Hepatitis C Virus NS5B and Host Cyclophilin A Share a Common Binding Site on NS5A*

Background: HCV replication requires the interaction of the viral polymerase NS5B with both viral and host proteins. Results: We performed the molecular characterization of the interactions between HCV NS5B, NS5A, and host CypA. Conclusion: HCV NS5B and host CypA share a binding site on HCV NS5A. Significance: The NS5A-D2 site, which interacts with both the HCV polymerase NS5B and the host CypA, might regulate the HCV replication. Nonstructural protein 5B (NS5B) is essential for hepatitis C virus (HCV) replication as it carries the viral RNA-dependent RNA polymerase enzymatic activity. HCV replication occurs in a membrane-associated multiprotein complex in which HCV NS5A and host cyclophilin A (CypA) have been shown to be present together with the viral polymerase. We used NMR spectroscopy to perform a per residue level characterization of the molecular interactions between the unfolded domains 2 and 3 of NS5A (NS5A-D2 and NS5A-D3), CypA, and NS5BΔ21. We show that three regions of NS5A-D2 (residues 250–262 (region A), 274–287 (region B), and 306–333 (region C)) interact with NS5BΔ21, whereas NS5A-D3 does not. We show that both NS5BΔ21 and CypA share a common binding site on NS5A that contains residues Pro-306 to Glu-323. No direct molecular interaction has been detected by NMR spectroscopy between HCV NS5BΔ21 and host CypA. We show that cyclosporine A added to a sample containing NS5BΔ21, NS5A-D2, and CypA specifically inhibits the interaction between CypA and NS5A-D2 without altering the one between NS5A-D2 and NS5BΔ21. A high quality heteronuclear NMR spectrum of HCV NS5BΔ21 has been obtained and was used to characterize the binding site on the polymerase of NS5A-D2. Moreover these data highlight the potential of using NMR of NS5BΔ21 as a powerful tool to characterize in solution the interactions of the HCV polymerase with all kinds of molecules (proteins, inhibitors, RNA). This work brings new insights into the comprehension of the molecular interplay between NS5B, NS5A, and CypA, three essentials proteins for HCV replication.

Hepatitis C virus (HCV) 5 is a small RNA virus from the Flaviviridae family. Since its discovery in the late 80s HCV has infected an estimated 130 -200 million people worldwide (1). HCV chronic infection is one of the main factors leading to liver cirrhosis and hepatocellular carcinoma and thus represents a serious public health challenge for almost all countries. HCV has been classified into six major genotypes (1)(2)(3)(4)(5)(6) according to its RNA genomic sequence. Among them, genotype 1 is the most prevalent all over the world except for Africa where genotype 4 represents nearly half of the infections (1). The standard of care to treat HCV infection consists of a combination of ribavirin and pegylated-interferon ␣ (2). This therapeutic strategy suffers both from a low efficacy in the case of genotype 1 infection and from non-negligible side effects for treated patients. The development of more efficient and safer new anti-HCV molecules is thus highly expected (3,4). HCV is an obligatory intracellular parasite and requires host cell factors in addition to its own components to achieve its complete life cycle (5). Consequently, anti-HCV drugs may be either directacting agents or host-targeting agents. Direct-acting agents have the advantage to be highly specific but may elicit resistance mutations selection, thereby limiting their long-time efficacy. The second category of molecules is thought to have a higher genetic barrier for resistance mutations but may interfere with host cellular processes. In both cases a better molecular characterization of the virus-virus and virus-host interactions mechanisms potentially will lead to more efficient anti-HCV drugs.
The HCV (ϩ)RNA genome encodes for a unique polyprotein precursor (ϳ3000 amino acids) that after processing by viral and host proteases results in 10 different proteins divided into structural (Core, E1, E2) and non-structural (p7, NS2, NS3, NS4A, NS4B, NS5A, and NS5B) proteins (5,6). NS3 to NS5B constitute the minimal set of viral proteins that assures the viral RNA replication; all are anchored into the endoplasmic reticulum host membrane to form the replication complex. NS3, which carries both RNA helicase and serine protease activities, is actually one of the main targets for direct-acting agent development, with two compounds (telaprevir and boceprevir) that were last year Food and Drug Administration-approved (4). In addition to NS3, NS5B constitutes another promising target for direct-acting agent development (2,7).
NS5B is the central protein (65 kDa) for the HCV replication process as it carries the RNA-dependent RNA polymerase enzymatic activity (8 -10). The last 21 residues at the C terminus of NS5B are required for endoplasmic reticulum membrane anchoring (11,12) and in vivo HCV replication (13), but they are not absolutely required for its enzymatic activity in vitro (14,15). NS5B exhibits a similar "right-hand" shape as other RNA-dependent RNA polymerases, with thumb, palm, and finger subdomains (16 -18). NS5B inhibitors are divided into nucleoside inhibitors (NIs), which bind into the active site and are terminator substrates of the RNA polymerase activity, and non-nucleoside inhibitors (NNIs) that bind into one of the allosteric sites of the polymerase. Numerous crystal structures of NS5B in complex with NNIs have led to the identification of at least four major NNI binding sites (for review, see Ref. 2). However, the conformational changes that follow NNI binding (19 -24) can interfere with proper crystal packing, thereby rendering crystallographic results difficult to interpret.
In vivo, HCV RNA replication occurs in a replication complex that contains endoplasmic reticulum-derived membrane(s), NS3 to NS5B viral proteins (25), and viral RNA as well as several host factors such as hVAP-33 (26), VAP-B (27), and cyclophilin A (CypA) (28). The latter is a major host factor required for HCV replication (29). It belongs to the cyclophilin (Cyp) family of proteins that bind cyclosporin A (CsA) and exhibit peptidyl-prolyl cis-trans isomerase activity (30). Cyps are able to catalyze the cis-trans isomerization of a peptide bond preceding a proline residue. Cyps have been identified as the main target of the early observed anti-HCV activity of CsA (31) or, more recently, of its non-immunosupressive derivatives (32)(33)(34). One of them, Debio-025 (alisporivir), is currently evaluated in a phase III clinical trial. Watashi et al. (35) initially proposed that cyclophilin B would be the target of CsA and proposed a mechanism whereby CypB would be a positive regulator of NS5B by modulating its RNA binding activity. Later, several members of the cyclophilins have been proposed to be involved into the HCV replication (29,31), and NS5B interactions with CypB (35)(36)(37)(38)(39) and CypA (28,29,36,40) have been reported. However, more recent evidence points to CypA as the major key player both for HCV replication and CsA sensitivity (29,41). The molecular mechanisms that confer to cyclophilin inhibitors an anti-HCV activity remain to be identified. In vitro selection for CsA-resistance mutations pointed out three viral non-structural proteins: NS2, NS5A, and NS5B as potential cyclophilin partners (37,41,42). Moreover, mutations located in the second half of NS5A confer the highest level of resistance to cyclophilin inhibitors (37,43).
NS5A is a 49-kDa phosphoprotein with RNA binding properties (44) that is required for HCV RNA replication and viral particle production (45)(46)(47) but for which up to now no enzymatic activity has been identified. The protein is involved in several steps of the HCV life cycle and interacts with almost all other non-structural proteins and with numerous host factors (48,49). NS5A is anchored at the cytoplasmic side of the endoplasmic reticulum membrane via an N-terminal helix (50) and comprises 3 cytosolic domains (D1, D2, and D3). NS5A-D1 carries most of the RNA binding properties of the protein. It has been shown that NS5A-D1 may adopt two different heterodimeric structures, at least in the context of a crystal (51,52). NS5A-D2 (250 -342), which is essential for RNA replication (47), and NS5A-D3 (356 -447), which is involved in the production and assembly of viral particles (45), have been shown to be natively unfolded (53)(54)(55)(56)(57). In previous studies we have shown that both NS5A-D2 (53) and NS5A-D3 (57) establish direct interactions with CypA in vitro and that these domains are substrates for the peptidyl-prolyl cis-trans isomerase activity of CypA. Using NMR spectroscopy, we have mapped onto NS5A-D2 the CypA interactions sites and found a correlation with CsA-resistance mutations or even mutations that impair RNA replication (53).
HCV NS5B and NS5A have been shown to interact together both in vivo and in vitro (58 -60). Conflicting results have been reported about the functional consequences of this proteinprotein interaction. Indeed, depending on the experimental setups used it has been shown that NS5A may either stimulate (59,61) or inhibit (59,62,63) the NS5B RNA-dependent RNA polymerase activity. At the molecular level, a strategy based on how NS5B mutations affect the NS5B-NS5A pulldown efficacy identified four discontinuous regions of NS5B (139 -145, 149 -155, 365-371, and 382-388) as essential for NS5A binding (60). Using truncated versions of NS5A, two regions of this protein (105-162 and 277-334) have been shown to be independently essential for the interaction with NS5B (59). Nevertheless, the absence of a structure of the complex leaves open many questions about the functional consequences of this interaction.
HCV NS5B, NS5A, and host CypA are three essential components of the viral replication complex and as such form three attractive targets for anti-HCV drugs. However, conflicting results have been reported as for the role of CypA for the proper association of these three proteins into the replication complex (28,40). Here we report NMR spectroscopy analyses at a per residue level of the molecular interactions between NS5A domains 2 and 3, NS5B, and CypA. We show that both HCV NS5B and host CypA share a common binding site on NS5A and use NMR of the NS5B enzyme to get a first localization of the NS5A-binding site on NS5B.
Surface Plasmon Resonance Analyses of the NS5B ⌬21 -NS5A-D2, CypA-NS5B ⌬21 , and CypA-NS5A-D2 Interactions-All materials and chemicals came from Biacore (GE Healthcare). Analyses were performed at 25°C on a Biacore 3000 system. HBS-P (10 mM Hepes, pH 7.4, 150 mM NaCl, 0.005% Surfactant P20) was used both as the running buffer at 10 l/min and as dilution buffer for analytes. To assess the NS5B ⌬21 -NS5A-D2 interaction, His-tagged recombinant NS5B ⌬21 was covalently coupled onto a Biacore CM5 (carboxymethylated) sensor chip using an amine-coupling kit (Biacore) according to the manufacturer's instructions. NS5B ⌬21 was immobilized at a concentration of 4 g/ml in a sodium acetate 10 mM buffer, pH 5, and at a 5-l/min flow rate of HBS. Covalent binding resulted in a signal of 2400 resonance units. To assess the CypA-NS5A-D2 interaction, the same procedure was applied to immobilized-CypA at 6 g/ml with a resulting signal of 2200 resonance units. A flow cell was submitted to a similar treatment but in absence of protein. This control flow cell was used as a control to assess bulk effects and nonspecific binding of the NS5A-D2. The NS5A-D2 was injected at concentrations ranging from 12.5 to 200 M in HBS-P at a 10 l/min flow rate during 4 min. Dissociation was detected during 11 min with 10 l/min flow of HBS-P. Regeneration was done by injecting 10 l of 10 mM NaOH with a 10 l/min flow rate. Results are represented as sensorgrams expressing the kinetics of NS5A-D2 binding (resonance units) to, respectively, immobilized NS5B ⌬21 and CypA. Experiments have been performed at least 3 times. To assess the CypA-NS5B ⌬21 interaction, NS5B ⌬21 was injected at concentrations ranging from 12.5 to 90 M in HBS-P supplemented with NaCl up to 300 mM at a 10 l/min flow rate during 4 min. Data were analyzed using BiaEvaluation 3.1 software, and both dissociation constant (K D ) and Req were calculated from kinetics parameters using method based on the Langmuir's 1:1 binding model.
NMR Data Collection on 2 H, 15 N-NS5B ⌬21 -NS5B NMR spectra were acquired at 305 K on a Bruker 900-MHz spectrometer equipped with cryogenic probe. All 2 H, 15

RESULTS
Interaction of NS5B with NS5A-HCV NS5B is the central enzyme for HCV replication, as it carries the RNA-dependent RNA polymerase activity (8 -10). NS5A is a multidomain protein that has been shown to be essential both for HCV replication and infectious particle production (45-47) but for which no precise molecular function has been evidenced. NS5B and NS5A are two central components of the replicase, where HCV RNA replication occurs, and have been shown to interact together. Here we use NMR spectroscopy to gain a very precise view at a per residue level of the direct molecular interaction between the domains 2 and 3 of NS5A (NS5A-D2 and NS5A-D3) and the HCV polymerase NS5B ⌬21 lacking the 21 C-terminal residues that correspond to its in vivo membrane anchor. All proteins have been produced in E. coli and then purified to almost homogeneity (see supplemental Fig. 1) as judged by Coomassie-stained SDS-polyacrylamide gels.
We previously assigned the spectra of the D2 and D3 domains of HCV NS5A when isolated in solution (53,56,57,67). To monitor by NMR spectroscopy the effects induced by NS5B ⌬21 on the spectrum of NS5A-D2 (or NS5A-D3), we mixed at equimolar concentrations (i.e. 50 M) 15 N-labeled NS5A-D2 (or NS5A-D3) with unlabeled polymerase and recorded 1 H, 15 N HSQC spectra. 1 H, 15 N chemical shift perturbations and/or signal broadening thereby should serve as indicators of a physical interaction. In the 15 N-NS5A-D2 spectrum in the presence of the polymerase, we observed numerous NMR peaks that were broadened up to near disappearance ( Fig. 1 and supplemental Fig. 4), but chemical shift perturbations were limited (see supplemental Figs. 2 and 5). Hence, in vitro NS5B ⌬21 establishes a direct physical interaction with NS5A-D2. Broadening of the NS5A-D2 peaks, which are usually very sharp due to the unfolded nature of this domain, is indeed expected upon interaction with the high molecular weight NS5B ⌬21 (65 kDa). Based on our previous assignment of the NS5A-D2 spectrum, we could plot on a per residue basis the ratio of the peak intensities in the absence or presence of NS5B ⌬21, and as such identified three different regions of NS5A-D2 that interact with the polymerase (Fig. 1, A and C, and supplemental Fig. 5). Two peptides of roughly 10 residues form the first anchoring points (residues 250 -262 (region A) and residues 274 -287 (region B)). A larger zone of interaction spanning residues 306 -333 forms the third zone of interaction (region C). Together, these three regions span roughly half of the NS5A-D2 domain, underscoring the extensive physical interaction between the D2 domain and NS5B ⌬21 . When looking at the physicochemical properties of the defined regions, we notice that the two first N-terminal ones are negatively charged, with three Asp plus one Glu and four Glu residues for regions A and B, respectively. Region C is different, as it carries besides the three negatively charged residues also four positively charged residues (four Arg). Although we previously showed that the D2 domain is essentially unfolded and that this property is conserved across HCV genotypes (53)(54)(55)67), we reanalyzed the experimental 13 C NMR chemical shifts corresponding to the C ␣ and C ␤ atoms of NS5A-D2 (Biological Magnetic Resonance Bank accession number 16165 (53)) with the SSP (secondary structure propensities) program (68). This novel algorithm detected a residual secondary structure in the three interaction regions of NS5A-D2 (Fig. 1, B and C). The SSP-positive score for region A corresponds to a helical tendency, the negative score for region B points to a ␤-strand or extended tendency, and region C shows a non-regular structural tendency with both positive and negative scores (Fig. 1B). This correlation suggests that the HCV polymerase interacts with the unfolded domain 2 of NS5A via its regions that have some intrinsic structural propensities.
As the NMR spectrum of 15 N-NS5A-D2 in the presence of NS5B ⌬21 suffered from excessive line broadening, thereby preventing a titration experiment, we used surface plasmon resonance (SPR) to measure an affinity constant for the complex. The purified NS5B ⌬21 was immobilized on a sensor chip and then increasing concentrations of NS5A-D2, from 12.5 to 200 M, were injected in the device (Fig. 2). Analysis of the corresponding sensorgrams showed that the signal response at equilibrium was initially dose-dependent and then saturated, which corresponds to a specific interaction between NS5A-D2 and NS5B ⌬21 . The Scatchard plot of these data is linear, and thus data are consistent with the Langmuir model (see supplemental Fig. 3). The measured association (k a ) and dissociation (k d ) rate constants allowed the determination of an equilibrium dissociation constant (K D ) of 21 ϫ 10 Ϫ6 M (see Fig. 2) for the interaction between NS5B ⌬21 and NS5A-D2.
We then used the same strategy to evaluate a potential interaction between the domain 3 of NS5A and NS5B ⌬21 . 15 N-Labeled NS5A-D3 was mixed in a 1:1 ratio (50 M) with unlabeled NS5B ⌬21 , and peak intensities of the 1 H, 15 N HSQC were compared with those of 15 N-NS5A-D3 in absence of the polymerase. In contrast with what has been observed for NS5A-D2, no significant peak broadening or even chemical shift perturbations were detectable on the NS5A-D3 spectrum in the presence of the polymerase (Fig. 3A and supplemental Fig. 2). We only observed a slight decrease for the peak intensities of 15 N-NS5A-D3 with NS5B ⌬21 that was nearly uniform all along the primary sequence of NS5A-D3 and that was probably due to an increased viscosity of the NMR sample when NS5B ⌬21 was added as the total protein concentration increased from 1 mg/ml (50 M 10-kDa NS5A-D3) to near 4.3 mg/ml (50 M concentrations of both NS5A-D3 and the 64.4-kDa NS5B ⌬21 ). These data showed that in vitro there is no direct molecular interaction between NS5A-D3 and the HCV NS5B ⌬21 and confirm the specificity of the previously detected NS5A-D2: NS5B ⌬21 interaction.
Interaction of NS5B with Cyclophilin A-As CypA is an essential host factor for HCV replication and as interaction with the NS5B RNA-dependent RNA polymerase has been repeatedly reported (28,29,36,40), we equally compared the 1 H, 15 N HSQC spectra of purified 15 N-CypA alone or in the presence of NS5B ⌬21 in a 1:1 ratio (50 M each). As for the individual domains of NS5A, we looked for chemical shift per-turbations and/or signal broadening in the CypA spectrum. We again detected the overall 10% reduction of peak intensities that we previously ascribed to the increased NMR sample viscosity ( Fig. 3B) but did not detect any region of CypA that underwent specific signal broadening when NS5B ⌬21 was added. The analysis of the 1 H and 15 N chemical shift perturbations also did not reveal significant modification of the CypA NMR spectrum in the presence of NS5B ⌬21 (see supplemental Fig. 2). These data hence show that in vitro there is no direct interaction between the HCV NS5B ⌬21 polymerase and the human CypA. We repeated the experiment with an excess of unlabeled NS5B ⌬21 over 15 N-CypA in a 5:1 molecular ratio (100 M to 20 M) but  15 N HSQC spectrum is shown as a black bar and has been normalized compared with the intensity observed for NS5A-D2 alone (C to E) or for NS5A-D2 in the presence of NS5B ⌬21 (F and G). The positions of proline residues are indicated by gray points as they do not have proton amide and thus do not have corresponding signal in HSQC experiments. In case of signal overlaps the intensities of the corresponding residues were divided by the number of peaks concerned. could still not detect any significant perturbation of the NMR spectrum (see supplemental Fig. 2). Even a weak affinity between both proteins thereby can be excluded. As our NMR results were in contradiction with previous reports in literature, this potential interaction was further investigated using SPR (supplemental Fig. 4B), but again we did not detect a specific interaction between NS5B ⌬21 and CypA in vitro.
Interplay between NS5B, CypA, and NS5A-D2-As we showed in this study that NS5A-D2 interacts with NS5B ⌬21 and as we had previously shown that NS5A-D2 also directly interacts with CypA (53), we set up an experiment with the three proteins present together. First, we analyzed the perturbations on the 1 H, 15 N HSQC spectrum of 15 N-NS5A-D2 induced by non-labeled CypA alone. In agreement with our previous findings (53) and confirmed very recently with a peptide-immobilized spot assay (69), two different regions of NS5A-D2 were broadened in the presence of CypA (Fig. 1D). The first comprises residues Gly-304 -Pro-323, whereas the second corresponds to residues Gly-337-Pro-341 located at the C-terminal end. One interaction region, centered at Pro-315, is hence shared between NS5B ⌬21 (Fig. 1C) and CypA (Fig. 1D), whereas the two N-terminal A and B regions are NS5B ⌬21 -specific (Fig.  1B), and the C terminus of the domain is solely recognized by CypA (Fig. 1D). The common binding region corresponds to the most conserved residues of NS5A-D2 over all genotypes, with five residues (Pro-310, Trp-312, Ala-313, Pro-315, and Pro-319) that are strictly conserved (see Hanoulle et al. (53) and further corresponds to the location where CsA-resistance mutations were identified (37,70). To further probe this common binding motif in NS5A-D2, we prepared a new NMR sample with the three different proteins in a 1:1:1 molecular ratio (i.e. ( 15 N-NS5A-D2:NS5B ⌬21 :CypA at 43 M each) and looked at the NMR spectral perturbations on NS5A-D2. In the HSQC 15 N-NS5A-D2 spectrum of the ternary mixture (Fig. 1E), we observed the sum of the individual effects caused by either the addition of NS5B ⌬21 (Fig. 1C) or the addition of CypA (Fig. 1D) alone. Indeed when the NS5A-D2 peak intensities in the experiment with both NS5B ⌬21 and CypA (Fig. 1E) were normalized to the ones when only NS5B ⌬21 (Fig. 1C) was added, we obtained the intensity profile (Fig. 1F) identical to the one whereby only CypA (Fig. 1D) was added. The addition of CsA, a CypA inhibitor, into the 15 N-NS5A-D2:NS5B ⌬21 :CypA ternary sample restored the pattern of the NS5A-D2 peak intensities to that when only NS5B ⌬21 was added (Fig. 1G). Thus, CsA inhibits the interaction of CypA with NS5A-D2 but has no influence on the molecular interaction between the polymerase NS5B ⌬21 and NS5A-D2. Our data hence do not favor the formation of a stable ternary complex between NS5A-D2, NS5B ⌬21 , and CypA but rather point to a system where NS5A-D2 can interact either with NS5B ⌬21 or with CypA with comparable affinities. For those NS5A-D2 residues in the Pro-306 -Glu-323 region that allowed the observation of a chemical shift difference, as they were not broaden beyond detection in the presence of NS5B ⌬21 and CypA, we found intermediate frequencies compared with   (1:1), then for each residue of these proteins the NMR intensity of its corresponding HSQC peak was measured and normalized to that in absence of the polymerase. Gray points indicate proline residues that do not have correlation signal in HSQC NMR experiment as they do not have amide proton. the ones in the spectrum with NS5B ⌬21 or CypA alone. Moreover, the minor peaks of the 1 H, 15 N HSQC spectrum of NS5A-D2, which correspond to residues in the vicinity of a proline in cis conformation (53), can be used as a further probe for CypA accessibility as they tend to disappear rapidly from the spectrum upon the addition of even a catalytic amount of CypA because of exchange broadening (53). Despite the fact that regions A (residues 250 -262) and B (residues 274 -287) of NS5A-D2 seem to be only involved in the binding of the polymerase (Fig. 1, C and D), we still observe an effect of CypA in these locations in the sample where both NS5B ⌬21 and CypA were present. Indeed, minor peaks in the A and B NS5B ⌬21 -specific regions still disappear from the NS5A-D2 spectrum when CypA is added (see supplemental Fig. 6), demonstrating that CypA can still access the regions A and B of NS5A-D2 in the presence of NS5B ⌬21 . These same minor peaks, assigned to residues in region B that neighbor a proline in cis conformation, are however, only weakly affected by the addition of NS5B ⌬21 (see supplemental Figs. 6 and 7), indicating in this case a transisomers specificity for the NS5A-D2-NS5B ⌬21 interaction.
NMR Analysis of NS5B-Numerous structures of NS5B in complex with nucleotides or with inhibitors have been solved by x-ray crystallography. However, molecular data that lead to the mapping of NS5B binding sites for interacting proteins and even for RNA molecules are mostly lacking. NMR spectroscopy on the purified polymerase NS5B ⌬21 could be useful in the characterization of such interactions, but it evidently faces the technical hurdle stemming from the high molecular weight of the protein (65 kDa). Recent technical developments, however, combining extensive protein deuteration and novel transverse relaxation optimized pulse schemes have shown promising results on systems of up to 1 MDa (71). We thus produced the double-labeled 2 H, 15 N-NS5B ⌬21 , and although the biomass of the bacteria grown in D 2 O minimal medium was reduced, we did obtain enough purified protein for several NMR experiments (see supplemental Fig. 1). The purified 2 H, 15 N-NS5B ⌬21 was kept almost 2 weeks in protonated buffer at 4°C to allow the amide deuterons to back-exchange with protons from water. Then we acquired a 1 H, 15 N-TROSY spectrum (66, 72) on a 90 M sample of 2 H, 15 N-NS5B ⌬21 using a high-field 900-MHz NMR spectrometer equipped with a cryogenic probe. The resulting heteronuclear NMR spectrum of the HCV polymerase was of high quality but evidently showed a large number of correlation peaks (Fig. 4). An automatic peak-picking procedure using TopSpin software (Bruker) led to the detection of nearly 490 peaks in the 1 H, 15 N-TROSY spectrum (Fig. 4). This number of peaks is lower than the 578 residues (including His tag) in the NS5B ⌬21 construct we used, mainly due to signals overlaps but still important regarding the coverage of the primary sequence of the polymerase, i.e. more than 80%. The assignment of the NS5B ⌬21 spectrum is ongoing in our laboratory but will require intensive biochemistry (through different labeling strategies) and NMR spectroscopy optimization. Nevertheless, as a first application of the spectrum, we tested if the interaction with NS5A-D2 could now be monitored directly on the 2 H, 15 N-NS5B ⌬21 NMR spectrum. We hence mixed 2 H, 15 N-NS5B ⌬21 with unlabeled NS5A-D2 at equimolar concentrations (70 M) and compared the resulting 1 H, 15 N-TROSY spec-trum with that of isolated 2 H, 15 N-NS5B ⌬21 . In the presence of NS5A-D2, several peaks in the 2 H, 15 N-NS5B ⌬21 spectrum underwent line broadening (Fig. 5, A and B, and supplemental Fig. 9A), thereby confirming our previous findings on a direct physical interaction between NS5B ⌬21 and NS5A-D2. Only a subset of peaks in the 2 H, 15 N-NS5B ⌬21 spectrum was affected, meaning that the NS5A-D2-binding site on the polymerase is rather localized and arguing against important conformational changes that would occur upon NS5A-D2 binding. To get information about the NS5A-D2-binding site on the HCV polymerase but in the absence of the assignment of the TROSY spectrum, we used an indirect strategy. We first analyzed the NMR spectral perturbations that were induced on the 2 H, 15 N-NS5B ⌬21 spectrum after the addition of a TCA inhibitor (Fig. 5, C and D, and supplemental Fig. 9B) for which the binding site onto NS5B is known as its structure in complex with the polymerase has been solved by crystallography (20, 65) (see supplemental Fig. 10). Then we compared these perturbations with the ones that were induced in the 2 H, 15 N-NS5B ⌬21 spectrum upon the addition of NS5A-D2 (Fig. 5, A and B). Some spectral perturbations on the 2 H, 15 N-NS5B ⌬21 spectrum were common upon the addition of both NS5A-D2 and TCA inhibitor, whereas others were specific to one or the other binding partner. These data suggest that the NS5A-D2-binding site on NS5B ⌬21 overlaps at least partially with that of the TCA inhibitor. NMR of NS5B might hence represent a highly valuable tool to study interactions mechanisms of this RNA-dependent RNA polymerase with different ligands or proteins.

DISCUSSION
The replication of HCV together with the polyprotein processing constitute main targets for development of anti-HCV drugs. As the replication process requires the formation of a membrane-associated multiprotein complex, its impairment can be done by hitting several individual targets, which can be viral or even host factors. The drugs that are currently in development and that act through the inhibition of HCV replication target either the HCV RNA-dependent RNA polymerase NS5B, NS5A, or CypA, an essential host factor. These three proteins are essential components of the replicase and have all been reported to interact two by two. Here we characterized these molecular interactions at a per residue level by NMR spectroscopy (Fig. 6).
Through the analysis of the chemical shift perturbations in the NS5A-D2 NMR spectrum that were induced after the addition of unlabeled NS5B ⌬21 , we identified three regions in NS5A-D2 that are involved in the interaction with the polymerase (see Fig. 1, A and C). These NS5B-binding sites correlate with the regions in the unfolded domain 2 of NS5A that exhibit some residual secondary structure propensity (see Fig. 1, B and C). NS5A-D2 thereby joins the growing list of unfolded proteins that interact with relevant biological partners through their residual structural elements (73,74). Using the same NMR strategy, we showed that the unfolded domain 3 of NS5A does not contribute to the binding of NS5B (see Figs. 3A and 6). Our results thereby complete the ones from Shirota et al. (59) who had performed a rough identification of the NS5B-binding site on NS5A using internal deletion mutants. They identified in genotype 1b two discontinuous regions of NS5A, residues 105-162 and 277-334, that are both required for proper interaction with NS5B. NS5A residues 277-334 in genotype 1b correspond to residues 273-330 in the genotype 2a we used in this study. As it has been later shown that region 105-162 belongs to the folded domain 1 of NS5A, its deletion may abolish the overall three-dimensional structure of this domain and render the pre-vious result difficult to interpret. The second described region 273-330 (genotype 2a numbering) encompasses two different binding regions, 274 -287 and 306 -333, that we identified at a per residue level in this work (see Fig. 1C). Shirota et al. (59) have reported that the regions 105-162 and 273-330 (genotype 2a numbering), located into domain 1 and domain 2 of NS5A, respectively, are independently essential for the binding of NS5B. We do not confirm this conclusion, as we showed that the isolated domain 2 of NS5A (residues 248 -341) is sufficient to interact with the polymerase. However, our NMR assay does not allow distinguishing whether regions A to C of NS5A-D2 interact at the same time with the same NS5B molecule or if individual regions A, B, and C interact at distinct moments with different NS5B molecules. Nevertheless, the Scatchard plot of the SPR data is consistent with a 1:1 Langmuir model (see supplemental Fig. 3). Considering the conservation level of the NS5A-D2 primary sequence among all HCV genotypes, the NS5B-interacting regions A, B, and C correspond to regions with less variability and even contain some strictly conserved residues (Ala-257, Asn-258, Asp-278, Pro-310, Trp-312, Ala-313, Pro-315, Pro-319, Trp-325 and Tyr-330) (see supplemental Fig. 11 and Ref. 53). This indicates that the three regions we identified in NS5A-D2 are functionally important for the virus; one possibility is that these regions are required for the proper interaction of NS5A-D2 with NS5B and hence for an efficient replication mechanism. The NS5A-D2 region (273-330, genotype 2a numbering), identified by Shirota et al. (59) to be required for NS5B binding, was further characterized in two subregions. Alanine substitutions in residues 273-305 (genotype 2a numbering) partially impaired the level of replication, whereas residues 306 -330 are essential as their substitutions led to absence of RNA replication (58). However, the absence of replication when NS5A-D2 residues 306 -330 were alanine-substituted may not be directly related to an impaired NS5A-NS5B interaction but may also be linked to the functional interaction with host CypA. Indeed we showed here that NS5A-D2 residues G-304 -Glu-323 and G337-Pro-341 constitute CypA-binding sites (see Fig. 1D). This result is identical with what we initially described (53) and is coherent with the tandem cyclophilin-binding site of NS5A that has recently been reported by Grisé et al. (69). We show here that HCV NS5B and host CypA share a common binding site on NS5A that contains residues Pro-306 to Glu-323 (see Fig. 1, C and D). Caution in the interpretation of replication assays with mutations and/or deletions in this particular NS5A region is thus required, as the observed effect may arise either from a modification of the NS5A-NS5B interaction and/or of the NS5A-CypA interaction (Fig. 6). We did not detect any direct physical interaction between CypA and NS5B (Fig. 6), in contradiction with several previous reports (28,29,36,40). The absence of chemical shift perturbations in the CypA spectrum upon the addition of NS5B (see Fig. 3B) was confirmed by the reverse experiment whereby the NMR spectrum of NS5B did not appreciably change upon the addition of CypA (data not shown). Although this apparent discrepancy may arise from the different constructs or genotypes that were used, a more plausible hypothesis is that the described CypA-NS5B interaction was indirect and mediated by a third interacting partner. Indeed, to our knowledge no studies have been done in vitro with only the two purified proteins but rather with cell lysate pulldown experiments (29,40), crude replication complex preparations pulldown experiments (28), or in in vitro translation reactions (36) that at least contain RNAs. Without interaction between NS5B and CypA, the interplay between HCV NS5B, NS5A, and host CypA, three essential viral replicase components, may be driven by the common binding site on NS5A-D2 that we identified. When we mixed NS5A-D2 with both NS5B ⌬21 and CypA in the same tube, we observed in the NS5A-D2 NMR spectrum the sum of the spectral perturbations (see Fig. 1E) that were individually observed by the addition of either NS5B ⌬21 (see Fig. 1C) or CypA (see Fig. 1D). Our NMR results are in favor of a model where NS5A-D2 would interact with similar affinities either with CypA or with NS5B (Fig. 6). We previously estimated by NMR spectroscopy the affinity corresponding to the CypA-NS5A-D2 interaction at 64 M (53), and in this work, using SPR, we measured a K D of ϳ18 M (see supplemental Fig. 4A). The fact that these two values were determined by different methods in solution or at surface may explain the relative variation. For the NS5B-NS5A-D2 interaction, using SPR we measured a K D of ϳ21 M (see Fig. 2). Thus the interaction strengths between NS5A-D2 and either CypA or NS5B are indeed quite similar. Moreover, we show that CsA only disrupts the molecular interaction between CypA and NS5A-D2 without interfering with the NS5A-D2-NS5B interaction (Fig. 6). Grisé et al. (69) recently described that they have preliminary data indicating the same findings. Chatterji et al. (75) have shown that CsA treatment has no influence on the association of NS5A and NS5B with the replication complex but rather depletes it from CypA. Their findings and our data are in favor of a model where the association of NS5A and NS5B into the replication complex would be mediated by direct interaction between these two viral proteins and that CypA incorporation in this functional complex would be done via its interaction with domain 2 of NS5A. This model is strengthened by our observation that CypA and NS5B have different cis-trans isomer specificities for NS5A-D2 binding in region B (see supplemental Figs. 6 and 7).
In this study we report the first high quality heteronuclear NMR spectrum of HCV NS5B (see Fig. 4 and supplemental Fig.  8) and show that this spectrum can be used to monitor molecular interactions directly onto the polymerase (see Fig. 5 and supplemental Fig. 9). Indeed we detected shifts and/or broadenings for a limited number of peaks in the NMR spectrum of a double-labeled 2 H, 15 N-NS5B ⌬21 in the presence of unlabeled NS5A-D2 (see Fig. 5, A and B, and supplemental Fig. 9B). These effects, which correspond to fast exchange, are in agreement with the K D of ϳ21 M that we measured by SPR. The limited number of peaks that were affected indicates that NS5B does FIGURE 6. NMR analyses of the interactions between HCV NS5A and NS5B proteins and host CypA. NS5A (in light gray) is shown as a dimer anchored to the cytoplasmic side of the endoplasmic reticulum (ER) membrane via a N-terminal amphipathic helix (AH). Domain 1 of NS5A (D1) is shown as a schematic from the PDB entries 1ZH1 and 1R7C. The disordered domain 2 and 3 of NS5A are represented by manually drawn lines. NS5B (in black) is shown as a schematic from the PDB entry 3I5K. Its C-terminal membrane anchoring helix is represented by a 20-residue ␣-helix. Human CypA and CsA are, respectively, shown as a schematic and sticks from PDB entry 1CWA. The direct in vitro molecular interactions between these proteins were investigated by NMR spectroscopy and are summarized in this figure. NMR-detected interactions are represented by solid arrows, whereas the absence of interactions are represented by dotted lines. The interactions that were found to be inhibited by CsA are highlighted with a ᮎ symbol. not undergo a global conformational change upon NS5A-D2 binding. This possibility has often been proposed to explain the effect, positive or negative, of NS5A-D2 on the RNA-dependent RNA polymerase enzymatic activity of NS5B (61,62). Currently, we have only started to assign the NMR peaks of NS5B ⌬21 , but it is already clear that even a partial assignment will require intensive biochemistry and NMR spectroscopy efforts. Nevertheless, we used an indirect strategy and showed that the NS5A-D2-binding site on NS5B partially overlaps with the binding site of a TCA inhibitor (see Fig. 5 and supplemental Fig. 9), which is known to bind into the thumb site 2 of the polymerase (2, 20, 65) (see supplemental Fig. 10). NMR hence should allow the direct mapping of the residues involved into various molecular interactions without using deletion or resistance mutants that may interfere with the global structure of the polymerase. Direct solution NMR spectroscopy on NS5B ⌬21 should also allow the study of compounds that induce relatively large conformational changes upon binding and that may interfere with crystal packing (19,20,23). Finally, the NMR spectrum of NS5B ⌬21 that we obtained may be used to characterized low affinity (over the micromolar range) interactions mechanisms as is the case with NS5A-D2.
In conclusion, we report here the first direct high resolution characterization of the molecular interactions between HCV NS5A, NS5B, and host CypA. We show that NS5B and CypA share a binding site on NS5A-D2 and that CsA only disrupts the NS5A-D2-CypA interaction. Further studies will be required to make a link between the interactions we observed in vitro and their potential effects on the RNA-dependent RNA polymerase activity of the HCV polymerase. The NMR spectrum of HCV NS5B ⌬21 furthermore constitutes a powerful tool to investigate its potential interactions with various interaction partners such as other proteins, small inhibitors, or even RNAs.