Polyclonal B Cell Responses to Conserved Neutralization Epitopes in a Subset of HIV-1-Infected Individuals

Broadly neutralizing antibodies (bnAbs) protect nonhuman primates from experimental simian-human immunodeficiency virus (SHIV) challenge (46–48, 61, 64, 81, 108) and are widely expected to be a key component of protective immunity conferred by an effective human immunodeficiency virus type 1 (HIV-1) vaccine (63). HIV-1 neutralization depends on the ability of Abs to recognize native envelope glycoprotein (Env) spikes, which consist of three surface gp120 subunits noncovalently linked to three membrane-spanning gp41 subunits (32, 70, 91). These spikes are compact structures protected by a heavy glycan shield that disfavors nAb recognition. Not all nAb epitopes are shielded, and in fact, the extent of shielding differs considerably among strains. Thus, some viruses are highly sensitive to neutralization by heterologous sera from HIV-1-infected individuals and are classified as having a tier 1 neutralization phenotype. However, most circulating strains are considerably less sensitive to heterologous neutralization and are classified as having a tier 2 phenotype (63).

A potent autologous nAb response against the early infecting virus is generally seen within the first year of HIV-1 infection (62, 83, 104, 113). Longitudinal studies have revealed that the virus and the host nAb response continually evolve (11, 73, 83, 86, 113). After several years of infection, an estimated 10 to 30% of individuals develop bnAb responses (4, 7, 28, 52, 57, 66, 72, 88, 89). Although these extraordinary responses have little impact in the setting of established infection, they could have a major impact on transmission if they could be induced by vaccines prior to virus exposure. Thus, the epitopes and biological processes that give rise to these exceptional bnAb responses are of considerable interest for rational vaccine design (114).

Until recently, only 4 monoclonal Abs (MAbs) (2G12, b12, 2F5, and 4E10), all from HIV-1 clade B-infected individuals, were known to potently neutralize genetically diverse isolates of HIV-1 (8, 13, 75, 107, 119). MAb 2G12 recognizes a tight cluster of glycans on the gp120 silent domain (14, 87, 93, 107). MAb b12 binds an epitope that overlaps the CD4 binding site (CD4bs) of gp120 (13). MAbs 2F5 and 4E10 (75, 119) recognize adjacent epitopes in the membrane-proximal external region (MPER) of gp41. Three of these MAbs (b12, 2G12, and 2F5) exhibit limited breadth and/or potency against non-clade B viruses, which account for the majority of infections worldwide (13). MAb 4E10 neutralizes more broadly, albeit usually at a lower magnitude. Other MAbs, including those directed to the V3 loop, usually exhibit a more limited profile of neutralization (51, 80), while MAbs directed to CD4-induced epitopes usually fail to neutralize tier 2 viruses altogether (8, 53, 115).

To date, vaccine immunogens targeting these MAb epitopes have not succeeded in eliciting bnAbs (42, 69, 82, 92). Consequently, there has been a resurgence of interest in isolating new bnMAbs, particularly from non-clade B-infected individuals, in the hope that these new epitopes will be more amenable to vaccine development (9, 12, 16, 18, 31, 95). Several new MAbs, most notably VRC01, VRC02, VRC03, PG9, PG16, and HJ16 (18, 111, 114, 118), each of which possesses potent activity against multiple genetic subtypes of the virus, have been identified.

These new MAbs were derived from donors of diverse geographic backgrounds infected with one of three different HIV-1 subtypes: A (MAbs PG9 and PG16), B (VRC01, VRC02, and VRC03), and C (HJ16). This emphasizes the importance of including globally diverse HIV-1-infected donor material in bnAb epitope discovery efforts (76, 102). MAbs VRC01 and HJ16 recognize epitopes that overlap the CD4 binding site (18, 114), while PG9 and PG16 recognize novel quaternary epitopes involving the V1V2 and V3 loops of gp120 (27, 111). Mutations that remove sequons at positions N156 and N160 in the V2 loop eliminate the binding of PG9 and PG16, as does enzymatic deglycosylation, suggesting that glycans are involved in these quaternary epitopes. Soluble CD4 (sCD4) ablates the ability of the latter two MAbs to bind trimers, suggesting that the conformational changes alter the quaternary epitope. These V2-V3 epitopes were somewhat surprising, considering that until recently, these loops were thought to protect underlying conserved epitopes and were considered too variable to be targets of broad neutralization. These findings suggest that despite the variability of the V-loops, conserved elements that can be targets of bnAb responses do exist.

Various technologies have been used to identify the epitopes of bnAbs (reviewed in references 1, 4, 5, 7, 21, 24–26, 37, 39–41, 57, 89, 101, and 117). An analysis of serum samples from approximately 200 individuals chronically infected with HIV-1 revealed that about one-third exhibit low titers of MPER nAbs, as detected by a sensitive HIV-2/HIV-1 MPER chimera assay in which the MPER is spontaneously exposed (3). However, in all but exceptional cases, these MPER nAbs do not neutralize circulating strains of HIV-1. In most cases, MPER nAbs were ascribed to 4E10-like epitopes, but 2F5-like nAbs have also been reported in rare cases (7, 39, 41, 89, 101).

Some studies suggest that bnAbs in chronic HIV-1 infection may be oligoclonal (2, 6, 23, 43, 95, 111). Indeed, recent work suggests that the responses are limited in specificity, often resembling PG9 and PG16 (111), are restricted in their IgG heavy and light chain usage, and require extensive somatic mutation (111). Furthermore, the observation that MAbs PG9 and PG16 largely recapitulate the neutralizing activity of the donor plasma provides evidence that broad neutralization can occur in some cases via a single specificity or multiple overlapping specificities (111). An overarching question from these studies concerns the number of nAb specificities that typically mediate broad plasma neutralization (72, 90, 95, 112). Here, to gain more insights into the frequency and titer of existing bnAb specificities and possibly to identify new bnAb targets, we analyzed the specificities of plasma samples from 9 HIV-1-infected subjects who exhibited high-titer bnAbs. Distinct bnAb specificities were present in the plasma from each subject, and in some cases, more than one specificity contributed to neutralization breadth.

A key question regarding broad plasma neutralization of HIV-1 is whether it is typically mediated by one (or a very few) broad and potent antibody or by the sum of the activities of a large number of narrowly focused neutralizing specificities. Studies exemplifying both of these scenarios have been reported (4, 95, 112). The present study represents an extensive collaborative effort of multiple laboratories employing a variety of mapping tools and provides evidence that two or more distinct antibody specificities can sometimes mediate breadth in a single subject (summarized in Fig. 10). Interestingly, bnAbs with the greatest magnitude and breadth targeted gp120, while those with less breadth recognized the MPER, and these two general specificities were largely mutually exclusive. In at least one case (C1-0219), two distinct bnAb specificities against gp120 were identified, one resembling PG9/PG16 and the other directed against the CD4bs. The fact that both specificities have been isolated as broadly neutralizing MAbs from this subject (10; J. R. Mascola, unpublished data) validates the presence of multiple plasma antibody specificities in a single individual, as well as the utility of mapping efforts to identify interesting epitopes for further study, including the isolation of novel broadly neutralizing MAbs. The presence of multiple broadly neutralizing antibodies circulating in plasma suggests that vaccines could, in theory, be designed to stimulate the immune response to productively develop multiple neutralizing antibody specificities that confer breadth.

We made use of a number of different methods to map the antibody specificities mediating broad neutralization in plasma from 9 chronically infected, antiretroviral-naïve subjects selected from a cohort of 308 patients infected mostly with subtypes A and C. The results of three different assays confirmed the presence of anti-CD4bs antibodies in plasma C1-0219. This included the use of ConC recombinant gp120 with a mutation at position 368 in the CD4bs that failed to deplete neutralizing activity, as well as the use of a YU2 resurfaced core protein that specifically binds CD4bs antibodies. A third method measured blue native gel band shifts and showed that C1-0219 efficiently shifted a JR-FL trimeric Env; this shift was lost when a protein containing the position 368 mutation was used, as with the prototype CD4bs MAbs, b12 and VRC03. None of the other 8 samples contained anti-CD4bs antibodies as determined by these methods, suggesting that broadly neutralizing CD4bs antibodies are induced relatively infrequently.

The presence of anti-MPER Abs was confirmed in at least 3 of the plasma samples in multiple assays. An HIV-2/HIV-1 MPER chimera detected high activity in 6 plasma samples, 3 of which were shown to mediate cross-neutralization via this specificity. These Abs were also detected in a fusion assay that showed that while they were a major component, they in fact represented only one-third to one-quarter of the total plasma neutralizing activity. A summary of the mapping data is shown in Fig. 10; it indicates that 8 of the 9 plasma samples contained at least 2 to 3 specificities that could be identified. However, in all cases, we could not account for all activity, suggesting that multiple nAb specificities of less breadth may also contribute to neutralizing activity.

Several other observations exemplify the complex nature of these plasma samples. (i) ConC gp120 adsorbed neutralizing activity against the matched isolate in all cases tested, yet only in 3 plasma samples was this truly cross-neutralizing. (ii) Plasma C1-0457 was effectively adsorbed using ConC gp120, which lacked a V2 loop, yet neutralization of JR-FL was knocked out by V2 loop mutations. (iii) Plasma C1-0763 was affected by mutants in V2 in a clade C background, but by V3 mutations and the N332A mutation on JR-FL. Part of this disparity is likely due to the backbone used to make these mutants, as clearly shown with N160 and N332 (Fig. 7 and 9), and thus, it is important to verify such findings in more than one background. A further caveat is that many Env point mutants are globally sensitive to distal specificities (Fig. 8 and 9). When this occurs, it suggests that the mutated residue is not entirely solvent exposed and may be involved in trimer packing. This is in fact a very common phenomenon that has also been reported in several recent studies (9, 74, 78, 100, 101, 112). Our observation that a relatively larger fraction of gp41 MPER mutants than of gp120 mutants cause global sensitivity (compare Fig. 5B and C to Fig. 8 and 9) is consistent with the idea that the MPER is partially sequestered and is involved in protein-protein or membrane associations, so that substitutions often lead to global effects on folding. In contrast, most of our gp120 mutants are likely to be surface exposed and therefore to minimally affect trimer packing. A previous bioinformatics study identified a series of residues that partitioned with broad neutralization, i.e., polymorphisms at these positions were associated with the presence or absence of broad neutralization (34). Several of these residues corresponded to the bridging sheet and coreceptor binding site and were found to cause global sensitivity when mutated. Therefore, it is possible that the development of broad neutralization might relate to trimer “compactness.”

In contrast to the findings of our study, Walker and colleagues found that the neutralization activity in 19 elite plasma samples was most often dominated by a single specificity and that MPER activity was minimal (112). It is not clear why the findings were different in the two studies. The Walker study used plasma from 14 subjects who ranked among the top 1% of neutralizers and from 5 additional subjects who ranked among the top 5% in a cohort of 1,798 subjects in International AIDS Vaccine Initiative (IAVI) protocol G (112). We used plasma from 9 subjects who ranked in the top 3% of a cohort of 308 subjects, including 3 subjects who were in the top 1%. The two studies also used different viruses and assay techniques. Previous studies have pointed out that detection of antibodies targeted to gp41 can depend on the target cells used in the assay. MAb 4E10 neutralization was found to be more sensitive in the TZM-bl assay (8), whereas other neutralizing antibodies directed to gp41 were better detected in the PBMC-based assay (30). These differences highlight the importance of examining multiple cohorts and using different approaches to capture the full spectrum of activities and to identify all possible targets on HIV-1 Env.

Various studies have hinted at the presence of unusual neutralizing specificities, some of which may be entirely novel. One report suggested that Abs overlapping the CCR5 binding site and sensitive to an I420R mutation in the bridging sheet can mediate broad neutralization (58). Although there is scant evidence for 2G12-like bnAbs from most mapping studies to date (4), recent studies revealed that a gp120 N332 mutation can sometimes ablate plasma neutralization, suggesting bnAb specificities that overlap the 2G12 epitope and may reside on the outer domain of gp120 (36, 38, 76, 109, 110).

In a recent review of mapping efforts to date, CD4bs activity accounted for a substantial proportion of the neutralizing activity in 12 of 43 plasma samples, whereas substantial MPER activity was present in 3 (4). Importantly, more than 50% of bnAb activity in 28 of the plasma samples could not be mapped. In approximately 50% of cases, gp120 failed to adsorb neutralizing activity (36, 37, 41). This suggests that much of the unmapped activity may target quaternary, trimer-dependent epitopes (35). Indeed, two recent studies (72, 112) have reported nAbs that recognize epitopes similar but not identical to those of MAbs PG9 and PG16. The results of another recent study suggested that the early nAbs preferentially target the CD4bs, while more broadly neutralizing Abs cannot be adsorbed by gp120 (i.e., PG-like or possibly MPER) and appear to be associated with CD4 T cell activation (66) and CD4 T cell decline upon infection (36).

Although MPER nAbs have been detected frequently in multiple studies, they only rarely exhibit neutralization breadth (4, 7, 37, 39, 41, 101). This may be because MPER nAbs tend to lack the potency of nAbs that target gp120. Thus, MPER neutralization may feature more prominently in plasma samples that exhibit somewhat weaker overall neutralization titers. It may therefore not be a coincidence that MPER accounted for little or none of the neutralization in our level 3 neutralizers (C1-0219, C1-0763, and C1-0457; top 1%) (findings similar to those of the IAVI study) but featured more prominently in at least 3 of our 6 level 2 neutralizers. The epitope of the MPER nAbs in these plasma samples resembled that of MAb 4E10, consistent with most previous studies (4). Despite its breadth, the lack of potency of 4E10 could explain why similar nAbs do not contribute significantly in most broadly neutralizing plasma samples. One possible reason why MPER nAbs are reported so commonly in mapping studies may simply relate to the fact that this is the only epitope to which we can measure neutralization directly, without the contribution of other nAbs that are present. With other epitopes, in contrast, mutant viruses, adsorption, and other techniques are needed in order to identify nAb activity against discontinuous conformational epitopes. Thus, we can detect exceedingly low titers of MPER nAbs, whereas at least half of the total neutralization activity must be affected to allow reliable measurement of other specificities using current technologies. In the case of plasma samples C1-0269, C1-0534, and C1-0536, this observation was consistent with MPER nAbs playing a significant role in the neutralization of most, but not all, viruses (Fig. 3C). Thus, the weak MPER activity detected in plasma samples C1-0175, C1-0219, and C1-0763 (Fig. 3A and B) is overshadowed by other nAbs targeted to gp120, as was suggested by the inability of MPER peptides to deplete neutralization of these plasma samples against most isolates tested (Fig. 3C and data not shown). An exception is C1-0175 neutralization of the Du156.12 isolate, which was largely explained by MPER nAbs (Fig. 3C). Thus, we can infer that other specificities contribute to the breadth of all 9 plasma samples, with MPER nAbs playing a supporting role in some, but not all, cases and that, in many cases, breadth is achieved by multiple specificities. Since we have identified several subjects whose samples naturally elicited MPER-neutralizing antibodies with some breadth, eliciting neutralizing MPER antibodies is a reasonable goal for HIV-1 vaccine design in concert with the elicitation of multiple broadly neutralizing specificities.

A substantial portion of ConC-neutralizing activity in the nine plasma samples was adsorbed by gp120 (Table 2), but only in the three level 3 neutralizers was cross-neutralizing activity against other viruses adsorbed. In the case of level 3 plasma sample C1-0763, V2-reactive activity was adsorbed by gp120, and the precise residues in V2 were mapped and shown to involve N160 and K169 (Fig. 6), which are also required by PG9 and PG16. Plasma C1-0219 was also found to contain V2 loop-reactive activity. This plasma sample was also sensitive to the N160 and K169 mutations, suggesting epitopes related to those of the PG MAbs. The frequency of this specificity is consistent with that in the IAVI study, in which 5 of 19 plasma samples were N160 dependent. It may not be a coincidence that these two plasma samples were derived from clade A infections, as were the PG MAbs, raising the possibility that such specificities evolve more often during clade A infections. The potency of the V2 specificities is exemplified by the dramatic effects of N160 mutants on neutralization of the Q23.17 and CAP45 viruses by plasma samples C1-0219 and C1-0763. In each case, this V2 activity accounts for most, if not all, of the neutralizing activity that imparts extremely high ID₅₀ titers, close to 10,000 (Fig. 2). It will be worthwhile to characterize this powerful activity more precisely in the future, especially if it cross-reacts with viruses from other clades.

Two recent plasma mapping studies further suggest that “PG-like” plasma nAbs may not precisely replicate the activities of the prototype MAbs (72, 112). Similarly, 2909-like quaternary nAbs isolated from SHIV-infected macaques exhibited specificities slightly different from that of the 2909 prototype (84). Taken together, these findings suggest that V2 or “PG-like” nAbs found in broad plasma samples can have various V-loop residue dependencies that may overlap but do not necessarily precisely replicate the known specificities. This is perhaps not surprising in view of the substantial variability of the V2 loop. The V2 dependencies of our three V2 specificity-containing plasma samples on ConC and JR-FL backgrounds are each substantially different, illustrating this point. One of these three plasma samples (C1-0457) was susceptible to a mutation at I165 of the JR-FL V2 loop (see Fig. S3 in the supplemental material) that was previously found to be a target of 5 plasma samples in the IAVI study, but distinct from PG-like nAbs, which are not sensitive to this mutation (112). This residue has been shown to regulate neutralization by trimer-specific MAbs 2909, 2.3E, and 2.2G (35, 84). Although these nAbs are strain specific, they recognize quaternary epitopes involving the V2 and V3 loops.

Three plasma samples (C1-0440, C1-0457, and C1-0763) were sensitive to the glycan at N332, depending on the virus strain. In at least four previous plasma mapping studies, this sensitivity has also been observed (76, 36, 112). One group has now isolated broadly neutralizing MAbs with this specificity via high-throughput screening of donor B cells (110). In our study, the neutralizing activity of at least one of our plasma samples, C1-0763, was broadly adsorbed by gp120 (Table 2), as observed for most of the previously reported N332-sensitive plasma samples (76, 112). However, there is a complexity of specificities in plasma C1-0763 (V2 linear, V3, MPER, N332A). Interestingly, unlike the N332A-dependent activity observed previously (112), plasma C1-0763 partially competed with biotinylated 2G12 for gp120 binding (Fig. 9B). It is worth noting that the N332 and N339 glycans are part of the alpha-2 helix, a known neutralizing epitope for clade C plasma samples (71, 85). Thus, it appears possible that, as with the V2 activity, broad neutralizing responses can ultimately develop from initially highly type specific responses, with time and sustained virus replication. Thus, N332A affects the neutralization by C1-0763 of clade B viruses (JR-FL and TRO.11) but not of the other viruses. This may be because this plasma is able to neutralize these other viruses via the V2 loop specificities discussed above, which may not cross-react with the clade B viruses.

C1-0219 is a clear example of multiple specificities in plasma mediating breadth. This plasma sample contained prominent neutralization activity directed to both the CD4bs and the V1V2 loops. Despite the breadth of the CD4bs nAbs, this activity did not explain all of the neutralizing activity against any of the isolates tested (Table 2). Indeed, the relative contribution of the CD4bs nAbs to the neutralization of a given virus appeared to depend on the overall sensitivity of the virus to the plasma sample. Thus, when C1-0219 exhibited modest neutralization titers (e.g., Du156.12 and SC422661.8), CD4bs nAbs played a major role (Table 2; Fig. 5A). However, when C1-0219 exhibited high neutralization titers, the contribution of CD4bs nAbs was less significant. Thus, while the partial adsorption seen with gp120 in this plasma was likely due to the D368R-sensitive CD4bs nAbs, the remaining activity was dependent on quaternary structures similar to those of the PG MAbs, which were sensitive to N160. Indeed, MAbs to both targets have now been isolated from this individual. These include antibodies CH01 to CH04, which map to the V2 region and are trimer dependent but are distinct from PG9/P16 (10), and three potent anti-CD4bs MAbs (113a) isolated using the RSC3 protein, as was done for VRC01 (114). Interestingly, isolates, such as CAP45.G3, that were only weakly neutralized by plasma CD4bs antibodies (Table 2) were very sensitive to the anti-V2 MAbs CH01 to CH04 (10), accounting for the potency of the plasma against this virus. Similarly, Du156.12 was sensitive to MAbs CH01 to -04 but not to the new anti-CD4bs MAbs. In contrast, isolates such as ZM109.4 were sensitive only to the anti-CD4bs MAbs, not to the anti-V2 MAbs isolated from C1-0219. These data confirm, through the use of monoclonal agents, that neutralization breadth can be mediated by two distinct antibody specificities in a single subject.

We were not able to determine the specificity of the cross-neutralizing activity in all the plasma samples. A case in point is that plasma C8-0258 was not significantly mapped by any method. This could be due to complexity of nAb specificities in this plasma or to our incomplete array of mapping tools. It is noteworthy that this individual was the only one infected with a clade B virus, and while the sample did show cross-clade neutralization (Fig. 2), it was weak against clade A. It is possible that breadth in this individual may be due to multiple narrow specificities or that we have overlooked CD4bs nAbs that are not D368R dependent (95), such as MAb HJ16 (18). Additionally, our study focuses on IgG-mediated neutralizing activities in the plasma and not on other antibody isotypes. We have found that purified IgA from the plasma of HIV-1⁺ subjects can neutralize (data not shown), which indicates that IgA, in addition to IgG, may be responsible for some aspects of virus inhibition in both broad neutralizers and other HIV-1-infected subjects.

Overall, our results support the idea that multiple distinct bnAb responses can evolve in some people. In cases where neutralization appears to be more clonal, this may not rule out multiple specificities; it means only that one specificity dominates neutralization. Only with time do bnAb responses finally develop in some subjects, and there is evidence that ongoing viral replication contributes to the development of these bnAbs (36, 66, 89). What remains unknown is why only certain individuals are capable of mounting a bnAb response. Is there something unique about the virus and the antigenic structure of its Env that favors antibody responses to conserved epitopes? The fact that the epitopes are present on most viruses would argue against this, unless regions outside the epitope can control immunogenicity. Another possible explanation is that the response is differentially controlled by host genetics. In this regard, our results suggest that the B cell response in certain individuals is privileged to target multiple unrelated bnAb epitopes that most likely require different IgG heavy and light chain gene usage. Identification of the biologic process governing these selective responses could yield valuable insights for vaccine design.

The significance of our findings also relates to the discovery of new neutralizing MAbs. In some instances, knowledge of the predominant nAb specificities may directly assist in MAb isolation. For example, in the present study, B cell receptor ligands consisting of parent CAP45.G3 gp120 and its K169E mutant might function effectively as positive and negative reagents for the selection of V2-neutralizing MAbs from the B cells of donor C1-0763 in a flow cytometry screening protocol similar to that reported recently (38, 114). Such rational approaches may accelerate the isolation of new bnMAbs, providing new tools for vaccine design.

This work was supported by the Bill and Melinda Gates Foundation Collaboration for AIDS Vaccine Discovery (CAVD-VIMC grant 38619), the National Institutes of Health (NIH/NIAID/DAIDS), and the Center for HIV/AIDS Vaccine Immunology (AI067854-05). Additional funding was provided by grants AI58C1-0763 (NIAID, NIH; to J.M.B.), Torrey Pines Institute's AIDS and Infectious Disease Science Center (to J.M.B.), and the South African HIV/AIDS Research and Innovation Platform (SHARP) of the Department of Science and Technology (DST) (to L.M. and E.S.G.). This work was supported in part by the intramural research program of the Vaccine Research Center, National Institute of Allergy and Infectious Diseases, NIH.

We are indebted to the CHAVI 001 and CHAVI 008 patients and clinical teams for numerous contributions to this project. We also thank Kelly Soderberg for CHAVI program management, Saidi H. Kapiga and Noel E. Sam for oversight of CHAVI clinical studies at the KCMC, the Duke Immunology Clinical Laboratory for anti-cardiolipin assays, Judith T. Lucas, R. Glenn Overman, Maphuti Madiga, and Robert Parks for technical expertise, and Xinzhen Yang for helpful discussions.