Antibody Response to Hypervariable Region 1 Interferes with Broadly Neutralizing Antibodies to Hepatitis C Virus

ABSTRACT Hypervariable region 1 (HVR1) (amino acids [aa] 384 to 410) on the E2 glycoprotein of hepatitis C virus contributes to persistent infection by evolving escape mutations that attenuate binding of inhibitory antibodies and by blocking access of broadly neutralizing antibodies to their epitopes. A third proposed mechanism of immune antagonism is that poorly neutralizing antibodies binding to HVR1 interfere with binding of other superior neutralizing antibodies. Epitope mapping of human monoclonal antibodies (HMAbs) that bind to an adjacent, conserved domain on E2 encompassing aa 412 to 423 revealed two subsets, designated HC33 HMAbs. While both subsets have contact residues within aa 412 to 423, alanine-scanning mutagenesis suggested that one subset, which includes HC33.8, has an additional contact residue within HVR1. To test for interference of anti-HVR1 antibodies with binding of antibodies to aa 412 to 423 and other E2 determinants recognized by broadly neutralizing HMAbs, two murine MAbs against HVR1 (H77.16) and aa 412 to 423 (H77.39) were studied. As expected, H77.39 inhibited the binding of all HC33 HMAbs. Unexpectedly, H77.16 also inhibited the binding of both subsets of HC33 HMAbs. This inhibition also was observed against other broadly neutralizing HMAbs to epitopes outside aa 412 to 423. Combination antibody neutralization studies by the median-effect analysis method with H77.16 and broadly reactive HMAbs revealed antagonism between these antibodies. Structural studies demonstrated conformational flexibility in this antigenic region, which supports the possibility of anti-HVR1 antibodies hindering the binding of broadly neutralizing MAbs. These findings support the hypothesis that anti-HVR1 antibodies can interfere with a protective humoral response against HCV infection. IMPORTANCE HVR1 contributes to persistent infection by evolving mutations that escape from neutralizing antibodies to HVR1 and by shielding broadly neutralizing antibodies from their epitopes. This study provides insight into a new immune antagonism mechanism by which the binding of antibodies to HVR1 blocks the binding and activity of broadly neutralizing antibodies to HCV. Immunization strategies that avoid the induction of HVR1 antibodies should increase the inhibitory activity of broadly neutralizing anti-HCV antibodies elicited by candidate vaccines.

U p to 170 million people worldwide are infected with hepatitis C virus (HCV), with significant risk for liver failure and hepatocellular carcinoma. The World Health Organization estimates an annual increase in the global burden by two million new infections (1), and in the United States HCV is increasing in young adults from injection drug use (2). Multiple lines of evidence suggest that CD4 ϩ and CD8 ϩ T cell responses are needed to control acute infection but insufficient to prevent long-term persistence (3). Accumulating data indicate that neutralizing antibodies are an important correlate of HCV clearance. In chimpanzee studies, an infectious inoculum obtained during acute infection from a patient who eventually developed chronic HCV hepatitis could be neutralized by in vitro incubation with plasma of the same subject collected at 2 years after the initial infection (4). A neutralizingantibody response measured against pseudotyped retroviral particles expressing HCV E1E2 glycoproteins (HCVpp) has been associated with control of infection in single-source outbreaks of acute HCV infections (5). In addition, antibodies to HCV E2 prevent infection (6,7) and clear established infection (8) in a human liver-mouse chimeric model. In spite of the relationship between antibodies and protection against HCV infection, 10 4 to 10 6 virions per milliliter of serum can be detected during chronic infec-tion even in the presence of high levels of neutralizing antibodies in serum.
One driver of persistent viremia is a high degree of viral variants, or "quasispecies." From a viral replication rate of 10 12 copies per day with an error-prone NS5B RNA-dependent polymerase, the estimated mutation rate is 2.0 ϫ 10 Ϫ3 base substitutions per genome per year (9). A major determinant of antibody-mediated neutralization is the first 27 amino acids (aa 384 to 410), i.e., hypervariable region 1 (HVR1), located at the N terminus of HCV E2 (10). This E2 segment is highly immunogenic, and antibodies against HVR1 can be detected in the majority of HCV-infected individuals (11). However, antibodies to HVR1 over time select for viral variants that escape the existing antibody response (12). The limited nature of the B cell response to this region is shown in studies of HCV evolution from acute to chronic disease (13,14). Sequential autologous serum antibodies inefficiently neutralize emerging variants, in contrast to their capacity to neutralize earlier quasispecies. Viral escape is associated with mutations within HVR1. Other negative modulators of antibody-mediated neutralization include cell-to-cell transmission, virion-associated lipoproteins, virus envelope protein-associated glycans, and HVR1 itself; HVR1 can mask epitopes or limit access of virus-neutralizing antibodies to their epitopes (15)(16)(17). Indeed, virions lacking HVR1 are less susceptible to neutralization by anti-SR-B1 antibodies but are more sensitive to antibody-and soluble CD81mediated neutralization (18,19). It has been suggested that HVR1 partly shields CD81 binding sites and conserved epitopes mediating virus neutralization (18).
The concept of interfering antibodies is controversial, but this remains a possible mechanism that contributes to the development of persistent HCV infection in infected subjects (20)(21)(22)(23). In this study, we performed more extensive epitope mapping of a panel of human monoclonal antibodies (HMAbs) to a highly conserved region on the E2 glycoprotein encompassing aa 412 to 423, designated HC33-related antibodies (20). Mapping by binding against a library of alanine substitution E2 mutants with mutations from aa 384 to aa 446 revealed two subsets of HC33 HMAbs. Although both subsets have contact residues within aa 412 to 423, one subset included a contact residue within HVR1. This raised the possibility that anti-HVR1 antibodies interfere with the neutralizing activities of antibodies to aa 412 to 423. Two murine MAbs were utilized to explore the relationship between anti-HVR1 and antibodies against aa 412 to 423. Surprisingly, the anti-HVR1 antibody inhibited the binding and neutralization of both sets of HC33 HMAbs. This interference of binding and neutralization also was observed against other broadly neutralizing HCV HMAbs that have contact residues outside aa 412 to 423. Thus, immunization strategies that avoid the induction of HVR1 antibodies should increase the inhibitory activity of broadly neutralizing anti-HCV antibodies.
Binding to E2 glycoprotein. A standard enzyme-linked immunosorbent assay (ELISA) (28) was used to compare HMAb binding to wild-type (wt) and ⌬HVR1 HCV E2 glycoproteins. In some experiments, incubations were performed at a different temperature or in a different sequence when paired with H77. 16. Briefly, ELISA plates were coated with Galanthus nivalis lectin (GNA) and blocked with 2.5% nonfat dry milk and 2.5% normal goat serum in phosphate-buffered saline (PBS) supplemented with 0.1% Tween 20. Lysates of cells expressing HCV E2 glycoproteins were captured by GNA onto the microtiter plate, followed by incubation with HMAbs and then washing. Bound HMAb was detected by incubation with alkaline phosphatase-conjugated goat anti-human IgG (Promega; Madison, WI), followed by incubation with p-nitrophenyl phosphate (Sigma) for color development. Absorbance was measured at 405/ 570 nm. The assay was carried out in triplicate in three independent assays for each HMAb.
Inhibition assay. An ELISA measured the inhibition by mouse MAbs of HMAb binding to GNA-captured E2 glycoproteins (30). Briefly, microtiter plates were coated with GNA and blocked with 2.5% bovine serum albumin (BSA) and 2.5% normal goat serum in 0.1% Tween-PBS. Pretitrated mouse MAb was added to each well at a saturating concentration. After 1 h, HMAb was added at a concentration corresponding to 65% to 75% of the maximal optical density (OD) level, incubated for 30 min at either room temperature (RT) or 40°C, and then washed. Bound HMAb was detected as described above. This assay was carried out in triplicate, a minimum of two times for each HMAb.
Epitope mapping. Epitope mapping was performed using alanine substitution mutants with mutations in a defined E2 region (aa 404 to 425) by ELISA. Alanine substitution mutants were constructed in plasmids carrying the 1a H77C E1E2-coding sequence (GenBank accession no. AF009606), as previously described (31). All mutations were confirmed by DNA sequence analysis (Elim Biopharmaceuticals, Inc. Hayward, CA) for the desired mutation and for exclusion of unexpected residue changes in the full-length E1-E2-encoding sequence. The resulting plasmids were transfected into HEK293T cells for transient protein expression using the calcium phosphate method, as suggested by the manufacturer (Clontech, Mountain View, CA).
Antibody cooperativity for virus neutralization. Synergistic, additive, or antagonistic cooperativity by two antibodies for virus neutralization was evaluated by the median-effect analysis method, as described previously (32,33), using the CompuSyn software (ComboSyn Inc., Paramus, NJ). The approach takes into account the potency, the shape, and the slope of the dose-dependent neutralization curve of each antibody alone and in combination, at a constant ratio, to calculate a combination index (CI). A CI value of 1 indicates an additive effect, a value of Ͻ1 indicates synergism, and a value of Ͼ1 indicates antagonism. For each antibody, dose-dependent neutralization was measured initially to determine the concentration that resulted in 50% reduction (50% inhibitory concentration [IC 50 ]). The constant ratio of the combined antibodies was set by the IC 50 s of the two antibodies. Neutralization of a serial 2-fold dilution of each antibody alone and in combination was then measured in a range of concentrations above and below the IC 50 s. The measured neutralization values were entered in the program as fractional effect (FA) in the range 0.01 Ͻ FA Ͻ 0.99 for each of the two antibodies alone and in combination. The software determines the linear correlation coefficient, r, of each curve to indicate the fit or conformity of the data with respect to the medianeffect method and calculates the CI values in relation to FA values. The 1a HCVcc and each HMAb or combination of HMAbs were incubated for 1 h at 37°C and then plated onto Huh7.5 cell monolayers (3.2 ϫ 10 4 cells/ well) that were grown in 8-well chamber slides (Nalge Nunc, Rochester, NY) for virus neutralization assay as described previously (20). These assays were carried out in four replicates for each HMAb and combination of HMAbs.
Production and purification of recombinant Fab molecules HC33.4 and HC33.8. Synthetic genes that were codon optimized for Drosophila melanogaster encoding heavy and light chains of the Fab regions of each antibody were cloned into a Drosophila S2 Fab expression vector containing a double Strep tag for efficient affinity purification. Drosophila S2 cells were transfected with these plasmids as reported previously (34). For large-scale production, cells were induced with 5 M CdCl 2 at a density of approximately 10 7 cells/ml for 7 days and pelleted, and Fabs were purified by affinity chromatography from the supernatant using a Strep Tactin Superflow column, followed by size exclusion chromatography using a Superdex 200 column. Purified monomeric Fab was concentrated to approximately 25 mg/ml. Complex formation, crystallization, data collection, structure determination, and refinement. A synthetic peptide comprising residues 406 to 425 (PGAKQNIQLINTNGSWHINST) of the H77 strain was synthetized by GenScript (Ͼ98% purity) and dissolved in water at 10 mg/ml. Fab-peptide complexes were formed overnight at 277 K with 12 mg/ml Fab plus 2 mg/ml peptide. HC33.4-peptide complex crystals were grown at 293 K using the hanging-drop vapor diffusion method in drops containing 1 l complex solution (14 mg/ml in 10 mM Tris [pH 8.0], 150 mM NaCl) mixed with 1 l reservoir solution containing 100 mM sodium citrate (pH 5.2), 300 mM ammonium sulfate, 100 mM potassium phosphate, and 1 M lithium chloride. Diffraction-quality crystals of the HC33.8-peptide complex were grown at 293 K as described above using a seed stock derived from HC33.4-peptide complex crystals and reservoir solution containing 32% polyethylene glycol (PEG) 4000, 100 mM Tris-HCl (pH 8.5), and 800 mM lithium chloride with 0.5 l of seed stock solution. Rock-like crystals appeared after 1 week and were flash-frozen in mother liquor with 20% glycerol. Space groups and cell dimension refinement statistics are summarized in Table 1. Data were collected at the Synchrotron Soleil (Proxima 1) for the HC33.4 complex crystals and at the SLS (PX 1) for the HC33.8 complex crystals. Data were processed, scaled, and reduced using XDS (35) and programs from the CCP4 suite (36). The crystal structures of the Fab complexes were determined by the molecular replacement method using Phaser (37). The molecular replacement for Fab HC33.4 was performed using separate variable and constant regions of a hypothetical Fab fragment assembled from the light chain of PDB accession code 4JZO (89% amino acid identity) and the heavy chain of PDB accession code 3KDM (91% amino acid identity) as search model. The molecular replacement for Fab HC33.8 was performed using Fab HC33.4 as search model. Model building was performed using Coot (38), and refinement was done using AutoBuster (39). Difference maps calculated after refinement of the Fab molecules revealed an unambiguous side chain density for a tryptophan residue close to the complementarity-determining regions (CDRs), which allowed us to manually build a partial atomic model for the peptide comprising aa 418 to 421 (HC33.4 complex) and aa 415 to 421 (HC33.8).
Crystal structure analysis. Peptides were aligned using the Match-Maker algorithm implemented in Chimera (40) and an iterative alignment process pruning long atom pairs until no pair exceeds 1 Å. Root mean square deviations were calculated using Chimera. Buried solventaccessible surface areas for the interfaces as well as for individual residues within the peptides were calculated using the PISA server (41). Interactions were determined using the protein interactions calculator (PIC) (42). Figures were prepared with PyMOL (http://www.pymol.org).
Statistical analysis. Statistical tests were two sided, and P values (calculated by GraphPad software) below 5% are considered significant.
Protein structure accession numbers. The atomic coordinates and structure factors have been deposited in the Protein Data Bank (www.pdb .org) under accession numbers 5FGB and 5FGC.

RESULTS
HVR1 has different effects on blocking antibody access to epitopes on E2. HVR1 attenuates the activity of broadly neutralizing antibodies to HCV (18,19). Its ability to decrease the potency of virus-neutralizing antibodies was established with representative HMAbs to clusters of overlapping epitopes, designated antigenic domains B and C (18). To test whether HVR1 acts by shielding access of these and other broadly neutralizing antibodies to their respective epitopes, studies of binding to 1a H77C recombinant E2 protein with and without HVR1 (⌬HVR1) were performed ( Fig. 1A and B). Representative antigenic domain B antibodies (HC-1 and HC-11 [43]) bound to higher levels to ⌬HVR1 E2 than wt E2 but with different degrees of increase (86% and 14%; P Ͻ 0.05). Antigenic domain D antibodies (HC84.20, HC84.24, and HC84.26 [28]) also bound more to ⌬HVR1 E2 than to wt E2 (56 to 75%; P Ͻ 0.05). CBH-7, an antigenic domain C antibody (25), also bound more to ⌬HVR1 E2 but with only a minimal increase (6%; P Ͻ 0.05).
We next interrogated HMAbs that recognize linear epitopes located in a highly conserved region on E2, encompassing aa 412 to 423, designated antigenic domain E (20, 44), and we observed two different patterns. HC33.1 behaved similarly to antigenic domain B and D antibodies, with greater binding to ⌬HVR1 E2 than wt E2 (50%; P Ͻ 0.001) ( Fig. 1A and B). Unexpectedly, binding by three other E antibodies, HC33.4, HC33.8, and HC33.9, to ⌬HVR1 E2 was reduced compared to that with wt E2 protein (65 to 78%; P Ͻ 0.05). The decrease in binding of these MAbs to ⌬HVR1 E2 suggests that part of their epitopes may fall within HVR1.
Epitope mapping of antibodies to a highly conserved region on E2. Prior mapping of HC33-related HMAb epitopes was limited to aa 411 to 446 by alanine-scanning mutagenesis analysis of H77C E2 (20). That analysis revealed loss-of-binding residues located at L413, G418, and W420. We expanded on these results using alanine-scanning mutagenesis of aa 384 to 446 on E2, which included the entire HVR1 region (aa 384 to 410) (Fig. 1C). The E2 mutants were expressed in 293T cells, and binding by HC33.1, HC33.4, HC33.8, and HC33.29 to cell lysates was measured by ELISA. Expression levels of the mutants were normalized by binding with CBH-17, a nonneutralizing HMAb that recognizes a different linear epitope on HCV E2 (25). A test concentration of 50% maximum binding of each antibody was selected for epitope mapping studies. This was determined by dose-dependent binding of each HC33 HMAb to E2 that showed the test concentration in the linear portion of the binding curve (data not shown). A greater than 80% reduction in binding was observed again for all four HC33-related HMAbs with alanine substitutions at L413A, G418A, and W420A (Fig. 1C), which suggests that they bind to the same or to nearly identical epitopes. Three of these antibodies, HC33.4, HC33.8, and HC33.29, also showed greater than 80% reduction in binding against a K408A mutant E2 protein. Taken together, these mapping studies revealed two subsets of HC33related antibodies and explain why the deletion of HVR1 led to decreased binding by HC33.4, HC33.8, and HC33.29 but not HC33.1. Anti-HVR1 antibody blocks the binding of broadly neutralizing antibodies. Epitope mapping of HC33-related antibodies revealed two subsets. One set, HC33.1, has contact residues that are restricted to aa 412 to 423. This is similar to other MAbs to this region, e.g., AP33, HCV1, and H77.39 (29,45). The other set, HC33.4, HC33.8, and HC33.29, includes a putative contact residue within HVR1 that is located at aa 408. This raises the possibility that at least some anti-HVR1 antibodies can interfere with the functions of antibodies to aa 412 to 423 because of their spatial proximity. To test this hypothesis, blocking studies were performed with two murine MAbs, H77.39 and H77.16 (29). H77.39 binds to an epitope centered at aa 412 to 423, whereas H77.16 recognizes an epitope within HVR1 (29). Epitope mapping by alanine-scanning mutagenesis of H77C E2 from aa 384 to 446 confirmed that H77.39 recognizes residues (N415, G418, and W420) restricted to aa 412 to 423, whereas H77.16 binds to HVR1 residues (G406, K408, and N410) (Fig. 1C). Each of these murine  (25). The E2 region encompassing aa 384 to 446 was analyzed. Binding when the residue was replaced by alanine (or glycine at aa 407), relative to binding to wt, is indicated as follows: red, 0 to 20% binding; orange, 21 to 40%; brown, 41 to 60%; white, 61 to 100%; green Ͼ100%. Data are shown as mean values from two independent experiments performed in triplicate.
MAbs was tested for blocking of binding by representative antigenic domain B to E HMAbs (Fig. 2). H77C E2 was first incubated with either H77.39 or H77.16 prior to adding the test HMAb. As expected, H77.39 inhibited the binding to E2 of both subgroups of HC33 HMAbs (67 to 75%; P Ͻ 0.05) ( Fig. 2A). Inhibition also was observed against representative neutralizing HMAbs to two other epitope clusters. HC-1 and HC-11 (antigenic domain B) were inhibited by H77.39 between 73 and 74% (P Ͻ 0.05), and HC84.24 and HC84.26 (antigenic domain D) were inhibited between 64 and 69% (P Ͻ 0.05). Against a neutralizing antibody (CBH-7) to a third domain C cluster, essentially no inhibition was observed (P Ͼ 0.05). The inhibition of domain D HMAb binding by an antibody to aa 412 to 423 has been observed previously and is unidirectional, which is consistent with proximity but not an overlapping nature of their respective epitopes (20).
Unexpectedly, H77.16, a murine MAb that binds primarily to 1a H77C HVR1 (29), also blocked both subsets of HC33-related HMAbs by 75 to 89% (P Ͻ 0.05) (Fig. 2B). Although we anticipated inhibition by H77.16 against HC33.4, HC33.8, and HC33.9 because of the effect of the K408A mutation on their binding, we did not expect a loss of HC33.1 binding (Fig. 1C). H77.16 blocked antigenic domain B antibodies HC-1 and HC-11 by 51% to 53% (P Ͻ 0.05) and antigenic domain D antibodies HC84.24 and HC84.26 by 39% to 44% (P Ͻ 0.05). In contrast, binding by an antigenic domain C antibody, CBH-7, was not affected by H77.16 (P Ͼ 0.05). Contact residues within HVR1 have not been identified for antibodies to antigenic domains B, C, and D in previous studies (25,28,43). Thus, the global blocking effect of H77.16 on the binding of these broadly neutralizing antibodies suggests that anti-HVR1 antibodies may interfere with the neutralizing antibodies of antigenic domains B, D and E.
Anti-HVR1 interferes with broadly neutralizing antibodies. We next examined the neutralization of 1a H77 HCVcc by representative antigenic domain B to E HMAbs in the presence or absence of "interfering" HMAbs. We assessed whether combinations of antibodies were antagonistic, additive, or synergistic using the median-effect analysis method, as described in Materials and Methods (32,33). A constant ratio between the HVR1 antibody H77. 16 (24). The IC 50 s for HC-11 (1.2 g/ml) and HC84.26 (0.080 g/ml) were previously established (28,43). Dose-dependent neutralization was performed for H77.16, HC33.1, HC33.4, and CBH-7, which determined their respective IC 50 s of 0.5, 3.24, 0.11, and 1.8 g/ml (data not shown). Dose-dependent neutralization was tested for each antibody alone and in combination in a range of 2-fold dilutions in concentration from 8ϫ IC 50 to 1/16ϫIC 50 . A representative set of analyses to determine cooperativity in virus neutralization is shown in Fig. 3A 3A); fractional effect (FA) values were plotted in relation to dose (Fig. 3B), and combination index (CI) values were calculated and plotted in relation to FA (Fig. 3C). These studies were performed initially for H77.16 in combination with HC33.1, HC33.4, HC-11, HC84.26, or CBH-7. For each set of analyses, the linear correlation coefficient r was greater than 0.95, indicating a high goodness of fit to the plots (data not shown). The CI values of the paired studies at FA values of the 50% effective dose (ED 50 ), ED 75 , and ED 90 were tabulated (Fig. 3D). Aside from CBH-7, the CI values for the remaining antibodies, HC33.1, HC33.4, HC-11, or HC84.26, in combination with H77.16 at the FA of ED 50 (range, 1.44 to 1.84), ED 75 (range, 1.31 to 2.76), and ED 90 (range, 1.18 to 2.32) were all above 1, which indicates antagonism. Moderate (CI ranging between 1.2 and 1.45) antagonism was observed between H77.16 and HC33.1 or HC84.26 (Fig. 4A). Stronger antagonism (ranging between 1.46 and 3.0) was observed between H77.16 and HC33.4 or HC-11. These findings are consistent with the ability of H77.16 to interfere with the binding of representative antigenic domain B (HC-11), D (HC84. 26), and E (HC33.1 and HC33.4) HMAbs. Additive cooperativity with CI values near 1.0 (defined as 0.9 to 1.1) was observed between H77.16 and CBH-7, which is consistent with the minimal competition between these two antibodies ( Fig. 2B and 4A).
To assess whether the mass of full-length IgG binding to HVR1 decreases access of broadly neutralizing antibodies to their epitopes on E2 by steric hindrance, combination antibody studies were repeated with H77.16 Fab 2 fragments (Fig. 3D). As documented for each pair, the CI values at ED 50 , ED 75 , and ED 90 were at 20 g/ml was first incubated with E2 that had been immobilized on an ELISA plate (28). Test HMAb at 1 g/ml was then added, and bound HMAb was measured as described previously (28). The percent inhibition was based on test HMAb binding in the absence of the murine MAb. Data are shown as mean percent inhibition Ϯ SD from two experiments performed in triplicate.
Temperature and time of addition alter anti-HVR1 interference of binding by broadly neutralizing antibodies. Previous studies with HCV (46) and other flaviviruses (47) demonstrated that higher incubation temperatures lead to greater antibody binding to cryptic epitopes because of enhanced viral protein motion. We hypothesized that at higher temperatures, binding by broadly neutralizing antibodies would be increased due to greater exposure of their epitopes and decreased steric hindrance by anti-HVR1 MAbs. Two sets of H77C E2 were first exposed to H77.16 (20 g/ml) at room temperature (RT). One (test) set was incubated at 40°C prior to addition of antigenic domain E, B, and D HMAbs; the other (control) set remained at RT prior to addition of the HMAbs. After 30 min at either RT or 40°C, detection of test HMAb binding was determined by ELISA. Indeed, greater binding by each antigenic domain E, B, and D HMAb was observed at 40°C than at RT, and this was associated with significantly less inhibition by H77.16 at 40°C than at RT (P Ͻ 0.05, Fig. 5A). Furthermore, we tested whether preincubation with either broadly neutralizing antibodies or H77.16 affected the ability of H77.16 to interfere with the binding of E, B, and D HMAbs. When E2 was first incubated with each E, B, or D HMAb and then with H77.16, virtually no inhibition of binding by H77.16 was observed, as expected (Fig. 5B). When E2 was first incubated by H77. 16 and then with each E, B, or D HMAb, the magnitudes of inhibition by H77.16 against the domain E, B, or E HMAbs were similar to the observed inhibition, as shown in Fig. 5A at RT. Binding of the control antibody, CBH-7, to E2 was not affected by H77.16 under either test conditions (Fig.  5B). These observations support the hypothesis of steric hin- drance of binding of broadly neutralizing MAbs by anti-HVR1 MAbs.
Determination of the structure of Fab-peptide complexes. Because the HC33.4, HC33.8, and HC33.29 epitopes appear to overlap between HVR1 and aa 412 to 423, structural analysis was performed. We expressed the Fab fragments derived from HC33.4 and HC33.8 and performed cocrystallization trials of complexes containing the Fab and a peptide comprising residues 406 to 425 (PGAKQNIQLINTNGSWHINST) of the genotype 1a strain H77. This peptide was chosen to include all putative contacts revealed by the alanine-scanning mutagenesis, i.e., residues K408, L413, G418, and W420. We obtained crystals diffracting to 1.65-Å resolution (Fab HC33.4) and 1.9-Å resolution (Fab HC33.8), respectively ( Table 1). The structures of both complexes were determined by the molecular replacement method (see Materials and Methods for more details). Since comparison of peptides from both complexes revealed an identical amino acid backbone conformation and the peptide in the HC33.8 complex structure is more complete, we concentrated our further analysis on this complex.
Molecular determinants of Fab HC33.8 interaction with its peptide epitope. The conformation of the peptide in complex with Fab HC33.8 resembles the recently described conformation of a similar peptide in complex with Fab HC33.1 (48). The interaction between the peptide and Fab HC33.8 is dominated by the side chain of W420 that protrudes and is deeply immersed into a cavity formed by the long complementarity-determining region 3 loop of the heavy chain (CDR-H3), the framework residues around CDR-H2, and the CDR-L3 loop ( Fig. 6B and C). This pattern agrees with the results of the alanine-scanning mutagenesis, with W420 being a primary determinant of antibody binding. The second residue that is suggested by the alanine-scanning mutagenesis as crucial for an antibody-E2 interaction is G418. This glycine residue makes three hydrogen bonds involving main-chain atoms; however, this cannot explain the amino acid specificity suggested by alanine-scanning mutagenesis. It is more likely that the extensive flexibility of the glycine residue is required to facilitate the protrusion of the W420 side chain into the described cavity in the paratope. In contrast to the case for the HC33.1 complex, in which electron density was observed for residues 412 to 423, peptide residues 412 to 414 of the HC33.8 complex are likely to be disordered, and no evidence for further direct interactions between antibody and peptide were observed.
The peptide in the HC33.8 complex is exposed to the solvent, and particularly N415 (i.e., the region directly downstream of the disordered part of the peptide) is located in a solvent channel (Fig.  6A), suggesting that crystal packing does not prevent a direct interaction between K408 and the antibody. However, the N-terminal peptide parts of the HC33.1 and HC33.8 complexes likely differ due to a short stretch at the end of ␤-strand F in the heavy-chain variable region of HC33.8 ( 101 VFTDS 105 , versus 101 VSSDI 105 in HC33.1). The distances between superposed C␣ atoms from the corresponding segments of HC33.1 and HC33.8 amount to 0.2 Å and 0.6 Å at residues 101 and 105, respectively, and to 4.3 Å in the middle, with the HC33.8 segment bulging out. Fab HC33.1 tightly interacts with I414 in the N-terminal part of the peptide via two main-chain hydrogen bonds from S H103 , presumably stabilizing the peptide conformation in the HC33.1 complex. The bulge observed in the HC33.8 complex prohibits this interaction, suggesting major differences in the N-terminal part of the interface. The complex structure can therefore neither confirm nor rule out a direct interaction between antibody and the N-terminal part of the peptide.
Conformation of the E2 peptide from aa 415 to 423 in complex with Fab HC33.8. The peptide in the HC33.8 complex adopts an extended conformation similar to the conformation observed in complex with the related HC33.1 Fab (48). However, this binding mode contrasts with the ␤-hairpin formed in complex with were first incubated with E2 that had been immobilized on an ELISA plate (28). Test HC33.4 HMAb at 1 g/ml was then added, and bound HMAb was measured as described previously (28). The percent inhibition was based on test HMAb binding in the absence of the murine MAb. Data are shown as mean percent inhibition from two experiments performed in triplicate. three independent neutralizing antibodies (49,50) or with the extended conformation observed in complex with MAb 3/11 (51) (Fig. 7). All three backbone conformations have been observed in complex with broadly neutralizing antibodies, underlining the structural flexibility at the surface of infectious virus particles. Our data suggest that different conformations of the antigenic region aa 412 to 423 are in equilibrium and that independent antibodies recognize this site following the principle of induced fit or conformational selection. The demonstrated conformational flexibility in this antigenic region supports the hypothesis of a temperaturedependent steric hindrance of binding of broadly neutralizing MAbs by anti-HVR1 MAbs. An intact IgG that dangles upstream of the flexible antigenic domain E (i.e., one that binds the C terminus of HVR1) sterically blocks binding of broadly neutralizing antibodies to adjacent epitopes.

DISCUSSION
The functional and biophysical properties of HVR1 on the E2 glycoprotein contribute to the ability of HCV to escape from immune recognition, which ultimately leads to persistent infection. The region is immunodominant and serves as a major decoy that diverts the B cell response from other, more conserved regions on E2. Antibodies elicited to HVR1 are neutralizing, but they are associated with rapid viral escape without compromising viral fitness. Prior studies have proposed that HVR1 structurally shields access by neutralizing antibodies to their respective epitopes (18). Our finding that broadly neutralizing HMAbs to overlapping conformational epitopes in E2 bind at higher levels to ⌬HVR1 E2 than wt E2 is consistent with this hypothesis.
The different binding patterns of a panel of HMAbs to mainly linear epitopes that are located adjacent to HVR1, encompassing aa 412 to 423 and designated antigenic domain E, led to a series of studies that identified a new mechanism for HVR1 attenuation of antibody-mediated virus neutralization. We showed that when an anti-HVR1 antibody occupies its site on HVR1, access of antigenic domain E antibodies (e.g., HC33.1 and HC33.4) is compromised.  24 and HC84.26) HMAb at 1 g/ml was then added, and after 30 min at either RT or 40°C, bound HMAb was measured at RT as described previously (28). The percent inhibition at either RT or 40°C was based on each HMAb binding in the absence of the murine MAb. (B) H77.16 (labeled H77.16 1st) or E, B, or D HMAb (labeled H77.16 2nd) was added to E2 and left for 30 min at RT. After washing, each E, B, or D HMAb was added to E2 preincubated with H77.16, or H77.16 was added to E2 preincubated with E, B, or D HMAb, and left for an additional 30 min. Control antibody was CBH-7, a domain C HMAb. Detection and calculation of percent inhibition were the same as for panel A. Data are shown as mean percent inhibition from two experiments performed in triplicate. View on the crystalline environment to determine crystal packing effects. The Fab is colored in light (light chain) and dark (heavy chain) gray and shown as molecular surface. The peptide (shown as a cartoon and ramp colored from blue to red through yellow from the N to the C terminus) interacts mainly with the heavy chain, and its N terminus is exposed in a solvent channel (arrows). (B and C) View on the paratope of the Fab HC33.8-peptide complex from two angles, illustrating the protruding side chain of W420 in a cavity formed by CDR 3 of the heavy chain 1, 2, and 3 (CDR-H3), CDR-L3, and framework residues surrounding the CDR-H2 loop. The Fab is colored as in panel A, CDRs 1, 2, and 3 are colored in cyan, light cyan, and dark green for the heavy chain and in sand, olive, and yellow for the light chain, respectively. The peptide is shown as a cartoon and colored by atom type (orange, red, and blue for carbon, oxygen, and nitrogen, respectively), and it binds to the paratope mostly between the long CDR-H3 loop (green) and the other heavy-chain CDRs. A red ellipse highlights a short stretch at the end of ␤-strand F in the heavy-chain variable region that bulges out in HC33.8, while it tightly interacts with the peptide in the HC33.1 complex (48). (D) View on the peptide (shown as molecular surface and colored as in panel B) rotated by 120°around the indicated axis with respect to panel C to illustrate how the W420 side chain protrudes from the bulge in the center of the peptide.
Diminished access against antibodies to antigenic domains B (HC-1 and HC-11) and D (HC84.20, HC84.24, and HC84.26) also was observed. The major binding regions for domains B and D have been mapped to aa 529 to 540 and aa 440 to 446, respectively (28,43). The effect of anti-HVR1 on neutralization by antibodies recognizing antigenic domain B, D, and E antibodies was apparent in the antagonism studies by use of the median-effect analysis method. When the mass of the anti-HVR1 antibody was reduced as Fab 2 fragments, the observed antagonism of anti-HVR1 was diminished. Furthermore, greater binding by each domain B, D, and E HMAb was observed at 40°C than at RT, which is consistent with increased exposure of their cognate epitopes due to enhanced viral, or in this case protein, "breathing" at higher temperatures (46,47). This led to decreased interference by anti-HVR1 antibodies. The drop in antagonism with H77.16 Fab 2 against both HC33.1 and HC33.4 appears to be similar, although the HC33.4 epitope includes a critical residue within HVR1 as determined by alanine substitution studies. A possible explanation is that structural studies with both antibodies are nearly identical, which means that HC33.4 does not include a true contact residue at aa 408 (48). Taken together, these findings suggest that anti-HVR1 Abs interfere with the binding of broadly neutralizing antibodies by steric hindrance. Finally, the interference by H77. 16 against domain E antibodies (HC33-related HMAbs) appears to be greater than that by H77.39, although H77.39 binds to the same region as antigenic domain E antibodies (Fig. 2). The difference is probably due to the different mechanisms of interference. H77.39 directly competes for the same binding sites as the HC33-related antibodies, which can be affected by the relative binding affinities of the two antibodies. In contrast, H77.16 binds to an adjacent site in which the bulk of the antibody molecule is preventing the binding of HC33-related antibodies to their respective epitopes. The extent of blockade by steric hindrance is less dependent on differences in the affinities of the competing antibodies, which could be more effective with the tested antibodies. The concept of antibody-mediated interference of HCV is not new. It has been proposed that epitopes within the E2 segment encompassing aa 434 to 446 elicit nonneutralizing antibodies and that these antibodies interfere with neutralizing antibodies directed at an adjacent E2 segment (aa 412 to 426) (22,23). However, other studies employing similar approaches of isolating polyclonal antibodies to synthetic peptides encompassing aa 412 to 426 and aa 434 to 446 showed no interference in virus neutralization (21). The lack of interference by antibodies to aa 434 to 446 was assessed by combination studies of HMAbs that bind to aa 412 to 426 or aa 434 to 446 (20,28). The median-effect analysis method was applied, and it was determined that the effect was additive and not antagonistic (20).
We performed structural studies to further understand how the binding of anti-HVR1 MAbs interferes with broadly neutralizing antibodies. The antigenic region aa 412 to 423 is positioned downstream of HVR1, which is a structurally flexible region at the N terminus of E2. HVR1 was reported to interfere with binding of neutralizing antibodies by shielding conserved epitopes (18,19). The different structures observed for peptides comprising aa 412 to 423 bound to multiple neutralizing antibodies suggest that this region also has considerable structural flexibility in the HCV particle. These observations indicate that there is a long, highly flexible region at the N terminus of E2, which extends beyond HVR1 and includes conserved residues strongly implicated in CD81 binding. Since the majority of neutralizing epitopes within E2 are located in close proximity to this flexible region, MAbs targeting epitopes within this region likely can interfere, even in the absence of direct competition for binding residues, with binding of neutralizing antibodies to other antigenic domains within E2 by steric hindrance.
Overall, our findings indicate that in addition to a direct shielding role of HVR1, an indirect shielding role is plausible and could be exerted by antibodies binding to the extended structurally flexible region at the N terminus of E2 (including aa 412 to 423). Thus, immunization strategies that avoid the generation of HVR antibodies may be needed to enable broadly neutralizing antibodies to function optimally and prevent new infections or control established infections.