Cyclophilin A allows the allosteric regulation of a structural motif in the disordered domain 2 of NS5A and thereby fine-tunes HCV RNA replication

Implicated in numerous human diseases, intrinsically disordered proteins (IDPs) are dynamic ensembles of interconverting conformers that often contain many proline residues. Whether and how proline conformation regulates the functional aspects of IDPs remains an open question, however. Here, we studied the disordered domain 2 of nonstructural protein 5A (NS5A-D2) of hepatitis C virus (HCV). NS5A-D2 comprises a short structural motif (PW-turn) embedded in a proline-rich sequence, whose interaction with the human prolyl isomerase cyclophilin A (CypA) is essential for viral RNA replication. Using NMR, we show here that the PW-turn motif exists in a conformational equilibrium between folded and disordered states. We found that the fraction of conformers in the NS5A-D2 ensemble that adopt the structured motif is allosterically modulated both by the cis/trans isomerization of the surrounding prolines that are CypA substrates and by substitutions conferring resistance to cyclophilin inhibitor. Moreover, we noted that this fraction is directly correlated with HCV RNA replication efficiency. We conclude that CypA can fine-tune the dynamic ensemble of the disordered NS5A-D2, thereby regulating viral RNA replication efficiency.

is bound to the ER membrane via an N-terminal amphipathic helix (16), is composed by a folded cytoplasmic domain (NS5A-D1) (11) and two intrinsically disordered domains (NS5A-D2 and -D3) (17)(18)(19) (Fig. 1a). No enzymatic activity has been identified for any of these domains. Different homodimeric structures of NS5A-D1 (11,20,21) suggested it might be implicated in RNA-binding (22). Importantly, this domain is the molecular target for direct-acting antivirals (DAAs) against NS5A (23,24). NS5A-D2 and -D3 are intrinsically disordered, and thus exist as a dynamic ensemble of conformers. NS5A-D2 is required for viral RNA replication (10), whereas NS5A-D3 is involved in viral particle production and assembly (12). We and others have shown that NS5A-D2 and -D3 interact with and are substrates of human cyclophilin A (CypA) (25)(26)(27), a peptidylprolyl cis/trans isomerase (PPIase) that is an essential host factor for HCV replication (28). Detailed analysis assigned the main CypA-binding site to a small region (about 20 amino acid residues) in NS5A-D2 that contains 5 strictly conserved residues across the seven HCV genotypes, among which are 3 proline residues (25,29). An even shorter peculiar structural motif, called PW-turn ( 314 PXWA 317 in the Con1 strain, genotype 1b), which is essential for HCV RNA replication was identified in this region (30) (Fig. 1a). Disruption of the PW-turn weakens the molecular interaction between NS5A-D2 and CypA, but is required for efficient prolyl cis/trans isomerase activity of CypA toward the 313 Met-Pro 314 bond, contrary to the other prolyl bonds of the domain (30). The interaction between NS5A-D2 and CypA can be inhibited by cyclosporin A (CsA) or its nonimmunosuppressive analogs such as alisporivir, SCY-635, or NIM-811, or by other small-molecule cyclophilin inhibitors (31)(32)(33)(34). These compounds thereby act as host-targeted antivirals (HTAs). NS5A hence can be equally targeted by DAAs (via NS5A-D1) or HTAs (via NS5A-D2 through its interaction with CypA). Of note, studies of HCV NS5A and CypA paved the way for the use of CypI for other viral infections such as human coronaviruses (e.g. MERS or SARS) that also require CypA for their replication (35).
The role(s) of CypA in HCV replication and the mechanism of action of cyclophilin inhibitors (CypIs) are complex and still not completely understood. CypIs bind host CypA in a single hydrophobic pocket that contains both its enzymatic PPIase and binding activities, and thereby disrupt both the molecular interaction between CypA and NS5A-D2 as well as the putative CypA PPIase activity toward selected prolyl bonds in NS5A-D2 (25,36,37). Therefore, all CypA catalytic mutants are also impaired in their binding properties. Several laboratories have demonstrated that CypIs inhibit the de novo formation of the DMVs, the HCV-induced membrane structures that hold the replication complex (38,39). A remodeling of the ER in the presence of CypIs has also been observed in HCV-infected cells (40). Most of the identified CypI-resistance mutations, such as D320E, D320E,Y321N, and Y321H (28,41,42), are located in the immediate proximity of the PW-turn in NS5A-D2. These mutations confer a moderate resistance (ϳ2-4 -fold) to CypIs and hence a partial CypA independence, but do not abolish the interaction between CypA and NS5A-D2 (27,43,44). A better understanding of the functional role(s) of both CypA and NS5A, and specifically of the disordered domains of the latter, in the HCV life cycle could shed light on the underlying molecular mechanisms of this resistance.
Intrinsically disordered proteins (or regions) (IDPs/IDRs) are functional despite not having a stable 3D structure (45,46). They are best described as dynamic ensembles of interconverting conformers. Their biological functions are usually related to their capacity to interact with numerous partners, with a high specificity often related to low affinity (even if subnanomolar affinities have been reported (47)). As a consequence, they are often described as molecular hubs. IDPs/IDRs, despite any enzymatic activity, are nonetheless involved in cell signaling and regulatory processes, which can be physiological or pathological. Indeed, their implication has been demonstrated in numerous human diseases, including cancer, neurodegeneration, diabetes, and viral infections (48). IDPs/IDRs can interact with a biological partner while remaining disordered and form fuzzy complexes (49), but they can also establish interactions using short peptide motifs, including short linear motifs (50), molecular recognition features (51), or pre-structured motifs (52), and then fold upon binding. The biological functions of IDPs/IDRs are further regulated by alternative splicing and post-translational modifications (PTMs), including phosphorylation, ubiquitination, and glycosylation (53). These PTMs can modulate the structural and conformational properties of IDPs/IDRs and thereby break or promote interactions. IDPs are also enriched in proline residues (54), but whether these have any functional role has not yet been described in a conclusive manner. More recently, the concept of allostery has been expanded from its original paradigm (55) and allosteric regulations have been described in disordered protein (56,57). It has been shown that allosteric perturbations (ligand binding, PTMs, mutations) can change the functional properties of IDPs/IDRS by remodeling their energy landscape or conformational ensemble (i.e. population shift). This extreme complexity allows IDPs/IDRs to exert a fine modulation of biological processes (58).
Here, we combine nuclear magnetic resonance (NMR) spectroscopy with molecular dynamics (MD) simulations to decipher the role of the proline residues on the structural disorder in HCV NS5A-D2, and link it to its functional consequences. The detailed conformational behavior of NS5A-D2 (JFH1 strain, genotype 2a) with its PW-turn structural motif centered on 310 PAWA 313 confirms its pan-genotypic importance for HCV. Using the Trp 312 resonance as a probe, we show that the NS5A-D2 region encompassing the PW-turn exhibits a peculiar dynamic behavior whereby it may adopt, at least 2 conformational states, one structured with the PW-turn conformation fully formed and one totally disordered. These two states are in the fast-exchange NMR regime. Unexpectedly, we find that the cis/trans isomerization state of the 5 surrounding proline residues affects the ratio of conformers that adopt or not the PW-turn motif in the NS5A-D2 ensemble. Hence, this structural motif is allosterically regulated by proline cis/trans equilibria. CypA, through catalyzing the interconversion for distinct prolines, can thereby connect these different ensembles on a subsecond time scale. Importantly, we find that the CypI-resistance mutations in NS5A-D2 correspond to similar allosteric perturbations as they favor ensembles with a Conformational regulation of NS5A-D2 by CypA decreased structured PW-turn population. Using a cell-based assay, we show that the population of the structured PW-turn motif in the ensemble directly correlates with HCV RNA replication efficiency. Our results reveal a complex mechanism in which CypA modulates NS5A-D2's function.

NMR characterization of NS5A-D2 and its PW-turn motif
The 1 H, 15 N-HSQC NMR spectrum of NS5A-D2 of the HCV JFH1 strain (genotype 2a) (Fig. 1b) displays a narrow 1 H chemical shift dispersion that confirms the high level of intrinsic disorder in this domain (Fig. S1a) (18). The secondary structure propensity analysis, based on the 13 C chemical shifts, indicates the presence of several residual ␣-helices and extended regions in the N-terminal half (Fig. 1c). This is similar to the observations made on NS5A-D2 from HCV strains HC-J4 and Con1 (both genotype 1b) (30,59). Comparison of the experimental 1 H and 15 N chemical shifts of NS5A-D2 (JFH1) with those predicted for a truly disordered protein (60) highlights unexpected values for the tryptophan 312 (Trp 312 ) residue in the C-terminal part of the domain ( Fig. 1d and Figs. S1 and S2). A similar observation was made for the equivalent Trp 316 residue in NS5A-D2 of the HCV Con1 strain and was attributed to the presence of the PW-turn structural motif in the most conserved region of NS5A-D2 (30). The primary sequence of the PW-turn in the Con1 strain is 314 PIWA 317 , whereas it is 310 PAWA 313 in JFH1. In this motif, all positions but the second are strictly conserved in all HCV genotypes. We used a 20-mer peptide (pepD2-WT) to obtain an atomic description of this region in NS5A-D2 (JFH1) (Fig. S3). All the NMR data ( 1 H, 15 N, 13 C chemical shifts, NOE contacts, as well 3 J HN-H␣ couplings) were used to calculate a NMR structural model of this PW-turn ( Fig.  2 and Fig. S4a). The local r.m.s. deviation values suggest that beyond the core of the PW-turn (i.e. 310 PAWA 313 ), where the ring of Pro 310 and the aromatic side chain of Trp 312 are engaged in a CH/ interaction (61), the residues at its C terminus also adopt a certain degree of order (Fig. 2b). The interaction of the PW-turn motif, and specifically the Trp 312 residue, with the Tyr 317 side chain and the rings of Pro 319 and Pro 320 (Fig. 2c) agree with a similar extension of the motif as described in the peptide when bound to the MOBKL1B protein (62). Conformations of pepD2-WT, free in solution (this study), are closely related to its structure in the crystallographic complex with MOBKL1B (Fig. S4b).
The plot of the 1 H, 15 N-HSQC peak intensity along the NS5A-D2 sequence reveals that the resonances in the N-terminal moiety are narrower than those in the C-terminal half (Fig.  1e). 15 N spin relaxation data on the full-length domain confirmed this, with higher R 2 values for the 304 -321 region of NS5A-D2. Heteronuclear NOE values in this region are also clearly positive, whereas they are negative or close to zero for most residues in the N-terminal half of the fragment. We hence conclude that this region, encompassing the PW-turn and corresponding to the host CypA-binding site (25), is characterized by an increased rigidity while simultaneously experiencing exchange broadening (63) (Fig. 1, f-h). We next measured the residual dipolar couplings (RDC NH ) of the individual HN vec-tors in NS5A-D2 using a partially oriented sample (Fig. 1i). The residues in the region 304 -321 display RDC NH values higher than expected for an IDR (64). The highest value (32.8 Hz) was observed for the Trp 312 residue in the PW-turn motif. The RDC NH values from the Trp 312 and Trp 325 side chains were 40 and 24 Hz, respectively (data not shown). Altogether, these data confirm that the CypA-binding site contains a structured PWturn motif with nevertheless a peculiar structural dynamics.
We have previously reported that the I315G mutation in a peptide derived from NS5A-D2 of the Con1 strain precludes the presence of the PW-turn motif without making any change on the conserved residues. To investigate the functional role of the PW-turn motif in the JFH1 strain, we introduced the analogous mutation, A311G, in NS5A-D2. We found that this mutation also efficiently disrupts the PW-turn in NS5A-D2 because in the different NMR spectra acquired on this NS5A-D2 mutant, the proton amide and 13 C␣ resonances of Trp 312 and Ala 313 move toward their expected frequencies for a fully disordered region (Figs. S1b and S5). The NS5A-D2 A311G mutant hence provides a way to assess the structural and functional role(s) played by the PW-turn motif in the CypA-binding site.

Proline conformations induced a linear chemical shift pattern for Trp 312
Upon closer examination of the 1 H, 15 N-HSQC of NS5A-D2 WT, we identified up to six resonances that could be assigned to the same Trp 312 residue (Fig. 3a). Based on the peak intensity, there are one major (W312_1) and five minor (W312_2 to W312_6) Trp 312 resonances. As the Trp 312 residue is surrounded by 5 different proline residues in NS5A-D2, we hypothesized that this conformational heterogeneity with slow exchange on the time scale imposed by the chemical shift differences between the individual resonances could correspond to the cis/trans equilibria of these individual proline residues. Evidence was found in the 1 H, 15 N, 13 C 3D experiments used for the assignment of the fragment. Starting from the proton amide W312_6 resonance and using the 13 C chemical shifts (C␣, C␤, and CO), we connected it, in a sequential fashion, to a minor resonance of the Ala 311 residue (A311_6), which is itself linked to a residue with 13 C chemical shifts typical of a cis-proline (62.4 and 34.3 ppm for C␣ and C␤, respectively) (65) (Fig. S6). Thus, the minor W312_6 resonance originates from the cis conformation of the Pro 310 residue. Likewise, the other minor Trp 312 resonances are also related to the cis conformation of proline residues. Indeed, in a 1 H, 15 N-heteronuclear zz-exchange experiment (66) in the presence of a catalytic amount of CypA, we identified exchange cross-peaks connecting each of the minor Trp 312 resonances (W312_2 to W312_6) to the major one (W312_1) (Fig. 3b). To link each of the minor Trp 312 resonances to a particular proline residue, we compared the 1 H, 15 N-HSQC spectra of NS5A-D2 Pro to Ala mutants (NS5A-D2 P306A, P310A, P315A, P319A, and P320A) with that of the WT construct (Figs. S7-S11) (67). In each mutant spectrum, a minor Trp 312 resonance is missing. We therefore could unambiguously assign the minor Trp 312 resonances in the 1 H, 15  All of these Trp 312 resonances fall along a line and form a linear chemical shift pattern (Fig. 3). A similar spectral behavior was observed for the residue Ala 313 (Fig. S12), which is also part of the PW-turn motif. A linear pattern of chemical shifts was previously interpreted as a proof that this residue exists in (at least) two conformational states with distinct chemical shift environments in fast exchange on the NMR time scale (68). Even if the W312_6 and W312_1 resonances do not exactly match with the two pure states of this fast-exchange system, these resonances can be used as proxies for the two conforma-tional states at the extremity of the linear chemical shift pattern. Considering the frequency difference between the W312_1 and W312_6 resonances (⌬ 1 H ϳ420 Hz, ⌬ 15 N ϳ300 Hz), the time scale of this exchange is faster than 1-0.1 ms. Moreover, the observed line broadening or equivalent enhanced R 2 rates for this residue suggest that the exchange is close to this microsecond-millisecond time range.

Allosteric regulation of the PW-turn motif by proline cis/trans equilibria
To further characterize the PW-turn in each of these distinct conformers, we analyzed the same NMR parameters as for the  Figure 1. NMR structural dynamics of NS5A-D2. a, schematic representation of HCV NS5A protein anchored to the cytoplasmic side of the ER membrane via a N-terminal amphipathic helix (AH). Its folded domain 1 (NS5A-D1) is shown as a white hexagon, whereas its disordered domains 2 and 3 (NS5A-D2 and -D3) are represented by white rectangles. The CypA-binding site, which comprises the PW-turn motif (dashed black line), is shown as a gray area. b, NS5A-D2 amino acid sequence (residue 248 -342) from HCV JFH1 strain (genotype 2a). Below is similarity deduced from the alignment of reference sequences from all confirmed HCV genotypes and subtypes. The light gray shaded region corresponds to the PepD2-WT peptide, which is also the CypA-binding site. c, secondary structure propensity analysis from experimental 13 C␣ and 13 C␤ chemical shifts of NS5A-D2. Positive and negative scores indicate helical tendencies and extended regions, respectively, whereas values close to 0 indicate a fully disordered state. d, deviation of experimental 1 H, 15 N-combined chemical shift values from neighbor-corrected IDP values calculated from the primary NS5A-D2 sequence. e, peak intensities in the 1 H, 15 N-HSQC spectrum, shown in Fig. S1a. f-h, longitudinal R1 (f) (Hz), transverse R 2 (Hz) (g), and heteronuclear NOE ( 1 H, 15 N) (h) relaxation rates in NS5A-D2, measured at 600 MHz and 298 K. R 1 and R 2 were calculated by fitting peak heights in a series of spectra to a decaying exponential, using Sparky. The error bars indicate the standard error of the fit. For 1 H-15 N NOE the error bars were calculated from the root sum square of (noise/signal) in the spectra with and without saturation. i, 1 D NH dipolar couplings measured using 2D 15 N IPAP-HSQC experiments recorded both in isotrope and anisotrope conditions. The error bars represent experimental errors, calculated from the linewidths at half-height in the 15 N dimension and the signal to noise ratio. The folding-unfolding of the PW-turn structural motif in the disordered NS5A-D2 thus is allosterically regulated by the cis/ trans conformation of 5 different proline residues, i.e. Pro 306 , Pro 310 , Pro 315 , Pro 319 , and Pro 320 . Adopting a cis conformation for any single proline residue reduces the structured population in the conformational ensemble. The impact of these cis-prolines on the PW-turn is not directly proportional to the distance to Trp 312 in the primary sequence. The cis-Pro 319 and cis-Pro 320 , which are at a distance of 7 and 8 residues, respectively, have for example a more pronounced effect on the PW-turn than the cis-P306 that is at a distance of 6 residues (Fig. S14). Interestingly, the three proline residues having a major effect on the structured PW-turn content, i.e. Pro 310 , Pro 315 , and Pro 319 , respectively ( Fig. 3 and Fig. S14), correspond to the ones having the most pronounced functional impact on HCV RNA replication (28).

CypA-inhibitor resistance mutations correspond to allosteric perturbations
The model of NS5A-D2 with a population-weighted average between a structured PW-turn motif and its disordered counterpart is further strengthened by the analysis of the NMR spectra of NS5A-D2 mutants. We examined the position of the major Trp 312 resonance (i.e. W312_1) in the 1 H, 15 N-HSQC spectra of two NS5A-D2 mutants with CypA-inhibitor resistance mutations (D316E or D316E/Y317N (DEYN), respectively). These mutations confer to HCV a moderate resistance (ϳ2-4 -fold) to CypIs and a reduced CypA dependence (27,41,43,44). Comparing them to the position of Trp 312 in NS5A-D2 Pro to Ala mutants (P306A, P310A, P315A, P319A, and P320A) (Fig. 5) and the A311G mutant, in which the PW-turn motif is absent (see Fig. S1b), we find that the Trp 312 peak in all these mutants displays a colinear chemical shift perturbation pattern (68) (Fig. 5a), which coincides with the one previously defined by the cis forms of individual prolines in NS5A-D2 WT (Fig.  4a). This further confirms the PW-turn as a dynamic, rapidly inter-converting ensemble wherein individual mutations directly influence the population of folded and unfolded conformers. The 1 H and 15 N chemical shift values of the Trp 312 resonance from the mutants correlate linearly with several structural NMR parameters ( 13 C␣, ⌬␦NH) (Fig. 5, b and c, and Fig. S15), and indicate that all but the P320A mutation reduce the population of the structured PW-turn motif in the dynamic ensemble (Fig. 5a). To verify whether the effect of the CypI- Conformational regulation of NS5A-D2 by CypA resistant mutants was direct rather than indirectly mediated through an altered proline cis/trans ratio, we explicitly measured the cis/trans ratio of each proline in the pepD2-WT or its D316E counterpart ( Fig. 5d and Fig. S16). Only minor increases of the cis content was measured for Pro 315 and Pro 319 , but the all-trans form remained the dominant fraction.
Our combined results demonstrate that the short structural PW-turn motif ( 310 PAWA 313 ) identified in NS5A-D2 is structurally coupled to a larger region encompassing the 304 -323 residues, whereby this larger peptide can adopt a structured motif (the PW-turn) in equilibrium with a disordered conformer. Any perturbation in this region has its impact on this equilibrium (Fig. S17). The different amino acid substitutions (A311G, D316E, DEYN, P306A, P310A, P315A, P319A, and P320A) and the cis-proline conformation of Pro 306 , Pro 310 , Pro 315 , Pro 319 , and Pro 320 , all constitute a library of allosteric perturbations. We used the chemical shift projection analysis (CHESPA) method (69, 70) on the two linear chemical shift patterns described in Figs. 4a and 5a, respectively (Fig. 5e). To this end, W312_6 (disordered) and W312_1 (structured) have been considered as the extremities of the linear pattern. The projection angle (cos) defines the direction of the perturbation and also shows if the observed residue (here Trp 312 ) is affected by the perturbation through nearest-neighbor effects. All perturbations resulted in a cos value of ϳ1, which is expected for a linear pattern. The only exception is for the A311G mutation that has a stronger nearest-neighbor effect on the Trp 312 resonance. The fractional shift (X) allows the quantification of the two states (here disordered/structured) in the presence of one of the perturbations (here, cis-Pro or mutations). The D316E and DEYN mutations, the cis-Pro 315 conformation, or the A311G mutation thereby correspond to ensembles wherein 82, 61, 55, or 7% of the conformers, respectively, adopt the structured PW-turn motif.

Molecular dynamics of the PW-turn
To validate our NMR-based conclusions by an independent approach, we performed Gaussian accelerated molecular dynamics (MD) simulations (ϳ280 ns, 140,000 frames) on five 20-mer peptides (pepD2-WT, -D316E, -DEYN, -A311G, and -P315cis). For each simulation, a dihedral principal component analysis (dPCA) (71,72) was performed on the backbone dihedral angles ( and ) of residues 309 to 316, and a clustering of the structures was performed using the two first vectors of the dPCA (Fig. S18a). The percent of the time that each cluster is present was determined. Then, using our NMR structure as a reference, we probed the presence of the PW-turn motif in every cluster of each calculation. In the MD simulations, the PW-turn motif was found in 90, 73, 66, 65, and 2% of the conformers (called the PW_fraction) for pepD2-WT, pepD2-D316E, pepD2-DEYN, pepD2-P315cis, and pepD2-A311G, respectively (Fig. 6a). Ensemble average proton amide ( 1 H and 15 N) chemical shift predictions, performed with ShiftX2 (73) on the MD structures, showed linear patterns for Trp 312 and Ala 313 , confirming the experimental observations (Fig. S18, b  and c). There is a strong linear correlation between the PW_fractions from the MD simulations and the experimental 1 H or 15 N chemical shifts of Trp 312 in the HSQC spectra of NS5A-D2 (Fig. 6, b and c). We also found a linear correlation (R 2 ϭ 0.94) between the MD PW_fractions and the fractional shift (X) from the CHESPA analysis (Fig. 6d), illustrating the validity of the model.

The interaction with CypA is modulated by the fraction of the structured PW-turn in NS5A-D2
We have previously shown that the disruption of the structured PW-turn motif weakens 3-fold the molecular interaction between CypA and NS5A-D2 (in the context of the Con1 strain) (30). As the CypA-inhibitor resistance mutation D316E reduces the fraction of the NS5A-D2 conformers that retain this structural element (Fig. 6), the molecular interaction between NS5A-D2 D316E and CypA was assessed. Using 15 N-labeled CypA, NMR titration experiments were acquired with increasing amounts of unlabeled PepD2-WT, PepD2-D316E, or PepD2-A311G peptides ( Fig. 7 and Fig. S19, a-c). The dissociation constants (K D ) were determined from the chemical shift perturbations of CypA induced by the addition of the peptides. The affinity between CypA and PepD2-WT (K D ϭ 0.74 mM),

Conformational regulation of NS5A-D2 by CypA
which hold the structured PW-turn, is ϳ2 times higher than the one with the fully disordered PepD2-A311G peptide (K D ϭ 1.37 mM). These affinities are similar to the ones that we have previously measured between CypA and similar peptides derived from the HCV Con1 strain (0.53 and 1.38 mM, respectively) (30). With respect to the interaction between CypA and the peptide pepD2-D316E, an intermediate K D value of 1.18 mM was measured. The affinities between CypA and the NS5A-D2-derived peptides, even being in the same order of magnitude, are correlated with the fraction of the conformers that own the structured PW-turn motif in the NS5A-D2 ensemble, as determined by NMR spectroscopy and MD simulations (Fig. S19, d and e).

The fraction of structured PW-turn motif in NS5A-D2 ensemble tunes the HCV replication level
The PW-turn motif in NS5A-D2 is essential for HCV RNA replication (10, 30), but has the dynamic equilibrium between the structured and disordered states that we identified in this N, 13 C-doubly labeled peptides. Error bars were calculated based on the signal-to-noise ratio with uncertainties propagations. e, CHESPA analysis. Both the fractional structuration (X, gray bars) and the projection angle (cos, white bars) were calculated for the linear chemical shift patterns shown in Fig. 3 and in panel a, respectively. In both cases, the W312_1 (All_trans) and W312_6 (P310cis) resonances of NS5A-D2 WT were taken as the extremities of the linear patterns.

Conformational regulation of NS5A-D2 by CypA
study have any functional relevance? To address this question, we measured the replication of subgenomic HCV replicons (JFH1 strain) encoding a firefly luciferase gene. These replicon RNAs were transfected into Huh-7 cells and replication was determined by quantifying luciferase in lysates of cells that were harvested at different time points after transfection. We have earlier shown that luciferase activity is a direct measure of viral RNA replication with the 4-h value serving as baseline because it reflects transfection efficiency. To study the importance of the equilibrium shift of the PW-turn motif, we compared replication capacity of the wildtype (WT) replicon or replicon variants containing distinct mutations in NS5A-D2 (A311G, P310A, DEYN, and A311I) (Fig. 8a). The A311G substitution in NS5A reduced RNA replication to background levels as determined with the replication-dead replicon encoding an enzymatically inactive NS5B polymerase (mutant ⌬GDD). By contrast, the replication efficiency of the A311I mutant was similar to that of the NS5A-WT consistent with the notion that the structural PW-turn motif in the disordered NS5A-D2 domain is essential for viral RNA replication. Indeed, in the HCV Con1 strain, position 311 corresponds to an isoleucine residue (30). Mutations P310A and DEYN in NS5A affected RNA replication, showing phenotypes intermediate between those of the WT and A311G mutant. Of note, we found a striking correlation between RNA replication efficiencies measured in cellulo and the NMR data acquired on purified NS5A-D2 in vitro (Fig.  8, b and c, and Fig. S20). Indeed, RNA replication levels of NS5A WT and its NS5A mutants correlate with the position ( 1 H and 15 N) of the Trp 312 resonance in their corresponding 1 H, 15 N-HSQC spectra. This strong correlation suggests that the structured PW-turn motif in the NS5A-D2 ensemble is required for robust HCV RNA replication, provided the CypA PPIase activity allows to reach, on a subsecond time scale, subensembles in which the structured PW-turn population is reduced. It means that the conformational equilibrium of the PW-turn motif in NS5A-D2 has to be finely regulated to be fully functional. Hence, our data also provide an explanation for the striking antiviral potency of CypIs.

Discussion
RNA replication, a central step in the HCV life cycle, requires the formation of a replication complex that includes the viral NS5B and NS5A proteins associated to membrane structures (DMVs) (7,74). NS5B with its RNA-dependent RNA-polymerase activity (8) is the catalytic core of this functional complex. In contrast, the absence of measurable enzymatic activity and the presence, alongside its first well-structured domain (11,20,21), of two intrinsically disordered domains (NS5A-D2 and -D3) (17)(18)(19), make the role of NS5A in the HCV replicase much less obvious. Human CypA, as an essential cellular protein required for the viral replication process (75), has been functionally linked to NS5A-D2 by the accumulation of CypI-resistance mutations in this disordered domain (41,76), and by the identification of a physical interaction between these two proteins (25,37). However, the role(s) of these two proteins in the replication of HCV remains elusive, which is in part because the PPIase activity and the binding properties of CypA cannot be uncoupled.
The precise role of human cyclophilins in a viral life cycle has been best studied in the context of HIV capsid (de)stabilization (77). CypA is incorporated into newly produced HIV virions through its interaction with a proline-rich loop (CypA-loop) in the viral capsid protein (CA) (78 -80). Although the Pro 90 residue in CA is required for both the interaction with CypA and viral replication, the importance of other surrounding proline residues (Pro 85 , Pro 93 , and Pro 99 ) in viral replication was also pointed out (79). CypA displays in vitro catalytic PPIase activity toward the Gly 89 -Pro 90 peptidyl bond in the CypA-loop of the CA protein in the intact HIV-1 virion (81), but the functional role of this activity remains uncertain. CA mutations such as A92E and G94D, obtained during HIV-1 passage in HeLa cells under CypA inhibition (82), allow these HIV-1 mutants to escape CypA dependence, without altering the interaction with CypA (83). However, the infectivity of these mutants drops by 90% in HeLa cells, which could be fully recovered upon CypA inhibition (83). Recently, it has been reported that the CypAloop of the A92E and G94D CA mutants in the assembled capsid structures adopts comparable dynamics as the loop in the WT CA when bound to CypA (84). Hence, both selected resistance mutations and CypA binding seemingly lead to the same dynamic effect on this CA loop, but cannot be combined without the risk of overshooting the dynamical requirements for optimal infectivity. Along these lines, the rhesus monkey Trim5␣ induces global attenuation of the capsid dynamics even beyond the CA loop and may ultimately promote its disassem-

Conformational regulation of NS5A-D2 by CypA
bly (85). From these data on HIV-1 CA-CypA interaction, one can conclude that fine tuning of dynamical aspects of viral proteins (here CA) is an additional mean used by viruses to optimize infectivity.
Our present work on HCV NS5A-D2 and its relationship with CypA points toward the same direction. We detect the PW-turn structural motif in the mainly disordered NS5A-D2 domain from the HCV JFH-1 strain (Figs. 1 and 2), thereby confirming our previous results in the Con1 strain (30). Moreover, we show that the structural motif extends beyond the 310 PAWA 313 sequence to include at least 4 more proline residues (Pro 306 , Pro 315 , Pro 319 , and Pro 320 ) ( Fig. 3 and Fig. S17). However, this motif is not static. Using NMR chemical shifts as atomic resolution sensors, we demonstrate that the PW-turn motif exists as an equilibrium between two states, a structured state and a disordered state, which interconvert in fast exchange on the NMR time scale (Fig. 4). This equilibrium, probed via the Trp 312 resonance, is modulated by the cis/trans equilibria of the 5 surrounding prolines (Pro 306 , Pro 310 , Pro 315 , Pro 319 , and Pro 320 ). This translates into one major and 5 minor resonances for the Trp 312 amide function that make a linear chemical shift pattern in the 1 H, 15 N-HSQC of NS5A-D2 WT (Figs. 4 and 5). The linear pattern is bordered on the structured side by the all-trans form used to derive the structure of the PW-turn (Fig. 3), and by the cis-Pro 310 form on the almost completely disordered side (Figs. 3 and 4). As these different equilibria are separated by the conformation (trans or cis) of distinct prolyl bonds, they are separated by the slow time scale of the spontaneous cis/trans isomerization of each of these bonds. CypA can lower this barrier through its PPIase activity, and thereby connect these different ensembles at a subsecond time scale (Fig. 3b) (25,30). As such, the PPIase activity of CypA do not alter the folded and unfolded states of the PW-turn but rather allows fast reach to the NS5A-D2 subensembles in which the structured PW-turn population is reduced (Fig. S21). The CypI resistance mutations in NS5A-D2, D316E, or DEYN lead to a similar effect as they reduce the population of the structured PW-turn motif in the dynamic ensemble ( Fig. 5 and Fig.  S21). The folded conformer population, with all prolines in trans, drops from nearly ϳ100% in NS5A-D2 WT to 82% in the D316E mutant (i.e. the same population as in the cis-Pro 319 subensemble in the WT), and to 62% in the DEYN mutant (comparable with that of the cis-Pro 315 subensemble in the WT) (Figs. 3 and 5). Both mutations confer some CypA inde-pendence, but lower the replication level when measured in cell lines where CypA is present (27,41,43,44). Moreover, in a similar way to what was observed for the mutations in the CypA-loop of HIV-1 CA, the replication capacity of the DEYN HCV mutant has been shown to be higher when CypA was silenced (28). As the Trp 312 chemical shift ( 1 H and 15 N) of the mutants correlates linearly with the RNA replication efficiency (log 10 scale) in a cell-based assay (Fig. 8), we conclude that CypA is fine tuning the dynamics of the PW-turn motif in NS5A-D2.
The relationships involving CypA/CA from HIV-1 or CypA/ NS5A-D2 from HCV are strikingly similar. A short sequence involving a conserved proline residue constitutes the CypAbinding site (Gly 89 -Pro 90 or 310 PAWA 312 , respectively); the residues C-terminal to this binding site trigger peculiar conformational properties and are involved in the replication efficiency of the viruses; CypI-resistance mutations conferring (partial) relief of the CypA-dependence are localized next to the CypA-binding site; and finally, CypA exerts a fine modulation of the dynamics of the viral proteins (the CypA-loop of CA or the PW-turn motif in NS5A-D2). HIV-1 and HCV seem to have evolved to use the host CypA as a fine-tuning rheostat, which allows them to keep their functional systems in a rather sharp optimal window. Finally, in the presence of the CypI-resistance mutations, CypA is detrimental as it over-attenuates the dynamics of the CypA-loop in HIV-1 CA or it over-reduces the fraction of the PW-turn in HCV NS5A.
Whereas we show that the fraction of the structured PWturn motif in the NS5A-D2 conformational ensemble, which is allosterically regulated by both the cis/trans isomerization of 5 prolines residues (Pro 306 , Pro 310 , Pro 315 , Pro 319 , and Pro 320 ) and by CypI-resistance mutations, correlates with the HCV RNA replication efficiency (Fig. 8), the question of how a structured PW-turn motif in NS5A-D2 contributes to viral replication remains open. NS5A was shown to play a role in the formation of DMVs and functional replication complexes contained therein, through remodeling the ER membrane, and CypI could interfere with this role for NS5A. Alternatively, NS5A-D2 directly interacts with the dynamic molecular machine that is the NS5B RNA polymerase (86), and thereby might allosterically regulate its RNA binding (15) and/or enzymatic activity (87). Different conformational ensembles of the disordered NS5A-D2, with varying fractions of (un)folded PW-turn motif, might be required for

Conservation of NS5A-D2 sequence among HCV genotypes
The NS5A-D2 sequence from the HCV JFH1 strain (AB047639, genotype 2a) is numbered as in the full-length NS5A protein. The amino acid repertoire was deduced from the ClustalW multiple alignments of 28 representative NS5A sequences from all confirmed HCV genotypes and subtypes (see the European HCV Database (88)) using the Network Protein Sequence Analysis webserver tools (89). Amino acids observed at a given position in less than two distinct sequences are not included. The degree of amino acid conservation at each position can be inferred from the extent of variability (with the observed amino acid listed in decreasing order of frequency from top to bottom) together with the similarity index according to ClustalW convention (asterisk, invariant; colon, highly similar; dot, similar).

NMR spectroscopy
All NMR experiments were performed at 298 K using Bruker Avance 600 MHz or 900 MHz NMR spectrometers, both equipped with a cryogenic triple resonance probe (Bruker, Karlsruhe, Germany). The proton chemical shifts were referenced using the methyl signal of TMSP (sodium 3-trimethylsillyl- [2,2,3,3-d 4 ]propionate) at 0 ppm. Spectra were processed and analyzed with the Bruker TopSpin software package 3.2. Data analysis, peak picking, and calculation of peak volumes were done with Sparky software (90).

Conformational regulation of NS5A-D2 by CypA
The 1 H, 15 N-combined chemical shift perturbations were calculated using Equation 1, whereby ⌬␦ ( 1 HN) and ⌬␦ ( 15 N) are the chemical shift perturbations in the 1 H and 15 N dimensions, respectively. The normalization factor of 0.08 for the nitrogen frequency shift derives here from the ratio of the maximum proton frequency shift over the maximum nitrogen frequency shift.

Residual dipolar couplings measurements
The RDCs were collected on a 100 M sample of 15 N NS5A-D2 aligned in a liquid crystalline medium consisting of 6.6% (w/v) polyoxyethylene 5-lauryl ether (C 12 E 5 ) and 3% (w/v) 1-hexanol (Sigma), yielding a D 2 O splitting of 35 Hz. 1 D NH dipolar couplings were measured on a Bruker Avance III 900 MHz spectrometer equipped with a cryogenic triple resonance probe, using 2D 15 N IPAP-HSQC experiments (92), which allow the spin coupling measurements in the 15 N dimension. The difference between the couplings acquired either in isotropic or anisotropic media were calculated to get the RDCs. 15 15 N NOE values were determined from spectra recorded either in the presence or absence of a proton pre-saturation period of 3 s, and a relaxation delay of 5 s.

NMR structure calculation
The NMR structure calculation was performed as previously described in Ref. 30. From the different NMR datasets acquired both on unlabeled and 15 N, 13 C-labeled peptide PepD2-WT (residues 304 -323, JFH1), distance-based (NOEs) and backbone dihedral angle-based experimental restraints were derived. NOE intensities used as input for structure calculations were obtained from the NOESY spectrum recorded with a 400-ms mixing time. According to their intensity NOEs were classified in three categories, which were then converted into distance restraints (1.8 -2.8 Å, 1.8 -3.9 Å, and 1.8 -5.0 Å). Protons without stereospecific assignments were treated as pseudo-atoms. From the 1 H, 15 N, and 13 C chemical shifts, dihedral angle constraints, calculated with Talos (93), were introduced. Peptide structures were generated from the experimental NOE distances and dihedral angles, using CNS (94) with the standard torsion angle molecular dynamics protocol, the standard force field and default parameter set. From the initial set of 100 structures that were calculated, with the dynamical annealing protocol to widely sample the conformational space, only structures with no distance restraint violations were retained. The 23 final selected structures, with the lowest energies, were compared by pairwise root mean square deviation over the backbone atom coordinates (N, C␣, and CЈ). Ramachandran analysis performed on the final structures showed that 89, 11, and 0% of the residues were in favored, allowed, and outliers regions, respectively. The PyMOL software (PyMOL Molecular Graphics System, version 1.8; Schrödinger) was used for molecular graphics (95).

NMR PPIase assay
PPIase activity of CypA on NS5A-D2 WT was assessed using 1 H, 15 N z-exchange spectra (66), where the exchange was monitored on the basis of novel cross-peaks connecting a trans and cis peak. 1 H, 15 N z-exchange spectra, with a 200-or 400-ms mixing time, were recorded on a 600 MHz spectrometer equipped with a cryogenic triple resonance probe. Exchange spectra were acquired on a 400 M 15 N-NS5A-D2 sample in the presence or the absence of 5 M CypA in 30 mM NaH 2 PO 4 / Na 2 HPO 4 , pH 6.4, 30 mM NaCl, 1 mM DTT.

Chemical shift projection analysis (CHESPA)
The NMR chemical shift projection analysis was performed as described in Refs. 69 and 70, from the 1 H, 15 N-HSQC of NS5A-D2. We calculated cos, which represents the angle between vectors A (defined by the W312_6 and W312_1 peaks) and B (defined by the W312_6 peak and the Trp 312 resonance from a mutant or corresponding to a minor form). Next, we calculated the fractional shift X, which corresponds to the projection of the vector B on the vector A. For the calculations, a scaling factor of 0.1 was applied to the 15 N chemical shift.
Briefly, the peptides were built using Tleap in AmberTools16 and all simulations were performed using pmemd.cuda of AMBER17 on graphics processing units P100 (98). Amber ff99SB*-ILDN force field (99,100) was used in all simulations. The peptides were then solvated in a cubic water box of 75.5 ϫ 75.5 ϫ 75.5 Å 3 pre-equilibrated TIP3P water molecules. The simulations were performed with 40 mM NaCl at 298 K, as in the experimental NMR studies. GaMD (101) simulations (ϳ280 ns) were used to explore the conformational space of the peptides and the coordinates were saved every 2 ps. CPPTRAJ (1) was used to analyze r.m.s. deviations, secondary structure, dihedral torsions, and hydrogen bonds from the GaMD simulation trajectories. The sampled conformations of the peptides were analyzed using the dPCA method (71,72) considering the backbone atoms of residues 309 to 316. The lowest energy conformations were identified by projecting the trajectories of the first two principal components onto a three-dimensional free energy (⌬G) (see Equation 2 below), in which R is the universal Conformational regulation of NS5A-D2 by CypA gas constant, T is the temperature, x, y, and z, are the calculated structural properties from the trajectory.
⌬G ϭ ϪRT ln ͩ P x,y,z P max ͪ (Eq. 2) Then, a clustering of the structures was performed using the two first vectors (PC1 and PC2) of the dPCA. For each MD simulation, the PyReweighting toolkit was used to reweight the GaMD simulation to compute the free energy landscape from PCA components PC1 and PC2. The percent of the time that a particular cluster is present was determined. Next, using our NMR PepD2-WT structure as a reference, the presence of the structured PW-turn motif was probed in every cluster of each simulation. For clusters of each simulations, the centroid structures were fitted on the NMR reference structure using all atoms of residues 310 -313. Then, backbone dihedral angles ( and ) of residues 310 to 313 and 1 and 2 angles from Trp 312 were measured using cpptraj in AMBER package. The presence of the structured PW-turn motif was assessed from the sum of squares of the differences between angles from the reference and centroid structures. Chemical shift predictions ( 1 H and 15 N) were performed with SHIFTX2 (73) from the MD simulations data. For each peptide, SHIFTX2 was used in the "ensemble mode" on all structures of each cluster. Then, from these predictions for a given peptide, a population-weighted average chemical shift value ( 1 H or 15 N) was calculated. The error bars represent the population-weighted standard deviations.

RNA replication assay
The protocol used for generation and electroporation of HCV RNAs has been described elsewhere (38). For transient replication assays, 400 l of single cell suspensions of Huh-7 cells (10 7 cells/ml) were mixed with 5 g in vitro transcribed subgenomic replicon RNA and transfected by electroporation. After transfection, cells were resuspended in 41 ml of complete Dulbecco's modified Eagle's medium, and 1.5 ml of the cell suspension was seeded in duplicate in each well of a 12-well plate. To measure luciferase activity, cells were washed with PBS 4, 24, 48, and 72 h after electroporation and lysed by addition of 350 l of lysis buffer (0.1% Triton X-100, 25 mM glycylglycine, pH 7.8, 15 mM MgSO 4 , 4 mM EGTA, and 1 mM DTT). Lysates were immediately frozen at Ϫ70°C, and after thawing, 100 l of the lysate was mixed with 360 l of assay buffer (25 mM glycylglycine, 15 mM MgSO 4 , 4 mM EGTA, 1 mM DTT, 2 mM ATP, and 15 mM K 2 PO 4 , pH 7.8). Luciferase activity was measured for 20 s in a luminometer (Lumat LB9507; Berthold, Freiburg, Germany) after addition of 200 l of luciferin solution (200 mM luciferin, 25 mM glycylglycine, pH 8.0). Replication efficiency was calculated by normalizing values of the different time points to the respective value obtained at 4 h, which reflects transfection efficiency.