Structural Basis for Penetration of the Glycan Shield of Hepatitis C Virus E2 Glycoprotein by a Broadly Neutralizing Human Antibody*

Background: HCV uses glycan shielding to mask epitopes recognized by neutralizing antibodies. Results: The structure of a human antibody bound to an HCV E2 epitope revealed how it penetrates a shield created by glycosylation shifting. Conclusion: Antibody binding induces an epitope conformation that accommodates multiple glycans. Significance: The structure provides a template for engineering E2 to elicit protective antibodies able to overcome glycosylation shifting. Hepatitis C virus (HCV) is a major cause of liver cirrhosis and hepatocellular carcinoma. A challenge for HCV vaccine development is to identify conserved epitopes able to elicit protective antibodies against this highly diverse virus. Glycan shielding is a mechanism by which HCV masks such epitopes on its E2 envelope glycoprotein. Antibodies to the E2 region comprising residues 412–423 (E2412–423) have broadly neutralizing activities. However, an adaptive mutation in this linear epitope, N417S, is associated with a glycosylation shift from Asn-417 to Asn-415 that enables HCV to escape neutralization by mAbs such as HCV1 and AP33. By contrast, the human mAb HC33.1 can neutralize virus bearing the N417S mutation. To understand how HC33.1 penetrates the glycan shield created by the glycosylation shift to Asn-415, we determined the structure of this broadly neutralizing mAb in complex with its E2412–423 epitope to 2.0 Å resolution. The conformation of E2412–423 bound to HC33.1 is distinct from the β-hairpin conformation of this peptide bound to HCV1 or AP33, because of disruption of the β-hairpin through interactions with the unusually long complementarity-determining region 3 of the HC33.1 heavy chain. Whereas Asn-415 is buried by HCV1 and AP33, it is solvent-exposed in the HC33.1-E2412–423 complex, such that glycosylation of Asn-415 would not prevent antibody binding. Furthermore, our results highlight the structural flexibility of the E2412–423 epitope, which may serve as an immune evasion strategy to impede induction of antibodies targeting this site by reducing its antigenicity.

Hepatitis C virus (HCV), 2 a member of the Flaviviridae family of positive-stranded RNA viruses, infects at least 2% of the world population and is a major cause of liver cirrhosis, liver failure, and hepatocellular carcinoma (1,2). The global burden is estimated at 170 million infected individuals with an annual increase of 3-4 million new infections (3). The recent introduction of successful HCV-specific direct-acting antiviral agents represents a major advance in HCV therapy (4). However, the high cost of direct-acting antiviral agents precludes their availability to the large majority of HCV-infected individuals living in developing countries. In addition, health care workers with occupational risk for blood-borne pathogens and injection drug users will remain at risk for repeated exposure to HCV, even after successful direct-acting antiviral agents treatment. Thus, there is a compelling need for a preventive HCV vaccine.
The key challenge in HCV vaccine development is to overcome the high diversity of this virus and its potential to escape from host immune responses. HCV is composed of a nucleocapsid core enveloped by a lipid bilayer in which two surface glycoproteins, E1 and E2, are anchored. Entry of HCV into hepatocytes is mediated by interactions between the E1E2 heterodimer and at least four cellular receptors: the tetraspanin CD81 (5), scavenger receptor class B type 1 (6), and the tight junction proteins occludin (7) and claudin 1 (8). An effective vaccine must include conserved epitopes of E1E2 that are able to elicit broadly neutralizing antibodies. Accordingly, much effort has been focused on the identification of conserved regions of E1E2 mediating virus neutralization through the isolation and characterization of mAbs from HCV-infected individuals and experimentally immunized mice (9). The large majority of these mAbs recognize E2, which is more immunogenic than E1. In addition, E2 contains binding sites for CD81 and scavenger receptor class B type 1 (5)(6)(7)(8).
Immunogenic determinants on the E2 glycoprotein are roughly segregated into the hypervariable region 1 (10 -13) and at least five clusters of overlapping linear and nonlinear epitopes, designated antigenic domains A-E (14 -17). Hypervariable region 1 is a major decoy that elicits isolate-specific neutralizing antibodies from which the virus can readily escape (11)(12)(13). Antigenic domain A is associated with non-neutralizing antibodies and constitutes another major decoy (15,16,18). By contrast, some antigenic domain B epitopes and most D and E epitopes elicit antibodies that are broadly neutralizing among the major HCV genotypes and subtypes (15)(16)(17). Antibodies to domain C epitopes neutralize HCV with more restricted genotype and subtype profiles (19).
The extracellular domains of both E1 and E2 are heavily glycosylated, with up to 5 sites on E1 and up to 11 sites on E2 modified by N-linked glycosylation (30). Some of these glycans are critical for protein folding and/or HCV infectivity (31). In addition, a number of studies have highlighted the role of N-linked glycosylation in shielding E2 epitopes from recognition by broadly neutralizing antibodies (31)(32)(33), as observed for HIV and influenza virus (34). In a particularly revealing recent example, deep sequencing analysis of HCV resistance to neutralization by mAbs HCV1 and AP33 identified mutations at asparagine 417 of E2 (N417S and N417T), an N-glycosylation site within the E2 412-423 epitope recognized by these mAbs, as responsible for virus escape (35). Comparison of the glycosylation status of E2 containing these resistance mutations revealed a glycosylation shift from Asn-417 to Asn-415 in the N417S and N417T E2 proteins. Crystal structures of the wild-type E2 412-423 epitope bound to HCV1 and AP33 showed that, whereas the side chain of Asn-417 points away from the antibody, Asn-415 is buried within the peptide-antibody interface (35)(36)(37)(38). Consequently, attachment of a glycan at Asn-415, but not at Asn-417, would create major steric clashes with HCV1 and AP33, resulting in abrogation of antibody binding and HCV escape, as observed in resistance selection studies (29,35,39). Notably, liver transplantation patients with HCV who were treated with HCV1 showed rebounds in viral load at various times following transplantation (39). These rebounds were consistently associated with mutations in the E2 412-423 epitope known to confer resistance to HCV1, in particular N417S.
In sharp contrast to mAbs HCV1 and AP33, HC33.1, which was isolated from an HCV-infected blood donor, can neutralize virus bearing the E2 N417S and N417T adaptive mutations with the glycosylation shift to Asn-415 (14,28). Moreover, infectious HCV virions containing the N417S mutation displayed increased sensitivity to neutralization by HC33.1 and related antibodies (28), suggesting that the glycosylation shift actually enhanced binding. To understand how HC33.1 is able to penetrate the glycan shield of HCV created by this shift, we determined the crystal structure of this broadly neutralizing human mAb in complex with its E2 412-423 epitope.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The HC33.1 antibody was expressed as a single-chain Fv fragment (scFv) by in vitro folding from inclusion bodies produced in Escherichia coli. The scFv construct consisted of the heavy chain variable (V H ) region (residues Glu-1-Ser-127) connected to the light chain variable (V L ) region (residues Gln-1-Leu-110) by an 18-residue linker (GSTGGGGSGGGGSGGGGS). The HC33.1 scFv was cloned into the expression vector pET-26b (Novagen) and expressed as inclusion bodies in BL21(DE3) E. coli cells (Novagen). Bacteria were grown at 37°C in LB medium to an absorbance of 0.6 -0.8 at 600 nm and induced with 1 mM isopropyl-␤-D-thiogalactoside. After incubation for 3 h, the bacteria were harvested by centrifugation and resuspended in 50 mM Tris-HCl (pH 8.0) containing 0.1 M NaCl and 2 mM EDTA; cells were disrupted by sonication. Inclusion bodies were washed extensively with 50 mM Tris-HCl (pH 8.0) and 2% (v/v) Triton X-100 and then dissolved in 8 M urea, 50 mM Tris-HCl (pH 8.0), and 10 mM DTT. For in vitro folding, inclusion bodies were diluted into ice-cold folding buffer containing 1 M L-arginine HCl, 50 mM Tris-HCl (pH 8.0), 1 mM EDTA, 3 mM reduced glutathione, and 0.9 mM oxidized glutathione, to a final protein concentration of 60 mg/liter. After 72 h at 4°C, the folding mixture was concentrated 50-fold, dialyzed against 50 mM MES (pH 6.0), and centrifuged to remove aggregates. Correctly folded HC33.1 scFv was then purified using sequential Superdex 75 HR and MonoQ columns (GE Healthcare).
Crystallization and Data Collection-For crystallization of the HC33.1-E2 412-423 complex, HC33.1 scFv (10 mg/ml) was mixed with E2 412-423 peptide (GenScript) in a 1:5 molar ratio. The complex crystallized in sitting drops at room temperature in 20% (w/v) PEG 8000, 3% (v/v) 2-methyl-2,4-pentanediol, and 0.1 M imidazole (pH 6.5). For data collection, crystals were transferred to a cryoprotectant solution of mother liquor containing 20% (v/v) glycerol, prior to flash cooling in a nitrogen stream. X-ray diffraction data were collected in-house at 100 K using a Rigaku R-AXIS IV ϩϩ area detector. The HC33.1-E2 412-423 crystal belongs to space group P2 1 2 1 2 1 with one complex molecule per asymmetric unit. Diffraction data were indexed, integrated, and scaled with the program CrystalClear (Rigaku). The data set is 94.9% complete to 2.0 Å resolution with R merge ϭ 7.9%. Data collection statistics are shown in Table 1.
Structure Determination and Refinement-The structure of the HC33.1-E2 412-423 complex was solved by molecular replacement with Phaser (40) using an ensemble search model constructed by combining aligned structures of the anti-severe acute respiratory syndrome spike protein scFv 80R (Protein Data Bank accession code 2GHW) (41), the anti-gankyrin scFv F5 (Protein Data Bank accession code 4NIK) (42), and the anti-influenza hemagglutinin scFv F10 (Protein Data Bank accession code 3FKU) (43). After model building for HC33.1 scFv was almost complete, extra density corresponding to the E2 412-423 peptide was very clear, and peptide residues were built manually without any ambiguity. All refinements were carried out using CNS1.1 (44), including iterative cycles of simulated annealing, positional refinement, and B factor refinement, interspersed with model rebuilding into A -weighted F o Ϫ F c and 2F o Ϫ F c electron density maps using XtalView (45). Refinement statistics are summarized in Table 1. Stereochemical parameters were evaluated with PROCHECK (46).
Modeling of Epitope Mutants and Glycans-Rosetta version 2.3 was used to perform a computational alanine scan and escape mutant simulations of the E2 412-423 epitope using the interface mutagenesis module (47), removing the N-terminal arginine residue from the HC33.1-E2 412-423 complex structure prior to modeling to avoid any influence of nonepitope residues on results. Extra side chain rotamers were included to ensure sufficient sampling (-ex1 -ex2 -ex3), and a separate simulation for non-alanine mutants was performed incorporating backbone and off-rotamer side chain minimization of the complex before and after mutation (-min_interface -int_bb -int_chi). Glycans were modeled using the GlyProt web server (48), using oligomannose glycans because high mannose glycans were found to be predominant in native E2 (49). As with the Rosetta simulations, the engineered N-terminal arginine residue was removed from the peptide in the complex structure prior to input into the server.
Protein Data Bank Accession Code-Coordinates and structure factors for the HC33.1-E2 412-423 complex have been deposited under accession code 4XVJ.

RESULTS
Overview of the HC33.1-E2 412-423 Complex-We expressed the HC33.1 antibody as an scFv by in vitro folding from bacterial inclusion bodies. The scFv construct consisted of the V H region connected to the V L region by an 18-residue linker. The HC33.1 scFv was crystallized with a 13-mer peptide (RQLINT-NGSWHIN) representing the E2 412-423 epitope to which an N-terminal arginine was added to increase solubility. We determined the structure of the HC33.1-E2 412-423 complex to 2.0 Å resolution by molecular replacement using an ensemble search model built from three scFv structures (41)(42)(43) (Table 1 and Fig. 1, A and E). Clear electron density corresponding to the E2 412-423 peptide was found in the binding site, as evident from the 2F o Ϫ F c electron density map (Fig. 2). All scFv residues were visible, except for six residues in the linker connecting V H and V L . The antibody-peptide interface was unambiguous.
The conformation of the E2 412-423 peptide bound to HC33.1 (Fig. 3A) is very different from the ␤-hairpin conformation of this same peptide bound to HCV1 (36) or AP33 (37, 38) (Fig.  3B) or to the humanized and affinity-matured mAbs MRC10.v362 (derived from AP33) and hu5B3.v3 (35). It is also distinct from the extended conformation of E2 412-423 observed in the complex with mAb 3/11 (Fig. 3C) (50). This rat antibody, like HCV1 or AP33 but unlike HC33.1, cannot recognize E2 with a glycosylation shift from Asn-417 to Asn-415 (29). In the HC33.1-E2 412-423 complex (Fig. 1, A and E), residues 414 -415 of the peptide form an anti-parallel ␤-sheet with ␤-strand F of the V H domain, whereas the rest of the peptide adopts a coil conformation. In the HCV1-E2 412-423 (Fig. 1, B and F) and AP33-E2 412-423 (Fig. 1, C and G) complexes, by contrast, these residues form part of the first strand of the ␤-hairpin (36 -38). In the HC33.1-E2 412-423 structure, the N-terminal coil preceding the ␤-strand of the bound peptide (residues 412-413) extends behind the FG ␤-sheet of the HC33.1 V H domain (Fig.  1A). The peptide makes a turn at residues 416 -419, surrounding the F strand of V H , whereas its C terminus sits loosely in front of the FG ␤-sheet. The peptide interacts predominantly with the HC33.1 H chain, which mediates 90% of the 186 total contacts. Interactions with the L chain are restricted to the side chains of Leu-413, Trp-420, and His-421 at the N and C termini of E2 412-423 . HC33.1 has an unusually long V H complementarity-determining region 3 (CDR3) (22 residues compared with 14 and 18 residues for AP33 and HCV1, respectively). In particular, the C terminus of strand F and N terminus of strand G, which constitute part of this V H CDR3, are each 3-4 residues longer in HC33.1 than in HCV1 or AP33. The extended F strand permits E2 412-423 to "clip" onto the antibody through formation of the anti-parallel ␤-sheet described above (Fig. 1A). V H CDR1 and V H CDR2 mainly contact the ␤-turn of the bound peptide.
Conformation of the E2 412-423 Epitope-Because the conformation of E2 412-423 bound to HC33.1 (Fig. 3A) is distinct from the ␤-hairpin or extended conformations of this peptide bound to other mAbs (Fig. 3, B and C), it would be of interest to know the conformation of unbound E2 412-423 in the context of the native E2 protein. However, residues 412-420 were disordered in one of the two recently reported E2 core structures (51) and were not included in the crystallization construct used for the other E2 structure (52). Accordingly, we used several prediction methods to assess the preferred unbound conformation(s) of the E2 412-423 epitope. Peptide structure prediction using the PEP-FOLD server, an ab initio peptide folding method, resulted in models with a ␤-hairpin conformation very similar to that in the HCV1 and AP33 complexes (36 -38) with a type IЈ ␤-turn at residues 416 -419 and a backbone root mean square distance (RMSD) of 2.3 Å between the HCV1-bound epitope and the top model. The I-TASSER server, with homologous templates excluded to prevent usage of E2 412-423 -containing structures as templates, produced a model that was nearly identical to the HCV1-bound ␤-hairpin (1.0 Å RMSD). Given these modeling results, in addition to the existence of the ␤-hairpin in the majority of mAb-bound structures, E2 412-423 likely adopts a ␤-hairpin conformation in the E2 protein and maintains a similar structure after binding to AP33 or HCV1. However, this ␤-hairpin structure can be disrupted through interactions with certain antibodies, such as 3/11 (50) or HC33.1. We next asked whether any epitope substructures are shared between the ␤-hairpin structure bound by HCV1 or AP33 and the HC33.1-bound structure. Indeed, secondary structural analysis identified a conserved ␤-turn conformation at residues 416 -419 (TNGS) of E2 412-423 in complex with each of these mAbs. Superposing the HC33.1-bound peptide onto the HCV1-bound peptide through residues 416 -419 gave an RMSD of 0.35 Å in the position of their C␣ atoms (RMSD ϭ 0.34 for a comparison of the HC33.1-and AP33-bound peptides), indicating close similarity (Fig. 3D). An analogous comparison of the HC33.1-bound and 3/11-bound E2 412-423 epitopes failed to identify regions of comparable substructural similarity. This result, together with the ␤-strand preceding the 416 -419 turn in the HC33.1, HCV1, and AP33, but not 3/11, complexes (Fig. 1, A-D), implies that the HC33.1-bound conformation of E2 412-423 is structurally intermediate between the HCV1/AP33-and 3/11-bound conformations.
In a previous study (14), alanine-scanning mutagenesis was used to identify three E2 412-423 residues as critical for binding of the E2 glycoprotein to HC33.1: Leu-413, Gly-418, and Trp-420. The HC33.1-E2 412-423 complex readily explains this result, because these epitope residues constitute key anchor points in the structure. Thus, the side chains of Leu-413 and Trp-420 point directly into the HC33.1 binding site, with Leu-413 92% buried and Trp-420 completely buried upon complex formation. Indeed, computational alanine scanning (47) of the HC33.1-E2 412-423 complex structure identified Leu-413 and Trp-420 as critical binding residues ( Table 2). The isobutyl group of Leu-413 is oriented parallel to the aromatic ring of V L CDR1 Tyr-33, with which it makes extensive hydrophobic interactions (Fig. 4A). In addition, the C␦1 atom of Leu-413 contacts the side chain of V L CDR2 Asp-51, whereas its C␦2 atom contacts the main chains of V H CDR3 Ser-102 and Ser-103 in the unusually long ␤-strand F of HC33.1. Together, these four antibody residues form a tight, mainly hydrophobic cage encasing Leu-413, which probably accounts for the extreme sensitivity of HC33.1 to even conservative substitutions (e.g. Ile) at this position (28). Computational simulations of this mutant ( Table 2) showed substantial predicted affinity loss, whether or not nearby residues were permitted to minimize to accommodate the Ile side chain; only the C␥2 methyl group fit into the pocket noted above, yielding a loss of favorable HC33.1 contacts.
The side chain of Trp-420 binds deep in a ␤-barrel composed of V H strands C, CЈ, CЉ, F, and G and V L strands F and G (Fig.  4C). It makes numerous hydrophobic interactions with side chains of residues lining the ␤-barrel, in particular Trp-47, Tyr-59, and Lys-112 from V H strands CЈ, CЉ,and G, respectively, and Trp-92 and Val-99 from the V L FG loop. These interactions would be mostly lost upon alanine substitution of Trp-420, in agreement with the finding that this mutation abolished binding of E2 to HC33.1 (14). Additionally, Trp-420 makes two main chain-side chain hydrogen bonds with HC33.1: Trp-420 N-O␥ V H CDR2 Ser-52 and Trp-420 O-N V H CDR3 Lys-112 (Fig. 4C).
Like Trp-420, Gly-418 is completely buried in the HC33.1-E2 412-423 interface, where it makes 21 close contacts with the backbones of V H CDR1 Asn-31 and Phe-32 and of V H CDR2 Ser-52, Ser-53, and Ser-54 (Fig. 4B). Modeling the structure of the G418A mutant indicated that the alanine methyl group could in principle be accommodated at this position without substantially clashing with HC33.1, as reflected by its predicted ⌬⌬G (Table 2). However, Gly-418 is located at the tip of the type IЈ ␤-turn of the E2 412-423 peptide (Fig. 3D), and glycine is a highly preferred amino acid at this position for this type of ␤-turn (54). Accordingly, substitution of Gly-418 by other residues could adversely affect folding of the ␤-turn, which is wedged between the V H CDR1 and V H CDR2 loops (Fig. 4B).
Accommodation of the Glycosylation Shift to Asn-415 of E2 412-423 -Escape of HCV from neutralization by the E2 412-423 -specific mAbs HCV1, AP33, and 3/11 is the result of adaptive mutations at Asn-417 (N417S and N417T) that cause a glycosylation shift from Asn-417 to Asn-415 in the E2 protein (29,35,39). In marked contrast to these mAbs, HC33.1 can penetrate the glycan shield at Asn-415 originating from the glycosylation shift (14,28). We do not expect that the N417S or N417T mutation would affect the overall conformation of the E2 412-423 epitope or the orientation of Asn-415 in the complex with HC33.1. First, the side chains of Asn-417 and Asn-415 are each ϳ60% solvent-exposed in the HC33.1-E2 412-423 structure. Second, serine and threonine, like asparagine, are polar residues whose side chains prefer to project into solvent.
The ability of HC33.1 to accommodate the glycosylation shift to Asn-415 is explained by the unique conformation of E2 412-423 induced (or captured) by this mAb, rather than by HC33.1 binding in a different way to the ␤-hairpin or extended conformation of E2 412-423 recognized by HCV1, AP33, or 3/11. In the HCV1-E2 412-423 and AP33-E2 412-423 complexes, the buried Asn-415 residue is part of the first strand of the E2 412-423 ␤-hairpin (Fig. 1, F and G) (36 -38), whereas in the HC33.1-E2 412-423 complex, Asn-415 is part of a strand that forms an anti-parallel ␤-sheet with strand F of V H , such that the Asn-415 side chain flanks the antibody and is 58% solvent-accessible (Fig. 1E). A glycan chain can be attached to the side chain amide of Asn-415, based on modeling using the GlyProt server (48), indicating that glycosylation of E2 at Asn-415 would not prevent binding of HC33.1 (Fig. 5C). Indeed, a glycan at this position could mediate additional productive interactions with the mAb, resulting in higher affinity, as suggested by the finding that the neutralization potency of HC33.1 against glycanshifted HCV virions is Ͼ10-fold greater than against wild-type virions (28). Although the simulated oligomannose glycan at Asn-415 was largely directed away from the antibody because of Asn-415 side chain movement by the GlyProt server (48), we generated an alternate conformation of the Asn-415 glycan by using the E2 412-423 structure alone as input to GlyProt (removing mAb HC33.1) (Fig. 5D). This resulted in a modeled Asn-415 side chain conformation closer to the antibody-flanking conformation observed in the crystal structure and attached glycan with substantial antibody contacts, including ones with V H CDR1, V H CDR3, and the L chain framework.
Importantly, the side chains of Asn-417 and Asn-423 are also solvent-exposed (60 and 95%, respectively) in the HC33.1-  a The Rosetta ⌬⌬G score was calculated with rotamer-based packing of the mutant residue, and for non-alanine mutants, backbone, and off-rotamer side chain, minimization of mutant and interface residues was also performed (in parentheses). Predicted substantially destabilizing mutants (⌬⌬G scores Ͼ1.5) are shown in bold type.
E2 412-423 structure, which would permit attachment of glycan chains to these residues as well, without interfering with mAb binding. Modeling of the Asn-417 and Asn-423 glycans using GlyProt (48) supports this (Fig. 5C). Thus, HC33.1 binding to the flexible E2 412-423 epitope induces a conformation that allows accommodation of multiple glycans in the interface, whether at Asn-417 and Asn-423 in wild-type E2 or at Asn-415 and Asn-423 in glycosylation-shifted E2 mutants.

DISCUSSION
We have identified the structural basis for human antibody recognition of HCV E2 412-423 that is capable of neutralizing clinically observed glycan shift mutants in this linear epitope. This provides new insights into the capability of antibodies such as HC33.1 to effectively target HCV, thereby overcoming previously identified escape mechanisms through mutation or glycan attachment at E2 residue 415 (35). Based on initial structures of mAbs bound to E2 412-423 , it appeared that this epitope was restricted to a ␤-hairpin conformation (36 -38). However, our study has revealed a third distinct conformation of E2 412-423 , intermediate between a ␤-hairpin and coil (50), that highlights the remarkable structural heterogeneity of this epitope.
Along with sequence variability and N-glycosylation, conformational flexibility has been proposed as a mechanism used by viruses such as HIV and influenza to evade humoral immunity (34). Conformational flexibility is thought to pose a barrier to effective elicitation of broadly neutralizing antibodies to a viral site because antibodies are unable to simultaneously target many possible epitope conformations (55,56). For example, murine norovirus was found to escape neutralization by antibodies against the capsid protein by undergoing mutations in two flexible loops of the capsid that switch their conformation to one with poor stereochemical complementarity to neutralizing antibodies (57).
Our results and those of others (50,51,58) suggest that HCV may also employ structural flexibility as an immune evasion strategy. The crystal structure of the HCV E2 core revealed that ϳ60% of all residues are either disordered or in loops, implying considerable overall flexibility (51). Neutralizing and non-neutralizing mouse mAbs specific for an E2 epitope comprising residues 427-446 were found to bind distinct conformations of this epitope that determined recognition specificity (58). Here we have shown that the E2 412-423 epitope can adopt at least three different conformations, which may contribute to reducing its immunogenicity in HCV-infected individuals (14).
Although structural flexibility may render E2 412-423 less antigenic, such flexibility could also be disadvantageous to HCV, insofar as conformational variability provides the immune system with multiple potential targets for antibody neutralization. Thus, if the E2 412-423 epitope were restricted to the ␤-hairpin conformation recognized by HCV1-like antibodies, it could not be targeted by HC33.1-like antibodies. As a consequence, the immune system would be unable to counter HCV variants that have undergone the Asn-415 glycosylation shift.
The HC33.1-E2 412-423 structure explains the ability of HC33.1 to accommodate glycans at Asn-417 and Asn-423 in wild-type E2 or at Asn-415 and Asn-423 in glycosylationshifted E2 mutants. A possible additional glycan has been proposed for the mutation S419N, which was observed in the context of other mutations, notably N417T, in a cell-based study of HC33.1 escape mutants (28). In combination, the S419N and N417T mutations may generate a new glycan at position 419 because of creation of a Thr-Xaa-Asn glycosylation sequon (59). Although far less common than the Asn-Xaa-Ser/Thr sequon, the Thr-Xaa-Asn sequon is one of several reported alternative motifs for N-glycosylation. Although the GlyProt server (48) does not recognize this sequon, in silico mutagenesis of S419N and N417T permitted modeling of a glycan at Asn-419. As expected from the high solvent exposure (60%) of the wild-type Ser-419 side chain in the HCV1-E2 412-423 structure (Fig. 1E), this glycan, if present, is not predicted to abolish antibody binding completely. Indeed, N417S/S419N escape variants could still be neutralized by HC33.1, albeit at higher antibody concentrations, and HC33.1 retained significant binding to recombinant E1E2 bearing these mutations (28). Apart from affecting interactions with HC33.1 directly, a glycan at Asn-419 may influence the structure or conformational dynamics of the E2 412-423 epitope, as noted in studies of peptide dynamics (60), or interact with other regions of E2. Another possibility is that the side chain change of Ser-419 to Asn has a direct effect on HC33.1 recognition; this would be supported by the hydrogen bond of Ser-419 with Asn-54 on the HC33.1 H chain, as well as some predicted loss of affinity upon in silico mutation (Table 2).
From the standpoint of vaccine design, the HC33.1-E2 412-423 structure provides critical information regarding a well conserved site and a new point of viral vulnerability. Having demonstrated how in principle antibodies can effectively target this epitope without including residue 415 in the inter- face core, vaccine design efforts can attempt to present this epitope structure to the immune system, rather than the likely preferred ␤-hairpin. Given that additive neutralization was observed between antibodies targeting this site (including HC33.1) and antibodies targeting another well conserved site that also contains residues critical for binding CD81 (14), simultaneously inducing robust immune responses to these epitope sites while minimizing the chances of escape (including glycan shifting to Asn-415) is key to an effective HCV vaccine.