT reaction with some modifications. The dNTP solutions and the primer/template/DNA polymerase mixture solutions were topped with mineral oil (30–60 µl) and pre-incubated at 70°C for 2 min. The polymerase reactions were performed at 70°C as well.
#text
xref
#text
xref
sc
#text
xref
sup
#text
NMR Spectra were recorded on a Bruker Avance II 500 NMR spectrometer. Chemical shifts δ are indicated in ppm relative to the solvent signals (1H and 13C) or HPO as external standard (P).
The P resonance of -Asp-dAMP could be observed at 6.90 p.p.m. and the assignment was confirmed using an H detected P-H COSY in which a clear scalar coupling could be observed between the P resonance and the α-proton of the -Asp moiety and between the P resonance and the 5′ protons of the deoxyribose moiety.
Degradation of -Asp-dAMP was monitored at pH 8.8 and pH 7 using the integral of the compound's 31P resonance peak. After each period of 45 min at 70°C the sample was cooled to 25°C and a P spectrum was recorded. The degradation could be fitted to , which indicates that the degradation follows a first order kinetics. represents the integral of the P resonance signal after one or several periods of 45 min at 70°C. is the integral of the P resonance signal at the start of the degradation and was set to 100.
Plotting the natural logarithm of the integrals against time allowed us to extract the degradation constant and half-life at both pH conditions ().
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
Protein–RNA complexes (ribonucleoprotein particles, RNPs) play a fundamental role in the control and regulation of gene expression in the eukaryotic cell. They participate in essential cellular processes such as pre-mRNA splicing, rRNA maturation, post-transcriptional control (mRNA stability), RNA export, translation and translational control. In the field of alternative splicing and of translational control by microRNAs, it was recently demonstrated that protein–RNA interactions and their dynamic changes provide a basis for the diverse and complex driving forces behind such processes ().
There are various approaches to identifying the proteins involved in these processes. One is the overall analysis of the proteins associated with the complexes by mass spectrometry [MALDI-MS (), Electrospray Ionisation (ESI)-MS ()], as was recently demonstrated by several proteomic studies of RNP complexes that play fundamental roles in (alternative) splicing () and siRNA- and miRNA-mediated translational repression (). However, in proteomic-driven studies, no information is gained regarding the question of which of the identified components interacts directly with RNA. A straightforward approach to identify proteins in direct contact to their cognate RNAs is protein–RNA cross-linking combined with MS (). An alternative/additional method for mapping protein–RNA interactions using MS is the dissociation of intact protein–RNA complexes in the mass spectrometer and the analysis of components that are still associated with RNA (,).
One possibility for protein–RNA cross-linking is the direct UV-irradiation of RNPs at 254 nm (), based on the natural UV-reactivity of the RNA nucleobases. Upon excitation, a covalent bond between a nucleobase and an amino-acid side chain of a protein is formed. This approach has several advantages over site-specific labelling (,) or over using heterobifunctional reagents [e.g. (,)]: (i) It can be applied directly to any native protein–RNA complex isolated from cells without reconstituting particles carrying site-specific cross-linkers (which can lead to a heterogeneous population and/or can reduce the yield of complexes for interaction studies). (ii) Zero-length cross-links have been proven to have a very high specificity, as demonstrated recently by the 3D structures of co-crystallized RNA–protein complexes (,). The site of contact identified in this way always reflects a structural (and functional) RNA interaction domain within or very close to the RNA-binding domain of the protein (). (iii) It obviates extensive probing experiments, in contrast to comparable protein–RNA cross-linking studies using heterobifunctional reagents, in which the optimal probing conditions have to be carefully adjusted (). (iv) No inter- or intramolecular protein–protein cross-links are generated (at least as reported so far), which reduces the number of putatively cross-linked species within the mass spectra and simplifies their interpretation.
However, combining protein–RNA cross-linking with mass spectrometry encounters several challenges: (i) because the yield in UV cross-linking is low(er), a purification strategy must be established that separates the cross-linked species from the excess of non-crosslinked species. (ii) MS must be adapted, since the peptide and RNA moieties of the cross-linked conjugates show divergent physico-chemical properties in the analysis. (iii) Enrichment and/or down-scaling strategies are required, both to reduce the amount of starting material (and thus to allow the study of material directly isolated from cells) and to increase the intensity of the peaks from (low-abundance) protein–RNA cross-links in MS.
In recent years, we have established a strategy for the purification and subsequent MALDI-Time-of-Flight (MALDI-ToF) MS analysis of cross-linked peptide–oligoribonucleotides derived from UV-irradiated native and reconstituted ribonucleoprotein particles (, and below). It comprises: digestion of the protein moiety of cross-linked RNPs with endoproteinases, removal of the excess of non-crosslinked peptides by size-exclusion chromatography, hydrolysis of RNA-containing fractions with RNases and subsequent fractionation of the resulting mixtures on a microbore liquid-chromatography (LC) system. Fractions that showed an absorbance at 220 nm (peptide moiety) and 254 nm (RNA moiety) were considered to contain cross-linked species and were subsequently analysed by MALDI-MS and -MS/MS using 2,5-dihydroxybenzoic acid (DHB) and/or 2,4,6-trihydroxyacetophenone (THAP) as matrices ().
On the basis of this work, we report here a down-scaling/enrichment strategy of cross-linked peptide–RNA oligonucleotide species from low amounts of starting material (≤50 pmol) obtained from UV-irradiated RNPs. The novel approach comprises enrichment of peptide–RNA cross-links by immobilized metal-ion affinity chromatography (IMAC) from capillary RP-HPLC fractions combined with treatment of the enriched species with calf intestinal alkaline phosphatase (CIP) to exclude false positives and the subsequent MS analysis by MALDI-ToF mass spectrometry.
In feasibility studies, we successfully applied this strategy to the detection of several peptide–RNA oligonucleotide heteroconjugates derived from (i) UV-irradiated partial complexes of the human minor spliceosome (), i.e. [15.5K-61K-U4atac snRNA] complexes (,), and (ii) from UV-irradiated native U1 small nuclear ribonucleoprotein (snRNP) particles () of the human major spliceosome () that had been studied before (,,). Moreover, we were able to enrich cross-links derived from a UV-irradiated [p14/SF3b14a-SF3b155] protein complex bound to a U2 snRNA oligomer that mimics the branch-site interacting region (BSiR) of the U2 snRNA ().
Reconstituted [15.5K-61K-U4atac snRNA] complexes () and native U1 snRNPs () were obtained and UV cross-linked as described previously (). [p14/SF3b14a-SF3b155-U2 snRNA BSiR] particles were reconstituted and UV-irradiated according to (). The protein moiety of the particles was digested with endoproteinase trypsin (Promega) or chymotrypsin (Roche) (ratio 1:20 for both) and the RNA with RNase T1 (Ambion; ratio 1:15) or RNases T1 and A (Ambion; ratio 1:20 for both). The applied purification scheme for the cross-links was according to () with the exception for the [p14/SF3b14a-SF3b155-U2 snRNA BSiR] particles. In this case, the non-crosslinked peptide moiety was separated from the non-crosslinked and cross-linked U2 snRNA BSiR oligomers by size exclusion chromatography on a Superdex Peptide column (300 mm × 3.2 mm) mounted on a SMART system (all GE Healthcare, Uppsala, Sweden). Peptide–RNA heteroconjugates were subsequently purified from the mixture by reversed-phase liquid chromatography (RP-LC) with an RP C18 column (150 mm × 0.3 mm; MicroTech Scientific, Vista, USA) coupled to a 140C microgradient system (Applied Biosystems, Foster City, USA) running at a flow rate of 2 µl/min. A gradient of water/0.1% trifluoroacetic acid (TFA) (solvent A) and 80% acetonitrile/0.085% TFA was used. Fractions of 8 µl volume were collected, evaporated to dryness and subjected to further analysis.
Peptide–RNA oligonucleotide cross-links were enriched from one-half of the cap-LC fractions by IMAC using POROS 20 MC beads (Applied Biosystems) loaded with Fe(III) ions (). For this, capillary LC fractions were redissolved in 50% acetonitrile (ACN) with 0.5% acetic acid (HOAc); one-half was incubated with the IMAC bead slurry for 30 min and the other half retained for hydrolysis (see below). The beads were subsequently washed with 25, 50 and 75% ACN in 100 mM HOAc. Bound cross-links were eluted from the beads with 1% phosphoric acid and subjected to MALDI-MS analysis. Peptide–RNA cross-links of the remaining half of the cap-LC fractions were IMAC-enriched as before and additionally incubated with 1 U of CIP (New England Biolabs) on Fe(III)-loaded POROS 20 MC beads. The beads were washed once with 50% ACN in 100 mM HOAc. Enriched and phosphatase-treated cross-linked species were subsequently eluted from the beads with 1% phosphoric acid as before and subjected to MALDI-MS analysis.
Cross-linked samples were analysed on a Reflex IV mass spectrometer (Bruker Daltonik GmbH, Bremen, Germany) with 2,5-dihydroxybenzoic acid (DHB) as MALDI matrix in the positive reflectron mode for the U1 snRNP and the [15.5K-61K-U4atac snRNA] particles and in the negative reflectron mode for the [p14/SF3b14a-SF3b155-U2 snRNA BSiR] complex, both under standard conditions, by summing up to 1000 laser shots.
We therefore introduced a capillary liquid chromatography (cap-LC) system for the final purification step of the peptide–RNA heteroconjugates [instead of the microbore HPLC used in previous experiments ()]. This allowed us to reduce drastically the amount of starting material of RNPs for cross-linking and for subsequent purification of the cross-linked heteroconjugates by LC to ≤50 pmol as compared to 1.5–3.0 nmol of starting material needed beforehand. An example of a cap-LC chromatogram derived from the purification of 50 pmol of native U1 snRNP particles after hydrolysis with trypsin and RNase T1 is given in . In contrast to our previous experiments (,), defined collection of isolated peptide–RNA cross-links is not longer possible. Fractions contain a mixture of residual non-crosslinked peptides and cross-linked heteroconjugates and unambiguous identification of the putative cross-links (for subsequent MS/MS analysis) is hampered.
One possibility to overcome this problem is the introduction of an enrichment strategy for putative peptide–RNA heteroconjugates, employing a technique such as IMAC (,). In this context, we make use of a feature shared by protein–RNA cross-links and phosphopeptides, namely the phosphate groups that both carry. As phosphopeptides, cross-links and free RNA oligonucleotides can interact with the IMAC material through their negatively charged phosphate groups [as also demonstrated for cross-linked protein–DNA complexes ()]. Since IMAC favours not only the enrichment of phosphate-containing species, but also a certain degree of unspecific binding of (acidic) peptides to the affinity matrix, the identity of enriched precursors must usually be verified by MS/MS experiments. Alternatively, additional treatment of the IMAC-enriched fractions with CIP can be used for the validation of the enriched species (), as it allows the exclusion of false positives, i.e. acidic peptides, by a characteristic mass shift of 80 Da (=HPO) in the corresponding MS spectra. CIP treatment together with highly accurate MALDI-ToF MS analysis in the reflectron mode and a computational database search () can reveal the cross-linked species in the spectrum without precursor selection and subsequent MS/MS analysis.
A–C shows the MALDI-MS spectra of capillary-LC fractions of UV-irradiated [15.5K-61K-U4atac snRNA] complexes treated with chymotrypsin and RNase T1 (A) or RNase T1/A (B) and of UV-irradiated native U1 snRNPs hydrolyzed with trypsin and RNase T1 (C) without IMAC treatment. D–F shows the MALDI-MS spectra of the same fractions treated with IMAC. In all three cases, MALDI-MS revealed enriched species with / values of 3071.92, 1801.68 and 2272.75, respectively. MS/MS experiments of these putatively cross-linked precursors cause difficulties, as the amount of precursor is not sufficient (and precursor selection even in the non-enriched samples is restricted by the technical specifications of the MALDI instrument). To prove without MS/MS experiments that enriched precursors carry a phosphate moiety ()—i.e. must contain a cross-linked RNA oligonucleotide with a 3′-phosphate (as derived from ribonuclease cleavage)—we treated the enriched precursors with CIP on the IMAC beads. G–I shows the MALDI-MS spectra of the IMAC-enriched capillary-LC fractions additionally treated with CIP. Enriched precursors with / = 3071.918, 1801.675 and 2272.749 (G–I) clearly show a shift of 80 amu, demonstrating that HPO was cleaved from the enriched species which thus must be a cross-linked peptide–RNA heteroconjugate. Comparison of the measured monoisotopic masses with calculated monoisotopic masses of all possible combinations of RNA oligonucleotides () revealed that the enriched species are no (modified) RNA oligonucleotides.
An ambiguity of our strategy is the fact that enriched phosphopeptides (derived from native protein–RNA complexes) would also show the loss of HPO and thus a mass shift of 80 amu in MALDI-MS. However, enrichment of residual phosphopeptides in our experiments is not very likely since the vast majority of the non-crosslinked peptides is separated during size-exclusion chromatography. The probability that residual non-crosslinked peptides is still—to a certain degree—phosphopeptides is very low, because of the low abundance of phosphopeptides in protein–RNA complexes . Nonetheless, to exclude any ambiguity between cross-links and phosphopeptides we envisage the possibility of incubating the enriched species with nucleases (e.g. nuclease P1) on the beads instead of CIP. Nuclease P1 hydrolyzes both 3′-5′-phosphodiester bonds in single-stranded nucleic acids and 3'-phosphomonoester bonds in mono- and oligonucleotides terminated by a 3′-phosphate without base specificity. The corresponding mass shift could be observed in the MALDI-MS as for the CIP-treated samples.
Moreover, by applying this strategy additional sequence information about the cross-linked RNA moiety could be obtained. Alternatively, cross-linked and enriched species could be hydrolyzed with HF (). Recent studies on protein–DNA cross-links demonstrated that HF treatment leaves only the nucleobase of the cross-linked nucleotides attached to the peptide ().
As no structural information about the cross-linked peptide moiety could be obtained by MS/MS experiments, we set out to calculate the sequence of the cross-linked peptide moiety. We compared the experimental masses of the enriched precursors (i.e. putative cross-links) with the monoisotopic masses of proteolytic peptides of all the proteins within the RNP plus the monoisotopic masses of RNA oligonucleotides derived from hydrolysis of the RNA molecules of the RNP (). Assuming that the cross-linked precursor mass is additively composed of the molecular masses of the cross-linked peptide and RNA moieties, the following constraints were made for the calculation: (i) mass deviation for searching ≤0.1 Da; (ii) specificity of the endoproteinase (e.g. trypsin: R/K, not after P; chymotrypsin: Y/F/W/M/L), allowing two missed cleavages; (iii) highest priority for specificity of the endonuclease(s) (RNase T1, RNases T1 and A), but also consideration of 3′ and 5′ non-specific cleavage/hydrolysis products () and (iv) for each experiment/search, the scope of the database was restricted to the sequences of proteins and RNA(s) specific for the investigated RNP particle. A peptide hit was considered to be genuinely positive when the computational database search from both the experiments with RNase T1 and RNases A and T1 gave the same peptide sequence.
The results of the computational database searches for the enriched cross-linked species derived from reconstituted [15.5K-61K-U4atac snRNA] complexes and native U1 snRNPs are summarized in . Strikingly, enriched precursor masses derived after digestion of the RNA moiety of the two cross-linked complexes with RNase T1 and RNases T1/A showed only one specific peptide sequence for each RNP particle (, bold peptide sequences). The first of these is SSTSVLPHTGY, encompassing positions 263–273 of protein 61K (hPrp31), with the 61K protein cross-linked to an AU dinucleotide within the U4atac snRNA (in the stretch CAUAG); and the second is RVLVDVER, positions 173–180 in the U1 70K protein cross-linked to an AU dinucleotide within the U1 snRNA (in ACU).
In general, unambiguous assignment of the cross-linking position on the RNA is more difficult. U4atac snRNA only contains a single specific RNase T1 fragment with the calculated composition of ACGU (AUCAG). The U1 snRNP cross-link with / 2272.279 (corresponding to a cross-linked RNA with the composition ACU) cannot be a specific RNase T1 fragment, as it does not contain a G (RNase T1 cleaves after G and RNase A after U and C). This fragment might result from the hydrolysis of a larger T1 fragment either during LC purification or from in-source gas-phase fragmentation during MS. However, the calculated nucleotide composition ACU is present in four RNase T1 fragments of the native U1 snRNA, namely nucleotide positions 21–28 (AUACCAUG), 29–33 (AUCACG), 111–117 (AAACUCG) and 122–130 (CAUAAUUUG). Without any further MS/MS experiments it is impossible to identify unambiguously the sequence of the cross-linked RNA.
Nonetheless, our enrichment/MS/computational database approach has proven—in this feasibility study—to be useful for the assignment of cross-linked proteins in RNPs and their cross-linked peptide region. Our calculations perfectly match the results obtained by MS/MS analysis of peptide–RNA heteroconjugates derived from the same UV-irradiated RNP particles in large-scale experiments ().
We extended our approach to a partial complex of the U2 snRNP, i.e. the ternary [p14/SF3b14a-SF3b155-U2 snRNA BSiR] complex, in order to identify the contact sites of the U2 snRNP-specific protein p14/SF3b14a and the so-called ‘branch-site interacting region’ (BSiR) on the U2 snRNA (,,).
During pre-mRNA splicing—that is excision of intronic sequences and ligation of exons of a pre-mRNA in order to yield mature mRNA—sub-spliceosomal U1, U2 and [U4/U6.U5] tri-snRNP particles assemble on the pre-mRNA in a stepwise manner to generate the catalytically active spliceosome (). A critical step in the assembly of the spliceosome is the recognition of conserved splicing motives of the pre-mRNA. Base-pairing of U2 snRNA (via its BSiR, nucleotide positions 33–38) with the branch-site sequence (BS) of the pre-mRNA forces an adenosine base (the so-called branch-point adenosine) to bulge out, defining a nucleophile for the first catalytic step of splicing. Both the p14/SF3b14a and the SF3b155 proteins are subunits of the U2 snRNP-associated splicing factor 3b (SF3b) that contacts pre-mRNA in the region of the branch-site, thus contributing to the recruitment of the U2 snRNP to the branch site (). Cross-linking studies demonstrated that p14/SF3b14a is in direct contact with the branch-point adenosine (,) and with the U2 snRNA at nucleotide G31 next to the BSiR (nucleotides G33–A38) (). In addition, p14/SF3b14a binds tightly to the SF3b155 protein between amino acids 255 and 424 () and its interaction facilitates binding to the BSiR on the U2 snRNA (). Recent structural studies without RNA further demonstrated that p14/SF3b14a contains an RNA recognition motif (RRM) that is partly involved in the interaction with SF3b155 (,). Interestingly, Schellenberg and co-workers showed by cross-linking experiments with a site-specifically P-labelled RNA mimicking the pre-mRNA branch-point region that a conserved tyrosine residue in the RNP2 motif of the RRM is in direct contact with the bulged-out adenosine or its adjacent nucleotides ().
A missing link in the overall picture in the absence of high-resolution structures is the identification of the contact sites between the p14/SF3b14a and its cognate U2 snRNA region, i.e. the branch-site interacting region. We therefore set out to identify the site(s) of such interaction(s) by applying our approach on reconstituted [p14/SF3b14a-SF3b155-U2 snRNA BSiR] complexes ().
shows spectra of a cap-LC fraction before (A) and after (B) IMAC, revealing enrichment of a putative cross-link with / 2832.061 (labelled ‘C’). Peak C disappeared after treatment of the sample with alkaline phosphatase, and a new signal at / 2752.036 arose (‘D’ in C). The characteristic mass difference of 80 amu between signals C and D was observed as before, consistent with an accessible phosphate residue.
We observed that MALDI-MS experiments performed in negative mode led to greater signal intensities, in particular when the cross-linked RNA moiety is larger and the cross-linking yield is low, which was the case in the latter experiment. A similar phenomenon has been described for the MALDI-MS analysis of phosphopeptides, where dramatically greater signal intensities are achieved in the negative mode (). It is therefore feasible during MALDI-MS of enriched cross-links to switch between positive and negative mode to increase the signal intensities of the cross-linked precursors. Importantly, this can be executed within one experiment without consumption of noticeably greater quantities of sample material, so that the use of both modes does not incur any need for additional preparative experiments or a larger-scale preparation.
A computer database search revealed that the masses associated with peaks C and D did not match any combination of nucleotides derived from the RNA alone (). However, it corresponded exactly to a p14/SF3b14a tryptic fragment encompassing positions 97–106 (AFQKMDTKKK) cross-linked to an RNA oligonucleotide with the composition ACGU. This composition matches two sequences within the oligonucleotide, namely GUAUC or UAUCG; of these, the former GUAUC is not a specific RNase T1 fragment. Owing to the low abundance of the precursor ions C and D (after IMAC enrichment, 400 a.i. with 1000 laser shots; after CIP validation, 100 a.i. with 1000 laser shots), MS/MS was not possible. Nevertheless, a cross-link of the p14/SF3b14a fragment AFQKMDTKKK to the RNA oligonucleotide GUAUC most probably explains signals C and D () for the following reasons: (i) the alternative RNase T1 fragment (UAUCG) carries a 3′-hydroxyl group and could thus not have been dephosphorylated by alkaline phosphatase. Note that the ions marked with ‘A’ (/ 1222.260) and ‘B’ (/ 1528.275) in B also show no loss of phosphate (although they are enriched by IMAC). Computational analysis of these mass peaks resulted in the oligonucleotides AUCG and UAUCG carrying a 3′-hydroxyl group, as G13 represents the extreme 3′ end of the oligonucleotide. Both these RNA sequences could be confirmed by MS/MS analysis of the pre-IMAC sample (A and data not shown). (ii) Notably, mass peak A is a hydrolysis product of B (the actual RNase T1 fragment). Hydrolysis is presumably a consequence of cap-LC separation at pH 2, as RNA becomes labile below pH 3. Thus, the putatively cross-linked RNA oligonucleotide GUAUC can similarly be explained as a hydrolysis product of a larger T1 fragment.
In a similar manner, we identified a second peptide–RNA cross-link (data not shown). The / value of this second cross-link (2680.094) was again shifted by 80 amu after CIP treatment (to / 2600.056), thus again representing a putative peptide–RNA heteroconjugate. These masses exactly matched a p14/SF3b14a fragment encompassing positions 101–111 (MDTKKKEEQLK) cross-linked to an RNA oligonucleotide with the composition GAU that can be found at the RNA positions 4–7 (UGUA), 6–9 (UAGU) or 8–11 (GUAU). Strikingly, the deduced peptide fragment partially overlaps with that of the first cross-link, AFQKMDTKKK. Taken together, these results argue that RNA can be cross-linked to p14/SF3b14a between positions 97 and 111 (AFQKMDTKKKEEQLK).
Importantly, the cross-linked peptide sequence encompasses parts of helix α3, helix α4 and the connecting loop between these helices (A and B). It harbours the basic residue Lys100 that is postulated to be involved in interactions with RNA due to its close neighbourhood to Tyr22 (). Furthermore, this particular residue is affected upon incubation with U2 BS-BSiR RNA duplex in NMR chemical shift experiments (). On the basis of our former work, in which we found that methionine residues within loop regions of (ribosomal) proteins are highly susceptible to UV cross-linking (), we postulate that Met101 in p14/SF3b14a is cross-linked to GUAUC (presumably to one of the uridine residues) within the BSiR (CGGUGUAGUAUCG). Our cross-linking data together with the data available from the 3D structures of p14/SF3b14a suggest a U2 BS/BSiR RNA binding interface on one site of the p14/SF3b14a encompassing β3′, β3″ and their connecting loop, β-sheet(s) of the RNP2 (and RNP1) and parts of the C-terminal helices α3 and α4 with their connecting loop (B).
We have developed an overall approach for the enrichment, validation and subsequent MALDI-ToF-MS-based identification of peptide–oligoribonucleotide heteroconjugates as obtained by UV cross-linking of RNP particles, both native and reconstituted —in particular when these are not available in large quantity.
Our approach can be applied to any protein–RNA complex for the identification of proteins in direct contact with RNA and for visualization of the site of interaction. Since it introduces capillary liquid chromatography for the final purification step of the cross-linked peptide–RNA heteroconjugates, it requires a relatively low amount of starting material for the protein–RNA cross-linking experiments (10–50 pmol of complex).
The strategy was first successfully applied to the purification of peptide–RNA oligonucleotide cross-links from two test systems that have been extensively studied before ([15.5K-61K-U4atac snRNA] complexes and U1 snRNP particles (,)). It further proved successful in the analysis of a contact site between the p14/SF3b14a protein and the U2 snRNA BSiR in partial U2 snRNP particles, i.e. [p14/SF3b14a-SF3b155-U2 snRNA BSiR] complexes ().
We assume that upon introduction of increasingly smaller chromatography systems, e.g. nano-LC, combined with enrichment strategies and MS(/MS)-based validation of peptide–RNA cross-links as presented here or elsewhere (), a comprehensive analysis of protein–RNA contact sites in larger RNP particles [e.g. spliceosomal B () and C () complexes] is possible.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
i
s
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
DNA double-strand break (DSB) is probably the most dangerous type of DNA damage among the various types of DNA damage that can affect a cell. They are formed by exogenous agents, such as ionizing radiation (IR) and certain chemotherapeutic drugs and by endogenously generated reactive oxygen species and chromosomal stress. The inability to respond properly to DNA DSBs and repair the damage may lead to genomic instability, which in turn may either lead to cell death or increase the risk of pathological consequences such as the development of cancer ().
Observations in yeast and mammalian cells suggest that sister chromatid cohesion is important for DNA repair as well as proper segregation of chromosomes. It has been proposed that cohesin facilitates DNA repair by holding sister chromatids locally at DSB sites to allow strand invasion during homologous recombination (HR) (). The cohesin complex of budding yeast, which consists of Smc1, Smc3, Scc1 and Scc3, forms a ring-like structure (). This holds the sister chromatids together by trapping the sister DNA molecules within its ring (,) and is essential for maintaining cohesion between sister chromatids until metaphase to assure equal segregation of sister chromatids (). Loading of the cohesin complex onto chromatin requires the Scc2–Scc4 complex, whereas Eco1/Ctf7 is required to establish sister chromatid cohesion during S phase (,). The interaction between Eco1/Ctf7 and PCNA, which acts as a clamp for DNA polymerases, is essential for sister chromatid cohesion (,). However, Eco1/Ctf7 is neither required for the loading of cohesin onto chromatin nor for the maintenance of cohesion in G/M phase (,).
Mutation of the gene causes a decrease in the fidelity of chromosome transmission or chromosome loss (). Mutations in and as well as , whose products form an alternative replication factor C (RFC) complex in conjunction with the four small subunits Rfc2-5 of RFC (Ctf18-RFC), cause a moderate non-lethal defect in sister chromatid cohesion (). Ctf18-RFC has been shown to interact with PCNA and to load it onto DNA (,). Moreover, Ctf18-RFC is capable of unloading PCNA (). Since Ctf18 physically associates with Eco1/Ctf7 (,), it seems likely that the moderate defect in sister chromatid cohesion of mutant cells maybe related to the function of Eco1/Ctf7. Of note, Ctf18 and Eco1/Ctf7 are found at replication forks, and Ctf18 is required for the efficient recruitment of PCNA onto replication forks in HU-arrested cells ().
Studies in yeast have revealed that normal loading of the cohesin complex onto chromatin during the progression of DNA replication is insufficient to hold DSB ends in close proximity. This suggests that the cohesin complex must be loaded within the vicinity of the DSBs following replication to facilitate the repair of the DSBs through sister chromatid recombination (SCR) (,,). Until now, the function of Ctf18 in homologous recombination has not been considered because of the synthetic sick or lethal interaction between the mutation of the gene and () which plays a major role in homologous recombination repair. In this study, we present evidence that Ctf18 is involved either directly or indirectly in recombination-mediated DSB repair.
The yeast strains used in this study are listed in Supplementary Table S1. Null mutants and Myc- or HA-tagged alleles were made using standard PCR-based gene disruption and insertion methods, as previously described (). Deletion mutants were replaced by and , which were amplified from pFA6aKANMX6, pAG32 and SHB1805, respectively, by using gene-specific primers consisting of 40–45 nt. The resulting PCR fragments were then transformed into yeast cells and colonies that appeared on G418-, hygromycin B-containing YPAD plates or SC-Trp plates were isolated. Gene disruption was confirmed by PCR of genomic DNA. The sequences of the primers used to generate either the DNA constructs used for gene disruption or for checking disruption of the genes as well as details on the yeast strains will be provided upon request.
For the analysis of sensitivity to continuous exposure to DNA-damaging agents, 10-fold serial dilutions of logarithmically growing cells in distilled water were spotted onto YPAD plates, or YPAD plates containing the indicated concentrations of methyl methanesulfonate (MMS), hydroxyurea (HU), phleomycin or camptothecin (Sigma-Aldrich). The plates were incubated for 3 days at 30°C and then photographed. For NHEJ analysis, strains containing galactose-inducible HO endonuclease were grown overnight to 1 × 10 cells/ml in YP medium containing 2% raffinose. After washing once with water, aliquots of 10-fold serial dilutions of cells were plated on YP medium, which contained 2% raffinose and was supplemented with either 2% galactose or 2% glucose. The plates were incubated for 3 days at 30°C and then photographed. For the analysis of sensitivity to phleomycin, logarithmically growing cells were diluted to 10 cells/ml and cultured in the presence of 15 μg/ml nocodazole for 3 h to induce G/M phase arrest. G/M-arrested cells were exposed to 100 μg/ml phleomycin for 0.5, 1, 1.5 and 2 h. The cells were washed to remove phleomycin and nocodazole, diluted and inoculated onto YPAD plates. After 3 days incubation at 30°C, colonies were counted.
To detect the modification of Rad53-13Myc in response to DNA damage, G/M-arrested cells were exposed to 100 μg/ml phleomycin or 0.1% MMS for 2 h at 30°C, and then the cells were harvested for immunoblotting analysis. The cells were incubated at 4°C for 15 min in 230 μl 1.5 M NaCl, followed by incubation at 4°C for 10 min after addition of 30 μl 55% Trichloroacetic acid, and spun at 14 000 r.p.m. for 1 min at 4°C. The cell pellet was resuspended in 20 μl of HU buffer (200 mM Tris-HCl (pH 6.8), 1 mM EDTA, 5% SDS, 8 M Urea, 1% BPB (Bromo-phenol blue) and 20 μl of 5× SDS–polyacrylamide sample buffer. Samples were boiled for 5 min, and spun at 10 000 r.p.m. before loading to SDS-PAGE gels. The proteins were detected by immunoblotting with an anti-Myc antibody (9E10) or an anti-histone H3 antibody (Abcam).
The strains constructed for detecting unequal SCR were previously described (). Diploid strains with the MR101 background were constructed such that recombination between the heteroalleles and could be detected by the restoration of histidine prototrophy (). The number of His+ colonies was scored for each of the 12 plates and the median number of His+ colonies for all 12 plates was determined. The rate of spontaneous recombination was then calculated by the median method (,). For detection of damage-induced recombination, logarithmically growing cells were inoculated onto SC-His plates and YPAD plates with or without MMS or phleomycin to evaluate the incidence of damage-induced recombination and colony forming cells, respectively. Alternatively, the logarithmically growing cells were diluted and arrested in G/M phase in the presence of 15 μg/ml nocodazole for 3 h at 30°C and were exposed to 100 μg/ml phleomycin for the indicated time at 30°C. The cells were subsequently washed to remove the phleomycin as well as nocodazole and plated on YPAD plates and SC-His plates. The recombination frequency after treatment with MMS or phleomycin was determined by dividing the total number of recombinants in the culture by the total corresponding number of surviving cells following treatment with MMS or phleomycin.
Logarithmically growing cells were diluted to 8 × 10 cells/ml, exposed to 0.1% MMS for 1 h, and then cultured in MMS-free medium for the indicated periods of time. In the case of phleomycin treatment on G2/M-arrested cells, logarithmically growing cells were diluted to 8 × 10 cells/ml and cultured in the presence of 15 μg/ml nocodazole for 3 h at 30°C to induce G/M phase arrest. G/M-arrested cells were exposed to 100 μg/ml phleomycin for 2 h at 30°C, washed to remove the phleomycin, and then cultured at 30°C in YPAD medium containing 15 μg/ml nocodazole for the indicated periods of time.
Agarose plugs containing chromosomal DNA were prepared as previously described, with minor modifications (). All plugs were subsequently treated with Zymolyase 100T (0.15 mg/ml) and proteinase K (1 mg/ml) at 30°C for 24 h. After electrophoresis, gels were stained with 0.5 μg/ml ethidium bromide for 30 min, destained in deionized water for 20 min, and then photographed.
Whole-cell extracts (WCEs) and chromatin pellets (ChP) were prepared as previously described () with modification. A total of 4 × 10 cells were harvested and sodium azide was added to 0.1%. Cells were incubated at room temperature for 4.5 min in 0.5 ml of prespheroplasting buffer [100 mM PIPES (pH 9.4), 10 mM DTT], followed by incubation in 0.5 ml of spheroplasting buffer [50 mM KHPO/KHPO (pH 7.5), 0.6 M Sorbitol, 10 mM DTT, 0.125 mg/ml Zymolyase 100T] at 30°C for 4.5 min with occasional mixing. Sphereoplasts were washed with 1 ml of ice-chilled wash buffer [100 mM KCl, 50 mM HEPES-KOH (pH 7.5), 2.5 mM MgCl and 0.4 M Sorbitol, 1 mM PMSF], pelleted at 4000 r.p.m. for 1 min in a microcentrifuge at 4°C, and resuspended in 200 μl of extraction buffer [EBX; 100 mM KCl, 50 mM HEPES-KOH (pH 7.5), 2.5 mM MgCl, 50 mM NaF, 5 mM NaPO, 0.1 mM NaVO], and protease inhibitors (1 mM PMSF, 20 μg/ml of leupeptin, 2 μg/ml of pepstatin, 2 mM benzamidine HCl and 0.2 mg/ml of bacitracin or 1× Complete (Roche), 0.25% Triton X-100). Spheroplasts were lyzed and incubating on ice for 5 min with gentle mixing. Lysate was underlayered with 100 μl of 30% sucrose, and spun at 14 000 r.p.m. for 3 min at 4°C. Pellet was washed with 100 μl of EBX, and spun again at 12 000 r.p.m. for 1 min at 4°C. All pellet fractions were resuspended in 20 μl EBX. Equal volumes of mixture of HU buffer [200 mM Tris-HCl (pH 6.8), 1 mM EDTA, 5% SDS, 8 M Urea, 1% Bromo-phenol blue] and 5× SDS–polyacrylamide sample buffer were added to each fraction. Samples were boiled for 5 min, and spun at 10 000 r.p.m. for 1 min before loading to SDS-PAGE gels. Histone H3 was used as a loading control for protein levels in WCEs and chromatin pellet (ChP) fractions. The proteins were detected by immunoblotting with an anti-Myc antibody (9E10; Santa Cruz) or anti-histone H3 antibody (Abcam).
ChIP was carried out as previously described, with minor modifications (). Briefly, asynchronous cultures were grown overnight at 30°C in YP medium containing 2% raffinose. Nocodazole (15 μg/ml) was added to the cultures when they reached 6 × 10 cells/ml, and they were incubated at 30°C for 5 h to induce G/M phase arrest. Expression of HO endonuclease was then induced by adding 2% galactose. The cultures were washed for 1 h after the addition of galactose and then treated with 2% glucose to repress the expression of HO endonuclease. Cells were harvested and incubated in 1% formaldehyde for 15 min to cross-link proteins to DNA, then the reaction was quenched by incubating the cells in 125 mM glycine for 5 min. Cells were lyzed with glass beads and extracts were sonicated to shear DNA to an average size of 1 kb. Extracts were then divided into two aliquots: Input DNA and immunoprecipitated (IP) DNA (1:20, respectively). Immunoprecipitation was carried out using a monoclonal anti-Myc antibody (9E10) (Santa Cruz Biotechnology, Inc.) or a monoclonal anti-HA antibody (12CA5) (Roche), and immune complexes were captured using Dynabeads Protein G (Dynal Biotech) for 4 h at 4°C. After a series of washes, proteins were released from the beads by incubation for 6 h at 65°C, and were treated with proteinase K. IP DNA was purified for PCR analysis using phenol extraction followed by ethanol precipitation.
Primers used for ChIP assay are listed in Supplementary Table S2. Amplified PCR products from IP DNA were separated by agarose gel electrophoresis, and quantitated using Scion Image software. To correct for different levels of IP efficiency at different loci, the signal from the target locus was first normalized to the signal from an independent locus () on chromosome VI by dividing each target signal by the corresponding IP signal. For analysis of different time points, or IP signals were normalized to the IP signal at 0 h, which was designated as 1.
DSB induction and strand invasion of the input DNA were detected by PCR. Primers are listed in Supplementary Table S2. The primer extension and ligation assays were carried out according to a previously described PCR-based method (). Amplified PCR products were separated by agarose gel electrophoresis and quantitated using Scion Image software. The signal from the target locus was normalized to the amplified signal from an independent locus () on chromosome VI. Primer extension and ligation were arbitrarily set at 100% for the highest wild-type level.
The cohesin complex is required for efficient repair of DSBs (,,). Ctf18 is involved in establishing sister chromatid cohesion (). However, the proteins that are involved in establishing cohesion for DNA repair have not been fully determined. In this study, we investigated the role of Ctf18 in DNA repair. We first examined the sensitivity of deletion mutants to DNA-damaging agents, methyl methanesulfonate (MMS), phleomycin, camptothecin and hydroxyurea (HU), which reportedly induce DSBs both directly and indirectly (). The mutant cells, like cells, were highly sensitive to these DNA-damaging agents compared with wild-type cells (A), which suggests that cells have defects in their DNA damage checkpoints or DNA repair mechanism. To test the former possibility, we examined the phosphorylation of Rad53, which is a hallmark of activation of the damage checkpoint (). Rad53 was phosphorylated in both and wild-type cells following exposure to MMS and phleomycin (B), suggesting that Ctf18 is not essential for the activation of DNA damage checkpoint. This is consistent with a previous report that cells showed only a slight defect in checkpoint function unless Rad24 was also absent (). The sensitivity of the 18 cells to these DNA-damaging agents suggests that Ctf18 is involved in DSB repair.
DSBs can be repaired by either non-homologous end-joining (NHEJ), which requires Yku70, or homologous recombination (HR), which requires Rad52 (). First, we investigated whether Ctf18 was required for NHEJ-mediated repair by inducing a single DSB in the locus using cells that express HO endonuclease. The HR-mediated repair pathway was precluded in these cells by deleting the homologous donor loci, and (C). Sustained expression of the HO endonuclease leads to a continuous cycle of DNA cleavage and ligation. Cells that are capable of error-prone repair by NHEJ grow in the presence of the HO endonuclease, while cells deficient in NHEJ do not survive (). Unlike cells, we found that and mutants grew as well as wild-type cells under sustained expression of HO endonuclease (D).
Next, we investigated if damage-induced homologous recombination was impaired in cells. We examined damage-induced unequal sister chromatid recombination (uSCR) in haploid cells (E). As shown in F, the mean spontaneous uSCR rate was slightly lower in cells than in wild-type cells; however there was no statistical difference in the rates between the wild-type and cells. In contrast, the frequency of MMS-induced uSCR was very low in cells compared with wild-type cells (G). This suggests that Ctf18 is involved in the recombination between sister chromatids following exposure to MMS.
HR-mediated DSB repair is a multiple-step process that includes the recruitment of proteins to sites of DSBs, resection of DNA, strand invasion of ssDNA on the homologous template, DNA synthesis and ligation. One of the best characterized systems for studying homologous recombination is HO endonuclease-induced switching, using as the donor template (). In this system, repair of HO endonuclease-induced DSB at the a locus is accomplished through recombination with the homologous locus. We investigated the role of Ctf18 in DSB repair by HR using the yeast strain JKM161, which contains a and as the donor template. DSB formation at the HO recognition site and invasion of the a recipient strand into the donor strand () were monitored by PCR using the primers shown in A. The amount of PCR product corresponding to the a locus was considerably reduced after 1 h following induction of the HO endonuclease in wild-type and cells. This indicated that cleavage at the a locus occurred with high efficiency. In addition, strand invasion was detected 2 and 4 h following induction of the HO endonuclease in wild-type cells, but not in mutants (B). To examine whether Ctf18 was recruited to both DSB sites in the and the donor loci during the process of recombination, we performed ChIP analysis using primers for either - or -specific sequences (C). We observed an association of Ctf18 with the DSB site in the locus 1 h after induction of the HO endonuclease in both wild-type and cells (D). In addition, Ctf18 was also associated with the donor strand 4 h after HO endonuclease induction (E). In contrast, an association between Ctf18 and the donor strand was not observed in cells (E). This data indicates that Ctf18 associates with recombination intermediates.
It was reported that recruitment of cohesin complex is dependent on Mre11. Therefore, we investigated whether or not the function of the MRX complex, a heterotrimeric protein assembly of Mre11, Rad50 and Xrs2, is required for the recruitment Ctf18 on the DSB site at locus. In response to DSBs, the MRX complex arrives first at the sites of DSBs and performs nucleolytic resection to generate single-stranded DNA (ssDNA). The ssDNA is recognized by the ssDNA-binding protein Rpa, and proteins involved in DNA damage checkpoint are recruited at DSB sites in an Rpa-dependent manner. Subsequently, proteins that are involved in HR are recruited. The assembly of Rad51 on ssDNA is initiated by Rad52, which in turn causes strand invasion. We found that the induction of DSBs occurred in a similar manner in both wild-type cells and cells arrested in the G/M phase (A). However, binding of Ctf18 around the DSB site was not observed in the cells compared with wild-type cells (B and C), suggesting that the generation of ssDNA by Mre11 is essential for the recruitment of Ctf18 to the DSB site. These findings indicate that Ctf18 is recruited to the DSB site in a manner dependent on the function of Mre11.
Since Ctf18 is required for the establishment of sister chromatid cohesion, and cohesin is recruited to sites of DSBs, we examined whether a defect in Ctf18 alters the recruitment of cohesin to the DSB sites. Cohesin is composed of Smc1, Smc3, Scc1 and Scc3. We investigate whether the recruitment of Scc1 onto DSB sites required Ctf18. To characterize the association of these proteins with DSB sites, we used the JKM179 strain () in which the and loci are deleted to prevent HO endonuclease-induced DSB repair by homologous recombination. In addition, to avoid any complication due to S-phase events, we used the cells arrested in G/M phase. Following induction of the HO endonuclease, DSBs at the α locus were induced rapidly and efficiently (D). Using ChIP analysis, we investigated the binding of HA-tagged Scc1 to sites around the DSB in wild-type and cells. As reported previously, the binding of Scc1 to the α locus was increased by 2- to 3-fold following induction of the DSBs, and Scc1 was localized to regions 2–10 kb away from the DNA break on both sides of the break in wild-type cells [E and F; (,)]. No difference was observed between cells and wild-type cells in the localization of Scc1 around the DSB site. This suggests that Ctf18 is not required for cohesin enrichment around the DSB site.
Recruitment of Ctf18 onto sites of DSBs and the association of Ctf18 with recombination intermediates prompted us to examine whether or not Ctf18 was required for the HO endonuclease-induced intrachromosomal recombination. We monitored the formation of DSBs at the HO endonuclease recognition site, primer extension following strand invasion on the donor strand, and ligation of synthesized DNA in the presence of PCR using the primers shown in A. We observed that the induction of DSBs, the extension of the invading DNA strand, and ligation occurred with similar kinetics in cells compared with wild type cells (B, C and D). This suggests that Ctf18 is not required for HO endonuclease-induced intrachromosomal recombination.
Although Ctf18 was associated with recombination intermediates (), no apparent defect was observed in HO endonuclease-induced DSB repair by either NHEJ or HR in cells (D and ). Therefore, we investigated whether or not Ctf18 was involved in DSB repair at the chromosomal level. Logarithmically growing cells were treated with MMS and then cultured in MMS-free medium. Chromosomal DNA was isolated from these cells and subjected to pulsed-field gel electrophoresis (PFGE). Immediately following the administration of MMS, the distinct chromosomal DNA bands were replaced by a low-molecular-weight DNA smear in both wild-type and cells (A). After culturing the cells in MMS-free medium, the chromosome-sized DNA bands were restored in wild-type cells, whereas the restoration of the DNA bands was impaired in cells. The amount of fragmented DNA might not necessarily reflect the level of DSBs because strand breaks can occur in MMS-treated DNA during the heat treatment of PFGE plugs (). However, the defect in restoring chromosome-sized DNA bands was only observed in cells and suggests a defect in post-replicative repair in these cells.
To further test the involvement of Ctf18 in DSB repair, we investigated the repair of DSBs induced by phleomycin in G/M phase-arrested cells. Cells arrested in G/M were treated with phleomycin and then cultured in phleomycin-free medium containing nocodazole to maintain G/M arrest. Chromosomal DNA was isolated and subjected to PFGE (B), and the cell cycle status of the cells was monitored by fluorescence activated cell sorting (FACS) (C). Distinct chromosomal DNA bands were absent just after exposure to phleomycin in both wild-type and cells and a low-molecular-weight DNA smear appeared, reflecting the occurrence of DSBs. The chromosome-sized DNA bands were restored in both wild-type and cells after culturing cells in phleomycin-free medium. However, the restoration of these bands was less efficient in cells compared to wild-type cells, which was reflected in the decrease in the viability of the cells compared with wild-type cells after a brief exposure of G/M phase-arrested cells to phleomycin (D).
We also investigated the association of Ctf18 with chromatin following exposure to either MMS or phleomycin. We observed an increase in the amount of Ctf18 in the chromatin-containing fraction after exposure of logarithmically growing cells to MMS (E). Moreover, the association of Ctf18 with chromatin was observed even in G/M phase-arrested cells (Supplementary Figure S1) after exposure to MMS and phleomycin (F and G). Taken together, these results suggest that Ctf18 plays a role in post-replicative DSB repair.
Ctf18 is involved in the establishment of sister chromatid cohesion, which is important for sister chromatid-based homologous recombination repair. Therefore, the defect of uSCR in cells (G) may be due to a defect in sister chromatid cohesion, even if Ctf18 is not required for cohesin recruitment around the DSB site (). In our previous study, we found that mutant cells show defects in damage-induced recombination between not only sister chromatids (data not shown) but also between homologous chromosomes (). Thus, we investigated whether Ctf18 is required for interchromosomal recombination between the heteroalleles in diploid cells (A). The spontaneous interchromosomal recombination rate of cells was comparable to that of wild-type cells (B). In the presence of either MMS or phleomycin, the interchromosomal recombination frequency in wild-type cells dramatically increased (C and D), whereas the interchromosomal recombination frequency in cells remained low. The interchromosomal recombination frequency of cells in G/M phase was also very low compared to that of wild-type cells after temporal exposure to phleomycin (E).
The gene was originally identified as the gene whose mutation causes a decrease in chromosome transmission fidelity or chromosome loss (). The function of Ctf18 in homologous recombination repair has not been considered because of the synthetic sick or lethal interaction between the mutation and the mutation, which plays a major role in homologous recombination repair (). In this study, we found that repair of DSBs induced by phleomycin (), recombination of sister chromatids induced by DNA damage () and interchromosomal recombination between heteroalleles () were defective in cells. This indicates that Ctf18 is involved in recombination repair. Although cells were nearly normal for HO-endonuclease-based NHEJ, this result does not necessarily rule out the possibility of Ctf18 involvement in the NHEJ pathway under other assay conditions, such as those used by Schar . () who reported that mutants were much less efficient in plasmid end-joining mediated by the NHEJ pathway.
Although Ctf18 was recruited to the DSB site at the locus () even in G/M phase-arrested cells, we could not detect an obvious defect in the processes of HO endonuclease-induced recombination, such as primer extension and ligation (). The discrepancy between the results obtained after inducing a single DSB with HO-endonuclease and those obtained after inducing DNA damage with phleomycin or MMS may be explained by the different mechanisms of recombination involved in repairing these two types of DNA damage; HO endonuclease induces intrachromosome recombination, whereas MMS and phleomycin induces interchromosomal or sister chromatid recombination. Alternatively, this discrepancy may be due to differences in the amount of DNA damage induced by these agents. For example, the function of Ctf18 maybe more important following the induction of multiple sites of DNA damage when the demands on the cell to execute HR-mediated DNA double-strand break repair are greater. Finally, the discrepancy may be due to the differences in the DNA ends created by HO endonuclease, MMS and phleomycin. Ctf18 could be required for the repair of DNA lesions induced by MMS and phleomycin but not for the repair of DSB induced by HO endonuclease.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
The ribosomal peptidyl transferase center (PTC) is the catalytic heart of the ribosome and plays a fundamental role in protein synthesis. It is a part of the large ribosomal subunit (50S in eubacterial ribosomes), a complex dynamic ribo-nucleoprotein ensemble with a molecular weight of ∼1.8 MD. The crystallographic structures compellingly confirmed that peptidyl transferase is an RNA enzyme [reviewed in ()]. This places the ribosome as the key entry on the list of naturally occurring ribozymes that outlived the transition from the pre-biotic RNA World to contemporary biology. The PTC is characterized by the most pronounced accumulation of universally conserved rRNA nucleosides in the entire ribosome. The prime functions of the PTC are (i) transpeptidation to covalently link amino acids via peptide bonds into polypeptides during the elongation phase of protein synthesis and (ii) peptidyl-tRNA hydrolysis, which is required for termination of translation and release of the fully assembled polypeptide from the ribosome. The termination reaction likely involves the transfer of the peptidyl moiety of P-site located peptidyl-tRNA to a water molecule (). In the course of the reaction, the nucleophilic attack of an activated water molecule in the PTC acceptor site onto the carbonyl carbon of the peptidyl-tRNA ester leads to formation of a tetrahedral intermediate. A proton is subsequently transferred to the 2′(3′)-hydroxyl of the 3′-terminal adenosine of peptidyl-tRNA breaking the ester bond between the peptide and tRNA. From a chemical point of view, peptidyl-tRNA hydrolysis is a more challenging reaction than transpeptidation because hydrolysis of the ester bond is driven by a significantly less nucleophilic water oxygen compared to the strongly nucleophilic α-amino group of aminoacyl-tRNA during peptide bond formation. It is therefore expected that the PTC-promoted reaction of peptide release should involve specific coordination and activation of the water molecule. The catalytic rate constant of peptide release has been estimated in an translation assay to be 0.5–1.5/s (), and is therefore clearly slower than transpeptidation (15–300/s) (,). The change of the mode of action of the PTC from amino acid polymerization to peptidyl-tRNA hydrolysis is triggered by class I release factors (RF1 or RF2 in bacteria) which bind in response to an mRNA stop codon displayed in the decoding A-site of the small ribosomal subunit.
The peptidyl transferase cavity is densely packed and decorated with nucleotides of the central loop of domain V of 23S rRNA (). The residues exposed on the surface of the catalytic center (A2602, U2585, U2506, A2451 and C2063; nomenclature is used here and throughout the manuscript) are universally conserved and are referred to as ‘inner shell’ PTC nucleotides. In the non-translating large ribosomal subunit, the peptidyl transferase cavity is hollow except for the 23S rRNA bases of nucleosides A2602 and U2585, which bulge into its core. The orientation of these two universally conserved residues depends on the functional state of the ribosome and the nature of the bound substrate () suggesting functional relevance of these nucleosides. Biochemical, cryo-electron microscopic and crystallographic data show that the tip of domain III (which harbors the universally conserved GGQ peptide mini-motif) of the A-site-bound RF reaches toward the bottom of the funnel-shaped active site crater of the PTC and is in immediate neighbourhood of A2602 and U2585 (). What happens in the PTC in response to the RF binding, what the role of the GGQ motif is and which functional groups are involved in the coordination and activation of the water molecule remain unknown. Models were proposed which suggest that the GGQ motif directly participates in peptidyl-tRNA hydrolysis by coordinating the water molecule (,,). However, mutational studies do not seem to support a direct participation of the GGQ sequence of the RF in catalyzing peptide release (,,). The fact that peptide release can be efficiently triggered even in the absence of the RF by A-site-bound deacylated tRNA (,,), strongly suggests that peptidyl-tRNA hydrolysis is a PTC-catalyzed reaction.
Mutational studies employing reconstituted 50S subunits showed that A2602 of 23S rRNA may be one of the critical components of the reaction for peptide release (). Whereas mutations of the other inner shell PTC residues C2063, A2451 and U2585 had only moderate effects on either of the reactions, substitution of A2602 with C or its deletion dramatically reduced the ribosome's ability to promote peptidyl-tRNA hydrolysis but had little effect on transpeptidation (). Furthermore, in experiments with affinity-tagged derived ribosomes, again mutations at A2602 had the most significant impact although also mutations at U2585 reduced, albeit to a lesser degree, the rate of peptide release (40-fold compared with 350-fold for the A2602C mutation) (). This indicates a possible additional contribution of U2585 for the peptide release mechanism. Based on the results of mutational studies, a model was proposed in which the class I RF triggers peptide release by reorienting A2602 in the PTC so that it can coordinate and possibly activate a water molecule for the attack onto the carbonyl carbon atom of the ester bond of the peptidyl-tRNA (). The structural flexibility and the central location of A2602 in the PTC are compatible with this proposed role (). The repositioning of A2602 for peptide release can potentially be coordinated with the movement of U2585, the second most flexible nucleotide in the PTC (,).
An important aspect that has not yet been directly addressed experimentally is the nature of the chemical group in the PTC that activates the water molecule or possibly stabilizes the oxyanion at the transition state. To address this issue we have employed a recently developed experimental system, the gapped-cp-reconstitution of 50S subunits, that allows non-natural RNA nucleoside analogs to be site-specifically placed into the 23S rRNA of the PTC (,). This experimental strategy expands the potential of standard mutagenesis approaches since it is possible to investigate PTC-catalyzed reactions on the functional group, and even at the single atom level of active site residues. The centerpiece of the gapped-cp-reconstitution is the use of circularly permuted 23S rRNA (cp-23S rRNA) that carries a short internal sequence deletion and a chemically synthesized RNA fragment (containing the desired nucleoside modification) that fills the 23S rRNA gap, for assembly of functional 50S particles. This system has already been successfully applied to demonstrate the importance of the A2451 2′-hydroxyl group as functionally critical component for efficient peptide bond formation (,). The A2451 ribose 2′-hydroxyl interacts with another important 2′-hydroxyl located at A76 of P-site-bound peptidyl-tRNA () via hydrogen-bonding. This interaction is considered to be pivotal for a pre-organized electrostatic hydrogen bond network that promotes transpeptidation (,), likely primarily via positional (,) or entropic forces (,). Here, we show that the A2451 2′-hydroxyl does not play an equally fundamental role in peptide release, whereas an intact ribose moiety at the universally conserved A2602 is absolutely required to trigger RF1-mediated peptidyl-tRNA hydrolysis.
Cp-23S rRNA constructs for investigating A2451, U2506, U2584, U2585 and A2602 in the gapped-cp-reconstitution were generated as described previously (,). To investigate the contribution of C2063, a cp-23S rRNA was constructed in analogy to the published procedure (). To generate the DNA template for cp-23S rRNA transcription, the forward and reverse PCR primer pair GAGCTTTACTGCAGCCTG and AGGCCGCATCTTCACGG was used on the plasmid pCPTaq23S carrying tandemly repeated 23S rRNA genes from . The T7 promoter sequences in the forward primer is underlined and the positions defining to the new 5′ and 3′ ends of the cp-23S rRNA are bold and numbered according to nomenclature. Cp-23S rRNA constructs allowing to study A2451 (cp2468-2440; whereas the first number always indicates the 5′-end and the second number the 3′-end of the cp-23S rRNA, respectively), U2506 (cp2523-2483) or C2063 (cp2069-2043) were assembled, and subsequently reassociated with native 30S subunits as described previously (). For the gapped-cp-reconstitution of 10 pmol cp2069-2043 RNA, 20 pmol of the synthetic RNA oligonucleotide (CCCGUGGCAGGACGAAAAGACCCCGU) was used to fill the sequence gap during reconstitution which was otherwise performed as described (). Prior to assembly of cp2623-2576, which allows to study U2584, U2585 and A2602, the chemically synthesized RNA fragment was ligated to the 3′-end of the cp-23S rRNA transcript. Therefore, 60 pmol of cp2623-2576 were mixed with 90 pmol of the 5′-phosphorylated synthetic RNA oligonucleotide AGCUGGGUUCAGAACGUCGUGAGACAGUUCGGUCUCUAUCCGCCAC and 60 pmol of the DNA splinter oligo TCACGACGTTCTGAACCCAGCTCGCGTGCCGCTTTAATGGGCGTTCTGCCC in 16.6 µl water. The mixture was incubated for 3 min at 90°C and subsequently placed on ice for 15 min. 2.4 µl of 10× ligation buffer (400 mM Tris/HCl pH 7.8, 100 mM MgCl, 100 mM DTT, 5 mM ATP, 2.4 µl of PEG 4000 (Fermentas), 40 u of RNase inhibitor (Fermentas) and 60 u of T4 DNA Ligase (Fermentas) were added and the reaction incubated at 30°C for 4 h. Subsequently, the DNA splinter was digested by the addition of 1 u DNase I (Fermentas) at 37°C for 30 min. The RNA was phenol/chloroform extracted, precipitated, dissolved in water and subsequently used in the gapped-cp-reconstitution. None-modified RNA oligonucleotides and those carrying a single deoxyribose residue or the C3-linker modification were purchased from , while synthetic RNAs that contained the abasic site analog were synthesized by solid-phase synthesis as described ().
The splinter ligation efficiencies between the cp2623-2576 transcript and all used RNA oligonucleotides was always checked via a modified primer extension reaction (). The 5′ P radio-labelled DNA primer CGACGTTCTGAACCCAG was annealed to the synthetic RNA oligo two nucleotides downstream of the ligation site and extended by 2 u AMV reverse transcriptase (Promega) as described () with the difference that cDNA synthesis was performed in the presence of 66.7 µM ddGTP, 833 µM dATP, 833 µM dCTP and 833 µM dTTP. Under these conditions the reverse transcriptase stops cDNA synthesis at the first encountered C nucleotide which results in a product of ‘primer plus 4’ nucleotides in the case of a ligated cp-rRNA species, whereas unligated rRNAs produce a ‘primer plus 2’ product. The cDNA products were resolved on a 15% denaturing polyacrylamide gel and the ligation efficiency quantified by comparing the intensities of the two RT products using the molecular dynamics Storm phosphorimager.
Formyl-[H]Met-tRNA (30 000 cpm/pmol) was prepared and purified by a reversed phase C4 HPLC column as described previously ().
70S ribosomes (containing ‘gapped-cp-reconstituted’ 50S subunits assembled from 10 pmol cp-23S rRNA and 4 pmol native 30S subunits from were programmed with 250 pmol synthetic mRNA UUCAUGUAA (Dharmacon Research, Inc.) and incubated with 0.3 pmol formyl-[H]Met-tRNA (30 000 cpm/pmol) for 15 min at 30°C in 15.9 μl of the reconstitution buffer containing 20 mM Tris/HCl pH 7.4, 20 mM MgCl, 400 mM NHCl, 5 mM β-mercaptoethanol and 0.2 mM EDTA (). The peptidyl-tRNA hydrolysis reaction was initiated by the addition of 15 pmol RF1 () and performed at 25°C in a final volume of 18.9 μl. The reaction was stopped and quantified by liquid scintillation counting as described (). Under these single turnover conditions (), reconstituted particles containing the full-length wild-type 23S rRNA transcript hydrolyzed the peptidyl-tRNA analog quantitatively, whereas in ribosomes containing gapped-cp-reconstituted 50S subunits the fraction of P-site substrate that was converted into product was in average between 20% (cp2623-2576) and 50% (cp2069-2043).
The release factor-independent peptidyl-tRNA hydrolysis assay using 70S ribosomes (containing ‘gapped-cp-reconstituted’ 50S subunits assembled from 10 pmol cp-23S rRNA and 4 pmol native 30S subunits from was performed as described () with the following modifications: 70S were programmed with 30 pmol of a 98 nt long mRNA transcript () placing a unique Met codon in the P-site, incubated with 0.3 pmol formyl-[H]Met-tRNA (30 000 cpm/pmol) and the reaction was initiated by the addition of 780 pmol of the synthetic CCA tri-nucleotide (Dharmacon Research, Inc.) and 30% acetone. The final reaction volume was 36 µl.
To measure transpeptidation activities of the gapped-cp-reconstituted 50S subunits carrying nucleoside analogs in the PTC, the puromycin reaction was employed. Therefore, 10 pmol reconstituted 70S were programmed with 60 pmol synthetic mRNA containing a unique AUG codon and incubated with 0.3 pmol formyl-[H]Met-tRNA (30 000 cpm/pmol) in order to fill the P-site. Puromycin was added to a final concentration of 2 mM, the peptidyl transferase reaction was performed, and the reaction product identified as described ().
The gapped-cp-reconstitution system offers the advantage of performing nucleoside analog interference studies to unravel crucial 23S rRNA groups for PTC functions (,). Here we applied this experimental tool to characterize the functionally pivotal groups on active site PTC residues for RF1-mediated peptide release. Previous standard mutagenesis studies highlighted the potential involvement of A2602 (and to a lesser extent U2585) in the catalysis of peptidyl-tRNA hydrolysis (,). We therefore constructed a cp-23S rRNA that harbors its novel 5′-and 3′-ends at positions 2623 and 2576 (cp2623-2576), respectively, which places a gap of 45 nt inside the PTC encompassing both prime candidates A2602 and U2585 (A and B). While this 23S rRNA variant has been shown to possess clear peptidyl transferase activity when the missing rRNA fragment was added during the assembly (), no RF1-mediated peptide release could be measured (). Obviously the precise positioning of helix 93 (harboring residue A2602) which packs against helix 74 of 23S rRNA via a GNRA tetraloop/helix interaction (), cannot optimally be achieved under these reconstitution conditions. In order to improve the catalytic performance of this gapped-cp-23S rRNA, the synthetic 45-mer RNA fragment was ligated to the 3′-end of the cp-23S rRNA transcript via the ‘splinter ligation’ approach (). To test whether this ligated cp-23S rRNA construct would be active in peptidyl-tRNA hydrolysis, we first generated a cp-23S rRNA placing its 5′- and 3′-ends at positions 2623 and 2622, respectively, by transcribing the corresponding PCR template (). This cp-23S rRNA mimics a 100% successful ligation event of the 45-mer RNA oligo to cp2623-2576 (A). Under the applied conditions, P-site-bound f-[H]Met-tRNA could be efficiently and almost quantitatively hydrolyzed (B). In accordance with our previous findings (), deletion of the entire nucleotide A2602 also in this cp-23S rRNA context, completely abolished the hydrolytic activity of the PTC (B). Thus, it appears that the ligation approach could potentially solve the functional problems of gapped-cp-reconstituted 50S containing the cp2623-2576 construct in peptidyl-tRNA hydrolysis.
Ligation efficiencies for the synthetic 45-mer RNA oligonucleotide (either containing the wild-type sequence or specific nucleoside modifications) were determined via a modified primer extension approach (see Materials and Methods section) to be in the range from 20% to 55% (data not shown). Importantly, while ligation efficiencies varied between different experiments, very similar efficiencies were obtained within one set of ligations performed in parallel (SDs < 23%). Therefore, we always compared the peptide release activities of gapped-cp-reconstituted 50S carrying different modifications that originated from the same set of parallel ligation experiments. This minimizes the possibility that potential differences in the peptide release activities between differently modified gapped-cp-reconstituted 50S result from varied ligation yields. Reconstituted 50S particles containing cp-23S rRNA with the ligated wild-type RNA fragment showed clear RF1-dependent peptide release activity (B). Importantly, control experiments revealed peptidyl-tRNA hydrolysis to be ribosome-dependent, RF-dependent and strictly required an A-site-bound stop codon on the 30S subunit (B and C). Furthermore, the peptidyl-tRNA hydrolysis activity was sensitive to known inhibitors of translation termination, such as sparsomycin and hygromycin A (C). These controls strongly indicate that an authentic peptide release reaction is monitored in our experimental system. With the ligated wild-type sequence the gapped-cp-reconstituted 50S particles catalyzed peptidyl-tRNA hydrolysis with a rate of 0.1/min, which is about 300-fold slower than reported termination rates in an optimized translation system employing native ribosomes and factors (). Despite these intrinsic limitations of the gapped-cp-reconstitution approach (,), our experimental system was able to assess severe reductions in RF1-triggered peptide release such as those observed with the 2602 deletion mutant (2602Δ) or the 2602C base mutation (B and ). Previously, very similar results were obtained with these two mutants in the context of assembled full-length 23S rRNA transcripts with native 5′- and 3′-ends () as well as with derived 2602 mutant ribosomes (). Therefore, the experimental system appears to report reliable PTC activities, and is therefore amenable for nucleoside modification interference studies.
We first introduced the abasic site analog at A2602 (C), the prime candidate in the PTC residue for peptide release. Unexpectedly, this modification did not interfere with the hydrolytic activity of the PTC. In fact, a slight but reproducible stimulation of peptide release was observed (A and ). It appears that a PTC that does not harbour any base at position 2602 is actually more capable of releasing the peptide from the peptidyl-tRNA than a ribosome with the C mutation (). This suggests that the 2602C base mutant is trapped in an unproductive conformation and explains why this mutant ribosome never produces the same product yield compared to wild-type ribosomes after 30 min of incubation [B and ()]. These data also exclude the possibility that the crucial functional group of A2602 resides on the nucleobase. Removal of the 2′-hydroxyl group either in the context of the abasic site analog (by introducing the deoxyribose-abasic modification) or in the context of the complete nucleoside (by introducing deoxy-adenosine) also did not show significant reductions (A and ). However, introducing the C3-linker modification, which lacks in addition to the entire nucleobase also the C1′, C2′ and the O4′ of the ribose sugar (C), almost completely deprives the PTC of any catalytic power to hydrolyze the P-site-located peptidyl-tRNA. Only marginal amounts of released f-[H]Met above the background could be measured after a prolonged incubation time (A). Compared to the gapped-cp-reconstituted 50S subunits carrying the ligated wild-type RNA fragment, the rate of peptidyl-tRNA hydrolysis with the C3-linker variant decreased by at least 33-fold (). Increasing the RF1 concentration 7-fold (from 0.8 to 5.6 µM) did not rescue any release activity of the 2602 C3-linker modified PTC, thus strongly indicating RF binding deficiencies not to be the reason for the diminished termination activity (data not shown). To more directly clarify this, we applied a peptidyl-tRNA hydrolysis assay that is independent of ribosome-bound release factors. In this set up, the hydrolytic power of the PTC is activated by A-site-bound deacylated tRNA or the 3′ terminal CCA end thereof (). We employed this RF-independent assay with the CCA tri-nucleotide to test the peptide release activities of all 50S subunits carrying modifications at position 2602. Also under these conditions the C3-linker modification at 2602 almost completely inhibited peptidyl-tRNA hydrolysis, while all other modifications showed no or only minor inhibitory effects (). This is independent evidence that the severely reduced activities of 50S subunits carrying the C3-linker at 2602 are resulting from a catalytic deficiency of the PTC to hydrolyze peptidyl-tRNA and is independent of the bound A-site substrate.
Like A2602, also U2585 has been shown to be a structurally flexible residue in the catalytic cavity of the PTC in various 50S subunit crystal structures (,,,,,). Furthermore, mutations at this base also showed reductions in peptide release activities, albeit to a weaker extent compared to A2602 (,). Here we introduced the abasic, the deoxy-uridine and the C3-linker modification at U2585 and tested the chemically engineered subunits in the termination assay. It turned out that none of the U2585 modifications interfered with RF1-medited peptide release. In fact all of them showed slightly enhanced (∼ 2-fold) peptide release rates (). Based on crystallographic studies it has been suggested that the base of U2585 protects the ester bond of P-site-bound peptidyl-tRNA from premature hydrolysis during the elongation phase of protein synthesis (). We therefore tested if the accelerated peptide release rates of the abasic 2585 subunit are the result of an RF-independent peptidyl-tRNA hydrolysis. To this end, the 2585-abasic particle as well as gapped-cp-reconstituted 50S carrying the wild-type sequence were incubated in the absence of RF1. Removal of the nucleobase at U2585 indeed slightly increased the premature background hydrolysis of peptidyl-tRNA. Compared to the wild-type control the background increased ∼ 2-fold in the 2585 abasic PTC (300 versus 600 cpm) after an incubation of 30 min. However, this amount of released f-[H]Met does not completely account for the observed release stimulation of the abasic 2585 50S subunits, since it only raises the measured product yield by 15% (data not shown).
To investigate whether any of the other active site residues carry functional groups essential for peptide release, we introduced deoxy-ribose nucleosides, the abasic analog as well as the C3-linker modification at U2506, A2451 and C2063 (). It is of note that all of the respective cp-23S rRNA constructs were active in RF1-mediated peptide release even when the synthetic RNA oligomer, that was required to fill the introduced gap, was added during reconstitution, and therefore not ligated to the 3′-end of the cp-23S rRNA transcript. Also the reaction rates increased compared to the cp2623-2576 construct used to study A2602 and U2585 and were in the range from 0.2/min (cp2468-2440) to 1.0/min (cp2523-2483). This likely indicates that none of these regions of the PTC is equally important to promote peptidyl-tRNA hydrolysis compared to the A2602 region. In agreement, none of the modifications introduced at U2506, A2451, or C2063 had equally severe effects on termination compared to the 2602 C3-linker (B and ). With the exception of the A2451 C3-linker modification (see subsequently), all chemically modified 50S particles reached essentially the same product yield compared to the wild-type control after 30 min of incubation and the peptide release rates were only slightly affected (B and ). The strongest effects in these experiments were seen with the PTC containing the C3-linker at position 2063 whose release rate was inhibited 5-fold (B). However, similar rate reductions were also measured in the puromycin reaction with this modified ribosome (6-fold), thus indicating a more general, probably conformational, defect of the PTC carrying this 2063 modification for 50S-catalyzed reactions (D). We would like to point out here that removing the 2′-hydroxyl group at A2451, which was shown before to be crucial for peptide bond formation (,), reduced RF1-mediate peptide release only marginally by 1.2-fold (). This indicates the catalytic mechanism for peptide bond formation to be distinct from that of peptide release, as suggested earlier ().
It is possible that any changes in the peptide release activities of chemically modified 50S particles, especially the 2602 C3-linker variant, are the result of varying reconstitution efficiencies. To investigate this possibility we employed the modified 50S particles in a peptidyl transferase reaction using puromycin as A-site substrate. It turned out that all the 2602 modified ribosomes, including the C3-linker variant, produced the same amount of reaction product in the peptidyl transferase assay compared to the wild-type particles (C). We have established previously that both peptide release and peptide bond formation are single turnover reactions under the applied conditions (). Thus, these data indicate that comparable fractions of catalytically competent 50S particles were assembled with all 2602 variants, and therefore exclude the possibility of gross reconstitution deficiencies of the C3-linker containing cp-23S rRNA. Similarly, ribosomes harboring modifications at position C2063, a residue not tested previously in the gapped-cp-reconstitution system, reached wild-type levels of product yields in the puromycin reaction (D). The effects of placing deoxy-nucleoside as well as abasic or C3-linker modifications at positions U2585 and U2506 have been shown previously not to affect the peptidyl transfer rates (). The only modified 50S particle whose reconstitution efficiency could not be assessed by employing the peptidyl transferase reaction was the 2451 C3-linker variant. This modification deprives A2451 of the key functional ribose 2′-hydroxyl, and therefore renders the PTC inactive for transpeptidation (). Nevertheless, since the product yield of the 2451 C3-linker variant in the peptide release assay was only reduced by ∼2-fold (data not shown) suggest that a catalytically competent 50S can even be formed in the presence of this rather rigorous nucleoside analog introduced at this central PTC residue.
Based on the results of previous mutagenesis studies an unequivocal functional role for the universally conserved inner shell nucleobases for PTC-catalyzed reactions could not be assigned [reviewed in ()]. Here we present evidence that none of the 23S rRNA nucleobases at A2602, U2585, U2506, A2451 or C2063 harbor fundamentally crucial functional groups for peptidyl-tRNA hydrolysis since high peptide release activities remained when abasic site analogs were introduced at those sites (). The severe reductions seen before in ribosomes carrying base changes at A2602 (,) or U2585 () are therefore likely the result of a conformationally distorted PTC that trapped the active site in a non-productive state for RF1-mediated peptidyl-tRNA hydrolysis. In support of this conclusion, we found strongly stimulated peptidyl-tRNA hydrolysis rates when the mutant base was entirely removed from the 2602C (B and ) or the 2585C (data not shown) mutant 50S particle. Likely these active site residues require a certain amount of structural flexibility for full peptide release activity which is obviously not necessarily the case when an incorrect base is attached to the ribose. However, full termination activities can be re-gained when the mutant bases are removed at 2602 or 2585 by introducing abasic site analogs thus hinting at an important contribution of the ribose moiety for translation termination. Introducing deoxy-ribose nucleosides at A2602, U2584, U2585, U2506, A2451 or C2063 did not interfere significantly with peptide release, excluding the possibility that any of these 2′-hydroxyl groups are crucial for peptidyl-tRNA hydrolysis (). However, further minimization of the ribose moieties of active site residues by removing in addition to the base also the C1′, C2′ and O4′ of the sugar by introducing the C3-linker (C) clearly emphasizes the importance of a single ribose, namely at position 2602 (A and ). Compared to the abasic 2602 ribosome, the C3-linker containing PTC showed at least 50-fold slower peptidyl-tRNA hydrolysis rates. Peptidyl-tRNA-binding deficiencies to the P-site in the 2602 C3-linker ribosome are unlikely since essentially wild-type levels of peptidyl transfer activities remained (C) and none of the interactions of P-tRNA with the 23S rRNA that have been identified recently to be important for efficient peptide release (), have been manipulated. Applying an RF-independent peptidyl-tRNA hydrolysis assay which was activated by the A-site-bound 3′-terminal CCA end of deacylated tRNA, again revealed a significantly decreased activity when the C3-linker modification was introduced at position 2602 (). This strongly suggests that the reason for the diminished peptidyl-tRNA hydrolysis activities of the modified PTC was not due to A-site substrate-binding deficiencies. Since the CCA tri-nucleotide required in the RF-independent assay interacts solely with the 50S part of the ribosomal A-site (), we can further conclude that no function of the 30S decoding center is causally linked to the observed compromised peptide release activity of the PTC with the C3-linker modification at position 2602. Thus, even though the P-site (peptidyl-tRNA) and A-site (RF1 or CCA) substrates are most likely present, the PTC carrying the C3-linker at 2602 was essentially unable to hydrolyze the peptidyl-tRNA, indicating that the nucleophile (the water molecule) is absent or strongly misplaced. The same C3-linker modification at all the other investigated PTC residues had no or relatively minor (2506, 2451 and 2063) inhibitory effects (). This highlights the distinct functional role of the A2602 ribose ring for peptidyl-tRNA hydrolysis.
How can these findings be explained in the context of the proposed model of translation termination, in which A2602 has been suggested to coordinate and possibly activate the water molecule for the nucleophilic attack on the ester bond of peptidyl-tRNA ()? With the exception of the O4′ position, which possesses lone-pair electrons, none of the crucial positions of the 2602 ribose (the C1′ and C2′) identified here has any functional groups to hydrogen bond to a water molecule or a hydrated metal ion. Furthermore, the distance of the 2602 ribose to the position where the attacking α-amino of aminoacyl-tRNA has been located in the crystal structure, and from where the nucleophilic water molecule is also supposed to launch its attack during peptide release, appears to be too large for the direct coordination of the hydrolytic water. The distance of the A2602 O4′ to the nitrogen atom of the α-amino group of puromycin is 11.5 Å (). Even though we cannot completely discard the possibility for an important direct water coordination by the ribose O4′ during the peptide release reaction, it seems more likely that A2602 functions as a molecular switch in the ribosome that regulates the specificity of the PTC from amide bond formation, in the case where aminoacyl-tRNA is located at the A-site, to peptidyl-tRNA hydrolysis when the RF is bound. For this switching function the nucleobase at 2602 is not actually required, however, an intact ribose moiety appears to be strictly necessary. For this crucial task during translation termination, however, it is not important whether the sugar at position 2602 is a ribose or a deoxyribose, since the PTC carrying the deoxyribose-abasic analog showed wild-type-like peptidyl-tRNA hydrolysis activities (). This modified model of peptide release is compatible with the structurally mobile nature of A2602 which involves movements of both the ribose and the base () as well as with the recent structure of RF1-bound 70S particles, showing close proximity of the universally conserved GGQ mini-motif of the RF to A2602 (). However, which group actually activates or positions the water molecule for optimal attack? The A2451 2′-hydroxyl which has been shown to be pivotal for catalyzing peptide bond formation (,) does not seem to play an equally important role in peptide release, and thus does not qualify for activating the nucleophile. Groups at other PTC residues are also not critical (), leaving the 2′-hydroxyl of A76 of P-tRNA, which plays an important role in peptide bond formation (), or a group on the RF as potential candidates for possessing the catalytic group for peptidyl-tRNA hydrolysis. Since all previous mutational studies suggest that peptide release is an RNA-promoted reaction rather than directly catalyzed by the protein RF [reviewed in ()], we favor the A76 2′-hydroxyl of P-site-bound peptidyl-tRNA. In fact, a clear inhibition of RF1-mediated peptide release has been observed when this 2′-hydroxyl was removed from peptidyl-tRNA (R. Green, personal communication). In this scenario, A2602 functions as a molecular trigger that is pulled in response to the A-site-bound RF to allow the nucleophilic water to access the active site. In our view, A2602 however does not simply function as a passive lid that prevents premature water access to the peptidyl-tRNA ester bond, since removal of the entire nucleotide in the 2602Δ mutant did not increase the background hydrolysis of P-site-bound fMet-tRNA in the absence of RF1 (data not shown). It is possible that the ribose at A2602 specifically channels or guides the hydrolytic water molecule into the PTC which is then activated and optimally positioned for the attack by a non-ribosomal group, most plausibly by the A76 2′-hydroxyl of peptidyl-tRNA. It is of note that as for transpeptidation where the 2′-hydroxyl at A2451 is crucial (,), the ribosome again provides a 23S rRNA backbone group, the ribose moiety at A2602, rather than any group on a nucleobase to promote the second chemical reaction of the PTC, namely peptidyl-tRNA hydrolysis. |
Dendritic cells (DCs) contribute to the non-specific innate immune response to viral infection, as well as to the development of viral antigen-specific adaptive immunity. A salient component of the early DC response to viral infection is the induction of interferon beta (), a secreted cytokine that complexes with type I IFN receptors to initiate a coordinated cellular program leading to widespread viral resistance (,). expression is a crucial step in both induction of innate immunity and in DCs maturation leading to induction of adaptive immunity. Because the induction of is so critical for the immune response, the mechanism underlying its control has been the subject of detailed study. The transcriptional activation of the gene requires the cooperative assembly of a multi-subunit enhanceosome, which includes an AP1 complex, IRF dimers, NFκB () and HMGI(Y) (,). The crystal structure of several components forming the enhanceosome has been solved ().
Pathogenic viruses produce protein antagonists that interfere with the innate immune response, for example by suppressing transcription factor activation and interferon induction. Newcastle disease virus (NDV) is an avian virus that lacks a functioning antagonist in human cells (). Because NDV infection efficiently stimulates the innate immune responses and the maturation of human DCs, it provides an ideal experimental perturbation with which to study unimpeded DC responses to viral infection. To assess the noise in gene induction, we quantified the variation in gene expression in individual primary human DCs following NDV infection using one-step quantitative reverse-transcription real-time PCR (A).
Gene expression noise has a significant effect on many biological processes, such as contributing to phenotypic variability of genetically identical organisms and determining cellular fate following viral infection in bacteria and eukaryotic cells (). Gene noise has been investigated using engineered gene reporters in unicellular organisms () and has recently been studied in metazoans (). Previous researchers distinguished two components of the total genetic noise to be intrinsic noise and extrinsic noise (). Intrinsic noise results from the probabilistic nature of molecular processes, such as transcription and translation resulting from the limited number of molecules reacting within an individual cell. Extrinsic noise results from cell-to-cell variations in components involved in generating the response. Total noise is the contribution of both processes to cell-to-cell variation.
Previous studies of noise have largely utilized protein reporter assays, which add translational noise to the data generated as well as measurement noise due to cellular autofluorescence and other limitations of direct fluorescence assays. In our experiments, we directly measured single cell mRNA expression in order to characterize the total noise of induction following virus infection of human dendritic cells (DCs). We employed a new real-time RT-PCR-based approach for quantification of transcripts. After virus infection, single DCs were screened and sorted directly into 384-well PCR plates, which contained cell lysis buffer and RNAse inhibitor in each well. One-step real-time reverse transcription PCR was performed and the total number of transcripts in each cell were measured. To determine the intrinsic components of total noise, we measured the differential expression from two alleles of heterozygous human DCs, using a common readout polymorphism of (rs1051922). The method of measurement utilized a second qRT-PCR (for details see B and Methods section) ().
Our results show a considerable cell-to-cell variability of gene expression, of which intrinsic noise is a significant component. Intrinsic noise cannot be due to cell-to-cell fluctuations in viral load, the number of transcription factors, signaling components or polymerases, since these would affect both alleles equally in a given cell. The source of intrinsic noise must lie within the transcription process itself, and therefore is likely associated with the complexity of enhanceosome assembly on the promoter (). The single cell model we have developed to interpret our results on noise is based on the premise that intrinsic noise and therefore allelic imbalance results precisely from the stochasticity of enhanceosome formation.
All human research protocols for this work have been reviewed and approved by the IRB of the Mount Sinai School of Medicine. Monocyte-derived DCs were obtained from healthy human blood donors following a standard protocol as described elsewhere (). Briefly, human peripheral blood mononuclear cells were isolated from buffy coats by Ficoll density gradient centrifugation (Histopaque, Sigma Aldrich) at 2300 r.p.m. and CD14 monocytes were immunomagnetically purified by using a MACS CD14 isolation kit (Miltenyi Biotech). Monocytes (0.7 × 10 cells/ml) were differentiated into immature DCs by 5–6 day incubation in 1 ml DC growth media with RPMI Medium 1640 (Invitrogen/Gibco) supplemented with 10% fetal calf serum (Hyclone), 2 mM of -glutamine, 100 U/ml penicillin and 100 g/ml streptomycin (Pen/Strep) (Invitrogen), 500 U/ml hGM-CSF (Preprotech) and 1000 U/ml hIL-4 (Preprotech).
The recombinant Hitchner strain of Newcastle disease virus (rNDV/B1) was prepared as described by Park . (). Aliquots of allantoic fluid were harvested, snap frozen and stored at −80°C. All virus preparations were free of bacterial contamination, as tested by the inoculation of blood agar plates. NDV virus was titered by immunofluorescence 18 h after infection of Vero cell plates using monoclonal antibodies specific for NDV HN protein (Mount Sinai Hybridoma Core Facility) followed by addition of anti-mouse IgG-FITC and visualization using fluorescent microscopy. NDV stocks were appropriately diluted in Dulbecco's Modified Eagle Medium (DMEM) and added directly into pelleted DCs at a multiplicity of infection (MOI) of 0.5 (). After incubation for 40 min at 37°C, fresh DC growth medium (without GMCSF and IL-4) was added back to the infected cells (1 × 10 cells/ml) for the remainder of the infection. Virus-free allantoic fluid was added to additional tubes of cells to serve as a negative control.
Cultured human DCs (1 × 10 cells) were divided into two samples (5 × 10 cells each). Cells in both samples were infected by NDV at an MOI of 0.5, and actinomycin D was added at 5 µg to one sample 9 h after virus infection to inhibit transcription. Total RNA samples were isolated using a Qiagen RNeasy mini kit every half an hour from 10 h through 12 h. The concentrations of RNAs were determined using a Nanodrop® ND-1000 Spectrophotometer. One hundred nanogram of each RNA sample was used as template for qRT-PCR to determine the expression levels of and . The data were normalized to , and . All PCR primer sequences are given in Supplementary Table 1.
Virus infected DCs were resuspended on ice in PBS at a concentration of 2–5 × 10 cells/ml. Cells were filtered through a 50 μm filter to remove aggregates prior to FACS (fluorescence activated cell sorting) sorting. Single DC was screened and sorted by visual light scatter (MoFlo high speed cell sorter) directly into 384-well bar-coded PCR plates (Applied Biosystems), which contained 5 μl cell lysis buffer [4 mM magnesium acetate (Sigma), 0.05% NP40 (Sigma), 0.8 U/μl Protector RNAse Inhibitor (Roche Applied Sciences)] in each well. Sorted DCs were immediately placed on dry ice and stored at –70°C to prevent RNA degradation.
One-step real-time reverse transcription PCR was performed on an ABI PRISM 7900HT sequence detection system (SDS) (Applied Biosystems) based on the manufacturer's recommended standard protocol except for cycling conditions. Following incubation at 65°C for 30 min, as required for the reversed transcription step, amplification was carried out for 50 cycles using 95°C for 30 s and 60°C for 30 s. The one-step protocol was made possible by using AccuRT, an aptamer-based hot-start, magnesium-activated thermostable DNA polymerase that extends a primer on either an RNA or DNA template (). An aliquot of 5 µl of 2× PCR reaction mix [2× buffer, 4 mM magnesium acetate, 1 μM each primer, 0.4 mM each dNTP (0.8 mM dUTP replacing dTTP) and 0.375 U/μl AccuRT with aptamer (provided by Roche Molecular Systems; research samples of this polymerase may be obtained from Dr Thomas Myers ()] was added into each well of a FACS single cell-sorted 384-well PCR plate using a multichannel pipettor. mRNA, mRNA, control gene ribosomal protein 9 () mRNA or NDV L gene RNA in each cell was directly amplified into 106, 156, 76 and 125 bp amplicons, respectively (primers’ sequences are found in Supplementary Table 1). At the end of each reaction, crossing threshold (Ct) was determined at a manually adjusted level that reflected the best kinetic PCR parameters, and melting curves were acquired and analyzed to validate the products. The number of transcripts in each cell was determined by relating the Ct value to the standard curve for the corresponding gene.
and control gene were genotyped at selected readout SNPs (rs1051922 for , rs1065744 for ) and DCs from heterozygotes were selected for single cell analyses. and amplicons from single cell real-time PCR measurements were used as templates to quantify the amplification from the two alleles using allele-specific PCR (ASPCR) (). ASPCR uses one common primer and a second allele-specific primer based on the 3′ nucleotide. Allelic imbalance (AI) is defined as the difference in the number of transcripts from the two alleles (M1−M2) divided by the total transcripts (M1 + M2), expressed as percentage. This difference was calculated based on the differences in Ct values (ΔCt) for the two allele-specific primers.
This ASPCR assay was performed on an ABI PRISM 7900HT SDS (Applied Biosystems). Single cell reverse-transcriptase PCR products were pre-selected by melting curve screening to eliminate those containing primer–dimer and diluted 1:10. Five microliter dilutions were added as template into the reaction mix of equal volumes as described elsewhere () [100 mM Tricine buffer, pH 7.5, 100 mM KOAc, 16.0% glycerol, 2% dimethyl sulfoxide, 0.4 mM dATP, dCTP and dGTP, 0.8 mM dUTP, 6 mM magnesium acetate, 2× SYBR Green (Invitrogen), 1.0 μM primers, 0.4 Unit ▵Z05 GOLD DNA Polymerase (a Hotstart DNA polymerase with improved discrimination against misextension kindly provided by Roche Molecular Systems; research samples of this polymerase may be obtained from Dr Thomas Myers ()] in 384-well-format PCR plates. The cycling conditions were 95°C for 12 min to activate the polymerase, 40 cycles of 30 s at 95°C, 30 s at 58°C and 30 s at 72°C. In an alternative PCR buffer [50 mM Tris buffer, pH 7.5, 50 mM KOAc, 2% glycerol, 1× BSA (0.1 mg/ml)], other hot-start Taq DNA polymerases perform well in ASPCR, but not as well as ▵Z05 (data not shown). ASPCR for and produces 66 and 59 bp amplicons, respectively, using the primers listed in Supplementary Table 1. All assays were replicated and normalized against heterozygote DNA as previously described ().
The systematic error of our measurements was determined by performing the entire qRT and ASPCR steps using serial dilutions of total RNA extracted from NDV-infected DCs from a single heterozygous donor as reaction templates. Total RNA was diluted from 5000 copies to 19 copies per qRT reaction. Twelve PCR products from each dilution were then used as the template for the ASPCR assays.
A multiplex single cell TaqMan assay has been established to study the correlation between the level of NDV L gene and transcription in single human DCs. The one-step real-time reverse-transcription-TaqMan assay was performed in single DCs by using AccuRT as described above for single cell real-time RT-PCR with the only variation being 0.4 μM primer pairs and 0.6 μM TaqMan probes and the absence of SYBR Green. The sequences of the primer sets and TaqMan probes for mRNA (FAM labeled) and NDV L gene (HEX labeled) are listed in Supplementary Table 1.
The model is based on the premise that the observed noise in mRNA is a result of stochasticity in the time of onset of transcription. More precisely, the complexity of enhanceosome formation prior to transcription initiation is assumed to provide the dominant contribution to the stochastic behavior. This assembly involves binding of the transcriptional activators, the heterodimers NFκB and AP-1 (ATF-2-c-Jun) and the interferon regulatory factors IRF to form the enhanceosome () as well as the architectural protein HMG-I(Y) (,). If these proteins bind weakly to their promoter region (,), the time at which enhanceosome formation is complete fluctuates between alleles in a given cell, providing a source of intrinsic noise. Once stabilized, the enhanceosome is assumed to persist and stimulate transcription during the course of the experiment. The complex processes that occur once the enhanceosome is stabilized, including the building of the Pol II complex etc., are not explicitly included and only a stochastic transcription of mRNA production is modeled.
For simplicity, in the following description of the model we will denote HMG-I(Y) by P1, NFκB by P2, AP1 by P3 and the IRFs by P4. The model thus comprises four proteins P1, P2, P3 and P4 that bind without interacting, as indicated by experiment (). Since the binding occurs through DNA conformational changes (), we represent the corresponding cooperative behavior through protein binding in a well-defined order. While more than one IRF factor may be required and two HMGI-(Y)s are needed, we include just four reactions in our simple model, since these account reasonably for the combinatorial nature of the sequential, cooperative assembly and consequent stochasticity in the time of transcription initiation. Since the experiment measures AI, we distinguish enhanceosome binding to either allele.
The relevant reactions are given by:
One has a similar set of equations for promoter region 2 of allele 2 where we have used the same reaction rate constants since the alleles are assumed to be functionally identical. Once the enhanceosome is bound (complex 13 and the corresponding complex for allele 2), transcription proceeds according to a Poisson process at the rate of ten mRNA molecules of per minute for both allele 1 and allele 2. Transcription itself is thus noisy; we have checked that this makes a small contribution to AI (intrinsic noise), which is dominated by fluctuations in the time of transcription initiation. The transcription by the gene in allele 1 is represented simply by:
Experimental measurements are made on single cells, not across a population. Therefore, we have employed the well-known Gillespie algorithm to simulate AI (). In the algorithm, the time intervals between reactions is chosen from an exponential distribution and the reactions themselves are chosen randomly according to weights that are proportional to rate constants and number of particles involved in the reaction.
Since measurements of the amounts of enhanceosome components do not exist, we take for simplicity the initial numbers of P1, P2, P3 and P4 to be equal to 12 000, which for a volume of 500 µ corresponds to a concentration of 40 nM. Within the implementation of the model, the rate constants and the corresponding copy numbers of the transcription factors enter as products and thence, independent specification of the two numbers is not crucial. For reasons of simplicity as well we take all the binding constants to be equal and given by 1 = 3 = 5 = 7 = 4.85 × 10 nM s, and the unbinding constants are equal and given by 2 = 4 = 6 = 24.25 × 10 s. We have also tried varying the rate constants within factors of 1/5 to 5 preserving the overall timescale and find that our results are robust at the semi-quantitative level.
There is an additional time delay between the initial infection and the subsequent formation of active transcription factors due to a number of steps: the viral RNA being presented to the cell, the activation and nuclear translocation of NFκB from its sequestered form by ubiquitination of IκB and activation of the other transcription factors. The experimental measurements, which show a strong component of extrinsic noise (see legend of ), suggest this directly since for different cells virus-induced transcription of mRNA starts at different times, with most cells being active after 8 h of viral infection, but some cells starting transcription hours earlier. To account for this, the starting time of chemical reactions in the model is chosen from a Gaussian distribution centered at 6 h with a width of 1 h, with the requirement that no cell starts transcription within 4 h of viral infection. We have also modeled a cascade of processes separately and parameterized the probability distribution for the time delay before the starting time for the reactions of our model as a Gamma distribution. The results obtained are quantitatively similar.
To maximize the fraction of infected cells while minimizing the number of cells infected by multiple viruses, experiments were performed at an infectious particles/cell ratio [multiplicity of infection (MOI)] below one. To select the time points for studying noise, we determined the fraction of cells expressing at various times following NDV infection. After NDV infection, the number of cells expressing increased between 6 and 10 h and plateaued at ∼10 h (A). Because the gene lacks introns, the PCR reaction cannot discriminate between the four strands of genomic DNA in each cell and the transcribed mRNA. Therefore a low level of signal was detected in the absence of viral infection (B inset). In contrast, the virus-exposed cells showed a second broad distribution of expression levels resulting from highly variable single cell induction (B). We also determined the individual cell-to-cell expression of a constitutively expressed housekeeping gene () as well as of an additional virus-induced gene (RIG-I). In contrast with in activated cells, showed a narrow distribution indistinguishable in infected and uninfected cells (C). Like , the virus-induced mRNA also showed a narrow distribution (Supplementary Figure 1). The measurement of , which also served to assess the efficiency of sorting of single, viable cells, revealed only 2% measurement failure due to sorting errors.
In a separate experiment, we used a single cell multiplex TaqMan assay to determine the covariation in expression of mRNA and NDV viral RNA. Regression analysis showed that the level of mRNA was independent of the level of NDV replication in the same cell (D). Furthermore, approximately one-third of the cells infected with virus, as shown by expression of NDV viral RNA, showed no significant induction of (E). Reflecting the high levels of cell-to-cell variation in mRNA in NDV-infected cells, we find that 7% of infected cells accounted for more than 50% of all mRNA synthesized (B). These findings show that the stochasticity of expression is not determined by the stochastic nature of viral RNA replication. It must therefore result from heterogeneity in individual DCs’ responses to virus infection and the process of transcription. We also found that some cells expressed mRNA in the absence of evidence of viral RNA replication. This may be due to the presence of undetected defective interfering particles in the NDV viral stock ().
RNA is stable for at least 2 h after synthesis, until the poly (A) tail is removed and exponential degradation commences (,). To confirm the kinetics of degradation in DCs, control studies were performed by adding actinomycin D 9 h after NDV infection, thus inhibiting transcription. mRNA levels were found to be stable for more than 2 h after actinomycin D treatment, followed by a rapid decay phase (Supplementary Figure 2A). degradation, which showed exponential decay immediately after addition of actinomycin D, served as a control (Supplementary Figure 2B). Coupled with the fact that mRNAs were mainly synthesized 8 h after NDV infection (A), these results indicate that the effects of degradation of mRNA on its level of expression do not affect the interpretation of the single cell assays.
Experiments were performed using DCs from individuals heterozygous for the readout polymorphism, which enabled mRNA from each of the two alleles to be distinguished. We quantified the level of single cell production by qRT-PCR (A). We also quantified the relative number of transcripts measured from each allele within each cell by allele-specific PCR (B). mRNA allelic imbalance (AI) is the difference in mRNA production from the two alleles divided by their sum () and is directly calculated from the ΔCt values for replicate allele-specific PCR reactions (see Methods section). The number of transcripts originating from each of the two alleles (M1, M2) was calculated from the total expression and AI for each cell (see Methods section).
We verified the equality of the distributions of M1 and M2 by a Kolmogorov–Smirnov test ( = 0.23 at 8 h, = 0.99 at 9 h, = 0.46 at 10 h). At low single cell expression levels, the majority of cells showed expression from one or the other allele at 8, 9 and 10 h after infection (A–C). When mRNA expression was greater, most cells showed expression from both promoters. Because these experiments use primary cells, studies were replicated using cells from different donors in order to assess the robustness of the distributions found. The distributions seen in A–C are comparable to those obtained in other experiments (Supplementary Figure 3). The running average of the absolute value of the difference between transcripts from the two alleles (〈 | 1 − 2.| 〉) was calculated and plotted versus the running average of the total transcripts (〈 | 1 + 2 |〉) (Supplementary Figure 4). The similarity of these plots representing different times after infection suggests that the intrinsic noise in (shown by (〈 | 1 − 2.| 〉) depends on the total transcription and is not dependent on time elapsed since viral infection.
In order to quantify the noise, we used the definitions proposed by Elowitz . (). In brief, if M1 and M2 designate the numbers of mRNA transcripts from the two alleles, intrinsic noise , extrinsic noise and total noise are given by:
The level of total noise reflected in the level of single cell expression (B) and the level of intrinsic noise reflected by the level of single cell AI were high, with intrinsic noise making a large contribution to total noise (e.g. η = 1.4, η = 2.0, η = 2.5 at 8 h). Total and intrinsic noise for the housekeeping control gene were much less than those observed for (D). We determined the accuracy of single cell allelic imbalance assays by simulating single cell experiments using dilutions of mRNA isolated from a large number of DCs. These assays showed that the AI standard deviation resulting from measurement error was much lower than the variation measured for mRNA, and that the error was independent of the initial copy number down to fewer than 20 target mRNA copies/cell (Supplementary Figure 5).
The experimental results showed that, in comparison with the other two genes studied, a high level of intrinsic and total noise was observed for mRNA expression. The high levels of single cell AI observed are incompatible with a model involving transient activation and reactivation of the promoters. If the enhanceosome could assemble, activate, disassemble and reactivate on a short timescale relative to these experiments, then the mRNAs originating from each promoter would equalize and single cell AI would not be observed. Therefore, we hypothesized that the differences in M1 and M2 result from differences in the initial formation of each functional transcription complex. To test this hypothesis, we explored whether a minimal stochastic model based on these assumptions would explain the single cell mRNA AI levels.
We simulated enhanceosome assembly and mRNA synthesis using Gillespie's algorithm (). The model we considered is based on the premise that the observed AI results from fluctuations in the time of formation of the enhanceosome complex. These fluctuations produce variability in the time when transcription of starts on each allele. The differences observed experimentally in M1 and M2 in individual cells suggest that the activation of the second allele is delayed by a period up to many minutes (A–C). The intricate series of processes leading up to the beginning of transcription are modeled as a sequence of ordered, slow binding of four proteins, NFκB, IRF3 dimer, AP1 and HMGI(Y), to the enhancer–promoter region of the gene. The proteins occur in adequate numbers so that their binding to the two alleles is uncorrelated. Single cell AI essentially results from fluctuations in the time of initiation of transcription. Once transcription starts, it continues unabated according to a Poisson process, and does not contribute significantly to the intrinsic noise. The model also incorporates variability in the time elapsed between virus infection and activation of enhanceosome components. This source of cellular variability affects the two alleles in a cell in the same way and does not contribute to intrinsic noise. The distribution of single cell AI reproduces the salient features of the measurements obtained experimentally as the comparison between model and experiment at two different time points shows (E–H). The model simulations were robust to a 25-fold variation in the rate constants (see Methods section). These results support the hypothesis that the sequential assembly of the multicomponent enhanceosome represents a major source of intrinsic noise in this system.
The AI (intrinsic noise) determined by simulations of the model and by experimental measurement show a close correspondence (E–H). However, the present simple model is far from complete as far as the sources of cell-to-cell variation (extrinsic noise) that lead to activation of the enhanceosome components are concerned. It only gives a qualitative description of the experimental results for total single cell mRNA yield (compare A–C and Supplementary Figure 6). Refinements of the model and further experiments will be necessary to clarify the extrinsic noise mechanisms contributing to cell-to-cell variability in the levels of mRNA.
The high level of variation in expression from cell to cell is influenced both by intrinsic stochastic processes and by extrinsic cell-to-cell variation in the initial concentrations of key signaling and enhanceosome components. The lack of correlation between single cell viral RNA and mRNA expression levels suggested that differences in viral replication do not contribute to the observed variability in induction. Variation of recruitment of RNA polymerase after enhanceosome formation is also not a significant contributor to the differences in single cell levels of since the level of housekeeping gene such as is tightly controlled across cells. The measurement of single cell AI provided an assessment of the level of intrinsic noise, as both promoters in each cell are exposed to the same environment. Transient differences in the rate of promoter recruitment or of RNA synthesis are likely to contribute little to the intrinsic noise, because any imbalances in transcription rate would average out over the timescale of these experiments. Our experimental data and simulations thus suggested that the intrinsic and extrinsic noise contributing to variations in expression result from variations in the time of formation of the functional enhanceosome.
The level of induction in DC cells responding to viral infection is very broad, ranging from few tens to several thousands of mRNA copies. A major element of this cell-to-cell variability is intrinsic noise, which we measured through a readout polymorphism. Our work is the first to measure cell-to-cell variability on gene expression in human DCs responding to virus infection. Its implication for differential immunity at the population level is unknown. We speculate that high levels of noise may possibly contribute to the overall robustness of the human antiviral response. Our experimental results show that only a few cells respond strongly to viral invasion. This limited response may be adequate to prime the immune system while at the same time avoiding the overreaction of a cytokine storm. The beneficial, neutral or detrimental effects of the noise on the overall immune system response to viral infection remain to be established by further studies.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
In addition to the structural and genes, the human immunodeficiency virus type 1 (HIV-1) genome contains several regulatory genes: and . The HIV-1 Vif (viral infectivity factor) protein, first considered as an accessory factor, is a positive modulator of viral infectivity in several cell types. In particular, Vif is required for efficient HIV-1 replication in so called ‘non-permissive’ cells such as primary CD4+T lymphocytes and macrophages and some T-cell lines (H9, HUT78), whereas HIV-1 lacking (HIV-1Δ) can replicate in fibroblasts and most T-cell lines termed ‘permissive’ ().
This cell type-dependent requirement for Vif can be explained by the expression of APOBEC3G and APOBEC3F, two cellular inhibitors of HIV-1 replication, in non-permissive cells (,). APOBEC3G/3F are members of a large family of cytidine deaminases (). In the absence of Vif, APOBEC3G/3F associate with Gag and RNA during viral assembly and are packaged into virions (). APOBEC3G/3F induce hypermutation of the HIV-1 genome by mediating deamination of cytidine to uracil on the newly synthesized (–) strand DNA during reverse transcription, thus leading to guanosine to adenosine transitions in the viral genome (,). Independently from their catalytic activity, APOBEC3G/3F factors also impair particle infectivity by affecting virion morphology and by destabilizing the reverse transcription complex (). Vif counteracts the antiviral activity of APOBEC3G/3F by several mechanisms. Vif seems to directly impair packaging of APOBEC3G/3F by an unknown mechanism (,), induces degradation of APOBEC3G/3F through the ubiquitination-proteasome pathway (,), and negatively regulates APOBEC3G/3F translation (,,).
Vif is part of a large cytoplasmic ribonucleoprotein (RNP) complex and it is now usually accepted that Vif is packaged into viral particles through interactions with the viral genomic RNA, co-packaged cellular RNAs and the nucleocapsid (NC) domain of Gag (,). Vif defective viruses produced from non-permissive cells display defects not only at early assembly events but also at post-entry steps of infection, resulting in a failure to complete reverse transcription and integration (). Moreover, viral particles produced in the absence of Vif show structural defects such as aberrant core morphology and reduced stability (,,). In particular, NC and reverse transcriptase (RT) were found to be less stably associated with viral cores in the absence of Vif (), explaining in part why Δ virions are defective in the reverse transcription step (,,,). Initiation of reverse transcription is completely impaired in Δ viruses, suggesting that Vif may serve as an auxiliary factor for HIV-1 RT () and allows formation of a functional RNP (). It has also been shown that Vif is able to modulate viral protease (PR) activity and the proteolytic processing of the Gag precursor at the p2/NC site, leading to the possibility that virion incorporation of Vif could stabilize NC intermediates ().
Considering that NCp7 () and Vif () are both involved in the reverse transcription process and in structural rearrangements of HIV-1 RNA (), we were interested to know whether Vif could modulate RNA dimerization, a prerequisite for RNA packaging (,), and the early steps of reverse transcription either alone or in presence of various intermediate processing products of Gag (NCp7, NCp9 and NCp15). In the absence of NC proteins, we found that Vif possesses RNA chaperone activity, resembling but distinct from the chaperone properties of NCp proteins with respect to RNA dimerization, hybridization of tRNA to the PBS, initiation of reverse transcription by RT and the first strand transfer. Surprisingly, Vif inhibited the initiation phase of reverse transcription. At high Vif/NCp ratio (1/3), Vif also inhibited NCp-induced maturation of the RNA dimer and tRNA annealing, whereas both NCp and Vif contributed to increase the processivity of RT. Vif had modest effects on the NCp-induced strand transfer reaction. At low Vif/NCp ratio (1/30), Vif had very limited effects on reverse transcription. Considering that Gag maturation is a highly ordered process that can be modulated by Vif (), and that Vif is a RNA chaperone influencing RNA dimerization, tRNA annealing, RT processivity, and the first strand transfer, our results suggest that during the assembly steps, Vif might be a negative temporal regulator of RNA dimerization and packaging, preventing premature initiation of reverse transcription, while promoting tRNA annealing, a process that could be affected by APOBEC3F/3G.
The donor and acceptor RNAs, corresponding respectively to nucleotides 1–311 and 8607–9229 of HIV-1 genomic RNA (Mal isolate) were synthesized by transcription of plasmids pJCB and pFB1 using T7 RNA polymerase and purified as previously described (). Plasmid containing a 3-nt substitution in the self-complementary sequence of the RNA dimerization initiation site (DIS) loop (pDIS-AAA) () was digested by RsaI prior to transcription to generate mutated RNA 1–311. RNA 1–615 used in the dimerization assays has been obtained after digestion of plasmid pHIV_615 by PvuII (). Prior to 5′-end labeling, 10 µg of tRNA purified from beef liver () were denatured for 2 min at 90°C in 25 mM Tris–HCl pH 8, 0.1% SDS (wt/vol), 15% (vol/vol) formamide, cooled on ice and incubated with 1000 U of BAP (Fermentas) at 70°C for 1 h. After phenol/chloroform extraction and ethanol precipitation, the dephosphorylated tRNA was purified by denaturing polyacrylamide gel electrophoresis (PAGE). Dephosphorylated tRNA (450 ng) was incubated 5 min at 70°C, cooled at room temperature and radiolabeled using 15 U of phage T4 polynucleotide kinase (PNK, New England Biolabs) and 100 µCi of [γ-P] ATP (Amersham), for 30 min at 37°C in the buffer supplied with the enzyme in a 30 µl final volume. Labeled tRNA was then purified by denaturing PAGE. ODN, an 18-mer oligodeoxyribonucleotide complementary to the PBS was chemically synthesized and 5′-end labeled for 45 min at 37°C using 100 µCi of [γ-P] ATP and 10 U of PNK (New England Biolabs). Internal labeling of RNA 1–615 was achieved by addition of [α-P] ATP (Amersham) during transcription ().
Wild-type HIV-1 Vif protein was expressed in with an N-terminal 6-His fusion tag and purified as previously described (). NCp7, NCp9 and NCp15 proteins from NL4.3 (55, 72 and 122 amino acids, respectively), expressed in and purified as described (), were reconstituted with one equivalent of Zn per zinc finger in milliQ HO (Millipore), aliquoted, layered with mineral oil and stored at −80°C (). Wild-type and RNase H(–) E478Q HIV-1 RTs were expressed in with a N-terminal 6-His fusion tag and purified as previously described ().
All annealing reactions were performed using 1 pmol of RNA 1–311 and 0.3 pmol of 5′-end labeled ODN or tRNA. Primer and template were denaturated 2 min at 90°C then cooled on ice. After addition of 0.1 M NaCl and 6 µM ZnCl, samples containing ODN or tRNA were incubated 20 min at 50 or 70°C, respectively, and cooled on ice.
Primer and template were first denatured separately by incubation 2 min at 90°C, ice-cooled and renaturated 10 min at 37°C in 50 mM Tris–HCl pH 7.2, 50 mM NaCl, 6 µM ZnCl and 5 mM MgCl (buffer H). RNAs were then mixed together, incubated 10 min at 37°C, 5 min at room temperature and put on ice while adding Vif and/or NC proteins. In parallel, heat-annealed complexes were formed as described above and adjusted to 50 mM NaCl, 50 mM Tris–HCl pH 7.2 and 5 mM MgCl. Heat- and protein-mediated annealing reactions (10 µl) were incubated 20 min at 37°C, and split in two equal volumes, in order to monitor hybridization and to assay extension by wild-type HIV-1 RT. To monitor hybridization, samples were deproteinized with 1.3 mg/ml proteinase K (Roche) in 8 mM NTPs, 1 mM spermidine, 1.3% SDS and 33 mM EDTA for 1 h at 37°C. Spermidine and NTPs help minimizing aggregation of nucleic acids by NC (). Volume was then adjusted to 60 µl with buffer H and following phenol/chloroform extraction, 50 µl of RNA-containing aqueous phase supplemented with glycerol-containing loading buffer were analyzed by non-denaturing 6% PAGE. Electrophoresis was performed at 4°C in 0.5× Tris–Borate buffer supplemented with 0.1 mM MgCl.
In a typical experiment, 100 nM of unlabeled HIV-1 1–615 RNA fragment were diluted in 10 μl of Milli-Q (Millipore) water with the corresponding labeled RNA (5000 c.p.m., 3–5 nM). Samples were denatured for 2 min. at 90°C, and snap-cooled on ice for 2 min. Dimerization was initiated by addition of Vif and/or NCp7 proteins in conditions disfavoring salt-induced RNA dimerization (50 mM sodium cacodylate pH 7.5, 50 mM NaCl, 0.1 mM MgCl). RNA samples were incubated 30 min at 37°C and deproteinized as above, then re-suspended in glycerol-containing loading buffer, split in two equal volumes and analyzed on a 0.8% agarose gel in native (Tris–Borate 0.5×, MgCl 0.1 mM, run at 4°C) or denaturating (Tris–Borate-EDTA 1×, run at 20°C) electrophoresis conditions. Gels were fixed in 10% trichloroacetic acid for 10 min. and dried for 1 h under vacuum at room temperature. Radioactive bands corresponding to monomeric and dimeric species were visualized and quantified using a FLA 5000 (Fuji).
To assay extension by HIV-1 RT, heat- or protein-annealed tRNA/RNA 1–311 complexes were first treated with 1.5 mg/ml proteinase K in 1.5% SDS for 1 h at 37°C. After addition of 0.3 M sodium acetate, complexes were phenol/chloroform extracted and precipitated in ethanol. Following centrifugation and vacuum drying, nucleic acids pellets were mildly solubilized in buffer E1 (50 mM Tris–HCl pH 8.0, 50 mM NaCl, 1 mM DTT, 6 mM MgCl and 1 mM spermidine). tRNA/RNA 1–311 complexes were incubated 4 min at 37°C with 70 nM HIV-1 RT and reverse transcription was initiated by addition of dTTP, dGTP, dCTP (100 µM each) and 15 µM ddATP as a chain terminator. After 45 min at 37°C, polymerase activity was blocked by adding 20 mM EDTA, samples were phenol/chloroform extracted and precipitated in ethanol. Purified nucleic acids were re-suspended in urea-containing loading buffer and analyzed by 8% PAGE. Radioactive bands were visualized and quantified using a BioImager BAS 2000 (Fuji).
To test the influence of Vif on the initiation step of reverse transcription, heat-annealed tRNA/RNA 1–311 complexes were incubated 20 min at 37°C in buffer H with increasing concentrations of Vif. Half of the reaction medium was treated as described above to monitor hybridization, then 90 nM HIV-1 RT and 1 mM spermidine were added on the remaining half and after 4 min incubation at 37°C, reverse transcription was initiated by addition of dTTP, dGTP, dCTP (50 µM each) and 7.5 µM ddATP as a chain terminator. After 20 min at 37°C, polymerase activity was blocked by adding 33 mM EDTA, 1% SDS and 1 mg/ml proteinase K. After proteolysis for 1 h at 37°C, samples were phenol/chloroform extracted and precipitated with ethanol. Nucleic acids were re-suspended in urea-containing loading buffer and analyzed by 8% PAGE.
For (–) strand strong-stop DNA (ssDNA) synthesis and strand transfer experiments, we used the 5′-end labeled ODN primer annealed to donor RNA 1–311. In both assays, NC and/or Vif proteins were incubated with 380 nM WT or E478Q HIV-1 RT for 4 min at 37°C in buffer E2 (0.1 M NaCl, 60 mM Tris–HCl pH 8.0, 6 mM MgCl, 50 µM dNTPs, 1 mM DTT). In strand transfer experiments, 2 pmol of acceptor RNA were denatured 2 min at 90°C, ice-cooled and re-natured at 37°C for 15 min in 0.1 M NaCl, 60 mM Tris–HCl pH 8.0, 6 mM MgCl. Acceptor RNA was then added to proteins and incubated 4 min at 37°C in buffer E2. Reverse transcription was initiated by addition of pre-incubated primer/template complexes and proceeded for 5–60 min at 37°C. Polymerization was blocked by addition of 10 mM EDTA, 1% SDS and 2 mg/ml proteinase K. Samples were treated as described above and analyzed by 6% PAGE.
To distinguish genuine strand transfer products from self-priming products, bands containing nucleic acids longer than ssDNA were cut out of the dried gels, eluted at room temperature in 0.5 M ammonium acetate, 0.1 mM EDTA and 0.1% SDS and PCR-amplified with sense (corresponding to nucleotides 9038–9055) and antisense (ODN) primers. Percentage of full-length transfer product (FL) was calculated according to:
Δ viruses exhibit defects in the early steps of reverse transcription (,,,) in cells expressing APOBEC3G/3F, and APOBEC3G has been recently shown to specifically inhibit tRNA-primed DNA synthesis, possibly by inhibiting tRNA annealing to the PBS, which occurs concomitantly to or immediately after budding (). Gag precursor, mature NCp7, or maturation intermediates are thought to be the main players of this step thanks to their RNA chaperone activity (). However, Vif is present in large amounts in the assembly complexes (Vif/Gag ratio ∼0.5) () and in stoichiometric amount with RT in the viral particles (∼100 molecules) () and could thus affect tRNA annealing. In favor of this hypothesis, Vif binds the 5′-end region of HIV-1 genomic RNA with high affinity ( ∼45 nM) and recognizes many secondary structures in the 5′-UTR, including the PBS region ().
To test the influence of Vif on tRNA placement onto the PBS, we performed annealing experiments using purified recombinant Vif, post-transcriptionally modified tRNA and a RNA template spanning the first 311 nt of HIV-1 genomic RNA. Hybridization was monitored by native gel electrophoresis, and to ascertain specific and functional placement of tRNA to the PBS, hybrids were tested for their ability to initiate reverse transcription.
In keeping with previous studies, annealing of tRNA at 70°C in the absence of proteins proceeded with >80% efficiency (A, lanes P and PT) (,). Remarkably, we observed that Vif was able to significantly promote tRNA annealing to the PBS at 37°C, in a dose-dependent manner. While no hybrid was observed at 37°C in the absence of Vif, ∼10% of tRNA was hybridized in the presence of 0.5 µM Vif, and a maximum of 30% of tRNA was annealed at 5 µM Vif (A and C).
To check the functionality of the Vif-annealed tRNA/RNA 1–311 complexes, we monitored reverse transcription of the viral RNA from the annealed tRNA by HIV-1 RT, after removal of Vif. Primer extension was performed using a mixture of dCTP, dTTP, dGTP and ddATP, which allowed addition of 6 nt to tRNA (B). We observed a good correlation between the percentage of hybrid and the amount of +6 extension products with both heat- and Vif-annealed complexes (B and C), indicating that RT can recognize these complexes and initiate DNA synthesis. However, for both types of complexes, the amount of +6 extension product was systematically 20% lower than the amount of hybrid, suggesting a systematic loss of tRNA/RNA 1–311 complexes during protein extraction and purification procedures. Taken together, these results clearly show that Vif is able to promote formation of a functional tRNA/RNA complex.
A recent study showed that Vif stimulates HIV-1 RT activity by enhancing both the polymerization rate and RT binding to nucleic acids (). Vif also stimulates DNA synthesis through abasic sites, a property that could play a role in counteracting APOBEC3G-mediated deamination of proviral DNA (,,). These results suggest that Vif is a co-factor of HIV-1 RT, although these experiments were performed using an artificial primer–template complex (). We previously showed that when using a RNA template corresponding to the 5′-end of the HIV-1 genomic RNA and tRNA as primer, (–) strand ssDNA synthesis proceeds through two distinct steps. Initiation corresponds to the addition of the first 6 nt to tRNA: during this step DNA synthesis is distributive, while the subsequent elongation is processive (,). Conversely, priming DNA synthesis with an 18-mer DNA complementary to the PBS (ODN) allows DNA synthesis to start in the elongation mode ().
To test the effect of Vif on the initiation of reverse transcription, heat-annealed tRNA/RNA 1–311 complex was incubated with increasing concentrations of Vif and reverse transcription was initiated by adding RT and a mixture of dCTP, dTTP, dGTP and ddATP that allowed to complete the initiation phase of reverse transcription (see Experimental Procedures section) (). Note that contrary to the +6 extension assay described above, Vif was present during DNA synthesis. Whereas increasing concentrations of Vif had no effect on the stability of heat-annealed hybrid (B), we observed a dose-dependent decrease in +6 extension product correlated with an increase in unextended tRNA primer (A). We did not observe any reverse transcription product between these two forms, indicating that inhibition of reverse transcription takes place before the addition of the first deoxynucleotide. These data suggest that Vif decreased initiation of reverse transcription either by inhibiting the addition of the incoming nucleotide or by preventing binding of RT to the tRNA/RNA 1–311:RT complex.
In order to test the effect of Vif on the elongation phase of reverse transcription, we primed ssDNA synthesis with an ODN annealed to the PBS of RNA 1–311. Whereas a low (0.2 µM) concentration of Vif had no significant effect on ssDNA synthesis, a 2 µM concentration of Vif increased ssDNA synthesis by ∼2-fold after a 60-min reaction time (). Analysis of the reverse transcription products by denaturing PAGE showed that the increase in DNA synthesis correlated with a decrease in RT pausing (A, asterisks), suggesting that Vif-enhanced ssDNA synthesis by facilitating reverse transcription through stable secondary structures present in the RNA template. Indeed, secondary structures slow down RT, inducing dissociation of the enzyme and subsequently pauses (,). Thus, our data suggest that stimulation of ssDNA synthesis results from Vif–RNA interactions.
Mature NCp7 is a well-characterized co-factor of HIV-1 RT (), and at least some of the Vif-binding sites in the 5′ region of viral RNA are also NCp-binding sites (). While the effects of NC on tRNA annealing and ssDNA synthesis have been well documented (), the combined effects of Vif and NC on these steps have never been studied. Furthermore, as Vif has been shown to modulate processing of the Gag precursor (), we also studied the effects of NCp9 and NCp15, together with Vif, on these steps. Primer tRNA annealing and extension experiments were performed as described above, except that NC proteins were present at a concentration corresponding to complete coverage of the RNA template (1 NCp/5 nt) () (). In the absence of Vif, this saturating concentration of NCp allowed very efficient hybridization of tRNA to the PBS, with a maximum of 90% hybrids with NCp9 and NCp15 and a slightly reduced efficiency with NCp7 (70% hybrids) (B, yellow bars), in agreement with previous studies (), Unexpectedly, increasing Vif concentrations had an inhibitory effect on the formation of tRNA/vRNA hybrids induced by NCp (A and B, yellow bars), with a corresponding decrease in +6 product formation (B, blue bars).
Moreover, while the tRNA/vRNA complex formed by heating or in the presence of Vif migrated as a single band, two tRNA/vRNA complexes could be observed in the presence of NC proteins, with the main one migrating slower than the complex formed in the presence of Vif (A). These two bands correspond to tRNA annealed to monomeric and dimeric forms of HIV-1 RNA. Indeed, when using a fragment of HIV-1 genomic RNA with point mutations (AAA) in the dimerization initiation site (DIS), which is crucial for RNA dimerization (,), the monomeric form of the hybrid became largely predominant, even in the presence of NCp7 (A, right panel). Interestingly, increasing Vif concentrations progressively inhibited NCp7-induced RNA dimerization, and inhibition was almost complete at the highest Vif concentrations (A, left panel). Vif had similar effects when tRNA annealing and HIV-1 RNA dimerization were induced by NCp15, i.e. Vif efficiently inhibited both processes (B right panel). However, tRNA annealing and especially HIV-1 RNA dimerization promoted by NCp9 were more resistant to inhibition by Vif (B, central panel).
These results prompted us to analyze more precisely the capacity of Vif to modulate HIV-1 RNA dimerization in the absence of tRNA and in absence or in presence of NCp7 (). Indeed, HIV-1 RNA can form two different kinds of dimers , termed loose and tight dimers (). Loose dimers are formed first and correspond to kissing complexes interacting by the DIS loop (,,). Formation of tight dimers depends on the presence of the sequences 3′ to the major splice donor site, the incubation temperature and the presence of NCp. Although several authors have proposed that tight dimers could correspond to extended duplexes, this has not been demonstrated conclusively on large RNA fragments, and alternatively the kissing complexes could be stabilized by tertiary interactions that remain to be identified (). Using an RNA fragment encompassing the first 615 nt of HIV-1 genomic RNA, we analyzed dimerization using two different electrophoresis conditions: (i) native electrophoresis in Tris–Borate magnesium (TBM) buffer at 4°C, under which both loose and tight dimers are stable (A) and (ii) semi-denaturing electrophoresis in Tris–Borate-EDTA (TBE) buffer at room temperature, under which only the tight dimers survive (B) (). In the absence of NCp, Vif stimulated RNA dimerization in a concentration dependant manner, and a dimerization yield of 30% was observed at 5 µM Vif (A). This Vif–induced RNA dimer is a loose dimer, as it dissociated during electrophoresis under semi-denaturing conditions (B) (,,). In the absence of Vif, a saturating NCp concentration induced >90% RNA dimerization, and the NCp-induced dimer was predominantly the tight dimer (compare A and B), in keeping with previous studies (,). However, increasing Vif concentration progressively decreased the dimerization yield from >90% to ∼50% (A), and the RNA dimer remaining at the highest Vif concentration was exclusively the loose dimer (compare A and B). Thus, Vif was able to promote formation of the loose dimer, and to inhibit formation of the tight dimer by NCp, indicating that, as in the tRNA annealing experiments, the effect of Vif was dominant over NCp. Taken together, these results suggest that the dominant effect of Vif over NC proteins is due to Vif/NC interactions, rather than to Vif–RNA interactions.
Next, we compared the efficiency of ssDNA synthesis in the presence of NCp15, NCp9 and NCp7 (). At a concentration of 1 NCp/5 nt, a ∼4.5-fold increase of ssDNA was observed with NCp9 after 60 min of reaction, which correlated with a strong decrease in the intensity of RT pauses (, middle panel, black triangles). Comparatively, NCp7 and NCp15 had a smaller stimulatory effect (∼2.5 and ∼3-fold, respectively) and less influence on RT pausing ().
Given the stimulatory effect of Vif on ssDNA synthesis (B), we tested whether Vif could act synergistically with NC to increase the yield of ssDNA. Compared with DNA synthesis performed with NC or Vif alone, reactions performed in the presence of both Vif and NC showed no significant synergy. Addition of 2 µM Vif to reactions containing saturating amounts of NC significantly delayed ssDNA synthesis, and similar yields of ssDNA were observed only at the last time point (60 min) (). However, RT pauses specifically induced by either NCp7 or NCp15 (, left and right panels, asterisks) were strongly diminished in the presence of 2 µM Vif. Note that NCp7 and NCp15 generated different pausing patterns, suggesting that these two proteins preferentially bind to different RNA motifs in the R region, but that Vif had similar effects in these two reactions. These results suggest that both NC and Vif contributed to increase the processivity of RT and that NC–RNA interactions are affected by Vif binding to RNA.
The presence of repeated (R) sequences at both ends of retroviral genomes allows transfer of the neo-synthesized ssDNA from the 5′ to the 3′-terminal region of the viral RNA (). It has been shown that the first strand transfer is strongly enhanced by NCp9 and NCp7 (,,), and interactions between NC and the TAR loop of the viral RNA or its complement cTAR on the ssDNA play an important role in this process (,). As we recently showed that the TAR apical loop is a high affinity Vif-binding site () (Bernacchi ., in press for publication), we tested the influence of Vif on the first strand transfer. For that purpose RNA 1–311 was used as the donor template, and RNA 8607–9229 as the acceptor RNA, as previously described ().
During ssDNA synthesis, as the RNA template is being degraded by RNase H, the cTAR sequence in ssDNA can fold into a hairpin structure that is able to ‘self-prime’ reverse transcription, yielding abortive products that are longer than ssDNA. First, we used an RNase H(–) HIV-1 RT that is unable to give rise to strand transfer and self-priming products to unambiguously identify ssDNA (A, first lane). Longer products were observed with wild-type RT in the presence of acceptor RNA but in the absence of Vif or NC proteins. These were self-priming products, since PCR amplification of the corresponding gel-eluted nucleic acid bands did not yield any products when using primers specific to strand-transfer products (see Experimental Procedures section, data not shown). On the other hand, strand transfer reactions performed with wild-type RT and increasing Vif concentrations yield a faint but reproducible product near the top of the gel, identified as full-length strand transfer product by PCR (A). Quantification showed that ∼5% of the ssDNA was converted into full-length transfer product in the presence of 2 µM Vif. Thus, Vif can stimulate the strand transfer reaction, albeit with a limited efficiency.
In the absence of Vif and in the presence of a saturating concentration of NC proteins, strand transfer proceeded with varying efficiency, depending on the protein: the highest efficiency was obtained with NCp7, giving rise to ∼25% of full-length transfer product, while NCp15 and NCp9 were less potent (∼14% and ∼8% of full-length product, respectively) (). This result clearly shows that mature NCp7 promotes strand transfer more efficiently than NC maturation intermediates, probably due to the different nucleic acids chaperone and aggregating activities of these proteins (). Adding increasing concentrations of Vif to a saturating concentration of NC led to a dose-dependent modulation of NCp7 and NCp15 activity, while no effect was observed on NCp9. Although Vif moderately decreased the amount of full-length transfer product in NCp7-mediated transfer reaction, Vif showed an opposite effect in presence of NCp15, where addition of 2 µM of Vif reproducibly increased the yield of full-length transfer product by ∼40%. Taken together, these results show that Vif alone slightly promote the first strand transfer reaction and that this protein is able to specifically modulate the activity of NCp7 and NCp15 in this process.
Several reports demonstrated that deletion of the gene affected reverse transcription during the entry phase of the viral life cycle (,,,) and prevented endogenous reverse transcription (,,), suggesting that Vif could interact with virion components involved in the regulation of reverse transcription such as RT (), NC (,), tRNA (,) and genomic RNA (,). However, the molecular mechanisms of Vif function in reverse transcription have remained unclear. Since we recently showed that Vif is an RNA-binding protein that preferentially binds to the 5′ terminal region of HIV-1 genomic RNA, including the PBS, we decided to examine for the first time the contribution of Vif to the initial steps of reverse transcription, either alone or in combination with NC proteins at different maturation stages. In addition, since Vif is present in the HIV-1 assembly complexes (), we studied the effect of Vif on RNA dimerization, as this step is a prerequisite to efficient HIV-1 RNA packaging (). In order to analyze the intrinsic properties of Vif, we first studied this protein in the absence of other viral proteins. Then, to evaluate the potential biological significance of our results, we studied the effects of Vif in the presence of NC proteins.
In the absence of NC proteins, Vif enhances several early steps of the reverse transcription process. First, Vif is able to anneal tRNA to a PBS-containing RNA fragment quite efficiently (A), even though Vif is less efficient than NCp7, NCp9 and NCp15 in this respect (B). In addition, the resulting primer–template complex is fully functional, since HIV-1 RT was able to initiate reverse transcription of these complexes (B). The annealing activity of Vif might be important for HIV-1 replication in non-permissive cells, since it was recently shown that inhibition of tRNA-primed reverse transcription of Δ viruses by APOBEC3G and 3F accounts for an important part of its antiviral effect, independently of its deaminase activity (,).
Second, Vif significantly decreases pausing of RT and enhances ssDNA synthesis (). RT pausing most often occurs when RT is blocked by stable secondary structures present in the RNA template (,). Indeed, we previously showed that Vif binds to several secondary structure motifs in the R and U5′ regions of the genomic RNA (), and our present results suggest that Vif is able to destabilize these structures, allowing a better processivity of HIV-1 RT. The effect of Vif on RT pausing might also be the result of an affinity increase of RT for the primer–template complex in the presence of Vif, as observed by Cancio . () using poly(rA)/oligo(dT). Even though the exact mechanism by which Vif increases RT processivity remains to be established, this activity is analogous to the previously described effect of NCp7 on RT pausing (,).
We also reproducibly observed that Vif stimulates the first strand transfer, albeit with a reduced efficiency compared to NC proteins: Vif was 2-fold less efficient than NCp9, 3-fold less efficient than NCp15 and 5-fold less efficient than NCp7 (). Nevertheless, Vif shares with NC proteins the ability to anneal tRNA to the PBS, decrease RT pausing, and promote strand transfer. These properties, especially the first and the third one, are characteristic of the RNA chaperone activity of NC proteins (). In addition, we found that Vif, like NCp (), promotes dimerization of HIV-1 RNA, even though the former induces formation of loose dimers, while the latter favors tight dimers. Therefore, our results show that Vif is an authentic RNA chaperone ().
These results prompted us to look for potential synergic effects between Vif and various Gag processing products such as NCp15, NCp9 and NCp7. Processing of the Gag precursor takes place in a sequential manner during maturation of HIV-1 viral particles due to differences in cleavage site efficiency (). Cleavage starts at the p2/NC junction, resulting in the production of MA-CA-p2 and NCp15 intermediates products. Secondary cleavage releases the mature MA protein as well as the NCp9 (NCp7-p1) intermediate after removal of p6 (). Tertiary cleavage produces mature CA and NCp7 proteins. Mature viral particles are not only composed of fully processed Gag and Gag-Pol products, but also contain residual amounts of Gag precursors and processing intermediates (). Moreover, Vif modulates the HIV-1 protease activity, leading to the possibility that virion incorporation of Vif could stabilize NC intermediates (). A number of molecular rearrangements occur in the RNP during or just after budding, including tRNA hybridization (), RNA dimer maturation () and initiation of reverse transcription ().
In the tRNA annealing reactions, Vif clearly has a dominant effect over NC proteins (). Indeed, increasing concentrations of Vif significantly inhibited NC-induced tRNA annealing (B) and RNA dimerization (A and ). This inhibitory effect of NC-mediated functions might explain why excessive Vif expression is detrimental to viral infectivity (). Importantly, the highest Vif concentration used in this study corresponds to a Vif/NCp ratio of 1/3, which is close to the 1/2 Vif/Gag ratio in the assembly complexes (,). Thus, Vif might initiate tRNA annealing in the assembly complexes, and since it is mostly excluded from the virions (the Vif/NCp ratio in virions is between 1/20 and 1/40), its inhibitory effect would be relieved after assembly of the virions, and NC proteins could complete tRNA annealing. In this context, inhibition of the initiation of reverse transcription by Vif () might prevent premature initiation of DNA synthesis in the assembly complexes of the producing cell. Interestingly, identical effects have been observed with the Tat protein (,). Similarly, inhibition of the NC-induced viral RNA dimerization by Vif () might temporally regulate RNA packaging and prevent premature maturation of the loose kissing loop complex into a more thermostable tight dimer. Thus, our results suggest that Vif is a negative regulator of several NCp-associated functions in the assembly complexes, and that the NC domain (either in Gag or its maturation product) becomes fully active in the viral particles, from which Vif is mostly excluded. Inhibition of the NCp-mediated tRNA annealing and RNA dimerization, as well as inhibition of the initiation of reverse transcription by Vif might allow temporal fine-tuning of these steps.
During ssDNA synthesis, both Vif and NC proteins decreases RT pausing. However, NC proteins also appear to induce pausing at sites that were specific for each NC species (). These pauses could result from increased RNase H activity of RT in the presence of NC proteins (). No additional pauses were observed with Vif; on the contrary Vif suppressed the NC-induced pauses. Thus, Vif act in cooperation with NC proteins to increase RT processivity and favor synthesis of long DNA products.
The effect of Vif on NC-induced strand transfer is limited, but, unexpectedly, Vif has either a positive, null or negative effect depending on the maturation stage of NC proteins (). Significantly, NCp9 is the most efficient NC species in ssDNA synthesis, whereas NCp7 has the most efficient strand transfer activity, suggesting that maturation of the NC intermediates might regulate the early steps of HIV-1 reverse transcription.
Thus far, it was unclear whether Vif has a direct role in reverse transcription. Our data strongly suggest that Vif does play a role during the early phase of this process in coordination with other components of the viral core such as Gag and its maturation products and during the dimerization of genomic RNA. Taken together, our data lead to the possibility that Vif might be a temporal regulator during viral assembly: (i) by interacting with genomic RNA and NC-derived products, Vif may prevent RNA dimerization/packaging and premature initiation of reverse transcription; (ii) however, still in conjunction with Gag precursors, Vif could promote the placement of tRNA on to the PBS, stabilizing NC intermediates to increase, at the right time, the efficiency of the early steps of reverse transcription. Obviously, a detailed temporal analysis of the effects of Vif on HIV-1 replication will be required to test our hypothesis. |
Nucleases play critical roles in DNA transactions including DNA replication, repair and recombination events (). Detailed analysis of the enzymatic properties of a newly discovered nuclease activity is required in order to help identify cellular function. Nuclease digestion patterns are generally studied using radiolabeled oligonucleotides as substrate followed by electrophoretic analysis of the reaction mixture. However, this method only detects fragments containing the radiolabel. Multiple labeling of the substrate DNA is required in order to ascertain the location of all the cleaved sites. Furthermore, electrophoresis does not reveal the precise nature of the 5′ and 3′-termini at the cleaved site. Alternatively, mass spectrometry can detect all fragments produced by the nuclease without the need for using radioisotopes. Electrospray ionization (ESI) is a soft ionization technique which is suitable for large biopolymers, such as oligonucleotides, because the multiply charged ions lower the / (). In addition, Fourier-transform ion cyclotron resonance mass spectrometry (FT ICR MS) is expected to achieve especially high mass accuracy in the analysis of mixtures of biopolymers and determination of the nature. In this study, we analyzed the degradation pattern of double-stranded DNA (dsDNA) by the catalytic domain of MutS2 (ttMutS2) using ESI–FT ICR MS. To the best of our knowledge, this is the first report that a nuclease activity was characterized by ESI–FT ICR MS.
Bacterial MutS2 possesses domains homologous to the MutS family proteins () which are involved in DNA mismatch repair, DNA recombination and other DNA modifications. Recent studies showed that bacterial MutS2 is involved in suppression of homologous recombination () or/and protection from oxidative DNA damage (). We previously revealed that ttMutS2 contains a nuclease activity (,), and the activity is confined to the C-terminal domain whose sequence is not conserved in the other MutS homologs. The amino acid sequences homologous to the C-terminal domain of bacterial MutS2 distribute in variety of proteins other than MutS homolog (,). It is also reported that the C-terminal domain of human BCL-3-binding protein shows sequence similarity to that of bacterial MutS2 and possesses a nuclease activity (), although there is no relationship between the biological function of human BCL-3-binding protein and that of bacterial MutS2. The precise characteristics of their activities had been unknown.
The 5′-hydroxylated 3, 7, 15, 21 and 37-mer single-stranded oligodeoxynucleotides were synthesized. Their sequences are: 5′-GCT-3′, 5′-GCTCGTA-3′, 5′-GCTCGTAGAGTGGTC-3′, 5′-CGGTATCTTGGCTATGACGGC-3′, 5′-GCTCGTAGAGCGGTCATAGTCAAGATACCG-3′ and 5′-ATGTGAATCAGTATGGTTACTATCTGCTGAAGGAAAT-3′, respectively. They were purified after synthesis by high-performance liquid chromatography (HPLC) and their concentrations were determined from their absorbance at 260 nm. The 5′-phosphorylated or 5′-hydroxylated single-stranded oligodeoxynucleotides, 5′-ATGTGACTCAGTATGGG-3′, were also synthesized and purified by HPLC. Then, they were annealed to their complementary 5′- phosphorylated or 5′-hydroxylated single-stranded oligodeoxynucleotides (5′-CCCATACTGAGTCACAT-3′) to obtain double-stranded oligodeoxynucleotides in TE buffer. Annealing was performed in a thermal cycler according to the following temperature profile: 5 min at 95°C, followed by a slow decrease from 95°C to 37°C over 60 min and from 37 to 4°C over 30 min. Oligonucleotide which contains the locked nucleic acid (LNA) at 3′-terminal end was also synthesized.
The C-terminal domain of ttMutS2 (CTD) was overexpressed and purified as described previously (). The substrate oligonucleoitdes (25 μM) were incubated with 0, 50, 100 or 200 nM CTD in a buffer containing 50 mM Tris–HCl (pH 7.5), 100 mM KCl and 5 mM MgCl at 37°C for 16 h. The total volume of reaction mixture was 30 μl. Reactions were quenched by addition of an equal volume of phenol–chloroform solution and the mixtures were then centrifuged at 15 000 r.p.m for 10 min. Supernatants were loaded onto 5 μl of SuperQ resin equilibrated with water in a 0.6 ml Eppendorf tube. After washing with 30 μl of water three times, DNAs were eluted with 30 μl of 0.75 M ammonium acetate (pH 7). The elutants were dried in a centrifugal evaporator.
The elutants were resuspended to a concentration of 1 μM in 50% methanol containing 25 mM imidazole and 25 mM piperidine. The 10 μl aliquots were loaded into quartz nanospray emitters. All measurements were performed on an Apex IV Fourier transform mass spectrometer (Bruker Daltonics, Billerica, MA, USA) equipped with a 9.4-T shielded superconductive magnet. The oligonucleotide solutions were infused into an external Apollo electrospray ion source at a flow rate of 13 μl/min with the assistance of N nebulizing gas. The off-axis sprayer was grounded, and the inlet capillary was set to 1.7 kV for generation of oligonucleotide anions. N drying gas was applied to assist desolvation of ESI droplets. Ions were accumulated in the hexapole for 0.2 s. All data were acquired in negative ion mode and processed using XMASS 6.0.1 (Bruker Daltonics).
Single-stranded DNAs were radiolabeled at the 5′-end with [γ-P]ATP using polynucleotide kinase before annealing. The 5′-labeled duplexes (10 nM) were incubated with CTD or king of DNA (KOD) polymerase (TOYOBO, Osaka, Japan) in a 50 mM Tris–HCl (pH 7.5) containing 100 mM KCl, 5 mM MgCl and 25 μM non-labeled substrate dsDNA for 16 h at 37°C. The enzyme concentrations are indicated in the legends to figures. Reactions were stopped with the addition of equal volume of phenol–chloroform solution and the solutions were centrifuged at 15 000 r.p.m. for 10 min. The supernatants were mixed with the sample buffer (5 mM EDTA, 80% deionized formamide, 10 mM NaOH, 0.1% bromophenol blue and 0.1% xylene cyanol) and heat-treated at 95°C for 5 min. They were loaded onto 11% polyacrylamide sequencing gels (8 M urea and ×1 TBE buffer, 89 mM Trisborate and 2 mM EDTA) and electrophoresed with × 1 TBE buffer. The gels were dried and placed in contact with an imaging plate. The bands were visualized and analyzed using a BAS2500 image analyzer (Fuji Film, Tokyo, Japan).
The substrate P-labeled DNA was base-specifically modified and digested according to the Maxam–Gilbert method (). In order to determine the length of the product DNA, these fragments were mixed with the sample buffer and electrophoresed along with the products on DNA sequencing gels.
First, we examined the validity of our method for the purification of DNA samples. Contamination of non-volatile salts such as sodium or potassium, which would otherwise prevent ionization of the biomolecules and cause the formation of series of metal adduct ions (), must be avoided. The single-stranded DNAs (ssDNAs) (3–30-mer) were prepared and analyzed as described in MATERIALS AND METHODS section. Raw mass spectra consisted of a series of peaks, corresponding to multiply charged ions of intact ssDNA having a specific number of protons removed from the phosphodiester groups (A). As shown in inset, few metal adduct ion species were observed, showing the efficacy of the purification method. When the same sample was purified by ethanol precipitation, even the charged ions of ssDNA were hardly detected (data not shown) probably because non-volatile salts were poorly removed. In B, signals representing the same DNA were deconvoluted to yield the molecular mass of the corresponding DNA. Comparison of the measured molecular mass against theoretical molecular mass enabled the length and sequence of the DNA species to be identified. A deviation between the measured and theoretical molecular mass of within ±2 p.p.m. was deemed acceptable. Our results show that the DNA desalting method can be applied to the analysis of extremely short oligonucleotides, at least to a 3-mer.
The 5′-phosphorylated 17-bp dsDNA incubated with CTD were analyzed as described above. We chose the oligonucleotide sequence shown in MATERIALS AND METHODS section as a substrate since it has left–right asymmetric nucleotide content. Symmetric distribution of the nucleotide content will cause difficulty in identification of the fragments since various potential structures correspond to a single mass. shows the deconvoluted mass spectra of reaction products. The substrate and product dsDNAs are completely denatured and detected as ssDNAs. Indeed, any remaining dsDNAs would have complicated the interpretation of the results. An increased number of peaks corresponding to DNA reaction products were detected as the enzyme concentration increased. Several peaks are more abundant than others. However, we cannot quantitatively interpret this in terms of selectivity because the nucleotide content and length seems to have an effect on the ionization efficiency of oligonucleotides (data not shown). Over 90% of these peaks were accurately identified as 5–12-mer fragments. The major fragments identified by mass spectrometric analysis are listed in . This result strongly suggests that the nuclease activity of CTD has no obvious sequence-specificity. All of mass spectrometric analyses were performed in triplicate.
There were a few unidentified fragments that corresponded in mass to two or more candidate fragments (A). Some of these peaks were subsequently identified by comparison with the result of a digestion of 5′-hydroxylated dsDNA (B). When the fragment contained an unreacted 5′-terminus, the corresponding peak would shift as far as the difference between masses of hydroxyl and phosphoryl groups. As shown in C, several peaks shifted and were identified.
All of the theoretical masses used to identify the peaks were calculated by assuming the 5′- and 3′-termini of cleaved sites were phosphorylated and hydroxylated, respectively. These results indicate that CTD incised the phosphate backbone of oligodeoxynucleotides at the 3′ side of the phosphates and the nicks generated by CTD could be ligated by DNA ligase. Thus, ESI–FT ICR MS can identify the nature of the cleaved sites with considerable accuracy.
The nuclease activity for 17-bp dsDNA was also examined by electrophoretic analysis using a sequencing gel. As shown in , the fragments were estimated to be 5–12-mers by comparison to the DNA size marker made by the Maxam–Gilbert method (). These results are entirely consistent with the analysis by mass spectrometry. Although a few fragments shorter or longer than 5- or 12-mer were observed when the enzyme concentration was increased (data not shown), main products were always 5–12-mer fragments. This result indicates that a certain length of dsDNA is required for the formation of a stable CTD–substrate complex.
All of the detected fragments retained an unreacted 5′- or 3′-terminus, indicating that CTD possesses a non-sequence-specific endonuclease activity rather than exonuclease activity. Gel electrophoretic analysis also ruled out the possibility of 5′ to 3′ exonuclease activity of CTD because there was no observable accumulation of short fragments during the incision of 5′-labeled substrates (). We then tested the oligonucleotide analog where the nucleotide at the 3′-terminal end was replaced with LNA, which contains an extra 2′-, 4′--methylene bridge on the ribose ring. It had been reported that this substrate shows a tolerance to exonuclease activity (,). Although the 3′–5′ exonuclease activity of KOD polymerase was affected by the replacement of the 3′-terminal end, the replacement did not affect the nuclease activity of CTD (). Taken together, electrophoretic analysis also demonstrated that CTD does not possess an exonuclease activity. Thus, we have confirmed the validity of the ESI–FT ICR mass spectrometric analysis.
The results in this study showed that ESI–FT ICR MS can precisely analyze the digestion pattern of a non-sequence-specific endonuclease. In comparison, the analysis of a sequence- or structure-specific endonuclease activity should be relatively straightforward using this methodology. The information about substrate specificity and the nature of cleaved sites would be a great help in understanding the mechanism of an enzyme and its role in a biological process. For example, when a newly discovered nuclease preferably incised at an abasic site yielding 5′-deoxyribosephosphate (5′-dRP) end, we should consider the possibility that the nuclease takes part in base excision repair through the repair of abasic sites and the repair pathway requires an enzyme that have a 5′-dRPase activity such as DNA polymerase β. |
DNA topoisomerase II (topo II) is an essential and ubiquitous enzyme for proliferation of eukaryotic cells (). It can alter the topological state of DNA and untangle DNA knots and catenanes (interlocked rings) via ATP-dependent passing of an intact double helix through a transient double-stranded break generated in a separate DNA segment, followed by religation and enzyme turnover (). In mammalian cells, topo II exists in two isoforms, α (170 kDa) and β (180 kDa), both having similar primary structure and almost identical catalytic properties, but differing in their production during the cell cycle (,).
Topo II is the primary target of a number of active agents currently used in the treatment of human cancers, such as epipodophyllotoxins (etoposide and teniposide), anthracyclines (doxorubicin and daunorubicin) and mitoxantrone (). These drugs (also termed topo II poisons) can stabilize the covalent enzyme-associated complexes and shift the DNA cleavage/religation equilibrium of the enzyme reaction toward the cleavage state, converting biological intermediates of topo II activity into lethal ones ultimately leading to triggering of programmed cell death pathways (,,).
HMGB1 is an abundant, ubiquitous and evolutionarily highly conserved non-histone chromatin-associated protein in mammals, which functions in a number of fundamental cellular processes such as transcription, replication, DNA repair and recombination (). HMGB1 is associated with chromosomes in mitosis and due to its extreme mobility in the cell, the protein is continuously exchanged between nucleus and cytoplasm (5 and references therein). HMGB1 also exhibits an important extracellular function in mediation of inflammation mechanisms, tumor growth and metastasis (,). HMGB1 binds relatively weakly to B-form DNA, but displays a high affinity for distorted DNA conformations [e.g. four-way DNA junction, DNA minicircles, hemicatenated DNA loops and cisplatin-modified DNA; ()]. Binding of HMGB1 to DNA causes local distortions by bending/looping or changes of DNA topology (,,). HMGB1 also interacts weakly with a number of proteins, including transcription factors, site-specific recombination and DNA repair proteins (). The importance of HMGB1 for life is supported by the phenotype of the HMGB1 knockout mice, which die 24 h after birth due to hypoglycemia and exhibit a defect in the transcriptional function of the glucocorticoid receptor ().
In the present study, we report a physical interaction between HMGB1 and human topo IIα . We show that HMGB1 promotes topo IIα-mediated catenation of circular DNA, enhances relaxation of negatively supercoiled DNA and decatenation of kinetoplast DNA. Stimulation of the catalytic activity of topo IIα by HMGB1 is mainly due to enhanced DNA binding and cleavage by the enzyme. Possible functioning of HMGB1 as a modulator of the cellular activity of topo IIα is discussed.
Experiments were carried out either with human topoisomerase IIα (topo IIα) purchased from Topogen, or with recombinant human topo IIα isolated from yeast strain JEL1Δtop1 (the strain was kindly provided by John L. Nitiss, Department of Molecular Pharmacology, St. Jude Children's Research Hospital, Memphis, USA) harboring plasmid YEpWOBalphaHT (kindly provided by Anni H. Andersen, Department of Molecular Biology, University of Aarhus, Aarhus, Denmark) as detailed in references (). topoisomerase IV (10 U/μl) and wheat germ topoisomerase I (2–10 U/μl) were purchased from Topogen and Promega, respectively. Kinetoplast DNA (kDNA) was isolated from (the cells were kindly provided by Paul T. Englund, Haskins Laboratories and Biology Department, Pace University, New York, USA) as detailed in (). kDNA was also kindly provided by Julius Lukeš (Institute of Parasitology, České Budějovice, Czech Republic). Antibodies against the following proteins were used: anti-HMGB1 (affinity purified rabbit polyclonal, BD Pharmingen), and anti-topo IIα (rabbit polyclonal, Topogen).
DNA plasmids were isolated by alkaline lysis method, followed by purification by two rounds of cesium chloride gradient or by the Qiagen plasmid kits. All purified plasmids exhibited ratios A/A higher than 1.85. In some cases (catenation assays), supercoiled plasmids pTZ19R or pBR322 were relaxed by wheat germ topoisomerase I, followed by deproteinization of the relaxed plasmid as detailed earlier ().
HMGB1 (residues 1–215), HMGB1 domain A (residues 1–88), HMGB1 domain B (residues 85–180) and HMGB1 di-domain A+B (residues 1–180) were derived from rat HMGB1 cDNA (the amino acid sequence of the rat HMGB1 protein is identical to that of the human HMGB1 protein). Alanine mutagenesis of intercalating residues Phe38 (domain A), Phe103 and I122 (domain B) of the individual HMGB1 domains or the full-length HMGB1 was carried out using PCR-based protocol generating ‘chimeric proteins’ (Štros, unpublished data). The introduced mutations were verified by dideoxi-sequencing of both strands. The DNA sequences coding for the HMGB1 and truncated forms were inserted into HI and I sites of the vector pQE-80L (Qiagen) which allows tightly regulated N-terminal 6xHis-tagged protein expression in . The N-terminally GST (Glutatione S-Transferase) tagged HMGB1 protein and its truncated forms were also synthetized from the pGEM-4T1 vector (GE Healthcare). Purification of HMGB1 proteins by FPLC-chromatography, and SDS–18% polyacrylamide gel electrophoresis was performed as described earlier (,,).
DNA supercoiling assay was carried out as previously described () with the following modifications. Negatively supercoiled plasmid pBR322 DNA (0.5 μg or ∼9 nM) was relaxed in the relaxation buffer (40 mM NaCl, 50 mM Tris-HCl pH 7.5, 1 mM EDTA, 20% glycerol, 1 mM DTT) by wheat germ topoisomerase I (2U, Promega) at 37°C for 90 min. Then a new portion of the enzyme was added and the relaxed DNA was subsequently mixed either with wild-type HMGB1 or mutated HMGB1 (0–6 μM) in a final volume of 20 μl. The reactions were allowed to proceed for 1 h at 37°C after which 5 μl of the termination mix (5× TBE, 5% SDS, 15% sucrose, 0.1% bromphenol blue, 0.1% xylene cyanol, 0.2 μg/μl proteinase K in 20 mM Tris-HCl pH 8.0,1 mM CaCl) was added and the samples were incubated at 45°C for 30 min, followed by phenol/chloroform extraction (). The DNA topoisomers were then resolved on a 1% agarose gel in 0.5× TBE buffer at ∼3 V/cm for 17 h, and the DNA samples were visualized by UV-illumination of the ethidium bromide-stained gel (0.5 μg/ml).
Supercoiled plasmid pTZ19R (∼15 nM) was linearized by dIII digestion, the deproteinized linear DNA was mixed with HMGB1 (2 or 6 μM), and pre-incubated on ice for 20 min. The DNA was then ligated with 0.2 U of T4 DNA ligase in a final volume of 20 μl at 30°C for 30 min in the presence or absence of 5% (w/v) polyethylene glycol (PEG 8000, Sigma). Deproteinized DNA samples were resolved on 1% agarose gels, followed by staining with ethidium bromide (0.5 μg/ml).
The following assays were used to assess topo IIα activity: catenation, decatenation, relaxation, cleavage and religation. Most experiments were carried out both with human topo IIα purchased from Topogen and with recombinant human topo IIα isolated from yeast (see ‘Materials and Methods section’). ‘Catenation assay’ was carried out with either negatively supercoiled or relaxed closed-circular plasmids. The plasmid pTZ19R (∼15 nM) was pre-incubated on ice in a total volume of 20 μl of topo IIα assay buffer (50 mM Tris-HC1 pH 8, 85 mM KCl, 10 mM MgCl, 1 mM ATP, 0.5 mM dithiothreitol, 30 μg/ml BSA) containing 0–10% PEG (w/v) with different amounts of HMGB1 (as indicated in the figure legends) for 20 min, followed by addition of topo IIα (typically 5 nM). The mixture was finally incubated at 37°C for 40–60 min. ‘Decatenation assay’ was performed in the topo IIα assay buffer by incubation of 0.2 μg of kinetoplast DNA (kDNA) with topo IIα (as indicated in the figure legends). No PEG was present in the decatenation assay buffer. ‘Relaxation assay’ was carried out in the topo IIα assay buffer by treatment of negatively supercoiled plasmid pBR322 (∼8 nM) with either a fixed amount of topo IIα and increasing amounts of HMGB1 or with increasing amounts of topo IIα and a fixed amount of HMGB1 as indicated in the figure legends. No PEG was present in the relaxation assay buffer. All other reaction conditions of the relaxation assay were identical to the catenation assay. The above topo IIα activity assays were terminated by rapid addition of 460 μl of 1 M NaCl/1% SDS/10 mM EDTA and 2 μl of 2.5% linear polyacrylamide as carrier, followed by vortexing with an equal volume of chloroform–isoamylalcohol (24:1, v/v). The mixtures were then centrifuged at room temperature at 10 000 g for 10 min, and the clarified upper layers were precipitated with 1 ml of absolute ethanol at −70°C for 20 min. The DNA precipitates were washed with 70% ethanol, air-dried and dissolved in 0.1× TE buffer. ‘DNA cleavage assay’ was carried out by incubation of plasmid pBR322 (∼4 nM) with topo IIα in DNA cleavage buffer (10 mM Tris-HCl pH 8, 50 mM KCl, 50 mM NaCl, 5 mM MgCl, 2.5% glycerol, 0.1 mM EDTA, containing either 1 mM ATP or 1 mM non-hydrolyzable analog ANP-PNP, Sigma) in a total volume of 20 μl at 37°C for 20 min. Some reactions also contained HMGB1 (1–3 μM) and/or etoposide (10–100 μM, Sigma) as indicated in the figure legends. The cleavage reactions were terminated by trapping the DNA cleavage complexes by addition of 2 μl of 10% SDS, followed by addition of 1.5 μl of 0.25 M EDTA and 2 μl of proteinase K (0.8 μg/μl in 50 mM Tris-HCl pH 8, and 1 mM CaCl) and incubation at 45°C for 60 min. ‘Religation assay’ was carried out in DNA cleavage buffer containing 30 ng BSA (bovine serum albumin)/μl in a DNA mixture with negatively supercoiled plasmid pBR322 (4 nM), 50 μM etoposide, and topo IIα (8 nM). In some cases, HMGB1 (1–3 μM) was pre-incubated with DNA for 20 min on ice before addition of the enzyme. The reactions were started by incubation at 37°C for 15 min, followed by shifting the incubation temperature to 65°C or 75°C (to promote only the topo IIα-mediated religation and the cleavage). Aliquots (20 μl) were then withdrawn at different times (typically 0–40 min), immediately mixed with 2.2 μl of 10% SDS, and finally digested with proteinase K (see ‘DNA cleavage assay’). The deproteinized DNA samples were finally resolved on 1% agarose gels containing 0.5× TBE. DNA was visualized by ethidium bromide staining which was either present in the gels prior to electrophoresis (DNA decatenation, cleavage and religation assays), or gels were stained after electrophoresis (DNA relaxation and catenation assays). DNA was quantified either using the ImageQuant TL software (GE Healthcare) or Multi Gauge software using imaging system LAS-3000 (Fuji).
Decatenation assay was carried out with 0.2 μg of kDNA in 40 mM HEPES-KOH (pH 7.6), 100 mM potassium glutamate, 10 mM magnesium acetate, 2 mM ATP, 10 mM DTT and 50 μg BSA/ml with topoisomerase IV (typically 0.1–1 U) in the presence or absence of HMGB1 (1 or 4 μM) in a final volume 20 μl at 37°C for 30 min, according to the manufacturer's instructions (Topogen). Analysis of decatenated products was performed as detailed for decatenation by human topo IIα.
The 36-mer oligonucleotides for EMSA containing a high-affinity topo IIα-binding site (,) were (5′ to 3′): oligo_1, ATGAAATCTAACAATGCGCTCATCGTCATCCTCGGC; oligo_2, GCCGAGGATGACGATGAGCGCATTGTTAGATTTCAT.
The oligo_1 was P-labeled at its 5′-terminus by T4 DNA kinase and [γ-P]ATP (specific activity 3000 Ci/mmol, GE Healthcare) and annealed with its corresponding complementary strand (oligo_2) to form a DNA duplex. Reaction mixtures for EMSA contained isolated topo IIα (75 nM) and increasing amounts of HMGB1 proteins (0.4–1.6 μM), ∼35 ng of P-labeled DNA duplex in EMSA buffer (20 mM HEPES, pH 7.6, 4% Ficoll, 0.02% Nonidet P-40, 1.5 mM spermidine, 0.1 mM EDTA, 0.125 M KCl and 0.5 mM DTT) in a total volume of 20 μl. In some EMSA assays, 0.2 μg of homopolymer poly (dI-dC) was used as a non-specific competitor DNA to reduce non-sequence specific binding of HMGB1. However, we have noticed that the presence of the competitor DNA significantly reduced the binding of topo IIα to the linear 36-bp DNA duplex as previously reported (). The proteins were pre-incubated with DNA on ice for 20 min (order of mixing had no effect on the outcome of the EMSA assays) and finally resolved on 5% polyacrylamide gels in 0.5× TBE buffer at 250 V at 4°C until the bromphenol blue reached the bottom of the gel. The gels were dried and the labeled DNA was imaged by Storm PhosphorImager (Molecular Probes).
ATP hydrolysis was studied in reactions containing 1 mM cold ATP plus 20 × 10 c.p.m. of [γ-P]ATP (specific activity 3000 Ci/mmol, GE Healthcare) and negatively supercoiled plasmid pBR322 (50 nM). Some reactions also contained HMGB1 (0.2–3 μM). ATP hydrolysis was initiated by addition of topo IIα (5 nM). Termination of ATP hydrolysis was accomplished by spotting the mixtures on TLC cellulose plates (20 × 20 cm; Baker-flex, Phillipsburg) at different times as previously described (). Products of ATP hydrolysis were separated by thin layer chromatography and the free phosphate was quantified by PhosphorImager.
Following incubation of the relaxed circular plasmid DNA with topo IIα and HMGB1, the DNA was deproteinized by chloroform:isoamylalcohol (24:1, v/v) extraction. RecA coating of DNA for EM was carried out by the protocols detailed in (). Briefly, for single-strand DNA coating with RecA (∼9.1 kb plasmid pAK-9.1), DNA was denatured at 63°C in the presence of glyoxal and purified on a Sephacryl S-500 minicolumn (0.4 × 4 cm) (GE Healthcare). Denatured DNA was then mixed with RecA protein at a molar ratio of 70:1 (∼2.8-kb plasmid pTZ19R) or 30:1 (∼9.1-kb plasmid pAK-9.1), cross-linked with 0.2% (v/v) glutaraldehyde, and purified on a Sephacryl S-500 minicolumn. Coating of double-stranded DNA (∼2.8 kb plasmid): following DNA coating with RecA, 0.5 mM ATP-γS was added to stabilize the complexes. DNA cross-linking with glutaraldehyde and purification of the RecA-coated dsDNA was carried out as for single-stranded DNA. RecA-coated DNA samples (15 μl) were adsorbed to a freshly glow-discharged () carbon-coated parlodion film on EM grids. Specimens were washed with 100 mM ammonium acetate, stained on drops of 5% (w/v) uranyl acetate and washed on the surface of 10 mM ammonium acetate (). The grids were both rotary and unidirectionally shadowed with platinum/palladium at an angle of ∼7° and observed with a JEOL JEM 1200EX electron microscope operating at 60 kV. The photographs were taken at 50 000 magnification.
GST alone or GST fused with full-length HMGB1 or its HMG-box domains at their N-termini were isolated as described (,). The purifed proteins (typically 1–2 μg) were bound to glutathione Sepharose 4B beads (40 μl of 1:1 v/v slurry), and incubated with purified human topo IIα (0.5 μg) in buffer PD [20 mM Tris-HCl pH 7.6, 0.2 M NaCl, 10 mM DTT, 0.2% Triton X-100, 0.2% Tween-20, 20% glycerol (v/v), protease inhibitors: 1 μg/ul aprotinin, 10 μg/ul leupeptin, 1 μg/ul pepstatin A, 100 μg/ul trypsin inhibitor, 0.1 mM TLCK and 20 mM benzamidine] in the presence or absence of 10 U of DNAse I for 15 min at 25°C, followed by rocking the samples for 2 h at 4°C. Subsequently, the beads were washed five times in 1 ml of 1× PD buffer. The proteins associated with Sepharose beads were finally eluted from the beads by addition of 20 μl of 10× SDS loading buffer and boiling the samples for 3–4 min. The beads-associated proteins were resolved on an SDS–7.5% polyacrylamide gel, transferred onto the PVDF membrane by western blotting and detected using polyclonal human topo IIα antibody (1:1000 dilution, Topogen), followed by incubation of the membranes with horseradish peroxidase-conjugated anti-rabbit antibodies (IgG-HRP) (1:2000 dilution, GE Healthcare) and ECL detection (GE Healthcare).
We have previously shown that chromatin architectural protein HMGB1 could stimulate T4 DNA ligase-mediated end-to-end joining of linear DNA molecules by promoting intramolecular association of DNA molecules via their ends (,). Ligation of diluted solutions of linearized plasmid DNA in the presence of HMGB1 resulted in preferential intramolecular ligation into closed-circular DNA (A, lanes 2–4). However, as both DNA and HMGB1 are present in eukaryotic nuclei in relatively high concentrations, the effect of the protein on DNA end-joining was re-investigated in the presence of a macromolecular crowding (volume excluding) agent, polyethyleneglycol (PEG), to mimic crowding . In the absence of HMGB1, PEG alone had only little effect on DNA ligation using limiting amounts of the enzyme (A, compare lanes 2 and 5). However, the simultaneous presence of HMGB1 and PEG suppressed intramolecular ligation and markedly promoted intermolecular DNA end-joining. Under these conditions, all monomeric DNA molecules were converted into linear DNA multimers, the majority of which migrating on agarose gels with the mobility of trimers or higher multimers (designated as L3 in A), and into a minor fraction of high molecular mass linear multimers (which were susceptible, like L3, to ExoIII digestion, data not shown) not entering the gel (A, compare lanes 5 and 7).
The ability of HMGB1 to promote intermolecular association of DNA prompted us to determine whether HMGB1 could also promote interlocking (catenation) of covalently closed DNA by human topoisomerase IIα (topo IIα). As shown in B (lanes 2 and 5), incubation of supercoiled DNA and HMGB1 with limiting amounts of topo IIα resulted in no formation of DNA multimers unless 5% PEG was included in the reaction buffer (B, lane 6). Most of the complex DNA multimers did not enter the agarose gel and remained at its origin. The apparent size of the complex DNA multimers was visibly decreased at 10% PEG, and a significant fraction of the DNA multimers was no longer trapped in the wells and migrated rather as discrete bands of lower mobility (B, lane 7).
Stimulation of topo IIα-mediated formation of DNA multimers by HMGB1 was independent of the topological state of the initial DNA (C). Interestingly, HMGB1 and topo IIα could induce small changes in the linking number of relaxed closed-circular DNA at the highest molar ratio HMGB1-to-DNA studied as evidenced by the formation of faster migrated DNA topoisomers (C, lane 5). The latter finding is reminiscent of the effect of HMGB1 on DNA supercoiling of closed-circular plasmid DNA by topoisomerase I (,).
In order to characterize the DNA multimers formed by plasmid DNA, HMGB1 and topo IIα in the presence of PEG, the deproteinized DNA multimers were treated with restriction nuclease HindIII or topo IIα. Digestion of the DNA multimers by HindIII resulted in a single band corresponding to the linearized plasmid DNA (C, lane 10). The DNA multimers could be resolved into relaxed circular DNA and a minor fraction of circular plasmid dimers upon treatment with topo IIα (D). The above results suggested that the DNA multimers formed by plasmid DNA, HMGB1 and topo IIα in the presence of PEG represented catenated DNA, i.e. double-stranded DNA molecules interlocked by a double-stranded DNA pass, rather than hemicatenated DNA molecules.
In order to further verify that HMGB1 promotes formation of fully inter-locked DNA molecules by topo IIα, and to assess the complexity of these multimers, the DNA complexes were visualized by electron microscopy (EM). Prior to EM, the DNA multimers (originating either from a 2.8-kb or a 9.1-kb plasmid for double-stranded or single-stranded DNA coating, respectively) were deproteinized, followed by coating with RecA protein for better imaging of the DNA at cross-over points. Examination of the samples by EM revealed that the DNA multimers formed by HMGB1 and topo IIα contained many linked DNA molecules (>60% of all visualized DNA samples), and that the majority of these catenated molecules formed heterogeneous population of DNA molecules inter-linked with one or several DNA molecules (; no DNA multimers were observed when DNA was incubated with HMGB1 in the absence of topo IIα, data not shown). The complex nature of these DNA multimers revealed by EM () corresponded to their migration in the course of agarose gel electrophoresis: they either remained at the origin of the gel or migrated with very low mobility (), suggesting their high-molecular mass.
To study the effect of HMGB1 on DNA decatenation, increasing amounts of the topo IIα were added to kinetoplast DNA (kDNA) in the presence or absence of HMGB1. As shown in A, the amount of decatenated DNA products increased proportionally with the amount of added topo IIα, and the decatenation activity of the enzyme was up to ∼10-fold higher (depending on the amount of the enzyme) in the presence of HMGB1 as revealed by the densitometry of the band intensities of the decatenated kDNA minicircles. No decatenation of kDNA was observed by incubation with HMGB1 alone in the absence of topo IIα (A).
HMGB1 protein consists of two HMG-boxes, domains A and B, and a highly acidic C-terminus (B). To find out which part of HMGB1 is responsible for stimulation of decatenation of kDNA by topo IIα, decatenation experiments were carried out with HMGB1 and its truncated forms. As shown in C, both HMGB1 and HMGB1-ΔC could enhance decatenation of kDNA to a similar extent, suggesting that the HMG-boxes of HMGB1, and ‘not’ the acidic C-tail, were responsible for the enhancement of the decatenation activity of topo IIα. Although both isolated HMGB1 domains, A and B, could decatenate kDNA, the observed stimulatory effect was significantly lower (up to ∼10-fold) than that observed with the full-length HMGB1 or the A+B di-domain (HMGB1-ΔC), C.
The fact that complex DNA multimers were formed under conditions when the relaxation activity of the enzyme was severely compromised by higher concentrations of PEG, suggested that HMGB1 could also stimulate the relaxation activity of the enzyme (B, compare lanes 6 and 7). As shown in D, when increasing amounts of topo IIα were incubated with negatively supercoiled plasmid DNA, DNA was relaxed more efficiently when HMGB1 was present in the reactions.
The above experiments provided evidence that HMGB1 could stimulate both DNA catenation/decatenation and relaxation of supercoiled DNA by topo IIα. While PEG was necessary to observe DNA catenation by low amounts of topo IIα in the presence of HMGB1 (), it was clearly dispensable for HMGB1-mediated stimulation of kDNA decatenation and relaxation of supercoiled DNA (), suggesting that PEG itself could not stimulate topo IIα activity.
HMGB1 has been known to affect DNA topology by unwinding or by inducing negative supercoiling (,,,). Recently we demonstrated that alanine mutagenesis of intercalating residues of HMGB1 domains, A (Phe 38) and B (Phe 103 and Ile122) (A and B), abrogated high-affinity binding of HMGB1 to bent DNA (minicircles) without affecting significantly its affinity to linear DNA (). Here we show that the HMGB1 mutant also lost the ability to supercoil DNA in a topo I-mediated supercoiling assay (C). To find out whether the mutated HMGB1 could still promote the catalytic activity of topo IIα, decatenation of kDNA was carried out. As shown in D, mutated HMGB1 could promote decatenation of kDNA by topo IIα indistinguishably from the wild-type protein, suggesting that the observed enhancement of catalytic activity of topo IIα by HMGB1 was not affected by the HMGB1-mediated changes in DNA topology.
The specificity of HMGB1 on the catalytic activity of human topo IIα was further investigated using type-II topoisomerase IV (). As shown in E, addition of increasing amounts of HMGB1 to the decatenation reactions did not enhance the activity of the prokaryotic enzyme. Thus, while the above data ruled out the effect of HMGB1-mediated changes in DNA topology on the stimulation of topo IIα activity (and also type-II topoisomerase IV), they may suggest an involvement of protein–protein interactions in the stimulatory effect of HMGB1 protein on the catalytic activity of human topo IIα.
To find out whether the stimulatory effect of HMGB1 on the catalytic activity of topo IIα could be due to a direct physical interaction of HMGB1 with the enzyme, the pull-down assay was used to investigate the possible existence of the HMGB1–topo IIα interactions. To address this question, recombinant GST alone or GST-HMGB1 were immobilized on glutathione-agarose and incubated with isolated human topo IIα. After extensive washing of the beads, the bound proteins were resolved by SDS-PAGE and subjected to western blotting and immunological detection using an antibody specific to topo IIα. While no topo IIα was found to be associated with GST alone, a significant fraction of topo IIα was bound to GST-HMGB1 (), suggesting that HMGB1 could interact with topo IIα. Using a similar approach we found that approximately similar amounts of topo IIα were associated with HMGB1 or HMGB1-ΔC, indicating that the HMG-boxes, and the polyanionic C-tail, of HMGB1 were responsible for the interaction with topo IIα (). From (upper panel), it also follows that both HMG-box domains, A and B, could interact with topo IIα. Similar results were obtained when DNAse I was included in the incubation buffer (, lower panel), demonstrating that the interaction of HMGB1 with topo IIα was independent of DNA that might have been associated with the isolated proteins. Specificity of the ‘pull-down’ assay was verified with other proteins known to interact with HMGB1 [p53, pRb, (,,)] as compared to those not binding to HMGB1 (E2F1, data not shown). Thus, the above ‘pull-down’ binding experiments strongly suggest that the interaction of HMGB1 with topo IIα is specific.
HMGB1 has previously been reported to promote binding of a plethora of transcription factors and other sequence-specific proteins to their cognate sites (). To find out whether the stimulatory effect of HMGB1 on the catalytic activity of topo IIα might have originated from enhanced binding of the enzyme to DNA, the EMSA technique (Electrophoretic Mobility Shift Assay) was used to study the effect of HMGB1 on topo IIα binding to DNA containing a strong topo IIα recognition sequence. A gradual increase in topo IIα binding to DNA was observed when increasing amounts of HMGB1 were added to a mixture of DNA and topo IIα (, arrow). Densitometric analysis revealed that the topo IIα binding to DNA was up to ∼10-fold enhanced by HMGB1. HMGB1 promoted topo IIα binding to DNA (Step 1 in ) without formation of a ternary complex topo IIα-HMGB1-DNA as no super-shift of the topo IIα–DNA complex was detected by specific anti-HMGB1 antibody (data not shown).
All catalytic activities of topo IIα are associated with the ability of the enzyme to form transient double-stranded breaks in the DNA duplex and to form covalent enzyme–DNA intermediates, referred as to the ‘cleavable complexes’. Topoisomerase IIα establishes two distinct DNA cleavage/religation equilibria: a ‘pre-strand passage DNA cleavage’ (corresponding to the ATP-free form of the enzyme), and a ‘post-strand passage DNA cleavage’ (corresponding to the ATP-bound form) (). To monitor a ‘pre-strand passage DNA cleavage’, ATP is omitted in the assays (Step 3 in ). A ‘post-strand passage DNA cleavage’ is measured in the assays employing a non-hydrolyzable ATP analog, adenyl-5′-yl imidodiphosphate (AMP-PNP) (Step 4 in ). When ATP is included in the cleavage assays, a mixture of the two equilibria (Steps 3 and 4 in ) is measured ().
To find out whether the observed stimulation of the catalytic activity of topo IIα by HMGB1 is due to enhanced cleavage of the DNA substrate, the topo IIα-mediated cleavage of the supercoiled plasmid DNA was carried out. This assay allows one to quantify the topo IIα-mediated double-stranded DNA cleavage by detecting the conversion of covalently closed-circular DNA to linear molecules. In the absence of drugs stabilizing the ‘cleavable complexes’, the evaluation of the DNA scission by topo IIα using ATP as a co-factor was complicated by the appearance of individual topoisomers in the course of relaxation of supercoiled plasmid DNA (A, +ATP). This was not the case when DNA cleavage assay was performed either in the absence of any cofactor (A, −ATP) or using the non-hydrolyzable ATP analog (A, AMP-PNP). The latter DNA cleavage experiments detected a significant (∼3-fold) HMGB1-mediated enhancement of DNA scission by topo IIα, while no visible DNA scission was observed when HMGB1 was incubated with supercoiled DNA in the absence of topo IIα (A).
The percentage of ‘cleavable complexes’ is usually very low, but the levels of the cleavable complexes are significantly enhanced in the presence of topo II poisons (such as etoposide) that affect the DNA cleavage/religation equilibrium (). Similarly to the DNA cleavage experiments in the absence of etoposide (A), HMGB1 could enhance >3-fold DNA cleavage by topo IIα when etoposide was included in the reactions (B). Densitometric analysis of the percentage of linearized DNA revealed that HMGB1 could stimulate more efficiently (>3-fold) the ‘post-strand passage DNA cleavage’ than the ‘pre-strand passage DNA cleavage’ (∼2-fold), regardless of the presence or absence of etoposide.
The effect of HMGB1 on DNA cleavage by topo IIα was also studied in reactions with increasing concentrations of etoposide (0–100 μM) and a fixed amount of the enzyme. Approximately ∼2–4 times higher percentage of linear DNA was observed in the presence of HMGB1, with the most prominent effect at lower concentrations of the drug (10–50 μM), C. The stimulatory effect of HMGB1 on formation of topo IIα–DNA cleavage complexes was detectable over a range of enzyme:plasmid ratios at fixed concentrations of the drug and HMGB1 (data not shown).
The ability of topo IIα to cleave and religate double-stranded DNA is essential for the regulation of DNA topology by the enzyme. As HMGB1 stimulates the DNA cleavage by topo IIα (Steps 3 and 4 in ), we have asked whether HMGB1 could also influence the ability of topo IIα to religate DNA using a heat-induced religation assay (Step 5 in ). This assay is based on the finding that religation activity of topo IIα is less sensitive to variations in temperature than DNA cleavage activity (). By shifting the temperature from 37 to 65°C before termination with SDS, linearized DNA plasmid (generated by topo IIα-mediated DNA cleavage in the presence of etoposide) is re-circularized in a time-dependent manner. As shown in , religation of DNA (as revealed by the decrease of linearized DNA) was very little, if any, enhanced by HMGB1, indicating that HMGB1 could not promote religation of the cleaved DNA by topoisomerase IIα. This was even more apparent when religation assay was carried out at 75°C (data not shown), a temperature which has been recently shown to inhibit more efficiently DNA cleavage by human topo IIα without significantly affecting the ability of the enzyme to religate DNA ().
Topoisomerase IIα changes the topology of DNA by coupling binding and hydrolysis of two ATP molecules to the passage of one DNA duplex through a transient double-stranded break in another DNA duplex. The ‘catalytic inhibitor’ ICRF-193 traps the topo IIα in the closed-clamp intermediate form in the presence of ATP (Step 6 in ), resulting in the inhibition of ATPase activity, strand passage and catalytic turnover of the enzyme (). As shown in A, HMGB1 could partially relieve (<2-fold) the inhibitory effect of ICRF-193 on decatenation of kDNA by topo IIα. A slight, but reproducible, decrease in the inhibitory effect of ICRF-193 on topo IIα-mediated decatenation of kDNA was also detected over a wide range of the drug concentrations (data not shown). The latter results prompted us to examine whether HMGB1 could modulate the rate of ATP hydrolysis by topo IIα using thin layer chromatography enabling to separate released phosphate from ATP upon incubation of [γ-P]ATP with topo IIα and negatively supercoiled plasmid DNA (). Quantification of the released phosphate revealed that HMGB1 could enhance (<2-fold) the rate of ATP hydrolysis by topo IIα (B). As the stimulatory effect of HMGB1 on ATPase activity of topo IIα is relatively modest, the DNA binding and cleavage are most likely the major control steps by which HMGB1 modulates the catalytic activity of the enzyme.
A distinctive property of living systems is that biochemical processes proceed in a medium which is highly crowded with macromolecules (). It has previously been shown that molecular crowding has a strong effect both on equilibrium and on the rate of reactions involving macromolecules (). Polyethylene glycol (PEG) as a volume excluding agent may imitate the conditions under which HMGB1 is able to facilitate (non-covalent) intermolecular associations of DNA helices with subsequent ligation by limiting amounts of enzymes (,,, and this study). The observed interlinking of circular, covalently closed DNA molecules by topo IIα in the presence of HMGB1 is likely a consequence of the PEG-mediated increase of local DNA and protein concentrations. Apart from increase of the DNA concentration, the presence of PEG obviously increases the concentration of HMGB1 in the catenation reaction (typically ∼1 μM in the test tube) which may then approach the expected actual concentration of HMGB1 in the cell nucleus, ∼25 μM (assuming ∼10 copies of HMGB1 per cell, the volume of the cell nucleus 6.5 × 10 ml, and disregarding the contribution of nuclear crowding agents which would otherwise likely lead to even higher concentration of HMGB1 in the nucleus). As higher amounts of topo IIα can promote DNA catenation in the absence of other factors (), there has been a possibility that the observed DNA catenation in the presence of HMGB1 might have had arisen solely as a consequence of a PEG-induced increase in the topo IIα concentration. However, this seems unlikely as PEG itself does not promote DNA catenation by low amounts of topo IIα when HMGB1 is absent. A variety of polycations, including spermidine and histone H1, promote catenation, and this is most likely due to compaction of the DNA substrate into aggregates with high local DNA concentration favoring the catenated state (). HMGB1 can promote interlocking of DNA rings by topo IIα only if a crowding agent is present in the course of the catenation reaction. This may reflect the fact that unlike DNA binding of the polyvalent cations, histone H1 or other nuclear proteins (such as nucleolin, DNA-PKcs and Ku70/80) (, and references therein), aggregation of HMGB1–DNA complexes is much less apparent in diluted solutions (,,).
HMGB1 promotes not only topo IIα-mediated catenation of circular DNA, but it can also enhance decatenation of kDNA and relaxation of negatively supercoiled DNA . The catalytic cycle of topo IIα depends on a coordinate function of several catalytic steps of the enzyme (). The first step includes binding of the enzyme to DNA (Step 1 in ) which has been demonstrated to be significantly enhanced by HMGB1. HMGB1 stimulates catalytic activity of topo IIα by promoting DNA cleavage (Steps 3–4 in ), without affecting the pattern of specific cleavage sites on DNA (data not shown). Although both ‘pre-strand’ and ‘post-strand passage DNA cleavage’ steps (Steps 3 and 4, respectively in ) are stimulated by HMGB1, the stimulatory effect by HMGB1 is more prominent at the ‘post-strand passage DNA cleavage’ step of the catalytic cycle of the enzyme. A possible explanation of the HMGB1-mediated stimulation of DNA cleavage by topo IIα is enhanced binding of the enzyme to DNA. The latter may be a consequence of the reported ability of HMGB1 to act as a ‘DNA chaperone’ [e.g. by bending DNA, ()] and/or due to direct interaction of the protein with topo IIα.
Topo IIα (,,), like HMGB1 (,,,,), can bind preferentially to supercoiled plasmid (including non B-DNA structures such as Z-DNA and four-way junctions) as compared to the linear B-DNA. Whether the reported preferential binding of HMGB1 to distorted or non B-type DNA structures contributes to recognition of topo IIα-specific DNA-binding sites/structures remains to be established. HMGB1 has also been previously reported to change DNA topology by unwinding the double-helical DNA or introducing negative supercoils in DNA (,,,). However, the HMGB1-mediated enhancement of decatenation by human topo IIα is similar when assayed using an HMGB1 mutant that does not affect DNA topology. In addition, the activity of topoisomerase IV, a prokaryotic type-II topoisomerase that has evolved independently from HMG-box proteins, is not influenced by changes in DNA topology induced by HMGB1. Therefore, we favor the hypothesis that the stimulatory effect of HMGB1 on the catalytic activity of the enzyme depends not only on DNA–protein interactions but also on protein–protein interactions. The fact that EMSA assay did not detect any ternary DNA–topo IIα–HMGB1 complex does not necessarily mean the lack of existence of such a complex. It is possible that HMGB1 can ‘deliver’ the topo IIα to its binding/cleavage sites by forming a ternary complex which is transient and unstable. This idea is in agreement with previous reports demonstrating the ability of HMGB1 to enhance the binding a plethora of sequence-specific proteins to their cognate DNA sites without formation of ternary complexes, despite the detection of the corresponding protein–protein interactions in pull-down assays (,).
ICRF-193 inhibits Step 6 of the catalytic cycle of topo IIα (), the ATP hydrolysis, by trapping the enzyme on the DNA in the closed-clamp form (). HMGB1 can partially relieve the inhibitory effect of ICRF-193 on decatenation of kDNA by topo IIα. HMGB1 could not significantly stimulate resealing of double-stranded breaks by topo IIα (religation Step 5 in ) using the temperature-induced religation assay. As the stimulatory effect of HMGB1 on the ATP hydrolysis by topo IIα (and even less on religation of double-stranded breaks) is relatively modest, the enhanced DNA binding of topo IIα and consequently higher DNA cleavage represent the major control steps by which HMGB1 modulates the catalytic activity of the enzyme.
An important question remains whether HMGB1 enhances the activity of topo IIα , and whether the possible activation involves direct stimulation of the enzymatic activity (as a result of mutual interactions) and/or synthesis of the enzyme. The latter possibility may be indicated by our luciferase reporter gene assays demonstrating a significant (>10-fold) transactivation of the human promoter by HMGB1 in Saos-2 cells (Štros ., unpublished data).
Topo IIα expression is associated with a higher cell proliferation rate, and the level of topo IIα is very low in G phase, it increases in S phase, and is maximal in G2/M (,,). Although HMGB1 is expressed throughout the cell cycle with no significant variations (), the protein is clearly over-expressed in most human tumors, including breast carcinoma, melanoma, gastrointestinal stromal tumors, colon carcinomas and acute myeloblastic leukemia (48 and references therein). Further experiments are needed to clarify whether this over-expression affects the cellular activity and expression of topo IIα, and if so, whether it contributes to the reported enhanced topo IIα activity in tumors (,). |
HIV-1 reverse transcriptase (RT) copies the viral genomic RNA into double-stranded DNA (dsDNA). Due to its essential role in viral replication and to the early availability of RT inhibitors, RT has been a leading target for anti-retroviral treatments. Currently, over half of the US Food and Drug Administration (FDA) approved anti-retroviral drugs target RT. These drugs fall into two categories: nucleoside analog RT inhibitors (NRTI) block extension of the template DNA upon incorporation into the replicating genome, and non-nucleoside RT inhibitors (NNRTI) bind a hydrophopic pocket near the RT active site resulting in allosteric inhibition (,). Although these small-molecule inhibitors have helped slow the progression of AIDS, their long-term utility can be compromised by cellular toxicity and the emergence of drug resistant HIV-1 strains (). The proven effectiveness of anti-RT therapeutics validates the push for new molecular inhibitors of RT. Antagonists that utilize novel inhibition mechanisms are especially attractive in that they may be less cytotoxic and may avoid the current escape mutations associated with NRTIs and NNRTIs.
High-affinity DNA and RNA aptamers have been selected to bind RT. These aptamers inhibit both the polymerase (pol) and RNase H functions of the protein () and have the potential to inhibit all steps of reverse transcription, including RNA- and DNA- primed extensions on either RNA- or DNA-templates, strand displacement and RNA cleavage by RNase H (). Half-maximal inhibition is observed in the picomolar to low micromolar range (), with RNA-primed reactions showing the greatest susceptibility to aptamer inhibition (). Aptamers appear to compete with primer/template for binding to RT (,,,), and have accordingly been referred to by some authors as template-primer analog RT inhibitors (TRTIs) (). Biochemical probing () and crystallographic studies () have shown that a canonical RNA aptamer folds into a pseudoknot structure and binds to RT in the primer-template binding-cleft. Because aptamers exploit inhibitory mechanisms that are distinct from those utilized by small-molecule inhibitors, they offer a unique opportunity in combating HIV.
Several studies have shown that intracellular expression of RNA aptamers to RT protects these cells from HIV-1 challenge and HIV-1 gene expression (), and that virus produced in cells expressing RNA aptamers are less infectious when applied to aptamer-naïve cells (). This protection extended across multiple HIV-1 subtypes and several drug-resistant viruses (). Other studies have identified ssDNA aptamers and double-stranded, sulfur-containing thioaptamers that bind the RNase H domain of RT (). Although the affinity of these aptamers for RT is much weaker than that of ssDNA aptamer RT1t49 (described below), these aptamers also afford protection to cells when administered prior to challenge with low to moderate levels of virus (,). The demonstrated antiviral efficacy of aptamers in three distinct modes—expression within target cells, co-packaging into nascent virus within producer cells and exogenous delivery to target cells—motivate further analysis of the molecular basis of RT inhibition by aptamers.
Aptamer RT1 is an 81-nucleotide ssDNA that was selected from a degenerate library containing 35 random positions. It has a reported Kd value of 1 nM and IC (reported as Ki) value of <0.3 nM, reflecting an RT-binding affinity that is more than 1000 times greater than that of the library from which it originated (). The authors introduced random mutations into RT1 and re-selected molecules that retained a high affinity for HIV-1 RT (2 nM after six SELEX cycles versus ∼1500 nM Kd for the partially randomized library). Comparative sequence analysis of the reselected species enabled truncation of the original 81-nucleotide aptamer to a 49-nucleotide version, denoted RT1t49, with similar affinity for HIV-1 RT (Kd ≈ 4 nM) (). A recent study using capillary electrophoresis has shown that RT1t49 binds to HIV-1 RT with a 1:1 stoichiometry, whereas two other DNA aptamers from the same selection (RT12 and RT26) appear to form complexes with two aptamers per RT (). Using enzymatic extension assays, we have previously shown that the potent inhibition of RT polymerization and RNase H activities by RT1t49 extends across multiple subtypes of HIV-1, HIV-2 and SIV RTs, whereas another ssDNA aptamer (RT8) inhibited only the HIV-1 RT subtype utilized in the original selection (). Fisher . () have demonstrated inhibition of multiple drug-resistant forms of HIV-1 RT by RT1t49, and observed that mutations in RT that give rise to biochemical resistance to RT1t49 come at a price of a severe replication deficiency in the virus (). Each of these observations highlights the importance of defining the molecular contacts that determine aptamer specificity.
There is little information on the specific nucleic acid structural requirements of RT1t49 for RT binding. Therefore, we analyzed a panel of RT1t49 variants to dissect the secondary structural requirements of RT1t49. We found that a stable, stem–loop structure with considerable sequence plasticity is important at the 5′-end, as is a shorter stem near the 3′-end, and that multiple regions of the molecule tolerate alterations from the primary sequence. The results of our mutational and deletion analysis of RT1t49 support a model wherein several alternative secondary structures contribute to the high-potency inhibition of HIV-1 RT by RT1t49. In addition, both solution and site-specifically generated hydroxyl radical probing enabled us to identify a specific interaction between RT1t49 position A32 and the pol-active site of HIV-1 RT. Finally, we show that a truncated version of RT1t49 that is unable to form the previously proposed stem II structure nevertheless retains an ability to inhibit RTs across multiple subtypes of HIV-1, HIV-2 and SIV. The implications of these results for drug resistance and for potential drug development are discussed.
Cy3 fluorophore-labeled and unlabeled DNA oligonucleotides were synthesized by Integrated DNA Technologies (Coralville, IA, USA). Nuclease S1 was purchased from Roche and used as recommended by the manufacturer at a dilution of 1:2000 in S1 nuclease buffer. Radioactive nucleotides were obtained from Amersham Biosciences (Piscataway, NJ, USA). RT from HIV-1 strain HXB2 (used in most experiments) and from other sources (used in ) were expressed in and purified as described ().
For RT1t49 mutational analysis, DNA-dependent DNA polymerization was monitored by single nucleotide (ddCTP) extension of a Cy3-labeled 18 nt primer (5′ Cy3- 3′) on a 40-nt template (5′ CAGTGTGGAAAATCTCTAGCAG-3′). A 1:1.5 ratio of primer to template (25 nM primer, 37.5 nM template) resulted in nearly complete annealing of primer with template (data not shown). Therefore, reactions were performed using 25 nM primer, 37.5 nM template and 100 ng/µl bovine serum albumin (BSA) in 1x reaction buffer (75 mM KCl, 5 mM MgCl, 50 mM Tris–HCl pH 8.3, 10 mM DTT) and 0.2 mM ddCTP. Primer and template strands were heated together at 95°C for 2 min and allowed to anneal at room temperature for 5 min. Synthetic ssDNA aptamers were re-suspended in water and added to the primer/template mix (final concentrations from 0.3 to 729 nM, in 3-fold serial dilution increments) prior to the addition of RT. Extension reactions were initiated by addition of HIV-1 RT (0.4 nM final active concentration) and incubated at 37°C. Reactions were quenched after 3 min with two volumes of stop buffer (urea-saturated 95? formamide, 50 mM EDTA and bromophenol blue). Reaction products were separated on 15? denaturing (7.5 M urea) polyacrylamide gels and scanned for Cy3 fluorescence using a Molecular Dynamics Typhoon Phosphorimager. Image analysis was performed using ImageQuant 5.2 software.
The percent extension was calculated by dividing the fluorescent signal from the N + 1 extended product by the sum of the signals from the unextended and extended primer bands. These data were normalized for extension without aptamer (100? relative extension) and without RT (0? relative extension). Normalized data were fit to a sigmoidal dose-response curve with a constant slope of 1 with GraphPad Prism software using Equation ():
Where is the log of inhibitor concentration and is measured relative percent extended at a given inhibitor concentration.
Aptamer inhibition of primer extension by RTs from across HIV-1, HIV-2 and SIV clades was performed as previously described (). Briefly, the Cy3-labeled 18 nt primer was extended on a 103-nt DNA template corresponding to the HXB2 5′ LTR U5 and PBS sequences (5′-AAGTAGTGTGTGCCCGTCTGTTGTGTGACTCTGGTAACTAGAGATCCCTCAGACCC TTTTAGTCAGTGTGGAAAATCTCTAGCAG-3′). The PBS sequence is underlined. Reactions were assembled in the 1x RT reaction buffer with 30 nM primer, 45 nM DNA template and 0.2 mM dNTPs. Aptamers were added to final concentrations of 0.3 to 243 nM, in 3-fold serial dilution increments. To maximize the dynamic range of polymerization reactions, extension proceeded until just prior to reaching the plateau phase of the product formation curve in unchallenged assays (10 min, data not shown). Reactions were initiated by the addition of enzyme to 0.4 nM final active site concentration, quenched after 10 min in 37°C with two volumes of 95? formamide, 50 mM EDTA and then analyzed as above. Three independent datasets were averaged to calculate the reported IC values; uncertainties reflect the standard deviations among IC values.
For S1 nuclease digestion, [P] 5′-end-labeled DNA aptamer (>10 c.p.m.) was unfolded at 95°C for 2 min with 1 x S1 nuclease buffer (Roche) in the presence of BSA (100 ng/μl), allowed to fold at room temperature for 5 min, and then kept on ice. Reactions were initiated by adding an equal volume of S1 nuclease (0.4 U/μL in 1 x S1 nuclease buffer.) After 15 min on ice, reactions were quenched with two volumes of stop buffer (urea-saturated 95? formamide, 50 mM EDTA and bromophenol blue) and electrophoresed on a 12? denaturing (7.5 M urea) polyacrylamide sequencing gel. The gel was dried, exposed to a storage phosphor screen and analyzed with the Molecular Dynamics Typhoon Phosphorimager.
Previously, Gotte . () showed that by replacing the Mg(II) in the RNase H active site with Fe(II) allows the site-specific generation of OH radicals that cleave the closest nucleotide in the DNA. Using the same methodology, we monitored both Fe(II)-dependent hydroxyl radical cleavage due to pol- and RNase H-specific cleavage as well as hydroxyl radical cleavage from bulk solution. Site-specifically generated hydroxyl radicals were generated as described (,). Briefly, 2.5 μl of [P] 5′-end-labeled DNA aptamer (>10 c.p.m.) was incubated in 10 nM sodium cacodylate, 0.1 mM EDTA, 50 mM KCl, 1 mM CaCl, 5 mM MgCl and 100 μg/ml BSA in a total volume of 16 μl. The sample was heated to 95°C for 2 min and cooled to room temperature for 10 min to facilitate aptamer folding. To a final concentration of 560 nM, 2 μl HIV-1 RT (or buffer) was added to the bottom of a 1.5 ml eppendorf tube. The hydroxyl radical solution was made by adding 1 μl of a freshly prepared Fe(II)-EDTA solution to the wall of the tube. A 1 μl drop containing 0.6? hydrogen peroxide (freshly diluted from a 30? stock) was added directly to the Fe(II)-EDTA solution, and finally 1 μl of 20 mM sodium ascorbate was added to the Fe(II)-EDTA/HO solution. The reaction was initiated by ‘flicking’ the hydroxyl radical solution into the DNA aptamer solution in the bottom and was quenched with stop solution (2.5 mM thiourea, 0.1 M sodium acetate and 20 μg glycogen as carrier) 2 min later. Reaction products were precipitated with three volumes of 100? ethanol, and the DNA pellet was re-suspended in 20 μl of loading buffer containing 95? formamide, 50 mM EDTA, bromophenol blue and xylene cyanole and electrophoresed on a 12? denaturing (7.5 M Urea) polyacrylamide sequencing gel and analyzed as above.
Single-stranded DNA aptamer RT1t49 has previously been shown to be an effective inhibitor of HIV-1 reverse transcriptase (). In the present study, we set out to determine the sequence and structural constraints associated with this inhibition. We have previously shown that RNA-dependent DNA polymerization is more susceptible to aptamer inhibition when compared with DNA-dependent DNA polymerization, therefore we have chosen to evaluate the inhibition of our aptamer constructs using DNA-dependent DNA polymerization assays (). Single nucleotide incorporation assays offer a convenient and readily quantifiable means of assessing DNA polymerization activity (). For these experiments we used a Cy3-labeled 18-nucleotide DNA primer corresponding to the 3′-end of human tRNA annealed to a 40-nucleotide DNA template containing the HIV-1 primer binding site (PBS) (A). Annealed primer/template was present in excess of primer/template over RT to allow multiple-turnover conditions. Increasing amounts of inhibitor were added to the reaction mixture prior to the addition of RT. Fitting the data for the RT1t49 titration to a sigmoidal dose-response curve gives half-maximal inhibition (IC) at 3.6 ± 0.9 nM (B and C). This equation assumes both a 1 : 1 aptamer:RT stoichiometry and that the aptamer is in excess of RT near the IC value. Because the total RT concentration is comparable with this value of IC, these data were also fit to the quadratic form of the binding equation and to a sigmoidal dose-response curve with a variable slope function. For each fit, the reported IC value fell within the error of the original fit (data not shown). Therefore, 3.6 nM was used throughout the rest of this study as the reference IC against which all others were compared.
The MFOLD algorithm () on the Integrated DNA Technologies web server () identified three potential secondary structures for RT1t49 (). The most stable structure (ΔG = −4.7) is also the one proposed by Schneider . () (A, hereon referred to as ‘structure A’). All three conformations share an identical stem I at the 5′ end, followed by a single-stranded connector and a variable stem II at the 3′ end. The nucleotides responsible for the alternative stem II structures are highlighted in according to base-pairing potential. Overall, the structure of stem II contributes significantly to the stability of RT1t49 as structure B (B) and structure C (C) are less stable (ΔG = −3.6 and −2.7, respectively) than structure A. To determine whether structural variability influences RT inhibition potency, a series of RT1t49 sequence and deletion mutants were created. Their effects on RT inhibition are described subsequently.
Three separate mutations were introduced into stem II of RT1t49 (). The first of these was designed to address the importance of the three-nucleotide loop capping the end of stem II in structure A (A). Converting the predicted tri-loop to an unrelated stable tetra-loop (AAC → TGCG) increased IC nearly 20-fold to 65 ± 10 nM. To determine whether this perturbation was due to loss of specific contacts between RT and the tri-loop at the end of stem 2, or to a decrease in the ability of RT1t49 to assume alternative stem II structures, the terminal pair of stem II was changed from G–T to a G–C Watson–Crick base pair. This mutation stabilized the proposed stem by −0.6 kcal/mol and had little consequence on the IC (5.5 ± 1.4 nM). While T45 is not predicted to be base-paired in the alternate stem II structures, the increased stability in structure A suggests a more-stable stem II is well tolerated. The third alteration disrupted the stem II in structure A by simultaneously introducing mutations G47A and G48A. Surprisingly, this mutation had no effect on the observed IC (3.1 ± 0.5 nM).
The results described above suggest that either the stem II associated with structure A is not an important structural element for RT inhibition or that alternative structures play a role in the high-affinity inhibition by RT1t49 and can be accommodated by these mutations. To differentiate between these possibilities, a series of RT1t49 variants were created, each with successive single nucleotide deletions from the 3′ end until the entire stem II of structure A was completely deleted. Up to five nucleotides (−1 through −5) could be deleted (creating a 44-nt species terminating with C44) without significantly affecting the IC values (C), which ranged from 5.1 to 8.1 nM. Clearly RT1t49 variants are still capable of a high level of inhibition even when the stem II of structure A is unable to form. However, removal of one additional nucleotide (−6) caused a dramatic jump in IC to 138 ± 68 nM. Further deletions of nucleotides −7 through −10 all resulted in IC values between 215 and 285 nM, while additional deletions (−11 through −13) eliminated nearly all inhibition and yielded IC values near 1 μM. These results strongly suggest that structures B or C (or both) contribute significantly to the inhibition of HIV-1 RT.
Several mutations and deletions were made to address the importance of the length, sequence and single-stranded character of the short connector region separating the two stems of RT1t49 (). Shortening this connector from six nucleotides to five by deleting A31 had only a minor effect, increasing IC to 8.3 ± 2.1 nM. However, deletion of both A31 and A32 raised the overall IC value nearly 40-fold to 127 ± 29 nM, and deletion of A31, A32 and A33 produced a 100-fold increase in IC (361 ± 81 nM) relative to that of RT1t49. Deletion of nucleotides C34 and T35 separately increased IC to 38 ± 18 nM and 58 ± 6 nM, respectively, but the simultaneous deletion of both nearly eliminated inhibitory function at all aptamer concentrations surveyed, yielding a calculated IC that approached 1 μM. Several base-substitution mutations were introduced to test the importance of individual nucleotides within the connection region. Mutation C34A increased the IC to 100 ± 35 nM, mutation T35A increased the IC to 228 ± 96 nM and the combination mutation of C34A/T35A increased the IC to >500 nM. These observations clearly indicate that the single-stranded connector between stems I and II plays an important role in RT inhibition by RT1t49. Several individual nucleotides are implicated as either making individual, specific contacts with the protein or in forming the alternative stem II structures found in structures B and C.
Three sets of mutations were introduced into RT1t49 to determine the consequences of weakening, strengthening or altering the sequence of stem I (). The C4A and T11A/T12A mutations weakened the base-pairing potential of stem I, and caused a nearly 5-fold increase in the IC value relative to that of RT1t49 (IC = 17 ± 1.2 nM and 21 ± 3.3 nM, respectively.) Disruption of T8–A23 and G9–C22 bp near the center of stem 1 prevented formation of stem I altogether (as predicted using MFOLD) and yielded a more substantial loss of inhibition (IC = 70 ± 29 nM).
Mutations that increased stem I stability had only modest effects on RT inhibition. Three of the four mutants surveyed gave IC values within 2-fold of the parental RT1t49. Especially interesting is the absence of significant effect on IC for the two constructs that fully pair the internal bulges. Crystallographic analysis has shown that the primer-template binding cleft between the pol and RNase H active sites of RT can accommodate between 17 and 18 bp of dsDNA (), which takes on an A-form/B-form mixed helix and has a bend of 40° to 45° at the junction between the A-form and B-form segments (). It has been suggested that aptamers with pre-existing bends may be recognized more easily than a simple uninterrupted helix by RT (,). In particular, the unpaired nucleotides in stem I of RT1t49 have been proposed to serve this role (). However, the results above suggest that the helical nature of stem I is more important for RT recognition than bending associated with the bulged positions within the stem.
To exclude the possibility of aptamer recognition through sequence-specific interactions, several mutations were introduced wherein nucleotides in one strand were switched with those in the other strand in various parts of stem 1 (B and C). Inverting the base-paired region in the beginning segment of stem I (RT1t49-Inv A) resulted in a moderate increase in IC (16 ± 4.9 nM). When the middle segment of stem I was inverted (RT1t49-Inv B) there was only a small increase in IC (10 ± 3.2 nM). The two alterations were not additive, as introduction of the two inversions simultaneously (RT1t49-Inv A + B) resulted in an IC of 9.3 ± 2.6 nM. Furthermore, changing most of the nucleotides in both stems to other Watson–Crick pairs (variant RT1t49-Change Sequences) yielded an IC value (8.8 ± 3.1 nM) that is in line with the other variants in this series. Thus, RT recognizes stem I of RT1t49 in a fashion that is predominantly independent of sequence.
To define more precisely the secondary structure of the folded aptamer, two versions of RT1t49 were subjected to enzymatic digestion with S1 nuclease (). For the full-length RT1t49, major cleavages are observed at nucleotides G5/G6 and from A31-G36, with secondary cleavages occurring at nucleotides C3, T12–T21, A40–G41 and A43–T45. These weaker cleavages could indicate that the dominant solution structure is in equilibrium with one or more minor, alternative folds. When higher molecular weight fractions of these same samples were separated by running the samples longer, no S1 cleavage was observed at nucleotides T46 through A49 of RT1t49, indicating that these nucleotides are base-paired or otherwise structured in the dominant conformation (data not shown). Thus, the S1 nuclease digestion of RT1t49 supports the structural prediction of structure A. For aptamer RT1t49(-5), which lacks the 3′-most 5 nucleotides, the cleavage patterns are essentially identical at G5/G6 and from A31–A36, although the secondary cleavages sites at nucleotides A17–T21 appear to be cleaved less efficiently than in RT1t49. Although both aptamers appear to form identical stem I and possess similar connection regions, the RT1t49(-5) deletion mutant lacks the capability of folding into a structure with the proposed stem II in structure A. Interestingly, the secondary cleavages observed beyond nucleotide T36 in RT1t49 are absent from the digestion pattern for RT1t49(-5). These data suggest that the 3′-terminal segment of RT1t49(-5) is base-paired or structured (as in structure B and C), and they support our hypothesis that alternative structures within stem II contribute to inhibition of HIV-1 RT.
We have recently shown that RT1t49 inhibits pol and RNase H activities of RTs from diverse clades of HIV-1, HIV-2 and SIV (). The present work demonstrates that multiple structures may contribute to inhibition of HXB2 RT, raising the possibility that multi-clade inhibition may have been dependent upon access to multiple structures. For example, some RTs may have been inhibited only by the dominant solution structure (structure A) while others were inhibited only by one of the alternatives (structure B or C). We therefore evaluated the effect of RT1t49(−5) on cross-clade inhibition of DNA-dependent DNA pol extension of an 18-nt primer on a 103-nt template. Inhibition by the full-length aptamer (RT1t49) is essentially identical to inhibition by RT1t49(−5) (). The difference of the IC values between pairs was not significant, and we found that RT's from HIV-1 subtypes C and G were more than 2-fold more sensitive to inhibition by RT1t49(−5) than by RT1t49. These data support our hypothesis that the stem II associated with structure A is not essential for the overall inhibition of HIV-1 RT and that the alternative structures associated with stem II must contribute significantly to the inhibition of lentiviral RT.
Previously Götte .() showed the RNase H domain of HIV-1 RT could utilize two distinct mechanisms to cleave both the RNA template of DNA/RNA hybrids and dsDNA (). Replacing the Mg(II) in the RNase H active site with Fe(II) allows the site-specific generation of OH radicals that cleave the closest nucleotide in the dsDNA. Building on this concept, we evaluated whether RT1t49 and RT1t49(-5) bound with HIV-1 RT could react with OH radicals generated from Fe(II) bound at the active sites of HIV-1 RT (A). Both aptamers show uniform, low-level reactivity to OH radicals in the absence of protein (lanes 2 and 7). Upon the addition of wild-type RT, new major cleavages are observed at nucleotides 11 and 32 (lanes 3 and 8). To determine whether these sites of hyperactivity were due to Fe(II) bound in the pol or RNase H active sites, wild-type RT was replaced with mutants that selectively disrupt metal ion binding in either the pol (D185N or D186N) or the RNase H (E478Q) active sites. Although RNase H catalytic activity can be rescued in the last mutant in the presence of Mn(II) (), this replacement abolishes RNase H-directed, Fe(II)-mediated radical cleavage of dsDNA (). Cleavage at position 11 remained strong for all three mutants, supporting aptamer being bound to the RT mutants. No change in the cleavage pattern of either aptamer was observed for the RNase H mutant (data not shown), ruling out radical cleavage mediated by the RNase H active site. In contrast, cleavage associated with nucleotide A32 was essentially eliminated in both of the pol-active site mutants for both aptamers. A plot of the intensity of cleavage from nucleotides 30–34 shows a dramatic increase in cleavage of wild-type versus pol-active site mutants at A32 (B). These results strongly suggest that A32 is located very near the pol-active site.
Hydroxyl radical footprinting is an established method for determining dsDNA binding domains and has been used to evaluate helicity of dsDNA (). Plotting the intensity of bands from nucleotides 6 through 36 in these OH radical probing experiments provide further insight into the structure of bound aptamer (A and B). Cleavage levels are above background for two regions (nucleotides 10–15 and 22–27) when exposed to OH radicals in the presence of protein. These nucleotides are located within stem I and are arranged such that these regions pair with (or are opposite of) regions found within the background intensity (C). Overlaying these sites with a crystal structure of HIV-1 RT bound to a dsDNA primer-template [2HMI and 1RTD, (,)] show this difference in intensities can be explained by the formation of a double helix and that these regions would be solvent exposed if they bound RT as primer-template mimics (A). Interestingly, this assignment places A32 within the dNTP binding cleft and in direct contact with the Mg(II) binding sites (B), supporting our results of hyperactive OH radical cleavage associated at the pol-active site. The lack of hyperactive OH radical cleavage at the RNase H active site can be readily explained by this model, as stem I is too short to reach the RNase H active site when A32 is in the pol active site. Our data ascribe no specific structural role for the 3′ stem II, although a reasonable extrapolation is that it extends beyond the active site to the ‘back side’ of RT, interacting either with the fingers (in analogy with the 5′ unpaired template) or thumb domain.
Further modeling suggests RT1t49-resistant mutations N255D and N265D (,) would be in direct contact with nucleotides 27 and 5, respectively, and place these amino acids in a position to influence interactions with RT1t49. Mutation R277K has been shown to be resistant to Type I RNA pseudoknot aptamers but not to RT1t49 [(), data not shown]. Based on our modeling, this mutation faces away from the helix and may have only minimal effect RT1t49 binding (data not shown). We believe understanding the molecular interactions between RT and RT1t49 may lead to new insights into the development of alternative therapeutics that can inhibit broadly across the primate lentiviral family.
bold
#text |
The development of high-density microarrays has enabled the study of thousands of individual transcripts in parallel and helped to identify the distinct transcriptional profiles of tumors (). As a result, hierarchical clustering of these tumor profiles has proven to be valuable for classification of cancers of the same tissue type. One such example has been the classification of gene expression patterns in primary breast tumors, which led to the identification of four distinct tumor subtypes subsequently linked with different clinical outcomes (). Such studies substantiate the use of molecular taxonomy in clinical medicine for accurate cancer diagnosis and especially for identification or design of suitable therapeutic approaches ().
Retrospective transcriptional profiling of archived tissues, especially those collected from subjects for whom long-term disease outcome is known, represents an attractive objective for identification of novel therapeutic targets. Unfortunately, the specimens that have been collected in surgical pathology have been routinely formalin-fixed and paraffin-embedded (FFPE), a preservation process that has been shown to induce the formation of cross-linkages between proteins and between proteins and nucleic acids (). This fixation method impedes the extraction of large amounts of RNA and is especially detrimental for single-stranded RNA molecules that appear fragmented and chemically modified (). These characteristics have been major roadblocks for systematic high-throughput analysis of archived tissues. In fact, multi-gene retrospective analyses of FFPE-RNA have been achieved only by hybridization or by relative quantification of target transcripts using real-time RT-PCR (). Although RT-PCR techniques have been enhanced for the study of larger gene sets, this method remains impractical for the analysis of tens of thousands of genes and especially for the identification of novel cancer-related genes, hence the need for developing more appropriate technologies (,).
One major disadvantage of microarray analysis is the requirement for significant amounts of high-quality RNA, which is essential for increased sensitivity and reproducibility, a standard that cannot be met by fragmented RNA extracted from archived tissues. Although a few commercial kits have been designed to amplify reliably small amounts of starting material, several studies have pointed out that FFPE-RNA is not a good substrate for cDNA synthesis (,). The RT of fragmented and chemically modified RNA molecules gives rise to a population of small cDNA transcripts that in turn produce short complementary RNA (cRNA) probes by transcription (IVT). The presence of undersized probes, along with amplified non-specific sequences, contributes to the increase of non-specific signal and thus dilution of specific signal in microarray experiments. There have been a few reports that described the high-throughput transcriptional profiling of FFPE-RNA, but the use of highly fragmented RNA limits studies to a small subset of genes (,). Recent studies performed with the Arcturus Paradise Reagents demonstrated that severely degraded RNA could not be analyzed, and showed that high-throughput analysis of clinical samples was limited to only 24% of specimens archived between two and eight years (,).
Considering that the fragmentation of RNA and the loss of valuable information increase with archival time, we have developed a reliable technique for microarray analysis of archived specimens that would otherwise be rejected because of the poor quality of the recovered RNA. This strategy has been designed for sequence selection and extension of short single-stranded DNA molecules, reverse-transcribed from FFPE-RNA, by copy of a complementary template (CT). We have termed this process, complementary-template reverse-transcription (CT-RT) because intact complementary sense-RNA templates are annealed to the short single-stranded cDNA primers derived from FFPE tissue, and reverse-transcribed. We investigated the correlation between a matched 10-year-old frozen and 10-year-old FFPE-RNA derived from human breast cancer tissue using the already established T7 IVT-amplification method and compared it to our strategy, by cDNA microarray analysis (). We demonstrated that the CT-RT process specifically increases the size of the cDNA transcripts utilized in the T7 IVT-amplification reactions. We demonstrated the high reproducibility of this process as well as the increase of specific signal in cDNA microarray experiments by comparison with direct T7 IVT-amplification of FFPE-RNA. We demonstrated that the recovery of single-stranded DNA primers from old FFPE tissue provides access to transcripts that are undetectable by IVT-amplification. Using the same RNA-template library for the restoration of cDNA primers reverse-transcribed either from breast or from cervical archived tissues, we further demonstrated the specificity of CT-RT. Our analysis reveals that highly fragmented poly(A) transcripts recovered from older archived tissues can successfully be restored and lead to the recovery of valuable transcriptional features.
We used universal human reference RNA (UHR; Stratagene), the quality of which was assessed on a Bioanalyzer 2100 expert (Agilent). The linear amplifications were performed using the MessageAmp II aRNA amplification kit from Ambion. The sense RNA template library was generated with the SenseAmp RNA amplification kit from Genisphere ().
Matched 10-year-old frozen and 10-year-old FFPE breast cancer samples were obtained from the Montefiore Medical center, Bronx, NY. Matched 3-year-old frozen and FFPE cervical samples were also obtained from the Montefiore Medical center. RNA from 10-year-old frozen tissue was extracted using TRIzol reagent following the manufacturer's instructions (Invitrogen). The 10-year-old breast cancer and the 3-year-old cervical FFPE tissues were macro-dissected and underwent the same RNA extraction procedure. Areas of interest were identified on H&E slides. Macro-dissection consisted of removing two to three punches of 1 mm from the FFPE tissue, which yielded 5 μg of RNA for the 10-year-old tissue, and between 5 and 20 μg of RNA for the 3-year-old tissue. The FFPE tissue was de-paraffinized using 500 μl of Hemo-De at room temperature on an agitator, three times (). The tissue was washed with 1 ml of 100% RNase-free ethanol three times followed by three washes with 1 ml of 95% RNase-free ethanol on ice, for 8 min each time (). The tissue was then washed with 1 ml of 1× PBS DEPC treated and incubated in 200 μl of RNase-Free 1× PBS and 6.5 μl of RNase-Out (Invitrogen) for 90 min on ice, for rehydration. Prior to proteinase K digestion, each tissue was homogenized in a 7 ml Wheaton homogenizer using 2.010 ml digesting buffer (50 mM Tris–HCl pH 7.5, 75 mM NaCl, 5 mM CaCl and 0.1% SDS). The homogenized tissue was separated in 15 tubes of 134 μl, to which was added 1 μl of RNase out (Invitrogen). A volume of 15 μl of proteinase K at 30 mg/ml was added to each tube (Roche Diagnostics). The digestion was carried out at 59°C for 1 h, upon agitation every 5 min. After 1 h, digests were gathered in two tubes and spun down at 12 000 r.p.m. for 1 min. The pellets were kept on ice, and 1 ml of butanol-1 was added to the supernatants. The tubes were vortexed and spun at 10 000 r.p.m. for 1 min. The butanol extractions were repeated to achieve partial removal of aqueous phase and obtain a final volume of 100 μl. This solution was used to re-suspend the tissue pellets, which led to a final volume of 150 μl. The solution was homogenized in 1 ml of TRIzol (Invitrogen) following the manufacturer's instructions. The RNA present in the supernatant was precipitated with 1 μl of 0.1 mg/ml of linear acrylamide and 3 M sodium acetate and 600 μl of isopropanol. The tubes were incubated for 12 h at −20°C and then spun at 14 000 r.p.m. for 30 min at 4°C. The precipitated RNA was washed with 200 μl of 70% RNase-free ethanol, dried and re-suspended in RNase-free water (Promega). The RNA was quantified on a Nanodrop ND-100 spectrophotometer and analyzed on Bioanalyzer using the Agilent-2100 software.
The synthesis of single-stranded DNA primers from FFPE-RNA was achieved by using 5 μg of RNA in a 20 μl reaction. For the RT of the FFPE-RNA, we used Arrayscript (Ambion), the same enzyme as the one provided in the MessageAmp II RNA kit (Ambion). Each reaction contained 5 μg of RNA, 2 μl of 10× arrayscript buffer, 1 μl of RNase inhibitor mix, 4 μl dNTP mix 2.5 mM each (Ambion), 1 μl (100 ng) of T7-Oligo dT-VN(5′-GGCCAGTGAATTGTAATACGACTCACTATAGGGAGGCGGdT (A/C/G)(A/C/G/T)-3′) and 1 μl of arrayscript reverse-transcriptase. The reaction proceeded at 42°C for 2 h. The 20 μl reaction was brought to 400 μl with RNase-free water (Sigma) for purification on a microcon YM-50 (Millipore), as instructed by the manufacturer. The purification was performed by spinning the YM-50 at 500 for 12 min to cut-off double-stranded cDNA/RNA fragments under 100-bp and single-stranded T7-oligo dT -VN. The filter was washed three times using 400 μl of RNase-free water (Sigma). A volume of 88 μl was recovered and incubated with 10 μl of 10× RNase H buffer and 2 μl of RNase H 10 U/μl (Ambion) for 30 min at 37°C. The solution was transferred to a boiling water bath for 3 min and transferred on ice for 5 min. The single-stranded cDNA was purified using a MinElute purification column, following manufacturer's instruction (Qiagen). The single-stranded cDNA was recovered in 10 μl and quantified on a Nanodrop NO-100 Spectrophometer.
The direct amplification of RNA extracted from matched 10-year-old frozen and FFPE breast cancer tissues and from the 3-year-old frozen cervical tissue were performed with the MessageAmpII aRNA kit from Ambion, using 5 μg of total RNA for each reaction, as instructed by manufacturer. The T7 IVT-amplifications proceeded for 14 h at 37°C. For the restorations, the CT-RT process was performed using the single-stranded DNA primers obtained from 5 μg of FFPE-RNA from the 10-year-old breast cancer and the 3-year-old cervical samples. In order to prevent the T7-oligo dT sequences of the purified cDNA primers from priming poly(A) sequences of random templates in the Sense-RNA template library, 1 μl of the Non-Sense Knock-out oligonucleotide (NSK), 5′-(A/C/T/G)(C/T/G)dA CCGCCTCCCTATAGTGAGTCGTATTACAATTCACTGGCC-3′ (0.5 μg/μl was added to the 9 μl single-stranded cDNA solution. The mix was incubated for 10 min at 70°C, 10 min from 70 to 37°C and at 37°C for 10 min for hybridization. The solution was then speed-vacuum-dried to obtain a final volume of 1 μl. The sense-RNA template library was prepared using 2.5 μg of fresh UHR RNA (Stratagene) and amplified using the SenseAmp RNA Amplification Kit from Genisphere, following the manufacturer's instructions. We made one minor change to the procedure, by adding DNase-I prior to the sense-RNA purification in order to remove any trace of DNA. The integrity of the UHR-RNA was checked on a Bioanalyzer (Agilent). The sense-RNA was quantified using the Nanodrop ND-100 and stored in 10 μg/5 μl samples at −80°C. For the CT-RT reaction, we annealed 8.8 μl of sense-RNA and 1 μl of purified cDNA/NSK1 primers in presence of 1.2 μl of 10× first strand buffer and one drop of RNase-free mineral oil (Sigma). The solution was incubated in a 0.5 μl microfuge PCR tube in a Perkin–Elmer Cetus DNA thermal cycler with a thermocycle file as follows: 70°C for 10 min, 70–42°C in 90 min. The first strand cDNA synthesis was prepared by adding 1.2 μl of RNase-free water, 0.8 μl of 10× first strand buffer, 4 μl of dNTP mix, 1 μl of ribonuclease inhibitor and 1 μl of arrayscript reverse-transcriptase from the MessageAmp II aRNA kit (Ambion) and added at the end of the cycle, at 42°C and incubated for 2 h at 42°C. At the end of the cycle, the tube was transferred on ice. The second strand cDNA synthesis mix was prepared (63 μl of RNase-free water, 10 μl of 10× second strand cDNA buffer, 4 μl of dNTPs, 2 μl of DNA polymerase and 1 μl of RNase-H) and added under the mineral oil to the 20 μl solution, following the manufacturer's instructions (Ambion). The tubes were incubated for 2 h at 16°C, the cDNA was purified and the aRNA synthesized following the IVT instructions for 14 h (Ambion). The aRNA was quantified on a Nanodrop spectrophotometer and analyzed on a Bioanalyzer.
Double-stranded cDNA obtained from 5 μg of UHR RNA, 10-year-old frozen RNA, 10-year-old FFPE-RNA and 10-year-old restored FFPE-cDNA was used for the PCR reactions. Half the volume of the cDNA recovered from MessageAmpII cDNA purifying columns was used to prepare a master mix for each set of PCR reactions (nine reactions). Three sets of four primers (Invitrogen), containing three forward and one reverse-primer, were synthesized for the three corresponding genes, human Cyclin D1 (Ccnd1; GenBank accession number: 053056), human tumor protein 53 (p53; GenBank accession number: 000546) and human tyrosine-kinase type-receptor (HER-2/neu/ERBB2; GenBank accession number: 004448). For Ccnd1, from 5′ to 3′ end, forward-primer 1; 5′-GTGATGGGGCAAGGGCACAAGTC-3′, Primer 2; 5′-CGGCTGGGTCTGTG-CATTTCTGG-3′, primer 3; 5′-CCCAGCACCAACATGTAACCGGC-3′ and reverse-primer 5′-TGGGGTTTTACCAGTTTTATTTC-3′. For p53, from 5′ to 3′ end, forward-primer 1; 5′-GCTGGTCTCAAACTCCTGGGCTC-3′, primer 2; 5′-GTGGAGCTGGAA-GGGTCAACATC-3′, primer 3; 5′-CCCACCCTTCCCCTCCTTCTCCC-3′ and reverse-primer; 5′-GCAGCAAAGTTTTATTGTAAAATAAG-3′. For Her-2, from 5′ to 3′ end, forward-primer 1; 5′-GCGACCCATTCAGAGACTGTCCC-3′, primer 2; 5′-GTGTCAG-TATCCAGGCTTTGTAC-3′, primer 3; 5′-GGGGAGAATGGGTGTTGTATGGG-3′ and reverse-primer; 5′-TGCAAATGGACAAAGTGGGTGTGGAG-3′. Each forward-primer was paired with the corresponding reverse-primer for each gene (Ccnd1, p53 and Her-2). All PCR reactions were performed using the Platinum DNA polymerase high-fidelity kit, in 50 μl reactions (Invitrogen). Each reaction contained 1 μl of cDNA, 2 μl of forward- and reverse-primer (1 μg/μl), 2 μl of 50 mM MgCl, 5 μl of 10× high fidelity buffer (600 mM Tris–SO pH 8.9, 180 mM ammonium sulfate), 5 μl of 0.2 mM of dNTP (2.5 mM), 33 μl of distilled water and a drop of mineral oil. Negative controls were performed using 1 μl of sterile distilled water instead of cDNA. Platinum high-fidelity polymerase (2 μl of 0.5 U/μl) was added after cDNA denaturation for 5 min at 95°C, reactions were performed in a Perkin–Elmer Cetus DNA thermal cycler for 30 cycles (95°C for 1 min, 50°C for 1 min 30 s and 68°C for 2 min, ending with a final extension step at 68°C for 10 min). The PCR amplicons were visualized using a UV light box after electrophoresis on a 1.5% agarose gel containing 0.5 μg/μl ethidium bromide. The gels were photographed using a Fluorchem Imager (Alpha Innotech Corporation, CA, USA).
The cRNA produced from each amplification was used to make fluorescent probes by RT. For each microarray experiment, 5 μg of cRNA synthesized from UHR RNA (reference) and 5 μg of cRNA from our samples were labeled. The aRNA was incubated in the presence of 2.67 μl of random primers (3 μg/μl) from Invitrogen in a final volume of 19 μl at 70°C for 10 min, spun down and put on ice for 5 min. Labeling reactions were performed by adding 8 μl of 5× first strand buffer, 4 μl of 0.1 M DTT, 4 μl of dNTP labeling mix (2.5 mM of each), 4 μl of 25 nM Cy3-labeled deoxyuridine triphosphate (Cy3-dUTP) for cRNA amplified from UHR (reference) or 4 μl of 25 nM Cy5-labeled deoxyuridine triphosphate (Cy5-dUTP; Amersham Pharmacia Biotech, NJ, USA) for cRNA from our samples, 1 μl of RNase-out at 40 U/μl (Invitrogen), 1.5 μl of Superscript II reverse-transcriptase 200 U/μl (Invitrogen) and incubated at 42°C for 1 h. After 1 h of incubation, 1.5 μl of Superscript II reverse-transcriptase was added for another 60 min at 42°C. The 40 μl reactions containing the labeled cDNAs were incubated in the presence of 44 μl of RNase-free water, 10 μl of 10× RNase-One buffer and 2 μl of RNase-One 10 U/μl (Promega) for 35 min at 37°C for removal of cRNA templates. The RNase-One was then inactivated by transferring the tubes at 95°C for 3 min and kept on ice. The 100 μl of Cy3-labeled UHR cDNA and the 100 μl of Cy5-labeled sample cDNA were combined in a tube containing 20 μl sodium acetate 3 M (PH5.1), 2 μl of 0.1 μg/μl linear acrylamide (Ambion) and 500 μl of 100% ethanol and precipitated at 14 000 r.p.m. for 30 min. The probes were washed with 200 μl of 70% RNase-free ethanol and air dried before being re-suspended in 16.5 μl of RNase-free water for 28 000 (28k) features microarrays and 6.5 μl of RNase-free water for 9000 (9k) features microarrays. The microarray slides were incubated at 50°C with 60 or 15 μl of pre-hybridization buffer for 28k or 9k microarrays, respectively. The 16.5 μl probes were incubated with 40.5 μl of hybridization buffer and 3 μl of human block solution for 28k microarrays and the 6.5 μl probes with 12.5 μl hybridization buffer and 1 μl of human block for 9k microarrays. The pre-hybridization and hybridization buffers were supplied by the Albert Einstein College of Medicine (AECOM) microarray facility, and human block solution prepared as described at: . After 1 h the slides were washed in distillated water and dried. The 60 μl labeled probe solution was added onto the 28k microarray slide and covered with a 22 × 60 premium cover glass, and the 20 μl of labeled probes added onto the 9k microarray slides and covered with a 22 × 22 Premium Cover Glass (Fisher). The microarrays were placed in a sealed chamber in a water bath at 50°C overnight, in the dark. The slides were washed the following day as described in Belbin . (). Dry slides were scanned using the GenePix 4000A microarray scanner (Axon Instruments, Foster City, CA, USA). The UHR Cy3 (Green) and the FFPE-RNA Cy5 (Red) images were acquired and gridded with GenePix Pro 6.0 Software to generate signal intensities.
Arrays used for the studies were designed and printed at the cDNA Microarray Facility, Albert Einstein College of Medicine (AECOM), Bronx, NY. The 28K arrays (28 704 spots) represent 14 725 human cDNA clones of the distinct genes with sequence annotation and gene ontology information, 8376 expressed sequence tags (ESTs) with incomplete sequence annotation, 5603 ESTs without any available annotation as of the date of the study and 192 bacterial sequences for quality control of the arrays. The 9K arrays contained 9216 spots (). The background-subtracted signals were normalized with a global LOWESS algorithm using in-house Perl scripts and the R statistical package. Two-parameter normalization was done between different arrays when necessary. Mean and SD of the signal intensity were calculated for each gene in every group of the repeated experiments. Missing and low intensity signals (<1000) were excluded from the correlation analysis.
The fragmented messenger RNA extracted from FFPE tissue is primed with a T7-oligo dT primer and reverse-transcribed. The RNA/cDNA duplex is filtered to allow the removal of single-stranded T7-oligo dT primers that have not been used up in the reaction, providing double-stranded fragments longer than 100 nt (a). The cDNA that has been synthesized and purified is single-stranded by RNA degradation with RNase-H. A blocking primer, complementary to the T7-oligo dT sequence of the purified cDNA, is added to prevent random annealing of the oligo-dT to the poly(A) tail of sense-RNA templates. The sense-RNA template library is obtained by IVT of a T7 promoter incorporated into the 3′ end of the first strand cDNA, which provides RNA with the same orientation as messenger RNA (b) (). Optimal annealing conditions between single-stranded DNA regions and sense-RNA templates are obtained in 90 min through a temperature gradient from 70 to 42°C (c). The RT of the hybridized sense-RNA templates allows for the extension and thus restoration of incomplete DNA sequences. The process of CT-RT is followed by double-stranded DNA synthesis, T7 IVT-amplification and cDNA microarray analysis of the cRNA ( Supplementary Data for the details of the method).
We tested our strategy on matched 10-year-old frozen and 10-year-old FFPE samples obtained from the same breast cancer tissue. After extracting RNA from the frozen tissue, we performed one round of T7 IVT-amplification with 5 μg of total RNA in four individual reactions (a). One of the reactions was stopped after purification of double-stranded DNA for PCR experiments, while the remaining three reactions underwent T7 IVT-amplification. Using the FFPE tissue, we extracted RNA from the same region as from the frozen tissue and obtained a 260/280 ratio of 1.90. We tested the direct T7 IVT-amplification of FFPE-RNA by preparing four reactions using each 5 μg of RNA (b). One of the reactions was stopped after purification of double-stranded DNA (dsDNA) and used for PCR experiments, while the three remaining reactions underwent T7 IVT-amplification. In order to evaluate the restoration strategy, we prepared four reactions containing 5 μg of FFPE-RNA each and performed RTs with the reverse-transcriptase provided in the MessageAmpII amplification kit from Ambion (c). After purification, for each reaction we obtained a set of T7-dT -single-stranded DNA primers that we hybridized with 10 μg of sense-RNA template library. The sense-RNA template library contained templates with sizes ranging between 250 and 1000 nt, peaking at 500 nt as observed on an Agilent bioanalyzer 2100 expert (a). The size of these transcripts was shorter than the expected size of transcripts obtain by IVT amplification of fresh RNA. The CT-RT was carried out by adding the reagents provided in the MessageAmpII aRNA kit (Ambion). The second strand DNA synthesis was carried out. One reaction was stopped after cDNA synthesis for PCR experiments, while the other three underwent T7 IVT-amplification.
In order to investigate the benefit of the CT-RT process over the single RT of short FFPE-RNA, we compared the size distribution of cRNA and cDNA products for each of our T7 IVT-amplifications on a bioanalyzer 2100 Agilent. The 10-year-old frozen RNA displayed a degradation pattern characterized by the absence of 28s ribosomal RNA and low abundance of 18s RNA, with a ladder-like pattern (b, lane 2). The T7 IVT-amplification of this frozen RNA gave rise to products ranging between 50 and 1000 nt, peaking at 200 nt (lane 4). Frozen RNA has been shown to provide shorter cRNA in previous studies (). The three amplifications generated 105, 88 and 88.5 μg of cRNA. The RNA extracted from the 10-year-old FFPE-RNA provided us with much smaller products, as most fragments were <200 nt (lane 3). The T7 IVT-amplification of this FFPE-RNA gave rise to products ranging between 50 and 250 nt, which peaked at 125 nt, in each of the triplicate reactions (lane 5). Although the RNA appeared largely degraded, the amplification reactions yielded 44, 40.4 and 32.9 μg of cRNA, thereby providing sufficient amounts for microarray analyses. We then analyzed the restored cRNA obtained by CT-RT reactions, performed on the same 10-year-old FFPE-RNA, and observed a large increase in the size of the products. The cRNA products ranged between 50 and 850 nt with a peak at 300 nt (lane 6). The CT-RT process produced lower amounts of cRNA in three reactions, with 28.2, 27.8 and 27.3 μg. Considering that the CT-RT process is based on the increase of available DNA sequences by a secondary RT of a longer complementary sense-RNA template, we compared the dsDNA obtained in our three different experiments on a Bioanalyzer 2100 Agilent (c). The dsDNA generated from 10-year-old frozen RNA appeared the largest with sizes ranging between 1000 and 2000 nt (lane 1). The dsDNA obtained by RT of FFPE-RNA appeared the smallest with sizes no larger than 200 nt (lane 2), whereas dsDNA obtained by CT-RT of sense-RNA templates exhibited products as large as 750 nt (lane 3). Taken together, these results indicate that frozen RNA provides a good template for IVT-amplification. Degraded FFPE-RNA, however, provides much smaller templates for IVT-amplification. The CT-RT of cDNA obtained from FFPE-RNA appears to provide lengthened transcripts.
In order to verify that the increase in cRNA size, for the restored material, was due to the RT of CTs, we designed a PCR experiment to test for the presence of gene and size-specific dsDNA molecules. We chose three genes of significance for breast cancer, the human Cyclin D1 (Ccnd1), human tumor protein 53 (p53) and human tyrosine-kinase type-receptor (HER-2/neu) (). For each gene, we designed three sense oligonucleotides increasingly spaced from the 3′end of the transcripts. Then, by combining each of these different sense primers with the anti-sense primer, in individual PCR reactions, we determined the amount of sequence available from the 3′ end of dsDNA for each gene (see Supplementary Data for details on the PCR). We first tested our PCR reactions in the absence of a known source of dsDNA, which did not generate any amplicons (, panel 1). Using dsDNA, obtained by RT and dsDNA synthesis of fresh UHR RNA with the MessageAmp II amplification kit from Ambion, we validated the presence of each of the three genes, and demonstrated that their RNA templates were larger than 250 nt by obtaining the largest PCR products for each of the three genes (panel 2, lanes 3, 6 and 9). We performed these PCR reactions with dsDNA obtained from RNA extracted from 10-year-old frozen tissue and detected each of the three different PCR products for each of the three genes, thus demonstrating the presence of RNA transcripts larger than 250 nt for each gene (, panel 3). Although the PCR experiments were performed with the same conditions as with dsDNA obtained from fresh UHR-RNA, we detected an increased number of non-specific products. These results suggested that fragmented frozen RNA may provide partial templates for annealing of the short PCR primers. We then performed PCR reactions with dsDNA obtained from RNA extracted from 10-year-old FFPE tissue, and verified the presence of the three genes by detecting the smallest PCR amplicons (, panel 4, lanes 1, 4 and 7). The detection of these products revealed that short dsDNA templates were present, for all three genes, but did not exceed at 150 bp, thus demonstrating the absence of dsDNA of 251 bp for Ccnd1, 214 bp for p53 and 225 bp for Her-2 (, panel 4). These PCR results corroborated the bioanalyzer analyses, which suggested that RNA from 10-year-old FFPE tissue contained transcripts around 100 nt and no larger than 165 nt. When we performed these PCR reactions with dsDNA obtained by RT of the FFPE-RNA and restored by the CT-RT process, we were able to detect much larger products for each of the three genes (, panel 5). We then cloned and sequenced the largest PCR amplicons of each of the three genes, and verified that these sequences were 100% specific to each of the three genes (, panel 5, lanes 3, 6 and 9). Taken together, these results demonstrated that the RT of older FFPE-RNA provides short but specific DNA transcripts that effectively match the ones detected in frozen RNA. When these cDNA transcripts were used for the CT-RT of cRNA templates, a physical restoration of gene-specific sequences was detected. The T7 IVT-amplification of larger cDNA templates produces larger cRNA transcripts.
In order to evaluate the robustness of the CT-RT process we analyzed the cRNA, obtained after subsequent IVT-amplification, on cDNA microarrays (). For quality control, each cDNA microarray was hybridized with both Cy5-labeled material (the experiment) and Cy3-labeled UHR cRNA (the reference). We first investigated the robustness of the IVT-amplification process. We examined the correlation in gene expression between technical repeats using cRNA obtained from fresh UHR RNA (Cy3) on cDNA microarrays (a). The coefficient of determination ( = 0.9495), which was high, demonstrated the strength of the linear relationship between repeats and thus the reliability of the IVT-amplification. We performed the same measurements with cRNA amplified from partially degraded RNA extracted from 10-year-old frozen tissue, and obtained similarly high coefficient of determination (b, = 0.9425). These results demonstrated that the MessageAmpII aRNA kit from Ambion provided high performance IVT-amplifications even with partially degraded RNA. We then evaluated the quality of the cRNA amplified directly from RNA extracted from 10-year-old FFPE tissue. The coefficient of determination was much lower ( between 0.5684 and 0.6201), reflecting that the correlation between technical repeats might be impeded by the degradation and/or the chemical modification of FFPE-RNA (c). We then evaluated the robustness of the CT-RT process, when performed with RNA extracted from the 10-year-old FFPE tissue, by cDNA microarray analysis of technical repeats (d). The coefficient of determination was significantly higher ( between 0.8621 and 0.8495) than the ones obtained by direct IVT-amplification of FFPE-RNA. Together, these results demonstrated that single-stranded DNA, obtained from highly degraded and chemically modified and older FFPE-RNA, provides a highly reliable material for CT-RT reactions and IVT-amplifications.
We then sought to evaluate the amount of information that could be retrieved from RNA extracted from the 10-year-old FFPE tissue by comparison with transcriptional information contained in messenger RNA from the 10-year-old frozen tissue, using either direct IVT-amplification or the CT-RT process followed by IVT-amplification. Using the microarray data, we determined that the number of observable spots with a signal higher than 1000 in each the red (Cy5) and the green (Cy3) channels was almost four times higher with the CT-RT process than by direct IVT-amplification of FFPE-RNA (1306; ). Our data also revealed that the number of spots, displaying <20% of variability with the signal of matching spots from frozen material, was five times higher by the CT-RT process than by direct IVT-amplification of the 10-year-old FFPE-RNA (). These results indicated that the CT-RT process allowed for a more reliable and a significantly higher recovery of features from the 10-year-old FFPE tissue. These features could not be detected by microarray analysis of material obtained by direct IVT-amplification of RNA from 10-year-old FFPE tissue. We chose one sample grid from our different microarray experiments to display the signal intensity generated by Cy5-labeled probes obtained by RT of the different cRNAs (a). We determined that the cRNA obtained by direct IVT-amplification of RNA from 10-year-old FFPE tissue generated the lowest intensity (FFPE-Amp 1–3), where cRNA obtained by CT-RT and IVT-amplification (FFPE-Restored 1–3) provided access to transcripts with matching signal intensities to the transcripts detected in 10-year-old frozen tissue (Frozen 1–3). In order to compare the amount of signal available on the arrays, we chose to analyze genes with expression ratios ranging between 0.5 and 2 in frozen tissue. We selected a set of genes detected in both frozen RNA and in UHR RNA, and generated a heat map for 1044 genes with expression ratios between 0.5 and 2 (b, Frozen-Amp 1–3). The UHR RNA was utilized to generate sense-RNA library that provided the templates used by CT-RT in restoration experiments. We observed that the restoration by CT-RT (FFPE-Restored 1–3) provided access to a much larger set of genes with expression ratios overlapping the ones from frozen material, than direct IVT-amplification of FFPE-RNA (FFPE-Amp 1–3). We determined that the total number of transcripts recovered from 10-year-old FFPE tissue with ratios matching the ones of transcripts from frozen tissue was three times higher after restoration than after direct IVT-amplification (1218, ). Furthermore, the coefficient of determination was four times higher ( = 0.38) than after IVT-amplification of FFPE-RNA ( = 0.10), despite the increased number of features. When we selected features with expression ratios ranging between 0.5 and 2, we detected after restoration more than three times the number of genes after IVT-amplification. For these genes, the CT-RT process provided a coefficient of determination 25 times higher ( = 0.50) than the one obtained by IVT-amplification ( = 0.02). Together, these results demonstrated that the restoration of FFPE material provides access to a larger set of transcripts, which display higher intensities possibly due to their elongation during the process of CT-RT. Our results demonstrate that the transcripts detected in restored experiments correlate better with the ones detected in frozen RNA than those detected by direct IVT-amplification of FFPE-RNA.
We then assessed the reproducibility and the reliability of the CT-RT process for distinguishing archived clinical samples that provide severely degraded RNA. We chose to compare a 3-year-old matched frozen and formalin-fixed normal cervical sample to the 10-year-old matched frozen and formalin-fixed breast cancer sample (). For these experiments, total RNA isolated from the 3-year-old cervical frozen and formalin-fixed tissues, as well as the cRNA obtained by IVT-amplification of frozen RNA or by CT-RT and IVT-ampliication of FFPE-RNA were analyzed on the Agilent 2100 bioanalyzer (Supplementary Data). In order to compare the cRNA derived from IVT-amplifications of 10-year-old breast cancer and the 3-year-old cervical frozen tissues, we performed a dye-flip experiment to minimize labeling bias. By overlapping the genes detected by Cy3 and by Cy5 dyes, for each sample, we identified the expressed genes in each tissue (). This dye-flip experiment was then performed with cRNA obtained by CT-RT and IVT-amplification of 10-year-old breast cancer and 3-year-old cervical FFPE RNAs. Comparison to the transcripts detected in frozen samples revealed that 41.3% of the breast transcripts and 35.7% of the cervical transcripts were restored (). The overlap between transcripts detected in frozen and restored FFPE materials was calculated to be ∼89% with ∼10% false positives (). We analyzed the transcriptional profiles of the two tissues for the presence of tissue-specific features. By comparing the microarray profiles of frozen breast cancer and cervical tissues, we identified 905 breast-specific genes and 1013 cervix-specific genes (). The comparison between restored transcriptional profiles from the breast cancer and cervical samples identified 42 breast-specific and 109 cervix-specific transcripts. Comparison of tissue-specific transcripts revealed that 3.4% of the breast-specific and 6.8% of the cervix-specific transcripts were restored. It is important to note that the sense-RNA library (UHR), which was used for the restoration of our FFPE samples, was obtained by pooling 10 different tissues and primarily provided high concentrations of transcripts common to all the tissues giving a selection bias for common transcripts. Thus, transcripts specific to one tissue were diluted by a 1/10 ratio and therefore might have been too low in abundance, which could explain the relatively higher yield for all transcripts (35–41%) versus the lower yield for tissue-specific transcripts (3–7%). Altogether, our experiments demonstrated that single-stranded DNA primers representing the 3′ region of mRNA, adjacent to the poly(A) tail, contain sufficient information for retrieval of tissue-specific transcripts by -CT-RT.
The successful analysis of RNA extracted from archived samples has been achieved in studies that targeted a limited number of genes. One such remarkable example is the resolution of the crystal structure of a major surface antigen of the extinct 1918 ‘Spanish’ influenza virus, which killed over 20 million people worldwide. This was determined after reassembly of the hemagglutinin gene from viral RNA fragments retrieved from 1918 formalin-fixed lung tissues (,). Although RNA from FFPE tissue is fragmented and chemically modified, RT-PCR experiments have successfully demonstrated the presence of valuable stretches of information spanning over a 100 nt (,,,,). Taking advantage of the presence of these sequences, we devised a novel strategy for the molecular restoration of valuable portions of information lost through fixation and extended storage. This novel technology has the potential to provide a tool for the retrospective high-throughput analysis of older archived samples and therefore the recovery of valuable transcriptional information.
The transcriptional profiling of moderately degraded RNA, submitted to multiple rounds of IVT-amplifications, has been shown to provide reasonable results (). Similar studies using IVT-amplifications based on the random priming of degraded RNA indicated also that microarray analyses might also be feasible with FFPE-RNA (,,). It has been shown, however, that the extent of fragmentation, which significantly increases with archive storage time, may become a detriment to the efficient detection of transcripts by microarray analysis (). Our results strongly corroborated these findings, as RNA recovered from a 10-year-old archived tissue appeared largely degraded, with fragments peaking at 165 nt and smaller than 200 nt. The microarray analysis of cRNA obtained by IVT-amplification of 10-year-old FFPE-RNA revealed that the correlation between experimental repeats was much lower than repeats performed on 10-year-old frozen RNA (). Although the amplifications provided sufficient amounts of cRNA, our experiments revealed that the short size of the products might have contributed to the loss of signal as well as the generation of non-specific signal. Our findings suggest that the process of T7 driven IVT-amplification may have primarily benefited short transcripts, non-specific messages or other species of RNA carrying poly(A) stretches, thereby severely decreasing the correlation between expression profiles of experimental repeats as well as the correlation with frozen material. It is important to note that T7 RNA polymerase has been shown to be abortive and less efficient with linear cDNA transcripts containing the T7 promoter, a process that might also contribute to the shortening of cRNA transcripts during amplification of small cDNA molecules (, , ).
The purpose of our study was to improve the quality and the specificity of microarray analyses performed with cRNA obtained from old FFPE material that contains severely fragmented RNA and that is routinely rejected from high-throughput analyses (,). RNA fragmentation and/or degradation are processes encountered with older FFPE tissues or samples that have been improperly fixed, and have remained major roadblocks for reliable microarray analyses (,). In fact, when Penland . used the state-of-the-art Paradise Reagent System from Arcturus on 2 to 8-year-old FFPE tissues, 76% of the archived clinical samples were rejected from microarray profiling due to poor RNA quality (). Therefore, we designed the process of CT-RT to strengthen transcript selection prior to IVT-amplification and in turn strengthen signal detection in microarray experiments, when using RNA of the poorest quality. This process was designed to investigate the small single-stranded DNA primers synthesized from highly fragmented messenger RNA, by hybridization and RT of sense cRNA templates. This strategy was built around the already established and widely used T7 IVT-amplification method in order to minimize variables in our experiments. Using PCR experiments, we showed that this process increased the 5′ region of cDNA molecules by at least 100 nt in length. Although the CT-RT process required the preliminary purification of DNA primers on a filter and a column, our results indicated that these steps contributed to the improvement of T7 IVT-amplification of older FFPE material. Our experiments also showed that the RT of CTs was a highly reproducible mechanism as we obtained coefficient of determination between 0.84 and 0.86, for material as old as 10 years. When we correlated the transcriptional profiles from frozen and FFPE tissues, our method allowed 35–41% of transcript recovery for gene ratios ranging between 0.5 and 2, where the direct IVT-amplification of FFPE-RNA only provided 2% of recovery. Our results support the hypothesis that short transcripts recovered from archival material might be a major source of non-specific signal as well as a contributing factor to the loss of signal. By contrast, transcript restoration by CT-RT dramatically improves transcript selection and in turn the correlation between transcriptional profiles of matched frozen and FFPE material, and this process has the advantage to provide T7 RNA polymerase with longer cDNA molecules for IVT.
The extension of short single-stranded DNA primers by RT of complementary sense-RNA templates, represented in excess in a library, is a reaction that gives access to genes that are still represented in highly fragmented FFPE-RNA but that cannot be detected by direct IVT-amplification. CT-RT is a multiplex process that can be performed in a single reaction and may lead to the recovery of thousands of biologically significant transcripts from valuable archived clinical samples. Sense-RNA templates are selected by the DNA primers, synthesized from the archived RNA, during the annealing reaction. These short DNA primers, although representative of the 3′ region of mRNA transcripts, adjacent to the poly(A) tail, demonstrated to contain specific and sufficient information to recognize CTs contained in a sense-RNA library (,). Therefore, the restoration of single-stranded DNA primers, synthesized from highly degraded RNA, is a reaction that has the potential to lead to the retrospective identification of transcriptional biomarkers that have remained in the archived tissues as fragmented sequences.
The focus of our work was to establish a reliable method for the high-throughput analysis of severely degraded RNA typically recovered in 76% of archived clinical samples (). Our results demonstrated that CT-RT provided access to biologically relevant genes, which could discriminate between two different tissues. It is important to note that the use of a generic library, where transcripts from 10 different tissues have been gathered, might not provide optimal tissue-specific sense-RNA concentration for the annealing of single-stranded DNA transcripts that are recovered in low abundance in a single tissue. Our experiments suggested that transcript recovery by CT-RT might depend upon the presence of the CT in the sense-RNA library as much as its representation in the single-stranded DNA primer pool. Therefore to improve the process of CT-RT as well as its efficiency, we propose that the selection of specific libraries containing a complete array of tissue, disease and/or tumor-specific transcripts might provide access to more informative and more complete transcriptional profiles when using highly fragmented FFPE-RNA from a specific tissue. The CT-RT process might also be improved by increasing the amount of recovered single-stranded DNA primers. The use of magnetic micro-beads may provide better mRNA recovery and thus increase the amount of primers for application of the CT-RT process to smaller FFPE-RNA fractions. (). It should be noted that the CT-RT process may bias alternative splicing profiles of transcripts containing identical 3′ regions, towards that of the reference template, but the overall cumulative expression level of the splicing variants in the group should remain intact. In summary, this novel technique provides the basis for a new molecular approach to the microarray analysis of severely degraded RNA, recovered from most archived clinical samples.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
The phenomenon of RNA interference was first observed almost a decade ago () and seems to have evolved as a defense mechanism against foreign double-stranded RNA. It is triggered by short RNA duplexes (∼21–23 nt in length), which are processed from longer double-stranded RNA transcripts (,). Two major classes of small double-stranded RNA molecules have thus far been identified: small interfering RNAs (siRNAs) and microRNAs (miRNAs). SiRNAs are fully complementary double-stranded molecules while miRNAs originate as duplexes that have bulges. Both classes of double-stranded RNAs assemble in the RNA induced silencing complex (RISC), which contains key proteins for the processing and functioning of double-stranded RNAs in RNAi. A member of the Argonaute family, Ago-2, has been identified as the catalytic core of this complex ().
Since their discovery, the use of siRNAs has quickly widened to diverse areas of research and as potential therapeutic agents (). High-throughput analysis based on the use of siRNA libraries is also revolutionizing the field of functional genomics (). However, RNAi is an important component of endogenous cellular processes () involved in post-transcriptional regulation of endogenous gene expression (,) as well as antiviral protection (). Interfering with the endogenous functions of RNAi could lead to severe toxicity as recently demonstrated ().
The key component for the cellular export of shRNAs and microRNAs is the nuclear karyopherin Exportin-5 (). In the presence of Ran-GTP, Exportin-5 binds to both shRNAs and microRNAs transporting them from the nucleus to the cytoplasm. Exportin-5 is a saturable transport pathway so the excessive production of small interfering RNAs could result in a decrease of cellular miRNA function (,). This export function is not required for the activity of synthetic siRNAs (). In this work, we show that the competition between shRNAs with cellular miRNAs is a general phenomenon that also takes place with synthetic siRNA sequences. The ability to compete varies with the sequence and does not depend solely on the saturation of Exportin-5. We show that siRNAs, which do not depend on Exportin-5 for their transport to the cytoplasm, retain their ability to compete against shRNAs and microRNAs. Ectopic expression of Exportin-5 only partially relieves the competition between shRNAs and endogenous miRNAs or exogenous siRNAs. Ectopically expressed shRNAs and synthetic siRNAs, but not ectopically expressed microRNAs are able to interfere with each other and with the endogenous microRNA pathway through their ability to be incorporated into RISC. However, the relative strength of competition can be equalized for all the tested siRNA sequences by reducing the cellular concentration of the HIV-1 TAR RNA binding protein (TRBP). Thus, our results suggest that TRBP is a sensor in the loading of the RNA guide sequence into RISC.
U6shRNA-expressing constructs were synthesized via a PCR-based reaction that includes primers encoding the shRNA with a 3′ region complementary to the U6 promoter and a primer complementary to the 5′ end of the U6 promoter as described previously (). All PCR reactions were carried out as follows: 1 min at 94°C, 1 min at 55°C and 1 min at 72°C for 30 cycles. The resulting PCR cassettes were cloned directly into the pCR2.1 plasmid (TA cloning vector, Invitrogen). The shRNA constructs used for the competition experiments are targeted against sequences present in the HIV- (shSII and shSI) HIV (shSI, shTAT), HIV (shVif) and the EnvPb1 retrovirus envelope (shL) genes. The shRNA Luc (Luc) is targeted to a sequence present in the coding region of the firefly luciferase gene. All shRNAs are expressed by transcription from the U6 promoter. The synthetic siRNAs (siH3, siH6 and siH1) are targeted against sequences present in the coding region of the hnRNPH gene. Refer to the Supplementary Information for specific sequences for the shRNAs and siRNAs. All other oligonucleotide sequences are available upon request.
The microRNA constructs were similarly constructed using overlapping primers containing the wild type or a variant mir30 sequence. The mir30 guide sequence was replaced with 21 nt complementary to a site present in the HIV- transcript (SII) by including the corresponding nucleotide mutations in the PCR primers. The resulting PCR products were cloned into pcDNA3.1 under the control of the CMV promoter.
The luciferase target (RLT) was generated by cloning a 21-nt HIV- sequence (SII) into the XhoI site located in the 3′-untranslated region (3′-UTR) of the humanized luciferase gene of plasmid psiCHECK-2™ (Promega).
Cells were grown in DMEM (Irvine Scientific, Santa Ana, CA, USA) supplemented with 10% fetal calf serum (Irvine Scientific), 1 mM -glutamine. To generate the HCT116-GFPmiR21 cell line, HCT116 cells were transfected (using Lipofectamine 2000, Invitrogen) with SapI linearized pcDNA4/GFPmir21 plasmid DNA () (generous gift from Thomas Tuschl) and cells with stable plasmid integrations were selected by applying 200 μg Zeocin per ml media for 10 days. Single colonies were picked and screened for expression of EGFP upon transfection with a 2′-methyl RNA complementary to miR21 relative to a control 2′O-methyl RNA complementary to miR33.
For the competition experiments, 293 cells were transfected with the shRNA expression constructs and the corresponding targets in 24-well plates using Lipofectamine Plus™ reagent (Life Technologies, GibcoBRL) as described by the manufacturer. One hundred nanogram of the FLT plasmid was co-transfected with 12.5, 25, 50 or 100 ng of the shRNA-expressing constructs and 0.2 ng of the luciferase plasmid which was included to normalize for transfection efficiency for each reaction. The synthetic siRNAs were transfected to achieve 10 nanomolar concentrations in all co-transfection experiments. However, 100 nanomolar concentrations of each siRNA were used for transfection in the HCT116-GFPmiR21 cell line. One hundred nanomolar concentrations of the 2′-Me oligonucleotides were also used for transfection in this cell line.
The TRBP down-regulation was achieved by transfecting 40 nM of anti-TRBP siRNAs () (siTRBP-A siTRBP-B and siTRBP1; sequences are available upon request) in a 100 mm plate of 293 cells. Cells were transfected in suspension at ∼50% confluency using the siQuest transfection reagent (Mirus, Madison, WI, USA) as suggested by the manufacturer. The following day cells reached ∼70% confluency and were transfected a second time using the siQuest transfection reagent. Twenty-four hours after the second transfection, cells were lifted, seeded in a 24-well plate, and transfected with the various constructs as described in the Results section. Cells were lysed 24 h after this final transfection and processed for luciferase assays.
A total of 293 cells were seeded in 24-well plates and transfected the next day at 80% confluency. Twenty-four or 48 h post-transfection, the medium was removed and cells were washed once with DPBS (Dulbecco's Phosphate Buffer Saline, Cellgro). One hundred microliter of 1× Passive Lysis Buffer (Promega) was then added to each well. Cells were lysed at room temperature with gentle shaking for 20 min. Ten microliter of cell lysates were assayed for dual luciferase activity according to the manufacturer's instructions (Dual Luciferase Reporter System, Promega). Changes in the levels of the FLT were calculated after normalization against the luciferase units. When the luciferase was used as the target, changes in expression were calculated relative to firefly luciferase (transfection control). The luminescence was determined using a Veritas luminometer (Turner BioSystems, Inc. Sunnyvale, CA, USA). Results are presented in relative light units and the SD is indicated with error bars as shown in the graph data.
Total RNA was isolated using RNA STAT-60 (TEL-TEST B Inc., Friendswood, TX, USA) according to the manufacturer's instructions. Fifteen microgram of total RNA were fractionated in 8 M–6% PAGE, and transferred onto a Hybond-N+ membrane (Amersham Pharmacia Biotech). P-radiolabeled 21-mer probes complementary to the siRNA-antisense sequences were used for the hybridization reactions, which were performed for 16 h at 37°C. A 21-mer DNA oligonucleotide was electrophoresed alongside the RNA samples and used as a size marker and hybridization control (not shown). For TRBP detection, 30 μg RNA was fractionated on a 1% MOPS agarose gel, transferred on a Hybond-N+ membrane and hybridized with a P labeled probe complementary to TRBP overnight at 37°C.
The immunodepletion was carried out by incubating 100 μl of the extract with 30 μl of anti-TRBP antibody () or 30 μl (2 μg/ml) of anti-β-tubulin (Abcam Inc., Cambridge, MA, USA) antibody. Following overnight incubation at 4°C, 30 μl of protein A agarose beads (Upstate USA, Inc. Chicago, IL, USA) were added and incubated for 1.5 h at 4°C with gentle agitation. Following the removal of the beads by a brief centrifugation, 3 μl of the supernatant were resolved in a 10% SDS-PAGE for western blot analyses, and 10 μl were used for the electrophoretic mobility gel shift assay. The pellet beads were boiled with 25 μl of 2× SDS loading buffer and 10 μl were resolved alongside the supernatant sample for western blot analyses. After transferring proteins overnight, the PVDF membranes were blocked with 5% milk in 1× PBS for 2 h at room temperature and washed with 1× PBST for 10 min three times. Primary and secondary antibody incubations were carried out each for 1.5 h in 5% milk in 1× PBST at room temperature followed by washing with 1× PBST for 10 min three times. The membranes were rinsed with 1× PBS for 5 min before drying. The secondary antibody labeled with a 700 channel dye, anti-rabbit-680 nm, was used to detect rabbit polyclonal anti-TRBP Ab672 and β-tubulin.
For each reaction, 10 μl of the TRBP or β-tubulin immunodepleted cell extract was incubated with 10 fmol of 5′ end-labeled siRNA for 30 min at room temperature. The samples were mixed with 2× native gel loading dye and resolved in a 5% non-denaturing polyacrylamide gel for 3 h at 200 V at 4°C. The HEK 293 total cell extract and the tubulin immunodepleted cell extract were used as controls.
The human T cell line CEM was maintained in RPMI 1640 medium supplemented with 10% FBS. A total of 1 × 10 transduced CEM T cells were infected with HIV-1 strain IIIB at an MOI of 0.01. After overnight incubation, cells were washed three times with Hanks’ balanced salts solution and cultured in medium with R10 (RPMI 1640 plus 10% FBS). At designated time points, culture supernatants were collected for p24 analyses (HIV-1 p24 antigen EIA kit, Beckman Coulter Inc.).
We have been interested in investigating the possibility that in the presence of multiple interfering RNAs, the saturation of RNAi components could diminish the activity of a particular sh or siRNA and generate misleading results. Most importantly, the degree of saturation could be such as to interfere with the endogeneous microRNA pathway, thus disrupting potentially critical regulatory functions in the cell. To test this possibility, we selected a series of shRNAs (shL, shVif, shSII, shTAT, shSI) and co-transfected each with a FLT and an anti-firefly-luciferase shRNA (Luc), which routinely generates ∼90% down-regulation of the target (a). Each shRNA was also co-transfected alone with the luciferase target (constructs 2–6, a) to monitor any non-specific reduction or increase of luciferase expression. The results clearly show that there is a direct competition among non-related shRNAs (a). The degree of competition varies with the shRNA used in the co-transfection. The down-regulation of Luc was reduced by co-transfected, non-specific shRNAs from 10 to 60%, yielding in two cases only 30% down-regulation of the luciferase target (, a Luc+shSI and Luc+shTAT) without any targeting of RNAi components and/or presumably other cellular genes. Northern analyses of total RNA from cells transfected with these shRNA constructs did not show a correlation between the degree of competition and their intracellular expression levels (, compare a and b). Thus, competition must be dictated by the ability of each shRNA to be processed into siRNA and incorporated into RISC. Many parameters have been analyzed in the literature that can improve shRNA designs and possibly their incorporation into RISC (). However, none of the popular rules for the optimal shRNA design can explain the different levels of competition exerted by each shRNA with the Luc shRNA construct (Supplementary Data).
Although the relative intracellular expression of each shRNA could not by itself explain the degree of competition for each of the shRNAs, the total amount of shRNA present in the cell can render Exportin 5 rate limiting and affect the overall shRNA activity. To test this possibility, we co-transfected low concentrations (either 12.5 or 50 ng) of three irrelevant shRNA expression plasmids (shL, shSII and shSI) with equal amounts of the Luc shRNA construct and its luciferase target. As expected, the higher amount of shRNAs present in the cell increased the competition while keeping the relative strength of each shRNA as a competitor unchanged (a). The Luc shRNA showed the same ability to down-regulate the target when 12.5, 25 or 50 total ng were transfected with the luciferase target (a, Luc-12.5, Luc-25, Luc-50). A double-stranded RNA stem-loop expressing construct (a mutated HIV TAR element that retains its structure) was also co-transfected with the Luc shRNA and the luciferase target to show that the competition is specific to shRNAs. The U6 TAR element is highly expressed in cells (not shown) yielding levels of cellular transcripts that are comparable to any of the shRNAs tested in these experiments. As shown in a, the TAR element did not significantly affect the efficacy of the Luc shRNA even when transfected at a 1:3 ratio (50 ng of Luc and 150 ng TAR). To test if the competition persists when the shRNAs are present in a linear range we performed a dose response curve using the Luc shRNA and the shSII as a competitor (b). We used 1, 3, 5, 12.5 and 25 ng of shLuc in a luciferase down-regulation assay to select the lowest amount of shRNA that still achieved optimal target down-regulation (b). Based on these results, we co-tranfected 5 ng of shLuc with increasing amounts of the shSII competitor. The results show that even 1 ng of added competitor is able to reduce the activity of shLuc (Luc-5 + shSII-1, b).
The observation that the shRNA competition increases with the total amount of transfected shRNA could be explained by the increasing saturation of Exportin 5, although the different relative strengths of each shRNA to compete against the shLuc when used in equal amounts suggest that they are more likely due to a preferential incorporation into RISC. Moreover, the shRNA can also compete when used in non-saturating conditions (b). However, to bypass Exportin 5 we repeated the co-transfection experiment using synthetic siRNAs (siH3, siH6 and siH1; a). The luciferase target was transfected alone or with each synthetic siRNA (a). The results of this experiment show that the synthetic siRNAs were able to affect the down-regulation of the luciferase target when co-transfected with the Luc construct. Similar to what was shown with the shRNAs, the siRNAs competed to differing degrees with the Luc shRNA. Since synthetic siRNAs do not need to be transported from the nucleus to the cytoplasm for their function and their activity is not affected by Exportin-5 (), the observed competition must be at the level of incorporation into RISC. Although a recent report shows that synthetic siRNAs can potentially shuttle from the cytoplasm back into the nucleus using Exportin-5 (), this should not be a major competition pathway for synthetic siRNAs. Therefore, these results show that sh and siRNAs compete at the level of incorporation into RISC but may also compete for the Exportin 5 pathway to some extent.
To characterize any contribution made by the saturation of Exportin-5 on the competition detected among interfering RNAs, we included a plasmid-expressing Exportin-5 in the co-transfection experiments. It has previously been shown that co-transfecting Exportin-5 relieves the saturation of transport and improves shRNA cleavage of its target (,). As expected, our results show that the overall activity of shLuc is improved, but competition among the different sequences still persists (b). It should be noted that for siRNAs, the effect could be indirect, and based only on more efficient export of shLuc.
Our results show that when different exogenously supplied interfering RNAs are introduced into cells, competition for export and incorporation into RISC can decrease or completely block the activity of some RNAi triggers. These findings suggest that transfected siRNAs and shRNAs can also compete with endogenous microRNAs. To test this hypothesis, we inserted the siRNA directed against the site II HIV--1 sequence (SII) into the backbone of the mir-30 microRNA. We made two different versions of this microRNA, one containing the entire wild-type backbone (Mir-B, and the other with some additional mutations designed to increase the stability of the microRNA structure (Mir-CG). Both microRNAs are processed (c and data not shown) and are effective in down-regulating a RLT which contains the corresponding complementary target sequence within its 3′ UTR (Mir-B, Mir-CG; a). When these microRNAs were co-transfected with an irrelevant shRNA-expressing construct (shSI), their inhibitory activity was dramatically reduced (Mir-B+shSI, Mir-CG+shSI; a). Conversely, the microRNA constructs did not exert significant inhibition on the activity of shRNAs as shown by co-transfecting Mir-B and Mir-CG with a U6 shRNA targeting the firefly luciferase gene (compare Luc to Luc+Mir-B, Luc+Mir-CG; b). Importantly, the guide sequence expressed by the microRNA becomes a good competitor when it is expressed in the form of a shRNA (shSII, a).
Finally, when Mir-B and Mir-CG were co-transfected with an irrelevant microRNA-expressing construct (Mir-Irr), there was no significant change in their ability to down-regulate their target (Mir-B+Mir-Irr, Mir-CG+Mir-Irr; a), indicating that perhaps the safest method for down-regulation of cellular targets is to express the siRNA of choice within a microRNA backbone.
To test if the differences among sh and microRNAs in their ability to compete resides primarily with their levels of expression, we repeated the competition experiment using concentrations that would generate comparable amounts of mature sequences as determined by Northern blot analysis (c). A 4-fold excess of mirCG over shSII was used for the northern blot. Based on the differences in the processed SII sequence, which is expressed from both constructs, a 100-fold excess of mirCG over shSII was used in a competition assay (c). The data show that a siRNA sequence is not an effective competitor when expressed from a microRNA and yet it is very proficient in down-regulating its target even when transfected at lower concentrations (a).
To confirm that si and shRNAs are able to interfere with the microRNA regulatory activities in an intracellular environment, we constructed a stable clonal cell line with an integrated EGFP gene that includes the target site for mir21 at its 3′ end (HCT116-GFPmiR21). In this cell line, the EGFP gene is silenced by the endogenous expression of the mir21 microRNA. Any inhibition with the processing, transport and/or incorporation of mir21 would cause a reactivation of EGFP expression. It has previously been shown that 2′-Me-oligonucleotides (Antagomirs) can block microRNA activity when designed to bind the microRNAs’ guide sequences (,). As expected, transfection of an Antagomir fully complementary to the mir21 guide sequence in the HCT116-GFPmiR21 cells reactivated EGFP fluorescence by blocking the activity of the endogenous mir21 (Mir-21–2′ -Me, d), while an irrelevant Antagomir used as control did not have any effect on the level of fluorescence (IRR-2′-Me, d). Both the non-specific sh and siRNAs were able to increase EGFP fluorescence, demonstrating that these constructs can interfere with the endogenous microRNA pathway (shSII, siH1, d). However, when the same siRNA guide sequence was expressed from a microRNA backbone, it did not affect the mir21 activity, as shown by the lack of EGFP fluorescence (Mir-CG, d). Two additional microRNA constructs were transfected in HCT116-GFPmiR21 with the same experimental outcome (data not shown).
To confirm that competition can alter the efficacy of shRNAs when used in a combinatorial fashion, we tested the anti HIV inhibitory activity of shSII alone (shSII) or in combination with two other shRNAs (shSII+2). CEM cells were stably transduced with a single shRNA or with three different shRNAs each independently expressed from a separate U6 promoter and then infected with HIV IIIB. The results from these experiments show that shSII results in potent inhibition of HIV replication whereas when combined with two additional shRNAs provide protection for the first seven days, but viral replication breaks through after one week (a). The loss of inhibitory activity by the triple construct was not due to loss of expression, since the levels of siRNAs were comparable for the single shRNA and triple constructs (b). The most plausible explanation for the viral breakthrough in the triple construct expressing cells is competition among the shRNAs/siRNAs for export and incorporation into RISC, allowing the virus to escape RNAi surveillance and replicate.
Our experiments show that each siRNA sequence competes to a different extent with other siRNAs and shRNAs. Incubation of P end-labeled siRNAs in HEK293 cytoplasmic extracts results in the formation of a gel-shifted complex using native polyacrylamide gel electrophoresis (a, lane 2). The amount of binding in this assay is directly proportional to the intracellular efficacy of the siRNAs analyzed (Sakurai ., unpublished data). We speculated that TRBP, a double-stranded RNA-binding protein that has been shown to be part of the minimal RISC (,), could contribute to the loading of siRNA guide sequences into RISC and that it would be part of this complex. Depletion of TRBP from the cellular extracts (anti-TRBP) resulted in a marked reduction of siRNA complex formation (a, lanes 4), while depletion of β-tubulin (anti-β-tubulin), which was used as control, had no effect (a, compare lanes 2 and 3). Identical results were obtained when depletion for β-actin was used as control (data not shown). Western analyses of the extracts showed that depletion of both TRBP and β-tubulin were nearly complete (b). However, the small amount of residual gel-shifted complex may reflect incomplete removal of TRBP or could be due to binding of the siRNAs by other components of RISC. We repeated our competition experiment using cells in which TRBP levels were reduced by siRNA-mediated targeting of the TRBP2 mRNA (see Materials and Methods section). TRBP levels were reduced by ∼80% as determined by western (c) and northern (data not shown) analyses. Depletion of TRBP resulted in more equivalent competition by each shRNA sequence (shL, shVif, shSII, shTAT, shSI) or siRNA sequence (siH3, siH6, siH1), but competition with shLuc still persisted (d).
We have shown competition between siRNAs and shRNAs when they are co-transfected into cultured cells. This competition is partially dependent upon saturation of Exportin-5 by the shRNAs and is largely sequence independent. The activity of a luciferase-targeted si or shRNA can be compromised by the ability of another siRNA to be preferentially loaded into RISC.
The competition among shRNAs raises some important issues. Our data show that both siRNAs and shRNAs can compete with the endogenous microRNA pathway (d) thus posing the potential for perturbation of endogenous cellular miRNA processes. In a recent publication, Grimm . () reported that some shRNA sequences were able to block the activity of the liver-specific miR122 and additionally caused severe liver toxicity and death of some of the treated mice. The authors attributed this adverse effect to saturation of Exportin-5, postulating that despite being expressed from the same AAV vector backbone, some sequences were expressed at higher levels and were more toxic. Our results demonstrate that competition with endogenous microRNAs can be a general phenomenon and that Exportin-5 is only one of the components involved, as shown by the inability to fully restore the U6sh-luciferase activity when Exportin-5 was ectopically expressed (b). Moreover, synthetic siRNAs, which do not rely on Exportin-5 for their transport, can still affect the activity of both shRNAs (a) and microRNAs (d). These results indicate that the ability of a specific sequence to compete is also determined by its ability to be loaded into RISC. Interestingly, the relative competitive strength of the synthetic siRNAs with the Luc shRNA correlated with their efficacy to down-regulate their cognate targets, which in turn is directly proportional to the complex formation in cell extracts (Sakurai ., unpublished data). These results are consistent with a previously published study () in which competition was detected between heavily modified and unmodified synthetic siRNAs. The competition was found to correlate with the potency of the siRNAs and it was attributed to a possible saturation of the export system or competition for other RISC components.
The affinity of each sequence for one or more RNAi component(s) may determine the efficiency of loading of that sequence into RISC, thereby potentiating the competition with other sh/siRNAs and miRNAs. TRBP is a double-stranded RNA-binding protein that has been shown to associate with both Dicer and Ago-2 (,). It has been proposed that TRBP couples the initiation and executions steps of RNAi () and that it identifies the guide strand prior for passage on to Ago-2 (). Our data reveal that cellular down-regulation of TRBP results in loss of differential competition by the various sh/siRNAs, although competition still persists (d), perhaps via a bypass mechanism for incorporation into the Ago2-binding domain. Our results support the role of TRBP as a sensor for the RISC loading or Ago-2 handoff (). However, it is possible that TRBP does not act alone and loading may include other RISC proteins.
In our experimental system, sequences expressed from a microRNA backbone are poor competitors relative to shRNA or siRNAs of the same sequences (). Two different sequences were expressed from three different microRNA backbones and none of them was able to compete ( and data not shown) despite the fact that these sequences were still effective in RNAi. The overall level of shRNA expression does not seem to be the primary reason for this lack of competition (c), which is probably determined by the overall kinetics of incorporation. Since miRNAs are being shunted through the endogenous miRNA-processing pathway, this may slow their entry into RISC relative to highly expressed shRNAs and transfected siRNAs.
Si and sh-RNAs are routinely used to down-regulate cellular genes, siRNA libraries have been widely employed to screen for gene function (), and siRNAs are already in human clinical trials for wet adult macular degeneration, respiratory syncitial virus (SRV) infection and acute kidney failure. The potential for competition of the clinically applied siRNAs with endogenous miRNAs is a concern as is the competition among combinations of si/shRNAs. Due to the rapid emergence of resistance mutations in viruses such as HCV and HIV, the use of combinations of sh/siRNAs represents a potential mechanism for circumventing this problem. If one of the siRNAs is more efficiently loaded into RISC, it could compromise the effectiveness of one or more of the other siRNAs. For example, our data clearly show that an effective shRNA targeted against the HIV-1 gene, when combined with two other hairpins, no longer provides stable, long-term inhibition of HIV replication (). This result is consistent with previous finding by Nishitsuji . (), which showed a negative effect on shRNA antiviral activities when used in a combinatorial fashion. However, based upon our findings, concern about obstructing the endogenous RNAi machinery can potentially be overcome by modifying the siRNA expression system to produce a microRNA, or by combining si and shRNAs that do not compete with one another. By finding siRNA/target combinations that work at the lowest possible concentrations, it should be possible to mitigate the potential for competition with endogenous miRNAs or among combinations of applied si/shRNAs.
p
p
l
e
m
e
n
t
a
r
y
D
a
t
a
a
r
e
a
v
a
i
l
a
b
l
e
a
t
N
A
R
O
n
l
i
n
e
. |
A potential site in DNA for the interaction of genotoxic species is the phosphodiester linkages between the 2′-deoxynucleosides, which constitute the sugar-phosphate backbone of DNA, resulting in esterification of the phosphate group and formation of phosphotriester adducts. The properties of phosphotriester adducts have been extensively studied using simple alkylating agents as model compounds showing that they represent long lived biomarkers of exposure (). In contrast the potential of polycyclic aromatic hydrocarbons (PAHs) to form phosphodiester adducts with 2′-deoxynucleotides as well as phosphotriester adducts in DNA has not been clearly ascertained ().
Studies involving human fibroblast cells and rodents have shown that alkyl phosphotriester adducts are more stable to DNA repair when compared to base adducts (). For alkylating agents it has been shown that the relative abundance of phosphotriester DNA adducts formed depends on the chemical nature of the genotoxic species (,,). The biological consequences of alkyl phosphotriester adducts is not fully understood, though they are chemically stable under physiological conditions and may potentially alter the binding/function of proteins such as DNA repair or replication enzymes (,). To date, the compounds investigated for the formation of phosphodiester or phosphotriester adducts include alkylating agents, such as dialkylsulphates, alkyl methanesulphonates and -nitroso compounds cyanoethylene oxide, cyclophosphamide and phenyl glycidyl ether ().
PAHs of which benzo[a]pyrene (B[a]P) is a well-studied example, represent an important class of compounds found as ubiquitous environmental pollutants or in certain occupational settings (,). PAHs are produced by the incomplete combustion of fossil fuels and the chargrilling of food as well as being present in automobile exhaust and cigarette smoke (). PAHs have been shown to be carcinogenic in animals and potentially carcinogenic to humans (). Benzo[a]pyrene-7,8-dihydrodiol-9,10-epoxide (B[a]PDE) is the reactive species formed by the cytochrome P450 mediated metabolism of B[a]P (,). The reaction of B[a]PDE via a B[a]Ptriol carbocation intermediate with purine and pyrimidine bases present in DNA has been well characterized, the predominant product formed is by reaction with the exocyclic amino group of guanine and to a lesser extent with the exocyclic amino groups of adenine and cytosine (). The base adducts formed by B[a]PDE exist as diastereoisomers following or addition at C-10 of the hydrocarbon and studies show that the (+)-anti-B[a]PDE isomer with the 7,8,9,10 configuration has greatest carcinogenic activity (). B[a]PDE produces concentration-dependent strand breaks in DNA with the fragmentation of the DNA being attributable to the formation of a phosphotriester adduct rather than a base adduct. A mechanism for DNA strand scission has been proposed that involves the C-9 hydroxyl group of B[a]PDE attacking the phosphotriester group and the formation of a cyclic triester intermediate as shown in ().
The technique of collision-induced dissociation (CID) tandem mass spectrometry provides structural information, which is important since both the base and phosphodiester adducted 2′-deoxynucleotides have the same molecular mass. This approach has been used to characterize the phosphodiester adducts formed by phenyl glycidal ether and ethylating agents (,). We investigated the formation of phosphodiester adducts resulting from the reaction of B[a]PDE with 2′-deoxynucleotides using negative electrospray ionization tandem mass spectrometry CID.
B[a]PDE .
The four 2′-deoxynucleosides 3′-monophosphates (1 mg dissolved in 750 µl of 0.1 M TRIS base, pH 7.0) were each incubated with B[a]PDE (250 µg; 1 μg/µl dissolved in methanol) for 18 h at 37°C. Unreacted starting material containing 2′-deoxynucleosides 3′-monophosphates was removed by subjecting the reaction mixtures to solid phase extraction using Oasis HLB columns (1 cc, 30 mg, Waters Ltd, Elstree, UK) connected to a vacuum manifold (Phenomenex, Macclesfield, UK) maintained at a vacuum of 5 mmHg. The columns were initially conditioned with 1.0 ml of methanol followed by 1.0 ml of HPLC grade water. The reaction mixtures were then loaded onto the columns and washed with 1.0 ml of 5 : 95 methanol/water (v/v). The reaction products were eluted from the columns with 2.0 ml of methanol. Samples were dried to completeness using a centrifugal vacuum evaporator (Speedvac, Savant, Farmingdale, NY, USA) and resuspended in 400 μl of 10 : 90 acetonitrile/HPLC grade water (v/v). Samples (20–50 μl aliquots) were analysed by HPLC-fluorescence detection using a Waters 2690 Separations Module coupled to a Waters 470 fluorescence detector (excitation 332 nm, emission 388 nm) connected to a Phenomenex Synergi Fusion-RP 80 C column (4 µm, 250 × 4.6 mm) and a Synergi Fusion-RP 80 C (4 µm, 4 × 3.0 mm) guard column. The column was eluted using a gradient of 0.05 M potassium phosphate buffer, pH 7.2 (solvent A) and acetonitrile (solvent B) at a flow of 1.0 ml/min with 0 min–10%B, 60 min–40%B, 65 min–80%B, 67 min–10% B, 70 min–10%B.
Collected HPLC product fractions were pooled together, evaporated to dryness and then dissolved in 1.0 ml of 5 : 95 methanol/water (v/v). The purified HPLC fractions were subjected to solid phase extraction using Oasis HLB columns (1 cc, 30 mg) connected to a vacuum manifold maintained at a vacuum of 5 mmHg. The columns were initially conditioned with 1.0 ml of methanol followed by 1.0 ml of HPLC grade water. The purified HPLC fractions were then loaded onto the columns and washed with 1.0 mL of HPLC grade water. The reaction products were eluted from the columns with two 750 μl aliquots of methanol, evaporated to dryness using a centrifugal vacuum evaporator and redissolved in 400 μl of methanol/HPLC grade water (20:80, v/v).
The mass spectrometer was tuned by using a dGp standard solution (10 pmol/μl) dissolved in methanol/HPLC grade water (45:65, v/v) introduced by continuous infusion at a flow rate of 10 μl/min with a Harvard model 22 syringe pump (Havard Apparatus Ltd, Edenbridge, UK). Initial analysis of the reaction products was performed using continuous infusion and full scan negative ESI-MS over the / range from 60 to 800, following a 1:10 dilution with methanol/HPLC grade water (45:65, v/v) of each of the purified reaction products.
A 20 μl undiluted aliquot of each purified reaction product was injected onto a HyPurity C (3 μm, 150 × 2.1 mm) column (Thermo Electron Corporation, Runcorn, UK) connected to a Uniguard HyPurity C (3 μm, 10 × 2.1 mm) guard cartridge attached to KrudKatcher (Phenomenex) disposable pre-column (5 μm) filter. The column was eluted isocratically with solvent A, methanol/HPLC grade water (45:65, v/v) at a flow rate of 120 μl/min for 45 min. It was then washed with solvent B, methanol at a flow rate of 200 μl/min for 10 min and then equilibrated to starting conditions with solvent A at a flow rate of 120 μl/min for 15 min. The collision gas was argon (indicated cell pressure 3.0–3.5 × 10 mbar) and the collision energy set at 21 eV. The dwell time was set to 200 ms and the resolution was one / unit at peak base. The samples were analysed in negative electrospray ionization (ESI) mode MS/MS CID for the deprotonated molecular ion [M−H] for each B[a]PDE adducted 2′-deoxynucleotide: 2′-deoxyguanosine 3′-monophosphate (dGp) [CHNOP-H]
/ 648.15; 2-deoxyadenosine 3′-monophosphate (dAp) [CHNOP−H]
/ 632.16; 2′-deoxycytidine 3′-monophosphate (dCp) [CHNOP-H] 608.14 and thymidine 3′-monophosphate (Tp) [CHNOP-H]
/ 623.14. The mass spectral data was acquired in continuum mode and processed using MassLynx version 4.0 (Micromass, Waters Ltd).
The reaction mixtures for the four different 2′-deoxynucleotides and B[a]PDE were initially subjected to solid phase extraction to remove any unreacted 2′-deoxynucleoside 3′-monophosphates, followed by separation using HPLC with fluorescence detection. The typical HPLC-fluorescence chromatogram of a control reaction mixture containing only B[a]PDE and 0.1 M TRIS base pH 7.0 buffer incubated at 37°C for 18 h and subjected to solid phase extraction is shown in . The typical HPLC-fluorescence chromatograms for 2′-deoxynucleotides plus B[a]PDE reaction mixtures are shown in A, A, A and A. Fractions corresponding to the peaks that eluted before retention time 35 min, since these were unique to the reaction mixtures and not present in the control reaction mixture, were collected, pooled and evaporated to dryness and then subjected to a further purification by solid phase extraction (to remove any salts) prior to analysis by continuous infusion full scan negative ESI-MS to determine the presence of the deprotonated molecular [M−H] ion for each adducted 2′-deoxynucleotides. Each fraction was then further characterized using LC-ESI-MS/MS CID.
For each 2′-deoxynucleotide, the phophodiester adducts eluted before the corresponding base adducts following analysis by HPLC-fluorescence and then by LC-ESI-MS/MS CID. Typically, the LC-ESI-MS/MS retention times ranged from 7 to 12 min using the microbore C column (data not shown). The LC-ESI-MS/MS CID product ion spectra for each adducted 2′-deoxynucleotide are shown in B, , B and B. Two distinct product ions at / 399 and 497 were observed corresponding to the [(B[a]Ptriol+phosphate)−H] and [(2′-deoxyribose+phosphate+B[a]Ptriol)−H], respectively, resulting from the fragmentation of the precursor molecular ion of each of the four adducted 2′-deoxynucleotides. The ion at / 381 corresponds to the loss of HO from the ion at / 399. These results indicated that the B[a]PDE modification was present on the phosphate moiety of the 2′-deoxynucleotide resulting in the formation of a phosphodiester adduct. Further product ions were observed corresponding to / 79 [PO], / 97 [HPO], / 195 [(2′-deoxyribose+phosphate)−H] that were common to the spectra for all four adducted 2′-deoxynucleotides. The spectra also contained product ions that correspond to the unadducted 2′-deoxynucleoside 3′-monophosphate and base at / 346 [dGp−H], / 150 [guanine−H], / 330 [dAp−H]; / 134 [adenine−H], 306 [dCp−H]; / 110 [cytosine−H] and / 321 [Tp−H]; / 125 [thymine−H]. The fragmentation pathway of phosphodiester adducted 2′-deoxynucleotides is shown in . It was noted that there was a complete absence of products ions corresponding to the adducted bases in the spectra.
For each 2′-deoxynucleotide, the base adducts eluted after the phophodiester adducts following analysis by HPLC-fluorescence and then by LC-ESI-MS/MS CID. Typically, the LC-ESI-MS/MS retention times ranged from 13 to 28 min using the microbore C column (data not shown). The LC-ESI-MS/MS CID product ion spectra for each adducted 2′-deoxynucleotide are shown in C, C and C showing adduct formation with the base. Product ions at / 452 and 550 were observed resulting from the fragmentation of the B[a]PDE adducted dGp precursor molecular ion at / 648 [M−H] (C), corresponding to [(B[a]Ptriol+guanine)−H] following cleavage of the glycosidic bond and [(B[a]Ptriol+2′-deoxyguanosine)−H] following loss of one HO molecule plus the phosphate group, respectively. The ions at / 434 and 416 correspond to the loss of one and two HO molecules from the ion at / 452, respectively. Similarly, the ions at / 532 and 514 correspond to the loss of one and two HO molecules from the ion at / 550, respectively. These results imply the presence of a B[a]PDE modification on the guanine base of dGp. A product ion at / 436 was observed resulting from the fragmentation of the B[a]PDE adducted dAp precursor molecular ion at / 632 [M−H] (C) corresponding to [(B[a]Ptriol+adenine)−H] following cleavage of the glycosidic bond. The ions at / 418 and 400 correspond to the loss of one and two HO molecules from the ion at / 436, respectively. These results imply the presence of a B[a]PDE modification on the adenine base of dAp. A product ion at / 412 was observed resulting from the fragmentation of the B[a]PDE adducted dCp precursor molecular ion at / 608 [M−H] (C) corresponding to [(B[a]Ptriol+cytosine)−H] following cleavage of the glycosidic bond. The ions at / 394 and 376 correspond to the loss of one and two HO molecules from the ion at / 412, respectively. These results imply the presence of a B[a]PDE modification on the cytosine base of dCp. The fragmentation pathway of base adducted 2′-deoxynucleotides is shown in . It was noted that there was a complete absence of products ions characteristic of an adducted phosphate group. All three spectra for the base adducted 2′-deoxynucleoside 3′-monophosphates contained a product ion at / 283 corresponding to B[a]Pdiol.
Confirmation that the structural identity of the base adducts was due to the reaction of B[a]PDE with the exocyclic –NH group was ascertained following analysis by positive ESI. The LC-ESI-MS/MS CID product ion spectra for each base adducted 2′-deoxynucleotide are shown in . The spectrum for each base adducted 2′-deoxynucleotide contained common product ions. The ion at / 303 corresponds to B[a]Ptriol. The ion at / 285 corresponds to B[a]Pdiol following loss of HO from the ion at / 303. The ion observed at / 257 corresponds to the loss of CO from the ion at / 285. The spectra also contained product ions that correspond to the unadducted 2′-deoxynucleoside 3′-monophosphate and base at / 348 [dGp+H]; / 152 [guanine+H], / 332 [dAp+H]
/ 136 [adenine+H], and / 308 [dCp+H]; / 112 [cytosine+H]. Product ions were observed corresponding to the adducted base following cleavage of the glycosidic bond at / 454 [(B[a]Ptriol+guanine)+H], at / 438 [(B[a]Ptriol+adenine)+H] and at / 414 [B[a]Ptriol+cytosine+H]. No product ions resulting from the neutral loss of 17 u corresponding to NH were observed, thus confirming that B[a]PDE adduct formation for each 2′-deoxynucleotides was by reaction with the exocyclic –NH group at position N for guanine, N for adenine and N for cytosine.
The results of this investigation provide the first direct evidence for the formation of phosphodiester adducts with 2′-deoxynucleotides that are normally present in DNA by the reactive PAH epoxide metabolite, B[a]PDE, thus providing supporting evidence to the supposition that B[a]PDE can react with the sugar-phosphate backbone of DNA resulting in the formation of phosphotriester adducts. Koreeda . were the first to provide indirect evidence that B[a]PDE may form phosphotriester adducts as well as base adducts following their experiments involving the reaction of tritium labelled B[a]PDE with polyguanylic acid (). Chan and Raddo hypothesized that the formation of phosphotriesters from the reaction of B[a]PDE with DNA was chemically feasible using initial model experiments investigating the formation of phosphotriesters following the reaction of cyclohexene oxides with dibenzyl or diethyl phosphates (). Subsequent experiments by the authors demonstrated that various PAH epoxides as well as B[a]PDE reacted with dibenzyl or diethyl phosphates resulting in the formation of phosphotriester adducts in a regio- and stereo-specific manner as determined by NMR spectroscopy (). We investigated the reaction of B[a]PDE with 2′-deoxynucleotides, which resulted in the formation of a number of additional early eluting product peaks when compared to a control reaction mixture containing only the B[a]PDE and buffer following HPLC-fluorescence analysis, which were further characterized using LC-ESI-MS/MS CID. The peaks present in the control reaction mixture () represent breakdown products such as B[a]Ptetraols following hydrolysis of the B[a]PDE. The direct HPLC-fluorescence analysis of (±)-anti-B[a]PDE following hydrolysis in aqueous methanol resulted in the detection of two major peaks (corresponding to the peaks labelled with an asterisk in ) and two later eluting minor peaks (data not shown). We hypothesize that the other later eluting peaks observed may be impurities derived from the interaction of B[a]PDE with the TRIS base reaction buffer. Previous studies have shown that hydrolysis of B[a]PDE in aqueous solutions at neutral pH results in the formation of two B[a]Ptetraol products with and configurations (). It was noted that each HPLC fraction analysed contained several product peaks, the reason being that each peak corresponded to the different stereoisomers of adducts resulting from the or addition at C-10 of the (±)-anti-B[a]PDE (). The interaction of the (+)- or (−)-anti-enantiomers of B[a]PDE with either 2′-deoxyguanosine or 2′-deoxyadenosine will result in the formation of two pairs of stereoisomer adducts, namely, (+)-anti--, (+)-anti--, (−)-anti-- and (−)-anti-- following reaction with the exocyclic amino group of the base (). Furthermore, the phosphate of the phosphotriester group is chiral with adduct formation resulting in two configurations, R and S (). Using LC-ESI-MS/MS CID reaction products were differentiated having a unique CID product ion spectra characteristic of a phosphodiester adduct. The presence of product ions at / 399 and 497 was observed for all four 2′-deoxynucleotides, corresponding to [(B[a]Ptriol+phosphate)−H] and [(2′-deoxyribose+phosphate+B[a]Ptriol)−H], respectively. There was an absence of product ions in the spectra corresponding to B[a]PDE adducted bases. For the three B[a]PDE and 2′-deoxynucleotide reaction mixtures the phosphodiester adducts eluted before the corresponding base adducts following HPLC-fluorescence analysis. Canella . noted the presence of uncharacterized products in their investigation of the synthesis of base adducts by the reaction of B[a]PDE with 2′-deoxynucleotides. Two early eluting products were detected following HPLC-UV analysis of a B[a]PDE plus 2′-deoxyadenosine 5′-monophosphate reaction mixture that were resistant to hydrolysis by alkaline phosphatase unlike the base adducts (). The identity of these products was not elucidated but it was concluded that their structure contained a nucleotide component that required the presence of 2′-deoxynucleotides in the B[a]PDE reaction mixture for their formation (). It is feasible to assume that these unidentified products were phosphodiester adducts.
Further reaction products were identified having CID product ion spectra characteristic of adduct formation with the bases of the 2′-deoxynucleotides resulting from cleavage of the glycosidic bond, with prominent product ions at / 452, 436 and 412 [(B[a]Ptriol+base)−H] corresponding to B[a]PDE adducts of guanine, adenine and cytosine, respectively. Consistent with previous reports in the literature B[a]PDE adduct formation with the base was confined to dGp, dAp and dCp (,,). Confirmation that the reaction with exocyclic –NH group of each base had occurred was obtained following positive ESI LC-MS/MS analysis of the reaction products. The CID product ion spectra revealed that no product ions resulting from the neutral loss of 17 u corresponding to NH, were observed. Thus, confirming that B[a]PDE adduct formation for each 2′-deoxynucleotide was by reaction with the exocyclic –NH group at position N for guanine, N for adenine and N for cytosine. For example, the CID product ion spectra of N-7 alkyl guanine adducts such as those derived from ethylating agents contain a product ion corresponding to the neutral loss of NH from the alkylated [M+H] ion (,). No base adducts of Tp were detected which is again consistent with previous findings. Due to later elution of the phosphodiester adducts of Tp following HPLC-fluorescence analysis (A) compared to the other three 2′-deoxynucleotides, any base adducts formed may have eluted in the region of the hydrolyzed break down products of B[a]DPDE. Studies have shown that the stereochemistry of B[a]PDE influences the extent of formation of the different base adducts following the reaction with DNA. The (+)-anti-B[a]PDE isomer with the 7,8,9,10 configuration has the highest reactivity resulting in 95% guanine and 5% adenine of total adducts being formed (,). The results described in the present study do not allow for the estimation of the extent B[a]PDE adduct formation with the phosphate group relative to the base since the HPLC-fluorescence detector response may be significantly different for the phosphodiester adducts compared to the base adducts.
The formation of B[a]PDE phosphotriester adducts in DNA has the potential for producing more profound adverse biological consequences when compared to the stable alkyl phosphotriesters adducts formed by simple alkylating agents. This is due to the presence of an oxygen atom in the β position with respect to the ester group that can lead to the alteration in the integrity of the DNA via mechanism of internal oxygen nucleophilic displacement at the phosphate atom, resulting in strand scission (,). The confirmation of the formation of B[a]PDE phosphodiester adducts should allow for the investigation of methodology for the detection of B[a]PDE phosphotriester adducts in DNA. The first methods developed to detect alkyl phosphotriester adducts in DNA relied on the production of non-specific alkaline hydrolysis-induced strand breaks at the site of the phosphotriester adduct (). Subsequent approaches to detect alkyl phosphotriester adducts in DNA relied on P postlabelling techniques (,,,). In the last few years, liquid chromatography coupled to mass spectrometry has been increasingly used for not only for the characterization of DNA adducts but also their detection (). Recently, LC-MS/MS methods have been described for the detection of alkyl phosphotriester adducts in DNA relying on selective enzymatic digestion (). The transalkylation approach involves the alkyl group of the phosphotriester adduct in di-2′-deoxynucleoside monophosphates, formed following enzymatic digestion of DNA, undergoing nucleophilic displacement by cob(I)alamin. The resulting alkyl-cob(I)alamin product can then be determined by LC-MS/MS (, ). The enzymatic digestion relies on the presence of a phosphotriester adduct in DNA generating di-2′-deoxynucleoside monophosphate adduct triesters, since the internucleotide bonds adjacent to a completely esterified phosphate group are resistant to enzymatic cleavage by nucleases (,). The method described by Haglund . uses enzymatic digestion with nuclease P1 in combination with 5′-phosphodiesterase and alkaline phosphatase for the direct detection of di-2′-deoxynucleoside monophosphate ethyl triesters by online column switching LC-MS/MS (). However, optimization of the enzymatic digestion may be required for the detection of B[a]PDE phosphotriester adducts, since it is found for bulky DNA adducts such as those derived from B[a]PDE, that di-2′-deoxynucleoside monophosphates are generated by the presence of base adducts, which are particularly resistant to enzymatic digestion (,). These B[a]PDE di-2′-deoxynucleoside monophosphate adducts have been observed to occur at a 50 times lower level than the main base adduct formed at the exocyclic –NH group of guanine in DNA ().
In conclusion, the formation of phosphodiester adducts following the reaction of B[a]PDE with 2′-deoxynucleotides has been confirmed, having been identified with distinct CID product ion spectra when compared to base adducted 2′-deoxynucleotides. |
1,4,8,11-Tetraazacyclotetradecane (cyclam)-based macrocycles form complexes of high thermodynamic and kinetic stability with transition metals. N-functionalised derivatives are selective metal ion chelators and are commonly used as the chelating component (known as a bifunctional chelator or BFC) of targeted radiopharmaceuticals in nuclear medicine. A bifunctional chelator contains a reactive site for attachment to a targeting group and, ideally, will fully retain the radiolabel under physiological conditions to allow localisation at the target site without transchelation occurring. We are particularly interested in copper radioisotopes Cu, Cu, Cu and Cu, which are positron emitters with half lives varying from 0.16 to 12.7 h that have been investigated for applications in positron emission tomography (PET) imaging.
Cu is a longer lived β emitter (
= 61.9 h) that has been used in targeted radiotherapy.
The 1,4,8,11-tetraazacyclotetradecane-1,4,8,11-tetraacetic acid (HTETA) chelator, where cyclam has been functionalised with four N-pendant acetate groups, is one of the most common chelator components in BFCs for copper() HTETA–BFC. The addition of pendant carboxylic acid groups further increases stability, forming six-coordinate copper() complexes where two of the pendant arms are coordinated to the metal centre. The charge on the resultant complex can vary with deprotonation of the remaining carboxylic acid pendant arms. We report X-ray crystal structures of the neutral copper() complex in which only the two coordinating arms are deprotonated (HTETA).
However, despite the relatively high stability of these complexes, retention of the radiolabel is still not optimum for medical imaging and therapeutic applications. Hence, we are interested in modified chelators based on the cyclam framework that also incorporate acetate arms but may have increased stability.
Complexes of topologically-constrained aza-macrocycles have improved kinetic stability. Weisman, Wong and Anderson have designed and characterised novel cross-bridged tetraza-macrocyclic chelators (CBTE2A and CBDO2A) and demonstrated increased stability of the radiolabelled complexes in comparison to TETA. For CuCB-TE2A, initial transchelation to ceruloplasmin, followed by transfer to Cu–Zn superoxide dismutase, is significantly reduced. We have recently reported the first BFC based on the CBTE2A framework, HCBTE2A–BFC. Other structurally reinforced macrocycles such as the side-bridged cyclam, with an ethylene bridge between adjacent nitrogens, are also of interest. In this work, we describe the synthesis and structure of copper() complexes of the first side-bridged macrocycle incorporating a coordinating acetate pendant arm.
For such an important chelator in biomedicine, TETA is relatively underrepresented in the Cambridge Structural Database (CSD). The solid state structure of neutral copper TETA complexes has not been investigated. Our main interest in TETA is to gather structural information about the flexibility of this chelator from the X-ray structural parameters and then to correlate this with the known kinetic stability issues.
The two copper TETA structures, reported by Kaden and co-workers, deposited in the CSD are [CuTETA] and CuTETA; both have fully deprotonated TETA chelators. The charge of the [CuTETA] anion is balanced by barium ions and extensive H-bonding is observed with lattice-incorporated water molecules. Two forms of this compound are known with an identical copper-containing anion but varying numbers of water molecules of crystallisation.
The CuTETA structure is of less interest as neither of the two metal ions is incorporated into the four nitrogen donor macrocyclic cavity. The structure of the neutral Cu(HTETA) complex has not been determined previously, although Kaden and co-workers report the isolation of this complex. To produce crystals for structural characterisation of the neutral compound, our initial approach was to follow Kaden's synthetic procedure and grow crystals from water at an appropriate acidic pH to ensure protonation of the two unbound carboxylates. Large blue crystals were produced by dissolving the compound in distilled water and adjusting the pH to 2.0 with hydrochloric acid. These crystals were of suitable quality for X-ray diffraction studies and revealed the expected neutral complex, Cu(HTETA) (A).
An alternative crystallisation technique was attempted by redissolving the isolated blue crystalline compound in acetonitrile. Slow evaporation of the acetonitrile solution resulted in formation of violet crystals that were of sufficient quality for X-ray crystallographic analysis. Once again, a neutral complex of Cu(HTETA) (B) was present in the unit cell. Some interesting structural differences were observed, with significant variation observed in the copper–ligand bond lengths. Both structures show the expected Jahn–Teller distortion around the d copper() centre. However, the orientation of the elongated axis varies; for one complex the increased bond lengths are observed across the macrocycle (N–Cu–N axis) and for the other along the O–Cu–O axis of the coordinated pendant arms, see . As expected, the axial elongation also results in compression of the four equatorial bonds. This demonstrates significant flexibility of the cyclam ring in the TETA ligand complex with copper().
The Cu(HTETA) (A) complex crystallised from water at pH 2.0 exhibits a distorted octahedral geometry with two weakly coordinating axial carboxylates and the four macrocyclic nitrogens in the equatorial positions, . This structure has both six-membered CuNC chelate rings in the chair conformation and the four nitrogens coordinated in a planar arrangement. There is crystallographic symmetry within the molecule and the copper atom lies on an inversion centre. The Cu–N1 bond lengths are 2.06 Å and Cu–N2 bond lengths are 2.16 Å, Jahn–Teller distortion is observed through the O–Cu–O axis with bond lengths of 2.27 Å, .
Similarly, the ‘ring-distorted’ compound, Cu(HTETA) (B), has the two six membered CuNC chelate rings in the chair conformation and the four nitrogens coordinated in a planar arrangement, see . Again, crystallographic symmetry is present within the neutral complex and the copper atom lies on an inversion centre. The Cu–O bond lengths are 2.020 Å and Cu–N1 bond lengths 2.002 Å. Jahn–Teller distortion is observed through the N2–Cu–N2 axis with bond lengths of 2.378 Å. Comparison with the published isomorphous structure of Zn(HTETA) is of interest as Jahn–Teller distortion will not be observed for a d complex. The Zn–O1 bond lengths are 2.13 Å and the Zn–N1 bond lengths are 2.11 Å. However, a smaller elongation is observed in the Zn–N2 bond lengths at 2.25 Å, probably due to a mismatch between the cavity size and the radius of the zinc() ion.
A linked 3D network of hydrogen bonded species involving both the water molecules of crystallisation and the uncoordinated pendant arms is observed for each structure (see Fig. S2 in the ESI). Structure (A) has four water molecules and structure (B) two water molecules present for every complete copper macrocyclic unit. Linear and bifurcated H-bonds are present in both structures.
There is a considerable amount of published data for copper() cyclam complexes going back over more than 30 years. We believe the distortion of the cyclam macrocycle to accommodate the ring-based Jahn–Teller distortion demonstrates flexibility that explains the increased kinetic lability of such complexes Over a hundred X-ray structures are known for copper() cyclam-based complexes and an analysis of all the deposited structures in the CSD was carried out using CONQUEST and VISTA to determine how common this phenomenon was for six-coordinate cyclam complexes with copper(). A parameter was specified for each bond length and the Cu–N ring bond lengths forming the square plane plotted in histogram . A greater range of bond lengths was observed for the non-ring donors as the nature of the donor atom ( N, S, O, P) varied. The distribution of bond lengths from the 118 X-ray crystal structures in the CSD
shows only one other example is known where the elongated axis from the Jahn Teller distortion is across the macrocyclic ring, see .
This other ring-distorted complex is Cu(HTETA–BFC), a neutral complex of the C-functionalised nitrobenzyl derivative, giving only two examples out of 120 structures. The energy difference between the two Jahn Teller distorted variations must be relatively low, as both structures are observed. However, the much higher frequency of occurrence of the non-macrocyclic Jahn–Teller distortion suggests it is the more stable arrangement. It may be that N–Cu–N elongation is only observed in structures where crystal packing forces and the optimisation of H-bonded networks compensate for the energy penalty of adopting this arrangement.
The existence of these two arrangements does suggest that the TETA conformation will be dynamic in solution and this flexibility may be the key factor that increases kinetic lability, accounting for the overall inadequate chelating ability of TETA It is certainly not valid to assume that the Jahn–Teller distortion is static along the axis between the two acetate arms.
The X-ray structure of the configurationally restricted and higher stability CuCBTE2A is known. In this case, the carboxylate oxygens are rather than due to the topological constraint of the macrocycle giving a folded conformation. The axial elongated bonds are present along one of the O–Cu–N axes, , with average values of Cu–O 2.314 and Cu–N 2.235. The more rigid backbone will prevent the macrocycle flexing and it is highly unlikely a cross macrocycle N–Cu–N Jahn–Teller distortion would be observed for this system.
Reduced flexibility is one feature of configurational restraint in bridged cyclam chelators. Due to the potential increase in kinetic stability and favourable results already observed for cross-bridged chelators, side-bridged chelators may also be of use in Cu radiopharmaceutical applications. We have produced the first example of a side-bridged cyclam chelators with a pendant acetate ‘arm’.
The side-bridged cyclam ligand has two adjacent nitrogen atoms linked by an ethylene bridge, forming a piperazine fragment within the macrocyclic ring, . The constraint imparted by the bicyclic structure results in restriction of the metal complex configuration solely to the type configuration (only the - or - forms, as described by Bosnich and co-workers, have been observed). Although interest has been expressed in this chelator class for use in radiopharmaceuticals, it has not been extensively studied.
The first syntheses of side-bridged cyclam ligands by Wainwright produced symmetrical, rigidified macrocycles with two secondary amine sites. The synthetic procedures reported in 1995 by Kolinski enable the mono-functionalisation of these macrocycles to produce asymmetrically-substituted analogues. A modification of this strategy has been used to produce a mono-substituted side-bridged cyclam with a -butyl ester-protected carboxymethyl arm, , . The substitution of -butyl bromoacetate by a tetracyclic bisaminal intermediate occurs efficiently to produce the mono-quaternary ammonium salt . Reductive ring opening with sodium borohydride produces the ethylene bridged cyclam, in total yield from of 85%. Acidic hydrolysis, using either trifluoroacetic acid (TFA) or hydrochloric acid, of the -butyl ester group in gives a carboxymethyl arm in chelator HSBTE1A, analogous to the pendant arms of TETA. It was expected that the carboxymethyl arm would deprotonate and coordinate to a copper() centre bound in the macrocyclic cavity. Complexes of and HSBTE1A with copper() have been produced by reaction with copper() salts at reflux in methanol. Single crystal X-ray structures were determined and show that the pendant arm is coordinated through an oxygen atom in both cases.
Complexes with the ester pendant ligand, , were prepared with copper() chloride or perchlorate. Single crystals of the sample prepared with the chloride counter anion were grown by diffusion of diethyl ether into a methanolic solution. On solving the structure, the counter anion present was found to be copper() trichloride, [CuCl]. The copper() trichloride anion must have been formed during the complexation of compound with copper() dichloride. [CuX] (where X = Cl, Br) has previously been observed in several crystal structures, and has been produced from CuX during complexation.
The [Cu] cation has a five coordinate copper centre in a distorted square-based pyramidal geometry (Addison and Reedijk's parameter shows no distortion towards trigonal bipyramidal, with a value of 0.01),
. The four ring nitrogens form a distorted square plane with Cu–N bond lengths in the range 1.980 to 2.063 Å. The side-bridged cyclam ring is in the - configuration. There is a longer bond to the ester carbonyl oxygen in the apical position at 2.414 Å. The Cu–O bond seems to be longer than expected for a five-coordinate square pyramidal species, for example, a mono-armed methyl ester cyclam complex has a Cu–O distance of 2.285 Å. The increased Cu–O bond distance in [Cu] is required to accommodate an additional long range interaction between the copper() and a chloride from the copper() trichloride anion at 3.048 Å. This results in the metal ion moving further into the macrocyclic cavity. This structure contains no solvent or water molecules within the crystal lattice and no H-bonded networks are present. The copper() trichloride counterions are packed into channels within the lattice.
A single crystal sample of [Cu(HSBTE1A)](ClO) was grown by diffusion of diethyl ether into a methanolic solution of [Cu](ClO) that hydrolysed over three weeks. Analysis of the methanolic mother liquor by mass spectrometry showed complete hydrolysis had occurred. The complex could also be formed by direct reaction with chelator HSBTE1A. The complex cation has a charge of 2+, with the pendant acetic acid arm having unexpectedly retained the proton. The proton on the carboxylate group was located on the difference map and the bond lengths within the carboxylic acid group are consistent with this assignment. Other examples are known of cyclam-type ligands with pendant acetic acid arms that coordinate to copper() through the carbonyl oxygen. IR and UV/vis. studies by Kang and co-workers show coordination of a protonated acetate pendant arm in acetonitrile solution and deprotonation to form a carboxylate complex when dissolved in water.
[Cu(HSBTE1A)] contains a five-coordinate copper() ion bound in a distorted square based pyramidal geometry with a considerable distortion towards trigonal bipyramidal (a parameter value of 0.37). The apical bond is between the copper ion and the carbonyl oxygen atom, with a Cu–O bond length of 2.224 Å, which is within the expected range for such an interaction and comparable to the value observed for the methyl ester compound. The secondary amine N–H forms an H-bond with the perchlorate counterion, which in turn H-bonds to a lattice water molecule forming a 1D H-bonded chain. The observed cyclam ring configuration is again -. Comparison of [Cu(HSBTE1A)] with [Cu] shows that the additional long-range interaction with [CuCl] forces [Cu] into a less-distorted square-based pyramidal structure with an unusually elongated Cu–O distance.
Although the side-bridged copper complexes have not been extensively investigated previously, data on six X-ray structures of copper() side-bridged cyclams are currently available in the CSD. All of these chelators are tetradentate with four N-donors with no intramolecularly-coordinating pendant arm donor atoms. Fabbrizzi and coworkers report a side-bridged cyclam complex formed with copper() perchlorate, which exhibits a distorted square-planar geometry that is attributed to the constraint of the ring geometry. Distortion from planarity of the four ring N-donors is also observed in our two structures. Fabbrizzi has already demonstrated that the non-functionalised side-bridged macrocycle does not benefit from a reinforced kinetic macrocyclic effect. Further investigations are required to determine whether this is also the case for functionalised ligands with coordinating pendant ‘arms’ and higher denticity.
The synthetic methodology used to produce the complexes [Cu] and [Cu(HSBTE1A)] (2 h reflux in MeOH) is not compatible with conditions required for radiolabelling. However, the colour changes observed suggest that the final complex is formed much more rapidly. For example, the addition of copper() perchlorate to ligand results in a blue-coloured solution immediately on addition of the metal salt, and within a few minutes of stirring at room temperature this has changed to a purple-blue colour. No further colour change was observed during the course of the reaction, suggesting any intermediate complex that is formed rapidly reacts to give the final product and that complexation is complete long before the two hour reflux. Further investigation of complexation under radiolabelling conditions and in aqueous solution is required.
The neutral copper() complex of the TETA ligand was synthesised and two different X-ray crystal structures obtained, showing Jahn–Teller distortion through either across the macrocycle or along the axis where the pendant arms were coordinated bonds. This reveals the flexible nature of the TETA chelator and provides evidence to explain the more rapid exchange kinetics observed for copper complexes of TETA in comparison to chelators with configurational constraint and reduced flexibility. Chelator flexibility, cavity size match and complexation/decomplexation kinetics are all key factors in determining the utility of these chelators as radiopharmaceutical components.
Side-bridged chelators have also been of interest as radiopharmaceutical chelators. We have reported a facile synthesis of the first derivative of this type to incorporate a coordinating acetate pendant arm. The X-ray structures show that the pendant donor atom is coordinated intramolecularly to give five-coordinate copper() complexes with the novel pentadentate chelators. Our further investigations of such chelators are underway, including attempts to synthesise side-bridged derivatives with two coordinating pendant arms. Radiolabelling conditions and the stability of the copper complexes of both one- and two-armed derivatives will also be studied.
All reagents and solvents were purchased from Sigma-Aldrich, Lancaster or Fisher and were used as supplied unless otherwise stated. : Perchlorate salts are potentially explosive when dry. Although no problems were encountered during this study, caution must be exercised when handling such substances.
All H NMR and proton decoupled C NMR spectra were collected on a Jeol JNM-LA400 spectrometer at 400 and 100 MHz respectively, and referenced against residual solvent signals. Mass spectrometry was carried out by electrospray ionisation with the data collected on a Finnigan LCQ spectrometer, or by the EPSRC National Mass Spectrometry Service at the University of Swansea on a Waters ZQ4000 spectrometer. UV/visible spectra were obtained using an HP-Agilent 8453 diode array spectrometer. Elemental analyses were carried out at the University of Hull. |
Science, broadly defined, is also the knowledge obtained by study of traditional practices after a careful trial showing the predictability of their effects in identifiable groups of patients.
Today, when functional dyspepsia is diagnosed, the drug treatment is often unsatisfactory (), many patients are advised to drink different waters and types of wine, to change their diet () and introduce vitamins or are referred to a spa without much effort to individualized care (–). Without clear descriptions of patient symptoms and syndromes, it is difficult for the primary physicians to lump together typical groups of patients in order to test their response to specific treatments and transform a traditional art in explicit and public scientific knowledge (–).
In the past three decades, following the example of psychiatrists and rheumatologists, symptom-based diagnostic criteria for functional gastrointestinal disorders have been suggested (–). When they allow the identification of natural groups of patients responsive to a specific kind of treatment (), they are extremely useful in the daily medical practice, considering that functional gastrointestinal disorders represent a large proportion—more than 50%—of gastroenterologists' referral, at least in the rich countries (,).
Since one in every four persons in these societies has symptoms compatible with one of the component functional diagnoses, these syndromes have been known for centuries, albeit with variable expressions of the primary symptoms. Rich Etruscans and Romans spent all day at the Terme and the Chianciano water is known for its gastroduodenal healing effects since then.
Why have we accrued so little knowledge of the specific type of patients that can take advantage and relief from this traditional style of treatment? There are both historical and scientific explanations: for centuries to spend days or months at the Terme was a luxury for few privileged riches, a sign of social distinction and a status symbol. After the Second World War, in both West and East Europe, thermal care was reimbursed by the Welfare State, but was considered a kind of non-specific panacea against any kind of undefined stress, or as a benefit for tired workers and exhausted housewives. No clear definition of functional bowel diseases existed, nor was there any recognition that patients should be selected in order to benefit from this traditional kind of therapy (–).
As a result, thermal care turned out to be considered non-scientific and its popularity declined with the economical restrictions of the eighties. Today spas are coming back with a revenge—they —partly because a growing number of patients affected by functional disorders is dissatisfied with the few available drugs (,).
A group of general practitioners of the Tuscany Region was trained in the use of the Rome II—and a preliminary draft of the Roma III criteria ()—for dyspepsia and other functional gastrointestinal disorders, in Continuing Medical Education courses (, ECM) focused on the classification of clinical cases ().
Patients considered to have functional dyspepsia on the basis of the Rome II criteria were eligible for the trial. Functional dyspepsia was diagnosed if persistent or recurrent upper abdominal pain or discomfort was the dominant complaint (). Pain consisted of epigastric pain or burning; discomfort was characterized by the presence of one or more symptoms that included postprandial fullness, early satiety, gastric distension, belching, nausea or vomiting. Symptoms had to be present for at least 12 weeks within the previous 12 months.
Biliary disorders, infection (), structural lesions or clinically significant biochemical abnormalities were excluded by recent available documentation or ecotomography, blood tests including fasting blood sugar and liver function tests and gastrointestinal endoscopy (,).
Patients were excluded if they were already taking other medications that may alter gastric function on a regular basis and if they were known by the general practitioners as heavy drinkers, heavy smokers or habitual drinkers of more than 2–3 cups of coffee every day ().
From a methodological perspective, traditional clinical trials are too expensive and difficult to apply to thermal care. In controlled clinical trials for the comparison of two drugs and for registration, the number of observation is decided in advance and is not affected by the observed results of treatment after each patient has completed therapy. Very often, when assessing a traditional practice of care, the efficacy is unpredictable before its start but depends on the observed results in a series of individual patients. The decision to stop the investigation depends on the results. A study of this type is called sequential (), following Wald (), Armitage () and Whitehead (). The main reason for using sequential methods are as follows: (i) economy emerging from the possibility to reduce the total amount of experimentation depending on the efficacy of the treatment under study and the results obtained in the patients already completed, (ii) the possibility to achieve a specified 0.05 sensitivity and 80% power of the study without being forced to anticipate a numerical estimate of treatment effect, (iii) ethical considerations preclude random allocation or the use of placebo when there is strong historical and anecdotal prior evidence—or common belief—in the efficacy of a traditional form of treatment and (iv) for the same reason it would be undesirable to continue following the tradition when the treatment is shown by sequential medical trials to be no better than the tossing of a balanced coin.
It is also available in glass bottles but in this case it is not classified as thermal water, but as a bicarbonate-sulphate-calcium mineral water, is usually drunk at ambient temperature, and in our experience does not have the same therapeutic efficacy as the thermal mother source for the patient population involved in this research. The chemical composition of the water is reported in .
A standardized questionnaire based on the Leeds Dyspepsia Questionnaire () and already tested for validity by our group in a previous study of mineral water () was filled by each patient before going to Chianciano and in the first week after the end of treatment.
p
The characteristics of each individual patient and the frequency and intensity of each symptom before (B) and after (A) treatment are reported for specific symptoms in and for accessory symptoms in . Since the trial focused on individual patients, no use has been done of averages or other summary statistics calculated on the whole group (the availability of complete raw data make it easy to do it if the reader consider it more informative).
shows the results of the sequential trial when a reduction of the global score for all symptoms (specific plus accessory) of at least 3 points (clinically relevant) as the effect of treatment is considered. The horizontal axis shows the number of patients; the vertical axis the excess preferences or successes. The continuous thick lines are the upper, lower and middle boundaries typical of a sequential plan for the required sensitivity and power of the trial. The efficacy of treatment on each patient of the series is analyzed sequentially. As soon as the effect on a new patient is known, the line move up one unit if the case is a success, move down one unit if it is a failure.
shows results for score based on symptoms specific for functional dyspepsia. A definitive significance was reached after 23 preferences.
shows the number of patients with no impairment of everyday activities as a result of treatment (symptoms with an intensity lower or equal 2), or asymptomatic—intensity ≤1—after treatment. A vast majority of patients benefited from an improvement of symptoms to the range of intensity not interfering with activities of daily living and statistical significance was obtained after nine preferences. On the contrary complete suppression of symptoms is not a realistic outcome of treatment and the series shows no statistical significance after 27 preferences.
shows the effect of water on the ulcer-like syndrome with a clinical success defined as a reduction of intensity of at least 3 points of symptoms as before. After 15 preferences the significance was reached. shows the effect on the dysmotility-like syndrome using the same criteria. The significance was reached after 17 preferences. For the associated symptoms, shows the non-significant effect of the water on heartburn and the abdominal syndrome. The sequential analysis is somewhat different from the previous ones because we wanted the middle boundaries to be very sensitive to the negative results suggested by previous experience ().
The Chianciano thermal water care is an effective short-term therapy for the specific and associated symptoms of functional dyspepsia in a carefully selected group of patients. Parallel results are obtained when only the specific symptoms are analyzed ( and ). For practicing physicians the most relevant result is the very small number of individual patients showing no improvement and the significant number benefiting from a marked clinical improvement. This is confirmed by the statistically significant number of patients in whom we saw a change with the disappearance of impairment in everyday activities because of symptoms intensity (). Vice versa, the complete disappearance of symptoms is not what physicians should promise to patients, as shown by the non-significant series in where no symptoms with an intensity ≤1 was the required outcome.
If the pattern of response to a specific therapy is accepted as evidence of the underlying disease mechanism and classification validity, thermal care emerges from the trial as a specific treatment for functional dyspepsia and associated symptoms, but does not confirm the possibility to differentiate between ulcer-like and dysmotility-like subgroups. On the contrary, the marked parallelism among the series of patients for the two syndromes suggests that as far as therapy is involved, they should be classified together ( and ) (,).
Thermal water and care has no significant effects on heartburn and the abdominal syndrome (), suggesting a specific effect of therapy at the gastroduodenal level and the possibility of different mechanisms for therapies targeted at the esophageal and abdominal walls (–).
The trial confirms the validity and practical utility of the Roma II criteria for identification of patients affected by functional dyspepsia and its inclusion among the gastroduodenal disorders, but does not confirm the more specific subgroups of ulcer-like and dysmotility-like dyspepsia as a criterion that makes a difference for treatment (–). Both functional esophageal disorders and abdominal disorders need a different and more specific type of therapy (–).
The study confirms the great practicality of the sequential trial approach to test the efficacy of traditional kinds of care on individual patients. If treatment is very efficacious, as in this case, a small number of carefully selected patients are sufficient to test the many hypotheses emerging from traditional wisdom or previous experience while avoiding the ethical, practical and economic difficulties of applying the more standard fixed-number trial approach.
The major weak point of the sequential approach is that the evidence for efficacy is often reached with very few patients; as expected, the confidence limits for the percentage of success obtained are wide. This is not critical when the efficacy of the treatment is great, but limit their application to the comparison of drugs or remedies when a small difference is expected (for example in the comparison of two statins or proton pump inhibitors).
As far as the practical physician is concerned, the sequential approach gives information at the individual level, avoiding the abstract presentation of table of averages and percentages without any precise indication of the concrete type of person that can take advantage of the tested practice of care. For this reason, the complete database of individual data is provided as part of the results and no summary measure has been calculated. The concrete individualized approach should help the physician in the application of the results to the precisely described natural kind of patient they can encounter in everyday ambulatory practice ().
Nobody can believe that 12 days of care, although very effective, can have positive effects lasting forever on functional gastrointestinal disorders that we know for their chronic course. The next studies should answer the many questions that the positive results reported make urgent: how long the positive effects of 12 days in Chianciano last? How often should the cycle be repeated in order to optimize the improvements? Is the practice best for prevention in mildly affected patients, or is it better to prescribe it when the clinical pattern is more severe, as in these patients? Does the availability of a regular and individualized schedule in each case improve the quality of life and reduce the direct and indirect costs of functional dyspepsia? Do these patients need the association of other kinds of treatment or is thermal care enough by itself for the long-term care of functional dyspepsia? Why thermal care is more effective than the equivalent mineral water based care? Is any extramolecular mechanism involved? ()
We believe that this study is only the first step toward a more scientific approach to the evaluation of many types of traditional care now and reimbursed by both the welfare state and insurance companies in many countries.
The emerging specificity of thermal care for a easily identified group of patients, affected by a specific functional disorder confirm our assumption that many kinds of traditional care are not panacea or placebos, but show specific activity at precise levels or organs of the body and on specific and recognizable symptoms, patterns and syndromes.
The results of the study suggest that it is important to test the efficacy of different practices of thermal and other traditional care on different disorders, upon different levels of the gastrointestinal tract and on different organs and systems of the body. This can be done easily and scientifically by the application of the sequential trial approach. Stimulated by the results of this study, a randomized controlled trials (RCT) is under way to compare the efficacy, direct and indirect costs, and duration of the positive effect of thermal water compared to standard treatments.
Why in our experience does thermal water and mineral water, with the same chemical composition give different results in the same group of patients? The available literature and the regulation for reimbursement are focused mainly or exclusively on the molecular effect of the minerals—chemical analysis—while in our experience other three factors should be integrated in future research as follows: (i) the temperature at the moment of ingestion, (ii) a possible homeopathic effect of thermal water but not mineral water due to the passage on diluted minerals and (iii) the relaxing effect of 12 days at a spa, that is missing when the patients drinks the mineral water at home. No data at the moment are available to estimate the integrated contribution of each of these three additional factors.
The traditional classification of therapeutic waters is based on the chemical characteristics, but experience with patients suggests that other factors should be taken into account in future research. The Rome II and Rome III definition of functional dyspepsia consider these symptoms as specific of the syndrome. Most patients affected by functional dyspepsia are affected also by other symptoms that do not dominate the clinical picture but are accessory and often present.
Thermal care is effective in reducing the global score for all symptoms in more than 80% of patients after only 20 trials. It targets the specific symptoms for functional dyspepsia and is effective in 80% of patients after only 23 trials. Thermal care is effective in reducing symptoms to a level not interfering with everyday activities—significant after nine patients—but is generally unable to completely suppress the symptoms—non-significance after 27 trials.
Rome II criteria defines ulcer-like syndrome as a subgroup of functional dyspepsia, but and show that they are not therapeutically distinguishable. The results are similar to those obtained with patients affected by ulcer-like dyspepsia, suggesting they are not therapeutically distinguishable. The Chianciano thermal care is ineffective against esophageal and abdominal symptoms, showing a specific effect on the wall of the stomach and duodenum. |
Complex treatment systems such as palliative care, public health, integrative medicine, rehabilitative medicine or traditional Chinese medicine, and interventions within those systems, are an important part of healthcare around the world. However, these approaches to healthcare are not always well served by the biomedical model of diagnosing, treating, understanding and evaluating diseases which emphasizes the evaluation of single-component interventions. The applicability of this model for investigating healthcare as it is actually practiced is limited. Hence, a broader perspective is necessary.
Complementary and alternative medicine (CAM) researchers have a particular interest in driving the debate about how best to assess complex healthcare systems, as they struggle with demands from regulators, insurers, purchasers, providers and patients for ‘evidence’ of effectiveness and efficacy in order to meet the standards of ‘evidence-based medicine’. The debate regarding these research design issues within conventional medicine has risen in parallel with the growing emphasis on team-based medicine and integrative medical teams, and, related to this, the increasing complexity of treatment interventions. In addition, there is increased recognition that explanatory (placebo-controlled) randomized clinical trials (RCTs) alone cannot adequately assess these interventions and their outcomes. Explanatory RCTs are conducted under conditions that are as controlled as possible and include the following characteristics: administration of a placebo to the control group in an attempt to hold all possible causal elements constant except for the intervention under investigation; standardization of inclusion and exclusion criteria; standardization of the intervention under investigation; randomized allocation of the participants to the intervention or the control group(s); blinding (allocation concealment) of the participants and investigators (if possible) ().
A number of descriptive articles have attempted to explain how one might begin to assess these complex interventions. Each approach has developed its own language and concepts to describe the phenomena, making communication and consensus building difficult across approaches. For example, a variety of terminologies has been proposed to describe what appears to be essentially the same phenomenon. In this paper we use the term ‘complex healthcare systems’. We define these as complex interventions to improve or enhance health and well-being as well as to prevent disease. ‘Complex’ denotes the entangled interrelationships among multiple ‘active’ components of the intervention. In addition, it highlights that, in general, the effects of the ‘whole’ intervention or system are interactive rather than additive (–), with the potential that the whole is more than the sum of the parts. This reasoning reflects the theory that complex systems have an inherent self-organizing property and that the elements of complex systems themselves interact in such a way that through the interplay of the elements new properties emerge that cannot be seen when investigating only the component parts. In our view, both the human body and systems of healthcare have to be seen as complex, self-organizing systems that create new, emerging properties through the interplay of their component elements (–).
The purpose of this paper is to compare, contrast and critique four different approaches to the assessment of complex healthcare interventions or systems, also identified as ‘whole’ systems. Terms such as whole systems, complex systems, CAM systems and whole medical systems appear to describe similar concepts. However, this divergence in terminology reflects some unique features with respect to how each is defined and also the cultural context in which each arose. The present paper aims to bring some clarity to the field and helps to establish a broader awareness and understanding of the issues, as well as to facilitate interdisciplinary research.
An international working group of researchers met in 2005 in Tromsø (Norway) to further develop whole systems research methodology and identified the need to bring clarity to the emerging conceptual and methodological literature. In order to address this, a subgroup of seven people reviewed literature relevant to designing appropriate research methods to assess complex healthcare systems. They used the following criteria:
The group did not attempt to be comprehensive in this selection, but rather strove to identify a range of documents from a variety of countries to explore the diversity of approaches to the issue. Four approaches were identified at the meeting:
After the original meeting in Norway, a fifth approach from the USA Institute of Medicine's Committee on the Use of Complementary and Alternative Medicine by the American Public was reviewed (). While this fifth approach does identify the need for appropriate study design for ‘complex treatment packages’, they indicate that a detailed discussion is beyond the scope of their report. Thus, this fifth approach is not reviewed for this paper.
A minimum of two members of the subgroup independently completed a focused qualitative content documentary analysis of each approach (,). Key documentation associated with each approach was reviewed to identify underlying principles and assumptions, models for assessing complex healthcare systems, and specific strengths or weaknesses of the approaches. Individuals assessing each approach then met face-to-face until they reached consensus on the key themes from the documents and prepared to present the approach to the rest of the authors. An additional face-to-face discussion of each individual approach was held to identify commonalities and differences amongst the approaches reviewed. Preliminary findings were then submitted by the subgroup to the entire group at the international meeting for discussion and feedback. (The subgroup working on this project was one of three groups working on different projects as part of the international meeting.) The findings from this process of interpretation and synthesis, with minor modifications resulting from subsequent electronic communications, are provided below.
table-wrap
#text
The most significant finding was that the International WSR group, the US NCCAM, the Norwegian National Research Center in Complementary and Alternative Medicine and the UK's Medical Research Council all acknowledge the need to investigate complex healthcare systems as integral systems. Other recent documents, such at the Institute of Medicine Report from the USA, have also identified this need (). There is considerable agreement that the classical pharmacological RCT model alone is not sufficient. Yet there does not seem to be universal agreement on what should be done instead. This may be partly because the perspectives analyzed here evolved from different cultural and social contexts to meet different needs. For example, the Medical Research Council guidelines were specifically developed from a health services perspective in order primarily to help researchers design higher quality RCTs. In contrast, the Norwegian National Research Center in Complementary and Alternative Medicine perspective was developed to guide a CAM research center's priority setting and research goals for therapies that were already in widespread use.
This raises the question: Is there a need (and is it possible) to develop a comprehensive, common set of guidelines for assessing complex healthcare systems? Or is diversity needed to match different social contexts and different funding agencies? For example, the WSR perspective clearly advocates a non-linear, iterative approach (). Prescriptive evidence hierarchies of research methods may not be helpful, but explicit guidelines of what needs to be incorporated or considered in complex healing systems research may help to increase the quality of future research.
Despite the use of different terminology, analysis of these four approaches suggests a growing international understanding of the need for a new conceptual framework for assessing complex healthcare systems. Multiple methods and integrated programs of research undertaken by interdisciplinary teams appear to be necessary. This field will benefit from additional international and interdisciplinary dialogue.
Competing Interests: None to declare. |
I would like to thank Mr James Flowers for his interesting commentary on Qi and his discussion of our work on Ki (called Qi in Chinese) (). I was impressed with his deep philosophical considerations and beautiful poetic explanation of Qi as being behind everything. In the same issue, Dr E.L. Cooper, the Editor-in-Chief of , wrote that, in the early stage of the development of this journal, more philosophically and historically oriented manuscripts were received for review than those of original scientific work. He believes that both philosophical consideration and rigorous scientific work should be blended with each other for the healthy growth of science (). Therefore, he was happy to notice that we contributed new experimental work on Ki, and subsequently, our papers stimulated the philosophical discussion by Flowers. These two articles prompted me to write this article covering the nature of Ki as well as its philosophical aspects and the significance in the modern civilization.
I am thoroughly grounded in modern science, but my roots are in Eastern culture. (I was born in Japan and received education there.) I have a degree in biophysics, but I also have a black belt in Aikido and taught it in the USA from 1967. I learned about the Nishino Breathing Method (NBM) through Mr Kozo Nishino's publications [() and many other Japanese books]. I first attended the school of NBM in Tokyo 10 years ago, and 3 years later, the collaboration with Mr Nishino started in order to find a scientific basis for Ki. After several years of struggle, our first paper on the effects of Ki on cultured cancer cells was published (), which Flowers cited in his commentary. Subsequently, we published two more papers (,). I would like to start my response with the data in these two papers, because they may have answers to some of the questions which he raised in his article. Then, I will discuss some philosophical aspects of Ki, and a possible role of the scientific study of Ki effects in broadening our Western perspectives.
#text
xref
#text
Flowers was asking whether Ki may be an ‘energy’. Many believe it is. However, it is interesting to note that there is a subtle difference in understanding between the Chinese and the Japanese. The former seems to believe that it is a ‘substance’ or ‘matter’ flowing in and through our bodies, and that it can be emitted from the body of a Qigong healer. In contrast, the Japanese considers that it is a form of energy. An interesting concept was published by Shinagawa's group that treats Ki as a form of ‘information’ (,). Flowers considers Ki as being about ‘relationships and patterning’. These concepts are another way of describing ‘information’.
The most interesting practice in NBM is called the Taiki-practice (a method developed by Kozo Nishino, the founder of the Nishino Breathing Method, to develop the level of individual's Ki through the Ki communication between an instructor and a student). In a class room, a student stands in front of a wall (which is covered with a soft cushion to absorb the shock when the student hits the wall) with his/her back facing the wall. The student and an instructor touch their hands (right to right or left to left). Then, they try to send Ki alternately from their hands. When the instructor sends strong Ki, the student feels the ‘energy’ and steps or runs backward toward the wall. With this practice, students can learn how to raise the level of their Ki energy. It is important to note that after training, the effect is the same even when the student is blindfolded and the instructor sends Ki without touching student's hand. Through the study on the Taiki-practice, we raised the possibility that Ki-energy may carry information, and that the information is in a form of ‘entropy’ (). If Ki consists of simple energy, then, Ki effects might be mimicked by an instrument. However, if Ki involves entropy, it may be difficult to artificially reproduce the entire Ki effect.
#text
xref
#text
xref
italic
#text
Flowers raised an interesting point as to whether Qi (or Ki) can only cause a desirable action. Since he is proficient in Chinese traditions, he may be discussing Yin and Yang (negative and positive) aspects of Qi.
As far as I know of, I have not heard of any negative effects of Ki in NBM during my 10 years association with the school. For this reason, Nishino's Ki seems to be very positive. I believe that the goal of our life is to have a healthy, happy life. A human being is happy and healthy when he/she is doing something good and enjoyable to himself/herself and, at the same time, beneficial to others. In other words, when individuals follow this humanistic path, they will become happy. If someone keeps following a negative or evil path, this person eventually loses good fortune, weakens his/her life force and becomes unhappy. I believe in the positive ethical doctrine that a human's inborn nature is fundamentally good. In this sense, I am an optimist, and I believe that NBM is a natural and humanistic way.
#text
Philosophy has been considered as the mother of all aspects of human culture, and indeed, science originated from philosophy. It took more than 2000 years until modern science proved Democritus’ atomic theory. As Flowers mentioned, in the Christian West, God was the center of everything as opposed to Ki (or Qi) being the center of everything in the East. Modern science and technology were largely developed in the West, where people were taught that the plants and animals on earth were to serve human beings. This seems to be related to the destruction of the ecosystem of which everyone is now more aware. In contrast, it has been taught in the East that the human being and its environment are one: Life is created from the environment, and when life ceases, it is returned to the environment.
In this sense, it can be said that modern science lacks humanistic philosophy. Science and technology became slaves of human desire, egoism and the pursuit of monetary gain. As a result, the pursuit of happiness was forgotten. Common problems in the world today, such as the fragmentation of the family, increase of mental diseases, depression and suicide, homicide, violence, terrorism and war (which is the ultimate form of homicide and terrorism) are the result of the loss of concern with ‘happiness’. In other words, the lack of guiding philosophy caused the loss of human dignity. This problem was mentioned in our first paper on Ki ().
The foundation of physics was laid by Descartes (1596–1650) who asserted that the movement of an object (for example, a canon ball) can be described by mathematical equations in Cartesian coordinates, and it is not influenced by the human mind. This separation of body and mind helped to develop mathematics- and technology-oriented modern culture. However, as I mentioned earlier in this commentary, this separation of mind may also have contributed to the lack of humanism in modern civilization.
Descartes’ mathematical physics was completed by Newton (1642–1727) who developed a mathematical technique called ‘differentiation’. It is based upon the concept that a part of a ‘curve’ can be considered as a ‘straight line’ when we look at an infinitesimally small portion of the curve. This can be regarded as an example of the Greek ‘reductionism’ in which a complex phenomenon can be reduced to simple, fundamental components. Using this mathematical tool, and based upon his discovery of universal gravitation, Newton completed a mechanistic world-view of the universe. His greatest success was that he was able to predict the movement of heavenly bodies, such as the sun, planets, moons and comets. The fact that Newtonian physics can predict the occurrence of solar and lunar eclipses with an accuracy less than a second, and predict the return of the comets with the orbital time span of hundreds of years symbolized the success of mechanistic and deterministic world-view of classical science.
The reason why Newton's physics was so successful in predicting the future motion of heavenly bodies was that the motion can be described by linear equations. Since the success was so great, no one paid attention to the finding of Poincaré (1854–1912), who discovered that the movements of three planets with similar sizes are unpredictable (because the motion is expressed by nonlinear equations). It was as recently as 1961 that the importance of the unpredictability in a nonlinear system was ‘re-discovered’ by Lorenz. He was experimenting with a nonlinear weather model on his computer. He was surprised to find that a change of the initial conditions, which is as small as only 1/5000 different from the previous run, yielded a dramatically differing weather pattern (). He called such an extreme sensitivity of a nonlinear system to initial conditions a ‘butterfly effect’. His result became well known as a joking notion that ‘a butterfly stirring the air today in Peking can transform storm systems next month in New York’. With his discovery, the Newtonian era of the predictability of a linear system ended, and it was replaced by the unpredictability of a nonlinear system, and the new science called ‘chaos’ was born (,). As I will describe in the rest of this article, this new science seems to hold a key in understanding Ki.
#text
As Flowers beautifully described, Ki may represent the entity of life itself. Then, the understanding of Ki may shed light on other aspects of biological sciences. When Darwin (1809–82) proposed the theory of evolution, many religious figures opposed his view. We thought that that battle was over a long time ago, but at least in the USA, it was not. Certain groups of people still oppose teaching the theory of evolution in the biology course in public school. Some people assert that an alternative ‘designer theory’ should also be taught in public school. Interestingly enough, the strongest opposition to the theory of evolution comes from the scientific community, not from the religious circle. Certain people argue that a big event, for example, the creation of complex organs (such as the brain or the eyes) cannot be explained as the result of the accumulation of a small ‘mutation’ and ‘natural selection’. If the entire change took place as the accumulation of small, step-by-step changes, then, intermediate species which correspond to each step should be found. They argue that ‘intermediate species’ between the unevolved and the fully evolved were never found. The theory of nonlinearity may have an answer to this long-lasting dispute.
A fascinating feature of a nonlinear system is its complex behavior, namely: (i) unpredictability of the future, (ii) extreme sensitivity to the initial condition and (iii) synchronization between two nonlinear systems. In other words, anything can happen over the life system because its essence is nonlinear. It has been proposed that ‘order’ may be created from the nonlinear chaotic system (). The spontaneous creation of life and its evolution on earth may be explained by the complexity and the infinite potential which a nonlinear complex system possesses (,). From the study of the evolution of an ‘artificial life’ (life-like behavior of a model on the computer screen), it was found that big, sudden changes occur sporadically in the nonlinear ‘evolution’ model (). If similar phenomena take place in the evolution of species on earth, then, we can explain why ‘intermediate species’ cannot be found. In summary, the nonlinear theory proposes that no ‘designer’ would be needed for creation and evolution.
#text
W
e
s
t
e
r
n
c
i
v
i
l
i
z
a
t
i
o
n
h
a
s
b
e
e
n
b
u
i
l
t
o
n
t
w
o
p
i
l
l
a
r
s
,
n
a
m
e
l
y
,
G
r
e
e
k
r
e
d
u
c
t
i
o
n
i
s
m
a
n
d
t
h
e
C
h
r
i
s
t
i
a
n
-
N
e
w
t
o
n
i
a
n
d
e
t
e
r
m
i
n
i
s
m
.
B
o
t
h
p
i
l
l
a
r
s
a
r
e
l
i
n
e
a
r
c
o
n
c
e
p
t
.
I
n
c
o
n
t
r
a
s
t
,
E
a
s
t
e
r
n
c
u
l
t
u
r
e
h
a
s
b
e
e
n
s
h
a
p
e
d
a
l
o
n
g
a
c
o
n
c
e
p
t
o
f
‘
d
e
p
e
n
d
e
n
t
o
r
i
g
i
n
a
t
i
o
n
’
p
r
o
p
o
s
e
d
b
y
G
a
u
t
a
m
a
(
6
t
h
c
e
n
t
u
r
y
B
C
)
.
T
h
i
s
c
o
n
c
e
p
t
h
o
l
d
s
t
h
a
t
e
a
c
h
c
o
m
p
o
n
e
n
t
i
n
t
h
e
u
n
i
v
e
r
s
e
(
i
n
c
l
u
d
i
n
g
a
n
i
m
a
t
e
a
n
d
i
n
a
n
i
m
a
t
e
)
i
s
i
n
t
i
m
a
t
e
l
y
r
e
l
a
t
e
d
w
i
t
h
a
l
l
o
t
h
e
r
c
o
m
p
o
n
e
n
t
s
.
H
e
a
s
s
e
r
t
e
d
t
h
a
t
e
a
c
h
a
n
d
e
v
e
r
y
h
u
m
a
n
b
e
i
n
g
i
s
e
q
u
a
l
,
a
n
d
t
h
a
t
t
h
e
r
e
i
s
n
o
a
b
s
o
l
u
t
e
b
e
i
n
g
.
T
h
e
m
o
s
t
i
m
p
o
r
t
a
n
t
f
e
a
t
u
r
e
o
f
t
h
i
s
s
y
s
t
e
m
i
s
t
h
a
t
i
t
i
s
n
o
n
l
i
n
e
a
r
.
T
h
e
r
e
f
o
r
e
,
t
h
e
f
u
t
u
r
e
o
f
t
h
e
s
y
s
t
e
m
i
s
e
s
s
e
n
t
i
a
l
l
y
u
n
p
r
e
d
i
c
t
a
b
l
e
.
T
h
e
d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n
t
h
e
c
u
l
t
u
r
e
s
i
n
t
h
e
W
e
s
t
a
n
d
t
h
e
E
a
s
t
m
a
y
b
e
d
e
s
c
r
i
b
e
d
b
y
t
h
e
d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n
‘
l
i
n
e
a
r
p
h
i
l
o
s
o
p
h
y
’
a
n
d
‘
n
o
n
l
i
n
e
a
r
p
h
i
l
o
s
o
p
h
y
’
.
S
i
n
c
e
K
i
p
h
e
n
o
m
e
n
a
a
r
e
e
s
s
e
n
t
i
a
l
l
y
n
o
n
l
i
n
e
a
r
,
a
s
t
r
a
i
g
h
t
f
o
r
w
a
r
d
a
p
p
l
i
c
a
t
i
o
n
o
f
l
i
n
e
a
r
p
h
i
l
o
s
o
p
h
y
m
a
y
n
o
t
b
e
e
f
f
e
c
t
i
v
e
i
n
i
t
s
s
t
u
d
y
.
W
e
h
o
p
e
t
h
a
t
t
h
e
s
t
u
d
y
o
f
K
i
p
h
e
n
o
m
e
n
a
m
a
y
h
e
l
p
b
r
i
d
g
e
s
o
m
e
o
f
t
h
e
a
p
p
a
r
e
n
t
d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n
W
e
s
t
e
r
n
a
n
d
E
a
s
t
e
r
n
c
u
l
t
u
r
e
.
The results of Ki studies force us to re-examine the very foundation of modern sciences which has been based upon Descartes’ assertion that seperates body and mind. As we wrote in a previous paper (), Ki-phenomenon is based upon the findings that our mind seems to influence the motion or vibration of a molecule, and as a result, it creates a motion of the entire human body. It is a challenge to traditional scientific thinking. Therefore, further analysis of Ki may help to create a Copernican paradigm shift in science. This shift may help better address some of problems in our civilization today.
Ki phenomenon seems to be characteristic to the nonlinear nature of life. If so, the study of Ki may help deepen our understanding on our life and the universe itself. Since Ki is related to our life activity, the understanding of Ki would contribute to the elucidation of the beautiful nature of life itself. I hope this shift from the linear, deterministic Newtonian world-view to a nonlinear, chaos-based world-view will highlight the infinite potential of each and every human being, and help affirm the dignity of life. This is so badly needed in today's troubled world. |
Mokuboito (Mu-Fang-Yi-Tang), a kind of Kampo formulations, has clinically been used for heart failure (). It has been shown that Mokuboito improves heart failure symptom, and reduces the level of New York Heart Association (NYHA) class and plasma brain natriuretic peptide (BNP) concentration ().
Mokuboito consists of 4 g Sinomeni Caulis et Rhizoma (rhizome of Rehdler et Wilson), 3 g Cinnamomi Cortex (bark of Blume), 3 g radix (roots of C. A. Meyer) and 10 g Gypsum Fibrosum. cortex and its constituent (cinnamic aldehyde) have been reported to have some pharmcological effects; antiplatelet and anti-inflammatory effects (,) and also antitumor activity (). possesses many pharmacological effects and exhibits some clinical efficicacies: care of general fatigue and stress (,). In addition, it possesses immunomodulating (,) and neuroprotective actions (). Finally, Gypsum has been used as one of the crude drugs in traditional Kampo medicine. Its pharmacological is reasonably clear, but previous reports have shown that Gypsum prevents the thermogenesis effect induced by ephedrine at an ambient temperature of 22°C ().
has traditionally been considered as a modulator of body's fluid and been applied for disturbances of body fluids. Sinomenine is one of alkaloids extracted from (). As an antirheumatic and anti-inflammatory regulator, sinomenine is clinically used for the treatment of rheumatoid arthritis (,). In basic researches, sinomenine reduces the production of prostagrandin (PG) E and nitric oxide (NO) from macrophages (). Sinomenine reduces inflammatory parameters and attenuates proliferation of synovial fibroblasts in rat adjuvant arthritis models (). In addition, acute and chronic cardiac allograft ejections are blocked by the immunomodulatory effects of sinomenine (). More recently, Satoh () has demonstrated that Mokuboito and its constituents exhibit cardiac electropharmacological actions and exert the conservative actions on cardiomyocytes. These findings indicate that Mokuboito, and sinomenine exert effective pharmacological actions that can improve chronic heart failure via the recovery of cardiac functions.
In general, the treatment of heart failure consists of (i) reducing workload of heart, (ii) protection of cardiomyocytes and (iii) restriction of sodium and fluid control. For reduction of the preload and afterload, the dilations of arterioles and veins are required in the case of elevated filling pressures and reduced cardiac output. We demonstrated that dilated vasoconstrictions (). Therefore, Mokuboito, or sinomenine may treat heart failure not only by improvement of cardiac effects but also by controlling the tension of blood vessels. In these experiments, therefore, we wanted to elucidate in more detail the vascular pharmacology of Mokuboito, and sinomenine, using rat aorta rings.
All experiments were carried out, according to the guidelines laid down by the Nara Medical University Animal Welfare Committee, and also under the terms of the Declaration of Helsinki.
Male Wistar rats (4–10 weeks old) were anesthetized with ether, and euthanized by exsanguination. The thoracic aorta was quickly removed, and the isolated aorta cut into rings of 3 mm in length. The rings were suspended between two triangular-shaped stainless steel stirrups in a jacketed organ chamber filled with 20 ml modified Krebs–Henseleit solution. The modified Krebs–Henseleit solution was comprised of (in mM); 118 NaCl, 4.6 KCl, 1.2 MgSO, 1.2 KHPO, 11.1 glucose, 27.2 NaHCO, 0.03 Na-ethylenediaminetetraacetic acid (EGTA) and 1.8 CaCl. The chamber solution was kept at 36.5°C and oxygenated with 95% O and 5% CO. The lower stirrup was anchored and the upper stirrup was attached to a force-displacement transducer (Nihon Kohden TB-652T, Tokyo, Japan) to record the isometric force. All rings were stretched to generate a resting tension of 1.2 g, which was optimal for contractions with α-adrenergic receptor agonist. After 40 min of resting, norepinephrine (NE, 5 μM) was added to the tissue bath. After the contractile response became steady, the drugs were cumulatively administrated into the bath solution. The effects of each concentration of the drugs were measured 6–10 min after the responses became steady. The responses were analyzed as a percentage change from the value before the application of drugs.
Mokuboito, the kampo medicine used in this study, has been supplied from Tsumura Co., Ltd (Tokyo) as a spray-dried powder which was extracted with boiling water of a mixture of respective ground raw materials, 4 g of Sinomeni Caulis et Rhizoma (rhizome of Rehdler et Wilson), 3 g of Cinnamomi Cortex (bark of Blume), 3 g of radix (roots of C. A. Meyer) and 10 g of Gypsum Fibrosum. This Mokuboito is exactly equaled to TJ-36 provided by Tumura Co., Ltd. extract was also prepared as a spray-dried powder extracted with boiling water of a ground raw material of Sinomeni Caulis et Rhizoma (rhizome of Rehdler et Wilson).
The other drugs used were sinomenine (7,8-didehydro-4-hydroxy-3,7-dimethoxy-17-methylmorphinae-6-one), indomethacin and staurosporine (Wako Chemical, Kyoto, Japan). -monomethyl--arginine acetate (L-NMMA), and nicardipine (Sigma Chemical Co., St Louis, MO, USA) were also used.
All values are represented as means ± SEM. The differences of data in mean values were analyzed by the Student's -test and ANOVA followed by tests (Dunn–Bonferonii test), and a -value of <0.05 was considered significant.
The aorta ring strip of rat exhibited a strong constriction after an initial application of 5 μM NE. Mokuboito applications (0.03–3 mg ml) markedly relaxed the constriction induced by NE in a concentration-dependent manner, as shown in . The significant relaxation was produced at concentrations of over 0.1 mg ml. Mokuboito at 3 mg ml decreased it by 98.9 ± 2.5% ( = 7, < 0.001). To examine the involvement with endothelium-dependent relaxation, 40 min pretreatment with L-NMMA (a non-selective NO synthesis inhibitor) was carried out. In the presence of L-NMMA (100 μM), the vasorelaxation induced by 0.03 and 0.1 mg ml Mokuboito was significantly attenuated from 5.0 ± 0.5% ( = 7) and 9.2 ± 1.6% ( = 7) to 2.3 ± 1.3% ( = 5, < 0.01) and 3.4 ± 1.7% ( = 5, < 0.05), respectively (). Also, Mokuboito was applied in the presence of indomethacin (an inhibitor of prostanoid production). Indomethacin (10 μM) significantly reduced the vasorelaxation of Mokuboito at 1 and 3 mg ml from 35.1 ± 7.1% ( = 7) and 98.9 ± 2.5% ( = 7) to 11.6 ± 8.1% ( = 5, < 0.01) and 57.5 ± 14.4% ( = 5, < 0.01), respectively (). In addition, the vasorelaxation was also examined in the presence of nicardipine to elucidate the involvement with Ca channels on the vascular smooth muscle cells. Nicardipine (0.1 μM) attenuated the vasorelaxation at 3 mg ml Mokuboito to 85.3 ± 3.7% ( = 5, < 0.01). The involvement with β-adrenoreceptor stimulation (cAMP-dependence) was also examined. Mokuboito (0.03–1 mg ml) further constricted the NE-induced constrictions in the presence of 0.3 μM propranolol (A). However, additional application of 3 mg ml Mokuboito reversely dilated the vasoconstriction to almost control level. The summarized data are presented in B.
Using a non-loaded aorta (not pretreated with NE), Mokuboito was cumulatively administrated to examine whether it caused vasoconstriction (A). Mokuboito at lower concentrations of 0.03–0.3 mg ml produced a vasoconstriction on the non-loaded aorta. At 1 mg ml, however, Mokuboito reversed to the vasorelaxation. Furthermore, the vasoconstriction induced by Mokuboito (at 0.03–0.3 mg ml) was blocked by 10 μM phentolamine (α-adrenoceptor blocker) (B and C). Mokuboito involved with β-adrenoceptor stimulation, as mentioned above. These results indicate that Mokuboito has both pharmacological characteristics for a vasorelaxation via β-adrenoreceptor stimulation and a vasoconstriction via α-adrenoreceptor stimulation.
also caused a concentration-dependent vasorelaxation on the NE-induced vasoconstriction (). at 3 mg ml dilated the constriction by 97.0 ± 4.8% ( = 6). In the presence of L-NMMA (100 μM), (at 0.1–3 mg ml) also significantly decreased the relaxation (). The vasorelaxation at 3 mg ml was attenuated from 97.0 ± 4.8% ( = 6) to 84.9 ± 1.9% ( = 5, < 0.05). Indomethacin (10 μM) and nicardipine (0.1 μM) also attenuated the -induced relaxation. However, propranolol did not affect it.
As shown in A, similar vasoconstriction on the non-loaded rat aorta was produced by 0.03–1 mg ml . At 3 mg ml reversed to dilate the vasoconstriction, like Mokuboito and the vasoconstriction is completely blocked by 10 μM phentolamine (B–D).
Sinomenine caused no effects on non-loaded aorta and never produced vasoconstrictive effect. Sinomenine (0.1–100 μM) potently relaxed the constriction induced by NE in a concentration-dependent manner, as shown in . Under the pretreatment with 100 μM L-NMMA, the vasorelaxation induced by 100 μM sinomenine was attenuated from 68.8 ± 5.1% ( = 6) to 25.3 ± 2.3% ( = 5, < 0.01) (). Indomethacin (10 μM) also strongly reduced it to 37.1 ± 9.3% ( = 5, < 0.001). The relaxation of sinomenine (0.1–100 μM) was significantly attenuated by nicardipine; at 100 μM sinomenine from 68.8 ± 5.1% ( = 6) to 35.5 ± 6.9% ( = 5, < 0.001). Propranolol (0.3 μM) also significantly attenuated the sinomenine (1–100 μM)-induced relaxation. The vasorelaxation at 100 μM sinomenine was attenuated to 45.2 ± 4.2% ( = 5, < 0.01).
The results from the present experiments were as follows: (i) Mokuboito, and sinomenine exhibited concentration-dependent vasorelaxation on NE-induced vasoconstriction, (ii) L-NMMA and indomethacin significantly attenuated the vasorelaxation induced by Mokuboito, and sinomenine, (iii) nicardipine also decreased them, (iv) Mokuboito and at lower concentrations caused vasoconstrictions on the non-loaded aorta, but at high concentrations oppositely dilated the vasoconstriction induced by lower concentrations, (v) sinomenine never caused the vasoconstriction on the non-loaded aorta, (vi) in the presence of propranolol, Mokuboito at lower concentrations reinforced further the NE-induced vasoconstriction, but at 3 mg ml it was reversed to decrease the vasoconstriction and (vii) propranolol also attenuated sinomenine-induced vasorelaxation, but caused no effects on the vasorelaxation induced by .
Mokuboito possessed the vasoconstrictive effect on the non-loaded rat aorta. Mokuboito at 0.03–0.3 mg ml caused vasoconstriction, but at 1 mg ml decreased the vasoconstriction induced by the lower concentrations and then recovered to control level. Since the vasoconstriction was blocked by phentolamine, Mokuboito produced the vasoconstriction mediated through α-adrenoceptors. Under the pretreatment with propranolol, also, Mokuboito at lower concentrations (0.03–1 mg ml) reinforced further the NE-induced vasoconstrictions. Therefore, these results indicate that Mokuboito possesses both β- and α-adrenoceptor stimulating actions. At lower concentrations of Mokuboito, α-adrenoceptor stimulating action overcomes its β-adrenoceptor stimulating action, resulting in an appearance of only vasoconstriction on the non-loaded rat aorta. At higher concentrations, the other actions including β-adrenoceptor stimulation of Mokuboito were exerted, and as a result, vasodilation would be produced.
Both L-NMMA and indomethacin attenuate vasorelaxation induced by Mokuboito. Thus, Mokuboito exerts the vasorelaxation by the secretions of endothelium-derived relaxation factor (EDRF, that is NO) and also of PGI. Furthermore, nicardipine attenuated the vasorelaxation induced by Mokuboito, resulting from Ca channel inhibitory action.
also possesses the vasoconstrictive effect on the non-loaded rat aorta; and at 3 mg ml reversed to decrease the vasoconstriction. Similarly, the vasoconstriction was attenuated by phentolamine, and the vasorelaxation was reduced by propranolol. never caused vasoconstriction in the presence of propranolol. Therefore, these results indicate that also possesses both β- and α-adrenoceptor stimulating actions. However, α-adrenoceptor stimulating action is not as strong as Mokuboito, and the vasoconstriction does not appear even under the pretreatment with propranolol.
Mokuboito has β- and α-stimulating actions. Since both α- and β-adrenoceptors are widespread in the whole body, Mokuboito might stimulate both adrenoceptors. In our study, it is still unclear whether Mokuboito, and sinomenine have more selective affinity to α- and β-adrenoceptors. The α-stimulating effects of Mokuboito and might be mediated through α-adrenoceptor on adrenergic nerve terminals.
Sinomenine possesses the strong vasodilating action (). However, sinomenine never causes the vasoconstrictive effect. The vasorelaxation was similarly attenuated by L-NMMA and indomethacin. These results indicate that sinomenine also produces EDRF and PGI releases. Furthermore, since nicardipine inhibited the sinomenine-induced vasorelaxation, the vasorelaxation is responsible for Ca channel inhibition.
The EDRF and PGI releases are activated by an increase of cellular Ca ([Ca]) in endothelium cells (). The detailed mechanisms for the endothelium-dependent relaxations are not yet unclear. However, Mokuboito and the constituents would increase [Ca] in endothelium cells and then activate NOS activity and PGI release. Thus, we conclude that Mokuboito, and sinomenine involve with EDRF and PGI, and regulate the Ca influx via L-type Ca channels on the cell membrane of aortic smooth muscle.
Mokuboito consists of (4 g), cortex (3 g), radix (3 g) and (10 g). Sinomenine is one of the constituents of . It is estimated that ∼66 μM sinomenine is contained in 1 mg ml of , corresponding to 2.2% (Pharmaceutical Institute of Tsumura Co.). at 1 mg ml caused the vasorelaxation by ∼46%. On the other hand, sinomenine at 66 μM estimated vasorelaxation by ∼60%. The relaxing strength of and sinomenine is not so different. Since both compounds may cause the vasorelaxation by similar mechanisms, sinomenine may play a main role for the vasorelaxation induced by . In the presence of propranolol, however, there were differences between the effects of and sinomenine. causes vasoconstriction, whereas sinomenine has no effect. This difference cannot be explained now, but there could be a component in different from sinomenine.
Although sinomenine may be considered as a key constituent of , we cannot necessarily explain the entire pharmacological effects of Mokuboito; because sinomenine is only 2.2% in , and only 0.6% in Mokuboito. Thus, there may be a great difference between the pharmacological potencies of Mokuboito and sinomenine included in Mokuboito. include only 20% in Mokuboito, but the potencies of Mokuboito and are almost similar. cortex and radix also possess vasodilating effects (,). Therefore, the complex interactions among cortex and radix would produce the pharmacological effects of Mokuboito as a whole. The whole effect as a Mokuboito is absolutely not a simple sum of each effect induced by its constituents.
Mokuboito is traditionally used for dyspnea and edema (,). Thus, Mokuboito may be expected as one of the therapeutic drugs for heart failure. Most recently, Satoh () has demonstrated the effective modulation of cardiac ion channels by Mokuboito, and sinomenine. Sinomenine inhibits I, and simultaneously produces the I decrease in cardiomyocytes, resulting in the APD prolongation. And sinomenine also exhibits an antiarrhythmic action on dysrhythmias under Ca-overloaded conditions. The antiarrhythmic effect would be due to the APD prolongation and the [Ca] reduction by the inhibitions in I and I channels.
Mokuboito, and sinomenine have multiple vasodilating mechanisms. The vasodilating actions are expected as one of the great useful tools for heart failure, and can regulate the preload and afterload of cardiovascular systems. The reduction of both loads is the essential therapeutic strategies for heart failure. Therefore, the vasodilating actions by Mokuboito, and sinomenine can sufficiently improve cardiac functions by reduction of the preload and afterload against heart failure.
In this study, Mokuboito, and sinomenine may regulate the Ca influx via L-type Ca channel, consistent with the previous report (). This may be a Ca channel inhibitor, because the effect of sinomenine was attenuated under the conditions of low Ca extracellular concentration (), as well as in the presence of nicardipine. Under ischemia and heart failure, cellular Ca overload of heart muscles elicits arrhythmias and dysfunctions (,). The Ca influx control may regulate cellular Ca, overload cardiomyocytes and produce the protective actions for Ca-overloaded cells (). Thus, Mokuboito, and sinomenine might protect cell damages of heart muscles via regulation of [Ca], and as a result, would exert a cardioprotective action. Further experiments are needed to elucidate in more detail the mechanisms for pharmacological actions. |
Inflammatory changes of the oral mucosa are a common side effect of cancer chemotherapy. According to the guidelines for prophylaxis and therapy of mucositis () of 2004, the risk to develop a grade 3 (severe) or grade 4 (life threatening) oral mucositis lies between 2 and 66%, depending on the type of chemotherapy (without additional radiotherapy), and between 1 and 53% depending on the type of tumor. In more than one-third of the patients who develop a grade 3 or 4 oral or gastrointestinal mucositis, the start of the next chemotherapy cycle needs to be delayed (), resulting in impaired cancer treatment.
Most of the cytostatic agents used for antitumor treatment exert their toxic properties especially on rapidly proliferating cell lines. Basal cells of the oral mucosa show a high rate of proliferation which is comparable to that of tumor cells. This is why chemotherapy affects both the oral mucosa by decreasing the regeneration of basal cells () and the immune system as a result of myelosuppression. Both factors together promote the development of diverse inflammatory changes in the oral mucosa under chemotherapy, will be summarized as ‘mucositis’. Even if the underlying interactions on the level of molecular immunology are not yet completely understood, there is some evidence that e.g. the administration of GM-CSF (granulocyte-macrophage colony stimulating factor) results in clinically relevant improvement of mucositis (). It has been shown that individual oral hygiene has considerable influence on symptom presence and severity of oral mucositis ().
Common dental care products offer several different active ingredients to reduce bacterial counts and plaque. Fluorides have antibacterial (sodium fluoride, amine fluoride, tin fluoride) () and plaque decreasing effects (amine fluoride and tin fluoride). Chlorhexidine is antimicrobial and effective in inhibiting plaque growth (). However, because of side effects (e.g. mucosal irritation) it is not advised for long-term use (,). The same holds true for sanguinarine, which is an antimicrobial alkaloid extracted from Canada Bloodroot or Paccoon plants () ().
Kationic detergents such as cetylpyridiniumchloride have antimicrobial and plaque-inhibiting effects (), and so do some volatile oils such as menthol, eucalyptol and thymol ().
The principal active ingredients of the preparations used in this study are extracts from rathania root and myrrh, which are indicated as phytotherapeutics for local treatment of slight inflammations of the mucous membrane in mouth and larynx. In Germany, those extracts are approved by the authorities as ‘Standardzulassung’ (). The primary effect of ratanhia () root extracts is tissue astringency. Myrrh tincture, an alcoholic solution of commiphora resins, shows additional antibacterial properties ().
Further ingredients of Weleda Ratanhia-Mundwasser® (ratanhia mouthwash) are horse chestnut extract, volatile oils to stimulate the mouth mucosa and homeopathic components used in anthroposophic medicine to strengthen dental structure and to promote tooth enamel production (Fluoride of Calcium D10, Argentum vitr. D15, Sulfate of Magnesium D20, horse chestnut bark D20).
Weleda Pflanzen-Zahngel® (herbal dentifrice, toothgel) contains extracts from ratanhia roots, myrrh and chamomile, delivering a broad range of effects with primarily anti-inflammatory, antibacterial and lesion healing properties (,). Volatile oils (from peppermint, spearmint and fennel seeds) are supposed to stimulate the mucosa.
It is the aim of this prospective study to document the effects of dental treatment and regular use of mouthwash and herbal dentrifice on incidence and severity of oral mucositis under chemotherapy.
#text
Since all 49 patients who were eligible insisted on participation in the active group, no control group could be established. This bias had unfortunately to be accepted, since blinding was impossible due to the astringent effects and the typical taste of the herbal ingredients, and active compliance of the patients was required to ensure correct administration.
Due to their poor overall condition (cancer patients), only 32 of the 49 patients enrolled in the medication phase were able to attend to all five study visits. Only these patients were included in the analysis ().
Our analysis was purely documentary in intention, not interventive. Additionally, incidence and severity of oral complications under chemotherapy strongly depend on the type of agents used, the therapeutical regimen administered, and on the patients' individual baseline conditions (). Still, considering the extensive and detailed representation of mucositis in the abovementioned guidelines (), a relative appraisal of the results of this study is deemed essential.
Professional tooth cleaning procedures immediately after the baseline visit served to provide equal and comparable dental baseline conditions for all participants. In order to ensure consistency in the evaluation of the dental parameters and indices, all investigations were conducted by the same treating dentist. Considering the mean patient age of 58.9 years, the baseline extent of caries as delineated by the DMF-T index was well within the normal range. Average values in Germany were listed as 16.1 for 35–44 year old persons and as 23.6 for 65–74 years of age in 1997 ().
Both plaque and gingiva indexes were below baseline values after four weeks of treatment and thus demonstrated good oral hygiene effectiveness of thrice daily administration of Weleda Rantanhia mouthwash and Weleda herbal tooth gel.
As expected, the patients showed various signs of damage of oral mucosal cells during the first week of chemotherapy. This is a common fact with cytostatic agents that are known to harm all kinds of mucosal cells. Until the final observation visit after four weeks, they tended to recede in spite of continued chemotherapy with the exceptions of gingival bleeding and mucosal redness, rendering an overall positive development.
More than 70% of the patients remained asymptomatic of mucositis even after four weeks. In only 15.6% of all patients mucositis was increased in severity by one degree, only one patient presented grade 2 severity and none grade 3.
On the whole, the symptoms could be described as moderate in severity. A basis for comparison is found which names a relative risk for developing grade 3 or 4 mucositis of 3 to 13% after administration of the cytostatic drugs that were observed. The overall risk for breast cancer patients was mentioned as being 8%.
Since for these data, as for non-study patients, basic common oral hygiene can be safely assumed, the observed mildly positive effects are probably due to thrice daily astringent mouthwash administration, which is not part of everyday dental care routine.
On the other hand, it is highly improbable that the observed symptoms of oral mucositis were caused by the mouthrinse itself. It is generally known that oral mucositis is an almost inevitable side effect of cancer chemotherapy. On the other hand, this mouthwash is widely sold all over Europe for normal daily oral hygiene and has not been reported to cause such side effects in thousands of administrations.
In their self-assessments, none of the patients complained about worsening symptoms between day 7 and day 28 of chemotherapy. Most patients (86.7%) judged the severity as unchanged, and 13.3% experienced some degree of relief. Almost all patients (93.3%) wanted to continue the use of the herbal mouthwash and tooth gel, and all of them voted in favor of professional dental assistance under chemotherapy.
Paralleling these results, the mucositis guidelines stress the necessity of basic oral hygiene and the efficacy of oral care under professional guidance (oral care protocols and patient guideline, evidence level III b) ().
The results of this study suggest that the use of Weleda Ratanhia mouthwash and Weleda herbal dentrifice may have a positive influence on the oral side effects of cancer chemotherapy, and that further investigations might be desirable. |
Stress is an extremely broad term that serves to define a wide variety of phenomena that humans are exposed to throughout their lives. The facets of stress are essentially limitless; however, broad categories include physical exertion, emotional upset, persistent psychological pressure, existential crisis and the residual effects of emotional trauma. The 1993 World Labour Report by the UN's International Labour Organization claimed that stress had become one of the most serious health issues of the 20th century (). Stress and its resultant comorbidity are responsible for a large proportion of disability worldwide. It has consistently been shown that individuals experiencing stress have impaired physical and mental functioning; more workdays lost, impairment at work and increased use of healthcare services. The disability caused by stress is just as great as the disability caused by workplace accidents or other common medical conditions such as hypertension, diabetes and arthritis (). The World Health Organization (WHO) Global Burden of Disease Survey estimates that mental disease, including stress-related disorders, will be the second leading cause of disability by the year 2020 ().
Since the characterization of the generalized adaptation syndrome (GAS) by Hans Selye, the issue of stress has come to be viewed as a risk factor in both the etiology and progression of a wide variety of diseases. Chronic and post-traumatic stress has been implicated in cardiovascular disease (–), gastrointestinal ulceration (), inflammatory bowel disease (), depression (,), anxiety (), fibromyalgia (), autoimmune disease () and schizophrenia (). In addition, chronic stress clearly demonstrates the ability to negatively impact immune system function and can predispose a person toward infection (–). Evidence also demonstrates that stress very early in life, i.e. within the first few years of childhood, may be an etiological agent for the development of type 1 diabetes () and also potentially in the development and severity of asthma ().
Normally the stress response is activated on an acute basis and the body is able to return to normal parasympathetic homeostasis. Stress becomes pathologic, however, when it is chronic or when the body acquires an inherent tendency towards activation of the sympathetic nervous system (,). Current therapies for dealing with ‘stress’ are extensive but perhaps not ideally targeted in most cases. Pharmacological approaches are principally focused on the treatment of depression and the manifestations of both acute and chronic anxiety disorders ().
A potentially beneficial use of herbal medicine involves the use of herbs as ‘adaptogens’ in order to prevent stress-induced morbidity. From an herbalist's perspective, the therapeutic intent of an adaptogen is to promote an optimal physiological response to both internal and external stresses. Essentially the adaptogen supports the body's ability to ‘adapt’ ideally to its environment. These herbs are believed to have a bimodal function of action either by providing a stimulant effect or a sedative effect depending on the needs of the individual in a particular situation. Common examples of adaptogens include Asian ginseng (), Siberian ginseng (), Ashwagandha root ( and rhodiola () (). Herbs are not the only potential adaptogens as other agents like vitamins and even amino acids have also exhibited this kind of activity ().
In the 1950s, Russian scientists set out specific criteria that an herb had to incorporate in order to be considered an adaptogen. The criteria were that the herb must (i) only cause minimal disorders in physiological function; (ii) must increase resistance to external and internal stresses through a wide range of physical, chemical, and biochemical factors; and (iii) must have a normalizing effect, improving all types of conditions while aggravating none (,). Specific examples of how adaptogens have been taken to improve health include their usage to: reduce fatigue, regain or improve athletic performance, regain or improve mental acuity, increase appetite, relieve insomnia, improve mood and increase resistance against infection (,,). For an intriguing discussion on how adaptogens may impact health, the reader is referred to a series of articles published by Olalde Rangel . in 2005 (–).
Relatively little clinical research has been conducted on the use of herbal adaptogens for stress, and still less on combinations of herbal medications. In this paper, we report on the results of an open-label uncontrolled longitudinal trial on the use of a proprietary herbal compound mixture, OCTA©, on subjective parameters of stress as well as objective measures.
Participants were provided with the proprietary herbal compound OCTA©, an aqueous-based liquid herbal preparation consisting of eight herbs as follows: and (listed in descending order according to respective concentrations). Participants were instructed to consume two tablespoons (30 ml) of the herbal compound per day each morning for a total period of 3 months.
The product has been through quality control testing and has demonstrated consistent concentrations of the herbal extracts. The two main ingredients are derived from standardized extracts ( 2.5% with anolides; and : 1% corosolic acid). OCTA© has been under review by Health Canada's Natural Health Products Directorate (NHPD) since 2004 but due to a backlog of requests has yet to be processed. The manufacturer adheres to GMP standards and is certified by Australia's Therapeutics Goods Administration (TGA), the Department of Health of the State of New Jersey, and has received an ‘A’ rating from the National Nutritional Foods Association (NNFA).
The SF36 Health Survey (SF-36v2) questionnaire is one of the most extensively used and validated quality of life measures. It is a self-administered, 36-item questionnaire that measures health-related quality of life in 8 domains that consist of both physical and mental-related issues. The physical and mental scores can be summarily added in order to obtain overall summaries of each of these two distinct areas. The reliability of the test is well documented with a Cronbach's alpha value greater than 0.85 ().
The State-Trait Anxiety Inventory (STAI) was initially conceptualized as a research instrument for the study of anxiety in adults. It is a self-reported assessment device that includes separate measures of state and trait anxiety. According to the author of this tool, state anxiety reflects a ‘transitory emotional state or condition of the human organism that is characterized by subjective, consciously perceived feelings of tension and apprehension, and heightened autonomic nervous system activity’. State anxiety may fluctuate over time and can vary in intensity. In contrast, trait anxiety denotes ‘relatively stable individual differences in anxiety proneness…’ and refers to a general tendency to respond with anxiety to perceived threats in the environment ().
Results were analyzed using one-way ANOVA. Scores for each test as well as the objective samples were independently assessed and only in the case of the SF-36v2 was the test divided into physical and psychological summary categories.
We enrolled 18 participants for the 12 week period. Seventeen participants completed the trial. One participant withdrew due to scheduling conflicts. displays the overall effects of OCTA© on primary subjective outcomes. reports results from the SF-36v2 Quality of Life results in all eight domains. displays selected objective outcome measures including serum DHEA levels, urinary creatinine and levels of the liver enzymes AST and ALT. No significant changes in blood pressure and cholesterol were observed in any of the participants (data not shown). None of the participants reported self-perceived adverse events, nor were any observed by the monitoring physician.
For our primary outcomes, we observed a 46% improvement in self-perceived stress. The state and trait anxiety scale reported overall changes from baseline of 40 and 26%, respectively. The Beck Depression Index improved by 67%. For the SF-36 questionnaire, the mental summary score improved by 72%. Each of these results was statistically significant with -values all <0.0001. The physical summary of the SF-36 demonstrated a modest improvement of 9%; however, this was not statistically significant ( = 0.228). ANOVA of serum DHEA demonstrated no change from baseline ( = 0.98).
e
f
i
n
d
i
n
g
s
o
f
t
h
i
s
t
r
i
a
l
s
h
o
u
l
d
b
e
o
f
i
n
t
e
r
e
s
t
t
o
p
h
y
s
i
c
i
a
n
s
a
n
d
p
a
t
i
e
n
t
s
a
l
i
k
e
.
W
e
f
o
u
n
d
a
n
o
v
e
r
a
l
l
i
m
p
r
o
v
e
m
e
n
t
i
n
p
e
r
c
e
p
t
i
o
n
s
o
f
s
t
r
e
s
s
a
n
d
q
u
a
l
i
t
y
o
f
l
i
f
e
d
u
e
t
o
c
h
a
n
g
e
s
a
s
s
o
c
i
a
t
e
d
w
i
t
h
s
t
r
e
s
s
a
s
m
o
n
i
t
o
r
e
d
b
y
v
a
l
i
d
s
t
r
e
s
s
a
n
d
q
u
a
l
i
t
y
o
f
l
i
f
e
m
a
r
k
e
r
s
.
e
l
a
c
k
o
f
a
d
v
e
r
s
e
e
f
f
e
c
t
s
f
o
u
n
d
i
n
o
u
r
s
t
u
d
y
a
s
w
e
l
l
a
s
a
n
a
b
s
e
n
c
e
o
f
c
h
a
n
g
e
s
i
n
b
o
t
h
l
i
v
e
r
a
n
d
r
e
n
a
l
f
u
n
c
t
i
o
n
t
e
s
t
s
t
e
n
t
a
t
i
v
e
l
y
s
u
p
p
o
r
t
s
t
h
e
n
o
t
i
o
n
t
h
a
t
t
h
i
s
h
e
r
b
a
l
m
i
x
t
u
r
e
i
s
n
o
n
-
t
o
x
i
c
.
W
e
l
l
-
c
o
n
t
r
o
l
l
e
d
a
n
d
l
o
n
g
e
r
-
t
e
r
m
s
t
u
d
i
e
s
a
r
e
n
e
e
d
e
d
,
h
o
w
e
v
e
r
,
t
o
e
s
t
a
b
l
i
s
h
t
h
e
l
o
n
g
-
t
e
r
m
e
f
f
e
c
t
s
o
f
t
h
i
s
m
i
x
t
u
r
e
b
o
t
h
w
i
t
h
r
e
g
a
r
d
s
t
o
e
f
f
i
c
a
c
y
a
n
d
t
o
x
i
c
i
t
y
.
I
n
a
d
d
i
t
i
o
n
,
t
h
e
s
m
a
l
l
n
u
m
b
e
r
o
f
p
a
r
t
i
c
i
p
a
n
t
s
i
n
o
u
r
p
i
l
o
t
s
t
u
d
y
d
o
e
s
n
o
t
p
r
e
c
l
u
d
e
t
h
e
p
o
s
s
i
b
i
l
i
t
y
t
h
a
t
s
e
r
i
o
u
s
s
i
d
e
e
f
f
e
c
t
s
m
i
g
h
t
o
c
c
u
r
.
W
h
i
l
e
i
n
c
o
n
c
l
u
s
i
v
e
,
t
h
e
r
e
s
u
l
t
s
f
r
o
m
t
h
i
s
s
t
u
d
y
a
r
e
e
n
c
o
u
r
a
g
i
n
g
.
A
l
b
e
i
t
u
n
c
o
n
t
r
o
l
l
e
d
,
t
h
e
p
i
l
o
t
s
t
u
d
y
i
n
d
i
c
a
t
e
s
t
h
a
t
a
p
o
l
y
h
e
r
b
a
l
c
o
m
p
o
u
n
d
,
O
C
T
A
©
,
p
r
o
v
i
d
e
s
a
b
e
n
e
f
i
c
i
a
l
a
d
a
p
t
o
g
e
n
i
c
e
f
f
e
c
t
.
T
h
e
l
a
c
k
o
f
o
v
e
r
t
a
n
d
m
e
a
s
u
r
a
b
l
e
t
o
x
i
c
i
t
y
i
n
a
l
l
p
a
r
t
i
c
i
p
a
n
t
s
i
s
a
l
s
o
e
n
c
o
u
r
a
g
i
n
g
a
n
d
s
u
p
p
o
r
t
s
t
h
e
n
e
e
d
f
o
r
m
o
r
e
r
e
s
e
a
r
c
h
t
o
b
e
t
t
e
r
e
v
a
l
u
a
t
e
t
h
i
s
p
o
l
y
h
e
r
b
a
l
c
o
m
p
o
u
n
d
. |
italic
xref
fig
#text
Swiss albino male mice (), 6–8 weeks old with body weight of 24 ± 2 g, were used. Mice were obtained from Hamdard University, New Delhi, India and maintained and bred in mouse house, as an inbred colony as per norms laid down by an Institutional Ethical Committee, given standard mouse food and water .
Mice were irradiated by Co source in the cobalt teletherapy unit (ATC-C9) at Radiation Oncology Department, Sawai Man Singh Medical College and Hospital, Jaipur, India. The mice were kept in ventilated box with a distance of 77.5 cm from the source to deliver a dose-rate of 1.33 Gy min.
Fresh leaves of () were collected from the Herbarium, University of Rajasthan, Jaipur, India and shade-dried. Dried leaves were subjected to three changes of 80% ethanol at room temperature. The extracts were pooled, lyophilized, weighed and preserved at 4°C until use.
Mice were randomly divided into following groups (five per group):
For survival studies mice of both control and experimental groups were exposed to whole body gamma radiation (6, 8 and 10 Gy) and were checked daily for 30 days. The survival percentage of mice up to 30 days of exposure against each radiation dose was used to construct survival dose–response curves. Regression analysis was done to obtain LD values and to determine dose reduction factor (DRF).
Mice (all groups) were injected intraperitoneally with 0.025% colchicine and sacrificed 2 h later by cervical dislocation. Femurs were dissected out and bone marrow cells were aspirated and washed in physiological saline, treated hypotonically (0.075 M KCl), and fixed in Carnoy's fixative and stained with 4% Giemsa. Metaphase slides were prepared by air-drying method of Savage (). Chromosomal aberrations were scored using oil immersion (with 100× object lens) under a light microscope.
GSH in liver was measured using the method described by Moron . (). Liver homogenates were treated with 0.1 ml of 25% trichloroacetic acid (TCA) and the resulting precipitate was pelleted by centrifugation at 3900 for 10 min. Free endogenous sulfhydryl was assayed in a total 3 ml volume by adding 2 ml of 0.5 mM 5, 5′-dithio-bis(2-nitro benzoic acid) (DTNB) prepared in 0.2 M phosphate buffer (pH 8) to 1 ml of the supernatant. The GSH reacts with DTNB and forms a yellow-colored complex with DTNB. The absorbance was read at 412 nm using UV-VIS Systronic spectrophotometer.
LPO levels in testis and liver were estimated by the method of Ohkawa . () as thiobarbituric acid (TBA) reactive substances. The liver and testis were dissected out and chilled in ice cold 0.09% NaCl. Homogenate of desired tissues were prepared in 1.15% KCl (1 g tissue in 9 ml of 1.15% KCl). Sodium dodecyl sulfate (8.1%; 0.2 ml) was added to 0.2 ml of sample in test tubes and pH was adjusted to 3.5 with 5 M NaOH. To this, 1.5 ml of 0.8% aqueous solution of TBA was added. The mixture was made up to 4 ml with distilled water and heated at 95°C for 60 min. After cooling under tap water, 1 ml of distilled water and 5 ml a mixture of -butanol and pyridine (15:1) were added and shaken vigorously. The solution was centrifuged at 3900 for 10 min. The upper organic layer was removed and absorbance was measured at 532 nm using UV-VIS Systronic spectrophotometer.
Acid and alkaline phosphatase activities in testes were estimated using the method described by Fiske and Subbarow (). The tissue homogenates were mixed with TCA and then centrifuged at 3900 for 10 min. The supernatant was then treated with molybdate solution. (Molybdate solution was prepared by dissolving 25 g of ammonium molybdate into 200 ml glass distilled water (GDW) and combining with 300 ml of 10 N HSO and then was made up to 1000 ml with GDW.) This resulted in the formation of phosphomolybdic acid from the phosphate present in the tissue. The phosphomolybdic acid was then reduced by 1-anilino-8-naphthalenesulfonic acid (ANSA) to produce a blue color whose intensity was proportional to the amount of phosphate liberated. The alkaline phosphatase activity is the difference between inorganic phosphate content of the incubated and control samples expressed as Bodansky units. One Bodansky unit corresponds to the liberation of 1 mg of inorganic phosphorous from the tissue in mg g h ().
Student's -test was employed to analyze the results (). -values <0.05 were considered significant. Regression analysis was done to obtain LD values and to determine DRF.
leaves () extract (AE) was given as 100, 200, 400, 800 and 1200 mg kg body weight of mouse per day in DDW orally to Swiss albino mice for 15 consecutive days. The extract was non-toxic and no mortality was observed till day 30. An optimum dose of 800 mg kg body weight of AE was selected against 8 Gy radiations on the basis of maximum survivability as depicted in . The percent survival was significantly increased in AE-fed mice subsequently treated with irradiation (AE-pretreated irradiated mice). When survival data were fit on regression line equation, LD values for control (irradiated alone) and experimental (AE plus irradiation) were computed as 6.47 and 9.3 Gy, respectively. On basis of these LD values, AE pretreatment produced a DRF of 1.43.
As depicted in , in AE-treated mice, there was no significant change in body weight on days 1, 3, 7 and 14 while a significant increase ( < 0.01) was observed on day 30 when compared with mice fed with distilled water alone. Notably, in irradiated mice, the body weight was drastically decreased on day 3, ( < 0.001), day 7 ( < 0.001) and day 14 ( < 0.001) when compared with mice treated with distilled water only. In AE-pretreated irradiated mice, the body weight significantly increased on day 3 ( > 0.05), day 7 ( < 0.01) and day 14 ( < 0.001) of observation as compared to irradiated group. To note, by day 30, all mice died in radiation-treated group, but in AE-pretreated irradiated group, 70% mice survived.
Mice treated with extract alone showed no significant change in weight of the testis on days 1, 3, 7 and 14 (). However, on day 30, a significant weight increase ( < 0.05) was observed as compared to mice treated with distilled water. Mice treated with radiation alone (8 Gy; Group 3) showed reduction in the testis weight during all days of observation (day 1: < 0.05; day 3: < 0.001; day 7: < 0.001 and day 14: < 0.001). Whereas in -pretreated irradiated group (AE plus radiation), there was a significant increase in testis weight after day 3 ( < 0.05), 7 ( < 0.01) and 14 ( < 0.001) when compared with irradiated mice (). In irradiated mice, there was a drastic depletion of spermatogonial population with necrotic and pyknotic nuclei were observed (B) when compared with mice treated with distilled water (A). The germinal epithelium was highly disorganized with shrinkage of tubules and cytoplasmic vacuolization (C). Total absence of sperm and spermatids were observed. Sertoli cells and Leydig cells showed shrinkage in their size (D). While in mice pretreated with AE, less damage to spermatogonial population and germinal epithelium was observed with more rapid recovery (E). In irradiated mice, there was significant decrease in number of spermatogonia type A and type B was noticed on all days of observation (B–D). Similar decrease was also found in the number of primary spermatocyte, secondary spermatocyte and spermatid. Notably, in mice pretreated with AE and then exposed to radiation dose, the quality (as determined by intact germinal epithelium, no pyknosis, necrosis, karyolysis present, less cytoplasmic vacuolization) and number of germ cells increased by day 30, the histology of testis revealed near normal histoarchitecture except some cytoplasmic vacuolization and lumen with full of sperms (G).
The exposure to radiation caused severe cytogenetic damages in bone marrow cells (B) when compared with mice treated with distilled water (A). Various types of aberration like chromatid breaks, chromosome breaks, fragments, and rings, chromosome exchange, and dicentric characteristics were observed in irradiated group (B). In contrast, in AE-pretreated irradiated mice, a significantly lesser degree of these aberrations were observed (C).
Glutathione level in liver was found to significantly increase at autopsy intervals of day 7 ( < 0.05), day 14 (0.001) and day 30 ( < 0.001) in -treated mice when compared with mice treated with distilled water (). In irradiated group, the GSH content showed significant decrease at all autopsy intervals. Notably, as shown in , in AE-pretreated irradiated mice, a significant increase in GSH content was observed at all intervals when compared with irradiated mice.
In extract (AE)-treated mice, a significant decrease in LPO level in liver was noticed on day 14 ( < 0.01) and day 30 ( < 0.05) as compared to those treated with distilled water (). In irradiated group, LPO level was increased ( < 0.001) at all autopsy intervals of observation in terms of thiobarbituric acid reactive substances (TBARS). On the other hand, in AE-pretreated irradiated mice, a significant inhibition in LPO level was observed ().
The LPO level in testis of -treated mice was significantly ( < 0.05) decreased on day 14 and day 30 as compared to those treated with distilled water (). In irradiated group, a noticeable elevation of LPO was observed at all intervals from day 1 to day 14. In AE plus radiation experiment group, a significant ( < 0.001) reduction in LPO level was seen from day 3 onwards as compared to irradiated group ().
Alkaline phosphatase activity showed no significant changes in -treated mice as compared to those treated with distilled water except on day 30 there was a significant increase (). In irradiated group, the alkaline phosphatase activity in testis showed remarkable and significant decline ( < 0.001) on all days of observation. In AE-pretreated irradiated mice, a significant recovery in alkaline phosphatase activity was observed.
The acid phosphatase activity in testis was found at normal level in group as compared to mice treated with distilled water at all autopsy intervals (). In irradiated group, a highly significant ( < 0.001) elevation in the enzyme level was observed. In AE-pretreated irradiated mice, a significant ( < 0.001) decline in acid phosphatase activity was observed at all autopsy intervals in comparison to irradiated mice.
italic
#text |
Cancer is the leading cause of mortality worldwide and the failure of conventional chemotherapy to effect major reduction in the mortality indicates that new approaches are critically needed (). An extremely promising strategy for cancer prevention today is chemoprevention, which is defined as the use of synthetic or natural agents to block the development of cancer in humans (). A variety of bioactive compounds and their derivatives have been shown to inhibit carcinogenesis in a number of experimental systems involving initiation, promotion and progression (,). Plants, vegetables and herbs used as folk and traditional medicine have been accepted currently as one of the main sources of cancer chemoprevention drug discovery and development (). In Brazil traditional system of medicine supports a growing interest in the pharmacological evaluation of various plants.
Species belonging to the Fabaceae, the Mill () have been used as an infusion or decoct (flavor extract by boiling 1 liter of hot water per 5 g of leaves) (). This plant is found in tropical and subtropical areas and well adapted to growth in semi-arid regions and soil of low fertility (–). It occurs in Northeast Brazil and has intensive popular use in the treatment of inflammation (,) and other diseases such as epilepsy in human () and in animal models (–). Recently, embryotoxic effects and antimicrobial activity have been reported (,). A chemical investigation of extract of leaves of in Natural Products Alert () and Chemical Abstracts databases have revealed the presence of alkaloids, flavanoids, steroids, proteins, carbohydrates and indigo. Antitumor activities have been reported in several plant species (,,–), however, up to now, few researches have been done to investigate this traditionally used plant in the recognition of their mechanism, guaranteeing in the future its scientific and therapeutic use. For allopathic drug development, even when traditional formulations are taken into consideration, traditional medical systems are very rarely regarded in the same way (). The aim of the present study was to carry out a brief basic toxicological analysis and establish the safety of aqueous extract of leaves of focusing on its cytotoxic and antitumor activities.
The leaves of were collected in June 2000 in São Caetano, State of Pernambuco/Brazil and authenticated by the Biologist Marlene Barbosa from the Botany Department, Federal University of Pernambuco (UFPE), Brazil. A voucher specimen number 32 859 has been deposited at the Herbarium of the Botany department.
Male Swiss albino mice weighing 20–25 g were purchased from the animal house of Centro de Pesquisas Aggeu Magalhães–Pernambuco, Brazil. They were housed in standard environmental conditions of temperature, humidity and under clear and dark cycles of 12 h. The mice were fed on diet of the biotery (LABINA Purina Brazil) and water . All procedures described were reviewed and approved by the University Animals Ethical Committee.
Sarcoma 180, solid tumor and HEp-2 (human epidermoid cancer cell) cell lines were obtained from Department of Antibiotics/UFPE, Brazil. The solid tumor was maintained in Swiss albino mice and the HEp-2 cells in minimum essential medium, DMEM (Dulbeccos' Modified Eagle medium).
Two extracts were prepared by infusion and maceration from 150 g of leaves of . The leaves were weighed; chopped and extracted with solvents and water. The infusion was prepared with 75 g of fresh leaves in 2 × 200 ml of increasing polarity solvents (hexane, ethyl acetate and methanol) at 40°C for 10 min and removing solid matter by filtration. After this preliminary step, the same plant material was extracted in boiling distilled water using the same conditions, and the maceration obtained following the above-mentioned process at room temperature (28°C) overnight. The solvents were removed by rotary evaporation. The yields (w/w) of the infusion and the maceration were hexane (0.67 and 0.74%), ethyl acetate (0.39 and 0.34%) methanol (3.9 and 1.88%) in terms of newly collected plant material. After lyophilization, the aqueous extracts yielded 4.20 and 1.75% and the dried material was stored at −20°C (,). The aqueous extracts by infusion and maceration were used for cytotoxic and antitumor activities.
The HEp-2 cells (human epidermoid cancer cells) were investigated by the MTT method [3-(4,5-dimetyl (thiazol-2-yl)-2,5 diphenyltetrazolium bromide)]. Extracts with concentrations more than 30 mg ml are considered citotoxic (). They were trypsinized, counted and prepared in a suspension with 10 cells per ml of DMEM and distributed in a plate with 96-wells, which was incubated at 37°C in a humidified atmosphere for 24 h. The aqueous extracts of leaves of obtained by infusion and maceration were dissolved in DMSO (dimethylsulfoxide) in concentrations of 6.25, 12.5, 25 and 50 μg ml and put in the wells with the HEp-2 cells. Each concentration was tested in quadruplicate. As the control, DMEM with DMSO was used. After 72 h, 25 μl of MTT and 5 mg ml of PBS was added to the wells and the plate was incubated for 2 h. The optical density was measured at 550 nm with ELX 800 reader.
Thirty-six healthy Swiss albino mice male, weighing 20–25 g, were divided in groups of six. The animals were on fasting for 18 h before being submitted to the experiment. The aqueous extract of by infusion was dissolved/suspended in distilled water and administered by the intraperitoneal (i.p.) route in doses of 50, 150, 300, 600, 1200 and 2400 mg kg. The general behavior of mice was observed continuously for 1 h after the treatment and then intermittently for 4 h, and thereafter over a period of 24 h (). The mice were further observed for up to 14 days following treatment for any signs of toxicity and deaths. The LD value was determined according to the method of Litchfield and Wilcoxon ().
Male albino Swiss mice were divided into six groups of five animals. In all groups 0.3 ml of Sarcoma 180 cells from a solid tumor (around 3 × 10 cells) i.p were injected. After 48 h of the implant, 0.2 ml kg i.p. of saline solution was administered in control group (groups 1 and 2). The chemotherapy was initiated making use of the aqueous extract of leaves of by infusion (groups 3 and 4) and by maceration (groups 5 and 6) in daily concentration of 50 mg kg i.p. for seven consecutive days. The dose of the extract was based on the LD On the 8th day, the mice were sacrificed for analysis of tumor development.
The experimental results of antitumor assays were expressed as Median (min–max). Data were assessed by ANOVA followed by Kruskal–Wallis. < 0.001 was considered as statistically significant.
Aqueous extracts of leaves of by infusion and maceration in concentrations of 6.25–50 μg ml were tested on HEp-2 cell lines by MTT method. The aqueous extracts did not produce any cytotoxic effect to HEp-2 cell lines (>30 μg ml) when compared with control DMEM and DMSO ().
There were no deaths but some low signs of toxicity were observed after i.p. injection of aqueous extracts of leaves of by infusion at any dose level up to the highest dose tested (2400 mg kg) whose effects were more pronounced (). Some adverse effects, such as agitation, piloerection, exhaustion and sleepiness, were seen immediately after the i.p. injection while others (irritability, exhaustion, agitation and spasm) were observed later, and they were more pronounced at the higher dose. The acute toxicity (LD) of aqueous extract of leaves of by infusion at different doses in mice did not show rates of mortality during 72 h of observation in the preliminary assay.
The effects of the aqueous extracts of leaves of by infusion and maceration on Sarcoma 180 is shown in . The extract by infusion reduced significantly the mean volume of Sarcoma 180 via i.p. administration in dose of 50 mg kg 64.53% [0.64 mg (0.01–22.10)] and the maceration extract reduced 62.62% [(0.64 mg (0.01–2.10)] at the same dose when compared to the mean volume of the tumor of control group treated with saline solution which showed a tumor development of 100% [1.97 (1.27–2.87)]. As shown, the inhibition was dose-dependent and similar to that promoted by the maceration and infusion extracts at the same dose.
The treatment of cancer may benefit from the introduction of novel therapies derived from natural products. Natural products have served to provide a basis for many of the pharmaceutical agents in current use in cancer therapy (). The use of chemotherapeutic drugs in cancer involves the risk of life threatening host toxicity. The search, therefore, goes on to develop the drugs, which selectively act on tumor cells. The plants belonging to family Fabacea have medicinal properties, especially the plant (,,). The present investigation shows that aqueous extracts revealed absence of cytotoxic effects, low toxicity dose-dependent and antitumor effect. Our results concerning the aqueous extracts of leaves obtained by infusion and maceration on HEp-2 cell lines did not reveal cytotoxic potential. Reports showed that cytotoxicity with extracts of leaves of were not encountered in the literature and it confuses the comparison of these results with others using the same conditions. Extracts of Umbelliferous plants retarded the development of solid and ascites tumors and increased the life span of these tumor-bearing mice. Tritiated thymidine uridine and leucine incorporation assay suggested that fraction acts directly on DNA synthesis (). Extract of was cytotoxic on tumor cells with reduction in tritiated thymidine but not to normal human lymphocytes (). Many pharmacological effects observed in animals can extend result in high value of application for the human species. Based on cytotoxicity bioassays, over 400 compounds have been isolated from plants, marine organisms and microorganisms between 1996 and 2000 (). As a result, this is the main reason for large use of toxicological test for determination of toxicity and safety when using drugs (). The aqueous extract of leaves obtained by infusion showed low order of toxicity in mice at least up to the maximum dose of 2400 g kg and no death but some low signs of toxicity.
A number of species of contain an amino acid indospicine (,), and nitropropanoyl esters of glucose (,) as some natural toxic product. Phytochemical analysis suggests that the presence of biologically activity compounds (alkaloids, steroids, flavanoids, proteins—lectin, carbohydrates, indigo, etc.) in the aqueous extracts of leaves of could be correlated to anti-inflammatory and antimicrobial activities (,). Biological activities of the compounds detected in the aqueous extracts of leaves of by infusion and maceration could be linked to antitumor activity.
Aqueous extracts of leaves of by infusion and maceration were used on Sarcoma 180 in mice at dose of 50 mg kg i.p. The aqueous extracts showed a tumor reducing activity. Our results concerning the aqueous extracts of leaves of corroborate with those found in aqueous extract of (Eupherbiacase), which showed a reduction of solid tumor in mice. The extract inhibits cell cycle regulating enzymes cdc 25 phosphatase (,).
Actual mechanism by which aqueous extracts of leaves of by infusion and maceration showed antitumor activity is not known. The results of the present study indicate that antitumor activity of aqueous extracts of may be due to its interference with cell development. This will be the object of future researches as modulating lipid peroxidation and augmenting antioxidant defense system (superoxide dismutase and catalase) (,,), DNA synthesis (), interaction with cell cycle regulation () and non-steroidal anti-inflammatory activity (). More recently, molecular biomarkers have been used for oncological management as prohibitin, mortalin and HSP 60/HSP 10 (); hyaluronic acid (HA) (); epidermal growth factor receptor (EGFR) and cytokeratins (); research of tumor angiogenesis with CD105 antigen (); alpha-fetoprotein (AFP) (); cyclooxygenase-2 (COX-2) () and immunohistochemical markers, such as CD34 and CD117 (). Besides assessment of prognostic factors as clinical (tumor growth rate and inflammatory signs), to also analyze parameters like hemoglobin content and red blood cell count, and histological (tumor stage, grading, tumor necrosis, lymph nodes status and margins status).
Same mechanisms of aqueous extract of leaves of by infusion may be involved in the embryo development in mice (). The first cleavages in mammalian embryogenesis are symmetrical mitotic divisions that increase the number of blastomeres by partitioning the oocyte without a net change in embryo size. Through the 8-cell stage the blastomeres of murine embryos are totipotent (–). The cell cycle of the two-cell embryos in the mouse has duration of 20–26 h with G and S phases that last ∼1–2 and 6–7 h; these phases are followed by an extended G phase (12–15 h) and M (mitosis) phase of 1–2 h (). Many plant extracts have been used as a source of medicinal agents to cure urinary tract infections, cervicitis vaginitis, gastrointestinal disorders, respiratory diseases, cutaneous problems, helmintic infections, parasitic protozoan diseases, and antitumor and antimicrobial (,–). The results of this investigation may improve our understanding in usage of this plant as an alternative anticancer therapy. Identification of the active principles and their mechanisms of action remain to be studied. This is a promising plant for further studies toward drug development. |
A certain type of empirical medicine called ‘Traditional Chinese Medicine’ or ‘Oriental Medicine’ is widely flourishing in the US and in Europe with the claim that it presents a frame of reference ‘completely foreign to the West’ (Kaptchuk) (). And for its exotic terminology, its peculiar diagnosis and its unconventional treatment modalities, Chinese Medicine is being perceived as such by the Western health-consumer. A cognitive-epistemic review, however, does not support this claim of originality and points to remarkable infrastructural similarities between Chinese Medicine and the once-prevalent pagan (Latin, , ‘from the country’) beliefs of well-being and safety in pre-Christian Europe. This resemblance has actually been the subject of a variety of comparative studies, which range from the hypothesis of ‘convergence’ (Lloyd), to that of ‘transmission,’ as argued by Paul Unschuld who, based on phonetic resemblances, maintains that the elusive Qí Bó, the main interlocutor in the Yellow Emperor's Canon of Internal Medicine, might have been Hippocrates ().
The intent of this article is to examine the fundamental archetypes/prototypes of pattern recognition and categorization in Chinese natural philosophy, and to point out that one of them is also found in the Pythagorean cult of and in the health and safety beliefs of the Old Religion of Northern Europe. Common inheritance, convergence or transmission, this architectonic similarity—in any case—further supports Paul Unschuld's view that ‘the notion of a vast dichotomy between Western and Chinese reactions to disease is completely unfounded’ ().
i
s
e
s
s
a
y
u
s
e
s
c
o
n
c
e
p
t
s
f
r
o
m
C
o
g
n
i
t
i
v
e
S
c
i
e
n
c
e
,
C
o
n
t
i
n
e
n
t
a
l
P
h
i
l
o
s
o
p
h
y
a
n
d
E
p
i
s
t
e
m
o
l
o
g
y
,
a
n
d
D
i
s
c
r
e
t
e
a
n
d
D
y
n
a
m
i
c
a
l
S
y
s
t
e
m
s
M
a
t
h
e
m
a
t
i
c
s
t
o
d
e
c
o
n
s
t
r
u
c
t
t
h
e
i
n
f
r
a
s
t
r
u
c
t
u
r
e
o
f
C
h
i
n
e
s
e
m
e
d
i
c
a
l
t
h
o
u
g
h
t
,
a
n
d
t
o
r
e
v
e
a
l
i
t
s
c
o
g
n
i
t
i
v
e
m
a
p
a
n
d
e
p
i
s
t
e
m
i
c
a
r
c
h
i
t
e
c
t
u
r
e
.
H
i
s
t
o
r
i
c
a
l
s
u
g
g
e
s
t
i
o
n
s
t
h
a
t
f
o
l
l
o
w
h
a
v
e
t
h
e
s
o
l
e
p
u
r
p
o
s
e
o
f
p
r
o
v
i
d
i
n
g
a
s
t
i
m
u
l
u
s
f
o
r
f
u
r
t
h
e
r
d
i
s
c
u
s
s
i
o
n
a
n
d
r
e
s
e
a
r
c
h
.
T
h
e
h
i
s
t
o
r
i
c
a
l
a
n
d
p
h
i
l
o
s
o
p
h
i
c
a
l
s
o
u
r
c
e
s
a
r
e
t
h
e
E
n
g
l
i
s
h
s
e
c
o
n
d
a
r
y
l
i
t
e
r
a
t
u
r
e
b
y
p
r
o
m
i
n
e
n
t
W
e
s
t
e
r
n
s
i
n
o
l
o
g
i
s
t
s
.
A
l
l
C
h
i
n
e
s
e
t
e
r
m
s
a
r
e
w
r
i
t
t
e
n
i
n
P
i
n
y
i
n
a
n
d
o
n
l
y
A
n
g
l
i
c
i
z
e
d
n
a
m
e
s
o
f
C
h
i
n
e
s
e
C
l
a
s
s
i
c
s
a
r
e
u
t
i
l
i
z
e
d
.
R
o
m
a
n
i
z
e
d
a
n
d
A
n
g
l
i
c
i
z
e
d
G
r
e
e
k
w
o
r
d
s
a
n
d
L
a
t
i
n
t
e
r
m
s
a
r
e
i
n
i
t
a
l
i
c
s
.
From the standpoint of knowledge architecture and representation, the core cognitive frame of reference in Chinese medical discourse is a combination of two key dynamical-processes pattern categorization schemas that seem to have been grafted together to form a hybrid system of thought. One emanates from ancient Chinese metaphysics that views ontogenesis (in the philosophical sense of the term) as the product of a wavering between an inhibiting and an activating force, called Yin and Yang, and symbolized by a broken (- -) and a solid (—) line, respectively. In this model, spatiotemporal events are assessed based on their dynamic state called ‘Qi’ (pronounced Ch’i), and plotted against a hierarchical and tree-like system similar to what is known as a ‘period-doubling bifurcation’ in Dynamical Systems Mathematics. For instance, at a second level of bifurcation, there are four possibe qualitative outcomes in a dynamic event, such as in the course of a disease: Yang-Yang, Yang-Yin, Yin-Yang and Yin-Yin periods (). According to the English secondary literature, one of the earliest references to this model is found in the I-Ching translated as the ‘Book of Changes’ or the ‘Classic of Changes,’ in which 2, or 64, topologies/combinatorics of Yin and Yang called ‘Hexagrams’ () are itemized and described, seemingly for the purpose of cleromancy and divination.
As for the epistemological merits of using a hierachical inhibiting-activating model to conceptualize biological processes, Weiss, Qu and Garfinkel of University of California point out that it can only constitute a ‘minimal model,’ with limited abilities to directly relate its predictions to nature's dynamical and emergent behavior, and that often biological events do not have explicit counterparts in it. Although their general assessment is not based on Chinese thought but on Alan Turing's purely mathematical model as described in his 1952 article ‘The Chemical Basis of Morphogenesis,’ it could nonetheless be applied to the Yin-Yang model, given that these empirical notions appear as being analogous to Turing's ‘short-range activating’ and ‘long-range inhibiting’ forces, and that within their contexts, both binary models seem to have the same epistemological limitations for their use of the minimum number of parameters needed to represent the dynamical nature of biological events ().
The other key categorization schema, also known as the 5-Elements/Agents or 5-Phases/Stages theory, is a non-linear syntactic (or structural) pattern recognition construct, which is known in Cognitive Science as an ‘associative multiparameter’ or a ‘Hopfield-style’ network. This type of congnitive organization consists of a set of essential parameters (variables/nodes) and a lattice of dependency-relationships (vector/arcs) that fully interlink them. This architecture seems to mirror the neural correlates of Hebbian learning. In the context of Chinese ancient metaphysics, the parameters are established by a ‘five-fold’ (Unschuld) discretization of all possibilities in every ontological domain into five mutually-exclusive variables, or paradigms, traditionally called Water, Wood, Fire, Earth and Metal (). Each parameter is attributed a finite number of potential states ranging from ‘deficiency’ to ‘excess,’ where events could be instantiated to. A system of cyclic vectors interlinks all variables by four series of ‘parental’ influences. There are two healthy series: the Generating (Sheng) and the Controlling (Ke) vectors, and two unhealthy ones, the Overacting (Cheng) and the Insulting (Wu) ones (). The pictogram of this semantic network amounts to a digraph (directed graph) with five nodes and ten vectors—a shape alike to a pentagram.
The cognitive value of this fully-connected semantic network resides in the fact that with sufficient training it can facilitate the quasi-mechanical structural recognition, classification and generalization of complex patterns based on the aggregation of variable-weights and relationship-strengths alone (,). Similar cognitive constructs are still widely used to design algorithms for machine learning and automated pattern recognition (). Due to its non-hierarchical structure, this model can be used as a global inference model in a variety of situations where detailed information about a phenomenon is incomplete, conflicting and vague, or where a phenomenon is culturally attributed to the occult and the supernatural. This conceptualization schema is also used in ‘holistic sciences,’ such as in Systems Biology and in Systemic Medicine for the overall representation of complex non-linear biological phenomena where the whole is more than the algebraic sum of the parts and where analytical approaches have failed to predict nature's complex, adaptive, and emergent behavior () ().
In the context of Chinese natural philosophy, this model is also paired with the ‘magick’ (exclusively meaning ‘paranormal’) belief in the existence of a second set of dependency-relationships between domains, where, by analogical reasoning, the manipulation of a variable in one domain is believed to cause changes in a parallel one () ().
Hence, by means of a multiparameter pattern recognition model, and with sufficient ‘supervised learning,’ a skilled folk healer can intuitively recognize a disease-pattern based on the instantiation of variable-weights (temperature, pulse and tongue conditions, etc.) and relationship-strengths (more pain than distention, less sweating than fever, etc.), and without any real knowledge of anatomy and physiology, or despite a lack of specific concepts and words for such a pattern (). Then, he or she can treat the disease with an empirically-established remedy with actions attributed to the magick correspondences between domains. As for the use of five parameters instead of another number, it might be due to numerology and geomancy, such as the practice of Feng Shui (). One might speculate that the notions of health and good fortune through sacred geometry and the divine or celestial harmony in terms of astrology, directions, location, the elements and proportions have determined its selection. The selection of five and the pentagram might also be related to the mathematical fact that the ratios of the lengths of the lines in a pentagram are all based on Phi (Φ), a ‘golden ratio’ close to 1.618, which also accounts for many ‘harmonious’ or ‘divine proportion’ found in nature () (,).
Overall, it seems reasonable to relate Chinese natural philosophy and medicine to what Isaac Bonewits has termed ‘Meso-paganism,’ meaning a group of spiritual or sacred beliefs and practices of natural or polytheistic beliefs that have been significantly influenced by monotheistic, dualistic or non-theistic worldviews. This group includes a vast array of beliefs and practices ranging from the Native American and Australian Aborigine spiritualities to Freemasonry, Rosicrucianism and Theosophy (,).
Paul Unschuld points out that the Yin-Yang model seems to be a purely Chinese construct; and that a ‘five-fold categorization of all phenomena’ seems to have been adopted at a later period (). Further, Maciocia relates that the dualistic model dates back to the Western Zhou Dynasty (c.1000–711 BC), while the first recorded reference to the pentic model dates only back to the Warring States Period (476–221 BC) (). It therefore seems reasonable to hypothesize that this late addition might have been the result of a cross-cultural exchange with the Mesopotamian or Hellenistic cultures. This suggestion relies on the fact that representations of a 5-pointed shape as a divine symbol date back to ancient Mesopotamia, where according to de Vogel it symbolized a ‘heavenly body’ (c.3500 BC). The same 5-pointed shape is found on Proto-Elamite tablets of the Susiana plain and the Iranian highlands east of the Tigris–Euphrates region where the pictogram signified the five directions: forward, backward, left, right and above, corresponding to the planets Jupiter, Mercury, Mars, Saturn, and Venus (3000–2500 BC) () (,). This semantic association is comparable to the Chinese correlation between directions and planets ().
The pictogram is also found in Greece among the Pythagoreans and in the context of the cult of Hygieia, (γιεια, in Latin) the Greek patron of well-being, sanitation and the prevention of disease. The words ‘hygiene,’ ‘salute’ and ‘salvation’ originate from her name. was the daughter of (’σκληπιóς), the God of medicine and healing, and the son of (’πóλλων or ’πέλλων), the God of life-giving Sun. Coincidently or not, the letters u-γ-ι-ει-α in her name also correspond to the initial letters of the five Pythagorean elements (): Hydor (δωρ, Water), (γαια, Earth), (ίδέα, Idea), (έιλή, Heat or Fire), and (άήρ, Air). The Pythagorean brotherhood which resembled a mystical circle for its ascetic living and its cultivation of health and ‘divine blessing,’ combined (Latin from the Greek άθαρσις, purification), numerology and geometric symbolism, and similarly, to the Chinese, believed in the existence of magick correlations between numbers, virtues, tastes, colors and sounds (,,).
According to de Vogel, the pentagram was equally used as a talisman in the pre-Christen Northern Europe in the context of health, good fortune and the prevention of disease. Its use may have been associated with the Druids, but de Vogel does not attribute it to the Pythagorean influence. It rather seems to have been used in the context of spell casting, ritual magick and folk healing as a protection against evil and malevolence, and was sometimes worn as an amulet for happy homecoming (). Today, a pentacle—a pentagram within a circle—remains a nostalgic symbol of Neo-pagan faiths such as the Reconstructionist and the Wiccan traditions. It is widely used in a number of rituals, and often worn as a symbol of recognition among the initiates of New Age witchcraft (). Even in urban shamanism, rave culture and other tribal/urban movements of the information age, the ‘primitives’ wear a pentacle talisman—possibly in defiance of Christianity—while they indulge in Chaos Magick and the sigil rituals of Pentagrammaton and Pentacle body-piercing (,).
Given this cluster of architectonic similarities, it is not inconceivable that pentic—or pentagramic—thinking in ancient Chinese natural philosophy and medicine is also historically related to pre-Christian Europe, the Pythagoreans and the pentagram. A cultural syncretism between the Greek culture and those of Central Asia started when Alexander the Great conquered Asia Minor and Central Asia in 334 BC. This cultural exchange gave rise to the Greco-Buddhist art and developed over a period of approximately 800 years in Central Asia between the 4th century BC and the 5th century AD. One might argue that this long period of intellectual exchange could also account for the gradual spread of a pentic star-shaped Mesopotamian or Greek archetype of health and wholeness in Central Asia and subsequently in China.
The adoption and the addition of this pentic archetype/prototype to a pre-existing binary model might also be due to an epistemological need, given that the Yin-Yang topological construct is a simple, hierarchical, tree-like and minimum model that only allows for one parental dependency-relationship per variable. It is therefore limited to the conceptualization of simple nonlinear structures and is ideal for binary taxonomies and ontologies (,). A non-hierarchical, multiparameter fully-connected semantic network, to the contrary, is an associative and ‘rhizomal’ model that allows for multiple dependency-relationships per variable, and is better suited to conceptualize dynamic systems where all components are inherently interlinked. In this model, discrete network-states can classify non-linear patterns based on the aggregation of parameter-weights and relationship-strengths alone (,). It would be interesting to see whether this epistemological shift from a purely tree-like model to a hybrid one, historically corresponds, or not, to a shift in perspectives from a purely ‘ontological’ view of disease as an external ancestral/demonological unifactorial phenomenon (‘Evil Qi’), to a view that also embraces a ‘functional-individualistic’ (Unschuld) outlook, and equally perceives disease as a multifactorial loss of an internal elemental/organic harmony (). If this were the case, such a paradigm drift would certainly require the adoption and the integration of an archetype/prototype of disease that would allow for multiple dependency-relationships per variable, and would also structurally reflect the link between health and wholeness and elemental/organic harmony. The selection of a pentagram would satisfy both requirements.
This cognitive and epistemological review points to remarkable infrastructural similarities between Chinese Medicine and certain esoteric beliefs of pre-Christian Europe. In both contexts, an associative five-parameter cognitive model paired with occultism, numerology, sacred geometry and magickal thinking underlies the belief system in health and the prevention of disease. This architectonic similarity is consistent with Paul Unschuld's view that the alleged dichotomy between Western and Chinese notions of disease is completely unfounded.
Nonetheless, Chinese Medicine is widely flourishing in the US and in Europe with the claim of being a unique, independent and comprehensive medical system based on alleged cognitive modalities ‘completely foreign to the West’. In the US alone, approximately 20,000 non-physician practitioners are licensed in over 40 states, and some deliver primary care, with insufficient training in modern health sciences, and with the reliance on ancient metaphysics, anachronous notions of health and disease, and a rudimentary and symptomological nosology ().
As for the prospect of an epistemological rupture with elementary cognitive models, and a paradigm shift that would lead to a firm biological basis for the testable remedies in such empirical medicines (,), one shall recall the words of Thomas Kuhn in Criticism and the Growth of Knowledge: |
The neurological complications of hyperammonemia in the central nervous system (CNS) are now receiving more attention. Ammonia is a neurotoxin that has been strongly implicated in the pathogenesis of hepatic encephalopathy (). Ammonia has also been a major pathogenetic factor associated with inborn errors of urea cycle, Reye's syndrome, organic acidurias and disorders of fatty acid oxidation (). Ammonia-induced neurotoxicity has been reported to include a dysfunction of multiple neurotransmitter system glutamate-mediated excitotoxicity, electrophysiological disturbances and defects in brain bioenergetics (,). In spite of extensive investigations, the precise mechanisms involved in ammonia neurotoxicity are not completely understood.
Oxidative stress is evolving concept in ammonia neurotoxicity. Its effect on the oxidative and nitrosative stress in the CNS has been recently reviewed (). Recent studies have reported an increased production of free radicals in cultured astrocytes after treatment with pathophysiological concentrations of ammonia (). A concurrent increase of superoxide production and a reduction in the activities of various antioxidant enzymes have been shown in animal models of acute ammonia toxicity (,). Oxidative stress-mediated lipid peroxidation was also shown as one of the characteristic features of hyperammonemia (,).
(Linn) (HS) (family Malvaceae), is an annual dicotyledonous herbaceous shrub popularly known as ‘Gongura’ in Hindi or ‘Pulicha keerai’ in Tamil. This plant is well known in Asia and Africa and is commonly used to make jellies, jams and beverages. In the Ayurvedic literature of India, different parts of this plant have been recommended as a remedy for various ailments such as hypertension, pyrexia, liver disorders and antidotes to poisoning chemicals (acids, alkali and pesticides) and venomous mushrooms (). Anthocyanins, flavonols, protocatechuic acid (PCA), along with others, have been identified as contributors to the observed medicinal effect of HS (). Anthocyanin and PCA have been shown to have antioxidant activity and to offer protection against atherosclerosis and cancer (). Compared to common antioxidants such as ascorbate, anthocyanins were found to be much more potent antioxidants (). It is well documented that most medicinal plants are enriched with phenolic compounds and bioflavonoids that represent potent antioxidants (). There is currently a growing body of evidence that supplementing the human diet with antioxidants is of major benefit for human health and well-being.
Nowadays, the use of complementary/alternative medicine and especially the consumption of botanicals have been increasing rapidly worldwide, mostly because of the supposedly less frequent side effects when compared with modern Western medicine. Both in conventional and traditional medicines, plants continue to provide valuable therapeutic agents (). Doubts about the efficacy and safety of currently available anti-hyperammonemic agents have prompted the search for safer and more effective alternatives (). To our knowledge, this report is the first study to investigate the effect of alcoholic extract of HS leaves (HSEt) on brain lipid peroxidation and antioxidant status in ammonium chloride-induced hyperammonemic rats. Therefore, we made an attempt to bridge the information gap. The present study was undertaken to investigate the effect of HSEt on brain lipid peroxidation and antioxidant status in ammonium chloride-induced hyperammonemic rats.
The mature green leaves of were collected from Chidambaram, Cuddalore District, Tamil Nadu, India. The plant was identified and authenticated at the Herbarium of Botany Directorate in Annamalai University. A voucher specimen (No. 3648) was deposited in the Botany Department of Annamalai University. The shade-dried and powdered leaves of were subjected to extraction with 70% ethanol under reflux for 8 h and concentrated to a semi-solid mass under reduced pressure (Rotavapor apparatus, Buchi Labortechnik AG, Switzerland). The yield was about 24% (w/w) of the starting crude material. In the preliminary phytochemical screening, the ethanolic extract of HSEt gave positive tests for glycosides, anthocyanins, polyphenols and flavones (). The residual extract was dissolved in sterile water and used in the investigation.
Male albino Wistar rats weighing 180–200 were used for the study. They were housed in polycarbonate cages under standard conditions (22 ± 2°C, humidity of 45–64%, 12 h light/dark cycles). They were given standard pellet diet (Hindustan Lever Ltd, Mumbai, India) and water All animal experiments were approved by the ethical committee (Vide. No. 273/2004), Annamalai University, India, and were in accordance with the guidelines of the National Institute of Nutrition (NIN), Indian Council of Medical Research (ICMR), Hyderabad, India. Ammonium chloride was purchased from Sisco Research Laboratories, Mumbai, India. All other chemicals used in the study were of analytical grade.
Hyperammonemia was induced in Wistar rats by daily intraperitoneal injections of ammonium chloride at a dose of 100 mg kg body weight for eight consecutive weeks (). Rats were divided into four groups, eight animals each. Group 1: control rats. Group 2: rats orally administered with HSEt (250 mg kg body weight) (). Group 3: rats intraperitoneally treated with ammonium chloride (100 mg kg body weight) (). Group 4: rats treated with ammonium chloride (100 mg kg) + HSEt (250 mg kg). At the end of 8 weeks, the animals were killed by decapitation and the blood samples were used for the estimation of circulatory ammonia, urea, uric acid, creatinine and non-protein nitrogen. Whole brain was dissected out and washed in ice-cold phosphate buffered saline. The brains were weighed and 10% tissue homogenates were prepared in 0.025 M Tris–HCl buffer, pH 7.5, and used to measure the activities of thiobarbituric acid reactive substances (TBARS), hydroperoxides (HP) and glutathione peroxidase (GPx). Enzyme activity was assayed in 10% brain homogenates prepared in 0.2 M phosphate buffer, pH 8.0.
Ammonia levels were estimated in the blood where blood, triethanolamine and NADPH/GLDH/buffered substrates were well-mixed and the absorbance was read at 470 nm using UV spectrophotometer-Hitachi 912. This procedure is detailed elsewhere (). Levels of urea, uric acid and non-protein nitrogen were measured in the plasma according to standard methods stated elsewhere (). Creatinine levels were estimated in serum following an alkaline picrate method [detailed in ref. ()]. The absorbance of each of the previous parameters was read at 480 nm using UV spectrophotometer-Hitachi 912. Blank and a series of standards were processed similarly.
Lipid peroxidation was estimated colorimetrically in brain by assessing TBARS and HP according to standard methods (,). In brief, for the estimation of TBARS the supernatant of tissue homogenate was treated with TBA–TCA–HCl reagent and mixed thoroughly. The mixture was kept in boiling water bath for 15 min. After cooling, the tubes were centrifuged for 10 min and the supernatant was taken for measurement. The developed color was read at 535 nm using UV spectrophotometer-Hitachi 912 against the reagent blank. For the estimation of HP, the supernatant of tissue homogenate was treated with Fox reagent () and incubated at 37°C for 30 min. The developed color was read at 560 nm using UV spectrophotometer-Hitachi 912 against reagent blank. TBARS and HP were expressed as mM per 100 g tissue.
Catalase (CAT) was assayed colorimetrically at 620 nm and was expressed as μmoles of HO consumed per min per mg protein as described by Sinha (). Superoxide dismutase (SOD) was assayed utilizing the technique of Kakkar . (). A single unit of enzyme was expressed as 50% inhibition of NBT (nitroblue tetrazolium) reduction per min per mg protein. GPx activity was measured according to the method described by Rotruck . (). Briefly, reaction mixture contained 0.2 ml of 0.4 M Tris–HCl buffer, pH 7.0, 0.1 ml of 10 mM sodium azide, 0.2 ml of tissue homogenate (homogenized in 0.4 M Tris–HCl buffer, pH 7.0), 0.2 ml glutathione and 0.1 ml of 0.2 mM hydrogen peroxide. The contents were incubated at 37°C for 10 min. The reaction was stopped by 0.4 ml of 10% TCA, and centrifuged. Supernatant was assayed for glutathione content by using Ellmans reagent (19.8 mg of 5,5'-dithiobisnitro benzoic acid in 100 ml of 0.1% sodium nitrate). Reduced glutathione (GSH) was determined by the method of Ellman (). One milliliter of supernatant was treated with Ellman's reagent and phosphate buffer (0.2 M, pH 8.0). The absorbance was read at 412 nm. The activities of GPx and GSH were expressed as μg of GSH consumed per min per mg protein and as mg per 100 g of tissue, respectively.
Data analysis was carried out by analysis of variance (ANOVA) and the groups were compared using Duncan's Multiple Range Test (DMRT).
shows the levels of blood ammonia, urea, uric acid, non-protein nitrogen and creatinine of control and experimental groups. Levels of blood ammonia, urea, uric acid, non-protein nitrogen and creatinine increased significantly in ammonium chloride-treated rats. These levels were significantly restored to near normal upon the administration of HSEt (). There is no significant change in bodyweight of animals in the experimental groups when compared with controls ().
The levels of lipid peroxidation products and antioxidants in brain of control and experimental groups are shown in . The levels of TBARS and HP were significantly higher and the levels of SOD, CAT, GSH and GPx were significantly lower in the brain of ammonium chloride-treated rats. These levels were significantly restored to near normal upon the administration of HSEt ().
This work is one of the series of studies showing that chronic hyperammonemia causes an imbalance in the oxidative status of nervous tissue and that the resulting free radicals damage the brain through a peroxidative mechanism. The elevated levels of ammonia and urea might vindicate hyperammonemic condition in rats treated with ammonium chloride (,,). The reduction in levels of ammonia, urea, uric acid, creatinine and non-protein nitrogen during HSEt treatment shows significant anti-hyperammonemic activity of this plant. This is probably indicative of antioxidant efficacy of this plant (). Phenolic compounds and flavonoids have the ability to remove excess ammonia, urea, uric acid and creatinine during hyperammonemic and nephrotoxic conditions and offer protection against hyperammonemia (,). Our results corroborate these previous findings.
Many studies have shown that oxidative stress and free radical production-mediated lipid peroxidation could be involved in the mechanism of ammonia toxicity (–,,). Elevated levels of ammonia in blood and brain result in derangement of cerebral function (–). A marked elevation in the concentration of TBARS and HP are observed in the brain of hyperammonemic rats. Excess ammonia induces nitric oxide synthase, which leads to enhanced production of nitric oxide and other toxic free radicals as well as thiobarbituric acid-positive compounds in brain and leads to oxidative stress and tissue damage (,,,). Administration of HSEt significantly decreased brain levels of TBARS and HP in Group 4 rats. HSEt could reduce levels of circulatory lipid peroxidation products during hyperammonemia (), which may corroborate our present findings. This may be due to the free radical scavenging property of HSEt and has been previously reported (,). In addition, a marked nitric oxide scavenging activity was observed for the alcoholic extract of HS flowers supporting the plant's potent antioxidant property (,–).
The level of lipid peroxidation in cells is controlled by various cellular defense mechanisms consisting of enzymatic and non-enzymatic scavenger systems (), the levels of which are altered in hyperammonemia (,). This might have decreased levels of antioxidants in brain such as SOD, CAT, GPx and GSH in ammonium chloride-treated group rats. Increased superoxide production and reduced activities of antioxidant enzymes have been reported in brains of rats subjected to acute ammonia toxicity (). In our investigations, the levels of both enzymatic and non-enzymatic antioxidants, which declined in the brain of hyperammonemic animals, were significantly restored to near normal after treatment with the HSEt extract. This restoring potential of HSEt might be due to active principles of HS. Preliminary phytochemical screening of HS showed the presence of flavanoids and phenolic compounds (like anthocyanins, glycosides, PCA and hydroxycitric acid) (), which have been reported to have antioxidant activity. It is believed that phenolic antioxidants can scavenge harmful free radicals and thus inhibit their oxidative reactions with vital biological molecules () and prevent development of many pathophysiological conditions, which can manifest into disease. Previous reports have shown that the calyces extract of HS decreased lipid peroxidation and cell damage (–). Previous studies show that the plant extracts and their active constituents have the ability to improve the antioxidant status and reduce the levels of urea, creatinine during various disease conditions (,). Therefore, it is possible that the mechanism by which the HSEt modulates brain lipid peroxidation and antioxidant status during hyperammonemic condition could be attributed to the presence of natural antioxidants, ammonia lowering effect and its free radical scavenging properties (). But the exact mechanism is still under investigation and isolation of contributing active constituents is required. |
Lipids containing polyunsaturated fatty acids are readily oxidized by molecular oxygen, and such oxidation proceeds by a free radical chain mechanism (). Lipid peroxidation can lead to aging, coronary heart disease, stroke, diabetes mellitus, rheumatic disease, liver disorders, multiple sclerosis, Parkinson's disease, autoimmune disease, Alzheimer's and carcinogenesis (,). An increasing number of investigations have been carried out to find antioxidative drugs, which not only prolong the shelf life of food products but also participate as radical scavengers in living organisms.
As with other synthetic food additives, commercial antioxidants have been criticized, mainly due to possible toxic effects. Therefore, there is an increasing interest in the antioxidative activity of natural compounds (,). They can be an alternative to the use of synthetic compounds in food and pharmaceutical technology or serve as lead compounds for the development of new drugs with the prospect of improving the treatment of various disorders.
Iranian conifers consist of two families: Cupressaceae and Taxaceae. Cupressaceae consist of one species of [ L. with two varieties: L. var. (Mill.) Aiton and L. var. with a cultivar namely L. cv. Cereifeormis], one species of ( Franco) and five species of [ L. subsp. (Presl) Nyman, M. B., L., Willd. and M. B. with two subspecies namely M. B. subsp. and M. B. subsp. (K. Koch) Takhtajan]. The Taxaceae consist of only one species of ( L.). and M. B. subsp. are monoecious and others are diecious ().
Iranian conifers are evergreen and aromatic plants are widely spread and grow in different parts of many countries including Iran. Each of them has its own Persian name (,). Most of these trees are medicinal plants and seeds, dried leaves and fruits are used to treat various diseases like bronchitis, common cold, nose bleeds, hypertension, inflammation and gout, and used as expectorant, contraceptive, diuretics, for rheumatic symptoms, to regulate menstruation and to relieve menstrual pain (). There are some reports on phytochemical and biological studies of some of these taxons as well as other related species ().
There are a few reports about the antioxidant activity of conifers. The antioxidant activity of methanol extract of heartwood was determined by DPPH method (). This plant revealed strong antioxidant activity. In another study, the antioxidant activity of fruit extracts were evaluated using different antioxidant assay. The results revealed that both the water and ethanol extracts exhibited strong total antioxidant activity (). The antioxidant and radical scavenging properties of (Cupressaceae) and essential oils were also tested by different methods (,).
There are no previous reports concerning antioxidant properties of the extracts of Iranian conifers. Therefore, this study evaluated the antioxidant properties of the methanol extracts of fruits and leaves of both male and female Iranian conifers.
Plant specimens were collected from different parts of the country () as follows:
Dr M. Assadi, Research Institute of Forest and Rangelands, Ministry of Jahad Keshavarzi, Iran, was identified these plants. Voucher specimens of the taxons have been deposited in the Herbarium of National Botanical Garden of Iran. The collected materials were stored at −20°C in order to avoid unfavorable changes in the chemical components ().
Individual fresh leaves of male and female of each of the dioecious plants and fresh leaves of monoecious of each taxon (100 g fresh wt.) as well as their fruits (100 g fresh wt.) were cut to small pieces and then ground by a blinder. Each sample was macerated in pure methanol for 24 h. The samples were then extracted using a percolator. The extracted solutions (27 samples) were concentrated at 50°C to dryness under reduced pressure. The methanol extracts of leaves and fruits of each taxon were evaluated for their antioxidant activity.
The volatile oils of fresh leaves and fruits of male and female of each taxon (200 g fresh wt.) were isolated by wet steam distillation for 4 h (). The oil samples were dried over anhydrous sodium sulfate. The yield percentages of the essential oils were expressed in ml/100 g of fresh plant materials.
The fruits and leaves of each plant (500 g) were dried at 50°C and then powdered separately. Each powder was defatted with petroleum ether (bp 40–60°C) using Soxhlet apparatus (6 h). The chemical components of defatted powders were extracted by maceration with methanol (four times). The methanol extracts were concentrated at reduced pressure and the presence of alkaloids (), flavonoids (), saponins () and tannins () were determined.
Several reports have evaluated the antioxidant activity of various essential oils and extracts of different plants (,,). In this study, the methanol extracts of leaves, of male and female, and fruits of eleven different taxons of Iranian conifers (27 samples) were evaluated for their antioxidant activity (final concentration 0.02% w/v). Ferric thiocyanate (FTC) and thiobarbituric acid (TBA) methods were used to evaluate antioxidant activity (). Vitamin E and butylated hydroxytoluene (BHT) (0.02%) were used as standards in both methods. One sample without antioxidant activity was also used as control.
In these experiments, for inhibition of linoleic acid peroxidation, the reaction mixture was composed of linoleic acid in ethanol, the sample solution, 0.05 mol/l phosphate buffer (pH 7.0) and water. Following incubation, the degree of oxidation was measured according to FTC and TBA methods. To calculate the percentage of antioxidant activity, after reading the absorbance of samples at 500 nm for FTC method and at 532 nm for TBA method, the percentage of activity was calculated according to the following equation:
The values obtained for the control samples were taken for 100% lipid peroxidation.
All experiments were repeated three times and an average used for calculating the antioxidants activity of each sample. The antioxidant activity of the extracts and positive controls were compared by ANOVA one-way test, ( ≤ 0.05), using SPSS program.
The leaves, of male and female, and fruits of all 11 different taxons of Iranian conifers (27 samples) evaluated for their antioxidant activity showed strong antioxidant activity by both FTC and TBA methods ( and ). The activity of BHT was ∼100% (the average absorbance of the BHT in both FTC and TBA methods was 0.002 in comparison with the control sample absorbances which were 0.868 and 0.635).
The amount of non-volatile components (from defatted methanol extracts) of the fruits and leaves as well as the yield percentages of the essential oils are shown in .
There is a strong need for effective antioxidants from natural sources as alternatives to synthetic food additives in order to prevent deterioration of foods, drugs and cosmetics. The extracts and essential oils of many plants have been investigated for their antioxidant activity (,,). In this study, the antioxidant properties of the methanol extracts of leaves, of male and female, and fruits of 11 different taxons of Iranian conifers; var. var. cv. Cereifeormis, subsp. subsp. subsp. and , were examined.
The inhibitions of activities against lipid peroxidation in linoleic acid were evaluated by measuring the concentration of the TBA-reactive substances and FTC.
The FTC method measures the amount of peroxide produced during the initial stage of lipid oxidation. Subsequently, at a later stage of lipid oxidation, peroxide decomposes to form carbonyl compounds that are measured by using the TBA method. The entire methanol extracts possessed strong antioxidant activity (low absorbance values) by both the FTC and TBA methods. The antioxidant activity was then compared with those of α-tocopherol (a natural antioxidant) and BHT (a synthetic antioxidant). Different extracts obtained from different parts of the plants exhibited strong antioxidant activity within the range of 78–99% by the FTC method (except for leaves of var. which was 55%) (). Using the TBA method, all extracts also showed strong antioxidant activity within the range of 60–99% (except for leaves of var. and male leaves of which were 44.5 and 51.5%, respectively), (). The pattern of activity was very similar for both methods. cv. Cereifeormis fruits’ methanol extract exhibited the highest antioxidant activity (quite higher than α-tocopherol) in both methods. Among the extracts examined, the leaves of var. methanol extract possessed the lowest antioxidant activity.
This may be partially due to the low amounts of flavonoids and tannins in the leaves of var.
A few studies on antioxidant activity of these plants’ extracts also showed the same results. The methanol extracts of heartwood revealed the strong antioxidant activity (). Results of a study of antioxidant activity of fruit extracts in support our findings ().
Antioxidant activity and its strength in each plant depends on the existence of various compounds in that plant. Antioxidative and radical scavenging activities of flavonoids are well studied (). Some of phenolic compounds (anthocyanidin, catechines, flavones, flavonols and isoflavones), tannins (ellagic acid, gallic acid), phenyl isopropanoids (caffeic acid, coumaric acids, ferulic acid), lignans, catchol and many others are antioxidants (). Several different essential oils obtained from various plants and their components have also been studied for their antioxidant activities (). and indicate different extracts obtained from the leaves, of male and female, and fruits of all 11 different taxons showed different strength of antioxidant activity by both FTC and TBA methods. The methanol extract of the tested plants contains various non-volatile and volatile compounds (). As it can be seen from , the amounts of non-volatile compounds in leaves and fruits of them vary significantly. In all them (except for ), the amount of alkaloids was not detectable while the amounts of flavonoids, saponins and tannins were very different. Variation in the amounts of various non-volatile and volatile compounds in different tested plant can be one of the reasons causing differences in antioxidant activity of the extracts obtained from different plants as well as from different parts of each plant.
Finally, while further investigation is necessary to separate the component of each extracted sample and then evaluate the antioxidants activity of each component using several different methods, at this stage, methanol extracts of these plants can be considered as a strong antioxidant agent. |
The 22nd International Symposium on Acupuncture and Electro-Therapeutics is the annual meeting and the main scientific event of research and teaching activity of the International College of Acupuncture & Electro-Therapeutics Research. The meeting is organized by Professor Yoshiaki Omura, Founder and Director of the College, Director of Medical Research in the Heart Disease Research Foundation. The Symposium is supported by the International College of Acupuncture & Electro-Therapeutics Research, by its official journal published by Cognizant Communication Corporation and by the Heart Disease Research Foundation.
The International College of Acupuncture & Electro-Therapeutics Research is a non-profit international medical society dedicated to scientific investigation and to education of physicians and other health care professionals on the latest research findings and practical methods in acupuncture and related techniques.
Scientific presentations at 22nd International Symposium on Acupuncture & Electro-Therapeutics addressed a multitude of topics relevant to physicians and scientists interested in acupuncture and related techniques. Presentations on advances in acupuncture research using the methods of basic science included the classical pharmacological experimental approach, structural and ultra-structural histological methods followed by clinical reports on the evaluation and treatment of a wide variety of diseases using acupuncture.
The Symposium was focused primarily on body acupuncture according to the traditional Chinese meridian system, ear acupuncture and diagnostic techniques combining Traditional Chinese Medicine (TCM) concepts with the modern scientific approach.
The program consisted of various types of presentations: lectures, workshops, live- and video-presentations of patients, and live demonstrations of diagnostic and treatment techniques.
The 4-day Symposium attracted researchers from Australia, Brazil, Canada, China, Germany, Japan, Korea, Russia, Serbia, Tibet, Turkey and USA. The intensive program was scheduled from 09:00 a.m. until 10:00 p.m. every day.
p
Dr Omura reported on the role of asbestos as a cause of chronic debilitating diseases, various malignancies, cardiovascular diseases, Alzheimer's disease and chronic intractable pain syndromes. Using extensive observational data, Dr Omura suggested that asbestos might have a much stronger pathological potential in the development of such diseases as malignant lung tumors (including small cell carcinoma, adenocarcinoma and mesothelioma), brain tumors (i.e. astrocytoma and glioblastoma multiforme), chronic intractable pain syndromes including fibromyalgia and some cardio-vascular pathology (). He compared the conventional method of asbestos detection using the transmission type electron microscope with the Bi-Digital O-Ring Test (BDORT) examination, which is inexpensive and easily performed in clinical routine. Dr Omura proposed prevention measures and ways to eliminate asbestos from the human organism.
Yasuhiro Shimotsuura, Shimotsuura Clinic, Kurume City, Fukuoka, Japan, presented exciting results of an observational case series study on the early prediction of cancer cases using the BDORT. The authors carefully described five cases where the early cancer diagnosis using BDORT predicted the localization of cancer 5–10 years before the cancer was detected using the common mainstream medicine diagnostic procedures.
Momir Dunjic, Department of Gynecological Endocrinology, School of Medicine Pristina, University of Belgrade, Yugoslavia, reported the intriguing clinical data where BDORT was applied for prenatal fetal gender determination and for diagnostics of pregnancy-related disorders of glucose metabolism. The researchers found that the BDORT had 95% accuracy of gender prediction in 149 cases and revealed all patients (8 out of 128 pregnant women) with glucose intolerance, later confirmed by standard oral glucose tolerance test. The same research group reported on the new approach in diagnostics and treatment of autoimmune Hashimoto thyroiditis (HT) in 292 patients using the BDORT and selective drug uptake enhancement method (SDUEM). The accuracy of HT diagnostic using the non-invasive BDORT ranged 90–95% referring to standard immunological diagnostic techniques. The authors noted that BDORT helped to reveal the possible etiology of HT, which improved the treatment effectiveness. The application of SDUEM led to the normalization of thyroid gland antibody levels in almost 90% of patients.
Muneyoshi Oka, Molecular Resonance Research Laboratory, Oita Oka Hospital, Japan, presented data on the coincidence of health hazards and an increased rate of endogenous depression. The main idea of his presentation is that the environmental pollutants (e.g. formaldehyde and dioxin) in rainwater markedly decrease cerebral neurotransmitters (e.g. dopamine, serotonin and GABA), which may contribute to the pathogenesis of depression.
Kemal Nuri Ozerkan, School of Physical Education and Sports, Istanbul University, Turkey, reported on new approaches to acupuncture diagnostics and treatment using the BDORT, applied to the so-called ‘jingwell’ acupuncture points on the tips of fingers and toes. He stressed that using the BDORT as a diagnostic tool helps precisely identify the distortion in the system of acupuncture meridians, to find more efficient acupuncture points for treatment of energy distortion in meridian system and to find the optimal time for this therapy.
Dr Usichenko presented a systematic review on pain relief when low-intensity electromagnetic millimeter waves (MW) are applied to acupuncture points. The evidence from 12 clinical trials (eight of them randomized) strongly suggests that the application of electromagnetic millimeter waves in the frequency range 30–70 GHz to acupuncture points might have alalgesic effects. The model of an ‘electromagnetic frame’ of the human body based on the principles of quantum physics and data from embryological, physiological and clinical research allows us to interpret the nature of acupuncture meridians and to explain the mechanism of MW applied to acupuncture points. Rigorous large-scale randomized controlled trials on the effectiveness of this non-invasive therapeutic technique are necessary (). The most intriguing question is to evaluate the role of the exposure site in relation to the topography of the acupuncture meridian system.
Narayan G. Patel, Ayurveda Center in Wilmington, Delaware, USA, reported on the implementation of energetic theories and techniques of ancient Tibetan medicine into modern health care. His lecture was followed by a demonstration of non-invasive diagnostics using the method and equipment developed by Reinhard Voll in the early 1950s in Germany (). Using volunteers from the audience, Dr Patel vividly demonstrated ways to combine the ancient theories and experience with modern methods of CAM applied to modern health care.
Avracham Henoch from Department of Family Medicine, New York Presbyterian Hospital, reviewed the historical development and clinical application of the Bi-Digital O-Ring Test and presented an updated literature review about the role of infection in the development of cancer.
The panel discussion on acupuncture training and licensing in various countries took place at the end of Symposium. Moderated by Dr Omura, this exchange of information among the representatives from different countries of the world contributed to a better understanding of the legislative situation of acupuncture in Brazil, Canada, Germany, Great Britain, Japan, Korea, Russia, Serbia, Turkey, Ukraine and some other counties. The Symposium ended with a planning meeting of the Council of the International College of Acupuncture & Electro-Therapeutics.
e
a
u
d
i
e
n
c
e
o
f
t
h
e
2
2
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
s
i
u
m
o
n
A
c
u
p
u
n
c
t
u
r
e
&
E
l
e
c
t
r
o
-
T
h
e
r
a
p
e
u
t
i
c
s
c
o
n
s
i
s
t
e
d
p
r
i
m
a
r
i
l
y
o
f
e
x
p
e
r
t
s
i
n
t
h
e
f
i
e
l
d
(
r
e
s
e
a
r
c
h
e
r
s
a
n
d
a
d
v
a
n
c
e
d
c
l
i
n
i
c
i
a
n
s
)
.
I
n
c
o
n
t
e
x
t
o
f
t
h
e
v
a
r
i
o
u
s
p
r
e
s
e
n
t
a
t
i
o
n
s
a
n
d
d
u
r
i
n
g
b
r
e
a
k
o
u
t
s
e
s
s
i
o
n
s
,
p
a
r
t
i
c
i
p
a
n
t
s
v
e
r
y
m
u
c
h
a
p
p
r
e
c
i
a
t
e
d
o
p
p
o
r
t
u
n
i
t
i
e
s
t
o
e
x
c
h
a
n
g
e
i
d
e
a
s
a
b
o
u
t
t
h
e
i
r
w
o
r
k
a
n
d
r
e
c
e
n
t
c
l
i
n
i
c
a
l
a
n
d
s
c
i
e
n
t
i
f
i
c
a
d
v
a
n
c
e
s
i
n
a
c
u
p
u
n
c
t
u
r
e
d
i
a
g
n
o
s
t
i
c
a
n
d
t
h
e
r
a
p
e
u
t
i
c
t
e
c
h
n
i
q
u
e
s
,
w
i
t
h
p
a
r
t
i
c
u
l
a
r
e
m
p
h
a
s
i
s
o
n
B
D
O
R
T
.
T
h
e
2
2
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
S
y
m
p
o
s
i
u
m
o
n
A
c
u
p
u
n
c
t
u
r
e
&
E
l
e
c
t
r
o
-
T
h
e
r
a
p
e
u
t
i
c
s
w
a
s
a
s
u
c
c
e
s
s
f
u
l
s
t
e
p
o
n
t
h
e
p
a
t
h
t
o
i
n
t
e
g
r
a
t
i
n
g
s
c
i
e
n
c
e
i
n
t
o
t
h
e
f
i
e
l
d
s
o
f
a
c
u
p
u
n
c
t
u
r
e
a
n
d
e
n
e
r
g
y
t
h
e
r
a
p
i
e
s
.
W
e
b
e
l
i
e
v
e
t
h
a
t
t
h
e
a
p
p
l
i
c
a
t
i
o
n
o
f
s
c
i
e
n
t
i
f
i
c
a
l
l
y
v
a
l
i
d
a
t
e
d
e
n
e
r
g
e
t
i
c
C
A
M
m
e
t
h
o
d
s
w
i
l
l
e
n
h
a
n
c
e
t
h
e
d
i
a
g
n
o
s
t
i
c
a
n
d
t
h
e
r
a
p
e
u
t
i
c
p
o
t
e
n
t
i
a
l
o
f
a
c
u
p
u
n
c
t
u
r
e
a
n
d
s
u
b
s
t
a
n
t
i
a
l
l
y
i
m
p
r
o
v
e
p
a
t
i
e
n
t
c
a
r
e
(
6
)
. |
Boiss (Lamiaceae) is a valuable medicinal plant grown extensively in Iran, Pakistan and Afghanistan (). The chemical compositions of extracts have been extensively characterized in Iran (–) and Pakistan (). The extract contains thymol, carvacrol (,), zatrinal, oleanolic acid, betulic acid, rosmarinic acid () and monoterpenoids, sesquiterpenoids, p-cymene, y-terpinene (,).
Aqueous and alcoholic extracts of have been therapeutically used for relieving nociceptive pain (,), recurrent aphthous stomatitis (RAS) (), and prevent growth of oral streptococci (), () and () as well as used as an insect repellent (). Fataneh () has investigated anti-fungal properties of extract o. In view of its potent antibacterial and anti-fungal activities, we hypothesized that extracts may possess anti- effects.
The plant was collected from Shiraz, Iran, and identified by Agricultural Research Centre, Ahwaz. The plant was identified in the Systematic Laboratory, Agricultural Sciences Centre, Ahwaz, Iran, where voucher specimens were deposited (ZM 1). Aliquots of 100 g of the dried powder of the plant were soaked in ethanol (1400 ml), methanol (1400 ml) and distilled water (2150 ml) for 24 h and then filtered with cloths. The extracts were concentrated to dryness in a at 53–55°C and yielded 11.39 g (aqueous extract), 15 g (ethanolic extract) and 13.3 g (methanolic extract).
Fourteen isolates of were studied including ( = 7), ( = 3), ( = 2) and ( = 2). All species were isolated from infected patients in the department of medical mycoparasitology, Jundishapour University of medical sciences, Ahwaz, Iran. All isolates were identified by CHROMagar Candida (CHROMagar Candida Company, Paris, France), germ-tube test, production of chlamydoconidia on Corn meal agar and growth at 45°C. Isolates were maintained on Sabouraud's dextrose agar (SDA) at 4°C. Organisms were subcultured on SDA and incubated at 37°C for 24 h. Several colonies of each species were collected in 2 ml of sterile PBS to prepare a suspension. The suspension was adjusted to 70% transmittance (T) by a spectrophotometer at 530 nm. This should result in a suspension containing about 1 × 10 cfu per ml.
A serial dilution of each extract was prepared in SDA plates. Aqueous, ethanolic and methanolic extracts were diluted by the same solvent. The same solvent, at an appropriate concentration was also used as a negative control. A plate was considered as positive control without extracts and solvents. Aliquots of 20 μl of standardised suspension of different species of were inoculated in to each plate. The plates were incubated at 30°C for 24–48 h. The lowest extract concentration where there was no visible growth was the minimal inhibitory concentration (MIC) when compared to control. All experiments were repeated three times and mean calculated.
sup
#text
Herbal and alternative medicines are popular in the general population worldwide. A great number of modern drugs are still derived from herbs (). Iranian scientist, Avicenna (980–1037) and Razi (846–930) published several books on herbal medicine a few centuries ago and are still in use in different libraries in Europe (). grows wild in central and southern Iran. is used in traditional herbal medicine for antiseptic, analgesic, and carminative properties (,,). was also used for treatment of ‘Women disease’ in Iranian folklore (). The leaf powder of is used as nutritional flavoring in Iran. It is important to investigate scientifically those plants, which have been used in traditional medicines as potential sources of novel antimicrobial compounds.
The presence of thymol, rosmarinic acid, and carvacrol in the different parts of the plant was observed (). The present results indicate that methanolic extracts of the aerial part of have marked activity against isolates of . Probably, the anti- activity of methanolic extract of is due to both rosmarinic acid and thymol that extracted only into methanol (). Probably, the anti- activity of methanolic extract of is due to above compounds. is used in traditional herbal medicine for women disease (). Fataneh () have shown that has anti-fungal activity. They tested several isolates of dermatophytes and saprophytic fungi against extract. Amanlou . () have shown that has antierythema in denture stomatitis compared to miconazole gel, however, gel did not reduce the colony count of the denture surface as efficiently as miconazole gel. The ethanolic extracts of aerial parts of showed antinociceptive activity (,). Phytochemical screening supported the presence of flavonoids in . Some flavonoids exert antinociceptive activity in mice (). Ramazani . () reported six fractions of the extracts of aerial parts of that have antinociceptive activity.
We conclude that represents an untapped source of potentially useful anti- and is worthy for future clinical study. In addition, measures must to be undertaken to preserve the traditional knowledge about medicinal plants. |
Moutan Cortex (MCE), the root cortex of Andrews, is a traditional Korean herb with various biological activities. It has been used in oriental medicine to remove heat from the blood, promote blood circulation and remove blood stasis (). Moutan Cortex has been shown to have antimicrobial actions against , typhoid, paratyphoid bacilli and , and antihypertensive efficacy including analgesic, sedative and anticonvulsant effects. Moutan Cortex exerts a protective effect against acetaminophen-induced hepatotoxicity by attenuating GSH depletion, cytochrome P450 2E1 activity and hepatic DNA damage (,), and scavenging effects on the DPPH (1,1-diphenyl-2-picrylhydrazyl) radical, superoxide anion radical (), hydroxy radicals () and intracellular ROS in oxidative condition ().
Lipopolyssacharide (LPS)-activated macrophages have usually been used for evaluating the anti-inflammatory effects of various materials. LPS is a principle component of the outer membrane of Gram-negative bacteria, is an endotoxin that induces septic shock syndrome and stimulates the production of inflammatory mediators such as nitric oxide (NO), tumor necrosis factor-α (TNF-α), interleukins, prostanoids and leukotrienes (–). Therefore, LPS plays a key role in not only eliciting an inflammatory response but also causing septic shock during a Gram-negative bacterial infection. Inflammatory responses are advantageous for eradicating bacteria, as long as they are under control. However, when out of control, deregulated inflammation leads to the massive production of proinflammatory cytokines such as TNF-α, interleukin-1 (IL-1) and interleukin-6 (IL-6) by macrophages (,), which can cause tissue injury and multiple organ failure ().
Recently, there have been many studies concerning natural products with anti-inflammatory activity, for example, (), (), (,), (), Ginsenoside Rg3 (), sauchinone (,), rhizomes () and Farfarae Flos (). Although Moutan Cortex has been used as an anti-inflammatory agent in oriental medicine, the mechanisms of anti-iflammatory activity of Moutan Cortex have not been studied scientifically. Paeonol is a major component of Moutan Cortex that may be largely responsible for the therapeutic effect of Moutan Cortex. Paeonol regulated histamine, IgE and TNF-α (). However, it is not enough to explain the anti-inflammatory activity of paeonol and Moutan Cortex. There is no evidence to explain the mechanism on the inflammatory cytokine regulation of Moutan Cortex. Including paeonol, there are many known chemical components of Moutan Cortex such as paenoside, paeonolide, paeoniflorin, oxypaeonolniflorin, benzoylpaeoniflorin, benzoyl-oxy-paeoniflorin and apiopaeonoside. Therefore, this study evaluated the effect of methanol extract of Moutan Cortex (MCE) on the regulatory mechanism of NO, cytokines and prostaglandin E (PGE) in the LPS-activated Raw264.7 cells.
Moutan Cortex (300 g, Wolsung, Daegu, Korea) was extracted with 1000 ml of methanol at room temperature for 24 h. The extract was filtered through a 0.2 μm filter (Nalgene, New York, NY, USA), lyophilized and stored at −20°C until needed. The amount of the extract was estimated by dividing the dried weight of Moutan Cortex into the lyophilized MCE. The yield of MCE from Moutan Cortex was 13.2%.
Raw264.7 cell, which is a murine macrophage cell line (KCLRF, Korean Cell Line Research Foundation, Seoul, Korea), was cultured in Dulbecco's modified Eagle's medium (DMEM, Cambrex Bio Science, MD, USA) containing 10% fetal bovine serum (FBS), 100 U ml penicillin and 100 μg ml streptomycin. For all experiments, the cells were grown to 80–90% confluence, and were subjected to no more than 20 cell passages. The cells were maintained at 37°C in a humidified atmosphere containing 5% CO. The Raw264.7 cells were plated at a density of 2–3 × 10 per ml and preincubated at 37°C for 24 h. After serum starvation for 12 h, the cells were exposed to either LPS (1 μg ml) or LPS + MCE for the indicated time periods (6–24 h). MCE was dissolved in a medium (EMEM, Cambrex Bio Science, MD, USA) and added to the incubation medium 1 h prior to adding the LPS.
LPS ( 026:B6) and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazoleum (MTT) were obtained from Sigma (St Louis, MO, USA). The FBS and antibiotics were purchased from Gibco/BRL (Eggenstein, Germany). The antibodies were obtained from BD Bioscience (USA), Cayman (USA) and Zymed (USA), and the NC paper used was Schleicher & Schuell (USA). The TNF-α, IL-6 and interleukin-1β (IL-1β) ELISA kits were purchased from Pierce endogen (Rockford, IL, USA).
The level of NO production was monitored by measuring the nitrite concentration in the cultured medium. Briefly, the samples were mixed with Griess reagent (1% sulfanilamide, 0.1% -1-naphthylenediamine dihydrochloride and 2.5% phosphoric acid) and incubated for 10 min at room temperature. The absorbance was measured at 540 nm using a Titertek Multiskan Automatic ELISA microplate reader (Model MCC/340, Huntsville, AL).
The Raw264.7 cells were plated at a density of 5 × 10 cells per well in a 96-well plate to determine the cytotoxic concentrations of MCE. The cells were exposed to MCE at concentrations of 0.1 and 0.3 mg ml at 37°C under 5% CO. After incubating the cells in the presence of MCE, the viable cells were stained with MTT (0.5 mg ml) for 4 h. The media were then removed and the formazan crystals produced in the wells were dissolved in 200 μl of dimethylsulfoxide (DMSO). The absorbance was measured at 540 nm using a Titertek Multiskan Automatic ELISA microplate reader (Model MCC/340, Huntsville, AL). The cell viability was defined as the % of untreated control cells [i.e. viability (% control) = 100 × {(absorbance of MCE-treated sample)/(absorbance of control)}].
The cells were lysed in the buffer containing 20 mM Tris (pH 7.5), 1% Triton X-100, 137 mM sodium chloride, 10% glycerol, 2 mM EDTA, 1 mM sodium orthovanadate, 25 mM beta-glycerophosphate, 2 mM sodium pyrophosphate, 1 mM phenylmethylsulfonylfluoride and 1 mg ml leupeptin. The total cell lysate was prepared by centrifuging the cells at 10 000× for 10 min and collecting the supernatant. The expression of iNOS, cyclooxygenase-2 (COX-2) and phosphorylated inhibitor of κBα (p-IκBα) was immunochemically monitored with the total lysate fraction using anti-mouse iNOS, COX-2 and p-IκBα antibodies, respectively. The bands for the iNOS, COX-2 and p-IκBα proteins were visualized using ECL western blotting detection reagents (Amersham Biosciences, NJ, USA) according to the manufacturer's instructions.
A double-stranded DNA probe for the consensus sequence of nuclear factor κB (NF-κB) (5′-AGTTGAAGGC-3′) was used for the gel shift analyses after end labeling the probes with [γ-P]ATP and T polynucleotide kinase. The reaction mixtures contained 2 μl of 5× binding buffer (20% glycerol, 5 mM MgCl, 250 mM NaCl, 2.5 mM EDTA, 2.5 mM dithiothreitol, 0.25 mg ml poly dI-dC and 50 mM Tris–Cl, pH 7.5), 2 μg of the nuclear extracts and sterile water in a total volume of 10 μl. The reactions were initiated by adding 1 μl of the probe (10 c.p.m.) and preincubating the resulting mixture for 10 min. Incubation was continued for 20 min at room temperature. The samples were loaded onto 4% polyacrylamide gels at 140 V. The gels were removed, fixed and dried, followed by autoradiography.
For the cytokine immunoassays, the cells (1 × 10 per ml) were preincubated with MCE for 1 h and further cultured for 6 or 12 h with 1 μg ml of LPS in 6-well plates. The supernatants were removed at the allotted times and the level of TNF-α, IL-6 and IL-1β production was quantified using an ELISA kit (Pierce endogen, Rockford, IL, USA) according to the manufacturer's instructions. Briefly, 50 μl of biotinylated antibody reagent and the samples were added to the anti-mouse TNF-α, IL-6 and IL-1β precoated 96-well strip plates. The plates were covered and kept at room temperature for 2 h and washed three times in a prepared washing buffer. This was followed by the addition of 100 μl of Streptavidin–HRP concentrate. After 30 min incubation at room temperature, the wells were washed three times, and 100 μl of TMB substrate solution was then added and developed in the dark at room temperature for 30 min. The reaction was quenched by adding 100 μl of TMB stop solution, and the absorbance of the plates was at 450 nm to 550 nm using an automated microplate ELISA reader. A standard curve was run on each assay plate using recombinant TNF-α, IL-6 and IL-1β in serial dilutions. Each kit was specific to TNF-α, IL-6 or IL-1β and does not measure other cytokines.
MCE was added to the culture medium 1 h before the addition of 1 μg ml LPS. LPS-treated cells were further cultured with vehicle or MCE for 24 h. The cultured medium was collected and assayed with ELISA kit (RnD Systems, Minneapolis, MN, USA). Cultured medium was incubated in goat anti-mouse IgG coated plate with acetylcholinesterase linked to PGE and PGE monoclonal antibody for 18 h at 4°C. The plate was emptied and rinsed five times with wash buffer contained in the kit. And then, 200 μl of substrate reagent was added to each well and incubated for 1 h at 37°C. The developed plate was read at 405 nm and the PGE concentration of each sample was determined according to the standard curve.
The data were expressed as a mean ± SD of the results obtained from a number () of experiments. One-way analysis of variance (ANOVA) was used to assess the significant differences between the treatment groups. For each significant effect of treatment, the Newman–Keuls test was used to compare the multiple group means. A value <0.05 was considered significant.
The inhibition of NO production by MCE was investigated by measuring the level of NO production in Raw264.7 cells treated with 0.1 and 0.3 mg ml of MCE. As shown in , in the LPS plus MCE groups, the level of NO production decreased in a concentration and time-dependent manner compared with the LPS only group. In the 0.1 and 0.3 mg ml of MCE group, the level of NO was significantly inhibited at 12 h and 24 h (A).
Cell viability was measured at different MCE concentrations for different times using an MTT assay in order to determine if the decrease in NO production was caused by a decrease in the cell population as a result of MCE-induced cytotoxicity. The results showed that MCE concentration dose revealed no cytotoxicity over a 6 to 24 h period (B).
We treated LPS to Raw264.7 cells, collected media every 3 h for 24 h, detected cytokines from the collected media, and then determined appropriate time for each cytokine assay. TNF- α, IL-1β and IL-6 were highly induced by LPS at 12, 12 and 6 h, respectively (data are not shown). TNF-α, IL-1β and IL-6 from the culture media were analyzed at the appropriate time. As shown in , 0.3 mg ml of MCE significantly reduced the level of TNF-α (A), IL-1β (B) and IL-6 (C) at 12, 12 and 6 h incubation. LPS-activated PGE also decreased by MCE as compared to control (D). PGE per 1 ml medium was 6.89 pg in control and increased to 48.6 pg in LPS group. However, PGE was reduced to 30.2 and 14.5 pg ml in 0.1 and 0.3 mg ml of MCE, respectively.
The expression level of the iNOS protein in the cytosol fraction was next examined using immunoblotting analysis. The iNOS protein was strongly induced by LPS. The groups of 0.1 and 0.3 mg ml of MCE showed a decrease in iNOS protein expression in a concentration-dependent manner. As shown in , 0.3 mg ml of MCE with LPS strongly suppressed the induction of iNOS.
COX-2 plays a key role in the development of inflammation (,). MCE was further investigated to determine if it affects the COX-2 expression levels. As shown in this experiment, the COX-2 protein was strongly induced by LPS. 0.1 mg ml of MCE slightly suppressed the induction of COX-2, but 0.3 mg ml of MCE with LPS strongly suppressed the induction of COX-2 ().
In order to determine if MCE can directly affect p-IκBα expression in macrophage cells, the level of p-IκBα protein expression was assessed immunochemically in Raw264.7 cells incubated with or without MCE. LPS increased the p-IκBα level. However, 0.1 mg ml of MCE reduced the LPS-inducible p-IκBα expression level, and 0.3 mg ml of MCE markedly reduced the protein levels of p-IκBα expression in a dose-dependent manner (A).
NF-κB is activated in cells challenged with LPS, and is involved in the transcriptional activation of the responsive genes. Gel shift analysis was carried out to determine if MCE alters the NF-κB DNA binding activity. LPS (1 μg ml, 1 h) increased the binding activity of the nuclear extracts to the NF-κB DNA consensus sequence. Although the treatment of macrophages with 0.1 mg ml of MCE for 1 h prior to the addition of LPS slightly inhibited the LPS-inducible increase in the band intensity of NF-κB binding, a pretreatment with 0.3 mg ml of MCE significantly (<50%) suppressed the band intensity that was increased by LPS (B).
Inflammation is the first response of the immune system to infection or irritation. It is caused by cytokines such as TNF-α, IL-1 and IL-6 (), and by eicosanoid such as PGE (). Thus, inhibitors of these cytokines have been considered as a candidate of an anti-inflammatory drug. Moutan Cortex has been used to diminish inflammation in oriental medicine. However, few studies have been conducted to evaluate the effects of Moutan Cortex on inflammation. Herbal medicine, keishibukuryogan (Gui-Zhi-Fu-Ling-Wan) containing Moutan Cortex, was reported to decrease disease activity evaluated by levels of inflammatory cytokines (). In this study, we show that MCE could modulate the regulatory mechanism of NO, cytokines and PGE in the LPS-activated Raw264.7 cells. MCE inhibited the level of NO, PGE, TNF-α, IL-1β and IL-6, and the expression of iNOS and COX-2 activated by LPS. These inhibitory effects were mediated through the inhibition of IκBα phosphorylation and NF-κB activation.
NO is a free radical produced from -arginine by nitric oxide synthases (NOSs), and an important cellular second messenger (). The modulation of iNOS-mediated NO release is one of the major contributing factors during the inflammatory process (). At adequate concentrations, NO can generate or modify intracellular signals, thereby affecting the function of immune cells, as well as tumor cells and resident cells of different tissues and organs. However, its uncontrolled release can cause inflammatory destruction of target tissue during an infection (–). Moutan Cortex has cytoprotective effects against NO-mediated neuronal cell death in cultured cerebellar granule cells (CGCs) (). In our study, MCE reduced the level of iNOS expression in a concentration-dependent manner.
PGE is generated by the sequential metabolism of arachidonic acid by cyclooxygenase and it is associated with inflammation (). Cyclooxygenase exists in two isoforms; COX-1 and COX-2. COX-1 is a constitutively expressed enzyme with general housekeeping function. COX-2 is an inducible isoform of cyclooxygenase (,), and its most important role is in inflammation. Herb medicine containing Moutan Cortex acts as a potent inhibitor of COX-1 and COX-2 (). Moutan Cortex also showed analgesic effect on the PGF2α-induced allodynia in mice (). We found that MCE could suppress the induction of COX-2 by LPS, and consequently reduce the production of PGE in a dose-dependent manner.
NF-κB has been implicated in the expression of iNOS and COX-2 protein. Activation of the NF-κB signaling cascade results in the complete degradation of I-κB via phosphorylation and ubiquitination. Paeonol showed regulatory effects on inflammatory cytokines-related anaphylactic reaction (), and cerebral disorder (). Recently, paeonol was reported to reduce TNF-α-stimulated iNOS protein and mRNA expression via inhibiting NF-κB activation (). In this study, MCE inhibited the phosphorylation of I-κBα and LPS-inducible increase in the band intensity of NF-κB binding. Thus, it seems that anti-inflammatory effect of MCE is partly responsible for paeonol. To evaluate the anti-inflammatory activity of paeonol in LPS-induced macrophage, studies on the effects of paeonol on the mechanism of LPS-induced NO, cytokines and PGE are needed.
In conclusion, we determined that the MCE can have anti-inflammatory activity by suppressing the phosphorylation of I-κBα and the activation of NF-κB, and by inhibiting the expression of iNOS and COX-2 in LPS-activated Raw264.7 cells. |
italic
xref
#text
italic
xref
fig
#text
italic
xref
#text
Twenty-two species (13.7%) in the genus , out of 160 species, have so far been phytochemically investigated namely: J.Sincl., Scheff, Mat-Salleh, . Hu, Finet & Gagnep., Hook. f. & Thoms., Hook. f. & Thoms., Hook. f. & Thoms., Merr., (W.T.) Wang P.T. Li, Hook. f. & Thoms., Craib, J. Sincl., Hook. f. & Thoms., Miq., Hook. f. & Thoms., J. Sincl., King and Airy Shaw. These phytochemical studies have resulted so far in the isolation of two very distinct classes of lipophilic secondary metabolites: acetogenins and styryl-lactones, both of them possessing complex stereochemistry and existing in different stereoisomeric forms (). Testing of these chemicals for cytotoxicity showed that both acetogenins and styryl-lactones are toxic for several human tumors cell lines. Note that both acetogenins and styryl-lactones are cytotoxic for mammalian cells as the result of distinct biochemical pathways which however take both their molecular origin near or in the mitochondrial membrane and or mitochondrial respiratory system (). To date some evidence clearly demonstrate that acetogenins have beneficial effects against the growth of tumors, including ovarian tumors (), gastric tumors () and multidrug-resistant cancerous xenografts () via the activation of caspases enzymatic cascades (). Most phytochemical reports found so far on species deal with the chemical constituents of a medicinal species: which abounds with cytotoxic acetogenins () as mentioned further.
Acetogenins are unusual series of polyketides which have so far only been characterized from members of the family Annonaceae including in the genus particularly , G. and (,). In the genus , acetogenins were first characterized as the active principles responsible for shrimp lethality from the bark of collected from Thailand. Extract of the bark showed toxicity in the brine shrimp test and showed murine cytotoxicity in the 3PS (P388) leukemia bioassay. The cytotoxicity of this extract compelled a series of phytochemical studies which resulted in the identification of a series of cytotoxic acetogenins including notably: (2,4- and -)-annomontacinones (i), annonacin (ii), giganenin (iii), gigantecin (iv), 4-deoxygigantecin, (2,4- and )-gigantecinones, 4-acetylgigantetrocin A (v), goniotrionin (vi), gigantransenin A (vii) and C (viii), gigantrionenin (ix), gigantetrocin (x), goniotetrocine (xi), (2,4- and )-gigantetrocinones, gonionenin (xii), (2,4- and )-gonioneninones (xiii), 4-deoxygigantenin (xiii), 4-deoxyannomontacin (xiv), goniothalamicin (xv), pyranicin (xvi), gigantriocin (xvii), goniotriocin, (2,4- and )-isoannonacins, longicoricin, longifolicicin, longimicin C, -gigantrionenin (xviii), pyragoniocin (xix), xylomaticin, and (2,4- and )-xylomaticinones () (). Gigantransenins A, and C showed selective inhibitory effects on the human breast tumor cell-line (MCF-7) comparable with the potency of adriamycin (). Both goniotetracin, and 2,4-- and -gonioneninone are selectively and significantly cytotoxic to the human pancreatic tumour cell line (PACA-2) (). Pyranicin exhibited a selective cytotoxic against the pancreatic cell line (PACA-2) in a panel of six human solid tumor cell lines, with pyranicin showing 10 times the potency of adriamycin ().
Jiang . isolated donhepocin (xx), goniodin (xxi), donhexocin (xxii) and donbutocin (xxiii), from Finet & Gagnep. collected from Guangxi Province, China (). Gardnerilins A (xxiv) and B (xxv) from Hook.f. & Thoms collected from DiaoLo mount, Hainan Province, China, gave cytotoxic IC values against Bel 7402 human tumor cell lines of 3.6 and 8.5 μg/ml, respectively (,) (). The mode of action of acetogenins is discussed next.
Acetogenins have very potent and diverse biological effects owing to the fact that they inhibit enzymatic activity of a key enzyme in Eukaryotic cells: mitochondrial NADH-ubiquinone oxidoreductase (complex I). To date the most potent existing inhibitor of this enzyme is an acetogenin known as bullatacin (). In regards to the precise molecular mode of action of acetogenins against the enzyme, there is an expanding body of evidences to suggest that the most lipophilic moieties are embedded in the mitochondrial membrane allowing suitable position of the pharmacophore. One might set the hypothesis that the tetrahydrofuran (THF) or tetrahydropyran rings as well as the free alkyl substituent fix the molecule, whereas the lactones maintained by an alkyl spacer acts on the active site of the enzyme as illustrated in (,). The work of Motoyuki . () lends strong support to that hypothesis. They synthesized series of acetogenins and assessed their activity against bovine heart mitochondrial complex I and showed that the length of the alkyl spacer and the polarity of THF surroundings were very important structural factor and that the γ-lactone and THF ring moieties act in a cooperative manner on complex I with the support of some specific conformation of the alkyl spacers as illustrated in .
The cytotoxic activity of acetogenin has prompted further work in an effort to discover synthetic acetogenins (). Oberlies . () studied the cytotoxicity of acetogenins toward cancerous and normal cells and showed that they are selectively cytotoxic against cancerous cells and also effective for drug-resistant cancer cells, while exhibiting only minimal toxicity to ‘normal’ non-cancerous cells. However, further work is needed to render acetogenins more specific to cancerous cells and very much less active against normal cells or significantly heavy side-effects will preclude clinical trials. A possible approach would perhaps be to use antigen-guided or receptor-guided forms of administrations by associating acetogenins to specific carriers, hemisynthesis could be of value in this instance. More specific cytotoxic principles from species are styryl-lactones reviewed in the next section.
Styryl-lactones are low molecular weight phenolic compounds, which, like acetogenins are essentially found in members of the Annonaceae family and present a lactonic pharmacophore (). Examples of styryl-lactones from species are goniothalamin (i), altholactone (ii) and cardiopetalolactone (iii) ().
Jewers . () first reported goniothalamin as the active constituent of the bark of Miq. and collected in the peat-swamp of Sarawak. Altholactone was characterized from Scheff. collected in the National Park of Varirata in the Central Province of Papua New Guinea and from the Mat-Salleh collected in Malaysia (,). Cardiopetalolactone was characterized from the stem bark of Hook.f. & Thoms. collected from Palaruvi forest in Kerala in India, with altholactone, goniopypyrone, goniothalamin, goniodiol (iv), goniofufurone (v) and goniofupyrone (vi) (,). Goniofufurone, goniopypyrone, goniothalamin, goniodiol, goniotriol (vii) and 8-acetylgoniotriol (viii) were isolated from the roots of (,). An isomer of altholactone, (+)-isoaltholactone (ix), was isolated from stem bark of , and from the leaves of J. Sincl. and the roots of Miq. (). Goniolactones A–F were identified from the roots of (). Digoniodiol, deoxygoniopypyrone A, goniofupyrone, goniothalamin, deoxygoniopypyrone A, gonodiol-8-monoacetate and gonotriol (x) and were characterized from the aerial parts of collected in the southern part of Taiwan near the coastal regions (). The petroleum ether extract of the stem bark of collected in Bangladesh yielded 5-isogoniothalamin oxide (). 5-Acetyl goniothalamin (xii) was characterized from King collected in Bangladesh (). Chen . () isolated howiinol A from Merr. (xii). The mode of cytotoxic action of styryl-lactone is described subsequently.
The evidence currently available clearly indicate that goniothalamin and congeners are toxic for several sorts of cancer cells cultured including HL-60 leukemia cells, breast cancer cell line MCF-7, liver cancer cell line HepG2, PANC-1, HeLa cell lines () (). Current paradigms of apoptosis suggest that styryl-lactones from activate in mammalian cells the caspases enzymatic cascades via a loss of mitochondrial transmembrane which results in the release of mitochondrial cytochrome (). To date, the very precise premitochondrial mechanism involved in this activation remains an enigma, and an exciting fact is that the activation of caspases, 3, 6, 7 and 9 is a sign of TRAIL receptors/Bax activation (). Other examples of goniothalameous styryl-lactones of possible chemotherapeutic value are altholactone, goniolactone B and howiinol. Altholactone is apoptogenic in HL-60 promyelocytic leukemia cells via oxidative stress and mitochondrial respiratory abrogation (,). Goniolactone B exhibited significant cytotoxicity against A2780, HCT-8 and KB cells with IC values of 7.40, 4.43 and 7.23 μM, respectively (). Howiinol A showed significant antitumor activities toward human tumor cell and (). A remarkable advance in the pharmacological knowledge of howiinol A has been provided by the work of He . Using techniques of cell growth curve determination, MTT test, soft agar colony assay and experimental therapy of transplantable tumors in mice, they showed that howiinol exerts potent inhibitory effect on cancer cells including drug-resistant cell line, KB/VCR 2000, whereas normal cells are less affected. Howiinol is active in rodents infected with H22 hepatoma and Lewis lung cancer and ascetic sarcoma 180. In addition to flow cytometry technique, they showed that the cycle of howiinol A block is used to analyze the cell cycle of L1210 cells from G1 phase to S phase with structural damage on DNA molecules. Tian . () showed that styryl-lactones which are cytotoxic against both HepG2 and HepG2-R cell lines show less toxicity on normal mice hepatocytes as the IC values of them on normal mouse hepatocytes were about 3 times of that on HepG2. They demonstrated that cells treated with goniothalamin and altholactone stopped to multiply at G/M and were apoptotic, whereas cells with chromosomes gathered at the equator were easily found in gonodiol-treated cultures.
Indicating that not all s styryl-lactones are exclusively apoptogenic, Zhong . () investigated the apoptosis-inducing effect of styryl-lactones from , on human promyelocytic leukemia HL-60 and showed the activation of caspase-3, reduced the expression of the anti-apoptotic gene Bcl-2, and increased the expression of the pro-apoptotic gene Box via cAMP-dependent protein kinase mechanism. Taking into consideration the available evidence, one might propose the hypothesis that goniothalamin and congeners induce apoptosis at the TRAIL-BAX system level via protein kinase modulation. Protein kinase has long been known to be involved in cell growth and proliferation. Wang . showed that protein kinase is involved in apoptosis mediated by TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) (). An example of styryl-lactone which inhibits kinase is flavokavain A from kava, or in the closely related family Piperaceae (). The shows the similitude of chemical structure between flavokavain and goniothalamin.
A possible mechanism of action for styryl-lactones would be a cAMP-dependent protein kinase-mediated TRAIL-induced apoptosis, by stimulating TRAIL-induced translocation of Bax from cytosol to mitochondria, loss of mitochondrial transmembrane potential, and subsequent release of cytochrome from mitochondria and activation of caspases, SMAC/Diablo, endo G and finally chromatin deterioration (). Protein kinase modulators are of immense therapeutic usefulness. Note that flavokavains are present in the genus , as discussed next.
The aerial parts of have yielded the known flavonoids 2′-hydroxy-4,4′,6′-trimethoxychalcone (flavokavain A), 2′,4′-dihydroxy-4,6′-dimethoxydihydrochalcone, 4,2′,4′-trihydroxy-6′-methoxydihydrochalcone, 5,7,4′-trimethoxyflavanone (naringenin trimethyl ether) and 7-hydroxy-5,4′-dimethoxyflavanone (tsugafolin) together with three novel compounds, the dimer characterized as ()-1β,2α-di-(2,4-dihydroxy-6-methoxybenzoyl)-3β,4α-di-(4-methoxyphenyl)-cyclobutane, 2′,4′-dihydroxy-4,6′-dimethoxychalcone and 2′-hydroxy-4,4′,6′-trimethoxydihydrochalcone (). A similar study of the aerial parts of led only to the isolation of the known flavonoids myricetin 4′--methyl ether-3--α--rhamnopyranoside (mearnsitrin) and myricetin-3--methyl ether (annulatin), together with a series of triterpenes friedelinol, friedelin and betulinic acid ().
Isoquinoline alkaloids were characterized from (). Other miscellaneous secondary metabolites isolated from members of this genus include goniopedaline, a phenanthrene lactam, aristololactam A-II and its -diacetyl derivative, taliscanine, aurantiamide acetate and β-sitosterol and its β--glucoside were isolated from the leaves and twigs of Hook.f. & Thoms. (). 3-Amino naphthoquinones were characterized from the stem bark of (). Alkaloids were characterized from G. and essential oils were distilled from and (,). In the genus , 138 species still await to be phytochemically investigated, including , the antibacterial property of which is reported in the next section.
or in Malay (climber of the elephant bringing forth) is a small tree found from Penang to Selangor and Pahang used by Malays apparently freely, either alone or with other substances after childbirth, and taken internally to prevent bacterial infection (). The plant is known to exhibit potent schizonticidal activity (). We report the first evaluation of the antibacterial activity of hexane, dichloromethane and aqueous fractions of . The plant was collected from 6° North and 98° East, near Kuala Kangsar, State of Perak, Malaysia in August 2004, 300 m above sea level. The plant material was identified on comparison with specimens available at the Herbarium of the ‘Forest Research Institute of Malaysia’, Kepong, Malaysia. A voucher specimen (number W1332) has been deposited in our Herbarium collection for future reference. Finely powdered, air-dried leaves of (800 g) were extracted with methanol (2 l) using a soxhlet apparatus. Hexane (250 ml), dichloromethane (250 ml), and water fractions (250 ml) were obtained by the partitioning of liquid methanol extract (250 ml) (yield: 5.52, 8.43, 64.5). The different fractions obtained were concentrated with a rotary evaporator and brought to complete dryness over water bath to yield the crude extracts. Hexane fraction (yield: 5.52) gave a positive chemical test for steroids, dichloromethane fraction (yield: 8.43) gave a positive chemical test for steroids and terpenes, and aqueous fraction (yield: 64.5) gave a positive chemical test for tannins (). These extracts were screened for antibacterial activity using the following antibacterial assay.
The crude methanol extract of and fractions were subjected to antimicrobial assay using the disc diffusion method of Bauer . (). Both Gram-positive and Gram-negative bacteria () were obtained from the stock cultures of the Department of Medical Microbiology at the University of Malaya. The organisms were of the American Typed Culture Collections (ATCC) and some nosocomial isolates. The organisms included sp., ATCC 25923, ATCC 29213, ATCC 24922, ATCC 27853, ATCC 25922, and a yeast (ATCC 90028). The organisms selected for testing in this experiment are commonly responsible for foodborn and nosocomial bacterial infections (). Mueller–Hinton agar was prepared according to the manufacturer's instruction. It was dispensed into sterile plates in 20 ml aliquots. After gelling and drying, the plates were seeded with appropriate organisms by streaking evenly in three planes onto the surface of the medium with cotton swabs. The inoculum was dried for 5 min. Sterile filter paper disks (6 mm diameter) soaked with 50 μl of extract (100 mg/ml) were placed onto the agar with flamed forceps and gently pressed down to ensure contact. Streptomycin (10 μg/disc) and nystatin (100 IU) were used as a positive standard against bacteria and fungi as they are both inexpensive and broad spectrum antimicrobials. The plates were incubated at 37°C for 24 h. The zones of inhibition were measured with a ruler. The experiment was carried out in triplicate. Results obtained for antibacterial activity of the crude methanol extract of and fractions are reported in . Methanol, hexane, dichloromethane and water used for reconstitution of the extracts showed no activity. Analysis of the data revealed that among the tested fractions, the dichloromethane fraction exhibited the highest rates of antibacterial activity. It showed antibacterial activity against ATCC 25923: 23 mm, ATCC 29213: 27 mm, ATCC 24922: 20 mm, ATCC 25922: 19 mm, sp.: 20 mm, : 13 mm (), : 28 mm, : 13 mm, and sp. 17 mm. The extract inhibited the growth of ATCC 90028. It was inactive against ATCC 27853.
This report is the first data available on the antibacterial activity of and lends support the traditional use of as post-partum remedy. An interesting development from these results would be first to identify the active constituents and next to study their precise molecular activity against bacteria. Note that mitochondria in eukaryotic cells take their origin in pro-bacterial ancestors from which they inherited NADH:ubiquinone oxidoreductase (). One can perhaps envisage a new antibacterial pathway that would encompass a ‘bacterial apoptosis’. One wonders.
was investigated as part of our study on the medicinal plants of Asia-Pacific (,,) A critical factor for Goniothalamus’ use as a medicinal herb is its content of styryl-lactones, which promote apoptosis in mammalian cells. One might propose the hypothesis that the abortifacient and/or post-natal and anti-inflammatory reported traditional uses of species might involve styryl-lactones since apoptosis is known to play a crucial role in trophoblasts of patients with recurrent spontaneous abortion of unidentified cause, and in T cells in the human decidua as defense mechanism against rejection of fetal allograft by the maternal immune system (,). In addition, goniothalamin induces apoptosis in vascular smooth muscle cells, the growth of which is required to allow embryo implantation and the development of the blood supply for fetal survival and inhibit the cell surface expression of intercellular adhesion molecule 1 and vascular cell adhesion molecule 1 on the surface of murine endothelial cells (,)
In regards to the result obtained for antibacterial activity of the crude methanol extract of and fractions, it can be concluded that the dichloromethane extract of is very active against both Gram-positive and Gram-negative bacteria. and the results obtained tend to answer positively the question of Chinnok . (). This work illustrates the fact that the careful study of the biochemical architecture of medicinal plants represents a fascinating and fruitful aspect of pharmaceutical research (). In regards , it will be interesting to know whether further studies on this plant disclose any molecules the treatment of nosocomial urinary, respiratory and wound nosocomial infections ( and ) which are developed by hospital patients.
In summary the evidence for the existence of anticancer, antibacterial and antiviral agents in the genus is strong and it seems likely that further consistent and systematic research on this genus of flowering plants will lead to the discovery of antineoplastic and antimicrobial agents. If enough botanical, phytochemical and pharmacological work is dedicated to this discrete tropical genus of flowering plants, a couple of drugs for the treatment of tumors and/or bacterial and even viral infections should be developed in the relatively close future. |
Amplification or deletion of chromosomal segments can lead to abnormal mRNA transcript levels and result in the malfunctioning of cellular processes. Locating chromosomal aberrations in genomic DNA samples is an important step in understanding the pathogenesis of many diseases, most notably cancers. Microarray-based comparative genomic hybridization (array CGH) is a powerful technique for measuring such changes (). To realize the promise of the array CGH technique, it is very important to develop effective methods to identify aberration regions from array CGH data. Existing analysis methods () can be roughly classified into two categories: smoothing-based () and segmentation-based (). The latter approaches explicitly model the observed array data as a series of segments, with unknown boundaries and unknown heights estimated from the data by employing certain optimization criterion. While the boundary points thus identified are reliable, the aberration regions identified may be less so, in the sense that some of them may be false positives. Smoothing-based approaches assume that true signals in a specific region, aberration or non-aberration, are smoother than any kind of noise superimposed on the signals. Therefore, they attempt to reduce noise by comparing individual data points to their adjacent ones and modifying them. While such methods can reduce the number of false aberration regions identified, the boundary points detected are usually less accurate than segmentation-based methods. It would be very desirable to develop new methods for analyzing array CGH data, with both the merits of smoothing and segmentation based approaches. Such a goal may not be fully accomplished by just incorporating mean or median smoothing to a segmentation-based method.
The term ‘noise’ is often used in biology to describe experimental measurement imprecision. In particular, when evaluating the accuracy of an algorithm for detecting aberrations, it is commonly assumed that noise in the data follows normal distribution. However, this important assumption has not been verified/falsified based on the analysis of experimental data. More importantly, one has to ask whether the performance of an algorithm depends on the character of the noise, and if yes, can one exploit the characteristics of noise to improve detection of aberrations?
To address the above questions, in this work, we treat any deviations from mean values as noise. Therefore, our noise essentially represents the array CGH measurements themselves, encompassing both global measurement imprecision and localized underlying biological alterations. Based on the analysis of bacterial artificial chromosomes (BACs) arrays with an average 1 mb resolution (), 19 k oligo arrays with the average probe spacing < 100 kb () and 385 k oligo arrays with the average probe spacing of about 6 kb , we show that when there are aberrations, noise in all three types of arrays is highly non-Gaussian and possesses long-range spatial correlations. We also show that such noise indeed leads to worse performance of existing methods for detecting aberrations in array-CGH than the Gaussian noise case. Fortunately, such noise can be exploited to devise a novel algorithm for analyzing array-CGH, which is capable of identifying both aberration regions as well as the boundary break points very accurately. We also address the fundamental question of how to assign certain confidence level to an aberration region and boundaries detected, by proposing a new concept, posteriori signal-to-noise ratio (p-SNR).
t
h
i
s
s
e
c
t
i
o
n
,
w
e
c
a
r
r
y
o
u
t
d
i
s
t
r
i
b
u
t
i
o
n
a
l
a
n
a
l
y
s
i
s
a
s
w
e
l
l
a
s
s
p
a
t
i
a
l
c
o
r
r
e
l
a
t
i
o
n
a
n
a
l
y
s
i
s
o
f
a
r
r
a
y
-
C
G
H
n
o
i
s
e
,
a
n
d
a
s
s
e
s
s
t
h
e
e
f
f
e
c
t
o
f
s
u
c
h
n
o
i
s
e
o
n
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
p
u
b
l
i
s
h
e
d
a
l
g
o
r
i
t
h
m
s
f
o
r
d
e
t
e
c
t
i
n
g
a
b
e
r
r
a
t
i
o
n
s
f
r
o
m
a
r
r
a
y
-
C
G
H
d
a
t
a
.
B
e
l
o
w
,
w
e
f
i
r
s
t
d
e
s
c
r
i
b
e
d
a
t
a
.
We now present an algorithm for detecting aberrations from array CGH data that has considerably taken into account the character of array noise. Since the method has merits of both smoothing and segmentation based methods, we denote it by CGHss. It consists of five steps. They are detailed below.
To make the above steps concrete, we have simulated an artificial chromosome data consisting of 100 probes, with the centering 10 probes having aberrations. The simulated data () is shown in a. b–g show the signals (), , (), , () and (), respectively. The red dashed lines in f correspond to a more or less arbitrarily chosen threshold value .
Note all the three thresholds, , and , can be defined by users. Also note that the ROC curves presented below do not depend on sensitively. After we describe the concept of p-SNR, we shall provide some guidelines as how to choose and .
We now compute the ROC curves for our method under the same setting as when we discussed the CGHseg algorithm (, green and purple curves). They are shown in as red and black curves, for array noise and Gaussian noise, respectively. First we note that comparing with the results of a recent comparison paper (), our method is comparable to the best smoothing based methods. Next, we make two interesting observations from (i) The ROC curves for our method are very similar for the array noise and the Gaussian noise. This is because our method has fully taken into account the character of the array noise. (ii) Our method is more accurate than CGHseg, especially when SNR is low and the aberration width is narrow. As we shall argue later, in the low SNR case, a good threshold to choose would correspond to the case of TPR + FPR =1 = 1. Under such a criterion, our method is almost 20% more accurate than CGHseg, for the case of SNR = 1 and aberration width = 5. Therefore, our method will be especially useful for processing array CGH data with small SNR [the 19 k oligo array data () belong to this category, as we shall show in the next section].
Next we examine the performance of our method for detecting the boundaries of aberrations. compares our result with those obtained using the CGHseg algorithm () The simulated data, shown in a, consists of a single aberration region, from 48 to 52. Due to fairly low SNR, visually it is hard to tell which region might be the aberration region. However, our method correctly detects both the aberration region and the two breakpoints, as shown in b. The CGHseg algorithm, however, does not seem to be able to cope with such low SNR data. This is evident in (c and d) (as well as Supplementary Figures 4–7): The CGHseg method may not only produce false aberration regions, but also fail to detect boundary points correctly. The reason that our method works better is that a few steps of our method have used smoothing. In particular, the second step of our method is a lossless smoothing. No other methods have done so.
Finally, we evaluate our method by analyzing 15 BAC array data (). compares aberration detection by using a and b our method and c and d the method of Olshen and Vankatraman Olshen () The cell line is GM05296, with known aberrations on chromosomes 10 and 11. We observe that the two methods perform similarly well. As will be explained later, these two datasets have fairly high SNR, and hence, aberration detection is not too difficult.
We now consider an important question: after one applies an algorithm to identify aberration regions and breakpoints from an array CGH data, how much confidence can one have on the final result? When there are a lot of data available, together with information on the background normal situations, one may follow the procedure discussed by Wang . Wang . (). Here, we consider the more challenging case of only one array data are available.
Our solution is quite simple. It amounts to utilizing the information summarized by ROC curves obtained by numerical simulations as much as possible. From , it is clear that the accuracy of detection depends on two critical parameters, the size of the aberration region and SNR. In fact, from , it is clear that SNR is even more important than the size of the aberration region. Therefore, a good starting point would be to estimate SNR after one identifies one or a few aberration regions. This can be achieved by using the following simple formula:
non-aberration region), respectively. When there are multiple aberration regions, - may be estimated for each aberration region identified. When this is the case, one needs to take the background region to be only around that aberration region. The - for the BAC array data () are listed in the second column of . We observe that they are quite large, indicating that one can have high-confidence in the final detection results for the BAC array data (). The - for the 385 k oligo array data are around 0.9 to 1.5, and are even smaller for the 19 k oligo array data () (around 0.7–1.1). In particular, we have estimated the - for the 385 k and 19 k oligo array data based on the chromosome using sex-mismatched samples. We have found that the p-SNR for the two array platforms are 1.26 and 1.03, respectively, both falling in the range for each type of data listed above. Although we do not have access to the sex-mismatched chromosome data for the BAC array data (), based on our analysis of the other two types of data, we have good reason to believe that the - for the BAC array data's sex-mismatched chromosome would be at least around 3, therefore, much larger than the - of the other two types of data.
Our concept of - suggests that a good starting point to choose the parameters and used in step () of our algorithm may correspond to signal intensity divided by -. This amounts to choosing one SD of the data. This rule suggests an iterative operation: starting from arbitrarily chosen and , calculate the corresponding -, then use the criterion discussed above to obtain a new estimate of and , finally calculate the new -′. If - and -′ are similar, then the two parameters have been chosen appropriately.
We emphasize that our method works excellently if - is high. However, if - is low, then one may choose threshold values that roughly yield TPR + FPR = 1, where TPR and FPR define the ROC curve. In this case, however, one should bear in mind that the classification may be incorrect with a probability of FPR.
In this paper, we have examined noise in array CGH data of three resolution, the BAC array data, the 19 k and 385 k oligo array data, and found that noise is highly non-Gaussian and possesses long-range spatial correlations. We have also developed a novel method for processing array CGH data. The method is a suitable combination of smoothing and segmentation, and has fully taken into account the characteristics of noise in array CGH data. We have shown that the method is as accurate as the best smoothing-based methods for detecting aberration regions, and as accurate as the best segmentation-based methods for finding boundary points. Furthermore, we have proposed a new concept, (-), to quantify the confidence level of aberration regions and boundaries detected. We have found that - for the 15 publicly available BAC array CGH data are all quite large, indicating it is a relatively easy matter to accurately detect aberrations and boundaries from those data. However, - for the four 19k oligo array data are quite small, suggesting it is considerably more challenging to detect aberrations from such array CGH data.
Although we have found that array CGH noise is highly non-Gaussian with long-range spatial correlations, we do not clearly know the mechanisms. A challenging task for future research would be to understand the biological mechanisms of such noise, as well as understand whether those mechanisms are differentially related to different types of diseases.
Being able to identify smaller copy-number changes that affect only a few probes is of particular importance in the field of copy number polymorphism. This is because inherited, germ-line copy number variants are typically much smaller than rearrangements in cancer genomes. For example, two recent papers, one by McCarroll . (), another by Conrad . (), identified a large class of inherited, multikilobase deletion polymorphisms that are predominantly smaller than 20 kb in size. We emphasize that in order to detect small copy number changes, the key is to improve the resolution of the array technology so that at least 2 points can be sampled for the region of interest. If only a single isolated point can be sampled, then it would be impossible by any analysis method to classify it as a true copy number change or just an outlier or noise.
Finally, readers interested in this method are strongly encouraged to contact with the authors (e.g. ) to obtain the source Matlab code. |
Mass spectrometry has much to offer the field of genomic analysis, particularly in terms of multiplexed analysis and accurate quantification. To date, many mass spectrometry-based approaches for genomic analysis have been based on direct detection of nucleic acids particularly using matrix assisted laser desorption ionisation time-of-flight (MALDI TOF) MS analysis. MALDI TOF is well suited to this approach due to the high mass range achievable by TOF analysis, however, MALDI TOF instrumentation is relatively expensive and sample preparation can be quite laborious. In addition, direct analysis of nucleic acids by mass spectrometry suffers from problems such as depurination leading to fragmentation () or cation adduct formation (,). These issues notwithstanding, MALDI TOF analysis of nucleic acids has been applied to DNA sequencing (,), RNA sequencing (,), analysis of DNA tandem repeats (). In particular, PNA probes, with and without non-cleavable mass modifiers, have been used for characterization of genomic DNA libraries (), detection of DNA methylation () and detection of single nucleotide polymorphisms (SNPs) (,) by MALDI TOF mass spectrometry.
Electrospray ionisation (ESI) mass spectrometry has also been used for direct detection of nucleic acids (). ESI-MS has some advantages over MALDI TOF MS particularly the availability of lower cost instrumentation. Moreover, sample handling can be simpler since most molecular biology assays are carried out in solution and such liquid samples are injected directly into the instrument. Furthermore, very high molecular weight species can be analysed due to the propensity for large molecules, such as PCR products, to form multiply charged ions that have relatively low overall mass-to-charge ratios under electrospray ionisation conditions. However, direct analysis of nucleic acids by ESI-MS still suffers from the same problems as MALDI TOF MS such as cation adduct formation (). In addition, the multiply charged ion spectra that are generated for large nucleic acid fragments can be very complicated reducing sensitivity and making multiplexing difficult ().
As an alternative to direct analysis of nucleic acids, mass spectrometry can also be used to detect nucleic acids indirectly through the use of cleavable mass tags, which avoids many of the limitations of direct analysis of nucleic acids while also offering numerous advantages such as ease of multiplexing, more robust detection of tag species and higher sensitivity and less demanding workup and sample preparation for analysis. Mass spectrometric analysis of mass tags, by ESI or MALDI also offers the possibility of accurate quantification through the use of isotopic tags. This capability of mass spectrometry has not really been exploited in genomic analysis but has been quite widely used in proteomic analysis () and is a key advantage of the mass spectrometric approach.
Again MALDI TOF analysis of nucleic acids with cleavable mass tags has been demonstrated by various groups () but it would be advantageous to be able to take advantage of lower cost ESI-MS/MS instruments and to avoid the laborious sample workup requirements of most MALDI approaches. A matrix-free laser desorption approach, which has reduced workup requirements has been demonstrated (,) but this still requires that the sample be spotted onto a MALDI target or hybridized to an array. To our knowledge, only one mass tagging approach employing ESI-MS analysis has been demonstrated (). In this approach, mass tags are photo-cleavably linked to oligonucleotides and tag detection requires a photo-cleavage step and a tag isolation step outside of the mass spectrometer prior to tag detection, i.e. the workup is not much simpler than that required for MALDI TOF analysis.
Here, we describe the synthesis of novel ESI-cleavable Tandem Nucleic acid mass Tag–peptide nucleic acid conjugates and their analysis by ESI-MS/MS. We demonstrate a novel mass tag cleavage method in which source voltages in the electrospray ionisation source are used to cleave an ESI-cleavable linker, by a collision-based process, releasing the mass tag from the oligonucleotide during ionization. This method allows for direct analysis of assay solutions without requiring complex workups to cleave and isolate tags. In principle, this cleavage method would also allow in-line separation, by capillary electrophoresis for example, of labelled nucleic acids with direct spraying of the separated material into the ion source where tag cleavage would take place automatically.
In addition, we demonstrate a novel MS/MS-based tandem nucleic acid mass tag (TNT) design and detection process that allows highly specific detection of TNTs in a complex background. The Tandem Nucleic acid mass Tag design also allows easy synthesis of large sets of isotopic tags supporting the development of multiplexed and quantitative assays. We demonstrate the quantitative nature of the TNT approach. Furthermore, we evaluate ESI-cleavable TNT–PNA conjugates as hybridization probes for the detection of target DNA sequences via the use of ESI-MS/MS.
The TNTs described here are constructed using FMOC peptide synthesis chemistry. Synthesis of peptide nucleic acid–TNT peptide conjugates is relatively straightforward as PNA synthesis can be achieved using the same FMOC protection groups that are used in peptide synthesis (). This means that peptide TNT–PNA conjugates can be synthesized on the same resin in a continuous process. PNA is a useful analogue of DNA and is advantageous for this application due to its high specificity and its neutral backbone, which means that it does not require high concentrations of salt to hybridize making it highly compatible with mass-spectrometry-based detection methods ().
The outline of the general approach for the detection of DNA sequences via ESI-cleavable TNT–PNA conjugates is presented in . The ESI-cleavable TNT–PNA conjugates consist of a PNA probe portion, which interacts with the immobilized target sequence (DNA or RNA) and a peptide tandem nucleic acid mass tag portion, which is ultimately detected. Note that the TNT shown is merely a representation of the tag and not a real structure (). An ESI-cleavable linker connects the PNA probe portion of the conjugate to the tandem nucleic acid mass tag peptide. The complete TNT ‘Parent Tag’ comprises the red ‘Tag Fragment’ portion and the blue ‘Mass Normalizer’ portion shown in . The TNT marker is designed to have a unique combination of parent tag mass and tag fragment mass, released during collision induced dissociation (CID), and it is this pair of masses that serves as the sequence identifier. In a typical scenario (), a set of PNA hybridization probes labelled with different TNTs is first hybridized to the captured target nucleic acids of interest (step). After stringent washes to remove the non-hybridized probes (step), the probes are denatured from the target (step) and injected into an ESI-MS/MS instrument for detection (step). In the mass spectrometer, the TNTs cleave from the PNAs during electrospray ionization (step). The TNT parent tag ions are then selected from the background, fragmented by CID and finally, daughter, tag fragment ions from the fragmentation are detected (step), confirming that the signals are indeed due to the presence of tagged probes, thereby detecting the presence of the target sequences.
The use of this MS/MS-based approach offers high specificity allowing TNT labels to be detected in a background of fragmentation noise. In addition, the MS/MS detection means that tags can share the same mass as long their tag fragment ions are distinguishable from each other. This means that many TNTs can be detected in a compressed mass range. This feature combined with the relatively low overall mass required for TNTs, means that TNT technology will be able to exploit lower cost compact and portable ESI-MS/MS instrumentation that is currently under development ().
FMOC-protected peptides were custom synthesized by PepSyn Ltd (Liverpool, UK) using commercially available FMOC-protected amino acids on a Beckman synthesizer. 4-FMOC-piperazin-1-ylacetic acid was obtained from Fluka (Sigma Aldrich, Dorset, UK). The amino acid isotope C, N-FMOC--alanine was obtained from Cambridge Isotope Laboratories, Inc (Andover, MA). The sequences of the TNT peptides used for the preparation of ESI-cleavable TNT-PNA probes are shown in .
Peptide nucleic acid oligonucleotide syntheses were carried out using a 2-µmol cycle on an Expedite 8900 synthesizer (Applied Biosystems, Foster City, CA). For the preparation of peptide–PNA conjugates, the FMOC-protected peptide TNT sequence was synthesized and left on the resin by PepSyn Ltd with the N-terminal FMOC left intact. The resin was extracted from the peptide synthesizer column and was then loaded into the Expedite synthesizer column (2 µmol per column). PNA synthesis was carried out as normal on the preloaded resin. The yield of each purified conjugate was in the range 4.3–32 OD, which corresponds to a 1.5–21% yield based on the 2-µmol synthesis scale. The sequences and yields of the TNT-PNA probes are shown in .
Biotinylated, fully complementary target 50-mer oligodeoxyribonucleotides for the hybridization experiments were synthesized by Yorkshire Bioscience (York, UK) on a 1-µmol scale. Target sequences are shown in .
Six aliquots of 20 µl of stock solution of the first TNT-PNA probe was mixed with 20, 10, 4, 2, 1 and 0.5 µl of the second TNT-PNA, respectively. To these, water was added to make up the solution to a total of 40 µl. The aliquots were then made up to 80 µl with methanol and formic acid to give a final solution of the TNT-PNA probes in 50:50 water:methanol with 1% formic acid. These samples were then analysed by direct injection ESI-MS/MS.
Five aliquots of 50 µl of MyOne Streptavidin C1 Dynabeads (10 mg/ml suspension) were separated from their storage buffer and washed with twice with 50:50 methanol:water to remove potential mass spec contaminants. The beads were then washed with 1 × Bind & Wash (B&W) buffer. B&W buffer for Dynabead incubation was made up according to the manufacturer's instructions: 20 mM Tris, pH 8.0, 2 mM EDTA, 2 M NaCl. Six aliquots of 20 µl (1 nmol) of stock solution of one biotinylated target was incubated with the Dynabeads. These aliquots were all made up to 40 µl with the addition of 20 µl of 2 × B&W buffer. The biotinylated targets were then incubated at room temperature with the streptavidin beads for 1 h according to the manufacturer's instructions to immobilize the targets on the beads. The target solution was then removed from the beads and the beads were washed twice with hybridization buffer (20 mM Tris, pH 7.5, 10 mM MgCl, 25 mM NaCl). A sixth aliquot of 100 µl of MyOne Streptavidin C1 Dynabeads was made up in the same way but this aliquot was incubated with 40 µl (2 nmol) of stock solution of the same biotinylated target and 40 µl of 2 × B&W buffer.
ESI-MS/MS spectra were obtained on a Micromass Q-TOF Micro mass spectrometer (Micromass (Waters), Wythenshaw, UK). The TNT-PNA oligonucleotides were denatured from the Dynabeads into 50:50 Methanol:Water with 1% formic acid. Mass spectra were externally calibrated using the manufacturer's standards and calibration protocols.
Two pairs of example TNT-PNA oligonucleotide probes are shown in part (i) of a and b. The TNTs are peptides comprising two parts, the tag fragment portion, which carries a charge due to the presence of a tertiary amino-functionality and the mass normalization portion, which remains essentially uncharged. These two portions of the tag are linked by a cleavage enhancement group, a piperazine ring, which also carries the charge of the tag fragment on its tertiary amino-group. The two tag portions both comprise a mass modifier component, which are isotopes of alanine in this tag although a large number of different mass modifiers could be used. It can be seen from that the two tags shown employ the same mass modifier components but the order differs between the tags. Thus, the overall masses of the tags are the same but the tag fragments have different masses, which are equalized by the mass of the mass normalization portion. These tags are designed so that on analysis by collision-induced dissociation (CID), the tag fragment is released to give rise to a uniquely resolvable ion. Thus, this pair of tags allows a pair of PNA probes to be distinguished by MS/MS analysis. Each tag is linked to a PNA oligonucleotide probe by a second linker, comprising aspartic acid and proline that is easily cleaved by CID (). As shown in , the aspartic acid/proline linker is used to cleave the tags from their oligonucleotides during electrospray ionization of the tagged oligonucleotides. The expected structures and mass-to-charge ratios of the cleaved parent tag ions generated by dissociation of the aspartic acid/proline linkage are shown in part (ii) of a and b (). Similarly, the expected structures and mass-to-charge ratios of the tag fragment ions and the structures of the neutral mass normalizer fragments, based on the predictions of the ‘mobile proton’ model of peptide fragmentation (,), are shown in part (iii) of a and b.
illustrates the ease with which a substantial number of tags can be synthesized from a small set of mass modifier components: nine tags, shown in b can be made from three mass modifiers, which are the three commercially available isotopes of alanine shown in a. In fact, the number of tags that can be synthesized increases as the square of the number of mass modifier components, e.g. there are at least five isotopes of alanine with different masses that are commercially available which would actually allow the synthesis of 25 tags using the design presented here.
One issue that emerged from the experiments presented here was a loss of C isotope from the carboxylic acid of alanine when the alanine isotope was present at the C-terminus of the peptide TNT, i.e. in TNT2 shown in . This loss is consistent across every TNT-PNA synthesized with TNT2 and may be an effect of the resin used, which relied on a 4-HydroxyMethyl-Phenoxy Acetic acid (HMPA) linker to the carboxylic acid group of the first amino acid to allow the peptide to be cleaved from the resin at the end of the synthesis. Other resins will be tested in future to avoid the loss of isotope. This has meant that the TNTs synthesized were not completely isobaric as shown in . The mass of the singly charged parent tag ion of TNT1 is 388.2 while that of TNT2 is 387.2. For the MS/MS analysis of these tags, both tags could still be selected simultaneously by the first quadrupole of the Q-TOF instrument, as the mass range that is gated is actually about 3 daltons for the default setting of the instrument. It was found that setting the nominal selection mass to 388 daltons facilitated transmission of both ions without significantly favouring the selection of either tag.
The TNT approach is similar in principle to other mass tagging techniques and enjoys the same features as other approaches, such as ease of multiplexing and the ability to design tag masses to suit applications, with some additional advantages. TNT tags can be made chemically identical, even sharing the same mass as long as the tag fragments are different, so they can act as more precise reciprocal internal standards, which leads to more accurate quantification and the same behaviour in analytical separations, hybridizations and labelling reactions thus avoiding ‘dye effects’ that plague fluorescent methods (,). The use of an MS/MS-based detection method allows TNTs to be selected from background noise thus improving signal to noise ratios. This allows untagged material to be ignored, greatly improving data quality.
To confirm that the TNT–PNA oligonucleotide conjugates cleave as they are expected to (see expected fragment structures and mass-to-charge ratios in ), the TNT-PNA oligonucleotides were analysed by ESI-MS/MS on a micromass quadrupole-time-of-flight (Q-TOF) mass spectrometer. The complete TNT-PNA probe molecules were initially ionized with the cone voltage (an accelerating voltage in the ESI source that can be varied on the micromass instrument to control the levels of collision induced dissociation) set to minimize fragmentation of the whole TNT-PNA probe conjugate ions. The whole molecular ions were selected using the first quadrupole of the instrument and were then subjected to collision induced dissociation at different collision energies. Fragmentation of the complete TNT–PNA conjugate ions was carried out as it allows both the cleavage of the TNT parent tag from the PNA and the cleavage of the tag fragment from the parent tag to be seen simultaneously in the TOF analyser of the Q-TOF instrument thus demonstrating both cleavage processes and their relative efficiencies. Typical results are shown in , in which it can be seen that, as the collision energy is increased, the whole TNT-PNA probe ion fragments to release the parent tag ion as the predominant fragmentation product. As the fragmentation energy is increased further, the parent tag ion undergoes subsequent consecutive fragmentation to give the desired daughter fragment ion. As the collision energy increases further, the intensity of the parent tag ion increases. Similarly, the ratio of the daughter ion to parent ion increases more. These results show that the TNT-PNA probe molecules fragment as anticipated with the aspartic acid/proline linkage cleaving more easily than the piperazine linkage. The higher collision energy spectra are quite noisy as these also contain other products from the fragmentation of the TNT-PNA probe molecule, which would not normally be present when the TNT parent tag ions are analysed by themselves.
The normal mode of analysis is shown in . Here the cone voltage in the electrospray ion source is increased to 25 V increasing the level of fragmentation during ionization thus releasing the parent tag ion from the TNT–PNA conjugate. The parent tag ion is then selected from background by the first quadrupole in the Q-TOF instrument. The parent tag ion is subsequently subjected to CID in the second quadrupole of the Q-TOF instrument. A collision energy of 25 V was used for CID. MS/MS spectra showing the detected tag fragment ions are shown in . These spectra show ratios of two tags and demonstrate the accuracy of quantification of the TNT technology, which is discussed in the next section.
A key feature of the Tandem Nucleic acid mass Tag design is the ease with which large numbers of chemically identical, isotopomeric tags can be made (). Sets of TNT isotopes should have almost identical behaviour in analytical separations () and during the ionization process. This means that it should be possible to use these tags to accurately quantify their associated oligonucleotide sequences as the ratios of the intensities of the TNT isotopomer fragments should reflect the ratios of the concentrations of the probes in solution or hybridized on their targets (). To demonstrate this feature of the TNT design, various experiments were conducted. In these experiments, pairs of TNT–PNA conjugates are analysed by ESI-MS/MS, where the parent tag ions are cleaved from their probes in the ESI ion source using a cone voltage of 25 V. The cleaved parent tag ions are subsequently selected from background by gating both 387.3 and 388.3 ions with the first quadrupole of the Q-TOF instrument. The gated parent tag ions are then subjected to CID in the second quadrupole of the Q-TOF instrument using a collision energy of 25 V followed by mass separation and detection in the TOF analyser to determine the ratios of the tag pairs. In the first set of experiments, pairs of TNT-PNA oligonucleotide probes of the same length and sequence, but with different tags were mixed in a predefined ratio and diluted to determine how well the ratios are conserved as the concentration of probes is decreased. Results are shown in . It can be seen that the ratios of the isotopic TNT-PNA probes are conserved over the range of concentrations investigated.
In a second experiment, pairs of TNT-PNA probes of the same length were mixed in various different ratios. The correlation between the expected and measured quantities of these different TNT-PNA ratios is shown in . It can be seen that there are simple linear correlations between the expected and measured ratios. The blue crosses in indicate the results of experiments where the two TNT-PNA probes with the same sequences but with different isotopomeric tags were mixed, i.e. the whole TNT-PNA probes were isotopes of each other. The measured ratios for these probes closely match the expected ratios.
The red squares in indicate the results of experiments where the two different probe sequences of the same length were mixed, i.e. although their TNT labels were different isotopes of each other, the complete TNT-PNA probes were not isotopes of each other. The red line represents a linear regression through these data points. Although the predicted and expected ratios do not match exactly, there is a good correlation between the results indicating that the measurements are quantitative. Since the TNT-PNA probes were actually different in these experiments, it is pleasing to see that there is correlation between the measured and expected quantities and the result suggests that, in future use, the measurements of the quantity of different targets in a sample could be calibrated against an internal control such as a housekeeping gene or, preferably, a known quantity of a spiked target sequence.
Quantification of hybridized probes was also evaluated. In the first experiment, external calibration of the quantities of hybridized probes was assessed, i.e. the amount of target in one sample was probed with a TNT-PNA whose abundance was then determined by comparison with a second reference sample comprising a predefined quantity of the duplex of the same target sequence and PNA probe sequence, but with a different TNT isotope conjugated to the probe, after the hybridization. In these studies, aliquots of a synthetic biotinylated 50-mer DNA oligonucleotide target was captured onto avidinated magnetic beads and hybridized with TNT-PNA probes with identical probe sequences but different tags. The hybridized beads were then mixed in different ratios. The target, arbitrarily selected, was a fragment of a sequence from the MOMP gene from Chlamydia pneumoniae. A fixed quantity of one TNT-PNA probe complementary to one of the targets was hybridized to the captured target sequences. The aliquots of beads were then washed extensively to remove probe that had not hybridized. The captured TNT-PNA probe mixture was then eluted into 50:50 water:methanol with 1% formic acid (a solvent suitable for ESI-MS/MS analysis) by thermal denaturation. The eluted TNT-PNA and the spike were then injected directly into the Q-TOF instrument for MS/MS analysis. The ratios of the intensities of the two tags derived from the TNT-PNAs should allow the amount of the target sequences in the pooled samples to be determined. shows the actual correlation between the expected and measured quantities of the target sequences. The results are very similar to the experiments where TNT-PNA probes with the same sequence but different tags are simply mixed together: the measured ratio matches very closely the expected ratio. Negative controls in which the target was absent do not show significant binding of TNT–PNA conjugates to the beads so the probe binding is sequence specific. This gives a clear indication that the probes behave quantitatively in hybridization assays and that TNT-PNA probes can be used for accurate quantification.
A further evaluation of the quantification of the TNT–PNA conjugates was carried out to determine whether accurate relative quantification can be derived from TNT-PNA pairs with different PNA sequences, i.e. can a reference sequence in a sample probed with one TNT-PNA be used to quantify a second sequence with a different PNA probe as long as the TNTs used in the probe pair are isotopes of each other. This would enable quantification by internal calibration using spiked sequences, housekeeping genes or similar controls in quantitative expression profiling or diagnostic assays. In these experiments, a pair of biotinylated 50-mer target oligonucleotides was used (MOMP-50 again and a sequence from , ANTHRAX-50; see ). These were captured onto streptavidin-coated magnetic beads. The quantity of one target was fixed while the relative quantity of the second was varied. The captured targets were then hybridized at room temperature with a probe solution comprising equal quantities of MOMP-12-TNT1 and Anthrax-12-TNT2 probes (see ). Probes of the same length were used together. After the hybridization, the magnetic beads were washed as before and the hybridized TNT-PNAs were eluted from their targets on the beads into 50:50 methanol:water with 1% formic acid and analysed as described earlier. Typical results are shown in . As observed in the simple mixture experiments, the measured TNT ratios show a linear relationship with the expected ratios but the measured quantities do not exactly match the expected ratios when the TNT-PNA probes being compared are not true isotopes but the linear relationship does mean that the measurements are quantitative. These data suggest that with appropriate choice of reference sequences, quantitative internal calibration should be achievable, which would be very useful in situations where suitable reference samples are not available for external calibration.
In this article the synthesis, characterization and application of ESI-cleavable TNT peptide–PNA conjugates to the quantitative detection of target DNA sequences by ESI-MS/MS has been described. The conjugates were prepared by first synthesizing the TNT tag peptide sequence in a peptide synthesizer, after which the peptide synthesis resin was transferred to a column compatible with a DNA synthesizer in which PNA can be prepared. The PNA sequence was extended directly from the peptide TNT. The use of PNA has several advantages for this application: (i) the oligonucleotide and the peptide TNT are generated in a single synthesis on the same resin, which means only a single purification step is required after the synthesis is completed; (ii) PNA is approximately 15% lower in mass than a corresponding DNA sequence, which enhances mass spectrometric sensitivity; (iii) PNA has enhanced binding affinity for its target compared with a corresponding DNA sequence, so a shorter probe can be used to achieve a corresponding level of specificity; (iv) PNA can hybridize under low salt conditions, which is more favourable for ESI analysis of samples as ESI-MS is susceptible to signal suppression by high salt concentrations.
We have shown that these TNT-PNA probes hybridize quantitatively and specifically to their targets and that the probes perform reliably over a wide dynamic range providing a new platform for multiplexed and quantitative genomic analysis.
The ESI-cleavable TNT mass markers described in this article have several properties that make them very useful for quantitative, multiplex assays, including the following: (i) The TNT portion can be elaborated into very large arrays of tags using only small numbers of starting components. The 20 standard amino acids as well as the large number of isotopic variants of these amino acids that are available provide the possibility of synthesis of numerous marker molecules that are easily resolved by the unique combination of parent and daughter ion mass-to-charge ratios; (ii) The ESI-cleavability allows direct analysis of solution phase assays without complex workup of the samples unlike MALDI, which requires that samples are spotted onto targets; (iii) The ESI-cleavable linker connecting the DNA and the TNT components cleaves virtually instantaneously during electrospray ionisation that will allow separations such as capillary electrophoresis to be performed in-line with the MS/MS analysis; (iv) In-line separation allows for a further level of multiplexing over and above the large numbers of available tags since the probes can be identified by their elution time as well as by their tag and in future work, we will explore the use of this feature to enable in-line coupling of analytical separations such as capillary electrophoresis; (v) sets of isotopic TNTs can be synthesized that behave identically during separations, hybridizations and labelling reactions enabling accurate measurements of quantities of target nucleic acids without ‘dye effects’ widening the range of applications for which mass tags can be employed.
The use of TNT–PNA oligonucleotide conjugates offer many of the advantages of fluorescent detection such as high specificity, ease and safety of handling and high sensitivity with the additional unmatched advantages that result from being able to generate large numbers of tags with predefined masses and from being able to construct these sets of tags with stable isotopes generating chemically identical entities that will behave the same in labelling reactions and in separation steps. This means that multiplexed analyses with accurate quantification are now enabled in a user-friendly format. Future experiments will be directed towards evaluation of the TNT-PNA probes for post-PCR amplicon detection. In addition, the development of TNT–DNA oligonucleotide conjugates and evaluation of these probes as primers for multiplexed PCR amplification and subsequent detection of PCR amplicons will also be pursued. The ability to employ in-line capillary electrophoresis with immediate cleavage and detection of tags will be of particular interest as many genomics assays, such as restriction fragment length polymorphisms, satellite marker analysis and multiplexed PCR employ size separations and the ability to perform such analyses with the higher levels of multiplexing enabled by this technology will be of great advantage. |
Proper gene expression depends on precisely orchestrated interactions between nucleic acids and nucleic acid-binding proteins. In fact, virtually every step of gene expression involves the activities of both sequence-specific and general nucleic acid-binding proteins. For example, transcription of DNA to RNA employs helicases, transcription factors, RNA polymerase, 5′-capping enzymes, poly (A) polymerase and numerous other proteins required to process the nascent mRNA transcript (,). All of these factors must recognize nucleic acids in some manner. Then numerous mRNA-binding proteins are required to escort transcripts out of the nucleus and, ultimately, to coordinate protein translation (). Thus, gene expression relies heavily on nucleic acid-binding proteins, with major roles for proteins that bind to mRNA. In keeping with this workload, proteins have evolved a vast array of domains that mediate interactions with RNA, including but not limited to the hnRNP K homology (KH) domain, RNA recognition motifs (RRMs), arginine-rich motifs, and zinc-finger motifs ().
Many RNA-binding domains are abundant and highly conserved across species. In fact, Chen and Varani have estimated that about 1.5–2% of the human genome is composed of proteins that contain RRM domains, the most abundant and well-characterized of the RNA-binding domains (). RNA-binding motifs are typically recognized and defined on the basis of their conserved primary amino acid sequence, but the degree of sequence conservation among different motifs varies widely (). Although some major residues are defined functionally, others have been identified strictly on the basis of sequence conservation or homology (,). A key question is whether these conserved residues are actually critical for protein function, and therefore whether sequence conservation alone can be used as a measure of functional significance.
One remarkably versatile and highly conserved RNA-binding motif is the KH domain (). KH domains, which were originally identified as a repeated sequence in the hnRNP K protein (), are ∼70 amino acids in length, with a characteristic pattern of hydrophobic residues, a GXXG segment and a variable loop. KH domains are often found in multiple copies per protein. There is strong evidence that KH domains are critically important to the function of those proteins that contain them. For example, fragile X syndrome, a leading cause of inherited mental retardation, can arise from a missense mutation of Ile304Asn in the second KH domain of the FMRP protein (). Indeed, the Ile304Asn mutation is associated with a severe clinical phenotype. Thus, there is evidence that mutations in KH domains can cause human disease, which underscores the importance of defining what constitutes a functional KH domain.
Some proteins contain both conserved KH domains that include the GXXG motif, as well as what have been termed KH domains, in which the GXXG motif is interrupted or altered (). Although the structures of a number of conserved KH domains have been solved (), there has been little functional analysis of diverged KH domains. Recent evidence suggests possible synergistic binding to mRNA targets by the three KH domains of hnRNP K protein (), thereby enhancing interaction with the nucleic acid substrate. Chmiel and colleagues () drew a similar conclusion from their work with the PSI protein, which contains four KH domains. While these findings do shed light on the functional importance of KH domain in proteins, they fail to address the functional significance of KH domain sequence conservation or divergence.
As a first step towards addressing whether GXXG is essential for KH domain function, we explored the roles of conserved and diverged KH domains in Scp160p, a multiple KH domain protein in . Scp160p includes 14 KH domains (), only seven of which contain a strictly conserved GXXG motif (KH 2, 8–12 and 14). The other seven KH domains are diverged, with interruptions or alterations of the GXXG motif. Although the exact function of Scp160p and its orthologs, known as vigilins in higher eukaryotes, is not known, recent work suggests a role for this protein in modulating the metabolism of specific mRNA targets in the cytoplasm (,). Consistent with this postulated function, much of Scp160p is found in large protein complexes associated with soluble or membrane-bound polyribosomes (,). The goal of the present study was to explore two questions. First, are diverged KH domains essential for Scp160p function? Second, can diverged KH domains functionally replace conserved KH domains? To address these questions, we deleted and/or interchanged conserved and diverged KH domains of Scp160p, then expressed these variant alleles in yeast. We applied a combination of previously defined genetic () and biochemical (,) assays to the strains expressing these mutated alleles to discern the functional capacity of each encoded Scp160p protein.
Our results demonstrated that the answer to each of our two questions was yes. Both conserved and diverged KH domains essential for Scp160p function, and diverged KH domains capable of functioning in place of conserved KH domains.
All yeast manipulations were performed according to standard protocols. The strains used for this study are listed in Supplemental Table 1 and, unless otherwise noted, were derived by two-step gene replacement () at the locus from the haploid parent strain W303 [MATa , a gift from Dr. R. Rothstein, Columbia University, NY].
Unless otherwise noted, all recombinant DNA manipulations were performed according to standard procedures () using the XL10-Gold® (Stratagene) strain of . To facilitate site-directed mutagenesis of , a 2.2 kb PstI/EcoRI fragment containing sequence from the beginning of KH8 through approximately 430 bp downstream of the stop codon was gel purified and ligated into pGEM3-zf(+) using T4 DNA Ligase (Invitrogen) according to the manufacturer's instructions. The resulting plasmid (JF4566, pGEM3zf +. .Eco.Pst) was confirmed by restriction analysis and sequencing and used as a template for the generation of all subsequent modified alleles, as detailed below.
To create alleles individually deleted for KH13 and KH14, we used the QuikChange™ Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer's instructions, with one exception: we increased the number of PCR cycles from 18 to 28. Relevant primer sequences are listed in Supplemental Table 2. Each initial deletion was designed to remove the appropriate KH domain precisely, but also to leave in its place a short “scar sequence” that would facilitate subsequent replacement by an alternate KH sequence via gap repair homologous recombination. Primers scpdelKH14.stu.xho.f1 and scpdelKH14.stu.xho.r1 were used to replace KH14 with a StuI-linker-XhoI scar, resulting in the allele . Similarly, primers scpdelKH13.stu.xho.f2 and scpdelKH13.stu.xho.r2 were used to replace KH13 with a StuI-linker-XhoI scar, resulting in the allele . Finally, alleles intended to carry either a clean deletion or KH domain replacement of KH13 or KH14 were generated by gap repair of the corresponding scar construct with an appropriate fragment, as described below. The gapped backbone fragment for each recombination procedure was generated from the appropriate plasmid (pGEM3-zf(+).FLAG.Scp160Δ14.scar or pGEM3-zf(+).FLAG.Scp160Δ13.scar) by digestion with StuI and XhoI, followed by gel purification.
The inserts for scar replacement were generated as follows: (a) to remove the StuI-linker-XhoI scar from each deletion template, the indicated oligonucleotides in Supplemental Table 2 were annealed and extended in a single round of polymerization using the TripleMaster PCR System (Eppendorf) in a thermocycler, (b) to create , a fragment encoding KH11 was amplified from a wild-type template using primers scpdelKH14.addback.KH11.f1 and scpdelKH14.addback.KH11.r1, (c) to create , the replacement sequence was generated by PCR using the primers scpdelKH14.addback.13.f1 and scpdelKH14.addback.13.r1, (d) to create , the replacement sequence was generated by PCR using the primers scpdelKH13.addback.KH6.f1 and scpdelKH13.addback.KH6.r1, and finally (e) to create , the replacement sequence was generated by PCR using the primers SCP.KH14.addback.f1 and SCP.KH14.addback.r1. All DNA fragments were gel purified using the QIAquick® Gel Extraction Kit (Qiagen), according to the manufacturer's instructions.
To accomplish gap repair homologous recombination, the appropriate purified PCR fragments were individually cotransformed into competent with gel-purified, gapped plasmid backbone. Transformants were cultured and the plasmids isolated using the Qiaprep Spin® Miniprep Kit (Qiagen). All resultant alleles were verified by restriction analysis and DNA sequencing across the entire region that had been generated by PCR to ensure sequence integrity.
Once the desired sequence manipulations were completed and verified, we next moved the relevant sequences into yeast-integrating plasmids via gap repair homologous recombination for subsequent insertion into the locus of the yeast genome. Towards that end, all verified plasmids were digested with PstI and EcoRI. The ∼2.2 kb band from each construct was gel purified and individually cotransformed into competent with pJF2147 (YIplac211.FLAG.) that had been gapped by digestion with AflII and XbaI. Resultant constructs were confirmed by restriction digestion, linearized by digestion with PstI, and transformed into the yeast strain JFy4493. Appropriate integration was confirmed by genomic PCR, followed by direct sequence analysis of each locus from KH12 to 600 bp downstream of the stop codon.
We assessed the ability of all modified alleles listed in to function by quantifying the ability of each to complement synthetic lethality (). Briefly, cells carrying both an “maintenance” plasmid and an variant allele in a 2 “test” plasmid were inoculated into liquid medium lacking both uracil and leucine, to ensure selection for both plasmids. Once grown to saturation, 300, 000 cells were inoculated into synthetic media lacking only leucine, to enable loss of the maintenance plasmid. When this culture reached an OD of ∼2.0, again 300, 000 cells were inoculated into synthetic medium lacking leucine and grown to an OD of 1.5–2.0. Finally, cells were counted, and appropriate dilutions were plated to synthetic medium lacking leucine ±5-fluoroorotic acid (5-FOA). 5-FOA is toxic only to cells that contain a functional gene (); therefore, the number of colonies that grew on the 5-FOA medium relative to the number of colonies that grew on medium lacking 5-FOA was a measure of the proportion of cells in the culture that had lost the maintenance plasmid. This proportion, in turn, was a measure of the ability of the variant test plasmid to complement loss of the wild-type maintenance plasmid. Empty test plasmid backbone served as the negative control, and test plasmid encoding wild-type FLAG.Scp160p served as the positive control. For comparison between experiments, the proportion of cells that lost the 3 plasmid in each strain was normalized to the corresponding proportion from the positive control strain. Therefore, the final values calculated indicated how effectively, relative to wild-type , each modified allele of functioned . Since the maintenance plasmid was the same in all strains and all experiments, any issues of plasmid replication efficiency or distribution, independent of sequence, should have cancelled out.
All cell lyses and subcellular fractionations were performed as described previously (,,). Briefly, mid-log-phase cultures were incubated with cyclohexamide at a final concentration of 100 μg/ml for 1 min at 30°C, and then for 15 min on ice with swirling. Cells were washed in cold water, then lysed in cold buffer (), each containing 100 μg/ml cyclohexamide. All cell lysis procedures were accomplished using a multihead vortex mixer that was set to the highest speed at 4°C for 20 min. The supernatant from the lysis was centrifuged at 3000 rpm at 4°C for 5 min in a tabletop microfuge. The supernatant from this step was transferred to a fresh tube and centrifuged at 12000 rpm, 4°C, for 8 min in a tabletop microfuge. The supernatant from this final spin was the soluble lysate. The protein concentration of this soluble lysate was quantified using the Bio-Rad Protein Assay (cat #500-0006) according to the manufacturer's instructions. Next, 15 μg of soluble protein from untagged wild-type, FLAG-tagged wild-type, and FLAG-tagged Scp160p variants were each run on a gel for subsequent Western blot analysis. FLAG.Scp160p band intensity, for the wild-type protein and all variants, was quantified by scanning densitometry, and FLAG.Scp160p signal was normalized to the intensity of a previously reported endogenous protein also recognized by the anti-FLAG M2 antibody ().
For analyses of polysome association, 200 μl of soluble lysate were loaded onto a 15–45% sucrose gradient and centrifuged at 39,000 rpm in an ultracentrifuge for 2.5 h at 4°C, as described previously (). Fractions (1 ml) were collected from the sucrose gradients and 400 μl of each were concentrated for Western blot analysis. In brief, samples were mixed with 400 μl MeOH and 100 μl chloroform and agitated vigorously. Precipitated proteins were pelleted by centrifugation at 13 000 rpm for 2 min in a tabletop microfuge; the top phase was discarded, and 400 μl MeOH was added to the pellet. Each sample was inverted to mix and then re-centrifuged at 13 000 rpm for another 5 min. This supernatant was discarded, and samples were dried in a speed vac with no heat for ∼20 min until no remaining liquid was visible. The protein pellet was resuspended in 1× loading buffer and applied onto a 10% SDS-PAGE gel for subsequent Western blot analysis using the monoclonal ANTI-FLAG® M2 antibody (Sigma-Aldrich) to detect FLAG-tagged Scp160p, as described previously (). Controls involved preincubation of lysates with either 50 units/ml RNase ONE™ (Promega, Inc.) or 60 mM EDTA for 30 min at 4°C with gentle rocking.
For analysis of mRNP complex formation, soluble lysate from 1 l of cell culture was size-fractionated over an S300 Sephacryl column, as described previously (). Fractions of 2 ml were collected, and 12 μl of each were analyzed via Western blot using the monoclonal ANTI-FLAG® M2 antibody (Sigma-Aldrich) to detect FLAG-tagged Scp160p.
As a first step to delineate the distinct versus potentially overlapping functions of conserved and diverged KH domains in Scp160p, we generated alleles encoding two modified variants: FLAG.Scp160Δ14p, in which conserved KH14 was deleted, and FLAG.Scp160Δ13p, in which diverged KH13 was deleted (). While previous studies from our lab and others (,) had tested the function of Scp160p missing KH14 or KH13 + 14, the allele created here was novel. From the and templates, we then generated an additional four modified alleles: two in which the deleted KH domains were replaced by the same class of KH domain (i.e., conserved for conserved, or diverged for diverged), and two in which the deleted KH domains were replaced by the opposite class of KH domain. The encoded proteins were designated FLAG.Scp160Δ14.11p, in which conserved KH11 replaced conserved KH14, FLAG.Scp160Δ14.13p, in which diverged KH13 replaced conserved KH14, FLAG.Scp160Δ13.6p, in which diverged KH6 replaced diverged KH13, and FLAG.Scp160Δ13.14p, in which conserved KH14 replaced diverged KH13 (). As a result of these manipulations, two of the modified proteins, FLAG.Scp160Δ14p and FLAG.Scp160Δ13p, each carried a total of 13 KH domains, and four of the modified proteins, FLAG.Scp160Δ14.11p, FLAG.Scp160Δ14.13p, FLAG.Scp160Δ13.6p, and FLAG.Scp160Δ13.14p, each carried a total of 14 KH domains. Each modified allele was sequence confirmed and introduced into the appropriate yeast genomic locus in place of the wild-type 160 allele by two-step gene replacement, as described in the Materials and Methods section.
To confirm the expression of each variant Scp160p protein, soluble lysates from mid-log-phase cultures of the appropriate strains were subjected to Western blot analysis with the M2 anti-FLAG monoclonal antibody. Representative results are presented in A. Due to the different lengths of the KH domains deleted or replaced, the protein sizes varied from about 70 amino acids larger than the wild-type protein (FLAG.Scp160Δ14.13p), to about 140 amino acids smaller than the wild-type protein (FLAG.Scp160Δ13p). Yeast expressing untagged Scp160p served as a negative control for specificity of the antibody, and a ∼100 kDa endogenous yeast protein that cross-reacts with the M2 antibody (,,) served as a convenient internal control for loading (A, bottom-most band in each lane). As illustrated in A (open circles), each modified protein was expressed and displayed the expected migration pattern relative to the wild-type FLAG-tagged protein. Furthermore, quantitation of the Scp160p signal in each lane, relative to the corresponding loading control, for this and replicate experiments, demonstrated that each modified isoform was expressed at a level indistinguishable from that of the wild-type protein (B).
To test the impact of KH domain loss or substitution on Scp160p function, we first assessed the ability of each variant allele illustrated in to replace (). As a control, we analyzed the variants that lacked either conserved KH14 or diverged KH13 in parallel with the other variants. Previous studies from our lab () had demonstrated the functional significance of KH14, and previous studies from Seedorf and colleagues () had demonstrated the functional importance of KH13 + 14 combined. For genetic testing, each modified allele was subcloned into a plasmid and transformed into cells maintained by a plasmid. Although is not itself essential in yeast (), deletion of in combination with deletion of is synthetically lethal ().
To assess how well each modified allele functioned , we applied a previously validated quantitative complementation assay (). As explained in Materials and Methods, this assay measures the frequency with which a mutant strain loses a wild-type maintenance plasmid in the presence of a modified test plasmid. Results are calculated as the relative percentage of maintenance plasmid loss, and the calculated value gives an indication of the functional capacity of the tested allele. For example, a poorly complementing sequence or empty test backbone will have a very low percentage maintenance plasmid loss, because few if any cells will survive loss of the maintenance plasmid. In contrast, strains that carry an test sequence that functions well will have a very high degree of maintenance plasmid loss, because most if not all cells will survive loss of the maintenance plasmid. As explained earlier, because the maintenance plasmid was the same in all strains, any issues of plasmid replication efficiency or distribution, independent of sequence, should have cancelled out. Further, given that each of the wild-type and variant test sequences was expressed from the same wild-type promoter and plasmid backbone, and that all of these alleles were demonstrated to express comparable levels of Scp160p protein when genomic (), there should have been no disparities in Scp160p expression levels for this experiment.
Results of these analyses for the control strains confirmed that both KH domains 13 and 14 are essential for Scp160p function. As illustrated in , the allele (conserved KH domain deleted) enabled only ∼14% wild-type levels of maintenance plasmid loss, and the allele (diverged KH domain deleted) enabled only ∼10% wild-type levels of maintenance plasmid loss. While the allele provided an anticipated result (), this was the first demonstration that loss of KH13 alone also compromised Scp160p function. Together, these results provide a foundation for the remainder of the study, which involved testing various add-back alleles for restoration of function.
As illustrated in , both of the add-back alleles derived from , namely , in which a conserved KH11 replaced a conserved KH14, and , in which a diverged KH13 replaced a conserved KH14, demonstrated significant complementation. In fact, demonstrated the highest degree of complementation of any of the modified alleles tested here. In contrast, while the add-back allele, in which a conserved KH14 replaced a diverged KH13, demonstrated significant complementation compared to the ΔKH13 variant, the allele, in which a diverged KH6 replaced a diverged KH13, did not. Indeed, the allele performed even more poorly than did either deletion variant alone.
Scp160p exists as a component of large mRNA/protein complexes (mRNPs) in yeast (). To assess whether conserved and diverged KH domains make specific contributions to these interactions, we tested whether the Scp160p variants created here could form mRNPs. For these analyses, soluble lysates prepared from mid-log-phase cultures of yeast expressing each Scp160p variant were size-fractionated over an S300 Sephacryl column, as described in the Materials and Methods section. The resulting fractions were subjected to Western blot analysis using the M2 anti-FLAG antibody to reveal the migration pattern of the Scp160p (A) (,,). Lysates from yeast expressing full-length FLAG.Scp160p protein were analyzed in parallel as a positive control. To facilitate quantitative comparison among the elution profiles of the different Scp160p variants, the signal intensity in each lane was quantified by scanning densitomety, and the fraction of total signal detected for each protein in the first 5 lanes versus the last 15 lanes is presented in B. These studies were replicated at least three times to ensure reproducibility.
As illustrated in , wild-type FLAG.Scp160p migrated through the column predominantly with the void volume, demonstrating that most of the protein existed as part of a large complex (≥1.3 mDa). Previous studies have confirmed that these large complexes are mRNPs by virtue of their sensitivity to RNase digestion and the presence of both poly (A) binding protein Pab1p () and polyadenylated RNA (,,).
Deletion of either conserved KH14 or diverged KH13 significantly shifted the Scp160p signal from the void volume to fractions representing smaller complexes, or even monomeric Scp160p (), which comigrates with globular proteins of ∼450 kDa (). While both variant migration patterns were clearly different from that observed for wild-type Scp160p, FLAG.Scp160Δ13p migrated even more aberrantly than did FLAG.Scp160Δ14p (B), suggesting that loss of diverged KH13 was even more disruptive to macromolecular interactions than was loss of conserved KH14.
When the missing conserved KH14 sequence was replaced, either by conserved KH11 or by diverged KH13 (), close to half of each add-back protein migrated correctly as a large complex, although a ‘tail’ of smaller complexes extended beyond the normal limit in both profiles. Similarly, when the missing diverged KH13 sequence was replaced, either by diverged KH6 or by conserved KH14, the majority of signal representing each modified protein again migrated as expected (). Indeed, the FLAG.Scp160ΔKH13.14p protein showed an S300 elution profile that was almost indistinguishable from that of wild-type Scp160p (B). Thus, the conserved and diverged KH domains studied here appear to be functionally interchangeable to a large extent for protein complex formation.
Scp160p has been implicated in the cytoplasmic metabolism of specific transcripts (). Consistent with this presumed function, Scp160p associates with polyribosomes (,,,). To examine this more specific aspect of Scp160p macromolecular interaction in the context of KH domain contribution, we assessed the migration through a sucrose gradient of each Scp160p variant from . Sucrose gradients enable the size separation of large complexes, such as ribosomal subunits, monosomes, and polyribosomes. Western blot analyses of the fractions collected from these gradients revealed the distribution profile for each Scp160p variant protein relative to the migration pattern of the FLAG.Scp160p wild-type control. Finally, to test whether co-migration with polyribosomes reflected association of each variant Scp160p with polyribosomes, as opposed to non-specific aggregation, we pretreated relevant samples with RNase or EDTA, as described below. To facilitate comparison between the different profiles, the FLAG-Scp160p signal intensity in each lane was quantified by scanning densitometry and plotted as the fraction of total signal migrating in samples 1–4 (complexes smaller than monosomes) versus fractions 5–11 (monosomes and polysomes).
As expected (,,), a majority (>80%) of the wild-type FLAG.Scp160p protein migrated with monosomes and polyribosomes near the bottom of the gradient, while more than half of the FLAG.Scp160Δ14p variant protein migrated near the top of the gradient (A), indicating a loss of association with polyribosomes. Loss of diverged KH13 produced an even more striking shift of Scp160p signal from the bottom to the top of the gradient (A), reconfirming the functional significance of this domain. Of note, although neither wild-type Scp160p nor any of the ΔKH13 variant proteins showed any signal in the top-most gradient fraction, representing free proteins and small complexes, all three ΔKH14 proteins did show signal in this fraction. The simplest explanation for this disparity is that the wild-type and ΔKH13 Scp160p proteins assembled entirely into complexes larger than those found in the first gradient fraction, while the three ΔKH14 Scp160p proteins each maintained at least a subpopulation that did not.
Add-back of conserved KH11 to the ΔKH14 protein had little if any impact on polyribosome association, although add-back of diverged KH13 restored monosome or polyribosome comigration to more than half of the molecules (A). A similar ‘partial rescue’ was seen following add-back of diverged KH6 to the ΔKH13 protein, such that more than half of the FLAG.Scp160Δ13.6p protein migrated with the denser gradient fractions. Finally, add-back of conserved KH14 in place of the missing diverged KH13 completely normalized the sucrose gradient profile of the variant protein (A).
To confirm that the pattern of Scp160p migration evident in these experiments corresponded to association with ribosomes, samples were either treated with EDTA, which dissociates polyribosomes and monosomes into small and large ribosomal subunits, or RNase, which dissociates polyribosomes into monosomes. Both treatments caused the signals for all Scp160p variants tested to migrate predominantly at the top of the gradient (B andC), confirming that the pattern in the absence of pre-treatment was due largely to association with polyribosomes, and not simply to aggregation. For all three variant proteins, however, especially Scp160Δ13.6p, a portion of the signal did remain in the denser fractions even after treatment with EDTA or RNase (B and C), indicating that these proteins may have aggregated to some extent. By extension, these results suggest that while the data presented in reflect predominantly mRNP formation by the variant Scp160p proteins, some degree of aggregation may also have occurred.
xref
#text |
xref
#text
A 75-year-old man with history of celiac disease presented with pallor, fatigue, and 20-pound weight loss of three weeks duration. CD was diagnosed two years prior, at which point the patient underwent extensive evaluation for unexplained microcytic hypochromic anemia and was found to have an elevated endomysial antibody titer. A small bowel biopsy showed celiac disease. The patient was then placed on a gluten-free diet and the anemia resolved completely.
On physical examination, a vague non-tender mass in the right hypochondrium was found and his stools tested positive for occult blood. The laboratory values were within normal range, except for hemoglobin (11mg/dL), MCV 75, SGOT (61 IU/L) and LDH (5000 IU/L).
CT scan of abdomen showed extensive carcinomatosis. There were also small pleural effusions and scattered lymphadenopathy. PET showed diffusely hyper-metabolic foci coinciding with CT findings. Colonoscopy revealed a friable nodular mass in the hepatic flexure and proximal transverse colon. However, histopathologically the mass consisted of a high-grade B-cell lymphoma. Flow cytometry following immunostaining reported positive CD10, CD19, CD20, CD45, CD79a, and Ki-67. FISH assay demonstrated t (14:18) translocation and bcl-2 rearrangement. The bone marrow biopsy showed evidence of disease []. The bone marrow biopsy was also positive for CD10, CD19, CD20, CD38, CD45 and HLA-DR.
The patient was treated with rituximab cyclophosphamide, Adriamycin, vincristine, and prednisone (CHOP-R) according to protocol, with intrathecal methotrexate prophylaxis. The patient is currently in remission.
This is one of the very few documented cases of an aggressive colonic Burkitt-like lymphoma in a patient with a history of celiac disease. CD is an inflammatory condition of the small intestine in which there is gluten-mediated damage to small intestinal villi. It is a common disorder affecting as much as 1 percent of the population [,]. The association between CD and intestinal lymphoma was first described by Fairley and Mackie in 1937 []. A recent large multi-center retrospective study involving over 9,800 patients in Europe indicates that celiac disease patients have an increased risk of NHL (Odds ratio 2.6, 95 percent CI 1.4-4.9) [].
According to Catassi et al., enteropathy-associated T-cell lymphoma is the most common with an incidence rate of 0.5- to 1 million, constituting 35 percent of all small bowel lymphomas, followed by B-cell and adenocarcinomas []. Koo reported a case of operable colonic lymphoma confined to the transverse colon []. We report this rare case in which a patient with a history of CD developed colonic BLL with extensive peritoneal involvement. There are three interesting and unique features: While CD is usually a disease more common in the fourth and fifth decades of life, our case was diagnosed in his eighth decade. Secondly, his lymphoma was diagnosed two years after the diagnosis of celiac disease, in contrast to literature, if celiac disease is diagnosed first, lymphoma occurs within five to 10 years with a probable time lag of up to 60 years []. Lastly, the median age for incidence of BLL is 55 years, while our patient was in his eighth decade of life.
Several oncogenic mechanisms might predispose the a digestive tract afflicted with CD to malignancy: increased permeability of environmental carcinogens like Epstein- Barr virus, chronic inflammation, release of pro-inflammatory cytokines, defective immune surveillance (post-kidney transplant and Human Immunodeficiency Virus patients) and nutritional deficiencies caused by the disease or the gluten-free diet []. Increased ribosomal DNA transcriptional activity in mucosal cells from patients diagnosed with CD has been demonstrated []. All the above-mentioned mechanisms culminate in proneoplastic mutations in B-cell populations such as deletion of 9p21, 17p23(P53) and gain of 1q chromosome. Howell and associates have previously demonstrated that apart from certain HLA genotypes (DQA1*0501 and DQB1*0201) predisposing to CD and EATCL, additional HLA-DR/DQ alleles might independently operate to cause lymphomas []. Likewise, further molecular typing studies are warranted to point out the nature of B-cell populations in the gut in patients who develop BLL. This can help to identify high-risk individuals among patient populations with diagnosed CD to aid in earlier therapeutic intervention and better prognosis.
i
s
i
s
t
h
e
f
i
r
s
t
c
a
s
e
o
f
a
g
g
r
e
s
s
i
v
e
B
L
L
o
c
c
u
r
r
i
n
g
i
n
a
p
a
t
i
e
n
t
w
i
t
h
c
e
l
i
a
c
d
i
s
e
a
s
e
i
n
h
i
s
e
i
g
h
t
h
d
e
c
a
d
e
o
f
l
i
f
e
.
I
t
i
s
p
o
s
s
i
b
l
e
t
h
a
t
c
h
r
o
n
i
c
i
n
f
l
a
m
m
a
t
i
o
n
f
r
o
m
C
D
p
r
e
d
i
s
p
o
s
e
s
i
n
d
i
v
i
d
u
a
l
s
t
o
t
h
e
d
e
v
e
l
o
p
m
e
n
t
o
f
b
o
t
h
h
i
g
h
-
a
n
d
l
o
w
-
g
r
a
d
e
B
-
c
e
l
l
l
y
m
p
h
o
m
a
s
,
a
l
t
h
o
u
g
h
t
h
i
s
h
y
p
o
t
h
e
s
i
s
i
s
y
e
t
t
o
b
e
p
r
o
v
e
n
.
P
r
o
g
n
o
s
i
s
o
f
t
h
e
s
e
p
a
t
i
e
n
t
s
c
o
u
l
d
b
e
i
m
p
r
o
v
e
d
b
y
i
d
e
n
t
i
f
y
i
n
g
c
e
r
t
a
i
n
u
n
i
q
u
e
H
L
A
g
e
n
o
t
y
p
e
s
a
s
s
o
c
i
a
t
e
d
w
i
t
h
d
e
v
e
l
o
p
m
e
n
t
o
f
h
i
g
h
g
r
a
d
e
B
-
c
e
l
l
l
y
m
p
h
o
m
a
s
b
y
c
l
o
s
e
r
m
o
n
i
t
o
r
i
n
g
a
n
d
e
a
r
l
i
e
r
i
n
t
e
r
v
e
n
t
i
o
n
. |
Despite the complexity of cancer as a disease, there are some common features shared by all cancer cells that distinguish them from their normal counterparts. They are immortal and have a high proliferative potential. They no longer require external growth signals to activate the proliferation program and are insensitive to growth inhibitory signals. They have developed mechanisms for evading senescence and programmed cell death, or apoptosis. In order to ensure tumor growth, the cells of solid tumors must promote the formation of new blood vessels for oxygen and nutrient supply. And, finally, malignant cells fail to exhibit normal cell-cell interactions, which results in their ability to invade and to metastasize and grow in an abnormal cellular environment []. The acquisition of these qualities is not a single event but rather the result of an accumulation of multiple genetic alterations leading to escape from normal control of growth regulation and differentiation [].
Genes with very diverse cellular functions have been implicated in carcinogenesis. This includes genes involved in intracellular signaling cascades, cell cycle control, cell differentiation, apoptosis, protein synthesis and degradation, organization of cytoskeleton, cell migration, cell-cell interaction, regulation of transcription, chromatin remodeling, and DNA damage sensing and repair [-]. Understanding how the disruption of these cellular processes leads to tumor formation is the key to rational development of effective cancer therapies.
Because the development of cancer is a multi-step process, it may take years between the occurrence of the first genetic lesion and the actual formation of malignant neoplasm. Thus, the incidence of cancer increases with age. In rare cases, however, individuals inherit a mutant allele of a cancer predisposition gene, which increases the mutation rate and promotes the development of cancer. Thus, individuals with cancer predisposition syndromes have both a greater risk of developing cancer and an average earlier age of onset compared to the general population.
In 1971, Knudson proposed his famous “two-hit” hypothesis []. By observing patients with familial and sporadic retinoblastoma, he suggested that individuals with familial cases inherited the first genetic lesion, or “hit,” and that they only needed to acquire one more hit to develop cancer. These patients developed retinoblastoma at an earlier age than individuals with sporadic tumors who needed to acquire two somatic hits in the same cell. This hypothesis was confirmed with the discovery of the retinoblastoma (RB1) gene in 1986 []. Indeed, individuals with familial retinoblastoma inherited a germline mutation in one allele of the RB1 gene with the remaining wild type allele being lost in tumors, often through a process involving large deletions with loss of heterozygosity (LOH) for surrounding polymorphic loci [].
Since the discovery of RB1, multiple tumor-predisposition genes have been identified by the study of familial cancer syndromes. It is estimated that 5 to 10 percent of all cancers develop in individuals who have inherited a mutation conferring cancer susceptibility []. Analysis of the genes responsible for hereditary cancer predisposition and the pathways in which they act has been critically important in providing insight into cancer formation, because the same genes that are affected in cancer-predisposition syndromes are often responsible for the majority of somatic cancer cases as well. For example, germline mutations in the APC gene account for less than one percent of all colorectal cancers, but APC is also inactivated in nearly all sporadic colon cancers [].
More than 50 cancer predisposition syndromes have been identified to date [,] with both dominant and recessive modes of inheritance (Table 1). Activating mutations in oncogenes occur only in a few cases, such as RET in multiple endocrine neoplasia type 2 (MEN2) [] and MET in familial papillary renal carcinoma []. Activated oncogenes function dominantly in the cell, and mutation of the second allele is not required. Most of the dominantly inherited cancer syndromes, however, have an inactivating mutation in a tumor suppressor, and disease formation follows the two-hit rule. The first hit is a germline mutation that exists in every cell of the body. Somatic inactivation of the second allele of the same locus in certain tissues may initiate neoplastic transformation. Thus, cancer predisposition syndromes often have dominant mode of inheritance at the level of the organism, but at the cellular level they are recessive because inactivation of both copies of the gene is required for cancer formation.
Genes whose inactivation leads to tumorigenesis are divided into two distinct groups: gatekeepers and caretakers. Gatekeepers are usually “classical” tumor suppressors that directly regulate cell growth, differentiation, and death in certain tissues. Dysregulation of a gatekeeper is necessary to initiate neoplastic transformation. Examples of gatekeepers include APC for colon epithelia, RB1 for retinal epithelia, NF1 for Schwann cells, and VHL for kidney cells [,]. Inactivation of both homologs of these genes is found both in hereditary forms of cancer and in the majority of sporadic cancers of the same type. In contrast to gatekeepers, caretaker genes do not directly regulate cell growth. Caretaker genes are involved in DNA repair and maintenance of genomic integrity. Inactivation of a caretaker does not directly promote tumor formation but facilitates the development of mutations in gatekeeper genes and other cancer-related genes. Caretakers, like gatekeepers, may be tissue-specific, but mutation in a caretaker is neither necessary nor sufficient for the development of cancer. Mutations in some caretaker genes cause human cancer by the two-hit mechanism. For example, patients with familial breast cancer syndromes inherit one mutant copy of BRCA1 or 2, and LOH for the second allele is found in tumors of breast and ovary [-]. Similarly, in hereditary non-polyposis colorectal cancer (HNPCC), one mutant copy of mismatch repair genes such as MLH1, MSH2, or MSH6 is inherited, and the wild type allele is lost in tumors of the colon []. However, unlike gatekeepers, mutations in caretakers are uncommon in sporadic tumors of the types seen in the hereditary disease [].
A number of human cancer predisposition syndromes have a recessive mode of inheritance (Table 1). These syndromes are also called genetic instability syndromes, because one of the hallmarks of these diseases is chromosomal instability and hypersensitivity to various DNA damaging insults. The genes mutated in these diseases, such as ATM, Fanconi anemia complementation group, NBS1, Xeroderma pigmentosum-associated genes (XPA, XPC, etc.), BLM, and RECQL4, function in DNA damage sensing or repair [,,] and can be classified as caretakers by analogy with the BRCA and HNPCC genes.
Since the identification of the first cancer predisposition gene (RB1) in 1986, tremendous advances have been made in understanding the function of multiple genes implicated in both sporadic and hereditary cancers. However, the mechanism of action of many more recently identified genes is not well understood. Further studies are necessary to provide new insights into human neoplasia and improve the diagnosis and treatment of both sporadic and hereditary cancers.
Multiple endocrine neoplasia type 1 (MEN1) is an autosomal dominant cancer predisposition syndrome characterized by parathyroid, pancreatic islet, and pituitary tumors as well as other neoplasms such as gastrinomas, carcinoids, lipomas, angiofibromas, angiomyolipomas and thyroid tumors []. Clinical manifestations depend on the affected organ. Most MEN1-associated tumors are not metastatic but create clinical effects by overproduction of hormones such as insulin, glucagon, prolactin, parathyroid hormone, gastrin, growth hormone, or adrenocorticotropic hormone [,]. The most common manifestation of the disease is parathyroid adenoma, with penetrance reaching almost 100 percent by the age of 50 []. Major causes of morbidity and mortality in MEN1 are gastrinomas and foregut carcinoids, which are often highly metastatic []. The estimated prevalence of MEN1 is one in 30,000 [].
was localized to chromosome 11q13 by linkage analysis, and the gene was later identified by positional cloning [,]. Tumors in MEN1 arise through the two-hit mechanism with germline mutation in one copy of the gene and somatic loss of the normal allele []. Germline mutations in are mainly nonsense or frameshift and are presumed to cause inactivation of the gene. The transcript of the gene contains an open reading frame predicted to encode 610 amino acid proteins. Analysis of the sequence of the protein menin showed two nuclear localization signals near the carboxyl terminus but has not provided any other clues to its function. Immunocytochemical studies of interphase somatic cells show that the protein is located in the nucleus []. Studies in human and mouse show that is expressed widely in adult and fetal tissues with high levels in the CNS, lung, liver, kidney, testis, and fetal thymus [,-].
Among the earliest studies of menin function was a yeast two-hybrid analysis showing interaction between menin and JunD, a component of the AP-1 transcription factor []. Studies in mammalian tissue culture systems showed that the N-terminal half of menin could inhibit JunD-mediated transcription. Menin did not bind to other AP-1 components such as c-Jun, JunB, or four members of the Fos family. Further investigation of menin’s involvement in JNK and other MAP kinase signaling cascades revealed that overexpression of in some cell lines inhibited ERK-mediated phosphorylation of JunD and Elk1 and JNK-dependent phosphorylation of JunD and c-Jun []. In addition, overexpression of menin in CHO-IR cells suppressed insulin-induced activation of AP-1 and transcriptional induction of c-Fos []. Similarly, menin was shown to bind to and inhibit transcriptional activation by several members of the NF-κB family of transcription factors []. Binding of menin to SMAD3 appears to enhance transcriptional activation function, and loss of menin blocks TGF beta signaling []. The homeobox-containing protein Pem [] has also been shown to bind menin in biochemical assays. These data suggest that menin may act as a transcriptional regulator. Our investigations of the homolog showed interaction of with JNK signaling pathway components. We also found that when tethered to DNA, menin acts as a transcriptional repressor. Menin does not have any recognizable DNA binding motifs, and data are conflicting as to whether menin can directly bind DNA and whether it binds in a sequence-specific manner [,].
Multiple lines of evidence exist to support menin’s involvement in transcriptional control and chromatin modification. Human menin was found to bind and control the activity of multiple transcription factors. Menin interacts with mSin3a, a component of the Sin3/Rpd3 HDAC complex [], and with a histone methyltransferase complex containing the trithorax family proteins MLL2 and Ash2L [-]. In addition, menin was found to bind to the promoter regions of several homeobox genes and to regulate their expression [,]. Genome-wide analysis of genomic binding sites showed menin’s localization to the promoters of multiple genes [,]. Recently, two other transcription targets of menin were identified. Cyclin-dependent kinase inhibitors p27 and p18 are downregulated in menin-deficient cells, which also exhibited increased proliferation []. Thus, menin may control cell proliferation through regulating transcription of genes involved in cell cycle control.
That tumors in MEN1 arise through two-hit mechanism suggests that menin may act in the regulation of cell growth or in the maintenance of genomic integrity, as discussed above. Several lines of evidence suggest that menin might participate in DNA damage sensing or repair. Peripheral blood lymphocytes from MEN1 patients have an elevated frequency of chromosomal abnormalities after exposure to the cross-linking agent diepoxybutane (DEB) []. Menin-deficient mouse embryonic fibroblasts (MEFs) are hypersensitive to DEB []. Menin interacts with FANCD2, a gene that underlies the genetic instability syndrome Fanconi anemia. Moreover, this interaction is strengthened after the exposure of cells to γ-irradiation []. In addition, menin was shown to interact with replication protein A (RPA) [], a protein that binds single strand DNA and is important for several DNA repair pathways []. Studies from our laboratory show a role for the homolog of in maintenance of genomic integrity. We found that inactivation of in results in flies that are sensitive to certain types of DNA damage, in particular to ionizing radiation and DNA cross-linking agents. In addition, -deficient flies have an elevated mutation rate both at baseline and after treatment with ionizing radiation and nitrogen mustard (a cross-linking agent) []. Menin-deficient tissues from both and mammals fail to undergo S-phase arrest in response to DNA damage by ionizing radiation []. These data suggest that tumor pathogenesis in the human disease may relate to genomic instability.
Studies also support a direct role for in the control of cell growth and differentiation. Overexpression of menin represses the proliferation and tumorigenesis of Ras-transformed NIH3T3 cells and insulinoma cells []. Data are conflicting as to whether menin-deficient mouse cells have increased proliferative capacity. One study found no cell-autonomous proliferative defects in embryonic stem (ES) cells or embryonic fibroblasts (MEF) []. However, another study showed that MEFs have increased proliferative capacity, and reintroduction of wild type menin suppresses their growth, possibly through an interaction with activator of S-phase kinase (ASK), a component of the Cdc7/ASK kinase complex []. Some phenotypes seen in mouse models of MEN1 suggest a direct role in growth control. For example, tissue-specific gene inactivation in pancreatic islet cells of the mouse leads to widespread hyperplasia, consistent with a similar role for the gene postnatally [].
A mouse model with targeted disruption of the gene showed that menin has an essential role in development []. The phenotype of heterozygotes was very similar to the human MEN1 phenotype with tumors of the parathyroid, pituitary, pancreatic islets, thyroid, and adrenal cortex. Homozygous mice died in utero. They appeared normal at E9.5 but by E11.5-12.5 were developmentally delayed and smaller than normal littermates. Virtually all fetuses were resorbed by E14.5. Morphological defects included an open neural tube, hypoplastic myocardium, and an extremely malformed liver with abnormal organization of the epithelial and hematopoietic compartments and increased apoptosis []. Conditional knockout of in mouse pituitary and pancreatic islets showed that both these organs were able to develop normally even if was deleted in early embryogenesis. Pancreatic islets developed β-cell hyperplasia and later, in six- to 12-month-old mice, pancreatic islet adenomas. Insulinomas and prolactinomas developed in some animals starting at nine months. The late tumorigenesis in these mice suggests that additional somatic events are necessary to promote cancer formation [,].
Two studies suggested a function of menin related to telomeres. Investigations in our laboratory [] showed that menin localizes to telomeres during meiotic prophase in mouse spermatocytes. Menin foci were seen only at this location and not in other regions of meiotic cells. The foci were apparent as soon as axial elements began to form, and signals increased in intensity throughout prophase. However, menin did not co-localize with telomeric proteins, such as TRF2, in somatic cells at any stage of the cell cycle. Furthermore, telomeres of MEN1-related tumors were no different in length than telomeres of histologically similar sporadic tumors arising without mutations. Overexpression of menin did not influence telomerase activity as assessed by TRAP assays. Another study, which systematically examined the effects of tumor suppressor genes on telomere maintenance, suggested that loss of menin function might stimulate telomerase (hTERT) expression [] and that menin might be involved in tumorigenesis through this mechanism.
Menin has also been found to associate with several cytoskeleton components. Menin’s interaction with the intermediate filament proteins GFAP and vimentin was suggested to control its nuclear import []. Another cytoskeletal interactor of menin is nonmuscle myosin II-A heavy chain protein, which was found to co-localize with menin at the cleavage furrow during cytokinesis [].
In the 10 years since the gene was first cloned, a wealth of information concerning its possible function in tumorigenesis has been obtained. The accumulated genetic and biochemical data show that menin has multiple binding partners in a cell. By virtue of interacting with several transcription factors, menin is proposed to be involved in signaling pathways such as JNK, MAPK, TGF beta, and NF-κB; all of which have a well-documented role in cancer formation [-]. At the same time, through its role in chromatin modification, menin has been shown to affect the expression of c-Fos, hTERT, p27, and p18, which are also linked to neoplastic transformation [,,,]. On the cellular level, menin has been shown to regulate cell cycle progression and genomic integrity, both of which are often associated with a cancer phenotype.
The multiple functions of are not altogether unconnected. As a regulator of transcription, menin could control the expression of multiple factors involved in cell cycle regulation and the DNA damage response (see the examples above). The transcription regulation is likely achieved through menin’s function in chromatin modification. At the same time, there is a wealth of evidence connecting the modification of chromatin with the maintenance of genome integrity []. Finally, the possibilities of MEN1’s involvement in cell cycle regulation and genome maintenance are not mutually exclusive. MEN1 was shown to regulate proper S-phase arrest in response to exogenous DNA damage []. Similar regulation is necessary in the cell as endogenous DNA damage occurs every cell cycle during DNA replication [].
Taken together, the accumulated knowledge of potential functions and cellular partners suggests several possible mechanisms of its involvement in cancer formation. The key question is which of these functions or interactions play critical role in cancer development. One possible way to address this challenging problem is systematic investigation of tumors obtained from Men1 heterozygous mice as well as from MEN1 patients. One such study generated p18(-/-)Men1(+/-) and p27(-/-)Men1(+/-) mice and found that p18 functionally collaborates with MEN1 in suppressing lung and neuroendocrine tumors [,]. Another important question is whether menin aids in the maintenance of genome integrity by regulating the cell cycle or by serving as a structural component of DNA repair machinery. To this end, analysis of mutational spectra could point to a DNA repair step(s) most deficient in cells lacking . Careful examination of interactions between menin and known DNA repair proteins is also necessary. Addressing these issues is of key importance for the understanding of -related carcinogenesis and for the development of effective therapeutic approaches. |
Cancer results from the outgrowth of a clonal population of cells from tissue. The development of cancer, referred to as carcinogenesis, can be modeled and characterized in a number of ways. One way to describe this process is to illustrate the essential features of both cancer cells and tumors: the “hallmarks” of cancer []. Cancer development requires the acquisition of six fundamental properties: self-sufficient proliferation, insensitivity to anti-proliferative signals, evasion of apoptosis, unlimited replicative potential, the maintenance of vascularization, and, for malignancy, tissue invasion and metastasis []. Cancer can also be considered with regard to a step-wise development functionally grouped into three phases: initiation, promotion, and progression []. Initiation is characterized by genomic changes within the “cancer cell,” such as point mutations, gene deletion and amplification, and chromosomal rearrangements leading to irreversible cellular changes. Tumor development is promoted by the survival and clonal expansion of these “initiated” cells. Progression encompasses a substantial growth in tumor size and either growth-related or mutually exclusive metastasis.
Essential to the development of cancer is the accumulation of genetic lesions in cells. Such events are obviously required for initiation but may also be involved in the promotion or progression of tumor development []. These genome-level events include the activation of cellular proto-oncogenes or inactivation of tumor suppressor genes, which act in a “cancer-cell intrinsic” manner bestowing these cells with certain properties. However, while these cell autonomous properties are necessary for tumorigenesis, they are not sufficient. Research over the last two decades has solidified the concept that tumor development and malignancy is the result of processes involving both the cancer cells themselves and non-cancer cells, many of which compose the heterocellular tumor. A clear example of this is illustrated by the requirement of neo-angiogenesis for tumor growth and thus the contribution of the vascular endothelial cells [].
An association between the development of cancer and inflammation has long-been appreciated [,]. The inflammatory response orchestrates host defenses to microbial infection and mediates tissue repair and regeneration, which may occur due to infectious or non-infectious tissue damage. Epidemiological evidence points to a connection between inflammation and a predisposition for the development of cancer, i.e. long-term inflammation leads to the development of dysplasia. Epidemiologic studies estimate that nearly 15 percent of the worldwide cancer incidence is associated with microbial infection []. Chronic infection in immunocompetent hosts such as human papilloma virus or hepatitis B and C virus infection leads to cervical and hepatocellular carcinoma, respectively. In other cases, microbes may cause cancer due to opportunistic infection such as in Kaposi’s sarcoma (a result of human herpes virus (HHV)-8 infection) or inappropriate immune responses to microbes in certain individuals, which may occur in gastric cancer secondary to colonization or colon cancer because of long-standing inflammatory bowel disease precipitated by the intestinal microflora [,]. In many other cases, conditions associated with chronic irritation and subsequent inflammation predispose to cancer, such as the long-term exposure to cigarette smoke, asbestos, and silica [,].
Observing signs of inflammation, such as leukocyte infiltration, at tumors infected with microbes or sites of chronic irritation is expected. However, as first observed by Virchow in the middle of the 19th century, many tumors for which infection or irritation are not necessarily a predisposing factor, such as mammory adenocarcinoma, show a “lymphoreticular infiltrate.” Many tumors of this type contain activated fibroblasts and macrophages, in addition to a gene expression profile with an inflammatory signature. Quantitative aspects of wound repair or inflammatory gene expression often correlate negatively with cancer stage and prognosis [-]. Further evidence for the role of inflammation has come from the use of non-steroidal anti-inflammatory drugs (NSAIDs) in the prevention of spontaneous tumor formation in people with familial adenomatous polyposis (FAP) []. Thus, cancer and inflammation are related by epidemiology, histopathology, inflammatory profiles, and the efficacy of anti-inflammatory drugs in prophylaxis. These observations have provided impetus for investigation and hypothesis on the mechanisms and semantics of the relationship between cancer and inflammation.
There is evidence to suggest the inflammatory and immune systems may inhibit the development of cancer. This may occur by two cancer-associated recognition events. In tumor immunosurveillance, the host may have a dedicated mechanism to perceive and eliminate transformed cells. Adaptive immune recognition of tumor-associated and specific antigens also may be an important means by which the immune system controls the development of cancer []. Such topics will not be covered here. However, it seems the net effect of the inflammatory system is to positively affect tumor development. The relationship between cancer and inflammation is not simple and cannot be reduced to one grand theory. Previous reviews have focused on various aspects of the relationship between cancer and inflammation, such as the role of various inflammatory cells [,,], mediators [,-], or signaling pathways [] in cancer. The focus of this essay will be to discuss the relationship between cancer and inflammation along the organizing principle that the inflammatory response maintains physiological processes such as tissue homeostasis and repair after injury and that in this role, inflammation may be an ancillary, or perhaps inseparable, aspect of tumor development.
As stated, long-standing inflammation secondary to chronic infection or irritation predisposes to cancer. How might such conditions lead to the emergence of initiated cells? In the case of some types of viral infection, it is clear that virally encoded genes can contribute to cellular transformation. An example of this are the transformative abilities of high risk HPV mediated by the oncoproteins E6 and E7 []. However, many microbes associated with cancer cannot transform cells. For example, while certain strains of contain factors that affect host cell signaling, they are not classical oncogenes [].
The chronic inflammatory states associated with infection and irritation may lead to environments that foster genomic lesions and tumor initiation. One effector mechanism by which the host fights microbial infection is the production of free radicals such as reactive oxygen intermediated (ROI), hydroxyl radical (OH•) and superoxide (O₂-•) and reactive nitrogen intermediates (RNI), nitric oxide (NO•) and peroxynitrite (ONOO-). Primarily thought to be anti-microbial, these molecules form due to the activities of host enzymes such as myeloperoxidase, NADPH oxidase, and nitric oxide, which are regulated by inflammatory signaling pathways. Importantly, ROI and RNI lead to oxidative damage and nitration of DNA bases which increases the risk of DNA mutations [].
Cells have intrinsic mechanisms by which to prevent unregulated proliferation or the accumulation of DNA mutations. These include tumor suppressor pathways that mediate DNA repair, cell cycle arrest, apoptosis and senescence. In the face of DNA damage or oncogenic activation, cells will either repair their DNA and prevent mutations or initiated cells will undergo cell death.
In the face of massive cell death as occurs in infection or non-infectious tissue injury, lost cells must be repopulated by the expansion of other cells, often undifferentiated precursor cells such as tissue stem cells. There are two requisites for this: Some cells must survive the injury, and cells must expand to maintain cell numbers for a proper functioning tissue. Many inflammatory pathways function to mediate these two prerequisites of tissue repair [,]. In an extension of its physiologic role in mediating tissue repair or as a strategy in host defense to infection, the inflammatory response may play a role in providing survival and proliferative signals to initiated cells, thereby leading to tumor promotion.
Direct evidence for a link between tumorigenesis and either host defense and tissue repair has come from a number of observations. Many molecules and pathways double-up, playing roles in homeostasis, tissue repair, and tumorigenesis. The Wnt/β-catenin pathway plays a critical role in both the maintenance of the steady-state proliferative compartment and tumorigenesis of tissues []. Molecules such as COX-1 and -2, involved in the synthesis of prostaglandins that mediate the tissue repair process in the alimentary tract [-], play critical roles in tumor development at these sites [,]. For the most part, these observations are only associations, based on an entity’s (enzyme or transcription factor) involvement in tissue repair and tumorigenesis. However, key supportive evidence to connect these processes has come from studies showing that dedicated tissue injury and wounding supports tumor growth and neoplastic progression. Injection of Rous sarcoma virus (RSV) into chickens leads to the growth of a sarcoma at the site of injection. Work over the years by the Bissell laboratory has shown that sarcomas may form at other sites of the chicken if that site is wounded []. The promotion of these wound-related tumors can be inhibited by glucocorticoids and may be mediated by the actions of transforming growth factor-β (TGF-β) and fibroblast growth factors (FGFs) [-]. In a B16 melanoma adoptive transfer study, tumor growth is enhanced in wounded limbs by the induction of paracrine factors such as TGF-β and bFGF in wound fluid [].
Recent studies investigating the role of NF-κB (a family of transcription factors central to the induction of inflammation) in tumorigenesis has provided some more detailed insights into the role of inflammation in tumor promotion. Greten et al. used a model of colitis associated cancer (CAC) induced by the intraperitoneal injection of the carcinogen azoxymethane (AOM), followed by multiple rounds of inflammation and leukocyte infiltration caused by administration of the colonic epithelial cell toxin, dextran sulfate sodium (DSS) []. In this system, it is clear that chronic inflammation augments tumorigenesis, as when one dose of AOM is given without DSS cycling, no tumors arise. How then does the inflammatory response affect tumorigenesis in this model? Inactivation of the classical NF-κB pathway in colonic epithelial cells by conditional deletion of the IκB kinase β (IKKβ) protein resulted in a substantial decrease in the frequency of visible tumors []. Importantly, NF-κB signaling in epithelial cells was required for inhibition of apoptosis shortly after administration of one round of AOM and DSS, perhaps by the induction of anti-apoptotic factors such as Bcl-X. Thus upon intestinal epithelial injury and the addition of a mutagen, NF-κB provides a survival signal to initiated cells. Importantly, IKKβ in the colonic epithelium is responsible for mediating epithelial cell survival in protection from both infectious and non-infectious injury [,,] and host defense pathways in intestinal epithelium []. A similar role for NF-κB in survival of initiated cells was demonstrated in a chronic inflammation model of hepatocellular carcinoma, which develops spontaneously in Mdr2-deficient mice []. In this model, when NF-κB activation was inhibited by use of a super-repressor of degradation of IκB selectively expressed in hepatic epithelial cells, there was an increased number of apoptotic hepatocytes, a finding which correlated with a decreased frequency of tumors compared to Mdr2-/- mice with degradable IκB [].
Tumor promotion requires not only the survival of initiated cells, but also their expansion. Many inflammatory mediators such as cytokines, chemokines, and eicanosoids are capable of stimulating the proliferation of both untransformed and tumor cell proliferation []. Mice deficient in TNF have fewer skin tumors upon administration of the phorbol ester TPA and the mutagen DMBA []. Investigation into how TNF regulates tumor progression in this model suggested that this inflammatory mediator acts as a tumor promotor, since soon after application of TPA/DMBA, the characteristic hyperproliferation of keratinocytes was shown to be dependent on TNF []. What induces this production of inflammatory mediators, such as TNF, which leads to expansion of tumor initiated cells? NF-κB activation in myeloid cells recently was shown to play a critical role in the production of inflammatory mediator in both the AOM/DSS model of CAC [] and mutagen-induced hepatocellular carcinoma upon administration of diethylnitroseamine (DEN) []. In both of these models, when myeloid cells were defective in activating NF-κB via the classical pathway, there was impaired production of inflammatory mediators, proliferation of dysplastic epithelium, and a decrease in both the frequency and size of tumors compared to WT mice []. Many of these factors, such as IL-6, are required for hepatic regeneration after injury []. An important finding of the DEN study was that when NF-κB was impaired in hepatocytes, there was increased epithelial cell death, yet a increased tumor burden []. This finding suggests that myeloid cells may lead to proliferation of initiated cells by detecting epithelial cell death []. Thus, in the presence of initiation and both tissue injury and massive cellular death, activation of an inflammation dependent tissue repair/compensatory proliferative response leads to tumor promotion.
Pre-malignant tumors are “wound-like” []. Such tumors are similar to healing or desmoplastic tissue in many ways, such as the presence of activated platelets [,]. As described by Coussens and Hanahan, tumor growth may be “biphasic” []. In the first phase, the body treats early tumors as wounds. This phase is characterized by tumor growth mediated by the actions of the stroma “indirect control” as occurs in physiologic tissue repair. For example, in murine models of skin and pancreatic carcinogenesis, bone-marrow derived cells, including mast cells, are responsible for providing matrix metalloproteases which convert VEGF into a biological active form to stimulate the pro-tumorigenic angiogenic switch [,,]. However, during later tumor growth, it appears pro-inflammatory factors, such as MMPs, come under direct control by the tumors themselves []. A similar transition in the regulation of inflammation by early vs. late tumors may be at hand in spontaneous intestinal tumorigenesis in both mice and humans. COX-2 is expressed by stromal cells in early tumors [,], perhaps as part of a response to early tumor associated wounding; in larger tumors, COX-2 is expressed by the dysplastic epithelium itself []. One intriguing hypothesis as to why this phenomenon occurs is that there are intact regulatory mechanisms present in tumor associated stromal cells, limiting their expression of tissue repair factors. This may lead to the selective emergence of tumor cells that autonomously can maintain these ancillary processes and are not dependent on the “wound-like stroma.” Eventually, however, the tumor associated stroma may undergo selective pressure, as there have been recent reports of genetic changes in tumor associated stroma [] and even loss of p53 in tumor-associated fibroblasts [].
In addition to playing a critical role in tumor growth, such as by mediating angiogenesis, the inflammatory response may have a role in other aspects of progression, such as tissue invasion and metastasis. Angiogenesis itself augments vascular invasion of migrating cells. Matrix metalloproteases and their inhibitors (TIMPs) are critical for angiogenesis and also in remodeling of extracellular matrix []. Infiltrating leukocytes may further aid tumor progression by blazing a trail through the ECM, the countercurrent invasion theory [].
As delineated above, the cell death and tissue injury that may occur right after initiation may induce tissue repair and homeostatic responses leading to tumor progression. In addition, signals derived from microbes also might activate innate microbial recognition pathways enhancing the development of tumors. In this manner, the inflammatory response plays a causative role in early tumor development via regulation of initiation and promotion.
Mutagens lead to DNA damage, and DNA damage may activate NF-κB via cell death or by DNA damage itself []. While such cell death is likely to be below the threshold required for compensatory proliferation and tumor promotion, repeated exposure to mutagens may cause enough cell death to warrant the survival of some initiated cells and their expansion. Repeated exposure of mice to mutagens alone (such as AOM) can induce tumorigenesis. Given that the life history of mammals is rife with both temporary (infection or burns) and constant environmental insults (ingested or airborne irritants), exposure to mutagens is likely to be coupled with some form of low level irritation or injurous agent. Thus, mutagens and irritants/infection may be at work in carcinogenesis in more cases and ways than previously estimated.
As cancers grow, tissues change in many ways. Much of these changes may resemble tissue injury and induce homeostatic mechanisms to maintain tissue function. For example, the ischemia at the core of a solid tumor may initiate a physiologic response to hypoxia, such as activation of the HIF1 pathway and subsequent angiogenesis. Growing tumors also might degrade epithelial barriers, the cellular architecture of tissues, or disrupt the extracellular matrix. All of these processes are likely to stimulate homeostatic processes of tissue repair, including the recruitment of inflammatory leukocytes. These responses lead to tumor growth itself, thus promoting a positive feedback loop [].
The next step in understanding the relationship between homeostatic processes, such as compensatory proliferation, tissue repair, and host defense, and tumorigenesis will be to determine whether pattern recognition of disturbances in homeostasis are involved in tumorigenesis. Are these pathways involved in intitiating tumor associated inflammation? Recognition of microbes or endogenous ligands by toll-like receptors is critical for tissue repair and regeneration at various organs [-]. Besides the TLRs, there are many microbial pattern recognition receptors for detection of virus and bacteria []. In addition, it has been purported that certain receptors may be involved in sensation of tissue injury and cell death, such as those that recognize extracellular matrix fragments (such as hyaluronan or fibronectin) or necrotic debris (including S100 and heat shock proteins and HMGB1). Any role for these recognition pathways in tumorigenesis is ripe for investigation.
An open question is what role innate pattern recognition pathways have in carcinogenesis secondary to microbial colonization, such as in -induced gastric carcinoma. It will be important to determine the relative contributions of -derived factors critical for microbial colonization and survival in the host (such as “virulence” or “adaptation” factors) and -determined responses (such as inflammation) in tumor development as well as to determine synergy between these factors. Such investigations may identify important information regarding why certain forms of chronic inflammation do or do not predispose to tumor development. |
xref
#text
A 68-year-old woman was admitted to our department after complaining of a persistent cough over the course of a week, hemoptysis, and a temperature that would peak at 37.8°C (100°F) over the last two months. She had noticed a painless swelling in the region of the thyroid gland and a palpable mass in the right side of her neck. Four years ago, the patient had been diagnosed with squamous cell carcinoma of the cervix stage ΙΙΙΒ, moderately differentiated, for which she had received external beam radiotherapy to the pelvis and brachytherapy.
Upon physical examination, an irregular, hard-fixed mass measuring 8.6 cm was felt in the right lateral region of the neck, along with a firm thyroid enlargement. Chest auscultation revealed altered breath sounds in the left lung, with rales and rhonchi. Her blood cell count and serum biochemistry were normal, and the rate of sedimentation was 65mm the first hour. The patient underwent a CT scan of the neck, thorax, abdomen, and pelvis that revealed a large irregular mass 9 cm in diameter in the right lateral region of the neck, enlargement of the thyroid gland, enlarged mediastinal and paraaortic nodes, and multiple patchy airspace infiltrates, without evidence of local recurrence. A bone scan was negative for metastases. Fiberoptic bronchoscopy revealed edematous and erythematous mucosa of the upper bronchus of the right lower lobe and erythematous edematous mucosa with probable subserosa infiltration of the lingual bronchus. Sputum cytology after bronchoscopy washing and brushing was positive for undifferentiated carcinoma. Since the cytology report could not establish a definite diagnosis, a biopsy of the thyroid gland was performed, which showed non-keratinizing squamous cell carcinoma compatible with primary squamous cell carcinoma of the cervix. The pathologist also reviewed slides of the cervical biopsy from four years previous and confirmed that the thyroid tumor was histologically identical to the initially documented cervical carcinoma.
The patient received six cycles of systematic chemotherapy with cisplatin and gemcitabine, with partial response. Subsequently, she received radiation therapy to the right side of the neck and the thyroid, with further decrease of the lymph nodes and thyroid mass. Five months after the completion of radiotherapy, the patient developed progressive disease and succumbed four months later. The overall survival after diagnosis of the thyroid metastasis was 16 months.
Metastases to the thyroid may occur with high frequency in patients with widespread metastatic lesions. When the thyroid lesions become symptomatic, patients usually present with thyroid enlargement and palpable nodules, cough, hoarseness, dysphagia, and dyspnea. Thyroid implication to other solid malignancies may arise either by direct spread from adjacent tissues or by lymphatic and hematogenous spread, which is not surprising, considering the thyroid gland is one of body’s most abundantly arterialized organs [].
Up to 70 percent of patients with cervical cancer who present nodal metastases and/or locally advanced disease will relapse. Cervical cancer gives distant metastases via the lymphatic spread, from the satellite nodes to the paraaortic and supraclavicular nodes. The route of the hematogenous spread is presumed to occur via the blood stream to the caval venous system through the lung parenchyma and systematic circulation. In patients who develop distant metastases, the most frequently observed metastatic sites are the lung (21 percent), the para-aortic lymph nodes (11 percent), the abdominal cavity (8 percent), and the supraclavicular lymph nodes (7 percent). Bone metastases occur in only 16 percent of patients. Metastatic carcinoma of the thyroid gland from cancer of the cervix is very rare. Upon reviewing the English literature, only five cases of cervical carcinoma are described as having metastasized to the thyroid gland [-].
The involvement of the thyroid gland usually becomes apparent some years after the diagnosis of the primary malignancy and is a feature of tumor dissemination, which is indicative of a poor prognosis. In the context of widespread metastatic disease, the clinical manifestation of thyroid lesions is usually of no importance. In fact, surgical statistics reveal that lesions of metastatic origin represent only 8.5 percent of all removed malignant neoplasms. It is unknown whether secondary tumors of the thyroid are more common in an already abnormal thyroid gland. Two studies have been performed with contradictory results as one group of researchers maintains that thyroid metastases are associated with underlying thyroid abnormalities, while the other failed to establish such a correlation [,].
Explaining the nature of thyroid nodules (whether malignant or benign) can become a difficult diagnostic task, especially if they occur many years after the initial tumor. A differential diagnosis must be made of multinodular goiter and benign thyroid nodules. If the malignant phenotype is confirmed, it must be clarified as to whether it is a primary or secondary neoplastic lesion. Since both entities have the same clinical signs and symptoms (if any), diagnosis relies upon fine needle aspiration cytology (FNAC), a simple, inexpensive, and safe procedure with high negative diagnostic value. In most cases, confirming the cytological origin of the neoplastic cells can be based on morphology and immunocytochemical staining; negative staining with anti-thyroglobulin and anti-calcitonin antibodies would exclude the thyroid as primary origin of the neoplastic cells. On the contrary, thyroid follicular neoplasms tend to maintain some cytoplasmic granularity and, as a result, usually stain positively for thyroglobulin. In the pathology examination, the presence of multiple deposits within the thyroid gland makes the diagnosis of a metastatic tumor more probable. Furthermore, a predominantly interstitial pattern of penetration, with deformation and dissemination of the follicles by the neoplastic cells, is more compatible with metastatic origin of the carcinogenic cells. On the contrary, primary thyroid neoplasms usually infiltrate the follicles. Despite all the above techniques, occasionally it may be difficult to determine from cytology whether the neoplastic cells originate from the thyroid gland, especially in cases of anaplastic carcinoma or the unusual clear cell variant of follicular carcinoma. In any case, diagnostic re-evaluation of the primary tumor and search for other metastatic sites are mandatory to establish metastatic derivation of thyroid nodules [,].
The management of patients with metastatic thyroid lesions is determined by the primary tumor type, the presence of other metastatic sites, the patient’s symptoms and the patient’s performance status. Aggressive surgical treatment of an isolated thyroid metastasis, which is most often from renal cell carcinoma, could be curative. The regional application of radiotherapy in symptomatic patients may achieve sufficient improvement of the quality of life and constitutes an acceptable alternative approach. Finally, the location of other metastatic sites will determine the manner of systemic management. Since the presence of metastases in the thyroid gland is indicative of disseminated disease, the prognosis of these patients is poor, regardless of the primary tumor site and therapeutic approaches. The time between presentation of thyroid metastasis and death ranged from three to 45 months in one study of 79 patients [].
In conclusion, patients who present unilateral swelling or palpable nodules in the thyroid and have a history of a previous malignancy must be considered for metastatic disease, despite the fact that in most cases thyroid lesions are occult and do not pose a clinical problem. FNAC provides a quick, easy, and reliable way of diagnosis, while biopsy of the lesion and histopathological examination is the most appropriate tool for the final diagnosis. Although detection of metastases to the thyroid gland is often indicative of a poor prognosis, aggressive medical treatment such as radiotherapy and systemic chemotherapy may be effective in offering a better quality of life. |
Each year, approximately 14,000 people are stricken with brain cancer. The disease occurs across both social and economic lines, with incidence rates peaking both in childhood and later in old age. Despite advances in imaging technology — which has led to earlier diagnosis of many tumors — the ability to treat the most aggressive form of brain cancer, glioblastoma multiforme (GBM), has not improved since 1980. The one-year survival rate for invasive central nervous system (CNS) cancer was 57.9 percent in 2002, and survival for GBM in particular is even lower [].
Gliomas are primary CNS tumors arising from the glial cells. Malignant gliomas have a characteristic ability to infiltrate healthy brain tissue and form satellite tumors. This capacity for migration makes them exceedingly difficult to treat and invariably fatal. Even after resection, invasive cells can give rise to tumors within centimeters of the resection site []. Untreated malignant tumors can eventually spread to the contralateral hemisphere []. Many forms of systemic chemotherapy are excluded from the CNS by the blood-brain barrier (BBB) []. A few compounds — such as the class of antiproliferative drugs called nitrosoureas (including carmustine and lomustine) or other alkylating agents (temozolomide) — have some ability to cross the BBB and have been used clinically []. Unfortunately, systemic delivery of these agents appears to offer modest benefit as a supplement to radiotherapy [,].
Over the past two decades, a variety of approaches to enhance the activity of systemically delivered chemotherapy drugs have been tested. Hyperosmolar BBB disruption has been used to enhance BBB transfer of chemotherapy agents, with mixed results. One study, using PET imaging to evaluate a combination of methotrexate and hyperosmolar BBB disruption, indicated a negligible effect in brain tumors, which is echoed by the marginal findings in clinical trials [,]. A variety of approaches have been tested for enhancing BBB permeability of systemically administered drugs — by modification with hydrophobic side groups, conjugation to ligands with known BBB carriers, such as transferrin, or encapsulation in liposomes or nanoparticles — but none of these approaches have impacted clinical treatment of glioma [].
The failure of conventional systemic drug delivery for glioma has motivated more direct approaches to drug delivery. Direct intracranial drug delivery would eliminate the need for a chemotherapeutic agent to cross the BBB. The ability to bypass the BBB would enable a wider range of agents — such as paclitaxel, doxorubicin, immunotoxins, and even gene therapy vectors — to be evaluated for brain cancer treatment. This review describes two of the most promising approaches for direct delivery of agents to intracranial tumors: polymeric-controlled release and convection-enhanced delivery.
Polymeric-controlled release has long been used for drug delivery. Some early systems — the five-year subcutaneous Norplant® contraceptive and the conjuctival Ocusert® system for glaucoma — have proven the effectiveness of this approach for both systemic and local therapy []. The controlled release of drug also protects it from elimination [].
Controlled release systems can be designed from both degradable and nondegradable polymers. When constructed from nondegradable polymers, drug release is usually governed by diffusion of the drug through the matrix. In contrast, release from degradable polymers is governed by a combination of drug diffusion through the polymer and erosion of the polymer matrix. Polymers can be combined into copolymers to tune degradation and release characteristics. Correctly designed, polymer and drug systems can provide reliable sustained release for periods of days or many years (see review in []).
#text
Local delivery to brain tumors can be accomplished using the two strategies of polymeric-controlled release and convection-enhanced delivery. Each technique strives to address the need for controllable intracranial drug delivery. The two technologies offer unique benefits and suffer distinctive limitations, which are listed in Table 2. The principle disadvantage to controlled release is the restricted drug distribution. The limitations imposed by diffusion within the brain limit drug penetration to a region smaller than the typical invasive area of malignant gliomas.
CED’s strength lies in the potential for large distribution volume of infused drugs. The long infusion times and unpredictable distribution have also caused an abundance of side effects in recent clinical trials. Trials using paclitaxel have noticed a high rate of drug related adverse events that include wound dehiscence, inflammation, and edema [,]. These side effects are attributed to drug backflow along the catheter, and drug localization in the perivascular and subarachnoid spaces. Infusing an encapsulated drug may decrease the incidence of these adverse events.
CED was developed as a method to address the shortcomings of polymeric-controlled release. Coincidentally, many of the flaws in CED are the strengths of a polymeric-controlled release approach. Encapsulation of drug would limit the reflux and promote better wound healing after delivery. Delivering encapsulated drug could shorten the infusion time. A shorter infusion time would deliver a lower maximum drug concentration which could address the instances of edema and inflammation. Moreover, a combined treatment strategy could truly localize the delivery by restricting it to a prescribed area. Nanoparticle technology has advanced such that high levels of targeting ligand can be expressed on the surface of the degradable particles []. These particles could then be targeted to a subpopulation of cells.
A combined approach is not without its potential pitfalls. The polymer particles would have to be large enough to deliver a clinically relevant dose. The increase in size could restrict the distribution of particles in the parenchyma. While the adjuvants could improve the distribution, a multi-focal delivery strategy may be necessary to circumvent the transport limitations.
Local delivery to brain tumors has already provided a modest increase in survival when used in addition to surgery and radiotherapy. Despite the advances, GBM remains fatal regardless of the mode of treatment. There is untapped potential in both of the local delivery strategies discussed here. Polymeric-controlled release and CED can be used as techniques to administer the standard agents, combinations of drugs, as well as new methods of anti-tumor therapy. |
A
l
t
h
o
u
g
h
i
t
i
s
c
o
n
v
e
n
i
e
n
t
t
o
c
o
n
s
i
d
e
r
h
u
m
a
n
t
u
m
o
r
v
i
r
u
s
e
s
a
s
a
d
i
s
c
r
e
t
e
g
r
o
u
p
o
f
v
i
r
u
s
e
s
,
t
h
e
s
e
s
i
x
v
i
r
u
s
e
s
,
i
n
f
a
c
t
,
h
a
v
e
v
e
r
y
d
i
f
f
e
r
e
n
t
g
e
n
o
m
e
s
,
l
i
f
e
c
y
c
l
e
s
,
a
n
d
r
e
p
r
e
s
e
n
t
a
n
u
m
b
e
r
o
f
v
i
r
u
s
f
a
m
i
l
i
e
s
.
T
h
e
p
a
t
h
f
r
o
m
v
i
r
a
l
i
n
f
e
c
t
i
o
n
t
o
t
u
m
o
r
g
e
n
e
s
i
s
i
s
s
l
o
w
a
n
d
i
n
e
f
f
i
c
i
e
n
t
;
o
n
l
y
a
m
i
n
o
r
i
t
y
o
f
i
n
f
e
c
t
e
d
i
n
d
i
v
i
d
u
a
l
s
p
r
o
g
r
e
s
s
t
o
c
a
n
c
e
r
,
u
s
u
a
l
l
y
y
e
a
r
s
o
r
e
v
e
n
d
e
c
a
d
e
s
a
f
t
e
r
p
r
i
m
a
r
y
i
n
f
e
c
t
i
o
n
.
V
i
r
u
s
i
n
f
e
c
t
i
o
n
a
l
s
o
i
s
g
e
n
e
r
a
l
l
y
n
o
t
s
u
f
f
i
c
i
e
n
t
f
o
r
c
a
n
c
e
r
,
a
n
d
a
d
d
i
t
i
o
n
a
l
e
v
e
n
t
s
a
n
d
h
o
s
t
f
a
c
t
o
r
s
,
s
u
c
h
a
s
i
m
m
u
n
o
s
u
p
p
r
e
s
s
i
o
n
,
s
o
m
a
t
i
c
m
u
t
a
t
i
o
n
s
,
g
e
n
e
t
i
c
p
r
e
d
i
s
p
o
s
i
t
i
o
n
,
a
n
d
e
x
p
o
s
u
r
e
t
o
c
a
r
c
i
n
o
g
e
n
s
m
u
s
t
a
l
s
o
p
l
a
y
a
r
o
l
e
.
Hepatitis C virus is an enveloped RNA virus of the flavivirus family. It is capable of causing both acute and chronic hepatitis in humans by infecting liver cells. It is estimated that approximately 3 percent of the world’s population are hepatitis C carriers []. Chronic infection with hepatitis C virus results in cirrhosis, which in turn can lead to primary hepatocellular carcinoma. Between 1 and 2 percent of infected patients with subsequent compensated cirrhosis will develop primary hepatocellular carcinoma per year []. Transmission of the virus occurs through the blood, with shared needles in intravenous drug abuse, sexual activity, and parturition being the primary routes.
The hepatitis B virus of the family hepadnaviridae is, by contrast, a DNA virus, but the features of its resulting disease share many similarities with hepatitis C virus. Hepatitis B virus also is a blood-borne pathogen that can result in acute and chronic hepatitis. Chronic hepatitis, that is, infections lasting more than three months, can lead to cirrhosis and liver failure. Chronic infection also can lead to the development of hepatocellular carcinoma []. Hepatitis B infections is a significant global health problem with an estimated 2 billion people infected and 1.2 million deaths per year attributed to subsequent hepatitis, cirrhosis and hepatocellular carcinoma [].
Hepatocellular carcinoma is an aggressive tumor that can occur in the setting of liver disease resulting from infections with hepatitis B and/or hepatitis C virus, although the exact mechanism of oncogenesis by these viruses is unclear. Diagnosis is usually made late in the course of liver disease and median survival ranges from six to 20 months after that time []. The traditional foundation of treatment is surgical, whether tumor resection or transplantation. However, nonsurgical options such as percutaneous ethanol injection, transarterial embolization, radiofrequency ablation, chemotherapy, and radiotherapy are also utilized. The choice of therapies frequently depends on the extent of disease and the amount of liver function the patient has in reserve [,].
Research into novel therapies have focused on the use of virally targeted and immunological strategies with an eye on preventing infection. Unfortunately, hepatitis C virus has proved to be poorly suited to vaccines because its genome possesses a very high mutation rate, especially in the hypervariable region of the genome coding for the envelope proteins allowing it to escape immune recognition and elimination by the host. There are 11 distinct genotypes and several subtypes identified.
The introduction of vaccines against hepatitis B virus in the early 1980s marked a major milestone with what might be considered the first cancer prevention vaccine, although the primary goal of this vaccine was to prevent hepatitis. Since that time, more than 110 countries have adopted a universal policy of immunizing all newborns, according to the World Health Organzation. Additionally, countries that have successfully implemented this program significantly have decreased the carrier rate and infection in their populations []. However, vaccine coverage is often low in many developing countries due to the cost, lack of heath care infrastructure for delivery of the vaccine, and the need for three needle injections over six months. Even in some developed nations, universal vaccination has not been implemented because of the belief that it is a limited public health problem and the expense is not justified [,].
New challenges for combating hepatitis B infection center around efforts to address the limitations of the current vaccine: the need for multiple injections, the presence of up to 10 percent nonresponders to the vaccine, the discovery of hepatitis B virus S gene escape mutants in infants that were infected despite an adequate response to the vaccine, and the cost for developing nations. The current multiple dosing schedule is being addressed with attempts to combine it with other required vaccines or decrease the number of doses. Oral vaccination also is being investigated as a way to obviate the need for trained personnel to administer injections. The World Health Organization estimates that from $8 to $12 billion will be needed to immunize children from the poorest countries from 2005-2010, which has prompted efforts from public and private organizations to advocate for funding to fill the need.
Medical therapy for patients infected with hepatitis B has focused on the use of interferon to reduce viral replication, which decreases the incidence of life-threatening liver complications in patients who respond to the treatment []. Interferon alpha treatment is effective in 20 to 30 percent of cases in inducing loss of the hepatitis B e antigen. However, the impact of interferon therapy on subsequent hepatocellular carcinoma rates is less clear [,]. Interferon therapy is also limited by cost and side effects.
The limitations of interferon therapy have been partly circumvented with the use of targeted antiviral agents. Lamivudine has been shown in a large multicenter randomized placebo-controlled trial to be effective in reducing both the incidence of hepatic decompensation and the risk of hepatocellular carcinoma []. Other antiviral agents continue to join the armamentarium; lamivudine, adefovir, entecavir, and telbivudine have been shown to be effective in hepatitis B disease. These agents are nucleotide analogues that exploit the need for the hepatitis B virus to use reverse transcriptase to replicate viral DNA. Since these agents specifically target the viral replication machinery and are given orally, they are better tolerated. However, it has been observed that long-term therapy with lamivudine can lead to the emergence of genotypic resistance mutations [], but this does not negate the benefits of lamivudine therapy in reducing the rates of hepatocellular carcinoma []. The success of these therapies has reached the point where patients with advanced cirrhosis secondary to hepatitis B can be treated and transplanted without the development of hepatitis B in the transplanted liver.
Medical treatment of infection with hepatitis C has not progressed at the same speed. Pegylated interferon with ribavirin, an antiviral agent that may act as a nucleoside analogue and inhibitor of RNA dependent RNA polymerase, has been shown to be successful in eradicating infection in half of patients []. However, therapy is expensive and side effects are significant. Phase II trials of oral antivirals such as protease inhibitors and polymerase inhibitors are currently under way []. Unlike hepatitis B, treatment of hepatocellular carcinoma due to hepatitis C infection with transplantation almost always results in recurrent infection of the transplanted liver [].
The search for targeted therapies that can block hepatitis C viral replication by selectively inhibiting viral replication has for many years been hampered by the lack of experimental infection systems, in either cell culture or animal models to test candidate therapies. The recent development of viral replicons, subgenomic RNAs that are expressed and autonomously replicate within cells, has led to the use of hepatitis C viral replicons that can replicate in human hepatoma cells lines [] and the development of mouse models of the disease [,]. These advances may herald more rapid progress in the development of virally targeted therapies such as hepatitis C virus specific protease and polymerase inhibitors.
EBV and HHV-8 (also known as Kaposi sarcoma herpesvirus) are both herpesviruses that possess large double-stranded DNA genomes. As with all herpesviruses, they encode enzymes involved in DNA replication and repair and nucleotide biosynthesis. They also both possess the ability to establish latency in B lymphocytes and reactivate into the lytic cycle. Both also are associated with naturally occurring tumors in humans.
EBV is a ubiquitous virus that is most commonly known for being the primary agent for infectious mononucleosis. Up to 95 percent of all adults are estimated to be seropositive, and most EBV infections are subclinical. EBV also is associated with a number of malignancies: B and T cell lymphomas, Hodgkin’s disease, post-transplant lymphoproliferative disease, leiomyosarcomas, and nasopharyngeal carcinomas. Of these cancers, Burkitt’s lymphoma, post-transplant lymphoproliferative disease, and leiomyosarcomas show an increased frequency in patients with immunodeficiency, suggesting a role for immunosurveillance in the suppression of malignant transformation.
The primary site of infection is the oropharyngeal cavity, and EBV is capable of infecting both B cells and epithelial cells and switching between the two []. The major surface glycoprotein, gp350/220, binds to the cd21 receptor on B cells. Transformation of B cells is a highly efficient process requiring a large portion of the EBV genome, which becomes circular for replication and latency. Virus will directly enter the latent gene expression state with suppression of the lytic cycle. Production of a number of latent gene products are required for immortaliztion.
Immune therapy of EBV-associated tumors has been target of research since standard therapy generally has entailed the use of multi-agent chemotherapy, radiation therapy, and surgery. This work has centered around adoptive transfer of EBV-specific cytotoxic T-cells [,] and shown success but must overcome obstacles such as potential graft vs. host disease and resistance due to mutation of selected EBV epitopes []. Vaccines capable of preventing primary EBV infection or boosting immune responses against EBV-associated tumors are under investigation. Much of the development thus far has focused on gp350/220 subunit vaccines [], since it is one of the most abundant proteins on the virus coat and also the protein against which the human EBV neutralizing antibody response is directed []. Another strategy involves the use of a recombinant vaccinia viral vector to express an EBV membrane antigen []. A successful vaccine would have the greatest impact in regions of the world that have an especially high incidence of specific malignancies. Burkitt’s lymphoma is the most common childhood malignancy in the central part of Africa where EBV and malaria are considered cofactors in its carcinogenesis and 95 percent of children are infected by age 3, compared to the United States, where infection is usually delayed until adolescence []. Nasopharyngeal carcinoma is relatively rare but has an exceptionally high incidence in southern China, approaching more than 20 times greater than that of most populations [].
In 1994, HHV-8 DNA was identified in biopsies from tumors of a patient with Kaposi sarcoma [], a relatively rare malignancy prior to the AIDS epidemic. In addition to it likely being an essential cofactor for the development of Kaposi sarcoma, HHV-8 also is believed to have a role in Castleman’s disease and primary effusion lymphoma []. The viral genome is expressed in these tumors and encodes transforming proteins and anti-apoptotic factors. The virus is also able to enhance the proliferation of microvascular endothelial cells []. As with EBV, the predominant infected cell is the B lymphocyte, although here the lytic cycle is embraced rather than repressed. This may play a crucial role in the pathogenesis of Kaposis sarcoma by elaboration of viral and host cytokines promoting cell proliferation, angiogenesis, and enhancement of viral spread.
Targeted antiviral agents such as ganciclovir directed against viral DNA replication have had a dramatic affect on decreasing the incidence of Kaposi sarcoma in AIDS patients through both therapy and prophylaxis []. Ganciclovir is phosphorylated into a GTP analog, which acts as a competive inhibitor of viral DNA polymerase resulting in termination of viral DNA elongation. Furthermore, a G protein coupled receptor (vGPCR) has been identified as a viral oncogene in HHV-8 infected cells that can exploit cell signaling pathways to induce transformation and angiogenesis []. vGPCR also has been proposed as a target for novel molecular therapies because of its key role in disease progression []. But the therapy regimen most responsible for the decreasing incidence of Kaposi sarcoma may well be the success of highly active antiretroviral therapy (HAART) regimens targeting HIV [], since it was the emergence of HIV that led to the increasing incidence of Kaposi sarcoma.
HPV are small non-enveloped DNA tumor viruses that commonly cause benign papillomas or warts in humans. Persistent infection with high-risk subtypes of human papillomavirus (HPV) is associated with the development of cervical cancer []. HPV infects epithelial cells, and, after integration in host DNA, the production of oncoproteins, mainly E6 and E7, disrupts natural tumor suppressor pathways and is required for proliferation of cervical carcinoma cells []. HPV also is believed to play a role in other human cancers, such as head and neck tumors, skin cancers in immunosuppressed patients, and other anogenital cancers.
Cervical cancer is the second leading cause of cancer mortality in women worldwide, causing 240,000 deaths annually []. Of approximately 490,000 cases reported each year, more than 80 percent occur in the developing world, where effective but costly Pap smear screening programs are not in place []. Early precancerous changes and early cancers detected by Pap smears are effectively treated and cured with surgical therapy or ablation. In the absence of effective screening, the disease is detected late. Traditional therapeutic options for cervical cancer that have advanced beyond definitive surgical treatment are chemotherapy and radiation therapy, which are associated with many toxicities and do not offer a lasting cure.
The immune system plays an important role in the prevention of persistent HPV infection and progression of precancerous lesions. Human papillomavirus is a poor natural immunogen; as a double stranded DNA virus, there is no RNA intermediate, nor does infection cause cytolysis, allowing initiation of innate immune responses []. HPV mainly encodes non-secreted nucleoproteins, which are poorly cross-presented and compared to other viruses its non-structural proteins are expressed at low levels. However, genital infection with HPV is usually transient. Additionally, inadequate T cell responses may lead to failure to clear HPV-infected cells. AIDS patients, renal transplant patients receiving immunosuppressive therapy, and individuals with T cell deficiencies have increased rates of HPV persistence, anogenital lesions, and cervical cancer [-].
In 2006, an effective prophylactic vaccine against HPV 16 and 18 based on virus-like particles (VLP) of recombinant L1, the major capsid protein [,], was approved for use by the FDA based on clinical trials that demonstrated nearly 100 percent protection from persistent infection through the generation of high levels of neutralizing antibodies. Since these types are the causative agent of approximately 70 percent of cervical cancers, development of such an effective vaccine holds much promise for the prevention of cervical cancer []. However, the vaccine currently costs $360 for a complete course of three injections given over six months, does not provide protection against other high risk HPV types, will presumably have limited benefit to women already infected, and has an unknown duration of protection.
Because of these limitations, therapeutic vaccination is being explored to treat women already infected and accelerate the impact of prophylactic vaccination in decreasing cervical cancer incidence. Traditional therapy for early cervical cancer and precancerous lesions involves surgical excision or ablation. Therapeutic vaccination seeks to generate a population of cytoxic T cells that will recognize and kill tumor cells. Since patients with T cell deficiencies are known to be more susceptible to HPV infection and disease progression, boosting T cell responses to HPV may be crucial to a therapeutic immune strategy. In the case of cervical cancer, E6 and E7 oncoproteins are expressed in all malignancies and are not found in uninfected normal cells. Therefore, they represent ideal targets for a therapeutic immune response. A number of strategies to generate immune responses against these antigens are under investigation. Viral and bacterial vectors have been used in mouse models to generate immune responses. Vaccinia virus delivery of HPV 16 and 18 modified E6 and E7 proteins has demonstrated safety and specific immune responses in early clinical trials []. DNA vaccination strategies also are under active investigation, and several are in various stages of clinical trials. Vaccination with plasmid DNA encapsulated in biodegradable micorparticles has shown histological and immunological responses when used to treat patients with high grade cervical dysplasia [-].
HTLV-1 is a slow transforming, single stranded RNA retrovirus and is associated with adult T-cell leukemia []. It possesses a diploid genome similar to other retroviruses: two long terminal repeats flanking gag, pol, and env genes as well as a number of accessory genes. HTLV-1 has a worldwide distribution, with an estimated 12 to 25 million people infected. However, disease is only observed in less than 5 percent of infected individuals. It is transmitted through blood transfusions, sexual contact, and during parturition. HTLV-1 displays a special tropism for CD4 cells, which clonally proliferate in adult T cell leukemia, though how this is effected is not known.
HTLV-1 infection has a very long latency period of 20 to 30 years, but once tumor formation begins, progression is rapid. Standard chemotherapy often can bring about an initial response with a partial or complete remission; however, relapse is common, and median survival is eight months. The HTLV-1 gene has been postulated to play an important role in tumorgenesis [] through the activation of viral transcription and the hijacking of cellular growth and cell division machinery, but the mechanisms leading to adult T cell leukemia are not well understood. It has been suspected that HTLV-1 infection may not be sufficient to transform, and recent evidence suggests that the decreased diversity, frequency, and function of HTLV-1 specific CD8 T cells in the host may play an important part in the development of adult T-cell leukemia []. Therefore, targeted therapies using peptide, recombinant protein, DNA, and viral vectors with the goal of generating neutralizing antibody against HTLV-1 and multivalent cytotoxic T cell response against are under investigation [].
The viruses reviewed here illustrate the diverse biological pathways to malignancy and the challenges of treating the resulting diseases. Yet the presence of the viral gene products in cancer and precancerous cells present attractive targets that may be exploited in novel therapies that distinguish these cells from normal cells. Antivirals such as lamuvidine used in heptatitis B and ganciclovir for Kaposi sarcoma specifically target the viral replication machinery. Targeting cancer cells specifically would have advantages over traditional modalities such as chemotherapy and radiation, which can include significant toxicities. Cervical cancer, because it retains HPV viral oncoproteins E6 and E7 and requires their continued expression for proliferation, provides an ideal model for cytotoxic immune therapies against these known antigens.
Given the prevalence of these cancers in the developing world and the limitations of health care infrastructure, strategies for vaccine design to prevent primary infection and targeted therapies for the treatment of disease must be carefully considered in this context. Use of needles, refrigeration, multiple doses, and cost are all significant barriers to the delivery of an effective vaccine []. Cost, need for trained personnel and sophisticated equipment and facilities may impede global use of the most advanced targeted therapies. These challenges suggest that exploration of prophylactic strategies and development of specific, targeted therapies are both necessary to decrease this portion of the global cancer burden. |
Based on the GLOBOCAN database, there were about 10,862,496 new cancer cases (excluding skin cancer) in the world in 2002. Of these, 5,801,839 (53.4 percent) were male and 5,060,657 (46.6 percent) were female. Nearly 45 percent of the new cases were diagnosed in Asia, 26 percent in Europe, 15 percent in North America, 7 percent in Latin America, and 6 percent in Africa. For males and females combined, the most common cancer site worldwide was lung (965,446 male and 386,875 female cases per year). The second most common site was colon (550,513 males and 472,743 females), followed by stomach (603,003 males and 330,290 females). Among women, the number one cancer site was breast (1,152,161 new cases per year), followed by cervix (493,100 cases), and colon (472,743 cases). Among men, the three most common cancer sites were lung (965,446 cases), prostate (679,060 cases), and stomach (603,003 cases).
The number of deaths caused by cancer worldwide in 2002 was 6,723,887, among which 3,795,991 were male and 2,927,896 were female. Lung cancer led to most cancer deaths in the world. In 2002, the total death toll due to lung cancer was 1,179,074, of which 848,321 were male and 330,753 were female. The second on the list was stomach cancer, which resulted in a total of 699,803 deaths, including 445,691 in males and 254,112 in females. Liver cancer was the number three cause of cancer mortality. A total of 598,412 deaths (416,926 male and 81,486 female) were attributed to liver cancer in 2002. For women, the top three sites for cancer mortality were breast (411,093 deaths), lung (330,753 deaths), and cervix uteri (273,449 deaths), while lung, stomach, and liver constituted the top three sites for cancer mortality in men.
Of the 21 regions listed in the GLOBOCAN 2002 database, East Asia had the largest number of incident cancer cases (all ages, all sites except skin) in 2002 (n = 2,890,311); North America and South Central Asia were second (n = 1,570,520) and third (n = 1,261,527) on the list, respectively []. The pattern of cancer sites varied substantially from region to region. For example, the three most common cancer sites among individuals 15 years or older in East Asia were stomach (18.9 percent), lung (17.1 percent), and liver (14.3 percent), whereas those in North America were prostate (16.5 percent), breast (14.7 percent), and lung (14.5 percent) [ and ].
For both males and females, the incidence rate of cancer increased substantially with age. For example, the annual male cancer incidence in the age group of 0 to 14 years was 6.45 per 100,000 in Western Africa, 9.07 per 100,000 in Eastern Asia, 14.10 per 100,000 in Western Europe, and 15.12 per 100,000 in North America; the rates in the same regions for those who were 65 years or older were 385.44, 1461.59, 2327.87 and 2958.14 per 100,000, respectively (Table 1). North America, Australia/New Zealand, and Europe had the highest overall incidence rates in 2002, while Northern and Western Africa had the lowest incidence rates (Tables 1-2). The geographic variation was rather substantial. For example, the age-standardized rate in North American males (398.4 per 100,000 person-years) was four times of the age-standardized rate in North African males (99 per 100,000 person-years).
The geographic disparity in cancer incidence is largely attributable to the various socioeconomic, environmental, and lifestyle factors in different regions of the world. Compared with developed countries, developing countries in general may lack the resources to ascertain incident cancer cases. For example, in developed countries, many cases of breast, prostate, colon, and cervical cancers are identified through screening (e.g. mammography, prostate-specific antigen test, colonoscopy, and Pap smear), whereas in developing countries, large-scale screening efforts are usually uncommon. Genetic factors also play a role, but the dominant effect of genetics is only observed in a relatively small percentage of the population. It is believed that the majority of cancer cases (over 90 percent) are due to the joint effect of genetic variations, environmental factors, and lifestyle choices []. Geographic factor per se probably has little influence on cancer risk except sunlight exposure and vitamin D metabolism, both of which have been linked to cancer risk. The major categories of cancer risk factors include tobacco use, occupational exposures, environmental contamination, infectious agents, and lifestyle factors.
#text
Cancer is primarily a disease of old age. If all other factors remain the same, the demographic change (population growth and an increasingly higher percentage of older individuals in the world population) will lead to a global increase of cancer incidence. The GLOBOCAN 2002 database provides a means to project cancer incidence in the future. In 2002, there were an estimated total of 5,801,839 male individuals diagnosed with cancer (all sites except skin). If the age-specific incidence rates remain the same, population growth and aging of the world population would result in a projected total of 6,993,778 in 2010, a 20.5 percent increase within eight years (Table 4). The picture is similar for females. The total number of females diagnosed with cancer (all sites except skin) is projected to change from 5,060,657 in 2002 to 6,037,753 in 2010, which represents a 19.3 percent increase (detailed data not shown). This underscores the importance to improve our understanding of the risk factors of cancer, design and implement practical prevention strategies, and develop better and more effective treatment options.
The population compositions are dramatically different between more and less developed countries. Less developed countries have a smaller percentage of older individuals []. As cancer incidence increases with age, less developed countries have a larger population base and potentially more room for population aging. The foreseeable increase in the global burden of cancer likely will be more profound in less developed countries. In a recently published report, Cancer Control Opportunities in the Developing World [], by the Institute of Medicine of the National Academy of Sciences, it is recognized that cancer is a significant disease burden in low- and middle-income countries, and the burden will become increasingly heavy for these countries not only because these nations are more populous, which give rise to more cases, but there are also more aggressive cancers and lower cure rates. The report also indicates that cancer causes and outcomes are very different between more and less developed countries. For example, one in four cancer cases in developing countries are related to infectious agents compared to less than one in 10 cases in the developed nations. These disparities suggest different cancer prevention and control strategies for more and less developed countries in directing resources into the field of cancer research with regard to etiology, prevention, treatment, and health policy to reduce the burden of cancer globally. |
Cancer is an extremely heterogeneous disease; tumors in different tissues display strikingly different behaviors. For example, tumors of the pancreas tend to be highly aggressive, while prostate tumors are more frequently organ-confined. Tumors that arise in the same tissue can even exhibit an array of cellular pathologies, ranging from benign hyperplasias to highly invasive malignancies [-]. Cancer is also a complex disease involving the deregulation of multiple signal transduction pathways. Since the discovery of the first tumor-promoting gene, hundreds of genes have been shown to play a role in tumor initiation and progression, and research continues to uncover many more. Microarray analysis has identified thousands of genes that are transcriptionally upregulated or downregulated in cancer samples [-]. It remains unclear which transcriptionally deregulated genes in an individual tumor play a causative role in tumor initiation and maintenance and which ones represent bystanders with no selective advantage. Regardless of the role specific genes may play in cancer progression, these studies underscore the fact that by the time a tumor is histologically identified, it has accumulated a large number of molecular lesions. In addition, a recent literature survey of all published cancer genes identified 291 genes for which there are molecularly characterized mutations and evidence of a causative role in tumorigenesis. These genes represent more than 1 percent of the human genome []. Yet this number is a conservative estimate, since the study did not consider epigenetic regulation of gene expression. The large number of mutations found in tumor samples raises the question of whether it is biologically meaningful to classify cancer as a single disease entity. Is there a common thread that underlies most, if not all, human malignancies? Are there biological rules that govern cancer initiation and progression?
Despite the heterogeneity observed in cancer, most tumors share certain characteristics: self-sufficiency in growth signals, insensitivity to anti-growth signals, evasion of apoptosis, acquisition of a limitless replicative potential, sustained angiogenesis, and tissue invasion and metastasis. These common traits, which have been termed “the hallmarks of cancer,” allow tumors to breach cellular barriers against expansion and metastasis []. The similarity in the cellular processes subverted in all cancer cells, regardless of their tissue of origin, likely indicates common tumor-initiation mechanisms for the complex pathologies observed in clinical tumor samples.
Cellular transformation — the process by which a normal cell is “transformed” into a malignant one — is thought to take place through the accumulation of mutations, as well as epigenetic changes, that activate oncogenes or downregulate tumor-suppressor genes and lead to uncontrolled clonal expansion. Oncogenes originally were identified as the transforming agents of tumor viruses. It was later found that oncogenes were mutated versions of normal cellular genes, or proto-oncogenes, which had been incorporated into the viral genome by recombination. Mutations or epigenetic events, leading to the deregulated activity or increased expression of cellular oncogenes, are found in most cancers. Oncogene activation is implicated in the positive control of cellular growth, and mutations in oncogenes are generally dominant. In contrast, tumor suppressor genes function as negative regulators of cellular growth. Mutations in tumor-suppressors usually inactivate gene function and are generally recessive. Thus, inactivation of both copies of a tumor-suppressor gene is usually necessary for tumor development []. Efforts to define the mechanistic origins of cancer have focused on identifying such genes, as well as the pathways they regulate.
Several lines of evidence indicate that mutations in a single oncogene or tumor-suppressor are insufficient to give rise to cancer. First, most cancers develop late in life and the incidence of disease increases dramatically with age. Statistical analysis of epidemiological data shows four to five rate-limiting steps as necessary for cancer to occur, implying that a cell needs to accumulate four to five sequential genetic lesions in key regulatory pathways in order to become malignant []. Second, experiments using cell lines, as well as models of cancer, confirm the multiple-hit hypothesis for the majority of cancers, with retinoblastoma and certain types of leukemia being exceptions to the rule.
Initial research using transforming retroviruses, which contained activated versions of normal growth-controlling genes, indicated that alterations in a single gene could lead to transformation of rodent cells in culture [-]. However, the cells used in these initial cancer studies were immortal and could therefore proliferate indefinitely. In addition, these cell lines most likely had acquired a series of other genetic alterations in culture. When these experiments were repeated with primary cell lines, it was found that activation of at least one pair of oncogenes was required for transformation. Research by Land et al. [] and Sinn et al. [], along with experiments carried out with other sets of oncogenes, confirmed that at least two cooperating mutations are required for cancer.
Clinically, tumors are histologically classified as presenting with different “grades,” which correspond to a set of physiological markers (such as loss of differentiation, abnormal ploidy, and morphology) and correlate with patient outcome. Higher grade tumors have a more negative prognosis, while low-grade tumors are often considered early lesions and may progress to more invasive, high-grade disease. These observations have led to the hypothesis that cancer progression can be dissected into a small number of crucial steps whose sequential deregulation is critical for the clinical progression from low- to high-grade cancer. Several pathways with key functions in normal cell biology are deregulated in tumors:
The multi-stage model of cancer postulates that a series of sequential events that progressively bypass cellular growth-control mechanisms are needed for tumorigenesis. In contrast, the cancer platform model posits that the principles governing cancer initiation can be further reduced to two interdependent conditions — stimulation of proliferation with a simultaneous block of cell death within a single cell. In this model, proliferation and death pathways are linked not only in tumor initiation, but also in normal development.
Early experiments with transforming retroviruses showed overexpression or activation of oncogenes frequently led to growth arrest or apoptosis. To date, most oncogenes have been found to either sensitize cells to cell death, directly cause cell death, or promote growth arrest. For example, activation of in rat embryo fibroblasts leads to growth arrest, while activation of and promotes cell death [-]. These observations suggest that uncontrolled cellular proliferation could trigger these pathways as a means to control unrestrained growth. However, proliferation and death also are part of normal development, and a high-proliferative rate is required for certain developmental periods. Therefore, a cell would have to be able to distinguish between a normal high proliferative rate and the abnormal cell proliferation characteristic of cancer cells in order to appropriately activate cell death pathways in response to excessive cell growth.
During development, environmental cues — most importantly, growth factor and nutrient availability — determine whether a cell is able to proliferate. Thus, it was proposed that the activation of pathways leading to cellular growth stimulates both proliferation and death; only when trophic environmental factors that support growth by blocking cell death are present is a cell able to proliferate. The ability to stimulate both proliferation and death has been recognized for several oncogenes, including , , , , CDKs and cyclins. Furthermore, it was found that in fibroblasts, -induced apoptosis could be inhibited by serum or IGF and that E2F induced high rates of cell death in the absence of serum. These results suggest environmental signals are indeed crucial for oncogenic stimulation of proliferation [,]. The interdependence between proliferation and death could represent an evolutionary response against cancer progression.
According to the cancer platform model, cancer is a rare occurrence due to the low statistical probability of one cell gaining two mutations simultaneously in a single, rate-limiting step of cancer initiation. Once a “cancer platform” of uncontrolled cellular expansion is established, the interaction of this expanding cellular mass with its environment will give rise to subsequent mutations that allow for other traits of malignancies to develop: angiogenesis, immune evasion, invasion, and metastasis. In addition, since apoptosis is frequently triggered in DNA-damage response pathways to eliminate unwanted cells, blocking apoptosis can lead to an increase in mutation rates. Finally, several of the characteristics of cancer cells may reflect intrinsic properties of proliferating cells and expanding tissues, rather than the accumulation of new mutations, as well as properties of the cells and/or tissue of origin. [,]. For example, in comparison to more differentiated cells, oncogenic mutations in a stem cell may give rise to more malignant, undifferentiated cancers with self-renewal capacity [].
Both models discussed above propose that the complexity of the cancer phenotype can be reduced to a small set of common pathways that must be deregulated for cancer progression. The cancer platform model identifies two processes of key importance in tumorigenesis: cellular proliferation and death. This has important therapeutic significance since it implies that cancer cells could be eliminated by targeting either the oncogenic lesion that confers proliferative advantage to the tumor cells or the apoptotic pathways deregulated in these cells. Understanding the mechanisms of oncogenic stimulation of proliferation and death is important to dissect specific cancer-initiation pathways and to develop therapeutics.
Even though the same processes seem to be deregulated in all cancer cells, tumors arising in different cells or tissues may preferentially deregulate specific pathways that contribute to these processes. In addition, specific oncogenes or tumor-suppressors may be more frequently mutated or exhibit altered expression in some tissues. It is possible there is a genetic signature in different types of cancer determined both by the tissue and cells of origin and by the oncogenic and tumor-suppressive lesions it has undergone. Moreover, if deregulation of an oncogene activates a specific cell death pathway, it is likely that tumors in which this oncogene is deregulated also successfully have blocked that pathway. While mutations occur at random, once the first (or first two) lesions have been selected for and fixed in a clonal population, the new mutations the tumor acquires could be influenced by external environmental selection pressures as well as internal selection pressures of the mutations already selected. Thus, tumor initiating mutations may predict what types of mutations may occur later in the life of a tumor.
Colon cancer is one of the few malignancies for which a genetic pathway has been defined. Colorectal cancers follow a defined histological pattern of development from adenomas to carcinomas; each of these histological changes is accompanied by mutations in specific genes in a large percentage of tumors []. More recently, microarray analysis has been used to generate expression-based classifications of different tumor types. Most tumors show characteristic expression signatures recognizable both for individual tumors and for tumor families with shared characteristics. Further, molecular classification of tumors has revealed different tumors show similarities that can be ascribed to the tissue or cell-type of origin. In addition, tumors’ molecular signatures can be grouped to predict clinical outcome. For example, analysis of histologically indistinguishable breast cancer samples identified four subgroups: ER+-luminal like, HER2+, normal breast, and basal-like; of these four, the last was a predictor of poor outcome. These findings suggest there are subsets of mutations that correlate with specific types of cancer, as well as subsets of genes that correlate with the degree of malignancy of specific tumors [,].
The concept that there is a genetic signature to cancer is compatible with all of the models discussed so far. In principle, it would be possible to describe pathways for tumors in different tissues and with different cellular origins, which could predict outcome and help design specific therapies. The existence of genetic pathways may imply that late-stage tumors are still dependent on the original lesions for survival. Alternatively, new mutations may not be influenced by earlier ones. Once a specific process is thwarted, as in the bypassing of barriers against uncontrolled proliferation, new mutations are selected independently of the original mutations. If the first approach is correct, understanding tumor-initiating events in the context of different molecular lesions will be crucial to develop effective cancer therapies.
If tumors remain dependent on their initial transforming oncogenic mutations for growth and survival, oncogene inactivation could lead to tumor regression, even in malignant cancers. This hypothesis has been tested using inducible mouse models of cancer []. In particular, several studies evaluating the overexpression of the oncogene in lymphoid and epidermal tissues showed that the inactivation of led to sustained tumor regression with concomitant promotion of either differentiation or apoptosis [-]. However, in other models, a fraction of tumor cells were found to be refractory to inactivation; these cells presumably had acquired new mutations that allowed -independent growth [-], suggesting that while mutations that give rise to tumors are often interdependent, new lesions also can arise independently of pre-existing ones, often replacing their function. Therefore, targeting tumor-initiating mutations may not eliminate all tumor cells.
Metastasis represents the main cause of treatment failure for cancer patients, since even complete resection of the primary tumor can leave behind undiscovered micrometastases. The traditional model of metastatic progression postulates that only a small subset of cells from the primary tumor have acquired the requisite mutations to metastasize to distant sites, where new mutations are accumulated as a response to the different selective pressures of a novel environment []. However, recent data suggest that most cells in primary tumors with metastatic potential already contain the lesions necessary for metastasis and, possibly, for survival in a foreign environment. Microarray analysis compared patterns of gene expression in lymph node-negative breast cancer patients with their known five-year survival and recurrence rates. Seventy genes were identified that could predict clinical outcome with a combined 83 percent accuracy []. In addition, it was found that solid tumors of different origin shared the same metastatic signature, implying there is a common set of molecules regulating metastasis in a variety of primary tumors [,]. If this model is correct, it follows that mutations involved in tumor initiation also may be predictive of clinical outcome. The ability to identify and understand the molecular signatures of metastatic and non-metastatic primary tumors would provide new prognostic markers. In addition, if mutations that confer metastatic potential are present in the primary tumor, and if metastatic lesions remain dependent on the original oncogenic mutations for their survival, targeting these genes also may be an effective therapy against metastatic spread. Delineation of the genetic pathways involved in specific tumors will be crucial for identifying these initial oncogenic mutations.
Different models of cancer initiation have focused on deregulation of proliferation and cell death as the main engines of cancer progression. However, impaired differentiation is a characteristic of most cancers, as a decrease in the degree of differentiation correlates with highly malignant lesions. Several oncogenes have been shown to regulate cell-fate decisions. Thus, depending on the cellular context, oncogenes can promote not only proliferation and death, but also differentiation, which can act as a failsafe mechanism against unrestrained growth []. Expression of in bone marrow cells leads to a loss of cell-renewal activity in hematopoietic stem cells leading to differentiation []; and are highly expressed in developing neurons and their overexpression leads to neurite outgrowth in PC12 cells [-]. In addition, oncogene activation does not always lead to cell death and may even protect against it. In these cases, terminal differentiation could be an effective mechanism to thwart tumor progression [].
The dual cancer platform may not be sufficient for cancer progression in all contexts, and a third axis may be needed: cellular differentiation. In this expanded model, only when oncogene-induced differentiation effectively is blocked by additional mutations or when the cellular environment fosters the proliferating function of the oncogene will tumors arise. Promoting proliferation while simultaneously preventing differentiation thus may constitute in specific situations a sufficient platform for cancer expansion. In others, the simultaneous blockade of apoptosis and differentiation, together with the promotion of proliferation, may be needed to establish a cancer platform. Mutations that block cellular differentiation likely will have oncogenic capabilities in the context of molecular lesions that deregulate proliferation and prevent cell death. Identification of genes responsible for cell-fate determination may thus provide new insights into mechanisms of cancer initiation as well as provide novel targets for cancer therapies. |
Subsets and Splits