|Home | About | Journals | Submit | Contact Us | Français|
The hepatitis C virus (HCV) NS5B protein is an RNA-dependent RNA polymerase essential for replication of the viral RNA genome. In vitro and presumably in vivo, NS5B initiates RNA synthesis by a de novo mechanism and then processively copies the whole RNA template. Dissections of de novo RNA synthesis by genotype 1 NS5B proteins previously established that there are two successive crucial steps in de novo initiation. The first is dinucleotide formation, which requires a closed conformation, and the second is the transition to elongation, which requires an opening of NS5B. We also recently published a combined structural and functional analysis of genotype 2 HCV-NS5B proteins (of strains JFH1 and J6) that established residue 405 as a key element in de novo RNA synthesis (P. Simister et al., J. Virol. 83:11926–11939, 2009; M. Schmitt et al., J. Virol 85:2565–2581, 2011). We hypothesized that this residue stabilizes a particularly closed conformation conducive to dinucleotide formation. Here we report similar in vitro dissections of de novo synthesis for J6 and JFH1 NS5B proteins, as well as for mutants at position 405 of several genotype 1 and 2 strains. Our results show that an isoleucine at position 405 can promote both dinucleotide formation and the transition to elongation. New structural results highlight a molecular switch of position 405 with long-range effects, resolving the implied paradox of how the same residue can successively favor both the closed conformation of the dinucleotide formation step and the opening necessary to the transition step.
The hepatitis C virus (HCV) is an enveloped positive-strand RNA virus belonging to the genus Hepacivirus in the family Flaviviridae (28). The genome of HCV is a 9,600-nucleotide (nt)-long RNA molecule encompassing a single open reading frame (ORF) that is translated primarily into one polyprotein and flanked by nontranslated regions (NTRs). The NTRs are the most conserved parts of the viral genome and play important roles in viral translation and replication. The polyprotein precursor is cleaved by cellular and viral proteases into at least 10 different mature proteins. The nonstructural proteins NS3 to NS5B are associated within a membrane replication complex in which NS5B is the RNA-dependent RNA polymerase (RdRp). NS5B copies the RNA genome into a complementary negative strand and subsequently uses this negative strand as the template for the synthesis of new HCV genomes (for a review, see reference 21).
In vitro, recombinant NS5B is a highly processive RdRp that copies RNA templates up to the size of the HCV genome (4). C-terminal deletions of the 21-residue transmembrane helix (NS5BΔC21) are fully active and much easier to produce and purify than the full-length enzyme (37), so that these deletions are used almost exclusively for structural and functional characterization of NS5B. These studies have shown that NS5B can initiate RNA synthesis by a de novo mechanism (19) allowing full copy of the template in a single polymerization round (14, 29). De novo initiation from the 3′ end of the viral positive- and negative-strand RNA is likely to be the physiological mode of initiation of RNA synthesis in infected cells (8). Recombinant NS5B is very poorly active overall despite its high processivity and a turnover number comparable to that of other polymerases (7). Dissections of the early steps of RNA synthesis in vitro (12, 14, 29) have shown that this is due to two inefficient steps in de novo RNA synthesis by NS5B; the first step is de novo initiation proper, i.e., formation of the first dinucleotide complementary to the last two 3′ bases of the template, and the second step is use of this dinucleotide as a primer for further RNA synthesis.
Structural analysis of NS5B led us to propose that these two steps are closely linked to two conformational states of NS5B (14). This RdRp is consistently crystallized in closed conformations that would not allow the egress of a double-stranded template-primer RNA during processive RNA synthesis (6). Key features of the closed conformations are closure of NS5B's so-called thumb domain and occlusion of the egress from the catalytic site by the segment (termed a “linker”) just upstream of the transmembrane helix. Hence, a conformational transition to a displaced linker and more open thumb must be postulated before NS5B can elongate a dinucleotide primer. On the other hand, dinucleotide formation itself seems to occur in a closed conformation, where NS5B likely provides a protein platform to stabilize the de novo initiation complex, similarly to homologous RdRps from double-stranded RNA viruses (5, 14, 35). In accordance with this model, we established a link between a very closed conformation and an unusual de novo RNA synthesis efficiency in JFH1 strain NS5B (34). Introducing the central JFH1 V405I polymorphism into closely related J6_NS5B recently allowed us to boost J6_NS5B's de novo RNA synthesis in vitro and J6_NS5B-based chimeras' replication in cells. An infectious J6_NS5B-based chimeric virus was thereby generated with just two nucleotide substitutions (30). We hypothesized that by stabilizing a very closed NS5B conformation, the V405I mutation favors the very first step in de novo synthesis (formation of the first phosphodiester bond), with limiting impairment of the subsequent steps (elongation of this dinucleotide primer).
In order to buttress or refute this model, we performed in vitro dissections of de novo RNA synthesis for J6 and JFH1 NS5B proteins, as well as for mutants at position 405 for several genotype 1 and 2 strains. Our functional results show that I405 can improve not only dinucleotide formation, as expected from our model, but also transition to elongation. New structural results, particularly the crystal structure of J6-V405I_NS5B, shed light on how position 405 is also involved in the transition from initiation to elongation.
How JFH1_NS5B, J6_NS5B and Con1_NS5B mutants were obtained was described in reference 30. The J4_NS5B I405V mutant was obtained by using the QuikChange site-directed mutagenesis kit (Stratagene) and oligonucleotides AGACACACTCCAGTCAACTCTTGGCTAGGC and TAGCCAAGAGTTGACTGGAGTGTGTCTAGC as forward and reverse primers, respectively.
NS5B with a C-terminal deletion of 21 residues (NS5BΔC21) from HCV strains Con1, JFH1, J4, and J6 was expressed as already reported (30).
A V405I mutant form of strain J6 NS5BΔC21 (J6-V405I_NS5B), used in structural work, and full-length NS5B of HCV JFH1 (JFH1_NS5B-FL), C terminally fused to a hexahistidine tag, were expressed in Escherichia coli BL21(DE3) or Rosetta(DE3) cells. Glucose (1%) was added to repress NS5B expression in all media except the induction medium. Carbenicillin was used as the antibiotic instead of ampicillin for BL21(DE3), and carbenicillin and chloramphenicol were used for Rosetta(DE3). Bacterial cells were grown to an optical density at 600 nm (OD600) of 0.6, expression was induced by the addition of 1 mM isopropyl-β-d-thiogalactopyranoside, and cells were further incubated for 5 h (JFH1_NS5B-FL) or overnight (J6-V405I_NS5B) at 23°C with shaking and sedimented for 10 to 15 min at 6,000 × g. For J6-V405I_NS5B, absolute ethanol was added to a final concentration of 3% at an OD600 of 0.3 to 0.4 and the culture was then placed at the induction temperature.
NS5BΔC21, used in biochemical analyses, was purified as previously described (3, 30). The enzymatic properties studied here were identical whether the purification protocol included a cation exchange chromatography step (3) or not (30). The proteins used in structural work (J6-V405I_NS5B and JFH1_NS5B) were purified as reported in reference 30.
Pellets from 1-liter cultures after the induction of full-length JFH1_NS5B were resuspended in 40 ml lysis buffer (50 mM Tris-HCl [pH 7.5], 500 mM NaCl, 20 mM imidazole [Sigma-Aldrich], 1% n-dodecyl-β-d-maltopyranoside [DDM], 20% glycerol, 1 mM dithiothreitol [DTT], 1 mg/ml lysozyme [Sigma-Aldrich], 25 U/ml benzonase [Merck, Darmstadt, Germany], and protease inhibitors [Complete EDTA free; Roche]). Incubation for 2 h with mild shaking at 4°C was followed by five freeze-thaw steps and centrifugation for 20 min at 10,000 × g at 4°C. The supernatant was incubated with 1 ml of Ni-Sepharose (GE Healthcare) for 2 h with mild shaking. One milliliter of Ni-Sepharose was then washed 2 times (2 × 20 ml) with buffer A (50 mM Tris-HCl [pH 7.5], 500 mM NaCl, 20 mM imidazole, 0.03% DDM, 20% glycerol, 1 mM DTT). The bound proteins were eluted with 4 ml of buffer B (buffer A with 750 mM imidazole) at room temperature. The fractions containing NS5B-FL were identified by SDS-PAGE and pooled. The pool was subjected to Superose 6 10/300 GL fast protein liquid chromatography (FPLC) at a flow rate of 0.4 ml/min. Resin was previously equilibrated with buffer C (50 mM Tris-HCl [pH 7.5], 500 mM NaCl, 0.03% DDM, 20% glycerol, 1 mM DTT). The purified fractions were pooled and diluted to a 150 mM NaCl final concentration and subjected to MonoS 5/50 GL FPLC at a flow rate of 0.5 ml/min. Resin was previously equilibrated with buffer C containing 150 mM NaCl. The bound proteins were eluted with 20 ml of a linear gradient of buffer C with 150 mM NaCl and buffer C containing 1 M NaCl at a flow rate of 0.5 ml/min at room temperature. The bound NS5B was eluted at about 700 mM NaCl at room temperature. The fractions containing JFH1_NS5B-FL identified by SDS-PAGE were pooled and concentrated on a 100-kDa-cutoff ultrafiltration unit. Purified JFH1_NS5B-FL was flash frozen by being dipped into liquid nitrogen and kept at −80°C until use.
Protein concentrations were determined by measuring OD280 with extinction coefficients calculated from the constructs' sequences.
Two templates were used in gel-based initiation and elongation assays. They correspond to 341 nt of the 3′ end of the minus strand RNA of HCV genotype 1 (G1-C) or 3 (G3-U). They were obtained by in vitro transcription as previously described (20). The assay was carried out at 30°C with 20 μl of 20 mM Tris (pH 7.5)–5 mM MgCl2–1 mM DTT–0.4 U/μl RNasin–10% dimethyl sulfoxide (DMSO)–0.7 or 1 μM RNA template–0.7 or 1 μM NS5B. As recombinant proteins were stored in a high salt concentration, in all assays, NaCl was adjusted to a final concentration of 80 mM. When the initiation phase alone was analyzed, G3-U RNA was used as the template and 0.5 mM ATP and 10 μM CTP with 4 μCi [α-32P]CTP (3,000 Ci mmol−1) were the only added nucleotides. At different time points, 4 μl was collected, the reaction was quenched by the addition of 1 μl of 100 mM SDS, and the reaction mixture was diluted in 8 μl of 90% formamide–10 mM EDTA–0.025% SDS–0.025% bromophenol blue–0.025% xylene cyanol loading buffer. After denaturation at 70°C for 5 min, samples were loaded onto a 22% polyacrylamide gel in TBE buffer containing 7 M urea. The migration was performed at 60 W for 4 h, and the gel was submitted to electronic autoradiography using a Pharos apparatus and the quantity one software (Bio-Rad). For analysis of the elongation phase, the template used was the G1-C RNA. The reaction was performed at 30°C in 20 μl containing 20 mM Tris (pH 7.5); 80 mM NaCl; 5 mM MgCl2; 1 mM DTT; 0.4 U/μl RNasin; 10% DMSO; 0.5 mM ATP, GTP, and 3′dUTP; and 100 μM CTP with 4 μCi [α-32P]CTP (3,000 Ci mmol−1; Perkin-Elmer), 0.7 or 1 μM RNA template, and 0.7 or 1 μM NS5B. At different time points, 4 μl was collected, the reaction was quenched by the addition of 1 μl of 100 mM SDS, and the reaction mixture was diluted in 8 μl of 90% formamide–10 mM EDTA–0.025% SDS–0.025% bromophenol blue–0.025% xylene cyanol loading buffer. Reaction products were fractionated and analyzed as described above. The molecular weight markers were synthesized as described in reference 14.
Crystals were obtained by the hanging-drop vapor diffusion method and flash cooled by being plunged into liquid nitrogen. Initial crystallization screens, set up using robotics (Cartesian MicroSys) with the vapor diffusion method from 200-nl sitting drops, produced small rods. After optimization of the conditions, the crystals with the best diffraction quality were grown from a 1:1 mixture of protein solution (3.5 mg/ml), and reservoir solution (2.17% polyethylene glycol [PEG] 4000, 0.05 M Tri-sodium citrate, pH 6.5) in 4-μl hanging drops. Before being flash cooled in liquid nitrogen, the crystals were briefly transferred into a solution containing 15% PEG 4000, 0.05 M Tri-sodium citrate, and 30% glycerol.
Crystals were grown as previously reported in 8 to 12% PEG 3350–0.2 M sodium phosphate, pH 6.5 (30). Before being flash cooled in liquid nitrogen, the crystals were soaked for 30 min in a solution at pH 6.5 containing 12% PEG 3350, 0.05 M ammonium acetate, and 35% glycerol. This yielded a crystal form different from that previously reported, with two molecules in the asymmetric unit (form pO).
Alternatively, the crystals were soaked for 10 min in a solution at pH 6.5 containing 12% PEG 3350, 0.05 M ammonium acetate, and 35% glycerol and then transferred for 17 min into the same solution with 1.3 mM phosphatidylinositol 4,5-phosphate (C4) (Tebu-bio). This treatment yielded the same crystal form as previously published (form cO with one molecule in the asymmetric unit).
X-ray diffraction data were collected at beamlines Proxima 1 of the SOLEIL Synchrotron (St. Aubin, Gif-sur-Yvette, France) or ID14 of the European Synchrotron Radiation Facility (ESRF; Grenoble, France). Diffraction data were processed with the XDS package (15). Structures were rebuilt with COOT (11) and refined with phenix.refine (2) with TLS refinement (one TLS group per NS5B molecule), and restrained individual B factor and positional refinement. X-ray weights were optimized for the final rounds of restrained refinement.
J6_NS5B-V405I crystallized in the same crystal form as J6_NS5B despite the point mutation and different crystallization conditions (30). The J6_NS5B-V405I structure was refined by starting with wild-type (WT) J6_NS5B (Protein Data Bank [PDB] accession number 2XWH) with initial rigid-body refinement. The same test set was retained as for the J6_NS5B data set.
Similarly, our previous JFH1_NS5B structure in crystal form cO (PDB accession number 2XXD) was used as a starting point for refinement against the higher-resolution data set. The test set for 2XXD was kept and extended to the new high-resolution limit.
Molecular replacement was carried out with the Phaser program of the PHENIX suite (2) using 2XXD as a search model. Soft (torsion) NCS restraints were enforced in refinement between the two molecules in the asymmetric unit.
Differences in main chain conformation between pairs of molecules were objectively assessed with the ESCET program (31). As in our previous analysis (30), we used significance values (ESCET “lolim” parameter) of 2σ to take into account all significant conformational differences. Differences in side chain rotamers were collated using the PHENIX structure comparison utility. For each residue with its side chain reported in different rotamers in different structures, we checked that all rotamers were well defined in electron density.
Our previous structure-function comparison of JFH1_NS5B and J6_NS5B indicated that JFH1_NS5B is highly efficient for de novo RNA synthesis and that amino acid 405 is a key residue in this reaction (30). Indeed, changing WT V405 of J6 or Con1 NS5B to isoleucine, as in JFH1, enhanced NS5B-directed de novo RNA synthesis. We did not, however, determine whether the dinucleotide formation or the transition-to-elongation step was affected by this change. To analyze these two initial stages of RNA synthesis more precisely, we performed gel-based initiation-elongation assays that allowed quantification of the initiation dinucleotide alone or the initiation dinucleotide and the elongation product. As in our work with genotype 1 enzymes (14), we used templates corresponding to the 3′ end of the HCV negative-strand RNA (i.e., the sequence complementary to the 5′ NTR). We previously established that these RNAs are suitable for in vitro study of the replication of HCV RNA by NS5B, allowing initiation from the 3′-terminal nucleotide with Mg2+ as the only divalent cation (14) as expected for synthesis of the positive strand in vivo (8). Synthesis of the dinucleotide alone was determined by using a 341-nt RNA template derived from the 3′ end of a genotype 3 HCV (G3-U RNA, Fig. 1A). RNA synthesis was run for 2 h in the presence of the first 2 nt incorporated, the lack of GTP preventing the transition to the elongation phase (14). Data from such an experiment, illustrated in Fig. 1B, clearly showed that, consistent with our expectations, JFH1_NS5B is 10 times more efficient in this first step of RNA synthesis than J6_NS5B or Con1_NS5B (Table 1).
Introducing mutation V405I into NS5B of genotype 1 strain Con1 also improved Con1_NS5B de novo RNA synthesis and Con1 HCV replication in cells, although not to the point of infectious particle production (30). Therefore, we next checked for both J6_NS5B and Con1_NS5B whether, as predicted by our model, a change of residue 405 from valine to isoleucine would specifically improve dinucleotide formation efficiency. To exclude any spurious effect of V405I on NS5B stability in our assays, we quantified products after 1 h and 2 h of incubation (Fig. 1B). The amount of dinucleotide increased at the same rate for WT and V405I RdRp (2.1-fold between 1 h and 2 h for WT and V405I Con1_NS5B and 2.3-fold for WT and V405I J6_NS5B), demonstrating that, at least in this time period, V405I mutation did not affect enzyme stability.
As predicted, J6-V405I_NS5B indeed synthesized 2 to 2.8 times more dinucleotide than J6_NS5B (Fig. 1B and andCC and Table 1). However, mutation V405I had no effect on Con1_NS5B's dinucleotide formation efficiency (Table 1).
We next determined the effect of the V405I mutation on the transition-to-elongation phase. For that, RNA synthesis was performed with G1-C RNA and the four nucleoside triphosphates, with 3′dUTP instead of UTP to arrest RNA elongation at position 11 on the template (Fig. 1A). Results obtained in such an experiment clearly showed that the V405I mutation boosted the synthesis of the elongation product of Con1_ΝS5Β, as indicated by the appearance of the expected 11-nt elongation product of Con1-V405I-NS5B in Fig. 1D. As this mutation had no effect on dinucleotide synthesis, we may attribute this effect to an improved transition to elongation in Con1_NS5B. Synthesis of the elongation product was also boosted in J6_NS5B by mutation V405I (Fig. 1D). The effect on the transition level was evaluated by determining the ratio of elongation to dinucleotide products after correcting for the number of labeled nucleotides in the products (Fig. 2A; Table 2). With the various enzymes used in this study, we found that there was usually less than 1% and at most a few percent of elongation product per dinucleotide synthesized, consistent with the transition to elongation being the rate-limiting step in de novo RNA synthesis by HCV-NS5B (Table 2). We also found a slight but significant increase in the transition by J6-V405I_NS5B over that by the WT (1.6-fold, P = 0.02, Student test). Most striking, though, is the effect of the V405I mutation on transition by genotype 1 Con1_NS5B (Fig. 1D). In this case, the V405I mutation multiplied by 5 the transition from initiation to elongation (Fig. 2A; Table 2). Altogether, these data indicate that mutation V405I improves both early steps of de novo synthesis of J6_NS5B whereas Con1_NS5B is markedly improved but only at the transition-to-elongation step.
Among 207 NS5B sequences of different genotypes recovered from the European HCV database (10), 191 sequences harbor V405 whereas I405 was found in only 16 sequences, all of them from genotype 1 strains except one (JFH1). The NS5B of HCV J4, a strain known to be infectious in chimpanzees (38), is part of this small genotype 1 NS5B group, with a WT isoleucine at position 405. We decided to analyze the early steps of de novo initiation with this polymerase. As illustrated in Fig. 1D and quantified in Fig. 2A, WT J4_NS5B displays as high a transition efficiency as Con1-V405I_NS5B (Table 2). This puts it on a par in this respect with JFH1_NS5B (see below). To confirm the involvement of residue 405 in the transition step, we introduced an I405V mutation into J4_NS5B and analyzed its initiation and elongation products as described above. The I405V change did not modify dinucleotide synthesis by J4_NS5B (Fig. 1C and and2B),2B), but it decreased elongation product synthesis (Fig. 2C). Thus, we found a slight but significant decrease in transition efficiency with J4-I405V_NS5B compared to that of the WT enzyme (1.5-fold [P = 0.005, Student t test]; Fig. 2A).
In order to clarify the involvement of residue 405 in de novo RNA synthesis, we solved the crystal structure of J6-V405I_NS5B to a resolution of 1.9 Å (Table 3). This point mutant crystallized in the same crystal form as the WT J6_NS5B protein we previously refined to a resolution of 1.8 Å (30). Nevertheless, there are a number of significant differences in main chain conformation as detected by the ESCET program (31). ESCET is the tool of choice here, as it explicitly takes experimental uncertainty into account by first computing estimates of positional errors for all alpha carbons (Cα) based on the structures' precisions and the Cα's individual temperature factors, allowing objective assessment of small but significant displacements between two crystal structures. Analyses such as these previously allowed us to pinpoint small but functionally important conformational differences in JFH1_NS5B (34).
At these resolutions, the coordinate errors for Cα are reported by ESCET as 0.075 ± 0.037 Å for J6_NS5B and 0.101 ± 0.038 Å for J6-V405I_NS5B, respectively. Only 83% of the 561 Cα's present in both structures are then in identical positions within experimental error with a cutoff of 2σ. We can thus map rigid blocks in NS5B that have moved with the V405I mutation (32). Apart from a local displacement of some 1.8 Å for the main chain of residues 405 and 406 (in red in Fig. 3), there is a general opening of the thumb domain (mostly in green in Fig. 3) relative to the palm-and-fingers domain (in blue in Fig. 3). Thus, the main chain at the top of the thumb moves by 0.8 to 1.0 Å, values similar to the differences in thumb orientation (0.7 to 1.1 Å) between JFH1_NS5B and J6_NS5B (30). Residues 95 to 98 at the top of the fingers (in yellow in Fig. 3) opposite residue 405 in the thumb also open by a small but significant amount. There is thus a general loosening of the thumb-fingers interaction in this region.
V405 contributed to buttressing the interaction with residues 95 to 98 in J6_NS5B, albeit less solidly than I405 in JFH1_NS5B (34). However, in J6-V405I_NS5B, I405 retracts against the beta flap, where it inserts between the apolar parts of the side chains of Asn444 and Glu446 (Fig. 3B). This displacement induces long-range changes in key interactions of the thumb with the other parts of NS5B. The side chains of Asn406 and Ser403 both flip and contribute to the rearrangement of two networks of interaction. The first, main network (Fig. 4A) goes across the beta flap to the linker. An alternate network of hydrogen bonds (Fig. 4A, top) starts from Asn406 and goes through Asn444, Ser453, and Asn442 and to Val564 (while electron density ended at Ser563 for J6_NS5B). At the base of the beta flap, the main chain shifts at residues 457 and 458. Asp458 makes a salt bridge to Arg517 in both J6_NS5B and J6-V405I_NS5B. In the latter, however, Arg517 is displaced and no longer makes a second salt bridge to Glu541 in the helix at 540 to 544 of the linker. A further weakening of the interaction of this helix with the thumb is seen in the loss of part of the hydrogen bonding with Arg465. Altogether, these rearrangements explain the concerted displacement of the upper beta flap and the helix at 540 to 544 (in magenta in Fig. 3). Simultaneously with this remodeling of hydrogen bonding, the new hydrophobic interaction between I405 and Glu446 induces (Fig. 4A, bottom) a displacement of 0.8 Å of the lower segment, 445 to 448, of the beta flap (in dark brown in Fig. 3) and a switch of rotamers for the side chain of Met447, a side chain that packs against the helix at 550 to 554 of the linker. The two-part hydrophobic packing of the linker against the beta flap on one side and the fingers on the other is thus altered, with shifts of the side chains of Ile560 and Phe193.
The second network of interactions that undergoes a marked rearrangement goes from residue 405 toward the fingertips (Fig. 4B). Flipping of the side chain of Ser403 is accompanied by a shift of segment 399 to 401. On one side, interaction of Arg401 with Glu18 of the fingertips is weakened, with the loss of a strong salt bridge. This change is transmitted to the base of the fingertips through Glu17 and Thr42. On the other side, the side chain of Thr399 flips, inverting the polar γ1 oxygen for the apolar γ2 methyl. As a consequence, the hydrogen bond to His428 is lost and the close hydrophobic packing with Leu26 (at the very tip of the fingertips) and Ile432 is loosened. The shift in position of Ile432 is transmitted (Fig. 4C) to the other side of the same helix through Met434 and induces a change in rotamers for the side chains of Arg514 and Leu439. The latter switch contributes to the displacement of the main chain at residues Leu457 and Asp458. Thus, the two networks to the fingertips and to the linker are connected at this point and both contribute to the release of the helix at 540 to 544.
In summary, mutation V405I has many subtle but important consequences on the overall conformation of J6_NS5B. The net effect, as shown by a comparison of the two crystal structures, is a slight opening of the thumb with concomitant loosening of both polar and hydrophobic interactions with the fingers (residue 405 with residues 95 to 98), outer fingertips (residues 401 and 399/428/432 with residues 18 and 26), and linker (residues 465/517 and 447 with the helices at 540 to 544 and 550 to 554).
We previously reported that JFH1_NS5B exhibits 5- to 10-fold higher overall de novo synthesis activity in vitro than the J6 enzyme (34). We show here that this high polymerase activity relies on more efficient dinucleotide synthesis as predicted (Fig. 1B), but data presented in Fig. 5C and andDD show that JFH1_NS5B is also unusually efficient at the transition step. Indeed, JFH1_NS5B is much more efficient at these steps than the other WT enzymes except J4_NS5B (Table 2). As the transition implies removal of the linker from the catalytic cleft and the linker itself connects to the C-terminal transmembrane helix in vivo, the transition step in particular will be likely affected by the C terminus of NS5B. Thus, we tested whether the dinucleotide formation and transition properties of the Δ21 construct also apply to the full-length JFH1_NS5B. As shown in Fig. 5 and Tables 1 and and2,2, under the same assay conditions, JFH1_NS5B-FL still displays about half of the transition and about one-third of the dinucleotide formation efficiencies of JFH1_NS5BΔ21.
In the light of our observations on J6-V405I_NS5B, we next asked whether we could find structural determinants for improved transition in the conformational variability of WT JFH1_NS5B. Two structures are newly available to us for this purpose, with identical sequences but in different crystal forms (Table 3). The first is a higher-resolution structure in the same crystal form as previously reported (cO). The second new structure is in a new crystal form (pO) and harbors two molecules in the asymmetric unit. These two molecules are in the same conformation as reported by ESCET (although differences may be blurred by the soft noncrystallographic symmetry restraints we used in refinement). However, one of the molecules in pO has different crystal environments than the one in cO near the helix at 540 to 544 and displays alternate conformations in both the helix at 550 to 554 and the immediate surroundings of residue 405. As a result, near residue 405, JFH1-cO_NS5B closely matches J6-V405I_NS5B while JFH1-pO_NS5B comes closer in an alternate Asn406 conformation to J6_NS5B (Fig. 6A). However, the main chain of JFH1-pO_NS5B comes closer to that of both J6_NS5B and J6-V405I_NS5B near the helix at 550 to 554, while for JFH1-pO_NS5B only the alternate path around the side chain of Trp550 is present (Fig. 6B). In all structures, the side chain of Trp550 remains in the same position and contributes to keeping the linker tethered to the thumb (1), but its environment in the hydrophobic cleft between beta flap and thumb changes (Fig. 3A, bottom). Differences are propagated upstream of the linker, up to Glu541 (Fig. 6C). There, J6_NS5B is the outlier, with the other three molecules again becoming superimposable.
HCV is a human pathogen of major importance (13). Numerous studies have therefore been devoted to the development of anti-HCV drugs, and these efforts are now beginning to bear fruit (24, 25). For years, though, research on the HCV life cycle was hindered by the lack of robust cell culture systems. Con1 was the first strain for which replication of HCV RNA in cells was achieved and a replicon system was developed (18). The first systems allowing a complete HCV cycle in cells were finally obtained thanks to the discovery (16) and characterization (17, 36, 39) of the HCV JFH1 (Japanese fulminant hepatitis 1) strain. It soon became apparent that the special properties of JFH1 are due to its abnormally high RNA replication capabilities and especially to its polymerase (22). Subsequent studies sought to characterize JFH1_NS5B by systematic comparisons with the closely related J6 strain (23, 30, 34). These studies combined replication in cell culture, activity assays of recombinant proteins, and, in the case of our own work, X-ray crystallography. They established that the key difference between JFH1_NS5B and other NS5B proteins from normal strains is much higher de novo RNA synthesis activity on single-stranded RNA templates by JFH1_NS5B. Our structural analysis allowed us to pinpoint the JFH1 polymorphism I405 as a key element of this JFH1_NS5B property (34) and thus to engineer a chimera based on J6_NS5B but infectious in cell culture with the single amino acid mutation V405I (30).
In the present study, we compared the de novo RNA synthesis activities of NS5B with either V405 or I405. Precise dissection of the early steps of synthesis of the positive strand from a negative-strand template establishes that for strains JFH1, J6, and Con1 also, the most inefficient step in vitro is not synthesis of the initial dinucleotide but its subsequent use as a primer, i.e., transition to elongation. Unexpectedly, mutation V405I in J6_NS5B improves not only dinucleotide synthesis but also transition to elongation. The crystal structure of J6-V405I_NS5B reveals the molecular basis of the effect on transition. It shows that residue 405 may switch from its buttressing position of the thumb-fingers interaction, where it stabilizes a closed conformation of NS5B and hence dinucleotide formation. The new I405 position is associated with a cascade of rearrangements that lead to an opening of the thumb of the same magnitude as the initial difference between JFH1_NS5B and J6_NS5B. A general weakening of the interactions of the thumb with the linker and fingertips is thus observed. The regions involved were previously reported as important for linker release (1) and for transition to elongation (9). Indeed, in the latter study, the authors specifically identified the outer fingertips as a locking mechanism for the initiation conformation and residue 428 as involved in initiation but not primer extension (9). The changes we saw in J6-V405I_NS5B are thus those that are postulated as the first steps in the opening of NS5B upon the transition to elongation (6, 14). The fact that they occur in the same crystal form as that of J6_NS5B indicates that they are not the effect of different crystal environments but that V405I stabilizes this more open pre-elongation conformation. However, V405I also improves dinucleotide synthesis by J6_NS5B and therefore also stabilizes its more closed conformation, as predicted. We conclude that recombinant NS5B exists in solution as a spectrum of very similar conformations in dynamic equilibrium. At any time, most of the population is neither closed enough for dinucleotide formation nor open enough to reach the conformation conducive to transition. V405I thus manages to improve dinucleotide formation, transition, or both by stabilizing one or both of the relevant conformations at the expense of the population in intermediate conformations. In accordance with this view, there is a similarly narrow range of crystallized conformations of NS5B_Δ21 of genotype 1b, and nonnucleoside inhibitors are seen to bind to such intermediate conformations as those postulated here (6). Further support for this model is brought by our present results with genotype 1b NS5B of strains Con1 and J4. They show that relative to V405, I405 improves the transition of both enzymes without changing the dinucleotide formation efficiency, showing again that one step may be improved without deterioration of the other.
Transition efficiency is improved in both strains (J6 and Con1) for which mutation V405I single handedly boosts both de novo initiation assays and HCV RNA replication in cells (30). Furthermore, JFH1_NS5B is more efficient in transition than these enzymes. The comparisons of JFH1_NS5B structures in different crystal packing environments show that in the JFH1 enzyme, the changes around residue 405 also correlate with changes in the linker region. Important differences with J6_NS5B, however, are seen in that the region of the helix at 550 to 554 is the most variable in JFH1_NS5B, while it is more subtly involved in J6_NS5B. In contrast, a key regulatory event in linker release in J6_NS5B is shown here to be the loss of a salt bridge to the helix at 540 to 544. This salt bridge is never formed at all in JFH1_NS5B because of the functionally important (30) JFH1 K517 polymorphism at a position that is an arginine in most genotype 1 and 2 NS5B proteins, including those of J6, Con1, and J4. Thus, the coordination of linker movement and thumb opening (14, 30) may be differently regulated in different HCV strains.
Two important questions addressed by studies of the JFH1 strain are the molecular determinants of JFH1_NS5B's replication properties and, by contrast, the low level of RNA synthesis (or rather the low numbers at any time of replication-competent NS5B proteins ) in other HCV strains such as J6. The former question is of paramount importance for the development of HCV cell culture systems not dependent on the JFH1 replicase. The latter question is a basic science puzzle that goes far beyond the field of HCV research. Indeed, all viruses that synthesize their replicase together with structural proteins as part of a polyprotein, including all members of the family Flaviviridae, are faced with the problem of subsequently preventing unregulated RNA synthesis by the vastly overexpressed replicase. In HCV's case, this is achieved by keeping nearly all NS5B molecules inactive in the cell. In the replicon system, for instance, a huge excess of nonstructural proteins is required to build up functional viral replication complexes but only a few are actually involved in RNA synthesis at a given time point (26). The present work confirms that the molecular basis of this regulation is a double block of de novo RNA synthesis at the dinucleotide formation and transition-to-elongation steps. It establishes that both of these steps can be greatly improved in isolated recombinant NS5B, even with a single point mutation. Transition is clearly the limiting step in our in vitro assays, a result consistent with our previous work with genotype 1 enzymes and the same templates used here (14). In contrast, numerous studies using nonviral simpler templates and dinucleotide primers consistently found that NS5B can extend exogenously added dinucleotide primers more efficiently than single nucleotides in de novo initiation (12, 27, 40). This discrepancy may arise from a conserved stem-loop in the 3′ end of the negative strand that leaves the last 4 bases unpaired (33) and likely favors dinucleotide synthesis but needs to be melted for transition. It is likely that in the context of HCV infection, both dinucleotide formation and subsequent extension actually rely on signals from specific partners for efficient performance at the right time and place.
We thank Volker Lohmann for the gift of plasmids encoding JFH1_NS5B, J6_NS5B, J6-V405I_NS5B, and Con1_NS5B. We acknowledge the Structural Biology and Proteomics pole of the IMAGIF integrated platform (https://www.imagif.cnrs.fr/?nlang=en) for access to crystallization and mass spectrometry services and synchrotrons SOLEIL (beamline PROXIMA 1) and ESRF for generous allocation of beamtime.
This work was supported in part by Agence Nationale de Recherche sur le Sida et les Hépatites Virales grants to S.B. (AO 2010-1, CSS4) and M.V. (AO 2010-2). C.C.S. acknowledges an Association pour la Recherche sur le Cancer postdoctoral fellowship.
Published ahead of print 24 April 2012