Structure and function of the membrane anchor domain of hepatitis C virus nonstructural protein 5A.

Hepatitis C virus (HCV) nonstructural protein 5A (NS5A) is a membrane-associated, essential component of the viral replication complex. Here, we report the three-dimensional structure of the membrane anchor domain of NS5A as determined by NMR spectroscopy. An alpha-helix extending from amino acid residue 5 to 25 was observed in the presence of different membrane mimetic media. This helix exhibited a hydrophobic, Trprich side embedded in detergent micelles, while the polar, charged side was exposed to the solvent. Thus, the NS5A membrane anchor domain forms an in-plane amphipathic alpha-helix embedded in the cytosolic leaflet of the membrane bilayer. Interestingly, mutations affecting the positioning of fully conserved residues located at the cytosolic surface of the helix impaired HCV RNA replication without interfering with the membrane association of NS5A. In conclusion, the NS5A membrane anchor domain constitutes a unique platform that is likely involved in specific interactions essential for the assembly of the HCV replication complex and that may represent a novel target for antiviral intervention.

Hepatitis C virus (HCV) nonstructural protein 5A (NS5A) is a membrane-associated, essential component of the viral replication complex. Here, we report the three-dimensional structure of the membrane anchor domain of NS5A as determined by NMR spectroscopy. An ␣-helix extending from amino acid residue 5 to 25 was observed in the presence of different membrane mimetic media. This helix exhibited a hydrophobic, Trprich side embedded in detergent micelles, while the polar, charged side was exposed to the solvent. Thus, the NS5A membrane anchor domain forms an in-plane amphipathic ␣-helix embedded in the cytosolic leaflet of the membrane bilayer. Interestingly, mutations affecting the positioning of fully conserved residues located at the cytosolic surface of the helix impaired HCV RNA replication without interfering with the membrane association of NS5A. In conclusion, the NS5A membrane anchor domain constitutes a unique platform that is likely involved in specific interactions essential for the assembly of the HCV replication complex and that may represent a novel target for antiviral intervention.
Hepatitis C virus (HCV) 1 infection is a major cause of chronic hepatitis, liver cirrhosis, and hepatocellular carcinoma world-wide (1). HCV contains a 9.6-kb single-stranded RNA genome of positive polarity that encodes a polyprotein precursor of about 3,000 amino acids (for review, see Refs. [2][3][4]. The polyprotein precursor is co-and post-translationally processed by cellular and viral proteases to yield the mature structural and nonstructural proteins. The structural proteins include the core protein and the envelope glycoproteins E1 and E2. The nonstructural proteins include the NS2-3 autoprotease and the NS3 serine protease, an RNA helicase located in the C-terminal region of NS3, the NS4A polypeptide, the NS4B and NS5A proteins, and the NS5B RNA-dependent RNA polymerase. As in all positive-strand RNA viruses, the nonstructural proteins form a membrane-associated replication complex together with replicating viral RNA, altered membranes, and additional as yet unidentified host cell components (5)(6)(7)(8). Determinants for membrane association of the HCV nonstructural proteins have been mapped (for review, see Ref. 9), but the protein-protein interactions involved in formation of a functional HCV replication complex are poorly understood.
HCV NS5A is a phosphoprotein of unknown structure and function (10,11). It is found in a basally phosphorylated form of 56 kDa and in a hyperphosphorylated form of 58 kDa. NS5A has attracted considerable interest because of its potential role in modulating the interferon response (for review, see Ref. 12), and numerous additional functions have recently been ascribed to this protein (for review, see Ref. 13). Interestingly, cell culture-adaptive changes cluster in the central portion of NS5A in the context of selectable subgenomic HCV replicons (14,15), indicating that NS5A is involved, either directly or by interaction with cellular proteins, in the viral replication process. This observation, together with the modulation of NS5A hyperphosphorylation by the nonstructural proteins 3, 4A, and 4B (16,17) and physical interactions described with other nonstructural proteins (18,19), strongly supports the notion of NS5A being an essential component of the HCV replication complex.
We have shown recently that the N-terminal 30 amino acid residues serve as a membrane anchor for NS5A (20). This domain was necessary and sufficient to target NS5A or a heterologous fusion protein to the endoplasmic reticulum (ER) or an ER-derived modified compartment by a post-translational mechanism, resulting in integral membrane association. Structure predictions and circular dichroism analyses indicated that this domain contains an amphipathic ␣-helix.
Here, we describe the three-dimensional structure of a synthetic peptide corresponding to the NS5A membrane anchor domain, NS5A , as determined by NMR in the presence of different membrane mimetic media. We report that the NS5A membrane anchor forms an in-plane amphipathic ␣-helix embedded in the cytosolic leaflet of the membrane bilayer. Moreover, based on targeted mutagenesis and RNA replication analyses we propose that the polar residues at the membrane surface define a unique platform that is involved in specific protein-protein interactions essential for the assembly of a functional HCV replication complex.

EXPERIMENTAL PROCEDURES
Peptide Synthesis-The NS5A  peptide, representing amino acids 1-31 of NS5A of the HCV H strain (GenBank accession number AF009606), was synthesized by the stepwise solid phase method of Merrifield and employing Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry. The peptide was highly purified by reverse phase high performance liquid chromatography and exhibited the expected molecular mass of 3,764 Da.
NMR experiments in H 2 O:TFE-d 2 were performed at 20°C, and those in SDS-d 25 and DPC-d 38 were recorded at 40°C. Multidimensional NMR experiments were carried out at 500 MHz on a Varian Unity-plus spectrometer equipped with a triple resonance 5-mm probe with a self-shielded z gradient coil. Double quantum-filtered correlation spectroscopy, nuclear Overhauser effect spectroscopy (NOESY), twodimensional homonuclear total correlation spectroscopy (TOCSY), and 1 H-13 C heteronuclear single quantum correlation experiments using conventional optimized pulse sequences as well as data collection, processing, and spectra analyses for spin systems identification and sequential assignments were performed as described previously (21-23 and references therein).
NMR-derived Constraints and Structure Calculations-NOE intensities used as input for the structure calculations were obtained from the NOESY spectrum recorded with a 100-or 150-ms mixing time. NOEs were partitioned into three categories of intensities that corresponded to distances ranging from a common lower limit of 1.8 Å to upper limits of 2.6, 3.8, and 5 Å, for strong, medium, and weak intensities, respectively. The cross-peak intensities of the H ␦ -H ⑀ protons of Phe 19 were used as reference distance (2.45 Å). No hydrogen bond or dihedral angle restraints were introduced. Protons without stereospecific assignments were treated as pseudoatoms, and the correction factors were added to the upper and lower distance constraints.
Three-dimensional structures were generated from NOE distances by the restrained dynamical simulated annealing protocol (24) with X-PLOR 3.85 using standard force fields and default parameter sets. Sets of 100 structures were calculated to sample the conformational space widely, and structures with restraint violations Ͻ0.5 Å and angle violations Ͻ0.5°were retained. The selected structures were compared by pairwise root mean square deviation (r.m.s.d.) over the backbone atom coordinates (N, C␣, and CЈ). Local analogies were analyzed by calculating the local r.m.s.d. of a tripeptide window slid along the sequence. Statistical and structural analyses were performed using AQUA and PROCHECK-NMR software (25). Superimposition of structures, three-dimensional graphic display and manipulations were carried out using MOLSCRIPT (26) (see Fig. 3A) and RASMOL 2.5 (27).
Replicon constructs harboring mutations within the NS5A membrane anchor domain of NS5A were prepared by PCR-based site-directed mutagenesis using oligonucleotide primers carrying the desired modifications ( Table 1 of the Supplemental Data). DNA fragments were inserted into replicon construct pFK-nt341-sp-PI-lucEI3420 -9605/ E1202GϩT1280IϩK1846T (15). The SfiI fragment of the reporter replicons was introduced into pFK1-389neoEI3420 -9605 (29) to yield selectable subgenomic replicons.
In Vitro Transcription, Electroporation, and Replication Assays-In vitro transcription, transfection of Huh-7 human hepatoma cells by RNA electroporation, and RNA replication assays using luciferase reporter replicons as well as selectable HCV replicons were performed as described previously (31,32).
Confocal Laser Scanning Microscopy-Immunofluorescence staining was performed as described previously (33). Coverslips were mounted in SlowFade (Molecular Probes, Eugene, OR) and examined with a Zeiss LSM 510 laser scanning system. Images were processed with Zeiss Image 3.1.0.99 and Adobe Photoshop 7.0 software.
Membrane Flotation Assays-Membrane flotation assays were performed as described previously (34) with minor modifications. In brief, 2 ϫ 10 7 transiently transfected U-2 human osteosarcoma cells were Dounce homogenized in a hypotonic buffer (10 mM Tris-HCl, pH 7.5, 2 mM MgCl 2 ) followed by centrifugation at 1,000 x g for 5 min to pellet nuclei, unlysed cells, and large debris. Nycodenz (Axis Shield, Oslo, Norway) was added to postnuclear supernatants to a final concentration of 37.5% (w/v) in phosphate-buffered saline. Four hundred l lysate was placed at the bottom of a 1.5 ml thick-walled ultracentrifugation tube and overlaid with 900 l 35% and 100 l 5% Nycodenz in phosphate-buffered saline. Equilibrium centrifugation was carried out at 100,000 x g for 20 h at 4°C in a Beckman TLS 55 rotor. Subsequently, 8 fractions of 175 l each were collected from the top, and equal volumes were subjected to immunoblot as described previously (33).

RESULTS
NMR Structure of NS5A -Previous circular dichroism analyses of the NS5A  peptide performed in the presence of SDS or DPC micelles and in TFE:water mixtures indicated an ␣-helix content of about 80% (20). Here, the conformational behavior of NS5A  in these commonly used membrane mimetic media (35) was investigated further by NMR to gain insights into peptide folding at the atomic level. The twodimensional homo-and heteronuclear NMR analyses of NS5A  dissolved in 100 mM SDS-d 25 , 100 mM DPC-d 38 , or 50% TFE-d 2 yielded well resolved spectra, as illustrated in Fig.  1A by an extract of the NOESY spectrum in 100 mM SDS. Sequential attribution of all spin systems was completed in the three membrane mimetic environments.
An overview of the sequential and medium range NOE connectivities for the peptide in SDS-d 25 is shown in Fig. 1B. NOE connectivities observed in TFE-d 2 and in DPC-d 38 are available in Fig. 1 of the Supplemental Data. Despite the lack of numerous connectivities caused by overlapping cross-peaks in SDS and DPC (indicated by asterisks), the NOE connectivity patterns are similar in the three membrane mimetic environments. The main body of the peptide (Leu 5 -Ala 25 ) displays typical characteristics of an ␣-helix, including strong dN-N(i,iϩ1) and medium d␣N(i,iϩ1) sequential connectivities, weak d␣N(i,iϩ2), medium d␣N(i,iϩ3), medium or strong d␣␤(i,iϩ3), and weak d␣N(i,iϩ4) medium range connectivities. In contrast, the residues flanking the N-and C-terminal ends of the Leu 5 -Ala 25 helix (i.e. amino acids 4 and 26 -28) remain more flexible with fewer medium range connectivities as a sign of fraying helix ends. Compared with those observed in TFE, several NOE connectivities are missing between Val 15 and Phe 19 in SDS and DPC. Cross-peaks in this region display a lower intensity compared with the other residues, indicating differences in the dynamic behavior in the Val 15 -Phe 19 region (see below). The N-and C-terminal ends of the peptide (i.e. amino acids 1-3 and 29 -31) are devoid of medium range NOEs and remain unstructured in all three media.
Differences of 1 H␣ and 13 C␣ chemical shifts from those found in a random coil conformation are additional indicators of secondary structure (36). Chemical shift differences for 1 H␣ and 13 C␣ are shown in Fig. 2B and Fig. 2 of the Supplemental Data, respectively. The long series of negative 1 H␣ (⌬ 1 H␣ Յ Ϫ0.1 ppm) and positive 13 C␣ chemical shift differences (⌬ 13 C␣ Ն 0.7 ppm) are typical of an ␣-helical conformation. However, the weaker ⌬ 1 H␣ observed around Ile 8 and Asp 18 suggests some flexibility of the ␣-helix around these residues. These data are in close agreement with the NOE connectivities described above. Taken together, these results clearly show that the peptide conformation is very similar in the three membrane mimetic media and that an ␣-helix extends from Leu 5 to Ala 25 .
NMR Structure Model of NS5A -The number of NOEderived interproton distance constraints used for the structure calculations are reported in Table 2 of the Supplemental Data. From the initial 100 structures calculated with X-PLOR, final sets of 51, 43, and 27 low energy structures were retained for the peptide in TFE, SDS, and DPC, respectively. 2 Structure selection was based on NOE violations Յ0.5 Å, and all three sets of structures fully satisfied the experimental NMR data. The final statistics are listed in Table 2 of the Supplemental Data. All structures show a regular ␣-helical conformation extending from residues 4 to 27 in 50% TFE, 5 to 25 in SDS, and 4 to 25 in DPC, but the best superimposition of calculated structures includes residues 5 to 25. A superimposition of the 2 The coordinates of the average structures and the NMR restraints of NS5A  have been deposited in the Brookhaven Protein Data Bank under the accession codes 1R7C, 1R7E, and 1R7G for the peptide in 50% TFE, 100 mM SDS, and 100 mM DPC, respectively. The accession codes 1R7D and 1R7F correspond to the sets of 51 and 43 calculated structures for the peptide in 50% TFE and 100 mM SDS, respectively. The 1 H and 13 C chemical shifts of the assigned residues in 50% TFE, 100 mM SDS, and 100 mM DPC have been deposited in the BioMagResBank (BMRB) under the accession number 5978. calculated structures obtained in SDS is shown as an example in Fig. 2A.
The ␣-helix is well defined between Leu 5 and Ala 25 , as reflected by the low global r.m.s.d. for the backbone atoms (CЈ, C␣, and N) of 1.11 and 1.12 Å for each set of structures in TFE and SDS, respectively (  (Fig. 2B). This dynamical behavior is less pronounced in the TFE and SDS structures (Fig. 2B), but it is supported by the lower chemical shift differences of ⌬H␣ (Fig. 2B) and ⌬C␣ (Fig. 2 of the Supplemental Data) observed in the three media compared with the surrounding residues that are clearly involved in stable helical segments (e.g. residues 12-14 and 20 -23). Although direct evidence of a dynamic behavior remains to be provided, these results suggest that the Val 15 -Ser 17 segment is an intrinsic flexible helical element of the NS5A[1-31] membrane anchor. This flexibility could lead to a variable orientation of the ends of the ␣-helix when NS5A  is bound to phospholipids (see below). Finally, the apparent structural instabilities observed on both edges of the Leu 5 -Ala 25 ␣-helix can be explained by the absence of stabilizing interactions with the rest of NS5A and/or membrane phospholipids in the isolated NS5A  peptide context.
The comparison of the average conformations of NS5A[1-31] obtained in TFE, SDS, and DPC shown in Fig. 3A highlights the close overlap in the folding of the Leu 5 -Ala 25 helices but also shows differences in the slight bending of the helices. This bending is mainly the result of the lack of long range distance restraints when calculating the structure of an isolated ␣-helix. Therefore, the relevance of such bending for membrane association and/or the structure of NS5A cannot be determined. As illustrated in Fig. 3B for the structure determined in SDS, charged and polar residues are exposed on the hydrophilic side, with a remarkable asymmetric distribution of positively and negatively charged residues along both edges of this side of the helix. On the other hand, aromatic residues Trp 4 , Trp 9 , Trp 11 , Phe 19 , and Trp 22 are well positioned on the hydrophobic side, suggesting an essential role in membrane association of the helix.
Positioning of NS5A  in Detergent Micelles-The positioning of NS5A  in the phospholipid bilayer was investigated by three different NMR strategies: 1) NOEs between residue side chains and detergent hydrophobic tails; 2) measurement of amide proton exchange with deuterated water; and 3) accessibility of residues to water by examining the broadening of proton signals caused by the proximity of paramagnetic Mn 2ϩ ions. The data obtained by these three strategies are summarized in Fig. 4A.
NOESY spectra were recorded at a 1:9 ratio of protonated DPC:DPC-d 38 to investigate direct contacts between detergent and NS5A   (22,37). Significant intermolecular NOE cross-peaks were observed between the hydrophobic peptide residues and various CH 2 groups of the hydrophobic tail of DPC. The hydrophobic region of the spectra was too crowded to yield unambiguous data. In contrast, the aromatic region provided NOE cross-peaks of DPC with side chains of Trp and Phe (data not illustrated). These spectra indicate that Trp 9 , Phe 19 , and Trp 22 are mainly buried in the DPC micelles, whereas Trp 4 and Trp 11 appear to be located at the hydrophobic-hydrophilic micelle interface.
Proton exchange analyses showed that all amide protons are exchanged after less than 24 h in TFE-d 2 , although some were still retained in SDS-d 25 (Trp 9 , Trp 11 , Ile 12 , Val 15 , Leu 23 , and Lys 24 as highlighted in Fig. 1B). In DPC-d 38 only the amide proton of Ile 12 appears up to 24 h. This indicates that the corresponding residues are likely buried in the micelle hydrophobic core.
At low concentrations, Mn 2ϩ primarily affects resonances of water-accessible residues at the surface of an SDS micelle (38). The paramagnetic broadening effect of Mn 2ϩ was studied by comparing TOCSY spectra in the presence or absence of 0.05 or 0.5 mM MnCl 2 . Residues for which cross-peaks disappeared in the presence of 0.05 mM MnCl 2 were considered to be wateraccessible (e.g. Asp 10 and Glu 14 , Fig. 4A). Conversely, residues for which cross-peaks were always visible in the presence of 0.5 mM MnCl 2 were considered to be buried. The unambiguous data obtained by this approach are summarized in Fig. 4A  (white arrows).
The combined results show that most of the hydrophobic residues of the helical region are buried in the detergent micelles. Some of the acidic residues are clearly surface-exposed (Asp 10 , Glu 14 ) whereas the side chains of basic residues Arg and Lys appeared mainly buried. Interestingly, two distinct regions connected by the Val 15 -Ser 17 flexible helical element emerge from these analyses. The Trp-rich, N-terminal portion of the helix (amino acids 5-12) seems to be located more closely to the micelle interface, whereas the Lys-rich, C-terminal portion (amino acids 19 -26) appears to be buried in the hydrophobic core of the detergent micelle. This is in line with the presence of short, negatively charged residues in the first portion and Lys side chains in the second, which permit a deeper embedding of the helix in the hydrophobic core. From these data, it is possible to position the average NS5A[1-31] structure approximately in a model phospholipid bilayer (Fig. 4B). It can be deduced that the amino acid 5-25 helix is located at the membrane interface, forming an in-plane amphiphatic ␣-helix embedded in the cytosolic leaflet of the membrane. Assuming that the charged amino acids interact with the polar head of phospholipids, and taking into account the length of their side chains together with that of the phospholipid polar head and hydrophobic tails, it appears that the whole helix could not be buried deeply in the hydrophobic core of the membrane. This positioning is quite similar to that deduced from the NMR analysis of the membrane-proximal region of the human immunodeficiency virus glycoprotein gp41 in DPC micelles where the Trp residues were proposed to be located within the membrane-water interface of the phospholipid bilayer (37). The putative flexible helical segment Val 15 -Ser 17 could play a role in such positioning, allowing either a lateral and/or a vertical bending of the helix with respect to the plane of the membrane. The polar side of the helix is likely accessible at the surface of the membrane, as tentatively illustrated in Fig. 4C.
Subcellular Localization of NS5A Membrane Anchor Mutants-The three-dimensional structure information was used to design a panel of NS5A mutants with amino acid substitutions, insertions, or deletions in the N-terminal ␣-helix as illustrated in Fig. 5A. The mutations were carefully designed to FIG. 4. Expected location and orientation of the NS5A amphipathic ␣-helix in the membrane. A, NMR analyses of NS5A  location in detergent micelles. Black arrows, aromatic residues exhibiting NOE cross-peaks between their side chains and DPC hydrophobic tails. Gray arrows, slow amide proton exchange with deuterated water in SDS-peptide micelles (reported from Fig. 1B). White arrows, residues buried in the SDS micelles for which TOCSY cross-peaks (H␣-HN, side chains) remain observable despite the presence of 0.5 mM MnCl 2 . Black diamonds, residues accessible to water for which TOCSY cross-peaks totally disappear in the presence of 0.05 mM MnCl 2 . Unlabeled residues correspond either to a partial broadening of their proton signal at 0.05 mM MnCl 2 , to the lack of information because of the proximity of their H␣ chemical shift with that of water, or to poorly resolved signals in crowded regions. The thick line indicates the helical segment. Residues corresponding to the putative flexible helical region are underlined. B, tentative positioning of the ␣-helix (ribbon diagram) at the interface between polar heads and hydrophobic tails of phospholipids. As the average structure of NS5A  was represented here (PDB entry 1R7E), the position of side chain residues is only indicative (note, e.g. that Trp 4 seems to be buried in the membrane although it was found to be poorly accessible to the hydrophobic tail of DPC, see A above). The phospholipid bilayer was drawn using the phosphatidylethanolamine models reported in Protein Data Bank entry 1BCC. A phosphatidylethanolamine molecule colored according to atom types (N, blue; O, red; P, yellow; C, H, gray) is given on the right to illustrate the polar head/hydrophobic tail interface. C, top view of the peptide embedded at the interface of a model phospholipid membrane.
FIG. 5. NS5A membrane anchor mutants. A, amino acid sequences of NS5A membrane anchor mutants. The amino acid conservation among different HCV isolates (20) is indicated by an asterisk (*), a colon (:), and a dot (.) for fully conserved, conserved, and similar residues, respectively. B and C, space-filling representation of theoretical helices for mutants 8ϩA and 11ϩA. An Ala insertion (shown in magenta) twists the helix by 110°. The C-terminal side of the helix is shown in the same orientation as in Fig. 3B to highlight the distortion of charged residues on the N-terminal side. Residues are colored as in Fig. 3B. These models were constructed by using the NMR average structure of NS5A  observed in 100 mM SDS as template and SwissPdb Viewer program (www.expasy.ch/swissmod/). preserve the overall fold of the amphipathic ␣-helix and taking into consideration the conservation of residues among different HCV isolates (20). Mutation of two charged residues at position 6 and 7 to Ala resulted in construct R6A/D7A. The absolutely conserved Trp 9 and Cys 13 were replaced by Ala, yielding constructs W9A and C13A, respectively. To perturb the asymmetric charge distribution on the hydrophilic side of the helix Ala residues were inserted at position 8 or 11, yielding constructs 8ϩA and 11ϩA, respectively. These insertions twist the helix by 110°, as illustrated in Fig. 5, B and C. Finally, construct ⌬5-11 was obtained by the deletion of two ␣-helix turns within the membrane anchor domain.
Subcellular localization of the NS5A mutants was analyzed by confocal laser scanning microscopy after transient expression in U-2 OS cells. As shown in Fig. 6, all membrane anchor mutants were found in a reticular staining pattern, which surrounded the nucleus, extended through the cytoplasm, and included the nuclear membrane. No nuclear or plasma membrane staining was observed. The staining pattern was indistinguishable from that of wild-type NS5A, which is associated with membranes of the ER or an ER-derived modified compartment (20). By contrast, an expression construct lacking the N-terminal domain of NS5A (NS5A45-448) showed a diffuse staining pattern with accumulation in the nucleus as reported previously (20). Thus, the carefully designed membrane anchor mutants retained their proper subcellular localization and apparent membrane association.
RNA Replication of NS5A Membrane Anchor Mutants-The above mutations were introduced into a reporter replicon carrying two cell culture-adaptive changes in NS3 and one in NS4B to analyze their effect on HCV RNA replication (Fig. 7A). In these replicons, translation of the firefly luciferase gene is directed by the internal ribosome entry site of poliovirus, whereas the internal ribosome entry site of encephalomyocarditis virus governs translation of the HCV nonstructural proteins. RNA replication was analyzed by monitoring luciferase activity at various time points after transfection into Huh-7 cells. The parental replicon and a replication-deficient RNA (GND) served as positive and negative controls, respectively. A representative result is shown in Fig. 7B. Single and double Ala substitutions for conserved amino acids located on the cytosolic side of the helix (R6A/D7A and C13A) had no influence on replication efficiency. The Ala substitution for the fully conserved Trp 9 decreased luciferase activity by about 10-fold. Insertion of Ala residues at amino acid position 8 or 11 (8ϩA and 11ϩA, respectively) abolished HCV RNA replication. The effect of these mutations was as dramatic as that observed for a deletion of two helix turns (⌬5-11).
In some experiments, luciferase activities obtained with the mutants 8ϩA, 11ϩA, and ⌬5-11 were slightly higher than that of the GND negative control (Fig. 7B). Although this difference was within the range of assay variability, these mutants were reanalyzed by a more sensitive replication assay using selectable replicons in which the luciferase reporter gene was replaced by the gene encoding for neomycin phosphotransferase. Upon transfection into Huh-7 cells and G418 selection, the number of G418-resistant colonies reflects the replication efficiency of a given replicon. As shown in Fig. 7C, no colonies were obtained with these mutants under these conditions, corroborating that the introduced alterations completely block RNA replication. In conclusion, these results indicate that insertions or deletions, leading to a twist of the helix or altering its size, have a dramatic impact on the function of NS5A in HCV RNA replication.
Membrane Association of NS5A Membrane Anchor Mutants-Membrane flotation analyses were performed to confirm membrane association of the replication-defective mutants 8ϩA, 11ϩA, and ⌬5-11. Mutants were expressed transiently in U-2 OS cells, followed by hypotonic cell lysis and equilibrium centrifugation in a Nycodenz gradient. Under these conditions, membrane proteins float to the upper, low density fractions, as indicated by the behavior of p63, an integral ER membrane protein (Fig. 8). As expected, wild-type NS5A floated to the membrane-containing low density fractions. Disruption of the membranes by 1% Triton X-100 results in sedimentation of NS5A and p63 into high density fractions. As shown in Fig. 8, the mutants 8ϩA, 11ϩA, and ⌬5-11 floated to the low density fractions as wild-type NS5A, confirming the unimpaired membrane association of these constructs.
More detailed analyses of 8ϩA, 11ϩA, and ⌬5-11 by differential membrane extraction revealed that deletion of 2 helix turns (⌬5-11) and, to a lesser degree, distortion of the helix by Ala insertion, attenuated the strength of membrane association, as reflected by an increased proportion of protein extracted into the supernatant fraction after alkali or 4 M urea treatment (Fig. 3 of the Supplemental Data). However, the 8ϩA and 11ϩA mutants fulfilled the criteria of integral membrane proteins. These results suggest that the dramatic effect of these mutations on HCV RNA replication is not primarily the result of an impaired membrane association of NS5A but rather impairment of additional functions of the N-terminal ␣-helix.

DISCUSSION
Three-dimensional NMR structure analyses of the NS5A  peptide performed in different membrane mimetic media revealed that the N-terminal membrane anchor domain of NS5A includes a long amphipathic ␣-helix (amino acids  which seems to be divided into two portions separated by a putative flexible region located in the center of the helix (amino acids [15][16][17]. By three different NMR analyses, the hydrophobic amino acids were found to be mainly buried in the hydrophobic core of the detergent-peptide micelles, whereas the polar and charged amino acids were mainly accessible at their surface. One can thus conclude that the NS5A membrane anchor domain is embedded in-plane in the cytosolic leaflet of the ER membrane, with a hydrophobic side buried in the membrane and a polar/charged side accessible from the cytosol. The Trprich, N-terminal portion of the helix (amino acids 5-12) seems to be located more closely to the membrane interface, whereas the Lys-rich, C-terminal portion (amino acids 19 -26) appears to be buried slightly more deeply into the hydrophobic core of the membrane.
The location of the Trp residues at the interface between the hydrophilic and hydrophobic sides of the amphipathic helix is a very typical feature. Indeed, in membrane proteins Trp residues are often located at the lipid bilayer interface but rarely within the hydrophobic core of the membrane (39,40). For example, in the membrane-proximal Trp-rich region of human immunodeficiency virus glycoprotein gp41 four of five Trp residues and one Tyr residue form a collar of aromatic residues along the axial length of the amphipathic helix, resulting in a Velcro-like interaction on the outer viral membrane (37). The full conservation of Trp residues in the NS5A amphipathic ␣-helix among the various HCV genotypes (20) suggests that in addition to ensure membrane association they may also participate in specific protein-protein interactions at the level of the membrane interface, possibly with other components of the HCV replication complex. The polar side of the ␣-helix exhibits a striking asymmetry in charge distribution with positively charged residues (Arg 6 , Lys 20 , and Lys 24 ) placed in-line along one border and negatively charged residues (Asp 7 , Asp 10 , Glu 14 , and Asp 18 ) along the other. This particular arrangement of these remarkably conserved charged residues (20) suggests that the polar side of the helix forms a specific interaction site. Hence, the helix appears as a multi-interaction platform with both the hydrophilic side and the borders of the hydrophobic side likely involved in specific protein-protein interactions.
No dimerization or oligomerization of the NS5A  peptide was observed, indicating that evolution toward an oligomeric transmembrane structure is unlikely and that the NS5A amphipathic helix remains a monomeric structural element located at the surface of the ER membrane. This is in line with a lack of glycosylation, i.e. lack of ER luminal translocation, of artificial glycosylation acceptor sites engineered to the N terminus of NS5A. 3 Based on the above structural features and the observation that NS5A expressed alone behaves as an integral membrane protein (20), we propose a scenario where the rest of NS5A folds onto the polar side of the N-terminal amphipathic ␣-helix membrane anchor. The three-dimensional structures of two other monotopic membrane proteins with a similar organization have been described, namely prostaglandin H synthase and squalene cyclase. In both cases, integral membrane association is mediated by in-plane amphipathic ␣-helices. Prostaglandin H synthase isoforms 1 and 2 bind to the ER membrane via four short in-plane amphipathic ␣-helices that form horseshoeshaped nonpolar protrusions and interact with one leaflet of the phospholipid bilayer (41)(42)(43). The membrane anchoring region of squalene cyclase forms a flat nonpolar plateau consisting of one amphipathic ␣-helix, one loop, and one segment between two helices (44). Thus, in both cases there are a few additional contacts between the protein and the membrane besides the amphipathic helix. In line with this notion, the hydrophobic and basic residues following the NS5A amphipathic ␣-helix could participate in membrane association. In addition, we recently found that a construct comprising the 3 V. Brass and D. Moradpour, unpublished data.
FIG. 8. Membrane association of NS5A mutants. Hypotonic lysates of U-2 OS cells transiently transfected with pCMVNS5Acon, -8ϩA, -11ϩA, and -⌬5-11 were analyzed by centrifugation through Nycodenz gradients as described under "Experimental Procedures." Fractions were collected from the top and analyzed by immunoblot using mAbs 11H against NS5A and G1/296 against p63. first 30 amino acids of NS5A fused to the N terminus of green fluorescent protein associates with ER membranes but is more easily extracted from these membranes when compared with the full-length NS5A protein. 3 This suggests that some other residues or segments located downstream of amino acids 1-31 interact with the membrane. Hence, although the N-terminal amphipathic ␣-helix is the main structural element anchoring NS5A to the membrane, the membrane binding region is likely more extended to ensure stable membrane association. In both prostaglandin H synthase and squalene cyclase, interactions between the membrane anchor and the rest of the protein are mainly ensured by charged and polar residues that define interaction sites on both parts of these proteins. A similar arrangement likely occurs in NS5A, as suggested by the platform of hydrophilic residues and the asymmetric distribution of charged residues on the cytosolic side of the N-terminal ␣-helix.
It was reported recently that disruption of the amphipathic nature of the N-terminal helix of NS5A by introduction of several charged residues into the hydrophobic side abolished membrane association of NS5A and, by consequence, HCV RNA replication (45). The more subtle mutational analysis reported here revealed that our NS5A membrane anchor mutants, including the 8ϩA and 11ϩA mutants defective in RNA replication, show proper subcellular localization and membrane association, with only slightly reduced strength of membrane binding. Moreover, polyprotein processing and incorporation of the mutants into dot-like structures, which represent membranous webs harboring HCV replication complexes (5, 6), were not affected when they were expressed in the context of a NS3-5B polyprotein (data not illustrated). However, structurefunction analyses of these mutants, which for HCV have become possible only recently through the development of the replicon system (14,29), demonstrated that perturbation of the relative positioning of conserved residues by Ala insertion at position 8 or 11, leading to a twist of the ␣-helix by 110°(see Fig. 5, B and C), completely abrogated RNA replication without significantly affecting membrane association. The discrepancy between RNA replication and membrane association phenotypes of these mutants suggests that the N-terminal region of NS5A has functions beyond serving as a membrane anchor and supports the idea of an essential interaction platform formed by the polar side of the amphipathic ␣-helix. In contrast, Ala replacement of selected residues (R6A/D7A, W9A, and C13A) had no or only moderate effect on HCV RNA replication. The absolutely conserved Cys 13 could be either involved in a disulfide bond or form a link to a lipid. In both cases, replacement by Ala should result in only a limited destabilization of protein folding and/or protein-protein or protein-membrane interactions. Similarly, replacement of either Arg 6 and Asp 7 or Trp 9 by Ala may be too subtle to destabilize essential interactions and may be compensated by the remaining residues involved in these interactions. Alternatively, the unimpaired replication of the R6A/D7A, W9A, and even C13A mutants raises the possibility that the N-terminal portion of the helix may not be required for an intramolecular interaction, as discussed above, but may be involved in intermolecular interactions that are not essential for replicon activity.
Interestingly, when the NS5A membrane anchor mutants were expressed in the context of a NS3-5B polyprotein and analyzed with respect to their phosphorylation state, it was found that the replication-defective mutants 8ϩA, 11ϩA, and ⌬5-11 were not hyperphosphorylated in contrast to the mutants R6A/D7A, W9A, and C13A. 4 This concordance between preserved hyperphosphorylation and RNA replication suggests that proper conformation of NS5A and/or the positioning of NS5A within the replication complex is altered in the replication-defective mutants, which in turn might interfere with kinase binding or accessibility of phosphoacceptor sites. Clearly, resolution of the three-dimensional structure of the entire NS5A protein will shed further light on these mechanisms.
A synthetic peptide representing the N-terminal amphipathic ␣-helix of NS5A was recently found to block membrane association of NS5A in vitro in a dose-dependent and sequence-specific fashion (45). The mechanism of inhibition remains to be determined, but this observation raises the interesting possibility of interfering with membrane association of NS5A as a novel therapeutic strategy. The detailed structure-function analyses reported here suggest the presence of distinct subsites (e.g. amino acids 5-15 and 17-25 segments) that will be interesting to pursue as antiviral targets.
Predicted amphipathic ␣-helices mediating membrane association have been described in other plus strand RNA viruses. For example, an amphipathic helix was predicted in the Nterminal region of poliovirus and hepatitis A virus 2C proteins (46 -49). However, no structural data have been reported for these picornaviral proteins. Investigation of Semliki Forest virus mRNA-capping enzyme nsp1 by NMR spectroscopy revealed a short amphipathic ␣-helix mediating monotopic membrane association (50). Sequence alignments suggested that a similar amphipathic ␣-helix is present in nsp1 of other alphaviruses. By analogy to HCV NS5A, it is reasonable to speculate that these amphipathic helices constitute the main structural determinants of membrane association and are essential for replication of the corresponding viruses. Hence, as for NS5A, these amphipathic helices constitute putative targets for the development of specific antiviral drugs.
In conclusion, the NS5A N-terminal in-plane amphipathic ␣-helix is not only a structural determinant for ER membrane targeting and binding via protein-phospholipid interactions but is also a multi-interaction binding platform thought to ensure the stabilization of NS5A folding and/or its positioning in the HCV replication complex via intermolecular interactions. These findings have implications for the functional architecture of the HCV replication complex and may define novel antiviral targets.