Antibodies to a Superantigenic gp120 Epitope as the Basis for Developing an HIV Vaccine

Abstract

Failure to induce synthesis of neutralizing antibodies to the CD4 binding determinant (CD4BD) of gp120, a central objective in HIV vaccine research, has been alternately ascribed to insufficient immunogen binding to antibodies in their germline variable (V) region configuration expressed as B cell receptors, insufficient adaptive mutations in antibody variable (V) regions and conformational instability of gp120. We employed peptide analogs of gp120 residues 421-433 (CD4BD^core) to identify antibodies produced without prior exposure to HIV (constitutive antibodies). The CD4BD^core peptide was recognized by single chain Fv (scFv) fragments from non-infected humans with lupus that neutralized genetically diverse strains belonging to various HIV subtypes. Replacing the framework (FR) segments of a V_H4-family scFv with the corresponding V_H3-family FRs from scFv JL427 improved the CD4BD^core peptide binding activity, suggesting a CD4BD^core binding site outside the pocket formed by the complementarity determining regions. Replacement mutations in the FR site vicinity suggested the potential for adaptive improvement. A very small subset of serum CD4BD^core-specific serum IgAs from non-infected humans without autoimmune disease isolated by epitope-specific chromatography neutralized the virus potently. A CD4BD^core-specific, HIV neutralizing murine IgM with heavy and light chain V regions (V_H and V_L regions) free of immunogen-driven somatic mutations was induced by immunization with a CD4BD^core peptide analog containing an electrophilic group that binds B cells covalently. The studies indicate broad and potent HIV neutralization by constitutive antibodies as an innate, germline-encoded activity directed to the superantigenic CD4BD^core epitope that is available for amplification for vaccination against HIV.

INTRODUCTION

The human immunodeficiency virus-1 (HIV) surface expresses noncovalently-associated oligomeric gp120-gp41 complexes that have been the targets of numerous experimental vaccines. Consensus has developed that inducing antibodies to one or more structurally conserved epitopes is required for protection against genetically diverse HIV strains responsible for the pandemic (broadly neutralizing antibodies). Most antibodies to gp120 bind its mutable epitopes. Conserved gp120 epitopes essential for the viral life cycle are poorly immunogenic (1, 2). Binding of the mostly conserved gp120 conformational determinant to the host CD4 receptors (CD4BD) initiates infection. Broadly neutralizing antibodies to the CD4BD have been identified in HIV-infected patients (2–4). Although the CD4BD is vulnerable to host immunity, antibodies with the correct specificity are not produced in sufficient amounts for effective protection against the infection. Various CD4BD epitopes are not equally susceptible to antibody neutralization. Some antibodies bind the CD4BD of monomer gp120 but fail to neutralize HIV (5). Deviations from the native CD4BD conformation were suggested to underlie the failure of monomer gp120 to induce the synthesis of broadly neutralizing antibodies (1). Genetically engineered oligomeric gp120 also failed to induce broadly neutralizing antibodies (6). The monoclonal antibody b12 to a conformational epitope that overlaps the CD4BD outer domain region expresses comparatively broad neutralizing activity (2, 7). An immunogen ‘reverse engineered’ to be conformationally complementary to the b12 binding site did not induce neutralizing antibodies (8).

These difficulties have inspired divergent hypotheses suggesting that the failure of the CD4BD to induce neutralizing antibodies may reflect an intrinsic weakness of the humoral immune system. Inducing antibody synthesis involves initial weak immunogen binding to B cell receptors with V regions encoded by germline genes (BCRs; antibodies associated with signal transducing proteins), followed by immunogen-driven clonal expansion of B cells and selection of BCRs containing mutated complementarity determining regions (CDRs) with improved antigen binding affinity. The BCR repertoire expressed constitutively prior to contact with immunogen is diverse and large, composed of paired light chain and heavy chain variable domains (V_L and V_H domains) generated from about 500 V, D and J germline genes by V-(D)-J gene recombination, a step that introduces sequence diversity in CDR3 (9). The diversity of constitutive BCRs is innate in the sense that it is generated randomly with no requirement for contact with immunogen. Dimitrov and coworkers suggested that an unusually low binding affinity of conserved HIV epitopes for constitutive BCRs precludes B cell recruitment into the adaptive differentiation pathway (Figure 1A; ref 10). Conversely, monoclonal neutralizing antibodies from HIV infected patients that bind discontinuous gp120 outer domain epitopes overlapping the CD4BD contain extremely dense V region somatic mutations (11, 12), prompting the hypothesis of an intrinsic incompetence of B cells in generating sufficiently mutated antibodies.

Figure 1. Alternative reasons underlying failed induction of antibodies to the CD4BD and constitutive antibodies as the basis for HIV vaccination.

A. X, Detrimental B cell event; √, Our proposed basis for HIV vaccination. Middle panel: B cell developmental events necessary for neutralizing antibody synthesis. Right panel: The CD4BD^core is a superantigenic epitope that binds the FRs of constitutive BCRs noncovalently, but this causes B cell down-regulation. Left panel: Alternatively, B cells are unable to accumulate sufficient V gene mutations needed for CD4BD binding (3), or the initial affinity of germline BCRs for the CD4BD is too low (10). NAbs, neutralizing Abs. B. Constitutive CD4BD^core-reactive antibodies pertinent to HIV vaccination. Top panel: The innate germline antibody repertoire contains a wide range of antibodies with varying CD4BD^core-reactivity and HIV neutralizing potency. A small antibody subset with the greatest CD4BD^core-reactivity is hypothesized to possess potent and broad neutralizing activity (black). Middle panel: An electrophilic vaccine expressing a CD4BS^core that mimics the viral CD4BD^core conformation accurately will selectively amplify the antibody subset with greatest CD4BD^core–reactivity and neutralizing activity effective against genetically divergent HIV strains. Bottom panel: A vaccine that mimics the viral CD4BD^core imperfectly will amplify an antibody subset with lesser CD4BD^core-reactivity and neutralizing activity (grey).

Our approach to HIV vaccine design focuses on the exposed, mostly-conserved 421-433 residues located in the β20/β21 units of the gp120 bridging sheet that provide essential contributions in high affinity CD4 recognition (CD4BD^core) as determined by mutagenesis and crystallography (13–16). Importantly, the CD4BD^core appears to contribute only minimally to the CD4BD-overlapping outer domain epitopes recognized by neutralizing antibodies from infected patients (3, 12). Several groups have reported that antibodies produced spontaneously by humans without HIV infection or vaccination (constitutive antibodies) recognize the CD4BD^core-spanning superantigenic determinant of gp120 and synthetic peptides containing the CD4BD^core (17–20). Superantigen epitopes are recognized by antibodies containing V regions in the germline configuration with no requirement for an immunogen-driven antibody response (21, 22). The binding occurs mostly at germline gene-encoded structural elements in the framework regions (FRs) and certain CDR residues outside the traditional antigen-binding pocket formed by the CDRs (22–24). Upon completion of the noncovalent CD4BD^core binding step, a subset of the constitutive antibodies proceeds to catalyze the hydrolysis of gp120 by a nucleophilic mechanism (19, 20). However, superantigen recognition by constitutive antibodies comes at a cost. Like other microbial superantigen epitopes (21), the CD4BD^core is poorly immunogenic. Infected humans and animals immunized with gp120 rarely generate neutralizing antibodies to the CD4BD^core by adaptive immune mechanisms (25, 26). Unlike the stimulatory effect of antigen binding at the CDRs, noncovalent superantigen-FR binding down-regulates further B cell differentiation (Figure 1A) (21). We and others have suggested that epitopes essential for microbial survival have evolved superantigenic character as an immune evasion mechanism that precludes an efficient adaptive antibody response (27, 28).

Capitalizing on CD4BD^core superantigenicity for HIV vaccination may be feasible if the constitutive antibodies neutralize the virus. Braun’s group reported highly variable binding of the gp120 superantigen determinant by serum IgGs in a large group of non-infected humans, with low binding activity serving as a predictor for increased incidence of subsequent HIV infection (29). However, the gp120 binding antibodies from the sera and a phage library from non-infected humans did not neutralize HIV detectably under commonly-employed tissue culture conditions (20, 22, 30), raising doubt about the functional relevance of the constitutive antibodies. We hypothesize the existence of a broad range of constitutive antibodies with varying CD4BD^core reactivity of which a small subset is capable of potent and broad HIV neutralization (Figure 1B, top panel).

The physiological restriction on adaptive antibody induction must be overcome to develop a CD4BD^core-based vaccine. There is no precedent for a clinically useful antibody response that targets a B cell superantigen epitope. We reported electrophilic analogs of CD4BD^core-spanning peptides that bind the constitutive antibodies noncovalently, coordinated with covalent electrophile binding to antibody nucleophilic sites (19, 20). Like the noncovalent superantigen binding sites, nucleophilic sites are expressed constitutively by secreted antibodies and membrane-bound BCRs (31). We also reported monoclonal antibodies to the CD4BD^core with HIV neutralizing activity raised by immunization with an electrophilic analog of full-length gp120. This suggested covalent stimulation of B cells as a strategy to induce the synthesis of antibodies to the CD4BD^core. Classical vaccines depend on the somatic hypermutation process to develop neutralizing antibodies. If a constitutive antibody subset with neutralizing activity directed against genetically divergent HIV strains exists, amplifying this subset could protect against HIV infection even without an improvement of the antibodies via somatic hypermutation (Figure 1B, middle panel). The extent to which the immunogen targets the minority of B cells producing strongly neutralizing constitutive antibodies is an important factor. An immunogen that amplifies only moderately neutralizing constitutive will be less effective vaccine (Figure 1B, bottom panel).

We describe evidence for epitope-specific constitutive antibodies from non-infected humans with potent ability to neutralize diverse HIV strains identified using the electrophilic CD4BD^core-spanning peptide probes. Structural analysis and mutagenesis studies of the constitutive human antibodies supported a contribution of the FRs in CD4BD^core recognition and the potential of adaptive antibody improvement. A monoclonal murine IgM isolated following immunization with an electrophilic CD4BD^core peptide recognized the CD4BD^core specifically and displayed broad HIV neutralization despite being free of V gene somatic mutations. The findings validate constitutive CD4BD^core recognition by antibodies as a novel principle supporting HIV vaccine development.

MATERIALS AND METHODS

Polypeptide probes

Proteins were obtained from these sources: ovalbumin, bovine serum albumin (BSA), thyroglobulin and calmodulin from Sigma (St. Louis, MO), soluble CD4 (4 domain) from Protein Science (Meriden, CT), recombinant gp120 (MN strain) from Immunodiagnostics (Woburn, MA), and biotinylated BSA from Pierce (Rockford, IL). Biotinylated gp120 (1.8 mol biotin/mol), E-gp120 (24.8 mol phosphonate/mol) and biotinylated E-gp120 (1.2 mol biotin/mol; 33 mol phosphonate/mol) were prepared by linking phosphonate diesters and biotinamidohexanoyl (Bt) to Lys side chains as described (31, 32). Non-electrophilic peptide 421-436—conjugated to BSA (NE-421-436; 6 mol peptide/mol), electrophilic E-421-433 with biotin at the N terminus, E-416-433 with biotin at the N terminus (E-416–433a), and E-416-433 conjugated to keyhole limpet hemocyanin (KLH) (E-416-433b; 2036 mol peptide/mol) or to BSA (E-416-433c; 6.1 mol peptide/mol) contained the consensus subtype B gp120 421-433 residues and were prepared as in (4, 31, 33). Cys418 in the 416-433 peptides was replaced with Ser to preclude intermolecular S-S bonding. Phosphonate diester groups were located at the C terminus in E-421-433 and at Lys421 and Lys433 side chains in E-416-433a, E-416-433b and E-416-433c. Control biotinylated E-VIP was prepared as in (34). The chemical identity of all peptides was verified by electrospray ionization mass spectrometry. Ala-containing mutant 416-433 peptides, non-electrophilic NE-416-433 peptide and sequence-shuffled 416-433 peptide (Sh416-433, GQKSWEIPAKNRLIMVIQ) were synthesized in the PEPscreen platform and their structural identity was verified by matrix assisted laser desorption time-of-flight mass spectrometry by Sigma-Aldrich (4).

Antibodies

The use of human subjects and vertebrate animals were approved by our Institutional Committee for Protection of Human Subjects and the Animal Welfare Committee. The animal protocol adheres to AAALAC International guidelines. Written informed consent was provided by humans who were studiied. The healthy humans had no evidence of infectious or other disease (age range 28–48 years; 5 males, 5 females). Antibodies were purified from sera or tissue culture fluids to electrophoretic homogeneity using immobilized Protein G (IgG), anti-IgA antibody or anti-IgM antibody (35, 36). All antibody subunit bands stainable with Coomassie Blue were stainable with antibodies to heavy (γ, α, μ) or light chain (κ, λ) subunits by SDS-electrophoresis. The IgA fraction from serum (2 mg) was purified further by affinity chromatography on E-416-433 conjugated to agarose (4).

scFv clones were selected from a phagemid library from 3 systemic lupus erythematosus patients free of HIV infection by binding to immobilized gp120 (36) followed by 2 further selection rounds in which 10¹² colony forming units of the phages were treated for 2 hours with Bt-E-gp120 (6.3 μg) in PBS (0.13 ml 10 mM sodium phosphate, 137 mM NaCl, 2.7mM KCl, pH 7.4), immune complexed phages were captured on agarose-conjugated anti-biotin antibody, and bound phages were eluted with acid pH buffer (37). gp120 (36) and E-gp120 (38) express the CD4BD^core as a component of the larger CD4BD in a conformation reactive with neutralizing antibodies. A subset of nucleophilic antibodies identified based on covalent binding to electrophilic E-gp120 express catalytic activity (32). The scFv clones contained a His6 tag and a c-myc tag. Secreted scFv expressed by E. coli HB2151 cells was purified by metal-affinity chromatography and quantified by c-myc dot blotting (37). Mutated scFv GL2_FR1m, GL2_FR3m and GL2_CDR2m were prepared by synthesizing the V_H GL2 domain genes in which the nucleotide sequences corresponding to amino acids 1–30, 66–94 or 50–65 (Kabat numbering) were replaced by the corresponding V_H JL427 sequence, followed by cloning of the mutant V_H domain gene in place of the wildtype V_H domain of scFv GL2 in pHEN2 vector via the ApaLI/NotI sites (Mutagenex, Hillsborough, NJ). scFv GL2_CDR1m was prepared by replicating the wildtype scFv GL2 vector with the mutated primers encoding JL427 V_H-CDR1 residues 31–35 (5′-GCATGCACTGGATCCGCCAGCA CCC, 5′-CATAGCTGCTGATGGAGCCACCAGA) (38). Monoclonal IgG1 b12 used as a reference reagent was from the Scripps Research Institute (courtesy Dr. D. Burton), and monoclonal IgG 4B6 directed to phosphatidylserine, from Abcam (Cambridge, MA).

Full-length IgG1 was prepared by inserting clone JL427 V_L and V_H domains adjacent to the human λ constant domain (BglII/NotI sites; pML-JL427-HL vector) or γ1 constant domains (EcorI/HindIII sites; pML-JL427-HG1), respectively (39). The light and heavy chain vectors were coexpressed in NS0 cells and IgG in the culture supernatant was purified as before. Monoclonal antibodies were prepared from female BALB/c mice (Jackson; 4–5 wks) immunized by nasal instillation of E-416-433b (100 μg peptide equivalents/20 μl in 5 μg heat-labile E Coli enterotoxin Arg192:Gly mutant adjuvant; kindly provided by Dr. John Clements, Tulane University) at weekly intervals for 6 wks. Hybridomas were prepared by fusing splenocytes and peritoneal cells 3 days after the final injection with NS-1 myeloma cells (32). Monoclonal cell lines were obtained by limiting dilution from wells secreting E-416-433c binding IgMs. IgM was purified from concentrated tissue culture fluid (Centriprep, YM10 filter; Millipore, Billerica, MA) by chromatography on immobilized anti-IgM antibody and a Superose 6 size exclusion column (19). Antibodies were dialyzed against PBS prior to neutralization assays or PBS containing 0.1 mM 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate prior to binding assays. Total protein was estimated using the microbicinchoninic acid assay kit (Pierce, Rockford, IL). LPS (endotoxin) was estimated by the Limulus amebocyte lysate test using the Endosafe®-PTS instrument and FDA-approved cartridges (Charles River, Wilmington, MA).

Binding and neutralization

Biotinylated peptides (10 μM) were incubated with antibodies in PBS containing 0.1 mM 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate followed by high sensitivity detection of covalent adducts by denaturing SDS-electrophoresis of boiled reaction mixtures and staining with peroxidase-conjugated streptavidin (31). ELISAs were performed using immobilized gp120 (strain MN), E-gp120, E-416-433b, E-416-433c or NE-421-436 (70–230 ng/well) treated with scFv or IgM as in (33, 38). Initial screening for gp120 binding was done using bacterial periplasmic extracts with an equivalently diluted extract from bacteria harboring empty pHEN2 vector as control (36). Mutant and wildtype scFv binding was compared by computing the ratio of scFv concentrations yielding 0.25 A490 determined from a standard curve fitted to the equation: $\text{observed}\text{A}490=\text{A}{490}_{max}\times [\text{scFv}]/(\text{Kd}+[\text{scFv}])$. Standard deviation for ELISA replicates were <10%. Non-specific binding on plates coated with KLH was <0.2 A₄₉₀ at the highest scFv concentration tested. Hydrolysis of biotinylated gp120 (strain MN) was determined by reducing SDS-electrophoresis and staining with streptavidin-peroxidase (32). Antibody neutralizing activity was measured using primary HIV-1 virus strains and PHA-activated peripheral blood mononuclear cells (PBMCs; pooled from 4–12 donors) as host cells (36). Infection was monitored by measuring p24. Unless indicated otherwise, neutralization assays were done by the commonly-used protocol in which HIV was preincubated with test antibody for 1 hour in medium containing 10% FBS (Life Technologies, Grand Island, NY), followed by addition of the mixture into culture wells containing PBMCs (designated “low-sensitivity” assay). To enhance detection of weakly neutralizing antibodies, some assays were done by preincubating the virus with antibodies for a prolonged duration (24 hours) in medium containing 0.66% FBS (designated “high sensitivity” assay) (20). The length of PBMC culture with the antibody-HIV mixture (4 days) and the FBS concentration (10%) in the culture were identical in both types of assays. In some assays, the antibody solution (95 μL) was pretreated with E-421-433 (5 μL; final concentration 100 μM) or an equivalent concentration of control E-VIP in a 1:1 PBS:RPMI diluent containing 1% dimethylsulfoxide. To study interference by β-chemokines, the neutralization assays were conducted in presence of goat polyclonal IgG antibodies to MIP-1α, MIP-1β and RANTES or an equivalent concentration of control nonimmune goat IgG (R&D Systems Inc, Minneapolis, MN) as in ref 40. We also measured inhibition of infection by the subtype C virus strain 97ZA009 by a purified lipopolysaccharide (LPS) preparation containing defined chemical constituents (O111:B4, Invivogen, San Diego, CA). Most HIV neutralization assays were done in Dr. Carl Hanson’s laboratory. Some confirmatory assays for scFv JL427 and IgM 2G9 neutralizing activity and control assays for study of host cell artifacts were run in Dr. Sudhir Paul’s lab. Neutralization of irrelevant viruses (West Nile, Dengue II, St. Louis encephalitis) was measured from the number of plaques formed in Vero African Green Monkey kidney cells infected in the absence or presence of antibodies (41).

Antibody structures

The V domains were sequenced by the dideoxynucleotide method using the primers in (38). Germline gene assignment, measurement of replacement/silent mutation ratios (R/S ratios) and analysis of V-(D)-J junctions was as in (38). The “Karray” residues deduced from antibody homology studies are described in (23). Models of the wildtype and mutant scFv clones were constructed based on canonical structure definitions using WAM at http://antibody.bath.ac.uk/(37). Surface visualization, model superimposition, and computation of root mean square (r.m.s.) deviation and translational Cα-Cα movements were with Accelrys DS Visualizer v2.0.1.7347.

RESULTS

Peptide and protein probes

Detecting and inducing vaccine-relevant antibodies depends on availability of a probe with sufficient conformational similarity to the native HIV CD4BD^core (Supplemental Figure S1A, B). We and others have reported assays for reversible gp120 binding and catalytic gp120 hydrolysis for detection of constitutive antibodies from non-infected humans directed to the CD4BD^core (19, 20, 23, 24). Use of E-gp120 (Figure 2A, 2B), a probe that forms covalently assembled oligomers, enabled improved detection of the antibodies (38). The peptide probes containing the 421-433 region are flexible and can assume alternate conformations with varying affinity for soluble CD4 (42–44). Electrophilic E-421-433 and E-416-433 were also validated previously as specific probes for constitutive antibodies with CD4BD^core specificity (4, 19, 20). The E-416-433 peptide containing the N terminal 416-420 extension binds the antibodies at somewhat superior levels compared to E-421-433 (38), and both peptides bind soluble CD4 with affinity lower than gp120 (4). The probes contain a peptide epitope that binds antibodies noncovalently coordinated with covalent binding of the electrophilic phosphonate to naturally-occurring nucleophilic antibody sites (Figure 2C). A subset of nucleophilic antibodies can catalyze the cleavage of peptide bonds. However, the nucleophilic sites are also expressed by non-catalytic antibodies deficient in active site structures needed to complete the catalytic reaction cycle (31). Consequently, the electrophilic peptides are combined probes for the non-catalytic and catalytic antibody subsets. The CD4BD^core sequence is largely conserved in diverse HIV strains (see Supplemental Table SI for example sequence divergences). The 421-433 CD4BD^core sequence in E-416-433 and E-421-433 corresponds to the consensus HIV-1 subtype B sequence. The 421-433 sequence in E-gp120 also corresponds to the subtype B consensus sequence except for a divergence at residue 429.

Figure 2. Probes for CD4BD^core binding antibodies.

A. Electrophilic peptide and gp120 analogs. The probes contain residues 421-433 constituting the CD4BD^core. The N-terminal LPSRI residues in E-416-433 stabilize the 421-433 epitope conformation (42). The electrophilic phosphonate is located at Lys421 and Lys432 side chains in E-416-433 and at the C terminus in E-421-433. Biotin placed at the N terminus of some probes enabled detection of immune complexes. BSA and KLH conjugates of the electrophilic peptides were used, respectively, for ELISA and immunization. NE-416–433 and NE-421-436 are the non-electrophilic peptide probes and Sh416–433 is the shuffled-sequence, control peptide. E-VIP is an irrelevant control peptide. E-gp120 contains electrophilic phosphonates groups at surface-accessible Lys side chains. B. Substituent structure. R1 and R2 denote, respectively, the electrophilic phosphonate and positively charged amidino-linker substituents. C. Coordinated noncovalent and covalent probe binding by antibodies. Initial noncovalent binding of amino acids constituting the CD4BD^core epitope (black ovals) to the antibody paratope imparts specificity to the subsequent covalent binding of the electrophilic phosphonate (E) by a nucleophilic (Nu) antibody site (31).

Recombinant CD4BD^core-specific antibodies

The paired V_L-V_H domains of single chain Fv (scFv) constructs are surrogates of antibody binding sites. We initially became interested in the potential protective role of antibodies to the CD4BD^core from evidence suggesting that these antibodies are increased in patients with the autoimmune disease lupus (45) and comments in the literature about the rare coexistence of lupus and HIV infection (46). A CD4BD^core-reactive scFv clone obtained by fractionating a phage library from non-infected humans (lupus patients) was described previously (36). Here, we describe the structure-function properties of 2 scFv clones with broad and potent HIV neutralizing activity from the same library (scFv JL427 and scFv GL2).

We fractionated the library by binding to immobilized gp120. Of 12 scFv clones obtained by gp120-fractionation, 5 displayed detectable binding to immobilized gp120 by ELISA. Four of the 5 gp120-binding clones neutralized HIV subtype C strain ZA009 in an initial screening assay consisting of 3-point concentration-dependence curves (example data in Figure 3A; Supplemental Figure S2 reports binding and neutralization data for individual clones). An identically purified extract from bacteria harboring the empty pHEN2 vector was devoid of neutralizing activity. scFv JL427 and scFv JL606, the 2 clones with greatest strain 97ZA009-neutralizing potency, also neutralized 2 additional HIV strains (subtype C strain BR004 and subtype B SF162; Supplemental Figure S2). As scFv JL427 displayed superior neutralizing potency across the 3 strains tested in the initial studies, this clone was picked for further analysis (mean IC50 for scFv JL427 and scFv JL606, respectively, 0.053 and 0.649 μg/ml). There was no correlation between the magnitude of gp120 binding and neutralizing potency of the 4 clones (e.g., scFv JL427 and scFv JL606 displayed nearly equivalent gp120 binding but differing average neutralizing potency). However, none of scFv clones devoid of gp120 binding activity neutralized HIV detectably (n=7; 2.5 μg scFv/ml; P<0.01 versus scFv group with binding activity, n=5, Fisher’s exact test).

Figure 3. HIV neutralization by scFv from non-infected humans.

Values are means ± s.e.m. of 4 culture replicates. A. Subtype C strain 97ZA009 neutralization data for a neutralizing and a non-neutralizing scFv clone purified by one metal-affinity chromatography cycle at the initial screening step. Inset, Schematic scFv structure. B. Slow hydrolysis of gp120 by scFv GL2. Reducing SDS-electrophoresis gel lanes showing streptavidin-peroxidase stained biotinylated gp120 treated with diluent (lane 1), an extract of E. coli harboring empty pHEN2 vector without the scFv gene (lane 2), noncatalytic scFv GL104 (lane 3), or the catalytic scFv GL2 (lane 4). Lane 5 is an overexposed gp120-scFv GL2 mixture showing product bands (nominal product mass in kDa is indicated). Lane 6 shows gp120 treated with the noncatalytic scFv JL427. scFv preparations were purified by one-cycle of metal affinity chromatography. Reaction conditions: gp120, 0.1 μM; scFv, 0.36 μM, 40h. C. Example scFv JL427 and IgG JL427 neutralization curves (subtype AE strain 90CM235 and subtype C strain 97ZA009). The scFv was purified by two metal-affinity chromatography cycles. The scFv and IgG were purified, respectively, by two cycles of metal-affinity and Protein G chromatography. Means ± s.e.m. Inset, Reducing SDS-electrophoresis of scFv JL427 stained with silver (lane 1) or anti-c-myc antibody (lane 2) and IgG JL427 stained with Coomassie Blue (lane 3) or an anti-γ/λ antibody mixture (lane 4; the 50 kDa and 25 kDa bands correspond, respectively, to the heavy and light chains).

We reported the ability of E-gp120 to induce the synthesis of antibodies that hydrolyze gp120 (32). To identify scFv clones with catalytic activity, the gp120-fractionated library was subjected to further fractionation using E-gp120. Of 21 scFv clones screened, one scFv displayed slow gp120 hydrolytic activity evident after prolonged incubation (Figure 3B; scFv GL2; 2.3% biotinylated gp120 hydrolyzed/hour at 0.36 μM scFv, gp120 substrate concentration 100 nM; measured as in reference 32). This clone was also positive for gp120 binding by ELISA (A490 0.25). scFv GL2 neutralized HIV strain 97ZA009 potently, and it was picked for further analysis (Supplemental Figure S2).

Example concentration-dependence data verifying the scFv neutralizing activity in independent scFv preparations are shown in Figure 3C). To verify HIV neutralization within the physiological antibody scaffold, we also prepared full-length IgG JL427 containing the scFv V_H and V_L domains. The purified IgG displayed concentration-dependent neutralizing activity (Figure 3C). A few neutralization assays included the reference monoclonal IgG b12 directed to the CD4BD epitope located in the gp120 outer domain (2). scFv JL427 displayed superior potency compared to IgG b12 (IC50 values in μg/ml, respectively: strain 98TZ013, 0.017 and 29.2; 92UG082, <0.004 and 4.8; strain SF162, 0.011 and 0.721). The viability of PBMCs treated for 4 days with scFv JL427 (60 μg/ml) or PBS was comparable, ruling out nonspecific scFv cytotoxicity (respectively, 84% and 74%). scFv JL427 did not neutralize irrelevant viruses (20 μg scFv/ml; Dengue II virus, St-Louis encephalitis virus and West Nile virus), suggesting specific HIV neutralizing activity.

Both scFv JL427 and scFv GL2 displayed E-416-433b binding activity, with scFv JL427 displayed 12.0-fold superior binding (Figure 4A). E-gp120 binding by scFv JL427 was also superior to the binding by scFv GL2 (by 2.5-fold, Figure 4B). The differing E-416-433b/E-gp120 binding ratios for the two scFv clones suggest a non-identical fine specificity of CD4BD^core recognition (scFv JL427, 7.3; scFv GL2, 1.5; determined from scFv concentrations yielding A490 values of 0.2). scFv JL427 was bound by E-421-433 but not the irrelevant probe E-vasoactive intestinal peptide (E-VIP) (Figure 4C, inset). Its HIV neutralizing activity was inhibited by E-421-433 but not E-VIP (Figure 4C). As further evidence of specificity, the binding of scFv JL427 to immobilized peptide NE-421-436 devoid of the phosphonate groups was inhibited competitively by soluble peptide NE-421-436 but not irrelevant proteins (Figure 4D). The scFv specificity behavior differs from the previously-described polyreactive binding of structurally unrelated antigens by non-neutralizing gp120-binding recombinant Fabs isolated from a non-infected human (30). Also, unlike the scFv, the previously-described Fabs did not display any clear specificity for the CD4BD^core within the gp120 sequence.

Figure 4. CD4BD^core peptide analog binding by scFv JL427 and scFv GL2.

A. E-416-433b binding. Binding of the purified scFv to immobilized KLH-conjugated E-416-433b (70 ng/well). Values are corrected for nonspecific binding to KLH. An identically-purified extract of E. coli harboring empty pHEN2 vector without the scFv gene did not display detectable binding. Values are mean ± s.d. B. E-gp120 binding. Binding of purified scFv to immobilized E-gp120 (100 ng/well). Values are corrected for nonspecific binding. An identically-purified extract of E. coli harboring empty pHEN2 vector without the scFv gene did not display binding. Values are mean ± s.d. C. Inhibition of scFv JL427 neutralizing activity by E-421-433. Neutralization of HIV subtype C strain 97ZA009 was determined in the presence of E-421-433 or control E-VIP (final concentration, 100 μM). Values are means ± s.e.m. of 4 culture replicates. Inset, Streptavidin-stained, reducing SDS-gel blots showing adducts of the scFv (30 μg/mL) with E-421-433 (lane 2, 10 μM) and E-VIP (lane 3, 10 μM) after incubation for 4 h. Lane 1, anti-c-myc staining of c-myc tag in the scFv. D. scFv JL427 binding specificity. Binding of the scFv (46 μg/ml, bacterial periplasmic extract) to immobilized peptide NE-421-436 was measured in the absence or presence of soluble NE-421-436 or the irrelevant proteins bovine serum albumin, thyroglobulin and calmodulin (1 μM). Binding (A490) in absence of soluble competitor was 1.8±0.1. Values are corrected for nonspecific binding observed for an extract of E. coli with empty pHEN2 vector devoid of the scFv gene.

The breadth of neutralization was tested using a panel of 17 HIV-1 strains. Both scFv clones neutralized all subtype A, B, C, D and AE strains tested. scFv JL427 displayed superior mean neutralization potency across the panel of strains (Table I; IC50 geometric mean and range, respectively – scFv JL427, 0.014 μg/mL and 0.001–0.169 μg/mL; scFv GL2, 0.337 μg/mL and 0.002–2.218 μg/mL; strains neutralized at >50% levels at the lowest tested antibody concentration tested in Table I were not included in geometric mean computation). Taken together, the data indicate potent and broad HIV neutralization due to specific CD4BD^core binding by the scFv clones.

Table I. Cross-subtype HIV neutralizing activity of scFv clones.

IC50 and IC80 values were interpolated from concentration-dependence curves using primary, coreceptor CCR5-dependent HIV isolates and peripheral blood mononuclear cells as hosts.

HIV subtype	Strain	scFv JL427		scFv GL2
		IC50, μg/mL	IC80, μg/mL	IC50, μg/mL	IC80, μg/mL
C	98TZ013	0.017	0.400	2.218	4.940
	98BR004	0.007	0.126	1.680	4.433
	97ZA009	0.169	0.715	0.020	0.639
	98TZ017	0.040*	0.040*	0.040*	0.079
	98IN022	0.001	1.012	0.002*	0.093
	97ZA012	0.009	8.319	NT	NT
D	92UG035	0.004*	0.004*	0.079	0.254
	93UG082	0.004*	0.004*	0.264	0.570
	94UG114	0.004*	0.004*	0.102	0.316
B	92BR021	0.004*	0.014	0.114	0.531
	SF162	0.011	0.195	0.036*	0.036*
	92BR020	0.004*	0.018	0.619	1.761
AE	CM235	0.001	0.360	0.633	3.097
	92TH005	0.022	1.581	0.797	4.184
A	97USSN54	0.048	1.190	0.449	3.459
	92RW008	0.159	6.651	0.844	2.777
	92RW024	0.004*	0.004*	0.344	0.446

^*Neutralizing activity at the indicated lowest scFv concentration tested was >50%.

NT, not tested. scFv concentrations tested ranged from a lower limit of 0.0004 —15 μg scFv/ml. Assay accuracy is greatest at 50% neutralization. For all strains neutralization exceeded 80% at the highest scFv concentration tested. r² values for fitted curves with an IC50 that could be determined by interpolation were 0.43–0.99.

scFv GL2 hydrolyzed recombinant gp120 slowly (see above). Its lesser average neutralizing potency compared to the non-hydrolytic scFv JL427 suggests that hydrolysis of gp120 does not contribute appreciably in virus neutralization. From the rate of recombinant gp120 hydrolysis, only 0.08% and 9.6% of the viral gp120 will be hydrolyzed, respectively, at the mean 50% neutralizing concentration of scFv GL2 (0.337 μg/mL) during its preincubation with HIV (1 hour) and the subsequent culture with PBMCs (4 days).

V domain structure

From sequence conservation studies, Karray et al (23) suggested 16 V_H region amino acids likely to underlie preferential binding of the gp120 superantigen site by V_H3-family antibodies (designated Karray residues; 5 located in FR1, 6 in FR3, 1 in CDR1 and 4 in CDR2 according to the Kabat database; nota bene – 3 of the 4 CDR2 residues have been designated FR3 residues in the IMGT database based on their wide-spread conservation in antibodies; ref 47). scFv JL427 derives from a V_H3 germline gene and contains all 16 Karray residues (Table II). Consistent with the constitutive origin of CD4BD^core recognition, 15 of the 16 Karray residues are encoded by its germline V_H gene. scFv GL2 displayed lower CD4BD^core peptide binding compared to scFv JL427. This scFv is derived from a minimally mutated V_H4 germline gene containing identities or conservative substitutions at 7 Karray positions (Table II). Substantial sequence conservation is evident between the human V_H1–V_H7 germline genes and between the V_H and V_L germline genes (47). Examples of CD4BD^core recognition by non-V_H3 family antibodies and rare V_L domains from non-infected humans have been reported previously (19, 48), and we do not exclude a role for the V_L domain in CD4BD^core recognition, particularly for non-V_H3 clones such as scFv GL2.

Table II. Antibody sequence properties.

Family assignments according to Kabat/Wu database (69). Germline assignments according to IgBLAST(blast.ncbi.nlm.nih.gov). V-(D)-J junctional deletions and insertions deduced using IMGT/V-QUEST (70). R/S ratio computation excludes mutations attributable to V-(D)-J junctional diversification and PCR primer-annealing region. The Kabat/Wu and IMGT/V-QUEST databases differ in demarcation of CDRs and FRs. R/S ratios computed from both databases are reported. R/S ratios predicted for a random mutational process are: V_L-CDRs, 3.56/1; V_H-CDRs, 4.16/1; V_L-FRs, 2.88/1; V_H-FRs, 2.84/1 (71). Genbank accession numbers (http://www.ncbi.nlm.nih.gov/genbank/) for scFv JL427, scFv GL2, IgM 2G9 V_H and IgM 2G9 V_L are, respectively: AF329462, JQ343847, JQ412130 and JQ343846. Replacement mutations in scFv V regions were (germline residues listed first, Kabat numbering): scFv JL427 V_H – G16:R, S33:G, N35:H, S52:G, S52a:R, S54:G, T56:H, I57:T, Y58:N, A74:S, S77:T, M82:I; scFv JL427 V_L, A11:V, T14:A, I28:F, S30:L, Q38:H, L39:F, S50:R, N52:D, D60:A; scFv GL2 V_H GL2: G10:R, E16:G, G32:S, S35b:G; scFv GL2 VL, none. IgM V_L and V_H regions contain no somatic mutations. scFv JL427 V_H contains all 16 Karray residues, 15 encoded by the germline gene and 1 acquired by somatic mutation (G10, Q13, R19, A23, T28, Y32, G54, K64, K65, Y79, Q81, N82a, R83, E85). scFv GL2 contains 3 identities (Y59, K64, K75), 4 conservative deviations (T28:S, Y32:S, Y79:S, N82a:S) and 9 non-conservative deviations at the Karray positions. Six of the 7 identities/conservations are germline gene-encoded. Conservation of Karray residues was computed as in (23). IgM 2G9 contains 7 identities (G10, Q13, Y32, G54, Y59, K75, N82a), 1 conservative deviations (T28:S) and 8 nonconservative deviations at the Karray positions.

Property	scFv JL427		scFv GL2		IgM 2G9
	VH	VL	VH	VL	VH	VL
V gene family	VH3	VL1	VH4	VK3	VH4	VK10
Germline genes, V	V3-48*01	LV1-47*01	V4-31*06	KV3-11*01	V2-2*02	KV10-96*01
D	D2-02*01		D5-05*01		D1-1*01
J	J6*02	LJ3*02	J4*02	KJ4*01	J4*01	KJ1*01
Number of R/S mutations (Kabat), CDRs	8/1	4/1	2/0	0/1	0/0	0/0
FRs	4/1	5/4	2/5	0/0	0/0	0/0
Number of R/S mutations (IMGT), CDRs	6/0	2/1	1/0	0/1	0/0	0/0
FRs	6/2	7/4	3/5	0/0	0/0	0/0
Junctional deletions/insertions (nucleotides)	39/4	30/28	26/8	3/0	5/0	6/5
VH Karray residue conservation, %	100	—	43	—	50	—

Neshat et al (24) reported loss of antibody binding activity for the gp120 superantigenic site by swapping of the FR1, FR3 and CDR2 segments. We tested whether replacing these segments in the V_H4-family scFv GL2 scaffold with the corresponding V_H3-family scFv JL427 segments results in gain of the binding activity. Compared to wildtype scFv GL2, the mutants scFv GL2_FR1m, scFv GL2_FR3m and scFv GL2_CDR2m containing the V_H3-family FR1, FR3 and CDR2 of scFv JL427, respectively, displayed improved E-gp120 and E-416-433b binding (Figure 5A–5C; by 18—40-fold for E-gp120 and 32—57-fold for E-416-433b; Figure 5D shows an example of scFv concentration versus binding data). The binding was influenced only marginally by replacing scFv GL2 V_H CDR1 with the corresponding scFv JL427 CDR (scFv GL2_CDR1m). The changes in E-416-433b and E-gp120 binding activity observed by introducing the mutations into the scFv were quantitatively non-identical, presumably because of fine differences in the CD4BD^core conformation expressed by the two test antigens. It may be concluded that FR1, FR3 and CDR2 of scFv JL427 provide important contributions in CD4BD^core recognition.

Figure 5. Contribution of scFv JL427 framework regions to CD4BD^core peptide binding.

A. scFv JL427 FR-containing mutants. The FR1, FR3, CDR1 or CDR2 segments of the V_H3-family scFv JL427 were inserted in place of the corresponding V_H4-family scFv GL2 segments, designated respectively, GL2_FR1m, GL2_FR3m, GL2_CDR1m and GL2_CDR2m. B. Increased E-416-433b binding by scFv JL427 FR-containing mutants. Values (means ± s.d.) represent fold increase of mutant binding relative to wildtype scFv GL2, computed as the ratio of scFv concentrations displaying A490 0.25 (A490 0.25 observed at 23±1 μg/mL wildtype scFv). scFv constructs were purified by metal-affinity chromatography.E-416-433b (70 ng/well). C. Increased E-gp120 binding by scFv JL427 FR-containing mutants. Values (means ± s.d.) represent fold increase of mutant binding relative to wildtype scFv GL2 computed as the ratio of scFv concentrations dipalying A490 was 0.25 (A490 0.25 observed at 36±2 μg/mL wildtype scFv). scFv constructs were purified by metal-affinity chromatography. E-gp120 (100 ng/well). D. Example E-gp120 binding data for scFv GL2_FR3m. Values are means ± s.d. corrected for nonspecific binding.

Molecular modeling of scFv JL427 suggested that the 30-residue FR1 and 32-residue FR3 are spatially separated from the antigen binding pocket composed of the CDRs (Figure 6A). Four of the 17 CDR2 residues approach the rim of the CDR-binding pocket, but the rest of CDR2 is well-separated from the pocket. In contrast, the CDR1 segment (which was without effect on the binding activity) lines the CDR-binding pocket. Local changes in backbone FR1, FR3 and CDR2 topography of the wildtype and mutated scFv GL2 were evident, but the overall structure of these segments was mostly maintained (Figure 6B–D; RMSD for FR1, FR3 and CDR2 mutants, respectively, 1.47, 0.86 and 1.10 Å). The overall wildtype and mutant scFv protein backbones were largely superimposable, suggesting the absence of global conformational changes (RMSD for scFv GL2_FR1m, scFv GL2_FR3m and scFv GL2_CDR2m backbones, respectively, 1.84, 0.14 and 0.32 Å). The levels of amino acid identities and conservations, respectively, between the wildtype and mutant segments was (ClustalW2 grouping; Supplemental Table SII): 19 and 6 residues in FR1 (length 30 residues), 20 and 8 residues in FR3 (length 32 residues), and, 6 and 8 CDR2 residues (length 17 residues; the wildtype scFv GL2 CDR2 is 1 residue shorter compared to the mutant). Ligand binding sites consist of non-contiguous or contiguous flat surfaces or concavities and protrusions (49). No marked concavity or protrusion was noted in the scFv FR1/FR3/CDR2 regions outside the CDR-binding pocket, arguing against a lock-and-key fit CD4BD^core binding model. This leaves the interpretation of surface chemical forces at a comparatively flat surface as the explanation for FR-CD4BD^core binding interactions.

Figure 6. scFv JL427 models.

A. Two scFv JL427 surface views. The CDR-binding pocket (green) is well separated from the V_H domain FR1 (white) and FR3 (pink). The V_H-CDR2 residues (brown) are located mostly outside the CDR-binding pocket. V_H-CDR1 residues are shown in magenta. Remaining V_H and V_L surfaces are shown in dark blue and light blue, respectively. B. Superimposition of scFv GL2 wildtype (red) and mutant (white) V_H-FR1. r.m.s. deviation 1.47Å. C. Superimposition of scFv GL2 wildtype (red) and mutant (white) V_H-FR3. r.m.s. deviation 0.86Å. D. Superimposition of scFv GL2 wildtype (red) and mutant (white) V_H-CDR2. r.m.s. deviation 1.10Å. E. Contiguous scFv JL427 V_H domain 14-residue surface composed of Karray residues and somatically replaced residues. The 8 Karray residues in this stretch belong to V_H FR1, FR3 or CDR2 (yellow – Arg19, Lys75, Tyr79, Gln81, Asn82a, Tyr59, Lys64, Gly65. Note that CDR2 residues Tyr59, Lys64 and Gly65 have been reclassified as FR3 residues in the IMGT database). The contiguity is extended on the left by 4 somatic replacements (Gly54, His56, Thr57, Asn 58) and on the right by 2 somatic replacements (Ser 74, Thr77) shown in red.

The V_H/V_L domain CDR3 sequences of both scFv clones have been diversified by V-(D)-J gene recombination, a process completed prior to the immunogen-driven phase of B cell differentiation. The role of immunogen-driven selection in antibody maturation can be reliably assessed from the replacement/silent mutation ratios (R/S ratios) outside the V-(D)-J junctions (50). R/S ratio for scFv JL 427 V_H FRs, V_H CDRs and V_L CDRs but not the V_L FRs exceeded the predicted ratio for the random mutational process (Table II), suggesting selection of the replacement mutations. A further indication of immunogen-driven selection of replacement mutations was the finding of a contiguous surface composed of 8 Karray residues belonging to FR1, FR3 and CDR2 extended by 2 FR3 and 4 CDR2 residues acquired by somatic mutation (Figure 6E; individual residues constituting the contiguous surface are identified in Supplemental Table SII). The probability of finding 6 somatic replacements in surface contiguity with the 8 Karray residues by random chance alone is minuscule [P=(13/82)⁶ =1.6×10⁻⁵; where 13 is the total number of V_H gene replacement mutations and 82 is the V_H gene length less the 8 Karray residues]. All 14 resides forming the contiguous surface are located well outside the CDR-binding pocket.

Constitutive CD4BD^core-specific polyclonal antibodies

We previously used electrophoresis assays to demonstrate specific covalent binding of E-421-433 and E-416-433 by a subset of constitutive IgAs purified from the pooled serum of non-infected humans without autoimmune disease (example in Figure 7A, inset; refs 19, 20). In the present study, we observed detectable E-416-433a binding by the individual polyclonal IgA preparations from all 5 healthy, non-infected human subjects tested (determined as in Figure 7A; combined intensity of the IgA heavy and light chain adduct bands with E-416-433a bands in arbitrary volume units 222,370 ± 96,185, mean ± s.d.; range 106,648–354,872). To evaluate the functional significance of the antibodies, we tested HIV neutralization by the IgA. The purified serum IgA fractions did not neutralize HIV detectably when tested by the “low-sensitivity” assay commonly employed to measure antibody effects on PBMC infection (Figure 7A, 1 hour HIV preincubation with IgA in diluent containing 10% FBS). Two CD4BD^core-specific antibody subsets were isolated from a pool of 10 subjects serum IgA by chromatography on immobilized E-416-433: (a) noncovalently bound antibodies recovered by acid elution; and (b) covalently bound antibodies recovered by treatment with 2-pyridine aldoxime, a reagent that cleaves the nucleophile-electrophile covalent bond. The same affinity chromatography procedure was applied previously to isolate epitope-specific antibodies from HIV infected patients (4). Only 0.04% and 0.005% of the total serum IgA from non-infected humans was recovered, respectively, in the noncovalently and covalently bound fractions. Both CD4BD^core-specific fractions but not the non-binding IgA in the column flow-through neutralized HIV (Figure 7A, IC50 for noncovalently bound and covalently bound IgA, respectively, 4.2 and 0.5 ng/ml). The affinity chromatography procedure entails extensive washing of the column that can be expected to remove antibodies with weak CD4BD^core-reactivity. The studies suggest that a small minority of highly CD4BD^core-reactive constitutive IgAs express sufficiently robust neutralizing activity for consideration as effective anti-HIV antibodies.

Figure 7. Neutralizing activity of constitutive CD4BD^core-specific antibodies.

A. Neutralizing CD4BD^core-specific IgA subset from sera of non-infected humans. Pooled serum IgA from 10 healthy humans without HIV infection was fractionated on E-416-433 conjugated to agarose, unbound IgA was collected, and noncovalently bound IgA was recovered by acid elution. Covalently bound IgA was recovered by treating the gel with pyridine 2-aldoxime methiodide and elution with the acid buffer. HIV neutralization was measured using human PBMC hosts and subtype C HIV strain 97ZA009 (coreceptor CCR5-dependent). Values are means ± s.e.m. of 4 replicates. Inset, Streptavidin-peroxidase stained reducing SDS-electrophoresis gels showing binding of E-421-433 by total serum IgA loaded on the E-416-433 column. Lanes 1, 2 and 3: IgA incubated with E-421-433 for 1, 4 and 20 hours, respectively. Lanes 4, 5 and 6: IgA incubated with control E-VIP for 1, 4 and 20 hours, respectively. IgA, 80 μg/ml; E-421-433 or E-VIP, 10 μM. Complexes of E-421-433 with the heavy chain subunit (65 kDa) and light chain subunit (25 kDa) are evident. B. Weak neutralization by unfractionated serum IgM from non-infected humans. Polyclonal IgM purified by chromatography on immobilized anti-μ chain antibodies from pooled human serum was tested for neutralization of strain 97ZA009 by the low-sensitivity method (preincubation of IgM-HIV mixture in 10% FBS for 1 h) or high-sensitivity method (preincubation of IgM-HIV mixture in 0.66% FBS for 24 h). Means ± s.e.m. of 4 culture replicates. C. IgM neutralization specificity. Inhibition of human serum IgM (10 μg/mL) neutralizing activity monitored by the high-sensitivity assay as in panel B in the presence of E-421-433 (100 μM) or irrelevant E-VIP (100 μM). The control incubation received diluent containing 1% dimethysulfoxide instead of the peptides.

Serum IgG (29) and IgM (17, 19) from non-infected humans are reported to bind the CD4BD^core-containing superantigen determinant. Similarly, monoclonal human IgMs from patients with Waldenström’s macroglobulinemia displayed variable but specific recognition of E-421-433 (19). Like the serum IgA prior to affinity chromatography on immobilized E-416-433, pooled serum IgM from non-infected humans (n=10) did not neutralize HIV in the commonly-used low-sensitivity assay, but the high-sensitivity assay detected IgM neutralizing activity (24 h HIV-IgM preincubation in 0.66% FBS; Figure 7B). The weak neutralizing activity was attributable to CD4BD^core recognition, evident from competitive inhibition of the activity by E-421-433 but not the irrelevant peptide E-VIP (Figure 7C).

Induction of constitutive CD4BD^core-specific IgMs

To obtain unambiguous evidence for HIV neutralization by CD4BD^core-reactive antibodies with no requirement for V region somatic mutations, we studied murine IgM class antibodies induced by immunization with E-416-433b. E-416-433b is an electrophilic immunogen that can bind covalently to nucleophilic BCRs. IgMs are the first antibodies secreted by B lymphocytes and often contain sparse V region somatic mutations (50). We proposed covalent BCR binding by electrophilic immunogens as a cellular stimulation strategy for inducing the synthesis of antibodies to superantigen epitopes (38). In principle, immunogen-driven clonal expansion of the pre-existing CD4BD^core-specific B cell subset alone may permit increased antibody synthesis with no requirement for further antibody improvement by the somatic mutation pathway (Figure 1A). Immunization of mice with E-416-433b induced the production of serum IgM (Figure 8A) and IgG with E-416-433c binding activity (serum IgM titer 1:2500, serum IgG titer 1:35,000; no E-416-433c binding of either antibody class detected at 1:100 preimmune serum). We focused on the IgM antibodies. Of 1,199 hybridoma wells screened, 77 wells were positive for E-416-433c binding IgMs (A490 0.2—3.0). The V domains of 6 IgMs with the highest E-416-433c binding activity were sequenced. Four IgMs contained V regions exactly identical in sequence to their germline V_H and V_L gene counterparts (clones 2G9, C11, 1F4, and 2G2; E-416-433c binding activity at 30 μg IgM/mL in A490 units, respectively, 0.49 ± 0.01, 0.45 ± 0.01, 0.51 ± 0.01 and 0.94 ± 0.04).

Figure 8. CD4BD^core-specific IgM induced by E-416-433 immunization.

A. Amplified E-416-433c binding by serum IgM from E-416-433b immunized mice. Serum was obtained on day 42 after initiation of immunization. Data are means ± s.d. of pooled serum dilutions (n=8 mice). B. Gel filtration of IgM 2G9. Superpose 6 FPLC column (0.4 ml/min; 50 mM Tris-HCl, 0.1 M glycine, 0.15 M NaCl, 0.1 mM CHAPS pH 7.8). Arrows indicate marker proteins: thyroglobulin (660 kDa), ferritin (440 kDa), catalase (232 kDa) and IgG (150 kDa). Inset, Reducing SDS-electrophoresis and immunoblots of IgM 2G9. Stained with Coomassie Blue (lane 2) and anti-μ/κ antibody (lane 3). Lane 1, Marker proteins with mass shown in kDa. The 50 kDa band in lane 2 and 3 was identified previously as a μ chain fragment (36). C. Neutralization of CCR5-dependent HIV subtype A, B, C and D strains by IgM 2G9. For strain 97ZA009, the activity of a second IgM preparation is included (IgM 2G9 preparation 2). Values are means ± s.e.m. of four independent culture replicates. D. Competitive inhibition of IgM 2G9 binding to immobilized E-416-433c by soluble proteins and peptides. IgM (30 μg/mL) binding to E-416-433c (70 ng/well) was measured in the absence or presence of increasing concentrations of E-416-433a, sCD4, gp120, control Sh416-433 and control ovalbumin. Residual binding was computed as percent of the binding in the absence of inhibitors (A490 0.67 ± 0.10).

IgM 2G9 was picked as a representative E-416-433c binding germline antibody for further studies). Its V_H and V_L regions are encoded, respectively, by germline genes V2-2*02 and KV10-96*01 (Table II). Twenty five of the 30 V_H-FR1 positions, 28 of the 32 V_H FR3 positions and 10 of the 16 V_H-CDR2 positions in IgM 2G9 are occupied by identical residues or conservative substitutions compared to the corresponding FRs of human scFv JL427 (Supplemental Table SII). The IgM was composed of heavy and light chain subunits with the anticipated mass, and a predominantly pentameric species was evident under non-denaturing conditions (Figure 8B). IgM 2G9 neutralized all four strains drawn from subtypes A, B, C and D (Figure 8C; IC50 range 0.002–0.51 μg/mL). A second independent IgM 2G9 preparation also neutralized subtype C strain ZA009 (IC50 2.9 and 8.5 μg/mL in 2 assays). It may be concluded that IgM 2G9 neutralizes genetically diverse HIV strains despite the use of germline V gene sequences free of somatic mutations.

Specificity of the IgM for the CD4BD^core was evident from competitive inhibition of E-416-433c binding by E-416-433a, gp120 and soluble CD4 (sCD4) but not ovalbumin (Figure 8D). As sCD4 binds E-416-433c specifically (4), it is predicted to inhibit IgM binding to E-416-433c by steric hindrance. We also evaluated the role of individual 416-433 residues by measuring competitive inhibition of IgM—E-416-433c binding in the presence of wildtype peptide NE-416-433 and mutant 416-433 peptides containing Ala replacements. Attenuated competitive inhibition by 2.1->166-fold was evident for 9 of the 17 amino acid positions tested (Table III). The control shuffled-sequence peptide (sh416-433) did not inhibit the binding (<8.4 ± 6.2% inhibition; mean ± s.d.). The amino acids at 5 of the 9 positions are also important for sCD4 binding (13, 15, 16).

Table III. Amino acid requirements for IgM 2G9 recognition of the CD4BD^core epitope.

Binding of IgM 2G9 (30 μg/mL) to immobilized E-416-433c was measured in the absence or presence of increasing concentrations of wildtype and mutant 416-433 peptides containing Ala substitutions at individual positions (0.4–50 μg/ml). A490 in absence of competitor peptides was 0.67 ± 0.10. IC₅₀ values were determined from plots of % residual binding (computed relative to binding without competitor peptide) versus competitor peptide concentration fitted to the equation: residual binding (%) = 100/(1+10^{(Log[competitor peptide]−Log[IC50])}). Loss of IgM binding activity is evident from an increased (IC50_{mutant peptide}/IC50_{wildtype peptide}) ratio. “None” indicates changes in binding ratios <1.5-fold. IC50 for the wildtype peptide was 27μg/mL. Residues reported to contribute in binding of gp120 to CD4 are marked “X”, as determined by crystallography (PDB 2B4C) (13) or site-directed mutagenesis (15, 16).

Amino acid	Fold loss of IgM binding activity	Contribution to CD4 binding
L416	3.7
P417	None
C418	None
R419	6.2
I420	>166.6
K421	>166.6	X
Q422	None
I423	>166.6
I424	>166.6
N425	None	X
M426	2.1	X
W427	>166.6	X
Q428	None	X
E429	None	X
V430	3.1	X
K432	5.0	X
A433	None

Host cell effects

Autoantibodies reactive with antigens expressed by PBMCs can inhibit HIV infection by an anti-host cell mechanism (51). A cross-reaction of anti-CD4BD^core antibodies with a PBMC antigen may explain their observed neutralizing activity. Antibodies to phosphatidylserine inhibit HIV infection in PBMC cultures by inducing β-chemokine release from monocytes, resulting in saturation of the chemokine coreceptors needed for HIV entry into the cells (40). Contaminant LPS (endotoxin) in antibody preparations can also inhibit HIV infection by inducing β-chemokine release from the cells (52–54).

Without the constant region (Fc), an anti-phospholipid antibody did not inhibit HIV infection of PBMCs (40). The scFv clones reactive with the CD4BD^core in the present study do not contain the Fc region, arguing against an anti-phospholipid antibody-like neutralizing mechanism. Inhibition of HIV infection by anti-phospholipid antibodies was completely blocked by antibodies to MIP-1α or MIP-1β (40). In contrast, the HIV neutralizing activity of IgG JL427 was maintained in the presence of antibodies to MIP-1α, MIP-1β and RANTES (Figure 9A, B). In the presence of excess anti-MIP-1α antibody, an increase of basal infection (no IgG JL427) was evident compared to the control non-immune IgG. In previous studies, MIP-1α inhibited the infection of PBMCs potently, and neutralizing antibodies to constitutively-produced β-chemokines in PBMCs and T cell cultures enhanced the infection, consistent with CCR5 availability as a factor regulating the infection (54, 55). The differences in IgG JL427 neutralizing activity in the control non-immune IgG versus experimental cultures containing antibodies to MIP-1α, MIP-1β and RANTES was within the range of experimental error (<25%). β-chemokine mediated inhibition of infection, therefore, is not a factor in HIV neutralization by the anti-CD4BD^core antibody.

Figure 9. Maintenance of IgG JL427 neutralizing activity in the presence of antibodies to β-chemokines.

Shown are p24 levels of PBMC lysates infected for 4 days by HIV strain 97ZA009. PBMCs were mixed with the 2.5 μg/mL (panel A) or 15 μg/mL (panel B) goat anti-β-chemokine IgG or control goat non-immune IgG and immediately cocultured with HIV pretreated with diluent or IgG JL427 (100 ng/mL, 1 h). The same anti-MIP-1α and MIP-1β IgG tested in this experiment completely blocked the β-chemokine mediated inhibition of HIV infection by anti-phospholipid antibodies (40).

To further rule out the β-chemokine mechanism and other potential neutralization mechanisms due to antibody recognition of a PBMC antigen, we tested whether the neutralizing activity is dependent on the duration of HIV preincubation with test antibodies to the CD4BD^core. The PBMCs were exposed to the antibodies for equivalent durations at all data points. Time-dependent neutralization is observed in this experimental protocol only if the antibody targets a viral antigen, whereas the level of neutralization remains constant if the target antigen is located on the PBMCs. Progressively increasing neutralization by IgG JL427 and IgM 2G9 as a function of increasing time of HIV-antibody preincubation was observed (Figure 10A, B). As expected, the time-dependency was lost at a saturating IgG JL427 concentration, and the curve shifted to the left at a sub-saturating IgM 2G9 concentration. Consistent with findings that the β-chemokine mediated inhibition of infection by antibodies to phosphatidylserine does not require antibody-HIV contact (40), the inhibitory effect of a control antibody to phosphatidylserine remained constant regardless of the length of antibody-HIV preincubation (Figure 10C). Another infection-inhibiting mechanism was reported for an antibody to phosphatidylinositol, internalization of antibody-coated HIV by macrophages, followed by release of β-chemokines (56). However, the infection was not impeded if macrophages were pretreated with this antibody prior to contact with the virus, suggesting functionally futile antibody consumption by excess cellular lipid. In contrast, IgG JL427 to the CD4BD^core pretreated with PBMCs maintained the HIV neutralizing activity (% neutralization observed in PBMCs cocultured with IgG JL427 3 hours prior to addition of virus versus simultaneous coculture with antibody and virus, respectively: 92 ± 2 and 93 ± 3; 0.5 μg IgG JL427/ml, strain 97ZA009). Taken together with finding that the neutralizing activity is maintained in the presence of anti-chemokine antibodies, the data indicate that HIV neutralization by the anti-CD4BD^core antibodies occurs by recognition of HIV free of host cell artifacts.

Figure 10. Dependence of antibody neutralizing activity on duration of pretreatment with HIV while holding constant the exposure of PBMCs to antibody at all data points (4 days).

A. IgG JL427 to the CD4BD^core. Shown are neutralizing activities (mean ± s.e.m) following preincubation of HIV strain 97ZA009 with the IgG (0.1 or 0.5 μg/mL) for 0, 1, 3 and 24 hours. PBMCs were maintained in the presence of antibody throughout until harvest of extracellular and intracellular p24 from the culture wells. B. IgM 2G9 to the CD4BD^core. Methods were in as in Panel A. HIV was treated with IgM 2G9 (0.3 or 1.3 μg/mL) in 0.66% FBS for the indicated durations. C. Control IgG 4B6 to phosphatidylserine. Methods were as in Panel A. As predicted for inhibition of infection via an anti-host cell mechanism, the effect of this antibody was independent of the length of HIV-antibody pretreatment.

Inhibition of HIV infection by LPS is highly variable depending on the LPS source, the test virus strain, host PBMC source and monocyte content. In the report of Geonnotti et al (54), most test LPS preparations did not inhibit infection in about half of the assays, but when observed, the inhibitory effect was potent (median IC50 for LPS in assays showing inhibition: 30–210 pg/mL). Another paper from the same group reported the requirement for very large amounts of LPS for inhibition of infection (IC80 4 μg LPS/ml; ref 40). In our hands, an authentic LPS preparation inhibited infection of 3 PBMC preparations by HIV strain 97ZA009 variably (IC50 1800 pg/mL for one PBMC preparations and >100,000 pg/mL for the other two preparations). The LPS concentration at the 50% neutralizing concentrations of purified scFv JL427 (0.2 pg/mL), IgG JL 427 (0.2 pg/ml) and IgM 2G9 (undetectable) were below the threshold needed to interfere with HIV infection. Unlike the inconsistent inhibitory effect of authentic LPS, the scFv clones consistently neutralized the infection of all test PBMC preparations (n=20) by all test HIV strains (n=17) with high potency.

The explanation of interfering host cell effects is also inconsistent with: (a) specific inhibition of scFv JL427 and IgM 2G9 neutralizing activity by the CD4BD^core peptide analog, and (b) potent neutralization by the CD4BD^core-specific IgA fraction from non-infected humans but not the flow-through from the affinity column. Similar studies ruled out host cell artifacts as the explanation for HIV neutralization by anti-CD4BD^core antibodies from HIV infected subjects (4) and mice immunized with E-gp120 (38).

DISCUSSION

Using the commonly-used low-sensitivity neutralizing assay, we isolated a very small subset of CD4BD^core-reactive IgAs produced spontaneously with no prior exposure to HIV that neutralized HIV potently. The IgAs were found in sera from humans without HIV infection or autoimmune disease. Like IgGs, class-switched IgAs usually contain somatically mutated V regions. Constitutive production of antibodies to the CD4BD^core is not restricted to the IgA classes. We and others reported constitutive antibodies of the IgG (29) and IgM (17, 19) class with this specificity. Rare scFv clones from humans without HIV infection also recognized the CD4BD^core specifically and neutralized genetically diverse HIV strains, verifying constitutive production of the neutralizing antibodies. Additional observations supporting the pre-existing CD4BS^core specificity include binding of the antibodies to CD4BD^core spanning peptides E-416-433 and E-421-433 but not irrelevant peptides, inhibition of the binding to E-416-433 by sCD4 and gp120 but not an irrelevant protein, and the previously documented catalytic cleavage of gp120 dependent on initial noncovalent binding to the CD4BD^core (19, 20).

Lupus patients produce CD4BD^core peptide-binding antibodies at increased levels (45), and clones with this specificity may be overrepresented in our scFv library. We noted the possibility of an unusual immune response to endogenous retroviral and exogenous bacterial antigens resembling the CD4BD^core as the cause (28). However, sera from non-infected humans without autoimmune disease also contain antibodies to the gp120 superantigen site (19, 20, 29), and finding of potent CD4BD^core-directed IgAs in the present study indicate that the autoimmune disease is not a prerequisite for constitutive production of the antibodies. Other groups have proposed insufficient gp120 binding affinity of antibodies containing V regions in their germline configuration and insufficient V region somatic mutations following exposure to HIV as explanations for infrequent production of adaptively matured neutralizing antibodies to conserved HIV epitopes like the CD4BD (10). Neither explanation applies to constitutive antibodies directed to the CD4BD^core epitope. We observed broad HIV neutralization by the CD4BD^core-specific scFv clones. One of the clones contained minimal V region mutations (scFv GL2) and the other was more extensively mutated. Although neutralizing antibodies to the CD4BD^core are difficult to induce adaptively, a small subset of these antibodies is present constitutively.

We conceive the constitutive BCR repertoire as a source of diverse antibodies with varying CD4BD^core reactivity, of which only a small subset displays sufficient fine specificity relevant to HIV vaccination. The specificity likely depends both on the initial events that commit B cells to use of a particular pair of rearranged germline V_H-V_L genes and the subsequent mutational changes. The gp120 superantigen site binds most strongly to human V_H3-family antibodies (22), but interactions with antibodies from other V gene families are also evident (19, 48). In the present study, the V_H3-family scFv (clone JL427) expressed superior CD4BD^core binding and HIV neutralizing activity compared to the V_H4-family scFv GL2. Swap mutation studies suggested that the FR1, FR3 and CDR2 segments of scFv JL427 lying outside the traditional CDR-based antigen-binding pocket provide important contributions in CD4BD^core binding activity. scFv JL427 contains all 15 germline V_H gene-encoded amino acids and 1 somatically-acquired replacement previously suggested to be important in binding the gp120 superantigen site (23). Mechanisms that can diversify the constitutive FR-based repertoire and explain the production of rare CD4BD^core–specific neutralizing antibodies are: (a) Varying binding activity of the 97 V_H3 germline genes and the remaining 166 V_H1–V_H7 germline genes; (b) A direct contribution from the V_L domain, suggested by the CD4BD^core recognition activity of certain free light chain subunits (48); (b) Binding site structural alterations caused by V_H–V_L domain pairing; and (c) Remote structural changes in the CDRs. In particular, V_H CDR3 is subject to substantial length and sequence changes during V-D-J gene rearrangement prior to contact with immunogen. Sub-Ångstrom-to-Ångstrom level movements of the backbone and amino acid side chains attendant to remote V domain structure alterations can impact their reactivity with antigens (37). The V_H CDR3 of both CD4BD^core-reactive scFv clones in the present report contains extensive structural changes due to the gene rearrangement mechanism.

Exploiting the constitutive neutralizing activity for HIV vaccination requires identification of an immunogen that stimulates clonal expansion and differentiation of B cells producing the CD4BD^core-reactive antibodies. Improving the CD4BD^core-reactivity by V gene somatic mutations is also desirable. Unlike stimulatory CDR binding of traditional immunogens, FR binding to superantigens down-regulates B cells (21). Only limited escape from the down-regulatory effect seems feasible under physiological circumstances. The Protein A superantigen site induced transient clonal proliferation prior to deletion of B cells (57), low-level CD4BD^core binding antibodies were detected in HIV infected patients (26) and CD4BD^core-specific class-switched IgAs with broad and potent HIV neutralizing activity were identified in patients with prolonged HIV infection over two decades (4). An immunogen-driven regulation of CD4BD^core-reactive antibody production by non-infected humans is also conceivable, as the CD4BD^core contains partial sequence identity with a protein from the commensal bacterium Streptococcus gordonii and certain endogenous retroviral sequences (28). Our structural studies suggested that scFv JL427 contains an FR-based CD4BD^core binding site with somatic replacement mutations, a sign of immunogen-driven selection. The spatially-separated CDR-binding pocket also contained somatic replacements. A structural precedent for two binding sites expressed within a single Fv is provided by monoclonal antibodies raised by immunization with E-gp120, the CD4BD^core binding site dominated by the FRs and a second CDR pocket that binds a physically distinct gp120 epitope (38). We hypothesized that simultaneous CDR engagement of the second epitope furnishes a stimulatory signal compensating for down-regulatory CD4BD^core binding to the FRs, thereby amplifying neutralizing antibody synthesis by B cells. In principle, binary engagement of the FR-based site and CDR-binding pocket by two epitopes expressed by a yet-to-be identified physiological immunogen may explain amplified production of antibodies without the occurrence of HIV infection.

Electrophilic immunogens are first-generation vaccine candidates holding the potential of inducing a broadly neutralizing antibody response. Immunization with E-gp120 induced a more robust class-switched antibody response compared to gp120 devoid of the electrophilic groups (58). In addition to the binary FR-CDR binding mechanism, our studies on the immunogenicity of E-gp120 suggested covalent electrophile binding at nucleophilic BCR sites as an independent stimulatory factor that may bypass the down-regulatory effect of noncovalent CD4BD^core-BCR binding (38). E-416-433 lacks a second CDR-binding epitope, but it contains the electrophilic phosphonate and peptide epitope available for covalent BCR binding coordinated with specific noncovalent binding at the FR-based site. IgMs often contain few or no V region somatic mutations, as they are produced from plasma cells developed early in immunogen-driven phase of the response, prior to extensive somatic V region mutations and IgG/IgA class-switching accompanying the germinal center reaction (50). To evaluate constitutive antibody amplification free of somatic mutation requirements, we analyzed the IgM antibodies in E-416-433 immunized mice. The mice mounted a polyclonal IgM response with E-416-433 binding activity. A monoclonal IgM induced by E-416-433 expressed specific E-416-433 binding activity that was inhibited competitively by gp120 and CD4 despite a complete absence of somatic mutations in its V regions. This IgM also neutralized diverse HIV strains. Clonal expansion of B cells with constitutive CD4BD^core recognition activity alone with no reliance on immunogen-driven V region mutations, therefore, is a valid basis to induce the synthesis of broadly neutralizing antibodies. With respect to effective vaccination, maintaining and improving the neutralizing activity as the immune responses matures to class-switched IgG/IgA antibodies, is an important objective. Others have reported both increased and decreased HIV neutralization by the IgM version of antibodies to certain HIV epitopes containing the same V domains compared to their smaller IgG/IgA counterparts (59, 60). The IgM scaffold supports comparatively rapid catalytic cleavage of gp120 compared to IgG (19), and multivalent epitope recognition by IgM may allow more avid noncovalent binding. However, the large IgM size may cause sterically-hindered epitope binding, and IgMs are found in blood at concentrations lower than the IgGs.

To selectively amplify the small subset of constitutive antibodies with the most potent CD4BD^core reactivity, the electrophilic immunogen must mimic the native HIV CD4BD^core conformation with sufficient accuracy and rigidity (Figure 1B middle panel). Imperfect conformational mimicry will amplify constitutive antibodies with lesser neutralizing potency. Similarly, an immunogen that expresses an insufficiently rigid CD4BD^core conformation can be anticipated to amplify a constitutive antibody subset with lesser neutralizing potency due to the induced-fit binding mechanism (Supplemental Figure S1B). Moreover, the process of immunogen-driven selection of mutations over the course of antibody maturation will improve immunogen binding activity, not native CD4BD^core binding activity, and a conformationally-deviant immunogen may select FR mutations that cause loss of neutralizing activity. Inducing a CDR-based neutralizing antibody response is even more onerous. Synthetic peptides such as the CD4BD^core peptides can assume alternate conformations depending on their microenvironment. The induced-fit binding mechanism will force a flexible immunogen that binds the CDRs into an inappropriate conformation, resulting in induction of non-neutralizing antibodies (Supplemental Figure S1B; see ref 61 for review). The importance of conformational considerations is evident from reports of vigorous IgG production induced by non-electrophilic CD4BD^core peptides that displayed inconsistent HIV neutralizing activity and varying binding specificity (33, 62, 63). Importantly, the immune response to the non-electrophilic peptide immunogens is likely dominated by traditional antibodies with CDR-based binding sites, as the noncovalent superantigen epitope-FR binding down-regulates B cells.

The utility of constitutive antibodies for HIV vaccination depends on the veracity with which the primary virus-PBMC neutralization assay reports a genuine anti-virus antibody effect. The assay is often cited as the “gold standard” for determining HIV neutralization in tissue culture because it approximates the natural process of infection in vivo. However, discrepant neutralization by antibodies to various HIV epitopes in the primary virus-PBMC assay versus the pseudovirion-engineered reporter cell assay have been noted (64–68), including the CD4BD^core epitope. The CD4BD^core-specific antibodies in long-term survivors of HIV infection (4) or induced in mice by E-gp120 immunization (38) displaying neutralizing activity in the primary virus/PBMC assay did not impede pseudovirion entry into genetically engineered host cells expressing CD4 and chemokine coreceptors (TZM/Bl cells) (68). scFv JL427 at concentrations showing readily-detectable neutralization in the PBMC-based assay was also ineffective in the TZM/Bl-based assay (12 μg/mL). The underlying reason remains to be elucidated, but overexpression of the HIV coreceptors and an altered CD4BD^core conformation on the pseudovirion surface are potential factors (68). Inhibition of HIV infection due to β-chemokine mediated antibody host cell effects and contaminant endotoxin was noted in the PBMC assay (40, 54). We ruled out these explanations from findings that neutralization by anti-CD4BD^core antibodies was preserved in the presence of anti-β-chemokine antibodies and occurred by a mechanism involving recognition of HIV, not host cells. The nneutralization increased progressively with increasing time of antibody contact with HIV while holding PBMC exposure to antibodies constant; enriched neutralizing activity was evident in the epitope-specific subset of polyclonal antibodies; CD4BD^core-reactive recombinant antibodies displayed neutralizing activity; the CD4BD^core peptide inhibited antibody neutralization competitively; a neutralizing monoclonal antibody was induced by immunization with a CD4BD^core-containing immunogen; and this antibody recognized amino acids within the CD4BD^core important for CD4 binding. It may be concluded that the PBMC-based assay is a valid measure of specific CD4BD^core recognition and HIV neutralization by constitutive antibodies.

In summary, our studies reveal the constitutive antibody repertoire as a novel basis for inducing broadly neutralizing antibodies expressing FR-based specificity for the CD4BD^core. The structural properties of a broadly neutralizing antibody fragment from non-infected humans were consistent with adaptive improvement of the constitute antibodies under the influence of an unidentified immunogen. Electrophilic immunogen-driven amplification of constitutive IgM antibodies with unmutated V genes appears feasible. Previous findings of slow appearance of broadly neutralizing anti-CD4BD^core antibodies in patients with prolonged HIV infection (4) and the more rapid induction of such antibodies observed in mice immunized with E-gp120 (38) also support a vaccination strategy that amplifies and improves the neutralizing function of constitutive antibodies.

The abbreviations used are

AVU

arbitrary volume unit

BCR

B cell receptor

BSA

bovine serum albumin

Bt

biotinamidohexanoyl

CD4BD

CD4 binding domain

CD4BD^core

gp120 region spanning residues 421-433

CDR

complementarity determining region

EU

endotoxin unit

FR

framework region

HIV

human immunodeficiency virus

KLH

keyhole limpet hemocyanin

MAb

monoclonal antibody

Ova

ovalbumin

PBMCs

peripheral blood mononuclear cells

R/S ratio

replacement/silent mutation ratio

sCD4

soluble CD4

scFv

single chain fragment variable

V_H

antibody heavy chain variable domain

V_L

antibody light chain variable domain

VIP

vasoactive intestinal peptide