INTRODUCTION
Hepatitis C virus (HCV), a positive-strand RNA virus belonging to the
Flaviviridae family, infects nearly 3% of the world's population (
1). In approximately 70 to 80% of patients, HCV establishes a chronic infection in the liver that can lead to cirrhosis, liver failure, and hepatocellular carcinoma (
2). HCV exhibits a high degree of genetic variability and is classified into seven major genotypes, each of which contains a large number of related subtypes (
3,
4). This diversity and the high level of intrahost variability (quasispecies) contribute to virus persistence in the infected hosts. The recently developed new therapies have profoundly improved cure rates. However, the higher costs associated with these new medications are expected to limit their wider utilization (
5–7). As yet, no vaccine against the virus is available.
HCV entry into target cells is believed to be mediated by a multistep process involving the interplay of the viral envelope glycoproteins E1 and E2 and several host cell factors, such as heparan sulfate, tetraspanin CD81, scavenger receptor class B type I (SR-BI), and the tight junction (TJ) proteins claudin-1 (CLDN1) and occludin (
8). E1 and E2 are transmembrane proteins with extensive N-linked glycosylation (4 and 11 N-linked glycosylation sites, respectively) consisting of a large N-terminal ectodomain and a C-terminal hydrophobic anchor (
9). The ectodomain of the E2 protein contains three highly variable regions. Hypervariable region 1 (HVR1; residues 384 to 411), located at the N terminus of E2, plays an important role in HCV entry, antibody binding, and disease outcome (
10). It is now well established that E2 binds CD81 and SR-BI and that these interactions are a prerequisite for virus entry (
10–13). However, the precise role of the E1-E2 envelope protein complex in HCV entry is still unclear.
The viral glycoprotein E2 is the major target for neutralizing antibodies. The majority of broadly neutralizing anti-E2 antibodies isolated to date target epitopes spanning the reported CD81 binding sites of E2. Importantly, mouse monoclonal antibody (MAb) AP33 (
14), rat MAb 3/11 (
15), and human MAbs HCV1 (
16), HC33 (
17), and Hu5B3.v3 (
18) block the interaction of E2 with CD81 by binding to linear epitopes located within the highly conserved E2 site encompassing residues 412 to 423, referred to as antigenic site 412 (AS412) (
19) or epitope I (
20). Other MAbs recognize discontinuous E2 epitopes overlapping the CD81 binding site on E2 and involving residues 395 to 424, 425 to 447, and/or 523 to 540 (
21,
22). Residues 412 to 423 have been proposed to be a potential target for HCV vaccine design (
18,
23–26).
Although two independent crystal structures of HCV E2 have recently been reported (
27,
28), they do not provide any structural data for the region from residues 412 to 423. Indeed, the construct used by Khan and coworkers (
27) spanned E2 residues 456 to 656, whereas in the structure described by Kong and coworkers (
28), the N-terminal portion encompassing the region from residues 412 to 423 is disordered in the crystal. Interestingly, a peptide representing this antigenic site adopts a β-hairpin conformation when it is complexed with neutralizing MAb AP33, Hu5B3.v3, or HCV1, in which Leu413, Gln415, Gly418, and Trp420 are key residues directly involved in the hydrophobic binding surface (
18,
24–26). Since the hairpin-like structure of the antibody-bound peptides suggests that the region from residues 412 to 423 adopts a similar conformation in the context of the protein, we designed and prepared a cyclic variant of the epitope to help the fragment to assume a hairpin-like structure. The cyclic peptide was used to immunize mice with a view to isolating novel anti-E2 MAbs. By a subtractive screening approach, we selected a subset of MAbs recognizing only the cyclic antigen to explore the possibility of generating AP33-like antibodies with an improved binding affinity and neutralization potency. We reasoned that such an approach might generate antibodies with the potential to more efficiently capture and lock the epitope in a conformation close to that which it putatively adopts in the context of the protein in complex with neutralizing antibodies. However, these MAbs unexpectedly failed to neutralize virus infection. The results of X-ray studies of the complex of one of these MAbs with the cyclic peptide, together with antigen-antibody reactivity data, provide possible explanations for the lack of neutralization activity and offer novel insights into the design of vaccine candidates targeting the antigenic site from residues 412 to 423.
MATERIALS AND METHODS
Reagents.
TRIzol was purchased from Invitrogen (Carlsbad, CA, USA). The protein G and A columns and reagents for surface plasmon resonance (SPR) were purchased from GE Healthcare. Freund's complete and incomplete adjuvant and all media, including Dulbecco's modified Eagle's medium (DMEM), Opti-MEM medium, and serum, were purchased from Gibco (Life Technologies, Italia) and Sigma-Aldrich (Milan, Italy). RPMI 1640 medium containing 10% fetal bovine serum (FBS) and 1% penicillin, streptomycin, and glutamine (PSG) was used for maintaining myeloma cells. RPMI-GM medium contained RPMI 1640 medium with 1% nonessential amino acids. For the fusion of splenocytes and myeloma cells, we used RPMI 1640 medium supplemented with 15% FBS, 2% hypoxanthine-aminopterin-thymidine (HAT), 1% PSG, and 10% hybridoma-enhancing supplement (HES). The same medium lacking HES was used for hybridoma clone selection. Reagents for peptide synthesis were from Novabiochem (Laufelfingen, Switzerland), Inbios (Naples, Italy), and GL Biochem (Shanghai, People's Republic of China). Solvents for peptide synthesis and purification were from Romil (Dublin, Ireland). Other reagents were from Sigma-Aldrich (Milan, Italy). Keyhole limpet hemocyanin (KLH) was from Pierce-Thermo Fisher (Milan, Italy). The broadly neutralizing mouse MAb AP33 has been described previously (
14,
29). It was generated following immunization of BALB/c mice with a mammalian cell-expressed recombinant secretory form of the HCV genotype 1a strain Gla E1E2 lacking their respective transmembrane domains (
30). Soluble E2 (sE2), comprising the region from amino acids 384 to 661 of genotype 1a strain H77 (GenBank accession no. AF009606) was expressed by infecting High Five insect cells with a recombinant baculovirus, and the protein secreted into the medium was purified by Ni-nitrilotriacetic acid chromatography.
Peptide design, synthesis, and purification.
Peptides were prepared under standard conditions of 9-fluorenylmethoxy carbonyl solid-phase synthesis (
31). A cyclic peptide corresponding to the region from residues 412 to 422 of E2 (sequence, QLINTNGSWHI) was designed to facilitate the adoption of a hairpin-like conformation by the linear epitope. The structures of the complexes of the linear peptides corresponding to epitope 412-422 with MAbs AP33, Hu5B3.v3, and HCV1 (
24–26) were used as a guide in the design. The peptide was generated by replacing Ile411 and Asn423 of E2 with two cysteines, which were then oxidized to form a disulfide bridge (
Fig. 1A). Two lysine residues were added at the N terminus to allow conjugation to KLH and to bovine serum albumin (BSA) via glutaraldehyde (amine-to-amine cross-linking). Here this peptide is named C-epitope I. The linear variant (here named L-epitope I) used in the experiments as a control was obtained by alkylation of the cysteine thiols. Alanine-mutated variants of C-epitope I were designed and prepared under the same conditions described above to investigate the contribution of specific residues to recognition by antibodies. To overcome the structural similarity of alanine with some of the native residues, we replaced the dyad Gly417-Ser418 with glutamic acids in peptide mutant III. After synthesis and purification, peptides were cyclized as reported elsewhere (
32). To suppress cysteine reactivity, the linear peptide was methylated with methyl iodide as reported previously (
33). Methylation was chosen as the modification introducing the minimum structural change within the molecule.
CD.
The purified cyclic peptide was characterized by circular dichroism (CD) using a Jasco J-710 spectropolarimeter (Jasco Corp.) equipped with a Peltier temperature control system and a 110-QS quartz cuvette with a 1.0-cm path length. The spectra of peptide solutions at 0.1 mM in 10 mM phosphate buffer, pH 7.0, were collected using the following settings: wavelength range, 190 to 280 nm; scanning speed, 20 nm/min; data pitch, 0.2 nm; bandwidth, 1 nm; and response time, 4 s. At least five independent readings of the spectra were averaged and smoothed.
Immunogen preparation.
One milligram of C-epitope I was conjugated with 3.0 mg of carrier protein (KLH or BSA) in 2.0 ml of 20 mM phosphate buffer (pH 7.0) containing 0.2% (vol/vol) glutaraldehyde by stirring the mixture for 3 h. One milliliter of 1 M glycine in water was added to block the reaction, and then the solutions were extensively dialyzed against phosphate-buffered saline (PBS; pH 7.4) before being lyophilized. The amount of peptide-protein conjugate was determined using a Bio-Rad kit (Bio-Rad, Milan, Italy). The same procedure was used to prepare glutaraldehyde-self-conjugated BSA (BSA
2). BSA
2 was used as a control in enzyme-linked immunosorbent assays (ELISAs) to exclude clones producing antibodies potentially recognizing the amine-cross-linked glutaraldehyde (
34).
Immunization of mice and generation and purification of MAbs.
BALB/c mice were housed and handled according to institutional guidelines (project identification code 2013/0038120, approved by the Ethical Animal Care and Use Committee of the University of Naples Federico II; date of approval, 24 April 2013). Four 5-week-old BALB/c mice (The Jackson Laboratory) were immunized with 100 μg of KLH-conjugated peptide emulsified with complete Freund's adjuvant. Four independent injections were carried out subcutaneously with 25 μl immunogen. Before immunization, 250-μl blood samples were taken from the caudal vein of each mouse and used as the preimmune control (time zero samples). Mice were boosted with the same amount of immunogen in incomplete Freund's adjuvant at day 30 after the first immunization. Blood samples were taken from the caudal vein (250 μl) before every subsequent boosting and tested by ELISA to monitor the antibody titer. A final antigen boost was administered intravenously to mice showing the highest antibody titer 20 days before they were sacrificed and splenectomized, as described below.
Cells harvested from the spleens of sacrificed animals were fused with myeloma SP2/0 cells (ATCC) at a ratio of 5:1 in RPMI-GM containing polyethylene glycol (PEG) 1300-1600 (Hybri-Max; Sigma-Aldrich, Milan, Italy) and 7.5% dimethyl sulfoxide (Sigma-Aldrich, Milan, Italy) as described previously (
35). The fused hybridoma cells were resuspended in 30 ml of selection medium consisting of RPMI-GM medium containing PEG 1300-1600, 10% FBS, 100 U/ml penicillin, 100 μg/ml streptomycin, 100 μM hypoxanthine, 16 μM thymidine, and 400 nM aminopterin (RPMI-HAT; Sigma-Aldrich, Milan, Italy). The cell suspension was dispensed into 96-well plates and incubated at 37°C in a 5% CO
2 atmosphere with periodic replenishment with fresh selection medium. After 12 to 14 days, the cell medium was screened by ELISA for binding to C-epitope I and to its linear analogue. Hybridomas secreting antibodies with strong reactivity with C-epitope I (but not L-epitope I) were recloned twice by limiting dilution, and their reactivity was reconfirmed by ELISA. Subcloned hybridoma cells were cultured in Opti-MEM medium containing 10% FBS, gradually adapted to serum-free cell culture medium, and then transferred to bioreactors (Integra Biosciences AG, Chur, Switzerland) for large-scale antibody production. The antibodies were purified to homogeneity by protein G affinity chromatography followed by gel filtration. Antibodies were characterized by SDS-PAGE, analytical size-exclusion chromatography, and Western blotting. The immunoreactivity of purified antibodies with antigen was determined by ELISA, as described below.
Fab fragment generation and purification.
Briefly, Fab fragments were prepared by papain digestion of purified IgGs. The reaction, monitored by SDS-PAGE, was optimized in 20 mM sodium phosphate and 10 mM EDTA (pH 7.0) buffer using papain at a 1:100 (wt/wt) ratio (Sigma-Aldrich, Milan, Italy) for 3 h at 37°C. The Fc portion was removed using a HiTrap protein G column (GE Healthcare, Milan, Italy), and then the Fab fragment was further purified by gel filtration on a Sephadex 75 column (GE Healthcare, Milan, Italy) in PBS or 25 mM Tris-HCl, 100 mM NaCl, pH 7.5. The concentration and purity of the antibodies and Fab fragments were estimated by determination of the absorbance at 280 nm using a NanoDrop 2000 spectrophotometer, SDS-PAGE, and size-exclusion high-performance liquid chromatography.
ELISAs.
Titrations of antibody in mouse serum and screening of hybridoma supernatants were performed by ELISA as described elsewhere (
36). Hybridomas were screened for their ability to secrete MAbs specific for the cyclic antigen. For this purpose, supernatants were tested by coating plates with both BSA-conjugated C-epitope I and BSA-conjugated L-epitope I. Also, BSA
2 was used as a control to exclude clones producing antibodies recognizing the BSA-linked glutaraldehyde. Positive clones were stabilized by 3 sequential rounds of limiting dilution passages in 96-well plates. Following incubation for 2 weeks at 37°C under 5% CO
2, the supernatant from each well was again tested for the presence of antibodies against the peptide antigen. Hybridomas secreting antibody at the highest titer were further subjected to 3 rounds of limiting dilution cloning. Finally, the antibodies secreted by selected hybridomas were purified and tested by ELISA for reactivity to the free C-epitope I and its linear variant. Briefly, the wells of ELISA plates (MaxiSorp; Nunc) were coated with the appropriate peptide at 0.5 μg/ml in PBS by incubation overnight at 4°C. The wells were blocked with BSA and then washed with PBS-Tween 20 as described above. Antigen-coated wells were incubated with 0.5 μg/ml of appropriate antibodies at 37°C for 60 min, washed as described above, and then further incubated with horseradish peroxidase (HRP)-conjugated anti-mouse IgG (1:1,000; Bio-Rad) at 37°C for 60 min. After washing, the bound antibody was detected by adding peroxidase substrate solution (prepared by dissolving
o-phenylenediamine dihydrochloride at 0.4 mg/ml in 0.1 M citric acid and 0.2 M Na
2HPO
4 buffer [pH 4.8] and adding 0.2 μl/ml of 30% H
2O
2). Following incubation at room temperature in the dark, the reaction was stopped with a 2.5 M H
2SO
4 solution, and the optical density at 492 nm was determined using a microplate reader (BioTek, Winooski, VT, USA). Specificity was assessed using an unrelated monoclonal antibody (trastuzumab; Genentech) as a negative control (NC) and MAb AP33 as a positive control. The negative and positive controls were used at the same concentration as the test MAbs.
An ELISA to detect MAb binding to mammalian cell-expressed sE2, full-length (FL) E2, and the E1E2 glycoprotein was performed essentially as described previously (
14,
37). Briefly, HEK293T cells were cotransfected with plasmids containing the appropriate sequence, and the expressed glycoproteins present in the clarified lysates of these cells were captured on
Galanthus nivalis agglutinin (GNA)-coated Immulon II enzyme immunoassay (EIA) plates (Thermolab Systems). Anti-E2 MAbs were added, and bound glycoproteins were detected with an anti-mouse immunoglobulin G-horseradish peroxidase (Sigma, United Kingdom) and 3,3,5,5-tetramethylbenzidine (TMB; Sigma, United Kingdom) substrate. Absorbance values at 450 nm were determined. Assays for the screening of the clones were performed at least twice in quadruplicate. Data are reported as the average of the results from all experiments and replicates ± standard deviation (SD). Assays for MAb binding to peptides and recombinant proteins were performed at least twice. Data are reported as the average of the results from all experiments and replicates ± SD.
Sequencing of antibody variable regions and isotype determination.
To clone the Ig heavy (H)-chain and light (L)-chain variable-region genes encoding a subset of the anti-E2 MAbs, total RNA from hybridoma cells was first used to generate cDNA libraries by reverse transcription reactions with a SuperScript III first-strand kit (Invitrogen) using random hexamers. The IgG Fab fragments corresponding to the antigen-binding variable regions were then amplified by PCR using a set of primers specific for mouse Ig heavy and light chains. The nucleotide sequences of the PCR products were determined, and their amino acid sequences were deduced using the program Translate from the ExPASy proteomic server. The isotypes of the MAbs were determined using a monoclonal antibody isotyping kit (IsoStrip; Pierce, Rockford, IL, USA) according to the manufacturer's instructions.
SPR analysis.
All SPR analyses were performed on a Biacore 3000 instrument from GE Healthcare, using CM5 sensor chips and the HBS buffer certified for use with this instrument (20 mM HEPES, 0.15 M NaCl, pH 7.2, 0.005% polysorbate 20), at 25°C. Immobilization was carried out by means of the canonical amine-coupling chemistry using the surface immobilization wizard procedure operating at 5 μl/min. Channels were activated with a mixture of 1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide and N-hydroxysuccinimide for 7 min; then, the ligand, which had been appropriately diluted in the preselected sodium acetate buffer, was coupled until a typical level of 500 to 600 response units (RU) was achieved. The remaining active ester groups were blocked with 1 M ethanolamine HCl, pH 8.5. Measurements of SPR binding to L-epitope, C-epitope I, and the alanine-mutated variants were carried out by immobilizing each antibody or Fabs on the CM5 sensor chips. Immobilization of the antibodies, including AP33, was efficiently performed at 5.0 μg/ml in 10 mM sodium acetate, pH 4.5, achieving immobilization levels of about 500 to 600 RU in every case. The same procedure was applied to prepare an sE2-functionalized sensor chip. An IgG1 isotype-matched antibody was used as a negative control at a concentration of 5.0 μM. For epitope mapping, the Fab fragment of the C2 antibody (Fab C2) at 5.0 μg/ml in 10 mM sodium acetate, pH 4.5, was directly immobilized onto the CM5 sensor chip (immobilization level, about 500 to 600 RU). An underivatized surface was prepared on every sensor chip and used as a blank control. All analyses were carried out at a flow rate of 20 μl/min, and a constant volume of 60 μl of protein or peptide solution appropriately diluted in the HBS running buffer was injected.
For every analysis, experimental sensorgrams were aligned, the blank signal was subtracted, and the sensorgrams were overlapped. All mathematical manipulations and fitting of the data were performed using BIAevaluation (version 4.1) software (GE Healthcare). All experimental data gave optimal fittings when they were processed by assuming a 1:1 Langmuir binding interaction.
HCVcc neutralization assays.
Cell culture-infectious HCV (HCVcc) was produced by electroporation of Huh7 human hepatoma cells with viral genomic RNA that was generated by
in vitro transcription using plasmid pUC-JFH1 or pUC-JFH1-N417T as the template as described previously (
38,
39). Virus neutralization assays were performed using Huh7-J20 reporter cells, and virus infectivity levels were determined by a secreted alkaline phosphatase (SEAP) reporter assay, as described previously (
40). Briefly, Huh7-J20 cells were plated out at a density of 5 × 10
3 per well in a 96-well plate. Virus was preincubated with the test antibody at 37°C for 1 h, prior to infection of the cells at a multiplicity of infection of 0.1. At 3 h postinfection, the inoculum was replaced with fresh DMEM. At 72 h postinfection, the SEAP reporter activity (which correlates directly with the virus infectivity level) in the medium of infected cells was determined as described previously (
40).
Crystallization and diffraction data collection.
Trials of the crystallization of the complex consisting of Fab C2 and the peptide were set up at 293 K using the hanging-drop vapor diffusion method. The peptide and the Fab were previously mixed at a molar ratio of 4:1. Preliminary screenings of the crystallization conditions were carried out using commercially available sparse-matrix kits (Crystal Screen I/II and Index kits; Hampton Research) (
41). These screenings yielded microcrystals that were optimized by fine-tuning of the protein and precipitant concentrations. Crystals suitable for crystallographic investigations were obtained using a Fab concentration of ∼5.0 mg/ml and 0.2 M ammonium sulfate and 25% (wt/vol) PEG 3350 in a buffer containing 0.1 M bis-Tris, pH 5.5.
Diffraction data were collected in-house at 100 K using a Rigaku Micromax 007 HF generator that produces Cu Kα radiation and that is equipped with a Saturn944 charge-coupled-device detector. Data were collected at 100 K by adding to the precipitating solution a solution of 20% (vol/vol) ethylene glycol as a cryoprotectant. The data set was scaled and merged using the HKL2000 program package (
42). Although two angles of the unit cell were numerically close to 90°, the crystals of the complex were triclinic. Indeed, all attempts to process the data at a higher symmetry yielded very high
Rmerge values. The analysis of the Matthews coefficient (
Vm) value of this crystal suggests the presence of two molecules in the asymmetric unit.
Crystallographic refinement.
The structure of the complex was solved by molecular replacement using phaser crystallographic software (
43). Starting models for the light and the heavy chains were selected by looking for structures in the Protein Data Bank (PDB) with the highest sequence identities with the sequences of the chains of our Fab. Using this approach, the starting models for the heavy and light chains were extracted from the structures with PDB accession numbers
2VL5 (identity, 82%) and
3DGG (identity, 94%), respectively. Taking into account the variability of the relative orientation of the constant and variable regions, both the heavy and the light chains were fragmented by considering the individual Fab domains. Therefore, since the asymmetric unit contains two independent molecules, an ensemble of eight individual fragments constituted the starting structures in the molecular replacement search. The application of this procedure provided a straightforward solution. This model was used for automatic rebuilding, which was carried out using the ARP-wARP model-building tool (
44). Crystallographic refinement was carried out against 95% of the measured data using the ccp4i program suite. The remaining 5% of the observed data, which was randomly selected, was used for
Rfree calculations to monitor the progress of refinement. Noncrystallographic restraints were applied in the Refmac macromolecular refinement program (
45) with medium restraints for the main chain atoms and loose restraints for side chain atoms. Manual modeling was performed using the Coot tool (
46). Water molecules were incorporated into the structure in several rounds of successive refinements.
MD studies.
In order to gain insights into the intrinsic conformational properties of the cyclic peptide, a molecular dynamics (MD) simulation was conducted using the structure of the peptide detected in the complex with Fab C2 as the starting model. The molecular dynamics simulation was performed using the GROMACS software package (version 4.5.5) (
47), the AMBER99sb force field, and TIP4P as the water model. The peptide was immersed in a cubic box of 4.50 by 4.50 by 4.50 nm
3 containing 2,933 water molecules. The simulation was run with periodic boundary conditions. The temperature and pressure of the systems were stabilized at 300 K and 1 atm, respectively. Energies were minimized by fixing the protein atoms, and then the simulation was run without restraints. The time scale of the simulation was 250 ns with a time step of 0.002 ps. The particle mesh Ewald (PME) method (grid spacing, 0.12 nm) was used to calculate the electrostatic interactions. A cutoff of 10 Å was applied to treat Lennard-Jones interactions. Bond lengths were constrained using the Linear Constraint Solver (LINCS) algorithm. Trajectories were checked to assess the quality of the simulation using GROMACS routines. H-bond interactions were identified on the basis of the cutoffs for the hydrogen donor-acceptor angle (∼30°) and the donor-acceptor distance (∼3.5 Å) by using GROMACS utilities (
47).
Protein structure accession number.
The coordinates of the model and the experimental structure factors of the complex have been deposited in PDB under accession number
5EOC.
DISCUSSION
The elevated costs associated with current antiviral therapies against hepatitis C and the high disease prevalence necessitate the urgent development of alternative therapeutic approaches. As for many other viral diseases, a vaccine would be the most obvious and a less expensive option. However, the high genetic variability of the virus is a major barrier that has so far prevented the generation of effective anti-HCV vaccines. Several studies have indicated that the HCV surface glycoprotein E2 is a major target for neutralizing antibodies. However, they are generally isolate specific and do not recognize E2 proteins from other HCV genotypes, thus preventing their use as broad-spectrum neutralizing reagents. The conserved region from residues 412 to 422 of E2, encompassing Trp420, a key residue for HCV recognition by the human receptor CD81 (
9,
11), is recognized by several neutralizing antibodies (
13–17). Despite the crucial role that the residues of this epitope play in HCV entry, no information on the conformation(s) that this region adopts in the context of the E2 protein is available. Structural studies on complexes of the E2 peptide from residues 412 to 423 with three neutralizing MAbs have shown a tendency of this fragment to adopt a β-hairpin conformation (
18,
24–26). In this framework, to gain insights into epitope 412-422 conformational preferences, we generated and characterized a set of MAbs that selectively recognize a conformationally restrained cyclic variant of this epitope but not its linear counterpart. Binding assays demonstrated that such MAbs were able to bind the soluble E2 protein with
KDs of about 50 nM. This finding holds an interesting implication. Considering that the MAbs are unable to recognize the linear peptide representing the sequence of epitope I, their ability to bind E2 suggests that the protein context has an impact on the epitope structure, likely shifting its conformation toward bent states. However, the affinity for E2 exhibited by these MAbs is significantly lower (approximately 500-fold) than that exhibited by the neutralizing MAb AP33, and this feature largely contributes to the lack of activity of the new antibodies.
A crystallographic analysis of the complex between the Fab fragment of the C2 MAb and the cyclic peptide was undertaken to unravel the basis of the reduced affinity of the MAbs for the E2 protein and of the consequent inability to neutralize either HCVpp or HCVcc. A comparison of the crystal structure of the C-epitope I–Fab C2 complex with the structure of the linear peptide bound to the Fab fragment of hu5B3.v3, AP33, or HCV1 (
18,
24–26) shows both analogies and differences. First, in all of these complexes, the peptide adopts a β-hairpin conformation stabilized by H bonds established with main chain atoms. Moreover, these peptide-Fab complexes feature a similar buried surface area (∼250 Å
2), with an important contribution to intermolecular interactions being provided by the side chain of Trp420 (
Fig. 6D and
E) (
18,
24–26). These observations indicate that the C2 MAb, despite its inability to neutralize virus infection, shares with hu5B3.v3, AP33, and HCV1 two important features related to the overall peptide conformation and to the tight binding to the Trp420 side chain.
A deeper comparison of these complexes underlines, however, some distinct features in the recognition of the cyclic peptide by C2. In particular, the binding of C-epitope I by Fab C2 appears to be rotated by 180° compared to the orientations of the other complexes (
Fig. 8A to
D). This leads to differences in the interacting surfaces between the linear and cyclic variant in the complex with their MAbs (
Fig. 9). In particular, residues such as Leu413 and Asn415, which interact with AP33 and HCV1, are fully exposed in the C2 complex. On the other hand, the side chain of Asn417, which is glycosylated in E2 and exposed to the solvent in the other complexes, is slightly buried upon C2 binding (
Fig. 6D). It is worth mentioning that the failure of C2 to neutralize variants such as N417T, which cannot be glycosylated at position 417 (
18), suggests that the partial burying of Asn417 does not play a major role in determining the distinctive behavior of C2.
Finally, there is a slight shift in the residues involved in the hydrogen bonding patterns of the hairpin in the peptide-C2 complex compared to the residues involved in the other complexes. The previously reported structures show that the hairpin is stabilized by H bonds between residues 414 and 421 and between residues 412 and 423, whereas the conformation of the cyclic peptide bound to C2 is stabilized by H bonds between residues 412 and 422 and between residues 414 and 420.
Predictive analyses, carried out using the PEP-FOLD server (
http://bioserv.rpbs.univ-paris-diderot.fr/services/PEP-FOLD/), suggest that the structure observed for the cyclic peptide in the complex is similar to the conformation intrinsically accessible to the linear sequence of epitope 412-422. Indeed, the RMSD values of the cyclic peptide structure against the top two solutions provided by the PEP-FOLD server are in the range of 1.9 to 2.0 Å (
50).
Together, these observations may in principle explain the reduced affinity of C2 MAb for E2 and suggest that the lack of neutralization may originate from a combination of a low antibody affinity and an inappropriate approach to the epitope. Indeed, despite the analogies with other complexes, the specificities of the C2 complex to the cyclic variant in terms of peptide H-bond patterns and side chain conformations, as well as in terms of the relative orientation of the epitope and MAb, may limit the ability of C2 to recognize E2. This implies that the conformations of the linear peptides observed in the complexes with HCV1 and AP33 represent a reliable model of the structure of the epitope in the context of real E2.
In this framework, the absence of a significant neutralizing effect of the MAbs generated against the cyclic variant is likely due to the inability of this peptide to fully reproduce the epitope conformation of the linear variant observed in the complexes with AP33 and HCV1, which was supported by our MD analysis. However, it should be noted that while the manuscript was in preparation, two novel complexes of neutralizing Fabs of MAbs 3/11 and HC33.1 with the linear epitope were reported (
51,
52). Surprisingly, in both of these new complexes, the bound peptides assumed rather extended conformations that were completely unrelated to those observed in the complexes with AP33 and HCV1. These new data strongly suggest that epitope 412-422 is, in the protein context, endowed with a remarkable structural versatility, a property believed to be an additional mechanism of neutralization escape (
51,
52). The observation that the epitope conformation recognized by the Fabs of 3/11 and HC33.1 may differ from that recognized by the Fabs of AP33 and HCV1 suggests that the specificities of the conformation of the cyclic peptide recognized by C2 may not be the only factor responsible for the inactivity of this MAb.
It is likely that our MAbs, which recognize with high selectivity a conformationally restrained variant of epitope 412-422, are unable to accommodate all of its accessible structural states. In other words, these MAbs are able to bind only a subpopulation of the diverse conformational ensemble adopted by epitope 412-422, a property that reflects the reduced affinity of our MAbs for the E2 protein and their inability to display any significant neutralizing activity. On the other hand, neutralizing MAbs such as AP33, which recognize both cyclic and linear variants of epitope 412-422, have a high affinity for E2 since they are able to capture different conformational states of this E2 region. These considerations should be taken into account in the selection of HCV-neutralizing MAbs and for the design of new potential vaccines.
In conclusion, our data corroborate the emerging notion that epitope 412-422 is characterized within the protein context by a high degree of conformational versatility which likely contributes to a mechanism of conformation-driven neutralization escape. Since rigid and flexible regions in a protein are typically characterized by conserved and variable sequences, respectively, further studies are needed to clarify why and how the E2 epitope 412-422 region combines a structural flexibility with a highly conserved local sequence.