Research ArticleVIROLOGY

An alternate conformation of HCV E2 neutralizing face as an additional vaccine target

See allHide authors and affiliations

Science Advances  24 Jul 2020:
Vol. 6, no. 30, eabb5642
DOI: 10.1126/sciadv.abb5642


To achieve global elimination of hepatitis C virus (HCV), an effective cross-genotype vaccine is needed. The HCV envelope glycoprotein E2 is the main target for neutralizing antibodies (nAbs), which aid in HCV clearance and protection. E2 is structurally flexible and functions in engaging host receptors. Many nAbs bind to the “neutralizing face” on E2, including several broadly nAbs encoded by the VH1-69 germline gene family that bind to a similar conformation (A) of this face. Here, a previously unknown conformation (B) of the neutralizing face is revealed in crystal structures of two of four additional E2–VH1-69 nAb complexes. In this conformation, the E2 front-layer region is displaced upon antibody binding, exposing residues in the back layer for direct antibody interaction. This E2 B structure may represent another conformational state in the viral entry process that is susceptible to antibody neutralization and thus provide a new target for rational vaccine development.


Hepatitis C virus (HCV) is a main cause of liver failure and hepatocellular carcinoma, infecting ~1% of the world population with an estimated 1.5 million to 2 million new infections each year (1, 2). Recent hepatitis data further highlight the need for an urgent global response ( HCV entry into host liver hepatocytes cells is a highly coordinated process, which is still not fully understood, but is mediated by the viral E1-E2 envelope glycoprotein heterodimer and involves interactions with a plethora of host factors [reviewed in (3, 4)]. The initial attachment is believed to be mediated by viral surface lipoproteins and apolipoproteins and heparan sulfate proteoglycans (HSPGs) followed by interactions with the low-density lipoprotein receptor (LDLR) and scavenger receptor class B member 1 (SR-B1). These interactions likely cause conformational changes in E2 to mediate interaction with the major receptor, the tetraspanin CD81. This interaction triggers lateral membrane diffusion of the HCV-CD81 complex toward the tight junctions followed by claudin-1 (CLDN1)– and occludin (OCLN)–mediated endocytosis [reviewed in (3, 4)]. A major challenge in the development of an HCV vaccine is the high genetic diversity of the circulating virus, which is grouped into six epidemiologically important major genotypes with >30% difference in genomic sequences (5). Moreover, rapid mutation of the virus in infected individuals results in viral quasispecies that can escape immune surveillance. Therefore, it is expected that an effective vaccine must elicit robust immune responses to conserved viral epitopes for protection against chronic HCV infection.

E2 is the main HCV receptor binding protein and the primary target for broadly neutralizing antibodies (bnAbs). The receptor binding domain (residues 384 to 645), also known as E2 core domain, is modified by up to 11 N-linked glycans [reviewed in (4)] and contains several variable regions (VRs; comprising ~25% of the E2 core sequence; Fig. 1A) that increase the genetic diversity of HCV for evasion of the immune system (6, 7). Crystal structures of E2 from different HCV strains are all very similar, with the E2 core domain consisting of a central immunoglobulin (Ig) β-sandwich fold (fig. S1A) (811) stabilized by eight disulfides and flanked by N-terminal hypervariable region 1 (HVR1; residues 384 to 410), a front layer (residues 421 to 459), and a C-terminal back layer (residues 597 to 645) (Fig. 1A). The high density of the disulfide bonds contributes to the high thermal stability of E2, although the receptor binding site in recombinant E2 exhibits high conformational flexibility [reviewed in (3, 9)]. It is unknown whether this structural flexibility is observed only in soluble E2 or a genuine molecular feature required for E2 function that might prevent or distract the immune system from eliciting bnAbs to this functionally conserved neutralizing target.

Fig. 1 Characterization of E2 VH1-69–encoded bnAbs.

(A) Schematic representation of the E2 extracellular region (amino acids 384 to 645) colored by structural components with VRs in gray, AS412 in pink, front layer in cyan, β-sandwich in red, CD81 binding loop in blue, and back layer in green. The E2 N-linked glycans and the conserved disulfide bonds are indicated by green branches and blue dashed lines, respectively (10). (B) Surface representation of E2c colored by hydrophobicity scale (36) with hydrophilic residues colored in green and hydrophobic residues in white. A hydrophobic pocket is formed by the front layer and CD81 binding loop (left), and the interaction between this pocket with AR3C CDRH2 hydrophobic tip is illustrated (right). (C) Schematic representation of the selected panel of VH1-69–encoded bnAbs targeting AR3 and AS434, colored based on the studies that reported the isolations of these nAbs (see also fig. S1C). CDRH3 length is indicated in brackets. (D) Neutralization breadth of the VH1-69–encoded bnAbs against HCVcc from genotypes 1 to 6. The chart displays the median inhibitory concentration (IC50) (μg/ml) titers of the antibodies (see also fig. S1D). The IC50 neutralization values of mAb AR3A, AR3B, AR3C, AR3D, and U1 and 212.10 and 212.1.1 were previously reported and are included here for comparison (11, 30). (E) Thermodynamic values (−TΔS, ΔH, and ΔG) of ITC measurements of E2c3 binding to AR3-targeted bnAbs. Values are averaged results from two or three ITC experiments. Representative isotherms are illustrated in fig. S1F.

A detailed understanding of E2 and bnAb interactions at the molecular level is imperative for rational design of HCV vaccine antigens. To date, most E2 bnAbs neutralize HCV by blocking E2-CD81 interactions and target three overlapping neutralizing sites [antigenic site (AS) 412 (amino acids 412 to 423), AS434 (amino acids 434 to 446), and antigenic region 3 (AR3)], which collectively form the E2 neutralizing face (fig. S1B) (12). AR3 is a cluster of discontinuous epitopes formed by the front layer and CD81 binding loop (amino acids 519 to 535) and is the target for bnAbs isolated from infected patients (12). In the published E2 structures in complex with different AR3-specific bnAbs, E2 adopts a nearly identical conformation at the binding interface with the bnAbs (8, 10, 11). Given that the E2 neutralizing face (and CD81 binding site) is likely to be structurally flexible (3, 12), the reported E2 structures appear to be in a “preferred” conformation recognized by bnAbs and thus a good template for structure-based design of vaccine immunogens.

Recently, genetic analyses indicate that bnAbs targeting AR3 are dominated by VH1-69–encoded monoclonal antibodies (mAbs), which are characterized by the hydrophobic tip of their germline-encoded heavy chain complementarity-determining region 2 (CDRH2) loop and by a longer-than-average CDRH3 loop (17 to 22 amino acids, using the Kabat definition) [reviewed in (12, 13)]. Previous structural studies of E2-bnAb complexes confirm that CDRH2 interacts with a highly conserved hydrophobic pocket formed by the front layer and CD81 binding loop (Fig. 1B) (8, 10, 11) and therefore overlaps with the anticipated CD81 binding site. In this study, we sought to investigate whether other HCV nAbs in the VH1-69 class also recognize the same E2 neutralizing face. We surprisingly discovered a different conformation of E2 that is also susceptible to antibody neutralization. These new E2 structures here support the hypothesis that the E2 neutralizing face (and CD81 binding site) is flexible and that these different conformations recognized by nAbs may represent various functional stages of E2 during viral entry. This discovery also provides an additional template for HCV vaccine design.


Characterization of VH1-69 class nAbs that target E2 neutralizing face of HCV

As no high-resolution structure of an E2-CD81 complex is available, expanding the structural knowledge of E2 recognition by bnAbs targeting the CD81 binding site may shed light on the interactions of HCV E2 with CD81, which is one of the principal receptors in viral entry. For this purpose, we selected a panel of bnAbs that neutralize HCV by blocking CD81 binding. These antibodies are encoded by the VH1-69 genes and target AR3 [reviewed in (13)] and AS434 (14, 15) (Fig. 1C and fig. S1, B and C). The antibodies contain similar hydrophobic CDRH2 tips but vary in their CDRH3 length (between 10 and 22 amino acids; fig. S1C). They cross-neutralized recombinant cell culture HCV (HCVcc) displaying E1E2 from diverse genotypes to varying degrees (Fig. 1D and fig. S1D).

Thermodynamic measurements were carried out for the genotype 1a H77 E2c3 core construct [E2c, comprising residues 412 to 459 and 486 to 645, with removal of the N448 and N576 glycosylation sites; E2c3 is an E2c variant without VR3 (amino acids 569 to 597)] (16) with bnAbs to provide insight into the mode of binding (e.g., nature of molecular interactions and conformational changes) (17). Isothermal titration calorimetry (ITC) measurements indicated a similar change in Gibbs free energy (ΔG) for all tested mAbs but compensatory differences in enthalpy (ΔH) and entropy (−TΔS) contributions (Fig. 1E and fig. S1E). While AR3A, AR3B, and HC11 binding resulted in modest differences in enthalpy and similar small change in entropy, the binding of 212.10, HC1AM, and 212.1.1 resulted in larger increases in enthalpy and corresponding unfavorable changes in entropy. These results are suggestive of conformational changes upon binding of bnAb 212.10 and HC1AM, as well as for nAb 212.1.1 (in E2c, mAbs, or both), thereby uncovering potential differences in E2c3 binding between these two groups of antibodies.

Crystal structures of VH1-69 class nAbs

To elucidate how these differences in the thermodynamic parameters of the selected bnAbs might be reflected in any structural perturbations, we determined crystal structures of HK6a E2c3-Fab U1, HK6a E2c3-Fab HC11, H77 E2c3-Fab 212.1.1-Fab E1, and H77 E2c3-Fab HC1AM-Fab E1 (table S1). Fab E1 is a non-neutralizing antibody (n.b. not the E1 envelope glycoprotein) that targets AR1 (figs. S1C and S2) and was used to enhance the crystallization of Fab 212.1.1 and Fab HC1AM with E2. E2 from HCV H77 and HK6a strains (genotypes 1a and 6a, respectively) produced crystals that diffracted well in our previous structural studies (10, 11). The antibody-antigen complex structures confirm that the four nAbs all bind to epitopes overlapping with AR3 (Fig. 2A). Superposition of the E2 residues from the E2c3-bnAb structures onto H77 E2c-Fab AR3C [Protein Data Bank (PDB) 4MWF] and HK6a E2c3-Fab AR3A (PDB 6BKB) indicates that HC11 binds E2 in a similar orientation as Fabs AR3A/AR3C, while 212.1.1, HC1AM, and U1 bind in different orientations (Fig. 2A).

Fig. 2 Crystal structures of selected HCV E2 VH1-69–encoded bnAbs.

(A) Crystal structures of H77 and HK6a E2c3 in complex with AR3A, AR3C, U1, HC11, 212.1.1, and HC1AM illustrate differences in the orientation of the VH1-69 bnAbs with respect to E2c. All structures are superimposed on the E2 of the HK6A E2c3-AR3A complex. For clarity, only the Fab V domains are shown. (B) Neutralizing epitopes of VH1-69–encoded bnAbs. The E2c structures are shown in surface representation, with the interacting residues (top) or CDRH2 footprint (bottom) in the epitopes colored and labeled.

It was previously shown that binding of VH1-69–encoded mAbs with long CDRH3 loops (18 to 22 amino acids; AR3 and HEPC bnAbs) to E2 is heavy chain (HC) dominant (8, 10, 11). The CDRH3 length of the VH1-69 antibody panel in this study varies between 10 and 17 amino acids (fig. S1C). Replacing the light chains (LCs) of the VH1-69 antibody panel with an LC originating from non-HCV mAbs resulted in loss of binding for all but AR3A and 212.10 nAbs (CDRH3 lengths of 18 and 17 amino acids, respectively) (fig. S3A). The results show that HC dominance of VH1-69–encoded anti-HCV antibodies is antibody specific, and LC can also be important for the VH1-69 class nAbs.

Buried surface area (BSA) analysis of the structures reported here confirms that the binding is dominated by the HC and mediated mainly by CDRH2 and CDRH3 (for Fab U1, binding is mediated only by the HC) (fig. S3B). However, some differences are observed in the HC BSA; while CDRH3 is the main contributor for AR3A, AR3C, HC11, and U1 (53 to 66% of total BSA), for 212.1.1 and HC1AM, CDRH2 increases its percent contribution (32 to 40%), whereas CDRH3 is correspondingly reduced (23 to 39%) (fig. S3B).

Footprint of VH1-69 class nAbs on E2

Epitope mapping, together with detailed analysis of the interactions between E2 and Fabs (Fig. 2B and fig. S4), indicates the dominance of the HC for all of the VH1-69 antibody panel. The interactions, footprint, and binding orientation of Fab HC11 with E2c3 (Fig. 2B and fig. S4) are similar to that of Fabs AR3A/AR3C (11). CDRH3 of HC11 is encoded by the D-gene IGHD2 allele [based on the IMGT/V-QUEST program (18)], containing an intrinsic disulfide bond, as seen also in AR3A, AR3C, and the HEPC3/74 bnAbs (8). The HC11 CDRH3 is bent in the conformation seen previously in AR3A and AR3C (Fig. 3). U1 shares similar CDRH1 and CDRH2 footprints with AR3A/AR3C and HC11 (11), but CDRH3 is shifted slightly more toward a groove between the front-layer N-terminal region (amino acids 421 to 430) and CD81 binding loop (amino acids 519 to 536) with fewer interactions with the front layer (fig. S4). Also, in contrast to AR3A/AR3C and HC11, U1 CDRH3 does not contain an intrinsic disulfide bond (fig. S1C). Substantial differences are observed for 212.1.1 and HC1AM in that their binding involves a shift of the front-layer C-terminal region and unexpectedly involves interactions with residues in the now exposed back layer (amino acids 597 to 645) (Fig. 2B and fig. S4). E2 recognition by the HC is mediated by CDRH2 and CDRH3 with no or one interaction with CDRH1 (fig. S4). As a result, CDRL3 gains interactions through a shift toward the CD81 binding loop (compared to LC interactions with the front layer for HC11 and AR3A; fig. S4).

Fig. 3 CDRH3 of VH1-69–encoded bnAbs to E2.

(A) Alignment of the CDRH3 sequences of AR3A, AR3C, HEPC3, HEPC74, and HC11, with the cysteines that form a disulfide bond highlighted in red. (B) Cartoon representation of the interaction of CDRH3 of the antibodies to E2, with the disulfide shown in yellow sticks. (C) Conformation of CDRH3 of AR3A, AR3C, HC11, and HEPC74 from the E2-bnAbs complexes. For clarity, only the HC VRs are shown.

Role of CDRH2 in antigen recognition

As previously suggested, the CDRH2 hydrophobic tip of VH1-69–encoded bnAbs has an important role in antigen recognition [reviewed in (13)]. Similar to AR3A/AR3C, the CDRH2 tips of U1 and HC11 interact with the front layer and CD81 binding loop (I53, F54 and M53, F54 respectively) (Fig. 4). In contrast, 212.1.1 and HC1AM use a more extensive region of CDRH2 (amino acids 52 to 58; fig. S4B) to interact with the hydrophobic pocket formed by residues from the front layer, CD81 binding loop, and back layer. This difference results from the different binding orientations of 212.1.1 and HC1AM on E2 (Fig. 2A). To further validate the role of CDRH2, the tip residues (amino acids 53 and 54) of selected VH1-69–encoded nAbs were mutated to alanine. Binding experiments indicated substantial reduction in affinity for both mutants (excluding AR3A M53A) with larger reduction for mutation of position 54 to Ala (54A) (Fig. 4B). Neutralization experiments indicated that, for AR3A, HC11, and HC1AM, the hydrophobic side chain at position 53 (Met) has no or relatively moderate influence, while position 54 (Phe) has a more notable influence on HCV neutralization (Fig. 4C). Double mutation at positions 53 and 54 resulted in a cumulative effect for AR3A and HC11 (Fig. 4, B and C). Moreover, for 212.1.1, mutation of L54 to Ala abolishes its neutralization capability, highlighting the critical role of the hydrophobic tip in neutralization by this antibody (Fig. 4B).

Fig. 4 CDRH2 of VH1-69–encoded bnAbs to E2.

(A) Interactions between the hydrophobic CDRH2 of selected VH1-69–encoded bnAbs with the E2 hydrophobic pocket. (B) Binding of AR3A, HC11, 212.1.1, and HC1AM IgGs to H77 E2c with CDRH2 hydrophobic tip residues mutated at positions 53 and 54 illustrates the role of the CDRH2 tip in E2 binding. Summary of the fold difference of dissociation constant (KD) values relative to the wild-type (WT) KD is shown below. (C) The role of CDRH2 tip residues 53 and 54 on neutralization of AR3A, HC11, 212.1.1, and HC1AM was tested by HCVpp neutralization assays with H77 (1a). The IC50 values are presented. The fold difference of IC50 values relative to the WT IC50 is summarized in the table. The 212.1.1 L54A and I53A-L54A mutants exhibit no neutralization (represented by dashed lines).

Conformational changes of E2 front layer

Structural studies of E2c3-bnAb complexes from this and previous studies (8, 10, 11) reveal substantial conformational changes of E2 in the different complexes. The β-sandwich loop residues 541 to 549 (between β6 and β7 strands) adopt one of two distinct conformations in all E2 structures determined thus far. Up to now, all genotype 1a E2 structures exhibit one conformation, while all genotype 6a structures adopt a different conformation. This region in the H77 E2c3-Fab 212.1.1 and H77 E2c3-Fab HC1AM structures matches those of other genotype 1a structures, while the loop in the HK6a E2c3-U1 and HK6a E2c3-HC11 structures is similar in other genotype 6a structures (fig. S5A). However, in two complexes of the identical 1b09 (genotype 1b) ectodomain with either Fab HEPC74 (PDB 6MEH) or Fab HEPC3 (PDB 6MEI), the loops in the two complexes adopt different conformations. The β-sandwich loop in the HEPC3 complex maintains the conformation found in genotype 1a E2, whereas in the HEPC74 complex, it adopts the genotype 6a conformation (fig. S5B), suggesting that this region may have some inherent flexibility. Superposition of all available E2 structures shows that the front layer (amino acids 429 to 451) for HK6a E2c3-U1 and HK6a E2c3-HC11 resembles other front-layer structures (fig. S5C) (11). However, differences are observed for the front layer in H77 E2c3-Fab 212.1.1 (Fig. 5A and fig. S5, C and D). The front-layer region is constrained by two disulfide bonds (C429-C503 and C452-C620), which covalently anchor the front layer to the β-sandwich and back layer (Figs. 1A and 5A). In the H77 E2c3-Fab 212.1.1 structure, there is substantial movement and conformational change of the front layer (e.g., 10 Å measured between the Cα’s of Y443; Fig. 5A) of the α1-helix (amino acids 439 to 443). In H77 E2c3-Fab HC1AM, only front-layer residues 429 to 432 have ordered electron density; however, superposition of H77 E2c3-HC1AM onto the H77 E2c-AR3C and H77 E2c3-212.1.1 structures (based on E2) suggests that HC1AM binds to E2 with a conformation different from that recognized by AR3C (fig. S5E).

Fig. 5 A and B conformations of HCV E2 front layer.

(A) Superposition of E2c from the crystal structures with AR3C and 212.1.1 indicating conformational changes of the front layer (amino acids 430 to 451). The two disulfide bonds (C429-C503 and C452-C620) that represent the boundaries of the conformational changes are labeled along with Y443. (B) Surface representation of E2c from the complex with AR3C (top) and 212.1.1 (bottom). The conformational changes in E2c in the 212.1.1 complex expose more of the β-sandwich and back-layer surfaces (colored in red and green, left panels) that include residues critical for CD81 binding (right panels). (C) Thermodynamic values (−TΔS, ΔH, and ΔG) of ITC measurements of E2c binding to Fc-CD81-LEL and MBP-CD81-LEL. Values are average results from two ITC experiments. Representative isotherms are illustrated in fig. S1F.

Exposure of the E2 back layer in the H77 E2c3-212.1.1 structure

The structures of E2c-Fabs AR3A-D and truncated E2 ectodomain-HEPC3/74 (8, 10, 11) reveal that the front layer covers some regions of the central β-sandwich and the back layer (fig. S1A). In H77 E2c3-212.1.1 and H77 E2c3-HC1AM (fig. S5E), this different conformation of the front layer (amino acids 429 to 451) exposes an extended surface of the β-sandwich and back layer that interact with CDRH2 and CDRH3 and can explain the much higher BSA of CDRH2 in 212.1.1 and HC1AM (fig. S3B). This newly exposed surface also includes Y613 and W616 and other E2 core residues that have been reported to be critical for CD81 binding (Fig. 5B) (19). In H77 E2c-AR3C, W616 is buried and Y613 is only partially exposed. In the H77 E2c3-212.1.1, conformational changes of the front layer lead to full exposure of W616 (Fig. 5B; the solvent-accessible surface area of W616 changes from 7 and 31 Å2 for H77 E2c-AR3C and HK6a E2cs-AR3A to 76 and 100 Å2 for H77 E2c3-212.1.1 and H77 E2c3-HC1AM, respectively). Together, structural analyses of these new and previous structures (8, 10, 11) suggest that the E2 front-layer region can adopt at least two conformations: The “A” conformation (in complexes with AR3A-D, HEPC3/74, U1, and HC11) where large sections of the E2 back-layer region are obscured by the front layer, and the “B” conformation (with 212.1.1 and HC1AM) where the back-layer residues are more accessible for direct interaction with nAbs. Notably, in the recently published structure of HK6a E2c3-AR3A-Fab E1 complex, E2c3 adopts the A conformation, indicating that a conformational change in the front layer is not a consequence of Fab E1 binding (20).

The ITC measurements of E2-bnAbs (Fig. 1E) indicate differences in some of the thermodynamic parameters of 212.1.1 and HC1AM compared to AR3A/B and HC11. Comparison of the crystal structure of unliganded AR3B to the E2-bound structure (11) reveals no large conformational changes of the HC CDRs [Cα root mean square deviation (RMSD) values of ~0.3 and ~0.6 Å for CDRH2 and CDRH3, respectively; fig. S5G], as previously reported for other bnAbs that bind to the E2 A conformation (8). In contrast, a notable change is indicated in the CDRH3 conformation of HC1AM (Cα RMSD values of ~3.7 Å; fig. S5G), suggesting that the unfavorable entropy could result from conformational changes of CDRH3 and/or E2 front layer after bnAb binding to E2 in the B conformation. ITC measurements of E2 binding to the large extracellular loop (LEL) of CD81 [CD81-LEL as an Fc or maltose binding protein (MBP) fusion protein; Fig. 5C] reveals a change in entropy (−TΔS = 7.9 or 5.3 kcal/mol) that is similar to that with the E2-antibody B conformation, suggesting that CD81 binding to E2 is also accompanied by conformational changes (in E2, CD81, or both).

Role of the back-layer residues in CD81 binding

The structural evidence that the back-layer Y613 and W616 are exposed in the E2 B conformation raises the question of whether Y613 and W616 interact directly with CD81 or indirectly through stabilizing the front layer (fig. S6A). Binding experiments using E1E2 [by enzyme-linked immunosorbent assay (ELISA)] or soluble E2c (by biolayer interferometry) show that mutating Y613 to Ala/Phe and W616 to Ala abolished CD81 binding, whereas mutating W616 to Phe reduced binding (Fig. 6A). In contrast, binding of AR3A, AR3C, or 212.1.1 was slightly reduced by only some of the mutations (Fig. 6A). These findings indicate either that Y613 and W616 side chains are involved directly in CD81 interactions or that the mutations indirectly disrupt the CD81 binding site. This hypothesis needs to be confirmed by determining the E2-CD81 complex structure.

Fig. 6 Preferred conformation of HCV E2 front layer for CD81 binding.

(A) Binding of Fc-CD81, AR3A, AR3C, and 212.1.1 to H77 E2 WT and back-layer Y613 and W616 mutants measured by biolayer interferometry (KD values against soluble E2c, left) and ELISA (relative to E1E2 WT binding, right). The bnAb AP33 (37) that targets AS412 was used as a control. (B) Binding of Fc-CD81, AR3A, AR3C, and 212.1.1 to H77 E2c3 G440C + W616C, G440S, and W616S mutants measured by biolayer interferometry (KD values against soluble E2c, left) and ELISA (relative to E1E2 WT binding, right). The bnAb AP33 was used as a control. (C) Binding of E2c3 WT and the G440C + W616C, G440S, and W616S mutants to WT and engineered rat hepatoma cell line BRL3A that expresses either CD81, SR-B1, or the four HCV entry factors (SR-B1, CD81, CLDN1, and OCLN; shown as 4Rs) as measured by flow cytometry experiments. Corresponding bar graphs show the mean fluorescence intensity (MFI) derived from at least three independent experiments. The parental BRL3A cells were used as a control for nonspecific binding. (D) Mutagenesis data-driven modeling of CD81-E2 interactions comparing the E2 A and B conformations (E2 from H77 E2c-AR3C and H77 E2c3-212.1.1 complexes). E2 residues from clusters 1, 2, and 3, and CD81 residues from helices C and D (fig. S6D) were used in the modeling. E2 interacting residues are colored based on the structural components shown in Fig. 1A and CD81 residues in yellow sticks. CD81 interactions with conformation A and B have binding energies of −90.7 ± 14.4 [arbitrary units (a.u.)] and −75 ± 15.4 (a.u.), respectively.

To evaluate whether one of the two conformations is preferred for CD81 binding, we introduced a disulfide bond between the front-layer G440 and back-layer W616 to lock the front layer in the A conformation (E2c3 G440C + W616C; fig. S6B). Examination of the thermal stability of the E2c3 G440C + W616C mutant by differential scanning calorimetry revealed an increase of 2.5°C of the thermal denaturation midpoint (Tm), supporting locking of the A conformation (fig. S6C). In comparison, no change in Tm was found for single mutation G440S, while a decrease of 5.1°C was observed for W616S, confirming the role of the back layer in E2 stability (19). Next, we tested CD81-LEL and bnAb binding (Fig. 6, B and C), where introduction of the G440C + W616C double mutation, but not G440S or W616S, resulted in a notable decrease in binding of 212.1.1 to soluble E2c3 or E1E2, further confirming locking of the A conformation (Fig. 6B). The double mutation resulted in reduced binding to AR3A and AR3C that was similar to the single mutations. Next, we evaluated binding of Fc-CD81-LEL to soluble E2c3 and full-length E1E2 captured on a solid support (Fig. 6B). The single mutations G440S and W616S, and double mutation G440C + W616C, greatly reduced CD81 binding. The double mutation decreased binding (KD) of E2c3 to CD81-LEL more than the single mutations (Fig. 6B, left). The effect of mutations on interaction with full-length CD81 was examined further by measuring binding of E2c3 to an engineered rat hepatoma cell line BRL3A expressing human CD81, SR-B1, or the four HCV entry factors (SR-B1, CD81, CLDN1, and OCLN; i.e., 4Rs) by flow cytometry. No binding to E2c3s was detected for cells that expressed only SR-B1 because E2c3 has a truncated HVR1, which would normally take part in SR-B1 binding (4). Substantial binding was detected for cells that expressed only CD81 or all four entry factors (4Rs) (Fig. 6C). Similar to the biolayer interferometry experiments, locking the E2 front layer in the A conformation resulted in additional reduction in binding compared to the G440S or W616S single mutants (Fig. 6C).

The binding experiments above support direct interaction of CD81 with the back-layer residues or, alternatively, disruption of the CD81bs by the mutations. Notably, in addition to exposure of the back-layer residues, conformational changes of the front layer result in a spatial change of front-layer residues that are likely required for CD81 binding (cluster 1; fig. S6D). In the absence of a high-resolution E2-CD81 complex structure, we used a mutagenesis data-driven modeling methodology to map CD81 binding to the E2 A and B conformations (Fig. 6D). On the basis of previously reported mutagenesis data (summarized in fig. S6D), residues that are critical for the E2-CD81 complex interactions were selected and used in the molecular modeling methodology implemented in the program HADDOCK (high ambiguity driven protein-protein docking) (21). The resulting models show clear differences in the interactions between the two E2 conformations: In the A conformation, CD81 interacts with the surface comprising the front layer and CD81 binding loop, and most of the predicted interactions are formed with the front-layer residues (cluster 1; fig. S6D). In the B conformation, due to conformational changes of the front layer, CD81 could then interact with the core region of E2 and directly with the back-layer (cluster 3; fig. S6D) and front-layer N termini, but not with cluster 1 residues. In both complexes, interaction with the CD81 binding loop W529 was predicted, but not with other CD81 binding loop residues (e.g., Y527 or D535; fig. S6D) that were previously reported as being critical for CD81 binding (cluster 2) (10, 19, 22).

These binding experiments (Fig. 6) are consistent with the notion that the back layer directly interacts with CD81 and suggest that flexibility of the front layer is required for CD81 binding. However, the binding experiments and molecular modeling do not answer the question of which conformation (A, B, or both) is the E2 conformation recognized by CD81. The answer will come when the E2-CD81 complex structure is eventually solved.


Entry of an enveloped virus into host cells to initiate infection involves complex events that include different conformational states as well as rearrangements of the viral envelope glycoproteins to mediate receptor binding, endocytosis, and membrane fusion of the virus with the host cell (23). A deeper knowledge of the transition between the different envelope glycoprotein states and conformations is critical for a better understanding of the entry mechanism; however, acquiring this knowledge is challenging due to the metastable nature of the envelope glycoproteins (23). The entry mechanism of HCV is still not fully understood and, therefore, it can be assumed that some of the epitopes that are vital for viral entry (e.g., receptor binding site or fusion peptide) might be transiently exposed but available for targeting by nAbs. Therefore, structural studies of nAb-antigen recognition can shed light on HCV entry mechanism and design of structure-based vaccines to elicit protective antibodies.

HCV entry into host cells is a multistep process that is still not fully understood and involves interactions of E2 with its CD81 receptor. In the absence of a high-resolution structure of the E2-CD81 complex, E2-CD81 interactions have been previously mapped by mutagenesis and antibody competition experiments (fig. S6D): E2 residues important for CD81 binding were mapped to the front layer (W420 and W437LAGLFY443), CD81 binding loop (Y527SWG530 and D535), and back layer (Y613 and W616) (Fig. 1A and fig. S6D) (10, 19, 22, 24, 25). Similarly, E2 interacting residues were mapped to helices C and D of CD81-LEL (fig. S6E) (20). Previous x-ray crystallographic and negative stain electron microscopy studies of an E2 core construct in complex with bnAb AR3C localized the CD81 binding site on E2c to a solvent-exposed hydrophobic surface that included the front layer and CD81 binding loop (10). However, these structures did not explain our finding here of involvement of the unexposed back-layer Y613 and W616 in CD81 binding (fig. S6D). On the basis of structural and high-throughput mutagenesis analyses, we have previously suggested that the back-layer residues indirectly affect CD81 binding by structural stabilization of the front layer (fig. S6F) (19). Our E2-nAb structures determined here reveal exposure of back-layer and β-sandwich residues, suggesting that the interactions with CD81 could potentially involve direct interaction with back-layer residues.

Exposure of back-layer and β-sandwich residues in the E2 B conformation raises the question of whether a change in the front-layer structure could be part of the entry mechanism. To date, 10 structures of E2-nAb complexes are available: AR3A-D, HEPC3/74, U1, HC11, HC1AM, and 212.1.1 (8, 10, 11); in 8 of them (except for HC1AM and 212.1.1), the front layer adopts the A conformation. Given now that nAbs isolated from infected humans can bind to both A and B conformations, it appears that both conformations are biologically relevant and exist, even if one is more transient, on the HCV envelope. The relevance of the B conformation that involves full exposure of Y613 (fig. S5H) is also supported by the report of Y613 phosphorylation by Lck tyrosine kinase in T cells for immune evasion (26). Yet, it is not clear which front-layer conformation is bound by CD81. Mutation of back-layer Y613 and W616 residues, as well as introduction of a disulfide bond to lock the front layer to the back layer, reduced CD81 binding, suggesting either direct interaction of CD81 with the E2 back layer or indirect disruption of the CD81 binding site.

Analyses of the structures here confirm that CDRH2 of VH1-69–encoded antibodies binds to a hydrophobic pocket formed by the front layer and CD81 binding loop; however, E2 structures with Fabs 212.1.1 and HC1AM indicate some variation in the interaction of CDRH2 with the hydrophobic pocket. The conformational changes of the front layer and exposure of the back-layer hydrophobic residues reveal a wider pocket for greater interaction with CDRH2 (Fig. 4A). Notably, the crystal structure of E2-HEPC3/74 complexes (8) also indicated some variation in CDRH2 binding as CDRH2 interacts with the front layer but not with the CD81 binding loop (8). In addition, binding and neutralization experiments indicate that the hydrophobic CDRH2 tip (residues 53 and 54) is critical for binding and neutralization of AR3-targeted VH1-69–encoded bnAbs (Fig. 4). Together, analysis of the structures here along with previous ones (8, 10, 11) suggests that the hydrophobic nature and plasticity of the neutralizing face allow it to mediate interactions with different bnAbs varying in CDRH2 sequence and CDRH3 loop length.

Previous structural studies have paved the way toward a molecular understanding of the immune recognition of HCV by bnAbs, thus providing a molecular template for rational vaccine design. Diverse strategies for structure-based vaccine design have been described and are under evaluation (20, 2729). Here, we investigated additional VH1-69 class bnAbs and identified a previously unknown E2 conformation, the B conformation. These structures provide additional high-resolution structural information on HCV vulnerabilities and an additional template to guide design of HCV vaccine antigens to elicit antibody responses that block binding of E2 to its CD81 receptor.

In summary, we have shown that E2 can adopt at least two conformations when bound to different bnAbs. Therefore, the two conformations are biologically relevant for understanding immune recognition of HCV and could be important for vaccine design. Whether these conformations are related to the conformational changes of E2 upon CD81 binding remains to be determined. Additional studies will therefore be needed to determine whether these or other conformational changes are involved in HCV entry to host cells and which conformations are able to elicit more potent bnAbs to aid in structure-based vaccine design.


Expression and purification of soluble E2, IgGs, and Fabs

The E2 constructs and mAbs AR1A, AR3A, and U1 were expressed and purified as previously described (10, 11). The HC and LC sequences of mAbs HC11, 212.1.1, 212.10, HC1AM, HC84.26AM, and HC84.1 were synthesized by GeneArt (Invitrogen, USA), cloned into the phCMV3 vector, and expressed in ExpiHEK293F or ExpiCHO cells (Thermo Fisher Scientific). The mAbs were purified on a protein G affinity column followed by size exclusion chromatography using a Superdex 200 column (Pharmacia) in 50 mM NaCl, 20 mM Tris-HCl (pH 7.2) buffer.

Generation of virus stocks of HCVcc recombinants

First-passage virus stocks of HCVcc recombinants were generated by passaging transfection supernatants onto naïve Huh7.5 cells and culturing until the virus spread to >80% of the cells, at which time supernatant was harvested and filtered. Collected supernatant was analyzed by infectivity titration and sequencing of the encoded envelope as described previously (30).

Neutralization assay

HCVcc neutralization was performed as previously described (31). In brief, JFH1-based Core-NS2 recombinant viruses were incubated with dilution series of the indicated VH1-69–encoded bnAbs in four replicates along with eight replicates of virus only. IC50 (50% inhibitory concentration) titers were calculated in GraphPad Prism 8 using three-parameter dose-response regression curve fitting. Control antibody b6 at the highest dose did not inhibit the viruses. The IC50 neutralization values of mAbs AR3A, AR3B, AR3C, AR3D, and U1 and 212.10 and 212.1.1 were previously determined by us using the same experimental protocol and research reagents (11, 30). HCVpp were generated by cotransfection of 293T cells with pNL4-3.lucRE plasmid and the corresponding expression plasmids encoding the E1E2 genes at a 4:1 ratio by polyethylenimine as previously described (32). Virus infectivity was detected based on the expression level of the reporter gene firefly luciferase (Promega), and neutralization was calculated as the virus infectivity inhibited at the antibody concentrations indicated divided by the infectivity without antibody after background subtraction. The background infectivity of the pseudotype virus was defined by infecting cells with virus made only with pNL4-3.lucRE. Pseudotype virus particles displaying the vesicular stomatitis virus envelope glycoprotein G were used as a control for nonspecific neutralization. The virus was incubated with the antibodies for 1 hour at 37°C before adding to Huh-7 cell monolayers and incubated for 5 hours at 37°C. Expression of the luciferase reporter gene in the infected cells was measured with a luminometer on day 3 after infection.

KD determination

KD values were determined by biolayer interferometry using an Octet RED instrument (ForteBio Inc.). IgGs at ~10 μg/ml in a kinetics buffer [1× phosphate-buffered saline (PBS) (pH 7.4), 0.01% bovine serum albumin, and 0.002% Tween 20] were immobilized onto protein G–coated biosensors. A twofold concentration gradient of E2, starting at 10, 20, or 40 μM depending on the binding affinity, was used in a titration series of five steps. The kon and koff values of each E2 protein for each IgG were measured in real time to determine KD values. All binding data were collected at 30°C in one experiment.

Enzyme-linked immunosorbent assay

Competition ELISA to map Fab E1 and U1 epitopes (fig. S2A) was performed as previously described (32). To study the role of the mAb LC in antigen binding (fig. S3A), recombinant E1E2 antigens (isolate H77) from cell lysate produced by transient transfection of human embryonic kidney (HEK) 293T cells were captured onto ELISA microwells precoated with Galanthus nivalis lectin (GNL) at 5 μg/ml in PBS overnight 4°C. After blocking with 4% (w/v) nonfat dry milk in dilution buffer (PBS + 0.02% Tween 20) for 30 min, titrated mAbs [wild type (WT) or IgGs in which the LC was switched to the HIV-1 b12 or influenza 2D1 LCs] in dilution buffer + 1% nonfat dry milk were added to the microwells and incubated for 1 hour at room temperature. The mAbs were detected with secondary antibody horseradish peroxidase (HRP)–conjugated goat anti-human IgG Fc (Jackson ImmunoResearch) and developed with 3,3′,5,5′-tetramethylbenzidine (TMB) substrate (Thermo Fisher Pierce). The reaction was stopped with 2 N sulfuric acid, and the plates were read at an absorbance of 450 nm. To study the role of E2 back-layer Y613 and W616 on CD81 and mAb binding (Fig. 6, A and B), recombinant antigens (E1E2 WT or mutants) were captured onto ELISA precoated microwells with GNL and blocked with 4% (w/v) nonfat dry milk. Fc-CD81-LEL or mAb IgGs (at 10 or 2 μg/ml, respectively) were added to the microwells, incubated for 1 hour at room temperature, detected with secondary antibody, and developed with TMB substrate. The relative binding to the mutated antigen was calculated as the percentage by comparing the optical density signal versus the WT signal after subtraction of the background signal (empty vector). The results are an average of three or two experiments.

Isothermal titration calorimetry

ITC binding experiments were performed using a MicroCal Auto-iTC200 instrument (GE Healthcare). Before titration, all proteins were dialyzed against a buffer containing 20 mM Tris and 150 mM NaCl (pH 7.4). Protein concentrations were determined by the absorbance at 280 nm. In the syringe, concentrations were between 79 and 126 μM for Fabs, 163 μM for Fc-CD81-LEL, and 57 μM for MBP-CD81. In the cell, H77 E2c3 was between 6.4 and 7.5 μM. Duplicate or triplicate experiments were performed with the following parameters: cell at 25°C, 16 injections of 2.5 μl each or 25 injections of 1.5 μl each (for 212.1.1 and 212.10), injection interval of 180 s, injection duration of 5 s, and reference power of 5 μcal. Origin 7.0 software was used to fit the integrated titration peaks using a single-site binding model.

Differential scanning calorimetry

Thermal melting curves of HCV E2 glycoproteins were obtained with a MicroCal VP-Capillary calorimeter (Malvern). The purified E2 and mutants from 293S cells were buffer-exchanged into 1× PBS and concentrated to 16 to 33 μM before analysis. Melting was probed at a scan rate of 90°C h−1 from 10° to 120°C. Data processing, including buffer correction, normalization, and baseline subtraction, was conducted using the standardized protocol from the Origin 7.0 software.

Crystallization and structural determination of Fabs and E2c3-Fab complexes

The E2-Fab complexes were formed by overnight incubation of purified E2 and Fabs in a molar ratio of 1:1.25 (E2:Fab) or 1:1.25:1.25 (E2:Fab:Fab) at room temperature followed by size exclusion chromatography (Superdex 200) to remove unbound Fabs using 20 mM Tris and 50 mM NaCl (pH 7.2) buffer. Crystallization experiments were performed using the vapor diffusion sitting drop method at 20°C. Crystals of HK6a E2c3-Fab U1, HK6a E2c3-Fab HC11, H77 E2c3-Fab 212.1.1-Fab E1, H77 E2c3-Fab HC1AM-Fab E1, Fab AR3B apo, and Fab HC1AM apo were obtained that diffracted to 2.40, 2.35, 2.62, 3.70, 1.65, and 2.60 Å, respectively (table S1). Crystals of the HK6a E2c3-Fab U1 complex were obtained using a reservoir solution of 20% (w/v) PEG-3500 (polyethylene glycol, molecular weight 3500), 0.2 M di-ammonium hydrogen citrate (pH 5.0); HK6a E2c3-Fab HC11 from 0.16 M ammonium sulfate, 20% (w/v) PEG-4000, 20% (v/v) glycerol, 0.08 M sodium acetate (pH 4.6); H77 E2c3-Fab 212.1.1-Fab E1 from 20% (w/v) PEG-3000, 0.2 M NaCl, 0.1 M Hepes (pH 7.5); H77 E2c3-Fab HC1AM-Fab E1 from 10% (w/v) PEG-8000, 0.2 M MgCl2, 0.1 M tris (pH 7.0); Fab AR3B apo from 20% (w/v) PEG-3350, 0.2 M Na thiocyanate, 10% ethylene glycol; and Fab HC1AM apo from 17% (w/v) PEG-4000, 15% (v/v) glycerol, 8.5% (v/v) isopropanol, 0.085 M Na-Hepes (pH 7.5). Before data collection, HK6a E2c3-Fab U1, H77 E2c3-Fab 212.1.1-Fab E1, and H77 E2c3-Fab HC1AM-Fab E1 crystals were cryoprotected with 10 to 15% ethylene glycol and flash-cooled in liquid nitrogen. Diffraction datasets were collected at the Advanced Photon Source (APS) and the Stanford Synchrotron Radiation Lightsource (SSRL) (table S1). Data were integrated and scaled using HKL-2000 (33). The structures were solved by the molecular replacement using Phaser (34) with the H77 E2c-Fab AR3C, HK6a E2c3-Fab AR3A, or Fab AR3B (PDB entries 4MWF, 6BKB, and 6BKC) as search models. Structure refinement and model building were conducted as previously reported (11). Final refinement statistics are summarized in table S1.

Cell lines

BRL3A rat hepatocytes (American Type Culture Collection CRL-1442) and Huh7.5 cells were grown in Dulbecco’s modified Eagle’s medium (Invitrogen) supplemented with 10% fetal bovine serum (FBS) and a mixture of 1% penicillin/streptomycin at 37°C and 5% CO2. Plasmids expressing full-length human receptor CD81, SR-BI, OCLN, or CLDN1 were transduced into BRL3A cells as previously described (35) to generate BRL3A-CD81, BRL3A-SR-BI, and BRL3A-4Rs (CD81-SR-BI-OCLN-CLDN1). Transduced cells were maintained in blasticidin (5 μg/ml; BRL3A-CD81), geneticin (500 μg/ml; BRL3A-SR-BI), or a mixture of blasticidin (5 μg/ml), geneticin (500 μg/ml), hygromycin (200 μg/ml), and puromycin (1 μg/ml; BRL3A-4Rs).

Flow cytometry analysis

To validate the expression of human receptors on BRL3A cells, ~106 cells were stained with corresponding mAbs for 30 min in the dark at 4°C and washed twice with fluorescence-activated cell sorting (FACS) buffer (2% FBS and 2 mM EDTA in PBS). The expression of human CD81 and SR-BI was detected by phycoerythrin (PE) anti-human CD81 antibody (clone 5A6; BioLegend) and allophycocyanin (APC) anti-human CD36L1 antibody (clone m1B9; BioLegend), respectively. Expression of OCLN and CLDN1 was detected using anti-CLDN1 antibody (clone 2H10D10, Life Technologies) and anti-OCLN antibody (clone OC-3F10, Life Technologies), followed by APC anti-mouse IgG antibody (Jackson ImmunoResearch). To analyze the binding of E2 to human receptors, ~106 of BRL3A WT cells or transduced cells were incubated with E2 proteins (5 μg/ml) (WT or mutants) for 1 hour at room temperature. The cell-bound E2 was incubated with human anti-E2 mAb AR1B for another 1 hour and then detected by APC anti-human IgG Fc antibody (clone HP6017, BioLegend). Samples were acquired with an LSR II flow cytometer (BD Biosciences), and the data were analyzed by FlowJo v10 (Tree Star).

Mutagenesis data-driven analysis of CD81-E2 complex

To determine possible modes of interaction of CD81 with the two possible conformation of E2, we used a mutagenesis data-driven modeling methodology (10, 21). Briefly, residues implicated in the binding interaction between E2 and CD81 (W420, W437LAGLFY443, Y527SWG530, D535, Y613, and W616 for E2; I162, K171, I181, I182, and L185FKED189 for CD81; fig. S6D) were input into HADDOCK (21). HADDOCK uses a knowledge-based approach, using experimental data in conjunction with structures to drive the docking of two macromolecules. Residues implicated for binding were designated active targets (residues on E2 and CD81), and neighboring residues were designated passive targets. For this study, combinations of active and passive targets were used over multiple HADDOCK runs to minimize bias. The first stage of docking consisted of randomization of orientations and rigid body energy minimization, which yielded hundreds of solutions. The resulting structures were subjected to semirigid simulated annealing in torsion angle space and final refinement in Cartesian space with explicit solvent resulting in 121 and 162 structures that were clustered into 15 clusters each for H77 E2c3 and E2c complexes, respectively. A representative structure from the highest scoring complex in the top-most clustered was used for the analysis (Fig. 6D).


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: This work was funded, in part, by NIH grants AI079031 and AI123861 (to M.L.), AI123365 and AI106005 (to M.L. and I.A.W.), the Independent Research Fund Denmark (DFF-4004-00598 and DFF-8020-00391B to J.B.), the Novo Nordisk Foundation (NNF17OC0029372 and NNF19OC0054518 to J.B.), the Lundbeck Foundation (R303-2018-3396 to R.V.-M. and R324-2019-1375 to J.P.), and the Candys Foundation (to J.P., E.H.A., and J.B.). We thank X. Dai and M. Elsliger for crystallographic and computational support and H. Tien in the Wilson laboratory for automated crystal screening. X-ray datasets were collected at the APS beamline 23ID-B (GM/CA CAT) and SSRL beamline 12-2. The use of the APS was supported by the U.S. Department of Energy (DOE), Basic Energy Sciences, Office of Science, under contract DE-AC02-06CH11357. The use of the SSRL Structural Molecular Biology Program was supported by DOE Office of Biological and Environmental Research and by the NIH NIGMS (including P41GM103393) and the National Center for Research Resources (P41RR001209). Author contributions: N.T., I.A.W., and M.L. designed the project. N.T., Y.H., E.G., and Z.-Y.K. expressed and purified the E2 proteins and mAbs. N.T. determined and analyzed the x-ray structures. N.T. and E.G. determined KDs and assessed binding. E.H.A., R.V.-M., J.P., and J.B. performed HCVcc neutralization assays. N.T. and K.N. performed HCVpp neutralization assays. N.T., F.C., and M.D. performed flow cytometry analysis. R.U.K. and N.T. performed mutagenesis data-driven analysis. N.T., R.L.S., S.K.H.F., J.B., I.A.W., and M.L. analyzed results and wrote the manuscript with support from all authors. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data and code to understand and assess the conclusions of this research are available in the main text, the Supplementary Materials, and the PDB (accession codes 6WO3, 6WO4, 6WO5, 6WOQ, 6WOS, and 6WOR). Additional data related to this paper may be requested from the authors. This is manuscript 29982 from The Scripps Research Institute.

Stay Connected to Science Advances

Navigate This Article