|Home | About | Journals | Submit | Contact Us | Français|
The identification and characterization of broadly neutralizing antibodies (bnAbs) against HIV-1 has formed a major research focus, with the ultimate goal to help in the design of an effective AIDS vaccine. One of these bnAbs, 2F5, has been extensively characterized, and residues at the apex of its unusually long complementarity-determining region (CDR) H3 loop have been shown to be crucial for neutralization. Structural studies, however, have revealed that the 100TLFGVPI100F apex residues of the CDR H3 loop do not interact directly with residues of its core gp41 epitope. In an attempt to gain better insight into the functional role of this element, we have recombinantly expressed native 2F5 Fab and two mutants in which either the apical Phe100B(H) residue was changed to an alanine or the CDR H3 residues 100TLFGVPI100F were replaced by a Ser-Gly dipeptide linker. Isothermal titration calorimetry (ITC) and competitive-binding enzyme-linked immunosorbent assays (ELISAs) rendered strikingly similar affinity constants (Kd [dissociation constant] of ~20 nM) for linear peptide epitope binding by 2F5 Fabs, independent of the presence or absence of the apex residues. Ablation of the CDR H3 apex residues, however, abolished the cell-cell fusion inhibition and pseudovirus neutralization capacities of 2F5 Fab. We report competitive ELISA data that suggest a role of 2F5 CDR H3 apex residues in mediating weak hydrophobic interactions with residues located at the C terminus of the gp41 membrane proximal external region and/or membrane components in the context of core epitope binding. The present data therefore imply an extended 2F5 paratope that includes weak secondary interactions that are crucial for neutralization of Env-mediated fusion.
Examples of diseases for which antibodies elicited by vaccines are able to protect against new infections include smallpox, diphtheria, tetanus, pertussis, pneumococcal pneumonia, hepatitis A, hepatitis B, varicella, measles, rubella, polio, rabies, and influenza (4, 38). In light of the predominant role played by the humoral branch of the immune system in protecting human hosts against these pathogens, it is evident that one of the major goals in the quest for an AIDS vaccine is the elicitation of neutralizing antibodies that would confer protection against HIV-1 infection. Past attempts at eliciting antibodies capable of neutralizing HIV-1 viruses of diverse genetic backgrounds have met with only limited success, although very recent studies have presented more encouraging results (5, 7, 8, 16, 19, 27, 39, 50).
One avenue guiding immunogen design that aims at eliciting neutralizing antibodies is to characterize the neutralizing ability of the sera found in some HIV-1-infected individuals. Indeed, recent studies have shown that a considerable proportion (approximately 25%) of HIV-1-positive subjects who have been infected for at least 1 year show moderate to broad neutralizing antibody responses (15, 43, 48, 51). Furthermore, approximately 1% of the HIV-1-infected subjects studied had an unusually high neutralizing potency against many HIV-1 clades (48). Characterizing the targets of naturally produced neutralizing antibodies could give important clues as to the necessary composition of an effective immunogen.
Although some HIV-1-infected individuals possess broadly neutralizing sera in which combinations of neutralizing monoclonal antibodies act together to neutralize a wide genetic range of viruses, monoclonal antibodies that are able to effectively neutralize various HIV-1 clades on their own are very rare (46). Of the few well-characterized ones, broadly neutralizing antibody (bnAb) b12 recognizes the site on gp120 responsible for binding to the CD4 receptor during the initial stages of fusion (14, 59). The epitope of bnAb 2G12 is a mannose cluster on the outer face of gp120, whereas bnAbs 2F5, Z13, and 4E10 recognize linear sequences of the membrane-proximal external region (MPER) of gp41 (9, 10, 29-31, 44, 45). In addition, two new HIV-1 broadly neutralizing antibodies, PG9 and PG16, have been reported; they recognize an epitope on the gp120 trimer and show particularly great potency and breadth (48, 55). Few would oppose the idea that understanding the neutralizing mechanism of these rare bnAbs at the molecular level could yield invaluable clues for the rational design of a vaccine immunogen.
The interaction between 2F5 and its minimal linear epitope (662ELDKWAS668) has been extensively characterized in numerous structural studies (6, 24, 37). Briefly, the primary gp41 epitope as recognized by 2F5 assumes a β-turn conformation, with core residues 664DKW666 in the center of the turn. These amino acids are flanked by residues adopting an extended conformation at the N terminus and a canonical α-helical turn at the C terminus, respectively (24). One interesting characteristic of bnAb 2F5 is its remarkably long complementarity-determining region (CDR) H3 loop, which contains 22 amino acids. On average, human CDR H3 loops encompass between 10 and 14 residues, although even larger CDR H3 loops, with up to 26 residues, have been reported (57). Although most residues at the base of the 2F5 CDR H3 are involved in direct recognition of the core 662ELDKWAS668 epitope, along with the other five CDRs (Fig. (Fig.1A)1A) (12, 24), the role of residues located at the apex of the CDR H3 loop is still a matter of debate. In a previous Ala scanning mutagenesis study, Zwick et al. demonstrated the involvement of this loop in neutralization and suggested that residues located at its apex play a role in core epitope recognition. This interpretation was based on the results of binding to both core peptide epitope and gp41 ectodomain in direct enzyme-linked immunosorbent assays (ELISAs) (60). However, such an involvement in core epitope binding is not apparent from structural studies (24, 37). Other studies documenting some affinity of 2F5 for membrane components have suggested that this loop interacts directly with viral membranes (2, 18, 26, 28, 42).
To investigate the role of the 2F5 CDR H3 apex residues in more detail, we first needed to produce sufficiently large amounts of recombinant 2F5 Fab of high purity. Expressing Fab in a bacterial system allowed us to easily create and produce mutants with changes in the 2F5 CDR H3 apex sequence. In an attempt to characterize the functional role of this element, two mutants were created. In one, the CDR H3 sequence of 100TLFGVPI100F was replaced by a single Ser-Gly dipeptide, bridging the ~5-Å distance between the Cα atoms of T99 and A100G, the last two residues represented by electron density in our structural analysis of the native 2F5 Fab′ (24). In the other mutant, the apical Phe100B(H) of the CDR H3 was changed to an alanine residue. Isothermal titration calorimetry (ITC) measurements, as well as competitive-binding ELISAs, confirmed that the 2F5 CDR H3 apex does not directly affect core epitope recognition. In contrast, the correct amino acid composition of this Fab element was required for potent neutralization of pseudovirus infection and inhibition of Env-induced cell-cell fusion. With the aim of unraveling the molecular interactions responsible for this loss of neutralization, we performed binding ELISAs of the 2F5 CDR H3 apex mutants with various extensions of the gp41 epitope, as well as with membrane components. We observed clear differences in binding affinities between native and mutant versions of 2F5 Fab when residues comprising the 4E10 epitope and beyond were added to the C terminus of the core 2F5 epitope and, even more so, if this extended 2F5 epitope was inserted into a phospholipid bilayer environment. In contrast, when probed against membrane components alone, no such differentiation was detected. The importance of these findings in further understanding the neutralization mechanism of the bnAb 2F5 at a molecular level is discussed in detail.
2F5 residues are numbered following the standard Kabat convention, and gp41 residues are numbered according to their positions in the full amino acid sequence of HXBc2 Env gp160.
Peptides representing different fragments of the gp41 MPER, 2F5ep (656NEQELLELDKWASLWN671), 2F5preTM (656NEQELLELDKWASLWNWFNITNWLWYIK683), and 2F5preTM(9,10)Ala (656NEQELLELAAWASLWNWFNITNWLWYIK683) were synthesized as C-terminal carboxamides by solid-phase synthesis using 9-fluorenylmethoxy carbonyl (Fmoc) chemistry and purified by high-pressure liquid chromatography at the Proteomics Unit of the University Pompeu-Fabra (Barcelona, Spain). Stock solutions for these peptides were prepared in dimethyl sulfoxide (DMSO; spectroscopy grade). The peptide 662ELDKWAS668, at a >75% purity level, was purchased from Genscript (Piscataway, NJ), and its stock solutions were prepared in 20 mM Tris, pH 8.0. 1-Palmitoyl-2-oleoylphosphatidylcholine (POPC), cholesterol (Chol), and 1-palmitoyl-2-hydroxy-sn-glycero-3-phosphate (sodium salt [PA]) were purchased in powder form from Avanti Polar Lipids (Birmingham, AL), and stock solutions were prepared in chloroform. Recombinant gp41 HXB2 (aa 541 to 682; glycosylated), produced in Pichia pastoris, was purchased from Viral Therapeutics, Inc. (Ithaca, NY). In this construct, the 2F5 core epitope sequence does not contain any glycosylation sites.
The genes encoding the 2F5 Fab (light chain residues Ala1 to Cys214 and heavy chain residues Ile2 to Cys216) were synthesized (BioBasic, Inc., Toronto, Ontario, Canada) and subsequently cloned into the pET-Duet 1 vector (Novagen, Gibbstown, NJ). Because the initial cloning of these genes made use of the constant region of another Fab (8F9) (53), the following deviations from the 2F5 consensus sequence are present in the 2F5 Fab constant region of all constructs: light chain 103RVDVR107 to 103KLEIK107 and E193A, as well as heavy chain 132AGGA135 to 132SGGT135, T114A, T116F, T195I, and R210K. The thrombin site-containing DKTHLVPRGSSSHHHHHH tag was added at the C terminus of the 2F5 heavy chain for subsequent nickel affinity purification. Site-directed mutagenesis was used to create 2F5 Fab mutants T15(H)A and F100B(H)A (BioBasic, Inc., Toronto, Ontario, Canada). For the 2F5 Fab delta CDR H3 mutant, residues 100TLFGVPI100F forming the top part of the CDR H3 loop were replaced by a simple Ser-Gly dipeptide linker. The resulting plasmids were transfected into Escherichia coli Rosetta-gami 2 competent cells following a standard heat shock protocol, and 2F5 Fab expression was subsequently induced with 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG). Cells were grown at 18°C for 36 h in Superbroth medium containing appropriate antibiotics. After centrifugation, cells were resuspended in lysis buffer (50 mM HEPES, pH 7.5, 500 mM NaCl, 5 mM imidazole, 5% glycerol, and 100 mM iodoacetamide) containing DNase, lysozyme, and EDTA-free Complete protease inhibitor (Roche, Mannheim, Germany). Cell lysis was performed by sonication. The cell lysate was loaded on a nickel affinity chromatography column (Qiagen, Mississauga, Ontario, Canada). After washing, the protein was eluted with 250 mM imidazole and further purified using a HiLoad Superdex 75 prep-grade gel filtration column (GE Healthcare, Piscataway, NJ). From 12 liters of culture medium, two to five milligrams of Fab were obtained reproducibly. Purity was estimated as approximately 80% by semiquantitative analysis of SDS-PAGE results using ImageJ software (1). Proper folding of the Fab fragments was confirmed by circular dichroism in a Jasco J-810 spectropolarimeter.
The nonglycosylated gp41 ectodomain construct used in this study was designed based on the one previously described by Tan et al. (52). It spans residues 535 to 669 (which includes the 2F5 epitope) of gp41, with an SGGRGG linker replacing native residues 580 to 627. An E. coli codon-optimized DNA sequence for this construct was synthesized (BioBasic, Inc., Toronto, Ontario, Canada) and cloned into the pEt-26b vector (Novagen, Gibbstown, NJ) using the NdeI and XhoI restriction sites. As a result, the LE(H6) sequence was added at the C terminus of the gp41 construct. The plasmid was transfected into E. coli BL21(DE3) cells using a standard heat shock protocol. Cells containing the plasmid were grown in kanamycin-containing LB broth, induced with 0.1 mM IPTG, and grown at 37°C overnight. After centrifugation, the cells were resuspended in a Tris-buffered saline (TBS) solution. Cell lysis was performed by sonication in the presence of lysozyme and DNase. After centrifugation, inclusion bodies were washed with TBS, centrifuged, and subsequently dissolved in binding buffer (50 mM HEPES, pH 7.5, 500 mM NaCl, 5 mM imidazole, 5% glycerol) containing 8 M urea. After centrifugation, the supernatant was loaded directly onto nickel affinity chromatography resin (Qiagen, Mississauga, Ontario, Canada). After washing, the protein was eluted with the binding buffer solution containing 250 mM imidazole. Fractions containing the gp41 construct were collected, combined, and dialyzed against 2 liters of 20 mM formate, pH 3.0, at 4°C for 16 h. Proper folding of the gp41 construct into its postfusion bundle of predominantly α-helical secondary structure was verified by circular dichroism using a Jasco J-810 spectropolarimeter.
Large unilamellar vesicles (LUVs) containing POPC, Chol, and PA (2:1:0.6) were prepared by the extrusion method (21, 22). For extrusion, a mini extruder set (Avanti-Polar Lipids, Birmingham, AL) with Whatman Nuclepore membranes with pores of 0.4, 0.2, and 0.1 μm was used according to the manufacturer's protocol. For producing peptido-liposomes, 2F5preTM peptide stock solution in DMSO was added to the LUVs at a 1:100 peptide-to-phospholipid ratio. The partitioning of peptides on the surface of vesicles using this method has been previously characterized (22).
Calorimetric titration experiments were performed with a high-sensitivity VP-ITC MicroCalorimeter (MicroCal, Northampton, MA) at 25°C. Proteins were dialyzed against 5 mM HEPES and 100 mM NaCl, pH 7.4 (dialysis buffer), overnight at 4°C. Protein and peptide solutions were degassed for 5 min before measurements were taken. The antibody solution (3 to 5 μM) in the calorimetric cell was titrated with 2F5ep peptide dissolved in dialysis buffer (concentration 20 to 40 μM) in successive injections of 10 to 12 μl. The corresponding heat of peptide dilution into buffer was used to correct the data. The experimental data were analyzed according to a 1:1 binding model by means of Origin 7.0 (MicroCal, Inc.). The fit of the binding curve yields the stoichiometry, the association constant (ka), and the enthalpy (ΔH), of the binding reaction. The free energy of binding (ΔG) and the entropy (ΔS) are determined by the basic thermodynamic expressions ΔG = −RTlnka = ΔH − TΔS, where R and T are the gas constant and the absolute temperature, respectively.
Syncytium formation assays were carried out using CHO-Env and TZM-bl as effector and target cells, respectively (obtained from the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, and contributed by C. Weiss and J. White and J. Kappes, respectively). Fusion was inhibited by incubating the 2F5 Fab constructs with CHO-Env cells for 90 min prior to coculturing them with TZM-bl cells, following the contributors’ specifications.
Pseudoviruses were produced by transfection of human kidney HEK293T cells with the full-length env clone pHXB2-env (AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, contributed by K. Page and D. Littman) using calcium phosphate, together with vectors pWXLP-GFP and pCMV8.91 encoding, respectively, a green fluorescent protein and an env-deficient HIV-1 genome (kindly provided by Patricia Villace, CSIC, Madrid, Spain). After 24 h, the medium was replaced with Optimem-Glutamax II (Invitrogen Ltd., Paisley, United Kingdom) without serum. Two days after transfection, the pseudovirus particles were harvested, passed through 0.45-μm-pore sterile filters (Millex HV, Millipore NV, Brussels, Belgium) and finally concentrated by ultracentrifugation in a sucrose gradient. Fab-induced neutralization was determined using TZM-bl target cells. Serial dilutions of Fabs were set up in duplicate in 96-well plates and incubated for 1 h at 37°C with a 10 to 15% tissue culture infectious dose of pseudovirus. After Fab-pseudovirus coincubation, 10,000 target cells were added in the presence of 15 μg/ml DEAE-dextran (Sigma-Aldrich, St-Louis, MO). Neutralization levels after 72 h were inferred from the reduction in the number of green fluorescent protein (GFP)-positive cells as determined by flow cytometry using a BD FACScalibur flow cytometer (Becton Dickinson Immunocytometry Systems, Mountain View, CA).
Binding of the recombinant 2F5 Fab to its gp41 epitope was evaluated by direct ELISA with standard 96-well plates (Cayman Chemical Company, Ann Arbor, MI) coated with gp41 ectodomain constructs. Briefly, 50 μl of the recombinant gp41 at a concentration of 20 μg/ml in 0.1 M sodium bicarbonate, pH 9.6, was incubated at room temperature for 2 h to coat the 96-well plates. Subsequently, the plates were blocked with 200 μl of 5% nonfat milk dissolved in TBS (blocking buffer) at 4°C for 16 h. Serial dilutions of 2F5 Fab ranging from 100 μg/ml to 0.001 μg/ml were set up in triplicate and added to the gp41-coated wells. After 2 h of incubation at room temperature and three washings with 200 μl of TBS, the secondary anti-human IgG (Fab specific)-alkaline phosphatase antibody produced in goat (Sigma-Aldrich, St-Louis, MO) was added to the plates at a 1:30,000 dilution in blocking buffer. Plates were subsequently incubated for 2 h at room temperature. After washing, the reaction was developed by adding 100 μl per well of alkaline phosphatase yellow (pNPP) liquid substrate system for ELISA (Sigma-Aldrich, Oakville, Ontario, Canada) and analyzed after 30 min using a Spectramax 340PC instrument (Molecular Devices, Sunnyvale, CA) at a wavelength of 405 nm.
Competitive ELISAs were performed using the coating and blocking steps outlined above. Prior to adding to the gp41-coated plates, 1 μg/ml of 2F5 Fab was preincubated for a minimum of 30 min with serial dilutions of competitor ranging from 1 mM to 10 nM and set up in triplicate. Competitors included gp41 constructs in 0.1 M sodium bicarbonate (pH 9.6), ELDKWAS peptide in 20 mM Tris (pH 8.0), 2F5ep peptide in DMSO, 2F5preTM peptide in DMSO, 2F5preTM(9,10)Ala peptide in DMSO, POPC/Chol/PA (2:1:0.6) LUVs, and POPC/Chol/PA (2:1:0.6) LUVs with 2F5preTM peptide. All LUVs were in a solution of 5 mM HEPES, pH 7.4, 100 mM NaCl. Incubation with secondary antibody and development with pNPP were performed as described for the direct ELISA protocol outlined above.
Direct ELISAs of binding to membrane components were performed following a protocol slightly modified from the one described previously (28). Briefly, chloroform stock solutions of POPC, cholesterol, and PA were combined and diluted with ethanol to 10 to 100 nmol/ml; a range of 1 to 10 nmol/well was used as the antigen. Control wells contained 0.05 ml of the appropriate solvent to match the lipid plated, and readings from these control wells were subtracted from the readings of matching experimental wells for analysis. After evaporation overnight, the plates were blocked with TBS-0.3% gelatin for 2 h. Subsequent steps were performed as described above, with the exception that the detection could only be performed after an overnight incubation with the pNPP substrate (as opposed to a 1-h incubation) due to a weak initial signal.
ELISA data were analyzed with the help of GraphPad Prism 5 software. For competitive ELISAs, absolute values were normalized and converted to the percentage of maximum response. Sigmoidal dose-response and log[inhibitor] versus response nonlinear regression analyses were used to derive 50% effective concentrations (EC50s) and 50% inhibitory concentrations (IC50s) for direct ELISA and competitive ELISA results, respectively. Finally, a one-way analysis of variance (ANOVA) followed by a Bonferroni post analysis test (significance level of 0.05) were used for the statistical analysis of IC50s obtained for the different Fab constructs.
2F5 Fab was expressed in E. coli Rosetta-gami 2 cells. To characterize the role of the 2F5 CDR H3 apex residues, a 2F5 Fab mutant (termed delta CDR H3) in which a Ser-Gly dipeptide linker replaces the seven CDR H3 apex residues 100 to 100F was designed (Fig. 1A and B). In a second mutant [termed F100B(H)A], the phenylalanine at position 100B of the heavy chain, located right at the tip of the CDR H3 loop, was replaced with an alanine, resulting in a minor change in overall hydrophobicity (Fig. (Fig.1A).1A). Finally, a Fab mutant with an unrelated threonine-to-alanine mutation at position 15 of the heavy chain [termed T15(H)A] was created to be used as a negative control (Fig. (Fig.1A).1A). Nickel affinity and size exclusion chromatography of native and mutant 2F5 Fabs resulted in the isolation of two to five milligrams per 12 liters of culture medium of correctly folded Fabs of approximately 80% purity (Fig. 1C and D).
To test for the epitope-binding capacities of the recombinant 2F5 Fab and its mutants, isothermal titration calorimetry measurements were performed. For this purpose, the peptide 2F5ep (656NEQELLELDKWASLWN671), representing the nominal 2F5 epitope, was titrated into a solution containing the 2F5 Fab constructs. The heat signal produced following epitope-peptide injection allowed the determination of the thermodynamic parameters of binding. Titration profiles are shown for 2F5ep/wild-type 2F5 Fab and 2F5ep/delta CDR H3 2F5 Fab (Fig. 2A and B, respectively), and the thermodynamic parameters of all 2F5 Fab constructs are presented in Table Table1.1. The binding of this complex is energetically driven by exothermic enthalpy (ΔH > 0) counterbalanced by unfavorable entropic (ΔS < 0) changes. Both the enthalpic and entropic terms contribute almost equally to the free energy of 2F5ep binding to wild-type 2F5 Fab and its mutants, resulting in very similar and tight binding affinities (Kd [dissociation constant] of ≈20 nM). These results constitute strong evidence that 2F5 CDR H3 apex residues are not involved in mediating core gp41 epitope interactions.
In an attempt to link the capacity for binding to the linear epitope with the functionality of the 2F5 Fab constructs, two types of assays were carried out. Performing both cell-cell fusion inhibition assays and pseudovirus neutralization assays allowed us to ascertain the activity of Fabs independently of the system or the detection methods used. As seen in Fig. 3A and B, the recombinant 2F5 Fab possesses the ability to abrogate gp41-mediated fusion with IC50s of approximately 10−7 M and 10−6 M in the pseudovirus infection and cell-cell mediated fusion assays, respectively. As expected, the negative-control mutant possessing the T15(H)A mutation, at a site unrelated to epitope recognition, is able to neutralize Env-mediated fusion at the same level as the wild-type 2F5 Fab. However, the Fab with the F100B(H)A mutation at the tip of the CDR H3 achieves half neutralization only at a concentration nearly 20-fold that of wild-type 2F5 Fab. Moreover, very limited, if any, inhibition of gp41-mediated fusion is observed with the delta CDR H3 Fab mutant at the highest concentration tested (i.e., 4 μM). Altogether, these striking functional results clearly indicate the importance of the proper residues at the 2F5 CDR H3 apex for 2F5 inhibition of Env-mediated fusion.
To support the ITC findings and to also allow for probing the binding of more complex systems, binding ELISAs were performed. First, direct-binding ELISAs were carried out by coating ELISA plates with two gp41 constructs: (i) a glycosylated gp41 ectodomain comprising residues 541 to 682 and (ii) a nonglycosylated gp41 construct, produced in E. coli cells, comprising residues 535 to 669 with an SGGRGG linker replacing the interhelix residues 580 to 627. As seen in Fig. 4A and B, the recombinant 2F5 Fab constructs bind to both gp41-coated plates. Interestingly, in this system, modifications of the CDR H3 apex seem to reduce the ability of the 2F5 Fab to bind to the gp41 constructs coating the ELISA plates. Furthermore, the extent of the binding difference seems to be influenced by which of the two gp41 constructs was used to coat the plate. It is unclear if this observed difference in binding is due to the presence/absence of glycans on gp41, extra residues at the C terminus, or less likely, to the deletion of the interhelix loop.
Next, competitive ELISAs were performed to assess the binding affinity of the different 2F5 Fab CDR H3 constructs for gp41 in solution. In this protocol, the 2F5 Fabs were preincubated with a gp41 competitor before being added to the gp41-coated plates. The nonglycosylated gp41 (535 to 669) construct was selected for coating the ELISA plates in these competition ELISAs because of the height of the signal observed in the previous direct ELISAs. As seen in Fig. 4C and D, no significant difference in binding was observed between the various 2F5 Fabs in competition ELISAs with both gp41 constructs. Although the apparent binding affinity in solution is low (no saturation of the signal is observed even at high μM concentrations of the constructs), it is clear from the competitive ELISA curves that all four Fabs behave in a similar way. The weak binding affinity observed for this gp41 construct stems from its six-helix-bundle postfusion conformation, which is a low-affinity conformation for recognition by 2F5, as previously described (17). These competitive ELISA results are in sharp contrast to what was observed in direct ELISAs. This difference between direct and competitive ELISA results for the behavior of the 2F5 Fab mutants strongly suggests that artifacts might be observed for the binding to the epitope adsorbed on the plate surface. Indeed, the height of the maximum signal observed for each Fab in competitive ELISAs was proportional to the apparent binding affinity observed in direct binding ELISAs (data not shown). This suggests that the 2F5 Fab CDR H3 mutants might have difficulty accessing their adsorbed core epitope. For this reason, competitive ELISA was the method of choice to subsequently investigate the relative binding affinities of the 2F5 Fab constructs to various epitopes.
Competition ELISAs were conducted with a variety of epitopes in an attempt to characterize the role played by the 2F5 CDR H3 apex residues. First, competitive ELISAs were performed with the ELDKWAS and 2F5ep peptides. As expected, no significant differences in apparent binding affinities (IC50s) were observed between the four Fabs with these peptides (Fig. 5A and B). This finding supports the results obtained by ITC and suggests that residues at the apex of the 2F5 CDR H3 do not affect binding to the core linear epitope. Subsequently, C-terminal residues 672 to 683 were added to the competitor peptide, making it a mimic of the full gp41 pretransmembrane segment (2F5preTM). With this extended construct, statistically significant differences in IC50s were observed between the wild-type 2F5 Fab and Fab mutant F100B(H)A (approximately one order of magnitude of difference) and between wild-type 2F5 Fab and Fab delta CDR H3 (approximately two orders of magnitude of difference) (Fig. (Fig.5C).5C). These results suggest that residues at the apex of the 2F5 CDR H3 loop play an important role in mediating interactions with the hydrophobic C terminus of this extended gp41 epitope peptide. To verify the specificity of 2F5 recognition in this system, the negative-control peptide 2F5preTM(9,10)Ala, which has two of the core epitope residues, Asp664 and Lys665, replaced by alanines, was used as a competitor and showed no apparent binding to the recombinant 2F5 Fab constructs up to the high μM range (data not shown). Thus, the contribution of the CDR H3 apex to the overall 2F5preTM binding process appears to depend on the initial specific recognition of the core epitope sequence. However, some caution must be exercised when interpreting these results due to the unknown folding state of this aromatic-rich peptide in aqueous solution (10% DMSO), e.g., the breaking up of hydrophobic peptide aggregates by the hydrophobic CDR H3 apex residues would in turn favor core epitope binding. Other possible roles of CDR H3 residues in mediating interactions with this C-terminally extended MPER peptide include unmasking the core 2F5 epitope from possible hydrophobic interactions with C-terminal residues, mediating conformation changes of the peptide that lead to 2F5 recognition of the core epitope in a β-turn conformation, and/or decreasing the off rate of Fab-peptide binding.
Furthermore, when the 2F5preTM peptide was inserted into the POPC/Chol/PA LUVs, the recombinant 2F5 Fab constructs showed even more significant differential binding (Fig. (Fig.5D).5D). Core epitope recognition in the context of this liposome system yielded statistically significant differences in IC50s that were of approximately one order of magnitude between the wild-type 2F5 Fab and Fab mutant F100B(H)A and of almost three orders of magnitude between wild-type 2F5 Fab and Fab delta CDR H3. Also of note, the overall binding of Fabs to the 2F5preTM peptide-liposome system was two orders of magnitude lower than the binding of Fabs to the 2F5preTM peptide alone. Whether this results from the concealed nature of the 2F5preTM peptide in a membrane environment decreasing its accessibility for 2F5 Fab recognition remains unclear.
The ability of the Fabs to interact with membrane components was assessed by both direct and competitive ELISAs. For the direct assays, ELISA plates were coated with POPC, cholesterol, and PA, and both native 2F5 Fab and the CDR H3 mutants displayed affinity for the coated membrane components (Fig. (Fig.6A).6A). Modifying the hydrophobicity of the 2F5 CDR H3 by either point mutation or deletion of the extended loop therefore did not abrogate affinity for the coated membrane components. Interestingly, it seems that the deletion of bulky hydrophobic residues in the extended 2F5 CDR H3 loop allowed for a better interaction between Fab and coated membrane components. Also of note in these experiments, the substrate development reaction needed to be carried out for an extended period of time before a signal could be observed, even when large amounts of membrane components were used to coat the ELISA plates, suggesting that these Fabs possess a rather low affinity for coated membrane components. Figure Figure6B6B depicts the results of competitive ELISAs using POPC/Chol/PA LUVs as competitor. Up to the high μM range, these vesicles were not able to compete off binding of the 2F5 Fab constructs to the gp41-coated plates. This finding further supports the previous observation that 2F5, when probed independently of core epitope binding, possesses a low affinity for membrane components (see Fig. Fig.5D5D for comparison).
The production of high-affinity antibodies to antigens is an essential step in protecting the host against undesired pathogens. In the course of HIV-1 infection, antibodies against HIV-1 epitopes are generated readily, although antibodies capable of neutralizing the virus generally only appear after prolonged infection (15, 43, 48, 51). Moreover, only in exceptional cases have infected individuals generated monoclonal antibodies that are capable of neutralizing HIV-1 potently across a broad range of clades. One of these antibodies, bnAb 2F5, has been extensively characterized in an effort to determine the molecular mechanism underlying its broad neutralization capability and with the objective of guiding vaccine immunogen design efforts. Although the core recognition site of 2F5 has been mapped to the 662ELDKWAS669 MPER linear sequence of gp41 and the atomic interactions of the core antigenic complex have been described in detail (6, 24, 37), the exact mechanism of broad HIV-1 neutralization by 2F5 remains elusive. Interestingly, no contacts have been identified between the core gp41 epitope and the central seven amino acids of the 22-residue-long 2F5 CDR H3 loop. The importance of these CDR H3 residues for neutralization has been documented previously (60). Many publications have hypothesized that this region of 2F5 might be involved in a secondary role other than core epitope binding (2, 18, 24, 28, 37, 41, 42). Indeed, it has been reported that antibodies elicited during the course of HIV-1 infection might display elevated polyreactivity compared to the reactivity of antibodies elicited during the course of other infections (33). It has also been suggested that the length of the CDR H3 loop, which plays a distinct role in determining antigen specificity, is related to the type of antigen recognized: antibodies raised against large antigens, such as viruses, have a tendency to have longer CDR H3 loops than antibodies responsive to smaller antigens, such as peptides (11, 23, 47). It is hypothesized that antibodies with longer CDR H3 loops have an extended binding site that allows them to insert into cavities within an antigen (47). The 2F5 CDR H3 might indeed play a crucial role in recognizing a recessed conserved epitope, stemming both from its location at the membrane-partitioning interface of the virus and from the potential oligomerization of the gp41 MPER.
In the present study, a 2F5 Fab mutant with replacement of CDR H3 residues 100 to 100F by a Ser-Gly dipeptide linker, termed delta CDR H3, and the single-site 2F5 Fab mutant F100B(H)A were purified in bacterial cells. Both cell-cell fusion inhibition and pseudovirus neutralization assays showed almost complete loss of neutralization capacity by the delta CDR H3 2F5 Fab mutant and more than an order of magnitude of reduction in the IC50 of neutralization by the F100B(H)A mutant. These assays provided an additional confirmation of both the functionality of the produced recombinant 2F5 Fabs and the crucial role the residues at the apex of the 2F5 CDR H3 loop play in virus neutralization.
In an attempt to characterize the binding determinants of the 2F5 CDR H3 apex residues that confer neutralizing capacity on 2F5, ITC, direct ELISA, and competitive ELISA measurements were performed. The ITC measurements of the different 2F5 Fab constructs clearly indicated that binding to the nominal epitope peptide was not affected by mutations at the CDR H3 apex and that affinity for all 2F5 Fab constructs was in the low nM range. The competitive ELISA measurements using core peptides as competitors also showed no difference in relative binding affinities for the different 2F5 Fab constructs. These results are in agreement with the structural data that show that the molecular interaction between the 2F5 paratope and core gp41 epitope residues is mediated by 2F5 amino acids other than those at the apex of the CDR H3 loop (6, 24, 37). Altogether, we believe that the present neutralization and binding affinity results represent strong evidence indicating that the 2F5 CDR H3 apex residues are essential for neutralization but are not involved in core epitope binding. Therefore, this suggests one or more secondary sites of interaction between 2F5 and its HIV-1 target.
In this study, we report differences in binding profiles from direct ELISA and competitive ELISAs for various 2F5 Fab mutants assayed with the same gp41 constructs. Indeed, 2F5 CDR H3 mutants were observed to bind differently to gp41 constructs in direct ELISAs, whereas they showed similar binding profiles in competitive ELISAs. Recognition of antibodies to the solid-phase adsorbed antigen may be limited by multiple factors, including the conformation of the antigens on the plate surface and steric repulsive and/or attractive interactions between the antibody molecules themselves (32, 35, 54). It is therefore not unlikely that removing a flexible hydrophobic moiety at the apex of the 2F5 antigen binding site would cause artifacts in apparent binding affinity for the gp41 epitope bound to the solid phase. In contrast, in a competitive ELISA, the free antibody available for binding to the adsorbed antigen is determined by the binding constant of the antibody to the competitor in solution (20). This allows for a more accurate comparison of the relative binding affinities of antibodies. Furthermore, this process is free of the mass transport limitations present in direct ELISAs (32, 34). Taken together, both the direct and competitive ELISA results form supporting evidence that the 2F5 CDR H3 apex is important for mediating access to a possibly recessed epitope, although any underlying specificity of the 2F5 CDR H3 apex residues is not located within the 2F5 core epitope itself.
In an attempt to determine if the 2F5 CDR H3 apex residues contributed to specific interactions with other components than its core epitope, additional ELISAs were performed. Experiments with the 2F5preTM(9,10)Ala peptide and POPC/Chol/PA liposomes as competitors revealed no such competition for 2F5 Fab binding up to high μM concentrations. These results indicate that 2F5 possesses low binding affinity, if any, for residues located at the C terminus of the gp41 MPER alone or for membrane components alone. In addition, the results of direct ELISAs suggest that the 2F5 hydrophobic CDR H3 apex contributes very little to interactions with lipid components on their own. In fact, it seems that removing the bulky 2F5 CDR H3 extended loop might even slightly improve the interaction between the Fab and lipid-coated plates, possibly by allowing easier access or by encouraging other types of interactions to take place. Such contacts might include ionic interactions between positively charged regions of the Fab and negatively charged phospholipid components of membranes, an interaction which has previously been proposed to be important in mediating possible viral membrane contacts by anti-MPER bnAbs (3, 24, 58). However, one should note that the binding observed in the present direct ELISAs probes the affinity for membrane components alone and outside a bilayer environment and, as such, the direct ELISA results might be significantly limited in their relevance for the description of binding events in vivo.
The results of the competitive ELISAs show that the hydrophobic CDR H3 apex residues only make a definitive contribution to binding when the residues 672WFNITNWLWYIK683 are added to the gp41 MPER C terminus and, even more so, if the resulting extended peptide is placed in a membrane bilayer environment. Indeed, quite significant differences in the binding affinities of the native 2F5 Fab and its CDR H3 mutants are observed for the extended epitope peptide 2F5preTM, both in solution and in a liposome environment. Taken together, our results and prior reports of weak interactions between 2F5 and both of these components individually (2, 18, 26, 28, 42, 56) allow us to hypothesize that the necessity of the apex of the elongated CDR H3 loop for 2F5 neutralization is caused by secondary interactions of much weaker affinity to either C-terminal MPER residues or components of a membrane bilayer or both, in the context of the energetically dominant core epitope binding. Indeed, a dual interaction of the extended CDR H3 of 2F5 with both membrane surfaces and C-terminally located gp41 MPER residues is feasible, since interfacial hydrophobicity is mainly based on aromatic and leucine residues, which can also contribute to establishing interactions among sequences embedded in a membrane milieu (40).
Recently, three other laboratories have reported on variations in the HIV-1 neutralization capacities of both 2F5 and 4E10 CDR H3 mutants with different hydrophobicity characteristics (3, 36, 58). All studies, including the one presented here, agree that the hydrophobic CDR H3 extended loops of 2F5 and 4E10 are critical for HIV-1 neutralization, although the proposed mechanisms by which this effect is to be achieved are different. From surface plasmon resonance measurements of 2F5 and 4E10 mutants, Alam et al. deduced a two-step neutralization model which postulates that the bnAb first preconcentrates on the membrane surface, mediated by hydrophobic residues in its extended CDR H3 loop. In a second step, the bnAb then interacts with its transiently exposed core epitope (3). A competing model, presented by Xu et al. and also suggested previously by others (49, 58), argues that the CDR H3 extended loop of 4E10, in conjunction with positively charged pockets near the paratope, is responsible for the extraction of the gp41 core epitope partitioned at the membrane interface. The data gathered in the present study are consistent with the hydrophobic potential of the residues contained in the extended CDR H3 loop of 2F5 playing a measurable role in antigen binding only after core epitope recognition has occurred, contributing in a rather limited fashion prior to the latter high-affinity interaction. Additional interactions of bnAb 2F5 with hydrophobic amino acids located C-terminally of the gp41 MPER and/or with lipid components in the context of core epitope binding could allow for a metastable immunological complex to form and prevent the fusion cascade from proceeding. Certainly for 2F5, the answer to the question of whether these additional hydrophobic interactions contribute to the extraction of its immersed MPER epitope or whether they support the prevention of repartitioning of the MPER to the membrane after being transiently exposed during the initiation of the fusion process requires further investigation.
Although humans possess the ability to make long CDR H3 loops, the unusual origins of the D(H) segment of bnAbs 2F5 (52 bp long) and 2G12 (31 bp long) suggest that making antibodies with specific CDR H3 loops that are effective in neutralizing HIV-1 is challenging for the immune system (25, 57). For 2F5, clues as to the requirement for such an unusual CDR H3 might be found in the present study, whose results clearly show that the high affinity of 2F5 for its primary epitope is not sufficient for neutralization but that one or more weaker, secondary binding determinants might be required for this crucial activity, a kind of “intramolecular avidity effect.” The present finding also helps to explain why past immunization protocols based on the presentation of the constrained core epitope of 2F5 have resulted in eliciting antibodies of high affinity but with strongly limited neutralization capacity. Our results support the hypothesis that immunogens seeking to elicit 2F5-like broadly neutralizing antibodies will require components in addition to the core epitope. The challenge of eliciting by immunization antibodies that are simultaneously highly specific, broad in recognition, slightly polyreactive, and able to access recessed epitopes is significant and exemplified by the obvious difficulty of generating such a response during natural HIV-1 infection. Structural studies looking at the interactions of broadly neutralizing antibodies with more complex epitopes in more complete and biologically relevant environments might be able to provide further clues about how to generate such a challenging immunogen.
We thank S. Bryson, M. Wierzbicka, and D. E. Isenman for help with experimental design and sample preparation and for critical discussions. We gratefully acknowledge the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH, for samples of CHO-Env (contributed by C. Weiss and J. White) and TZM-bl cells (contributed by J. Kappes), as well as for clone pHXB2-env (contributed by K. Page and D. Littman). We thank Patricia Villace, CSIC, Madrid, for the gift of an env-deficient HIV-1 genome.
This work was supported by the Canada Research Chairs Program (E.F.P.) and a predoctoral fellowship from the Canadian Institutes for Health Research (J.-P.J.), as well as by the Spanish MICINN (grant BIO2008-00772) and the University of the Basque Country (grants GIU 06/42 and DIPE08/12) (J.L.N.). R.M. was the recipient of a predoctoral fellowship of the University of the Basque Country. This research was funded in part by the Ontario Ministry of Health and Long Term Care.
The views expressed do not necessarily reflect those of the OMOHLTC.
Published ahead of print on 10 February 2010.