Antibody polyspecificity and neutralization of HIV-1: A hypothesis

Abstract

HIV-1 has evolved many ways to evade protective host immune responses, thus creating a number of problems for HIV vaccine developers. In particular, durable, broadly specific neutralizing antibodies to HIV-1 have proved difficult to induce with current HIV-1 vaccine candidates. The recent observation that some broadly neutralizing anti-HIV-1 envelope monoclonal antibodies have polyspecific reactivities to host antigens have raised the hypothesis that one reason antibodies against some of the conserved HIV-1 envelope trimer neutralizing epitopes are not routinely made may be down-regulation of some specificities of anti-HIV-1 antibody producing B cells by host B cell tolerance mechanisms.

1. Introduction

One of the problems for HIV-1 vaccine development is to determine why it is so difficult to induce broadly reactive neutralizing antibodies to the HIV-1 envelope, and to devise ways to overcome this problem. Several escape mechanisms of HIV-1 from neutralizing antibodies have been described including emergence of a glycan shield over a large portion of gp120 [26,27,63,67], entropic barrier and conformational shifting [7,26], and high Env mutation rate [44,63]. This review explores the hypothesis that in addition to these established mechanisms of HIV-1 immune evasion, some species of broadly neutralizing antibodies may not be readily made because they are subjected to negative B cell immunoregulatory control.

1.1. Anti-HIV-1 broadly neutralizing human mabs

In spite of the difficulty in inducing broadly reactive neutralizing antibodies with immunogens, one of the positive findings for HIV-1 vaccine development has been the isolation of several rare human monoclonal antibodies (Mabs) that broadly neutralize HIV-1 [6,12,52,56]. Mab IgG1b12 binds to gp120 at the CD4 binding site [45], Mab 2G12 binds to Manα-1,2Man oligomannose residues on the immunologically silent face of gp120 [8], and Mabs 2F5 and 4E10 mabs bind to linear epitopes (2F5-ELDKWAS; 4E10-WFNITNW) within the tryptophan-rich (aa 660–683) membrane proximal external region (MPER) of HIV-1 gp41 [36,52].

To explore the possibility that some broadly reactive neutralizing antibodies might not be routinely made because they are derived from B cell populations that frequently make polyspecific antibodies, we have tested these four rare broadly reactive neutralizing human Mabs in clinical autoantibody assays [19]. We found that Mab IgG1b12 reacted with dsDNA, centromere B and histones, while Mab 2G12 was not autoreactive in these assays. The two MPER Mabs, 2F5 and 4E10, both showed polyspecific responses; 2F5 reacted with cardiolipin, Ro, centromere B, and histones, and 4E10 reacted with cardiolipin and Ro [19]. Mab 4E10 also reacted with other phospholipids including phosphatidylserine and phosphatidylethanolamine, and had lupus anticoagulant activity, while mab 2F5 did not have these reactivities. Moreover, Mab 4E10 reacted with cardiolipin with nM apparent affinity (K_d), while 2F5 reacted much weaker to cardiolipin with μM apparent affinity [19].

To determine how similar 2F5 and 4E10 were to autoantibodies from autoimmune patients, we have compared 2F5 and 4E10 Mabs with two well characterized anti-cardiolipin Mabs, IS4 and IS6, derived from a primary anti-phospholipid syndrome (APS) patient [69,70]. We found that IS4 and 4E10 antibodies were similar in surface plasmon reasonance binding assays, and had nM apparent affinities for lamellar cardiolipin, but did not require β-2-glycoprotein-1 for binding to cardiolipin. In contrast, mabs 2F5 and IS6 had μM apparent affinities for lamellar cardiolipin, and binding to cardiolipin was enhanced by β-2-glycoprotein-1 (Alam, M, Haynes, BF, Chen, P., et al. manuscript in preparation).

Anti-cardiolipin antibodies from APS patients have similar CDR3 regions: they are long and have characteristically spaced CDR3 arginine residues [43]. Analysis of the CDR3 region of 2F5 demonstrated three characteristic cardiolipin antibody CDR3 arginines. In contrast, 4E10 does not have the characteristic CDR3 arginines of anti-cardiolipin antibodies (Alam, M, Gewirth, D, Kepler, T. and Haynes, B, manuscript in preparation). Finally, the original 2F5 and 4E10 Mabs as well as IS4 and IS6 are IgG3 isotype, suggesting they may be derived from a B cell population with restricted class switching such as the transitional-1 or marginal zone B cell pools [33,54,57,65]. Both 2F5 and 4E10 are somatically mutated indicating their origin from an antigen driven process.

While Mab 2G12 was non-reactive in clinical autoantibody assays [19], Scanlan et al. have suggested the rarity of 2G12 antibodies in HIV+ patients and immunized subjects may relate to two factors [46]. First, the HIV-1 carbohydrates to which 2G12 binds may be recognized as “self” carbohydrates since HIV-1 has no glycosylation machinery of its own [46]. A second reason HIV-1 virion carbohydrates may be poorly immunogenic is antigen microheterogenicity, ie a single protein sequence would be expected to express multiple carbohydrates forms leading to dilution of an immune response [46].

1.2. Relevance of polyspecificity to neutralization capacity of anti-HIV-1 antibodies

While the significance of the relevance of the polyspecificity of 2F5 and 4E10 Mabs is not yet known, that some broadly neutralizing human Mabs are polyspecific antibodies with characteristics of antibodies that are frequently deleted during B cell development [35] suggests two related hypotheses.

First, some species of broadly reactive neutralizing antibodies may not be routinely made because the B cells that make them are polyreactive and are subjected to B cell negative selection, either during the establishment of the B cell repertoire or during the B cell response to antigen. Second, animals or humans with autoimmune disease syndromes and defects in B cell deletion, anergy induction, or receptor editing may be able to make antibodies that broadly neutralize HIV-1, either after immunization or after HIV-1 infection.

These hypotheses do not suggest that all anti-cardiolipin antibodies, all anti-phospholipid antibodies, or sera from HIV-1 uninfected patients with autoimmune disease will neutralize HIV-1. We have tested anti-phospholipid syndrome-derived antibodies IS4 and IS6 for HIV-1 neutralization, and as expected, these anti-cardiolipin antibodies from HIV-1 uninfected patients do not neutralize HIV-1 (Haynes B.F., Chen, P., et al unpublished observations).

Rather, the hypotheses suggest that some autoimmune patients upon infection with HIV-1 or after HIV-1 Env immunization, may be able to make specificities of anti-HIV-1 antibodies that normally would not be made.

1.3. Incidence of AIDS in systemic lupus erythematosus (SLE)

Regarding the hypothesis that SLE and other autoimmune disease patients will have less clinical AIDS than expected, several investigators have noted the paucity of coincident SLE and HIV-1 infection [9,15,41]. In 1993, Barthel and Wallace calculated that at that time 400 cases of coincident SLE and HIV-1 disease would have been expected based on the prevalence of each disease [2]. In 2004, Palacios and Santos noted that there were only 32 reported cases of combined SLE and HIV-1 disease [40]. Further, only 21 of the 32 SLE/HIV-1+ cases met full diagnostic criteria for SLE. Of the total, the diagnosis of HIV-1 infection and SLE was made concurrently in three cases, HIV-1 preceded SLE in 14, and 15 of 32 were diagnosed with SLE before the diagnosis of HIV-1. Kaye suggested that the dearth of SLE patients with AIDS may relate to the spectrum of antibodies that SLE patients can make [25]. In the NIAMS Lupus repository, 2 of 5,000 are known to be HIV-1 infected (Harley, J, Haynes, B, unpublished).

To study these issues, we have begun to prospectively study SLE and APS patients both for the incidence of HIV-1 infection and for the quality of anti-HIV-1 neutralizing antibodies that these patients make after HIV-1 infection (Moody, MA, Haynes, BF, unpublished). Thus, whether there is a true paucity of HIV-1 in autoimmune disease patients, and whether autoimmune patients can make protective anti-HIV-1 antibodies after HIV-1 infection are important questions.

Production of broadly reactive neutralizing antibodies by HIV-1+ autoimmune disease patients, would provide strong evidence for the suppression of neutralizing antibodies by B cell tolerance mechanisms. If HIV-1 infected autoimmune disease patients and animals with autoimmune diseases do not make broadly reactive neutralizing antibodies, then it will point to other mechanisms of control of these antibody species.

Since autoantigens are conserved throughout phylogeny, it is also important to study the ability of autoimmune mice to make anti-cardiolipin and anti-MPER antibodies both before and after immunization with oligomeric gp140 Envs that express the 2F5 and 4E10 gp41 epitopes. We have found that B cells making antibodies against the 2F5 versus the 4E10 MPER peptide epitopes appear to be differentially regulated in both normal (BALB/c) and autoimmune (MRL/1pr−1/−) mice (Verkoczy, L., Haynes, BF, et al., unpublished). For example, antibodies against a 2F5 gp41 epitope are constitutively produced in MRL/1pr−/− mice while antibodies against the 4E10 peptide epitope are not. The nature of the 2F5-epitope reactive antibodies produced in MRL/1pr−/− mice are being probed by making murine Mabs reflective of these antibody reactivities followed by cloning and sequencing of the Mab CDR3 regions.

These issues are also being studied in acute and chronic HIV-1 infected patients and in rare HIV-1 infected autoimmune disease patients by determining the presence or absence of 2F5 and 4E10-reactive B cells using the antigen specific liposome probes described above as markers of antigen specific B cells. It is important to determine in HIV-1 patients if 2F5 and 4E10 B-reactive cells are made, and if present, determine their B cell subset of origin.

1.4. Prevalence of anti-cardiolipin antibodies in HIV-1 infection

There are two types of anti-cardiolipin antibodies, “autoimmune” and “infectious” [21,34]. “Autoimmune” anti-cardiolipin antibodies are found in diseases such as SLE and primary APS. They are associated with clotting syndromes and are frequently dependent on β-2-glycoprotein-1 for antibody binding. In contrast, “infectious” anti-cardiolipin antibodies arise as a consequence of polyclonal activation induced by infectious agents such as EB virus, leishmaniasis, and HIV-1, and are not usually β-2-glycoprotein-1 dependent, nor associated with clotting syndromes [21].

Early on after the recognition of AIDS, it was found that up to 40% of HIV-1 infected patients make anti-cardiolipin antibodies [25]. However, Bibollet-Rusche et al have shown that most chronically infected HIV-1 patients do not make anti-2F5 or 4E10-like neutralizing antibodies [3]. Interestingly, Cavacini et al. have shown that there is a failure to make robust IgG3 responses to neutralizing epitopes on HIV-1 Env during natural infection [11]. Thus, the anti-cardiolipin antibodies that are present in HIV-1 chronically infected patients likely will not be reactive with HIV-1. Rather, the occurrence of these “infectious” anti-cardiolipin antibodies in chronic HIV-1 infection is likely due to disordered immunoregulation from T cell dysfunction induced by HIV-1 infection [47]. Nonetheless, the spectrum of cardiolipin and other polyspecific antibodies from HIV-1 infected patients should be probed to evaluate their potential for contributing to a salutary anti-HIV-1 antibody response.

1.5. Lipid reactivity and neutralization of Anti-MPER antibodies

The role of polyspecificity and lipid reactivity in HIV-1 neutralization is being studied by analysis of the mode of binding of 2F5 and 4E10 to artificial liposomes that project the nominal 2F5 and 4E10 gp41 epitopes from the liposome surface to mimic the relationship of the gp41 MPER to the viral lipid bilayer. Both 2F5 and 4E10 mAbs bind to their respective nominal epitope peptides that is described by a simple Langmuir equation. However, our initial studies with bivalent Mabs suggest that 2F5 and 4E10 both bind to MPER liposomes by a sequential two-step conformational change binding model [29]. The first step of this linked biphasic binding mode would involve the encounter of the Mab with membrane lipids preceding the subsequent more stable docking of the Mab to gp41 (Alam, M, Haynes, B, manuscript in preparation). This model is in keeping with the observations of Ofek et al. [38] and Cardoso and colleagues [10], who have shown in crystal structures of 2F5 and 4E10 Fabs with the nominal gp41 peptide epitopes, that the gp41 contact points of both mab CDR3s are small, and postulated that 2F5 and 4E10 CDR3s could also bind to the viral lipid bilayer.

1.6. Candidate B cell subsets from which broadly neutralizing antibodies might originate

Several B-cell compartments, each with characteristic properties of specificity, response and location, are found in peripheral lymphoid tissues. All B cell types develop from committed lymphocyte progenitors in fetal liver or bone marrow via a process of sequential immunoglobulin (Ig) gene rearrangements that produces a functional B-cell antigen receptor complex (BCR) [18,30]. BCR assembly from Ig gene segments is random and inescapably, autoreactive B cells are generated. These self-reactive B lymphocytes are deleted or silenced by a variety of tolerizing mechanisms [1,35,37] and only a fraction of the newly formed B cells produced in the bone marrow leave as IgM⁺IgD⁻ transitional type 1 (T1) B cells [18]. T1- and the more mature (IgM⁺IgD^lo) T2 B cells represent the final stages in the maturation of immature B lymphocytes and they travel via the blood to peripheral lymphoid sites, primarily the spleen, where they undergo final maturation and selection to become mature B cells [33,50,57,62,66].

The largest compartment of mature B cells, the mature follicular or B2 compartment, comprises some 80% of peripheral B cells. B2 B cells express IgM and IgD (IgM^loIgD^hi), receptors for complement (CD21) and IgE (CD23), are long-lived, recirculate through secondary lymphoid organs, and dominate T cell-dependent antibody responses to protein antigens (Table 1) [33,62]. A second B-cell type, marginal zone (MZ) B cells strongly express IgM and CD21 but little IgD, and are characteristically distributed along the periphery of splenic follicles. MZ B cells constitute 10%–15% of mature B lymphocytes and are specialized for rapid humoral responses to microbial pathogens and blood-borne antigens [33,62] (Table 1). The third subset of mature B cells is the B1 compartment. B1 B-cells are not continuously supplied from bone marrow progenitors, but rather arise during fetal development and are continuously self-renewing in adults. B1 cells persist in the spleen and in the pleural and peritoneal cavities where they produce the great majority of natural serum antibody. Typically, B1 cells bear and secrete low-affinity, polyspecific IgM (21) but can be induced to undergo Ig class switching (CSR) [24,33,62] (Table 1).

Table 1.

Characteristics of transitional and mature B cell subsets

Characteristics	Transitional 1(T1)	Transitional 2 (T2)	Marginal Zone (MZ)	B1	B2, Mature Follicular (MF)
Percent of B Cells (in spleen)	∼2%	∼2%	∼10%	∼5%	∼80%
Phenotype	CD19+, CD21−/lo CD23−,CD24hi IgMhi	CD19+,CD21hi CD23hi,CD24hi IgMint/IgDlo	CD19+,CD21hi CD23lo/−CD24hi IgMhi/IgDlo/int	CD19+,CD5hi CD11b+,IgMhi CD1c+	CD19+,CD21int CD23hi,CD24lo IgMlo/IgDhi
Location (Most to least frequent)	Bone marrow, Peripheral blood, Spleen	Bone marrow, Peripheral blood, Spleen,	Margins of splenic follicles	Spleen, Pleural and peritoneal cavities	Spleen and Lymph nodes (follicles, germinal centers), Bone marrow
Half Life	Short	Short	Long	Short	Long
Time to Cell Cycle (LPS)	Short	Long	Short	?	Long
Polyreactive CDR3s	++	+/−	++	++	+/−
IgG3 Germline Transcripts	+*	− *	+ + *	?	− *
IgG3 CSR (γ3 circular transcripts)	+*	− *	−/ + *	?	− *
First (Innate) Response to Bacteria	Yes*	?	Yes	Yes	No
Respond to T Independent Antigens	Yes*	?	Yes	Yes	No
Response to T Dependent Antigens	?	?	Yes	No	Yes

^*Ueda, Y., Liao, D., and Kelsoe, G., submitted.

Activation-induced cytidine deaminase (AID) is the enzyme responsible for the two key process of antigen driven B-cell differentiation: somatic hypermutation (SHM) and Ig CSR [39]. AID is primarily expressed by B cells in germinal centers, histologic sites of intense SHM and CSR [23] and the loci of antibody affinity maturation [22]. In mature B2 cells, AID expression is induced by BCR signaling in conjunction with CD154-dependent T-cell help [13,31] or by activation through several toll-like receptors (TLR) [20,68]. Interestingly, AID expression driven by BCR/CD154 signaling results in both SHM and CSR whereas TLR signals support CSR only [20].

Two exceptions to this paradigm for induced AID expression have been reported. Weller et al. [64] identified extensively mutated Ig genes in MZ-like B cells from CD154 deficient patients (type-I hyper IgM syndrome) incapable of germinal center responses and concluded that an alternative pathway for SHM must exist. Secondly, the laboratories of Imanishi-Kari [32] and Melamed [48] identified AID expression in developmentally immature B cells that was sufficient to support limited CSR [48] or, in certain conditions of B-cell deficiency, SHM [32].

Ueda, Y., Liao, D., and Kelsoe, G., (submitted) have followed up these reports with studies of developmentally regulated AID expression and learned that murine T1 B cells, whether in the bone marrow or spleen, constitutively express AID at levels similar to that of B2 cells after 24 h exposure to TLR ligands [57]. AID message in T1 B cells is correlated with active γ3 germline transcription and CSR and with low but significant frequencies of SHM. Surprisingly the T1 compartment also is capable of efficient antibody responses to microbial vaccines (Ueda et al., submitted). In fact, the response of T1 B cells to microbial products is similar to that of MZ B lymphocytes; both responses are rapid, both are enhanced by BAFF, both comprise a (weakly) self-reactive component, and both produce substantial quantities of IgG3.

These similarities between the T1- and MZ antibody responses suggest some specific connectivity between MZ B cells and the T1 compartment [16,33,57]. In mice, inflammatory responses to adjuvants or infections elicit the emigration of B-cell precursors from the bone marrow and the establishment of extramedullary B lymphopoiesis that dramatically expands the numbers of peripheral T1 B cells [58,59]. The production of AID⁺T1 B cells at sites of inflammation/infection appears well regulated and may, like B1 B-cell responses [33,62], represent an innate component of humoral immunity that, while T-cell independent, is capable of SHM and CSR [57]. A recent report by Sims et al. demonstrating circulating human transitional B cells in the blood of SLE patients indicates that a similar developmental path/innate humoral response might exist in humans as well [50]. Could these circulating T1 cells represent the precursors of the mutated MZ B lymphocytes reported by Weller et al. [66]? This possibility is being actively pursued (Y. Ueda and G. Kelsoe, unpublished).

Alternatively, murine MZ B cells can generate germinal centers with SHM and the production of membory B cells in response to some T-dependent antigens [51]. Although germinal centers founded by MZ B cells require CD154+ T-cell help, T-cell independent germinal centers that support limited SHM – but not B-cell memory – have been reported [28,55].

The polyspecificity, frequent point mutations, and IgG3 isotype of the 2F5 and 4E10 Mabs could indicate their origin from the polyspecific, autoreactive B cells found in the T1- or MZ B-cell compartments. The inability to induce or detect similar antibodies in mice, non-human primates, or humans could be the consequence of clonal rarity, of immunization protocols that do not target T1 and MZ B cells, or from tolerizing events that remove or anergize activated B-cells with polyspecific/autoreactive BCR. Alternatively, anti-MPER B cells may originate in the mature follicular B2 cell pool, but also be subjected to peripheral tolerance mechanisms that delete or revise self-reactive/polyspecific BCR generated by SHM in germinal centers [60]. Indeed, purifying selection against self-antigen surrogates has been demonstrated in germinal centers by several groups [17,42,49].

If HIV-1 broadly neutralizing epitopes indeed are poorly immunogenic because they activate polyspecific/self-reactive B1, T1 or MZ B cells, then key areas of research should include the design of immunogens that recruit B2 cells to MPER epitope responses and/or to identify adjuvants that safely induce non-pathogenic antibodies from the B1, T1 or MZ B-cell pools. Alternatively, if the problem of generating ineffective antibody is caused by the gp41 MPER epitope mimicking conserved self antigens and selecting autoreactive B cells from various B cell pools, then gp41 MPER structures must be found that induce non-polyspecific antibodies capable of binding to neutralizing epitopes. However, as noted above, the apparent lipid binding requirement of 2F5 and 4E10 Mabs to neutralize HIV-1 may make this latter goal difficult. In this case, an important goal of HIV-1 vaccine development would be to devise strategies to recruit “innate” B1, MZ and transitional 1 B cells to be primed to mediate rapid protective anti-HIV-1 antibody responses at mucosal sites.

It should be noted that at least two species of anti-MPER antibody types have been described, those that bind to the nominal peptide alone but do not neutralize HIV-1 [14,71], and the unique 2F5 and 4E10 Mabs themselves. Most of the induced MPER antibodies that do not neutralize HIV-1 bind N-terminal to the ELDKWAS sequence [14,61,71]. We have recently made a murine anti-MPER mab (13H11), that binds to the sequence QQEK seven amino acids N-terminal to the 2F5 epitope ELDKWAS (Alam, M., Scearce, R.M., Haynes, B.F., unpublished). Mab 13H11 does not react with cardiolipin nor neutralize HIV-1 in vitro, yet blocks the binding of 2F5 Mab to oligomeric gp140. Similarly, Earl et al., have shown that the anti-MPER antibody D50 also binds to the N-terminus of the MPER region, and does not neutralize HIV-1 [14]. Thus, not all B cells that bind to nominal 2F5 epitopes, either in an ELISA peptide assay, or in assays to detect 2F5-specific B cells, will be truly 2F5-like and neutralize HIV-1.

1.7. Do other pathogens evade protective immunity through tolerance mechanisms?

Campylobacter jejuni is one of the most common causes of diarrhea world-wide, and is complicated in −0.01% of infections by the autoimmune neuropathy, Guillain – Barré syndrome (GBS), that is caused by C. jejuni-induced anti-ganglioside autoantibodies [5]. A target of protective antibodies in C. jejuni is lipopolysaccharide (LPS) [4,5]. GBS following C. jejuni infection is caused by a cross-reactive antibody repertoire that arises as a result of molecular mimicry between the C. jejuni LPS and GD3 and GM1 ganglioside on neural tissues, and that is not produced in patients with uncomplicated C. jejuni infection [4]. When pathogenic cross-reactive human anti-ganglioside/LPS antibodies are made and cause GBS, B cell tolerance is broken, and the induced pathogenic anti-ganglioside/LPS antibodies are somatically hypermutated and are of the IgG1 and IgG3 subclasses [5]. In animal models of this syndrome, similar anti-ganglioside/LPS antibodies have features of antibodies derived from “innate” B cells (B1, marginal zone, T1) i.e. polyspecificity, low affinity, and high frequency of IgG3 subclass [4,5]. In mice, these pathogenic anti-ganglioside/LPS autoantibodies have been shown also to be regulated by tolerance mechanisms and require evasion of tolerance mechanisms for antibody induction [4,5]. Thus, C. jejuni appears to utilize host B cell tolerance mechanism similar to what is proposed here for HIV-1 for evasion of protective neutralizing antibody responses to C. jejuni LPS.

2. Summary

HIV-1 has evolved multiple and varied ways to evade protective antibody responses. One of these mechanisms may be the requirement for MPER antibodies to be polyspecific and be able to react with host cell-derived lipid of the viral membrane in order to effectively neutralize HIV-1. The hypothesis that some species of anti-HIV-1 broadly neutralizing antibodies may be controlled by tolerance mechanisms is an hypothesis and requires in-depth analysis and testing. The ultimate goal however, is to determine if study of this hypothesis can teach us how to safety induce broadly reactive protective anti-HIV-1 antibodies. If immunoregulatory control mechanisms are found to be limiting induction of some species of broadly reactive antibodies, then recruitment of otherwise non-responsive B cell pools capable of inducing protective immune responses may be achieved by adjuvant formulations aimed at breaking tolerance, by structural studies designing new HIV-1 Env structures that recruit B cells to make the desired immune responses, or by a combination of both strategies.