Discussion
In natural infection, broadly neutralizing antibodies (bNAbs) are generated too late to halt early infection events, and the effectiveness of humoral immunity is further hampered by virus escape in response to developing immune pressure. However, the generation of bNAbs via preventive vaccination could possibly block HIV acquisition. Thus, much effort is being placed on defining optimal immunogens to elicit effective bNAb responses. As the number of identified T/F env genes continues to grow, a detailed understanding of whether T/F strains may share ― within or across clades ― certain global features affecting neutralization sensitivity will underpin discovery of suitable neutralization targets and, thus, development of a preventive vaccine inducing effective virus neutralization.
The question whether combination of broadly reactive antibodies directed against distinct epitopes may have synergistic, additive or antagonistic effects on neutralization potency has not been adequately addressed. Thus, in this study, a limited scope assessment of such effects on the neutralization of five Env-pseudotyped viruses and six T/F ICMs by six human broadly neutralizing antibodies was performed. All but one of the HIV-1 strains have been ascribed a Tier 2 neutralization phenotype in TZM-bl/PV assays; only SF162 possesses a Tier 1A phenotype (Neutralizing Antibody Resources Tools, at
www.hiv.lanl.gov) [
23]). To our knowledge, this is the first study to test human bNAbs individually and in pair-wise combinations against a panel of clade B T/F viruses; among them, three T/F virus strains, REJO.c, RHPA.c and THRO.c, were juxtaposed with pseudoviruses with early infection
env genes derived from the same patients, respectively.
Interestingly, the three pseudoviruses and T/F IMCs sharing the nearly identical
env sequence, i.e. REJO, RHPA and THRO (with two, two, and one amino acid differences, respectively, between early infection and T/F
envs; Additional file
1: Figure S1), displayed overall similar patterns of neutralization by single antibodies, however, IC50 values for IMC were generally higher, or not reached (i.e. IC50 > 66 μg/ml) (Figure
1). For example, the Env proteins in REJO PV and IMC have no substitutions and a shared insertion (Ile) in the b12 epitope, but differ from one another in aa 255 (Ala vs. Val), 2 positions upstream of Ser
257-Thr
258 residues which are part of the b12 epitope (Additional file
1: Figure S1); this variation may contribute to the very different IC50s observed for these viruses (2.58 ug/mL vs >66 ug/mL). Similarly, Env proteins in RHPA PV and IMC differ from one another immediately following the LTRGD
437 portion of the epitope, and resulted in different IC50 values (0.15 vs 1.08 ug/mL). However, Env in THRO PV and IMC were identical to one another in and around the b12 epitope but still displayed different IC50 (0.83 vs 12.87 ug/mL, respectively). Thus, Env sequence alone cannot fully explain the sensitivity of a specific viruses to a given antibody, nor differences between pseudoviruses and IMCs sharing the same or highly similar
env sequence. Of note, previous studies also reported different sensitivity (IC50) of IMC and pseudoviruses with identical
env genes to single antibody neutralization, regardless of virus clade [
24],[
25]. In both reports, the pseudoviruses were found to be less sensitive than IMCs to specific mAb neutralization [
24],[
25]. However, because of small sample numbers in each study it cannot be ruled out that these results are
env-strain specific rather than PV versus IMC specific. In our study including early-infection
env PV and T/F IMC, the IMCs generally showed higher IC50 values in single NAb neutralization assays (Figure
1). However, importantly, antibody pair synergy was observed at a higher proportion in IMC than in pseudovirus assays (Figure
4), suggesting that IMCs were as susceptible to synergistic antibody activity as pseudoviruses. Moreover, substantial similarity between IMCs and PV emerged when the distribution of all individual CI values within each group was compared (Figure
5), and no significant differences between IMCs and PV were documented.
Differences in neutralization sensitivity between IMCs and PV could be due to the two genetically distinct proviral backgrounds since IMCs encompass a complete autologous viral genome from which
env is expressed in
cis under the control of the autologous LTR [
2],[
4]. In contrast, pseudoviruses are derived by complementing a common
env-defective backbone with heterologous
env genes expressed in
trans[
2],[
4]. Not surprisingly, other studies have reported that different ratios of backbone and
env-plasmids transfected in host cells were found to give rise to pseudovirus particles endowed with different envelope features, such as the proportion of
env protein cleavage and the level of gp120 surface expression [
26]; such changes in envelope features were found to affect pseudoviruses infectivity, and, possibly, antibody reactivity [
26]. Indeed, host cells are known to impact biochemical and structural features of virus particles, e.g. in terms of protein processing, folding and glycosylation patterns [
25]. Viruses cultured in PBMC or in primary cells were found to be more resistant to antibody neutralization than those obtained from laboratory-adapted cell lines, for example due to a different glycosylation pattern shielding key epitopes and preventing antibody neutralization [
24]. However, previous studies investigating structural changes among virus structure or protein composition, failed to associate differences observed in IC50 values or infectivity with any well-defined structural or biochemical feature [
26]. In our study, both pseudoviruses and T/F IMCs were produced in 293 T cells, therefore diversity in antibody sensitivity cannot be ascribed here solely to the effect of host cells. We also strove to minimize other possible sources of variability in neutralization result by choosing a standardized, validated method, the TZM-bl assay [
9], to perform all assays. Prior to standardization, unsatisfactory assay equivalency among laboratories had been observed even when reagent batches were shared [
27].
As was expected, no single antibody neutralized all virus isolates, neither in the pseudovirus nor the T/F IMC group (Figure
1). The b12 antibody, targeting a conserved epitope within the CD4 binding site, achieved 50% neutralization on most T/F strains (5/6), and all PV strains (5/5). The 2G12 antibody was poorly reactive against 9 out of 11 viruses, neutralizing only SF162 pseudovirus and CH058 T/F IMC; this finding is in concordance with the absence of critical amino acid residues (N295, N332, S334, N339) of the 2G12 epitope, and glycosylation, in the resistant strains, respectively. PG16 was more reactive than PG9, neutralizing nearly half of virus strains in both panels (Figure
1). Sensitive virus strains in the study do share N156 and N160 glycosylation sites, which are crucial for PG9/PG16 binding (Additional file
1: Figure S1). Conversely, SF162 PV Env, and CH058.c and CH040.c IMC, lacking N160, showed resistance to these mAbs (Figures
1 and
2). THRO PV and IMC were resistant to PG9/PG16 mAbs, albeit the presence of both N156 and N160 glycosylation sites, possibly due to a K178R mutation in the epitope (Additional file
1: Figure S1).
MPER antibodies 4E10 and 2F5 each neutralized 3/6 T/F IMC, and 5/5 (4E10) and 4/5 (2F5) PV, respectively (Figure
2). Of note, IC50 values for both bNAbs differed seven- to ten-fold between patient-matched
Env proteins expressed in either the PV or IMC context (IC50 values higher or not reached in IMC) despite identical MPER sequence. The MPER domain in gp41 is usually weakly recognized by neutralizing antibodies in native virus particles [
28],[
29]. Since 2F5 and 4E10 mostly recognize MPER epitopes when gp41 is conformed in pre-hairpin intermediate [
28]-[
30], the higher IC50 values obtained against T/F REJO, RHPA and THRO strains may be explained by a more compact and stable conformation of Env expressed in
cis from T/F IMCs as compared to their respective pseudovirus counterparts [
31], a feature that could increase binding restriction and result in poor accessibility to antibodies [
32].
Since humoral responses to pathogens are usually polyclonal, synergy and antagonism between antibodies may naturally occur. Due to their dimensions, neutralizing antibodies do not cluster on one unique Env molecule within the trimer, but are likely to bind distinct monomers within a single ― or within two proximal ― trimer spikes [
30],[
33]. Electron microscopy and mathematical modeling have not yet determined the spike number required to carry out infection successfully, however, HIV particles are studded with only a low number of spikes (between 4–45), sparsely distributed on the envelope membrane [
34]. Therefore, synergy and antagonism would result from the interaction of two or more antibodies with a population of molecular targets, where each single virus particle can carry a number of trimer spikes as well as Env dimers or monomers [
33]-[
35], and it cannot be readily assumed that two antibodies would have synergistic or antagonistic effects because they were bound to the same Env molecule. Due to the inherit variability of the envelope protein, a relevant question is whether in the
in vivo context the presence of prolonged Nabs activity may play a role in modulating evolution of the disease. Although it is worth mentioning that Nabs have been associated with control of the disease in Long-Term Non-Progressor subjects, where an equilibrium has been established between virus and host, we cannot exclude that over time mutations occurring within the envelope can affect neutralizing activity thus resulting in an antagonism rather then synergy. This latter situation could occur when the patients’ clinical status changes to rapid progressors, thus loosing the previously established equilibrium. In this regard, the different density level of envelope spike could play a crucial role as well. The low density of envelope spikes, a distinguishing feature when compared with viruses to which protective neutralizing antibody responses are consistently raised, directly impedes bivalent binding by IgG antibodies. The result is a minimization of avidity, normally used by antibodies to achieve high affinity binding and potent neutralization, thereby expanding the range of mutations that allow HIV to evade antibodies. Understanding limitations to avidity may be essential to establish whether specific antibodies combination can differentially modulate their activity, in particular upon variability on the density of envelope spike during the course of chronic infection.
Not all antibody combinations were tested against all PV and T/F IMC strains since not every antibody had reached IC50 individually. In all cases in which antibody combinations were assayed (n = 33 for PV; n = 17 for T/F IMC), inhibition levels of at least 50% were reached. Remarkably, synergy was observed in 42 out of 50 assays. CI values indicative of additivity were seen in 5 cases (one with T/F ICMs; four with pseudovirus assays). Low level antagonism was observed in three assays (one with T/F IMC, two with PV, respectively) and involved antibody combinations 2F5 + 4E10, and PG9 + PG16 which target related epitopes (Figures
2,
3,
4, Table
2). Nearly all antibody pairs achieved synergistic inhibition of both pseudoviruses and T/F IMCs, respectively, with a few and possibly virus-strain specific exceptions (Table
2). Findings from the bNAb pair assays, thus, suggest that synergy usually occurs when antibodies targeting different
env domains were involved (e.g. 4E10 ― or 2F5 ― with b12 or with PG16). In other words, association of two suitable antibodies could induce a favourable conformational change, when binding the same monomer in a trimer or even when binding different monomers, therefore creating favourable conditions for synergic activity. From this point of view, synergy between b12 and MPER-targeting antibodies is not surprising, because CD4 binding takes usually place
before gp41 exposure and promotes Env refolding into the intermediate, extended conformation [
29],[
36]. Similarly, b12 binding could enhance accessibility of 4E10 (or 2F5) antibodies to MPER domain by inducing suitable conformational changes involving both gp120 and gp41 glycoproteins [
30],[
37],[
38].
Combinations of PG9 + PG16 and the 2F5 + 4E10 antibodies, with members of each pair targeting overlapping or adjacent epitopes, were tested as controls. Surprisingly, their mean CI values ranged from synergy to antagonism depending on virus isolates (Table
2). In some cases all 12 individual CI values for each bNAb pair/virus combination (Figure
4) fell into only one category (e.g. for 4E10 + 2F5 vs RHPA PV), while in others the individual CI values differed over a wide range depending on NAb concentrations (e.g. synergy for 4E10 + 2F5 versus REJO PV). While antagonism observed with the PG9 + PG16 pair may be explained by steric hindrance or target competition, since both of them bind V1-V2 loops of gp120 or quaternary structures exposed on the top of the gp120 trimer [
39],[
40], it is noteworthy that antagonism was seen in only one out of four tested virus strains. The PG9/PG16 binding site determinants on gp120, and Env trimers, are not fully resolved; the N160 glycosylation site, shared by most HIV isolates, is one unique feature precisely attributed to both binding sites, and its mutations are known to affect PG9-PG16 neutralization [
41]. All viruses tested in the study share N160 glycosylation site within their gp120 sequences (see Additional file
1: Figure S1), however they differ in the amino acid positions in the adjacent contact sites and displayed different neutralization sensitivity to PG9 and PG16, possibly because not all gp120 molecules and Env trimers could effectively bind PG16 and PG9. In the case where PG9 and PG16 neutralize individually, but their activity is antagonistically affected in combination it is possible that the PG9 and PG16 pair compete for the same binding site when both present, and thus causing antagonism [
34] .
The 2F5 and 4E10 antibodies recognize two contiguous, linear epitopes along MPER, which are especially ― but not exclusively ― accessible in pre-fusion gp41, i.e. the pre-hairpin intermediate. Differently from PG9-PG16, the two MPER epitopes are close, but not overlapping; moreover, 2F5 and 4E10 antibodies target epitopes which are made accessible on different conformations of gp41 [
36] therefore, their binding may not be competitive under some conditions [
42], and in an Env strain dependent manner. Due to the nature of the epitope conformation and to MPER refolding, the 4E10 epitope may be accessible on native gp41 and throughout gp41 refolding, while the 2F5 epitope is accessible only during early phases of hairpin formation [
36]. In addition, mutations involving the CDR-H3 region in 2F5 and 4E10 are known to reduce their interaction with lipids without altering epitope binding, but make these antibodies non-neutralizing [
28]. The notion of the better and more prolonged accessibility of the 4E10 epitope versus the 2F5 epitope was supported by studies in which 4E10 showed a broader neutralizing activity than 2F5.
Due to misfolding, symmetry within Env trimers may be disturbed, making MPER epitopes ― as well as any other Env epitope ― more easily accessible to antibodies [
43]. Furthermore, 2F5 antibodies representing different isotypes (IgA2 and IgG1) displayed synergic neutralizing activity even though they were directed against the same 2F5 epitope , probably by accessing and blocking 2F5 epitopes on distinct gp41 molecules within or between trimers [
44]. Hence, the 4E10-2F5 range of synergy-additivity-antagonism observed in the study may result from binding to individual monomers in single or multiple trimers as well as from strong membrane interactions, with unexpected effects on virus infectivity [
12],[
17],[
28],[
36],[
45]-[
47].
In future work it will be of interest to explore whether antibodies that individually are poorly inhibitory and fail to reach IC50 could nevertheless be more potent in combination, due to synergic effects. To further validate the findings from our study that bNAb synergy may be a rather ubiquitous occurrence and thus may be harnessed to inhibit HIV-1 infection, it would be ideal to test a larger panel of circulating HIV-1 strains against additional bNAbs from the ever-growing reservoir. The recently described multi-clade Global Panel of 12 Env clones from the Neutralization Serotype Discovery Project (NSDP) was shown to represent the continuum of neutralization phenotypes observed for globally circulating HIV-1 strains [
7]. Thus, testing for bNAb synergy against the Env Global Panel would be highly relevant and timely to gain a deeper understanding of the prevalence and potential of synergic effects on neutralization.
In conclusion, we submit that immune strategies eliciting synergic antibody responses have the potential to augment inhibition of transmission and early virus infection, provided that polyclonal responses are employed and that their synergic potential can be fully exploited. Although many open questions remain regarding bNAb synergy, exploiting synergy between more easily inducible individual broadly neutralizing antibodies with more limited potency holds promise for effective vaccination strategies.