The major finding of this study is the identification of a gp41-specific cross-reactive HIV-1-neutralizing hMAb, which binds to a novel conformational epitope on gp41 and does not exhibit reactivity to self-antigens. A number of anti-gp41 hMAbs have been identified and characterized, but only several of them are neutralizing (51
). Of those, 2F5, 4E10, and Z13 bind peptides and denatured gp140s (Fig. ) (52
), and another anti-gp41 HIV-1 NAb, CL3, recognizes a linear epitope that includes the disulfide bond in the immunodominant loop of gp41 (GCSGKLICTT) (12
). It appears that immunogens based on these linear epitopes do not lead to elicitation of antibodies neutralizing primary isolates. For example, the use of ELDKWA inserted into a carrier protein did not induce HIV-1 NAbs, possibly due to the lack of an appropriate environment (absence of the lipid of the viral membrane) to support the structure of this peptide in the context of gp41 (11
). In contrast to these previously characterized neutralizing anti-gp41 hMAbs, the newly identified antibody m44 does not bind peptides and Envs without conformational integrity, nor does it bind to lipids or the other autoantigens tested. Its epitope is conformational and conserved. The m44 VH
is derived from the M29811 IGHV4-61*01 germ line, and the VL
is derived from the X12686 IGKV3-20*01 germ line. m44 has a relatively short heavy-chain CDR3 (HCDR3) (m44 HCDR3, ARGTRGGSTLDS) compared to 2F5 and 4E10, which have long HCDR3s (2F5 HCDR3, AHRRGPTTLFGVPIARGPVNAMDV; 4E10 HCDR3, AREGTTGWGWLGKPIGAFAH). Arginines in HCDR3s were reported to be associated with antibody polyreactivity (18
). 2F5, 4E10, and m44 have three arginines, one arginine, and two arginines, respectively, in the HCDR3s. While several groups have found the lipid reactivities of 4E10 (1
) and 2F5 (1
), one group has not found 2F5 reactivity with lipids, although they found 2F5 reactivity with glyceraldehyde phosphate dehydrogenase (43
). This difference is likely due to the weak apparent affinity of 2F5 for lipids and to assay differences. In both ELISAs and surface plasmon resonance assays, m44 reactivities in direct comparison to 2F5 and 4E10 reactivities were negative. The structural basis of lack of polyreactivity of m44 remains to be determined, and as well, it remains to be determined if m44-like antibodies will be difficult to elicit. Meffre et al. have suggested that most hMAbs with long CDR3s are depleted in the bone marrow (30
). Certainly, the lack of binding to self-antigens raises the hope that elicitation of m44 or m44-like antibodies may not be regulated by tolerance mechanisms in humans.
m44 does not compete significantly with 2F5 or 4E10 or with many of the mouse antibodies previously developed to map epitopes on gp41 (Table ) (5
) but competes strongly with the cluster IV conformation-dependent mouse MAb T3 (16
), probably due to steric interference. In our study, we found that T3 bound to 5HB and 6HB (Fig. ) in a way that is very similar to the binding of NC-1. However, the m44 epitope is distinct from the epitopes of NC-1 and T3, based on their different capacities for binding to 5HB, N36/C34-formed 6HB (Fig. ), and gp140 (Fig. ). To localize the m44 epitope on gp41, we constructed a gp41-Fc fusion protein and performed alanine-scanning mutagenesis. We characterized the wild-type gp41-Fc for antigenicity and immunogenicity prior to generation of a panel of mutants by alanine-scanning mutagenesis. Our preliminary data based on immunization of rabbits with purified gp41-Fc showed that the gp41-Fc is highly immunogenic. The titers of immune serum reached 102,400 for the gp41-Fc fusion protein and 5,120 for the recombinant gp14089.6
. In addition, purified rabbit IgG from one of the four immunized rabbits showed broad neutralization activity against several HIV isolates from different clades in a pseudovirus assay (data not shown), suggesting that gp41-Fc retains its antigenicity and immunogenicity. We emphasize that the alanine-scanning mutagenesis data presented in this work suggest but do not definitely prove that the m44 epitope is located in the C-HR and the upstream adjacent region. However, the agreement of the mutagenesis data with the data obtained by measuring binding to 5HB and 6HB (Fig. ) increases our confidence that the m44 epitope involves portions of the C-HR and adjacent regions. Only the crystal structure of m44 in complex with gp41 will definitely and precisely define its epitope.
Alanine-scanning mutagenesis focused on six residues in the loop region (M61, G62, L70, W78, V80, and W82) also caused significant decrease in the 2F5 binding. It was previously reported that alanine-scanning mutagenesis focused on residues M61, L70, W78, and W82 resulted in abolishment or significant reduction of viral entry (21
). Thus, mutations at these positions could change the overall structure of gp41, and it is difficult to evaluate their contribution to the antibody epitope. Mutations at G62 and V80 may also affect the gp41 structure, but this fact does not exclude the possibility that these residues are involved in m44 and 2F5 binding. In the C-HR region, five alanine mutations (M94, W96, M97, R101, and I103) also affected 2F5 binding. These residues could be involved in stabilizing the gp41 structure, and/or their mutations may affect the exposure of the 2F5 epitope in the MPER.
It has been previously reported that some antibodies exhibit large variations in neutralizing activities, depending on the assay used. For example, 4E10 exhibits much broader neutralizing activity when tested in an assay based on pseudovirus entry into a cell line than when tested in a PBMC-based assay (4
). We have also found that in some cases, 4E10 exhibits different activities when measured by the assays used in this study in comparison to previously reported results (e.g., it did not neutralize the 92UG029 isolate in our PBMC-based assay but did neutralize the same isolate in another PBMC-based assay, although relatively weakly [46
], and two other isolates [BR025 and G3] were better neutralized by 4E10 in another study [52
]). These differences could be related to the donor of PBMCs used or other details of the assays used, including the use of p24 as a marker of replication instead of RT activity, as in our assay. The RT activity is correlated with the number of virions in culture supernatant (14
), while p24 can also be released by infected cells as a free non-virion-associated protein. In addition, other details in the protocol, including whether the infection is spreading (e.g., if it includes at least two infection cycles) and whether the antibody is present for the subsequent cycles of infection could significantly affect the neutralizing activity of the antibody (for a recent analysis of possible causes of discrepancies, see reference 38
). Our data from three different assays based on spreading HIV-1 infection in PBMCs consistently showed that the neutralizing activity of m44 in such assays is on average comparable to or better than that of 4E10 or 2F5. However, the potency and breadth of neutralization exhibited by m44 in cell line/pseudovirus-based assays are significantly lower than those it exhibited in PBMC-based assays. Experiments with animal models would be required to evaluate the in vivo neutralizing efficacy of this antibody.
We noticed that m44 was weak in neutralizing HIV-1 primary isolates from clade A. In an attempt to find possible cause of this difference between clade A and other clades (B, C, and D), we performed an analysis of all HIV-1 Env sequences available in the HIV Sequence Database (http://www.hiv.lanl.gov/content/sequence/HIV/mainpage.html
) by calculating the frequency of most common residues in each position (data not shown). We specifically looked at the regions corresponding to the alanine-scanning mutagenesis we performed and noticed two unique differences between the clade A consensus sequence and two clade A primary isolates, 92UG037 and 92UG029, that were not neutralized by m44. One position with different residues is 629, for which the clade A residue is leucine, while the clade B, C, and D residues are methionine. The other one is position 633, for which the residue for clade A is lysine, while the residues for clades B, C, and D are arginine. We have no evidence to directly show that these changes are related to the lack of neutralization activity of m44 against clade A isolates. Further experiments with more clade A isolates are required.
The identification of this new antibody with unique properties could be attributed to the antibody library we used for panning, the antigens, and the CAP methodology we have developed and used here. The antibody library was previously constructed (Tom Evans, personal communication) by using bone marrow obtained from three long-term nonprogressors (A, H, and K) (15
) whose sera exhibited the broadest (against six primary isolates) (32
) and most potent (against JR-FL at a 1:40 dilution) HIV-1 neutralization among 37 HIV-infected individuals (49
), suggesting that m44 may exist in vivo. CAP methodology facilitates the identification of gp41-reactive antibodies in the context of native Env (48
). Further studies are needed to elucidate the mechanism of neutralization by the newly identified anti-gp41 antibody m44. One can speculate that it does bind to native Env and sterically prevents virus from binding to receptor-expressing cells, as previously proposed, as a mechanism of virus neutralization by antibodies (9
). Another possible mechanism is that this antibody binds structures important for fusion either before or after engagement of CD4 and coreceptors or both (3
). This unique antibody could help in the elucidation of the mechanisms of evasion of immune response by HIV and in the development of new candidate vaccine immunogens.