A large (1.5 × 10¹⁰ members) phage-displayed naive human antibody antigen-binding fragment (Fab) library (designated m21) was constructed from peripheral blood B cells of 22 healthy donors, spleens of 3 healthy donors, and lymph nodes of 34 healthy donors as described in Materials and Methods. The phagemid used for library construction, pZYD-N1, was synthesized and briefly described previously [17]. To approximately estimate the sequence diversity of m21, 44 clones were randomly selected from the library and sequenced. No identical sequences of heavy and light chains were found; more than 80% of the clones are productive. Of the 44 clones 26 had kappa light chains and the other 18 - lambda light chains. M21 was panned against several HIV-1 Envs and human cancer-related antigens; significant enrichment was obtained with all the pannings (data not shown). These results indicate that the quality of the library is likely to be good.

To identify human IgM-derived antibodies specific for HIV-1 Envs, we panned m21 against the gp140 of a clade B isolate, R2 (gp140_R2). One of the selected antibody clones, R3H1, contained a TGA stop codon at the very beginning of the light chain due to a nucleotide deletion that resulted in a reading-frame shift. The heavy chain variable domain (VH) of R3H1, designated m0, differs from the closest human germline sequence (VH3–23) by mutations distributed in all frameworks (FRs) and complementarity determining regions (CDRs) (Figure 1a). M0 contains 15 mutations in nucleotide sequence of the variable (V) region resulting in 9 mutations in amino acid sequence. We corrected the reading frame of the light chain by site-directed mutagenesis and the corrected antibody, designated R3H1m, was expressed at high levels in bacteria as an Fab. Fab R3H1m specifically bound with relatively high affinity to gp140_R2 (EC₅₀, ~60 nM) and the gp140 of a clade C isolate, GXC44 (gp140_GXC44) (EC₅₀, ~100 nM).

Because the antibody was initially selected as a functional heavy chain-only antibody, we hypothesized that the binding activity of R3H1m should be attributed mainly to the heavy chain of the antibody and the heavy chain could be paired with light chains derived from different germlines while retaining cross-reactivity with HIV-1 Envs. We cloned m0 as an isolated antibody domain and found that it was highly soluble, stable, monomeric, and expressed at high levels in bacteria [18]. In agreement with our hypothesis it specifically bound to gp140_R2 although with low affinity (EC₅₀, >1000 nM). We further constructed a light chain-shuffling Fab library (3 × 10⁷ members) based on the heavy chain of R3H1. To increase the probability for selection of cross-reactive antibodies the library was panned sequentially against gp140_R2 and the gp140 of a clade F isolate, 14/00/4 (gp140_14/00/4). Five unique clones were identified that contained exclusively lambda light chains (Figure 1b). They bound to gp140_R2 and gp140_14/00/4 with EC₅₀ ranging from 2 to 80 nM (Table 1) while no significant binding to an unrelated antigen, bovine serum albumin (BSA), was observed (data not shown). One of the antibodies, m19, also bound with high affinity to the Envs of other three clade B isolates, gp120_Bal, gp140_JRFL and gp140_Con-S[19]. These results suggest that m19 and possibly the other selected antibodies target conserved epitopes on gp120.

To test whether the five newly selected antibodies are capable of neutralizing HIV-1 primary isolates, we used viruses pseudotyped with Envs from HIV-1 isolates representing clades A, B and C and using either CCR5 (R5) or CXCR4 (X4) or both (R5X4) as a coreceptor. None of the antibodies exhibited potent broadly neutralizing activity. Four antibodies (Fabs m19, m19b, m19c and m19d) enhanced infection by Bal even at very low antibody concentration while a positive control antibody, m36 [17], showed efficient neutralization and a negative control antibody, 1A1, did not affect entry (Figure 2am19d enhanced infection by another clade B primary isolate, JRFL, at high concentrations (Figure 2b). Interestingly, the three antibodies that enhanced significantly infection by the Bal weakly neutralized IIIB which is a TCLA clade B X4 isolate (Figure 2c

To find whether the activity of the antibodies is related to antibody size and how the viral infection could be affected by cross-linking of HIV-1 Envs, we generated a single-chain Fv fragment (scFv) (scFv m19) of m19 and a human IgG1 Fc-fusion protein (m19Fc) of scFv m19; m19 was selected for further characterization because its light chain was relatively less divergent from the germline (Figure 1b) and was the only one which did not contain any somatic mutations in the CDR3 of the light chain. When Bal was tested, the enhancing activity of scFv m19 was comparable to that of Fab m19 and slightly lower than that of m19Fc (data not shown). These data indicate that the antibody access to the epitope was not size-restricted and that the bivalency of the Fc fusion protein strengthened the antibody activity probably due to avidity effects. The ability of m19Fc to modulate HIV-1 infection in vitro was further tested with six additional clade B isolates and an isolate each from clade A and clade C, respectively (Figure 3). Enhancement was observed with JRCSF (clade B, R5) at very low m19Fc concentration and the enhancement was decreased with an increase in antibody concentration. M19Fc also conferred slight enhancement of infection by JRFL (clade B, R5) and 92UG037.8 (clade A, R5) at about 200 nM antibody concentration. It neutralized the clade B X4 isolates IIIB and NL4-3. It also neutralized R2, which is a clade B R5 isolate, displays some CD4 independence and is moderately neutralization sensitive, as well as GXC44 which is a clade C R5 isolate. No significant antibody activity was observed with the dual tropic clade B isolate 89.6. These results indicate that the antibody activity in terms of neutralization or enhancement varies when different isolates are tested and is not significantly correlated with the coreceptor usage and the CD4 dependency of the viruses in entry.

To approximately localize the antibody epitopes and begin to elucidate the underlying mechanisms of antibody neutralizing or enhancing activity, we measured the antibody binding to Envs from different isolates alone and in complex with CD4 as well as the antibody competition with well-characterized antibodies. M19Fc bound to a single-chain polypeptide analogue of the HIV-1 gp120-CD4 complex (gp120_Bal-CD4) [20] with threefold higher affinity (EC50, ~2.5 nM) than to gp120_Bal (EC50, ~10 nM) alone as measured by an ELISA (Figure 4a). It did not react with sCD4 suggesting that the antibody targeted a structure on gp120 that was outside the CD4-binding site (CD4bs) and could undergo CD4-induced (CD4i) conformational changes. The conformational changes on gp120_Bal after CD4 binding were confirmed by a CD4i bnAb, m9Fc [21], which exhibited dramatically higher binding to gp120_Bal-CD4 than to gp120_Bal (Figure 4a_Con-S, which was a synthetic Env designed by aligning the consensus Env sequences of group M [19]. This Env, by design, is composed by shorter V1, V2, V4 and V5 loops and therefore, might expose regions around the variable loops that might contain conserved neutralizing determinants. However, m9 epitope as a representative for CD4i epitopes is still completely hidden on gp140_Con-S because m9Fc did not bind in the absence of sCD4, it did bind with high affinity in the presence of sCD4 (Figure 4b). However, m19Fc showed comparable binding to gp140_Con-S with or without sCD4 (Figure 4b) suggesting that the antibody does not precisely target CD4i epitopes but other structures on the gp120. Although m19Fc was not directed against the CD4-binding pocket, it strongly competed with the CD4bs bnAb, scFv b12, which has binding surface larger than that of CD4 [22] (Figure 5aM19Fc also competed although weakly with two CD4i antibodies, scFv m9 and m36 [17] (Figure 5b). These results suggest that the epitope of m19Fc is in very close proximity to the CD4bs and the coreceptor-binding site (CORbs) which overlaps with CD4i epitopes on gp120 (Figure 6).

To estimate the possibility of rapid elicitation of these antibodies in vivo when an HIV-1 gp120-related immunogen is administered, we generated a germline-like scFv of m19 (scFv m19gem) and measured its binding to Envs from different isolates in the presence or absence of sCD4. No obvious interaction was observed with scFv m19gem when several Envs that bound the original m19 were tested (data not shown). Previous studies showed that the binding activity of an antibody in the form of Fab or scFv could be increased up to thousands of times when it is converted to dimeric formats such as an IgG. We therefore fused scFv m19gem to the Fc portion of a human IgG1 to gain possible avidity effects. Still, the Fc-fusion protein of scFv m19gem (m19gemFc) did not show measurable binding to the Envs with or without sCD4 (Figure 7). These results suggest that significant somatic diversification is required for the m19 corresponding germline antibody to achieve recognition of the m19 epitope on gp120. They also indicate that in some individuals certain IgM antibodies could undergo somatic hypermutations to relatively high-affinity binders to the Env in the absence of HIV-1 infection or immunization with Env.

We purchased the 293T cells from ATCC. Other cell lines and plasmids used for expression of various HIV-1 Envs were obtained from the National Institutes of Health AIDS Research and Reference Reagent Program (ARRRP). Recombinant gp140s were produced in our laboratories. Gp120_Bal and the single-chain fusion protein gp120_Bal-CD4 were kindly provided by T. Fouts (Institute of Human Virology, Baltimore; currently at Profectus, Baltimore, MD). Horseradish peroxidase (HRP)-conjugated anti-FLAG tag antibody and HRP-conjugated anti-human IgG (Fc-specific) antibody were purchased from Sigma-Aldrich (St. Louis).

M21 was constructed by using phagemid pZYD-N1 according to the reported protocols [52]. The sources for amplification of antibody gene fragments are commercially purchased poly A+ RNAs (BD Biosciences, San Jose, CA) from peripheral blood B cells of 22 healthy donors, spleens of three donors and lymph nodes of 34 healthy donors, respectively.

M2l was used for selection of antibodies against HIV-1 antigens conjugated to magnetic beads (Dynabeads M-270 epoxy; DYNAL Inc., New Hyde Park, NY) as described previously [53]. 5, 5, 2.5 and 1 µg of gp140_R2 were used in the first, second, third and fourth round of panning, respectively. Clones that specifically bound to gp140_R2 were identified from the third and fourth round by using monoclonal phage ELISA (mpELISA) as described [53].

A light chain-shuffling Fab library (3 × 10⁷ members) was further constructed based on the heavy chain of R3H1. The light chain repertoire was also harvested from a naive human Fab library (5 × 10⁹ members) constructed from peripheral blood B cells of 10 healthy donors [53], in addition to that from m21. The new library was panned sequentially with two gp140s from different clades. For the sequential panning, 5 and 2.5 µg of gp140_R2 were used in the first and third round, respectively; antigens were alternated with 5 and 2.5 µg of gp140_14/00/4 during the second and fourth round, respectively. Clones that bound to both gp140_R2 and gp140_14/00/4 were identified from the fourth round by using mpELISA as described [53].

The following primers were used: m36F, 5’-TGG TTT CGC TAC CGT GGC CCA GCC GGC CCA GGT GCA GCT GGTG-3’ (sense); m19FcR: 5’-GTG AGT TTT GTC GGG CCC TAG GAC GGT CAG CTT GG-3’ (antisense).

For construction of scFv, the full-length original or germline scFv gene fragment was synthesized (GenScript, Piscataway, NJ), digested with SfiI and cloned into pComb3X. To generate Fc-fusion protein, the scFv gene was PCR (primers m36F and m19FcR) amplified by using the synthetic scFv gene fragment as a template. The new scFv products appended with SfiI and ApaI restriction sites on both sides were digested and cloned into pSecTagB-Fc.

The scFv and Fab were expressed in E. coli HB2151, as described previously [53]. The bacterial pellet was collected after centrifugation at 5,000 × g for 10 min and resuspended in PBS (pH 7.4) containing 0.5 million-unit polymixin B (Sigma-Aldrich). After 30 min incubation with rotation at 50 rpm at room temperature, it was centrifuged at 25,000 × g for 25 min at 4 °C. The supernatant was used for purification of scFv and Fab by immobilized metal ion affinity chromatography (IMAC) by using Ni-NTA resin (Qiagen, Valencia, CA) according to manufacturer’s protocols. Fc-fusion proteins were expressed in 293 free style cells. 293Fectin (Invitrogen, Carlsbad, CA) was used to transfect 293 free style cells according to the instructions of the manufacturer. Three days post-transfection, the culture supernatant was harvested and used for purification by using nProtein A Sepharose 4 Fast Flow (GE Healthcarezcomx, Piscataway, NJ).

Binding and competition ELISAs were performed as described previously [17].

Viruses pseudotyped with HIV-1 Envs were prepared by cotransfection of 70–80% confluent 293T cells with pNL4–3.luc.E-R- and pSV7d constructs encoding HIV-1 Envs by using the PolyFect transfection reagent (Qiagen) according to manufacturer’s instruction. Pseudotyped viruses were obtained after 48 h by centrifugation and filtration of cell culture through 0.45 µm filters. For neutralization, viruses were mixed with different concentrations of antibodies for 1 h at 37 °C, and then the mixture was added to about 1.5 × 10⁴ HOS-CD4-CCR5 (used for all R5 and dual tropic viruses) or HOS-CD4-CXCR4 cells grown in each well of 96-well plates. Luminesence was measured after 48 h by using the Bright-Glo Luciferase Assay System (Promega, Madison, WI) and a LumiCount microplate luminometer (Turner Designs). Mean relative light units (RLU) for duplicate wells were determined. Relative infectivity (%) was calculated by the following formula: (average RLU of antibody-containing wells/average RLU of virus-only wells) × 100.