Minding the gap: the impact of B-cell tolerance on the microbial antibody repertoire

Summary:

B lymphocytes must respond to vast numbers of foreign antigens, including those of microbial pathogens. To do so, developing B cells use combinatorial joining of V-, D-, and J-gene segments to generate an extraordinarily diverse repertoire of B-cell antigen receptors (BCRs). Unsurprisingly, a large fraction of this initial BCR repertoire reacts to self-antigens, and these “forbidden” B cells are culled by immunological tolerance from mature B-cell populations. While culling of autoreactive BCRs mitigates the risk of autoimmunity, it also opens gaps in the BCR repertoire, which are exploited by pathogens that mimic the forbidden self-epitopes. Consequently, immunological tolerance, necessary for averting autoimmune disease, also acts to limit effective microbial immunity. In this brief review, we recount the evidence for the linkage of tolerance and impaired microbial immunity, consider the implications of this linkage for vaccine development, and discuss modulating tolerance as a potential strategy for strengthening humoral immune responses.

1. INTRODUCTION

Vaccines are among modern medicine’s triumphs. Vaccination has eradicated smallpox globally,¹ removed the threat of poliomyelitis from many nations,² and greatly reduced infant morbidity and mortality world-wide.³ Virtually all vaccines act by eliciting protective antibody (Ab) responses that persist to provide long-lasting immunity against infection.⁴ There are, however, notable exceptions to the vaccine paradigm. For example, rapidly mutating viruses such as HIV-1 and influenza escape from pre-existing humoral immunity (within individuals, HIV-1; within populations, influenza),^5–7 making it difficult to generate vaccines that elicit sufficiently broad humoral responses to protect against viral variants.^8,9 Whereas immunization with HIV-1 envelope (Env) antigens can initiate broadly neutralizing Ab (bNAb) lineages, it does not drive their maturation to neutralization breadth and potency.^10–14 Similarly, contemporary influenza vaccines are seasonal because they are primarily strain-specific and require annual revision to elicit protective Abs against the variant viruses in circulation.⁹

The recovery of broadly protective Abs against HIV-1 or influenza indicates that “universal” vaccines might be possible for these infectious agents.^15–20 Broadly protective Abs often target conserved epitopes that are critical for virus infectivity, assembly, or stability, so that mutations in these epitopes/structures substantially compromise viral fitness and are therefore disfavored.^21–23 However, the recovery of broadly protective Abs presents a conundrum: if such Abs can be elicited, then why are HIV-1 and influenza broadly protective Abs significantly rare?^8,24–26 Whereas a number of non-mutually exclusive hypotheses have been offered to account for the barriers to generation of HIV-1 and influenza bNAbs,^27–29 in this review, we focus on the tolerance hypothesis, a model of immune evasion by molecular mimicry.^30–32

To meet the challenge of providing a defense against a numerous and diverse assortment of pathogens, B cells use random V(D)J gene recombination³³ and pairing of immunoglobulin heavy- and light-chains to generate a clonally distributed pool of as many as 10¹⁸ unique BCRs, a potential repertoire so large that it becomes limited by the size of the B-cell compartment.³⁴ An inevitable consequence of generating such a diverse array of BCRs is that a substantial fraction of them bind self-antigens. Indeed, repertoire analysis of the human early immature B-cell population shows that about 75% of newly generated BCRs bind autoantigens.^35,36 Self-reactive B cells are potentially dangerous, as their dysregulation can result in systemic autoimmune diseases, including systemic lupus erythematosus (SLE) and rheumatoid arthritis (RA).^37–40 Proper control of self-reactive B cells mitigates the risk of pathologic autoimmunity and aids in the maintenance of homeostasis.^{35,36,39,41,42} Developing B cells are screened at tolerance checkpoints, during which autoreactive cells are culled via deletion (apoptosis),^{35,39,43–46} receptor editing (secondary rearrangements to produce a non-autoreactive BCR),^47,48 or anergy (physiologic inactivation of B cells making them resistant to activation and plasmacytic differentiation).^49–51 Studies in mice have estimated that of the 10–20 million IgM⁺ immature B cells generated daily in the bone marrow, only ~10% reach the spleen and ~2% enter into pre-immune, mature B-cell pools.^52–54 Thus, ~98% of BCRs generated during B-cell development are lost to immune tolerance or other mechanisms that control B-cell maturation⁵⁵ and, normally, are unavailable in the pre-immune, mature B-cell repertoire.

What is the impact of these losses on the primary BCR repertoire? Do tolerance controls eliminate BCR specificities that are important for Ab responses to microbial pathogens? If epitopes of pathogens structurally resemble those of autoantigens, then tolerance controls might purge or inactivate many of the very B (or T) cells most fit to respond (Fig. 1).^30,56 This is the central premise of the tolerance hypothesis and the focus of our review. We will discuss recent studies in the field of B-cell tolerance and review their implications for humoral immunity to pathogens, with specific emphasis on how tolerance controls constrain the BCR repertoire to foreign antigens. We will also discuss how accessing the “forbidden” BCR repertoire might be useful for vaccination strategies.

Fig. 1. B-cell tolerance controls create “gaps” in the B-cell receptor (BCR) repertoire.

Each stacked bar graph represents the BCR repertoire – i.e., the pool of all BCR specificities – at a given stage of B-cell development. The repertoire comprises BCRs that bind only epitopes from foreign antigens (blue), BCRs that bind epitopes shared by foreign and self-antigens (yellow), and BCRs that bind only self-epitopes (red). Collectively, early immature B cells with newly rearranged BCRs bind a diverse array of foreign and/or self-antigens (left). Tolerance checkpoints purge the repertoire of most self-reactive B cells (right), creating a gap or blind-spot (empty dashed box) that is exploited by HIV-1, Campylobacter jejuni, and (possibly) influenza and other pathogens.

2. B-CELL TOLERANCE CHECKPOINTS

Tolerance checkpoints reduce the frequencies and avidities of self-reactive B cells and maintain homeostasis.^{35,36,39,41,42} The first B-cell tolerance checkpoint (i.e., central B-cell tolerance) occurs as developing bone marrow B cells begin to express cell-surface immunoglobulin.^57,58 This checkpoint primarily removes B cells that bind nuclear proteins, reducing the frequency of self-reactive cells by half (to 40%), as determined by cloning recombinant Abs (rAbs) from single B cells and assaying for their reactivity to HEp-2 cells/lysates.^35,36,39 Additionally, the first checkpoint decreases the frequency of polyreactive B cells [defined as those that bind two or more elements from this list: double-stranded DNA (dsDNA), single-stranded DNA, lipopolysaccharide, and insulin] from 60% to 7%.^35,36,39

Studies of mice and humans deficient in AID, MyD88 or IRAK4 have revealed that these molecules are critical for purging poly- or autoreactive (poly/autoreactive) immature B cells at the first tolerance checkpoint.^{40,41,59–63} Mice deficient in AID and MyD88^41,61 offer a different view of how the first checkpoint reduces autoimmunity. In contrast to the clonal elimination characteristic of BCR transgenic mice expressing autoreactive BCRs with a single specificity and affinity,^43–46 in mice with diverse BCR repertoires, the first checkpoint does not remove all self-reactive B cells but rather acts to shift autoreactive BCR avidities.⁴¹ In a study of the role of AID and MyD88 in central tolerance, we compared the frequency and avidity of single-cell cloned populations of immature/transitional 1 (imm/T1) B cells that bound dsDNA.⁴¹ Whether from MyD88-sufficient or deficient mice,^41,64 the frequencies of all dsDNA-binding imm/T1 B cells differed only slightly – and not significantly – in MyD88-sufficient (23%) mice with an intact first checkpoint from MyD88-deficient mice (26%) with impaired tolerance.⁴¹ In addition, the maximum BCR avidities for dsDNA also did not significantly differ between MyD88-deficient and MyD88-sufficient mice. Instead, we found that the avidity distribution of dsDNA imm/T1 BCRs was shifted to significantly higher values in MyD88-deficient mice compared to MyD88-sufficient mice.⁴¹ The avidity distributions of these imm/T1 populations effectively map the parameters of the first tolerance checkpoint for dsDNA reactivity. Thus, in mice at least, central B-cell tolerance does not act by the exhaustive depletion of autoreactive B cells or by purging self-reactive B cells with BCR avidities above some threshold. Instead, the first checkpoint determines avidity set points about which autoreactive BCRs cluster.⁴¹ In this way, the central B cell tolerance checkpoint maintains diversity in peripheral BCR repertoires, including those that bind self-antigens but offer possible protection against pathogens that mimic their host. Presumably, the potential risk of developing autoimmunity is balanced by more effective responses to viruses, bacteria and parasites.

The second tolerance checkpoint occurs as B cells complete their maturation in the periphery (i.e., from early bone marrow emigrants to mature B cells in humans or from T2 to mature B cells in mice).^35,46,51 This second checkpoint removes or inactivates self-reactive B cells that were not removed by first checkpoint,^{35,36,42,46,51} reducing the frequency of autoreactive mature B cells to ~20%.^35,36,39,42 The first and second tolerance checkpoints appear to differ in their targets, with central tolerance focused on the removal of cells that bind the ligands for endocytic TLRs or are polyreactive,^35,36,41,63 and the second checkpoint focused on removal of cells that bind non-mitotic antigen ligands.^35,39 Follicular exclusion^65,66 and enhanced BAFF dependence⁶⁷ represent examples of the second checkpoint in which BCR activation by autoantigens results in apoptosis in the absence of survival signals provided by T-cell help or insufficient BAFF signaling, respectively.^65–67

It is somewhat surprising that 20% of circulating mature naïve B cells retain self-specificity; however, two independent approaches obtained similar estimates of this value: Wardemann and Meffre cloned recombinant IgGs from single B cells, and measured the reactivity of the IgGs to HEp-2 cells/cell lysates,^35,36,39 while we have assayed clonal IgGs from single B-cell Nojima cultures against panels of self-antigens in Luminex binding assays.⁴²

Interestingly, the second tolerance checkpoint in healthy humans does not appear to counterselect all self-specificities with equal efficiency.⁴² We observed that for most of the self-antigens on our screening panel – including SSA, HEp-2 nuclear antigens, Centromere B, Sm, Histone, SSB, and insulin – the frequency of reactive BCRs was ≥2-fold lower in mature B cells than in transitional B cells.⁴² However, reductions in frequency were more modest (20% to 35%) for ribonucleoprotein (RNP), Scl-70, or Jo-1, and for some autoantigens [dsDNA or kynureninase (KYNU)] showed no significant reductions.⁴² Given that BCRs specific for dsDNA or KYNU have been shown to be culled by the first checkpoint,^68–70 the stability (or even increases) in frequencies across the second checkpoint suggest that the targets of central and peripheral B-cell tolerance differ.^35,41,63

In contrast, as B cells matured in SLE patients with impaired second checkpoint activity, the frequencies of autoreactive BCRs in transitional and mature B cells were not significantly different.⁴²

As noted above, it is somewhat unexpected that 20% of circulating, mature, naïve B cells in humans exhibit self-reactivity.^35,36,39,42 The 20% of self-reactive mature B cells appear to remain silent in healthy donors; however, it is possible that through incidental and/or chronic exposures to self-antigens and cognate T-cell help, these B cells can be induced to produce mutated, high affinity autoAbs. Nevertheless, neither high affinity autoAbs nor symptomatic autoimmunity are present in the great majority of healthy individuals,^71–74 indicating that regulatory activities other than those that control BCR repertoires constrain autoimmunity.^35,43–51 Recently one of these constraints against autoimmunity was demonstrated by the “redemption” of autoreactive B cells in germinal centers (GCs) by the acquisition of BCR mutations that decrease self-reactivity but heighten binding to foreign antigen.^75–77 Reciprocally, GC B cells that acquire de novo self-reactivity as a result of BCR hypermutation are eliminated by apoptosis or silenced by other means.^78–84 Finally, it has been proposed that tolerance controls restrict the differentiation of self-reactive IgG⁺ human memory B cells into Ab-secreting, bone marrow plasma cells.⁸⁵

3. THE TOLERANCE HYPOTHESIS: ORIGIN AND FIRST EVIDENCE

A necessary evil of immunological tolerance is that it reduces the immune repertoire, creating “gaps” that are potentially vulnerable to exploitation by molecular mimicry (Fig. 1). This prediction, the tolerance hypothesis, was formulated in the early 1960s by Damian,⁵⁶ Sprent,⁸⁶ and Dineen,^87,88 who reasoned that a pathogen might camouflage its critical structures as host autoantigens to evade control by the host’s immune system. Collectively, they advanced the idea that the evolutionary pressure of an immune response might drive the selection of mutant pathogens whose “fitness antigens” (i.e., structures essential for the pathogen’s propagation) more closely resembled immunologically forbidden host structures.^56,87 In consequence, over iterative rounds of selection, successful pathogens would better disguise their vulnerable epitopes as autoantigens, allowing them to slip into the “gaps” in the Ab repertoire. The camouflaged antigens were called “eclipsed” antigens.⁵⁶

At the time the tolerance hypothesis was formulated, many instances of antigenic similarities between hosts and pathogens were already known,⁵⁶ and more cases of molecular mimicry have been described since.^89,90 However, in nearly all of these cases, the antigenic similarities were discovered due to the positive identification of Abs that cross-reacted with host and pathogen antigens and that frequently were associated with autoimmune pathology. These data did not directly support the tolerance hypothesis, whose fundamental prediction was that molecular mimicry would be an advantageous adaptation for pathogens because eclipsed antigens would avoid provoking immune responses.⁵⁶ However, studies⁹¹ of the humoral response to Campylobacter jejuni (C. jejuni), a causative agent for Guillain-Barré syndrome (GBS),⁹² eventually provided the first evidence for linkage between molecular mimicry and impaired Ab responses.

Guillain-Barré syndrome (GBS) is a neurodegenerative autoimmune disorder that manifests a few weeks after a viral or bacterial infection.⁹² A class of Abs elicited by the infection cross-react with gangliosides (sialylated glycolipids abundant in nerve membranes) and can initiate an autoimmune response that causes nerve damage and paralysis.⁹² Of the ~50% of GBS cases in which the antecedent pathogen can be identified, C. jejuni is the culprit in at least one-third.⁹³ C. jejuni is likely a common instigator of GBS because sialylated moieties on C. jejuni lipopolysaccharides (LPS) closely resemble those of human gangliosides (e.g., GM1).⁹⁴ Importantly, however, ≤0.01% of people infected with C. jejuni proceed to develop GBS.^95,96 One potential explanation for this is that in most C. jejuni-infected persons, tolerance controls prevent the generation of Abs to host gangliosides, as well as to the C. jejuni antigens that molecularly mimic those gangliosides. To test this hypothesis, Willison and colleagues used C. jejuni LPS to immunize wildtype mice or congenic mice deficient for N-acetylgalactosamine transferase (GalNAcT^−/−). Wildtype mice are immunologically tolerant to complex gangliosides, while GalNAcT^−/− mice lack complex gangliosides -- including GM1 and GD1a – and therefore are not tolerant to them.⁹¹ Immunization elicited weak anti-GM-1 and anti-GD1a responses in wildtype mice, with minimal IgG class switching; moreover, on rechallenge, no memory responses were observed.⁹¹ In contrast, GalNAcT^−/− mice produced significantly stronger Ab responses to GM1 and GD1a, with substantial IgG class switching and clear memory responses after boost immunizations.⁹¹ Humans infected by C. jejuni with uncomplicated enteritis shed bacteria for about 16 days and are characterized by low levels of ganglioside/LPS Abs.^97–99 Thus, it seems likely that host mimicry by Campylobacter “fitness antigens” plays a substantial role in prolonging bacterial transmission by reducing effective humoral responses.

4. MOLECULAR MIMICRY AND IMMUNE EVASION BY HIV-1

We have recently discussed the evidence for (and implications of) immune evasion by molecular mimicry in the field of HIV-1 vaccine research;^100,101 however, we will recapitulate the main ideas here.

The first indication that tolerance controls might proscribe the generation of some HIV-1 bNAbs came from the work of Haynes and colleagues,³⁰ who were studying the first known human HIV-1 bNAbs. Two of these bNAbs, 2F5 and 4E10, both target epitopes in the membrane-proximal external region (MPER) of HIV-1 Env, and both bNAbs have long, hydrophobic heavy-chain third complementarity-determining regions (HCDR3s), a common feature of poly/autoreactive human Abs.¹⁰² Haynes then proposed that the 2F5 and 4E10 Abs might also be poly/autoreactive, providing an explanation for why 2F5- and 4E10-like bNAbs were so challenging to elicit by vaccination or infection.³⁰ This prediction proved correct: 2F5 and 4E10 both exhibited substantial reactivity to cardiolipin (a host phospholipid), and also bound some other self-proteins in vitro.^30,70,103

Having established that at least some HIV-1 bNAbs are autoreactive, the next test of the tolerance hypothesis was to determine whether these self-specificities were physiologically relevant for B-cell tolerance controls. This was determined in knockin mice expressing the heavy- and light-chain (HC+LC) variable regions of mature bNAbs or their germline (gl) precursors. Mice expressing the HC+LC of 2F5, gl2F5, 4E10, gl3BNC60, or HC of 2F5 or gl3BNC60, showed clear evidence of one or more of the characteristics of immune tolerance, including: clonal deletion during B-cell development, BCR editing, and peripheral B-cell anergy.^{13,69,104–108} These specific events define mechanisms of B-cell tolerance observed in the many studies of transgenic mice expressing autoreactive BCRs.^{43,45,47,48,50,51,109} Therefore, the self-reactivity exhibited by several mature bNAbs or even their gl precursors is sufficient to activate tolerance counterselection.

Our group then began to identify the autoantigens bound by HIV-1 bNAbs. In early studies, 2F5 or 4E10 were used to immunoprecipitate potential autoantigen ligands from human cell lysates.⁷⁰ Candidate autoantigens were then identified by mass spectrometry fingerprinting, and the candidate list was narrowed by stringent immunosorbence assays. Separately, 2F5 and 4E10 were used to probe autoantigen microarrays displaying >9,400 human protein targets. The results of these complementary experiments were unanimous: the principal protein self-antigen for mature or germline 2F5 was KYNU, while 4E10 strongly bound splicing factor 3b subunit 3 (SF3B3).⁷⁰ Strikingly, the 6-amino acid linear epitope for 2F5 (ELDKWA) is exactly shared by the HIV-1 gp41 MPER and most mammalian orthologues of KYNU.⁷⁰ A notable exception is in the opossum, where a single substitution in the KYNU epitope (ELEKWA) abolishes 2F5 binding. This proved a fortuitous model for testing whether the suspension of tolerance to the 2F5 epitope would permit the generation of 2F5-like Abs. As expected, vaccinating laboratory opossums (Monodelphis domestica) with MPER-derived antigens elicited robust 2F5 epitope serum Ab responses that were ≥100-fold higher than in vaccinated mice.⁷⁰ The specificity of this increase was shown in that immunization did not induce Abs to the adjacent 4E10 epitope, consistent with the high homology between human and opossum SF3B3 and the broad polyreactivity of 4E10-like BCRs.^70,103

The opossum vaccination experiment left open the possibility that opossums might produce 2F5-like Abs for some reason other than the absence of ELDKWA-specific B cells due to tolerization. However, we later demonstrated that normal C57BL/6 mice also had the latent capacity to mount 2F5 Ab-like responses and that the capacity to do so was lost as B cells traversed the first tolerance checkpoint.¹¹⁰ Significantly, if developing B cells were allowed to bypass the central tolerance checkpoint via in vitro culture, the BCR repertoire retained specificity for the 2F5 epitope.¹¹⁰ In mice reconstituted with these cultured cells, vaccination with MPER antigens resulted in robust GC responses, whereas immunization of control animals did not. Moreover, secondary immunization of reconstituted animals produced MPER-specific serum IgG titers 12-fold higher than control animals.¹¹⁰

Recently, our group has screened thousands of single B-cell Nojima cultures⁶⁴ to determine the BCR repertoire before and after the tolerance checkpoints in 2F5 HC+LC mice (Finney et al., manuscript under revision). These experiments showed that BCR specificity for MPER and KYNU are almost perfectly correlated in 2F5 HC+LC mice. Importantly, KYNU/MPER-specific B cells existed only prior to the first tolerance checkpoint or in the peripheral anergic (IgM^-IgD⁺) B-cell pool. Interestingly, although the vast majority of peripheral mature B cells were purged of KYNU/MPER reactivity via LC editing, the cells retained the knockin 2F5 HC and maintained substantial avidity for cardiolipin. This implies that tolerization by the lipid-binding activity of the 2F5 bNAb is substantially less effective than is the elimination of KYNU-binding B cells; this dichotomy may be an important consideration when designing vaccination strategies to elicit 2F5-like Ab lineages.

Collectively, the experiments described above provide strong evidence that at least some broadly conserved HIV-1 epitopes evade immune control by mimicking autoantigens and hiding in the “gaps” that immunological tolerance creates in the Ab repertoire.

These studies, nonetheless, left open the possibility that bNAb poly/autoreactivity is a by-product of chronic infection due to persistent inflammation and prolonged antigenic exposure^111–113 rather than a common requisite for broadly neutralizing activity. To test this, we used microarrays displaying >9,400 human proteins to screen panels of bNAbs and non-broadly neutralizing Abs (nNAbs, i.e., non-neutralizing Abs and autologous neutralizing Abs) for auto- and polyreactivity.¹⁰³ We defined Abs as polyreactive if their averaged array binding was >2-fold greater than a standard, non-polyreactive control Ab. nNAbs and bNAbs that were not polyreactive were considered to be autoreactive if they bound one or more array protein targets with ≥500-fold avidity than did the standard control Ab.^70,103 Using these criteria, we found that the frequency of poly/autoreactive nNAbs (~20%, 2/9) was no different from that of poly/autoreactive mature, peripheral B cells in healthy humans.^35,103 In contrast, ~60% of bNAbs were poly/autoreactive, including at least one example from each of the major bNAb classes (i.e., CD4 binding site, MPER, variable loops 1 and 2, or variable loop-associated glycan).¹⁰³ Importantly, bNAbs were also enriched for poly/autoreactivity compared to nNAbs isolated from infected patients (i.e., excluding nNAbs elicited by vaccination). Additionally, the frequency of V_H somatic mutations was not correlated with poly/autoreactivity.¹⁰³ Thus, bNAb poly/autoreactivity is not simply a result of chronic infection or extensive SHM. Instead, these data indicate that poly/autoreactivity is probably inherently linked to broadly neutralizing activity, supporting the hypothesis that HIV-1 disguises its most sensitive epitopes as host molecules.¹⁰³

If only ~60% of bNAbs are poly/autoreactive when assayed for self-protein binding, then why are the remaining 40% also challenging to elicit?¹⁰³ One explanation is that the protein microarray underestimates the prevalence of bNAb autoreactivity because it cannot identify Abs that bind non-protein self-molecules. For example, the PGT121 bNAb binds self-glycans, even in the absence of protein determinants.^114–116 Additionally, there may be other obstacles to bNAb generation, including conformational masking of broadly neutralizing epitopes,^117,118 immunological dominance of non-broadly neutralizing epitopes,⁸ the sparsity of Env spikes on virions,^119–121 and the requirement of some bNAb lineages for specific alleles of V-, D-, or J-genes.¹²²

5. POLY- AND AUTOREACTIVITY IN INFLUENZA HEMAGGLUTININ bNAbs

Circulating influenza viruses escape humoral immunity by acquiring mutations – usually in the head domain of the influenza hemagglutinin (HA) – at epitopes targeted by neutralizing Abs.¹²³ In consequence, annual changes in the circulating strains of influenza require seasonal re-vaccination to provide protection against the predominant contemporary variants and closely matched strains. A major goal of infectious disease immunology is the development of a “universal” influenza vaccine that would provide protection against many viral strains by eliciting bNAbs targeting conserved viral structures, including the receptor binding site (RBS) on the trimeric head of influenza HA, or the membrane-proximal HA stem region.⁹

Whereas RBS-specific bNAbs may be more common than previously appreciated, stem-directed bNAbs are rare.^124,125 Recent studies by us and others indicate that the rarity of stem-directed bNAbs may be due in part to counterselection by immunological tolerance controls.^32,124 Andrews et al. observed that ~90% of HA stem-binding human Abs were polyreactive toward dsDNA, LPS, and insulin, whereas <10% of HA head-binding Abs were polyreactive.¹²⁴ Importantly, the subset of stem-directed Abs with little neutralization capacity were only modestly enriched (~25%) for polyreactivity, implying that this feature may be intrinsically linked with broadly neutralizing activity, rather than being a general property of stem-binding Abs.¹²⁴ This conclusion was further supported by the determination that stem-binding Abs encoded by V_H1–69 (commonly used by stem-directed bNAbs) were far more enriched for polyreactivity than were V_H1–69-encoded HA Abs that did not bind the stem.¹²⁴ Thus, polyreactivity in HA-specific Abs is not a trait derived from a specific gene, but rather a feature tightly associated with neutralization breadth.

These results were generally supported by those of a second study, which assayed Ab binding to various lipids, fixed HEp-2 cells, and human protein microarrays to assess the poly/autoreactivity of 6 RBS-specific and 6 stem-directed bNAbs.³² The majority (5/6) of stem-directed bNAbs exhibited clear poly/autoreactivity in one or more of the Ab binding assays. Interestingly, half (3/6) of the RBS-specific bNAbs were also polyreactive, as determined by >2-fold greater binding to human protein arrays than the non-polyreactive control Ab.³² Collectively, these findings and those of Andrews et al. are consistent with the possibility that immunological tolerance controls reduce specific subsets of HA bNAb precursors – particularly stem-directed ones – from the functional BCR repertoire. In doing so, tolerance may thereby raise a barrier to the development of broad humoral immunity against influenza.^32,124 It will be exciting to see whether a direct in vivo test of HA bNAb purging confirms or excludes this hypothesis.

6. EVIDENCE IN HUMANS: IMPACT OF THE SECOND B-CELL TOLERANCE CHECKPOINT ON FOREIGN-REACTIVITY OF B CELLS

6.1. Peripheral tolerance removes B cells that react with microbial epitopes

As the studies described above have demonstrated, the tolerance hypothesis is supported by a growing body of evidence from animal models and in vitro experiments.^{30,32,70,91,103,124,126} However, these studies have focused on a few important pathogens, leaving undetermined the degree to which B-cell tolerance, by the restriction of autoreactive BCRs, also constrains the BCR repertoire to foreign antigens (both innocuous and pathogen-associated). In other words, the extent of the overlap between self-epitopes and foreign epitopes is unknown.^127,128 However, given that fundamental molecular structures of nucleic acids, proteins, carbohydrates, and lipids are common to all organisms, it is likely that overlap between self-epitopes and foreign epitopes extends to more cases than just C. jejuni, HIV-1 Env, (possibly) influenza HA and a few infection-triggered autoimmune diseases.^{30,32,70,89–91,103,124}

To i) determine the extent to which foreign-specific BCRs are restricted at the second tolerance checkpoint, and ii) test the tolerance hypothesis in humans, we isolated human transitional B cells (pre-checkpoint) and mature naïve B cells (post-checkpoint) from a cohort of healthy donors. Individual B cells were clonally expanded in in vitro cultures that promote robust proliferation, class-switching to IgG, and plasmacytic differentiation.^18,19,42,64 Then, we harvested the clonal IgGs from these cultures and measured their reactivity to a defined set of foreign antigens, including HIV-1 Env subunits gp41 and gp140, nitrophenyl acetyl-conjugated bovine serum albumin, ovalbumin (OVA), keyhole limpet hemocyanin (KLH), recombinant Anthrax protective antigen, streptavidin and influenza HAs.⁴²

At the checkpoint between human transitional B cells and mature, naïve B cells, the frequency of clonal IgGs that reacted with any of the foreign antigens (F-reactive) decreased by ~40%.⁴² Most of this reduction was due to a 50% loss in clonal IgGs that bound both foreign- and self-antigens (F+S-reactive), whereas the frequency of F-reactive clonal IgGs that did not react with self-antigens (F-only-reactive) only decreased 15% as transitional B cells traversed the checkpoint and became mature, naïve B cells.⁴² The loss, therefore, of F+S-reactive BCRs at the second tolerance checkpoint resulted in a significant reduction in F-reactive mature, naïve B cells. We interpret these results as a demonstration that the second checkpoint identifies almost half of all F-reactive BCRs in human transitional B cells as potentially autoreactive. Removal of 40% of F-reactive BCRs during human B-cell maturation surely identifies a significant overlap between self- and foreign epitopes and reduces the breadth of possible Ab response to microbial pathogens.⁴²

In contrast to healthy donors, frequencies of F-reactive and F+S-reactive BCRs were no different between the transitional and mature, naïve B cells of SLE patients, who are defective in the second tolerance checkpoint.^{36,37,39,40,42} F+S-reactive transitional B cells that are normally excluded in healthy individuals by the second tolerance checkpoint can enter into mature B-cell pools in SLE patients.

6.2. Mature, naïve B cells in healthy donors and SLE patients show distinct reactivity to foreign antigens

F+S-reactive B cells can provide BCR specificities that are distinct from those provided by F-only-reactive B cells. Therefore, because defective tolerance checkpoints in SLE patients do not efficiently remove F+S-reactive mature B cells, we hypothesized that the naïve BCR repertoire for foreign antigens in SLE patients would differ significantly from that in healthy controls. This hypothesis is supported by the fact that SLE patients can establish B-cell responses distinct from healthy donors in response to vaccination or infection. For example, in a cohort of people who had all received seasonal influenza vaccination, recombinant IgGs cloned from circulating plasmablasts of SLE patients bound to autologous influenza virus more avidly than did IgGs cloned from healthy controls.¹²⁹ Furthermore, the anti-influenza Abs from SLE patients possessed long HCDR3s, with frequent J_H6 gene usage.¹²⁹ As we will discuss later, these are features common to F+S-reactive BCRs. It has also been reported that a class of HIV-1 bNAb isolated from an SLE patient also react with DNA.¹¹¹

To test our hypothesis, we compared between healthy donors and SLE patients the distributions of foreign-reactivity and self-reactivity among BCRs from mature, naïve B cells. We found that SLE patients contained a greater frequency of mature, naïve B cells that bound certain self-antigens (Sm, Scl-70 and insulin) and foreign antigens (OVA, KLH, influenza HAs and HIV-1 gp-140) than did healthy people.⁴² In contrast, the BCR-reactivity distributions of transitional B cells did not significantly differ between healthy donors and SLE patients. Thus, in association with an impaired second tolerance checkpoint that fails to remove F+S-reactive clones from mature naive B cells, clonal IgGs expressed by mature naive B cells in SLE patients show distinct patterns of reactivity against foreign antigens that are not present in healthy controls.⁴² Therefore, the primary BCR repertoires of SLE patients are substantially different from those of healthy controls.

6.3. Features of F+S-reactive BCRs

That F+S-reactive B cells can provide a class of BCRs that bind foreign antigens and self-antigens through overlapping BCR paratopes was formally demonstrated by inhibition assays for F+S-reactive Abs.⁴² While incubating F+S-reactive Abs with immobilized target antigens, we titrated in soluble foreign antigens (i.e., potential competitive inhibitors) and determined the impact on Ab binding to the immobilized antigens. As expected, soluble foreign antigens inhibited F+S-reactive Abs from binding to immobilized versions of the same foreign antigen (i.e., homologous inhibition).⁴² In addition to the homologous inhibition, in some cases, soluble foreign antigens also inhibited Ab binding to immobilized self-antigens.⁴² This indicated that at least a fraction of F+S-reactive BCRs bind self-antigens and foreign antigens at the same paratopes or at proximal paratopes that are close enough to cause steric hindrance. Combined with the observation that the peripheral tolerance checkpoint reduces the frequency of F+S-reactive B cells, these results imply that the linkage of foreign and self-specificities causes immune tolerance controls to restrict the foreign antigen repertoire.⁴² In this section, we describe features of F+S-reactive BCRs and discuss types of BCR specificities to foreign epitopes that can be provided by F+S-reactive B cells.

1). Foreign epitopes that mimic epitopes on self-antigens –

F+S-reactive B cells can provide BCR specificities to foreign epitopes that structurally resemble self-epitopes. As we described above, one example is the HIV-1 bNAb 2F5, which avidly binds a rare peptide motif (ELDKWA) present in both the MPER of the HIV-1 gp41 ectodomain as well as in KYNU from humans and many other mammals.⁷⁰ Other examples include the HIV-1 bNAb CH98, which also binds to double stranded DNA;¹¹¹ and influenza nAb HC19,¹³⁰ which recognizes the HA RBS and also significantly binds to RNP, although with much lower avidity (~5,000-fold) than binding to HA.⁴²

2). Foreign specificities associated with long HCDR3s –

Self-reactive BCRs frequently incorporate long HCDR3s.^35,131–133 To determine whether this was also true of F+S-reactive BCRs, we recovered V(D)J rearrangements from clones that react with Sm and/or RNP self-antigens (as representative of self-reactive BCRs) and from non-self-reactive clones (exhibiting no binding to a defined panel of self-antigens) and analyzed the properties of their HCs.⁴² In self-reactive BCRs, including S-only reactive BCRs (those that bind a panel of self-antigens but not foreign antigens) and F+S-reactive BCRs, the HCDR3 length was significantly longer than that of non-self-reactive BCRs (~18 amino acids vs. ~15 amino acids, respectively). Regardless of the source of B cells (transitional B cells or mature B cells from healthy donors or SLE patients), self-reactive BCRs carried longer HCDR3s than did non-self-reactive BCRs.⁴²

With regard to V_H and D_H gene usage, no significant differences were found between self-reactive and non-self-reactive Abs: V_H3 and V_H4 genes and D_H3 genes appeared most frequently in both B-cell groups.⁴² In contrast, the distribution of J_H gene usage differed significantly between self-reactive and non-self-reactive B cells. While the J_H4 gene was used most frequently (about 50%) by non-self-reactive Abs, the J_H6 gene was recovered most frequently (55%) from S-only and F+S reactive B cells.⁴² Thus, self-reactive BCRs, including F+S-reactive BCRs, have long HCDR3s that are associated with frequent J_H6 gene usage.

A long HCDR3 using J_H6 is a common trait among certain classes of bNAbs for HIV-1 and influenza.^134–138 For these bNAbs, the extended HCDR3 loop appears crucial for interacting with conserved viral epitopes, including the CD4-binding pocket of HIV-1 Env and the RBS of influenza HA.^18,135–140 That this feature is counterselected during B cell development suggests that influenza or HIV-1 bNAb precursors might be among the F+S-reactive BCRs eliminated by the peripheral checkpoint.

3). Foreign specificities associated with polyreactivity –

In addition to counterselecting BCRs with strong avidity for one or a few related autoantigens, immune tolerance controls also reduce the frequency of cells with polyreactive BCRs, i.e., those that promiscuously bind several unrelated antigens.^35,36,39,141 Thus, tolerance can also restrict foreign specificities that are intrinsically linked with polyreactivity. A class of influenza bNAbs that react with the HA stem or RBS also exhibit polyreactivity to human self-antigens,³² suggesting a biological significance that can be provided by polyreactive Abs. By definition, F+S-reactive B cells are polyreactive, as they bind to two or more antigens. Indeed, most (about 80%) of F+S-reactive B cells identified in our study bound multiple foreign-antigens and/or self-antigens, respectively. This is in contrast to the fact that only about 20% of F-only or S-only reactive B cells bound multiple antigens.⁴² It is important to mention that the polyreactivity of F+S-reactive clonal IgGs did not uniformly extend to all antigens (e.g., OVA, KLH) we used in our assays, but rather exhibited “restricted polyreactivity.” Therefore, F+S-reactive B cells are not just polyreactive, but they bind selectively to antigens with avidity sufficient to exceed a threshold required for B-cell activation.^142–144

7. APPLICATION TO VACCINE STRATEGIES

In light of the role that tolerance controls play in purging the B cells potentially most fit to respond to host-mimicking pathogens, several questions with important implications for vaccination are raised. First, can tolerance controls be relaxed or circumvented to increase the frequency of peripheral B cells expressing desirable F+S cross-reactive BCRs? Second, can peripheral autoreactive B cells be recruited into humoral responses? Third, are relaxation of tolerance controls and activation of autoreactive B cells unacceptably risky for use in human vaccination? In the following sections, we address these questions.

7.1. Can tolerance controls be relaxed or circumvented to increase numbers of B cells with desirable F+S specificity?

Because immunological tolerance precludes Ab responses to otherwise vulnerable conserved microbial targets, there are at least two divergent strategies for developing universal vaccines against pathogens “camouflaged” by molecular mimicry. The first option is to accept the limits imposed by tolerance controls and to focus instead on eliciting Abs to the non-disguised epitopes. The second option is to modulate or “break” tolerance to regain access to the forbidden Ab repertoire. Unlike the second tactic, the first carries no additional risk of instigating autoimmunity, and as such, probably would face lower hurdles to regulatory approval and wide use. However, the potentially significant disadvantage of the first strategy is that it attempts to achieve immunity while targeting only a subset of conserved protective epitopes. For vaccination to HIV-1 (and likely other pathogens), this would require the elicitation of broadly protective Abs from an even more restricted pool of already rare precursors. Such a constraint could be problematic, since precursor B-cell frequency is likely an important determinant of B-cell competitiveness in humoral responses.^145,146 Moreover, some broadly protective Ab lineages require specific allelic variants of immunoglobulin V-, D-, or J-genes;^29,122 in individuals lacking the requisite alleles, the potential path to broad protection by vaccination would be severely narrowed or even non-existent. Nevertheless, it remains possible that these issues can be resolved and effective vaccines developed while staying within the confines established by tolerance controls.

The second option for vaccination against pathogens camouflaged by autoantigen mimicry is to interfere with immune tolerance. This might occur either by intentionally “breaking” tolerance to specific antigens, or else by transiently relaxing the stringency of tolerance controls to increase the frequency of B cells specific for otherwise forbidden epitopes. An obvious concern with these approaches is their potential to induce autoimmune disease. We address these concerns in a later section, and focus here on tactics to influence tolerance controls to elicit desirable F+S-reactive humoral responses.

A potential first step for accessing the forbidden Ab repertoire would be to increase the frequency of peripheral B cells bearing F+S-reactive BCRs. This could be achieved by transient relaxation of tolerance controls, e.g., by administration of chloroquine, a widely used, inexpensive anti-malarial drug that inhibits endosome acidification and thereby interferes with the first tolerance checkpoint.⁴¹ Our lab showed that such treatment impaired the negative selection of imm/T1 B cells expressing the 2F5 BCR or the dsDNA-reactive 3H9 BCR.⁴¹ Other potential targets for modulation of tolerance controls are the c-Cbl and Cbl-b ubiquitin ligases, which are reported to enforce B-cell anergy.¹⁴⁷ We emphasize here that relaxation of tolerance controls would have to be temporary to limit the risk of developing autoimmune disease. However, we propose that judicious use of chloroquine or other drugs, carefully timed in conjunction with vaccination, might be an appropriate tactic for increasing the peripheral frequency of normally proscribed BCRs and permitting their maturation into protective Abs.¹⁰⁰

7.2. Can F+S-reactive B cells be recruited into humoral responses?

Although F+S-reactive B cells are readily recoverable from the peripheral blood of both healthy donors and SLE patients, the biological significance of the F+S-reactive BCRs is unclear. It is important to consider whether self-reactive B cells (including F+S-reactive B cells) can respond to antigenic stimulation by foreign antigens and be recruited into humoral responses.

Most of F+S-reactive BCRs examined in our study had very low avidity for foreign and self-antigens,⁴² raising the question of whether they could be recruited into immune responses or provide protection against foreign threats. Studies have shown that a class of influenza or HIV-1 bNAbs do not bind their nominal antigens when these Abs are expressed as their reverted, unmutated germline forms.^148–154 This implies that pre-immune, naïve B cells expressing BCRs with unmeasurably low avidity for antigen can be recruited to GCs, where they undergo rounds of V(D)J SHM and selection, and eventually express mutated, high affinity BCRs/Abs that serve physiologically important functions. By extension, B cells expressing F+S-reactive BCRs with measurable (albeit low) avidity for foreign antigens may also be capable of participating in T-cell dependent humoral responses.^64,142–144

Another consideration is that peripheral autoreactive B cells (including F+S-reactive B cells) are often IgD⁺IgM^lo/-, a phenotype historically believed to be indicative of anergy (i.e., unresponsiveness to antigen stimulation).^{50,51,155,156} However, several recent studies demonstrated that appropriate stimulation can activate anergic cells and recruit them into immune responses. To break tolerance to the ELDKWA epitope shared by KYNU and HIV-1 MPER, 2F5 knockin mice were repeatedly immunized with MPER peptide-conjugated liposomes (designed to mimic MPER epitopes present on HIV-1 virions) and toll-like receptor (TLR) agonists.^13,157 In 2F5 HC+LC knockin mice, this vaccination regimen succeeded in selectively activating MPER-binding B cells and inducing their proliferation, class-switch recombination, and differentiation into plasma cells.¹⁵⁷ In consequence, vaccinated mice achieved substantial serum titers of MPER-specific neutralizing IgG. Similarly, MPER-liposome vaccination of gl2F5 mice resulted in the selective proliferation of MPER-binding B cells.¹³ Disappointingly, however, this strategy did not induce CSR or SHM, and elicited only low serum titers of anti-MPER IgM.¹³ Likewise, immunizing gl3BNC60 HC+LC knockin mice with highly multimerized immunogen (but not trimeric immunogen) reliably produced serum Ab responses to the HIV-1 CD4 binding site; however, activated B cells acquired few (if any) mutations, indicating poor affinity maturation.¹⁰⁴

Promisingly, vaccinating macaques with MPER liposomes and TLR agonists succeeded in eliciting ELDKWA-specific serum Abs.¹³ Thus, in mice and macaques, tolerance to self-antigens can be overcome by the right immunization regimen, permitting the generation of normally forbidden Abs against epitopes shared by foreign and self-antigens. Unfortunately, the MPER-specific Abs acquired only limited potency for virus neutralization, because the vaccine-induced SHM failed to achieve sufficient HCDR3 hydrophobicity for effective binding of virion lipids, which is required for HIV-1 neutralization.¹³

Recent studies by Goodnow and colleagues^75,77,158 have described an alternative outcome for antigen-activated self-reactive B cells that may be a relevant strategy for eliciting influenza or HIV-1 bNAbs to epitopes that imperfectly mimic host antigens.¹⁰⁰ This pathway involves recruitment of autoreactive precursor B cells into GCs, where ongoing SHM and clonal selection “redeem” autoreactive BCRs by reverting their autospecificity.^75,77,158 In proof-of-concept studies using a mouse model where a triply-mutated form of hen egg lysozyme (HEL^3x) was expressed as a ubiquitous neo-autoantigen, HEL^3x-specific B cells were anergic.⁷⁷ However, these cells could be activated and recruited into GCs by particulate immunogens decorated with high densities of a closely related foreign antigen, duck egg lysozyme (DEL).⁷⁷ Inside DEL-induced GCs, SHM and antigen-driven selection enriched for clonal lineages with increased binding to the foreign antigen (DEL) and reduced binding to self-antigen (HEL^3x).⁷⁷ Importantly, B cells with higher affinity to DEL and decreased affinity to HEL^3x could differentiate into memory B cells or Ab-secreting cells.⁷⁷ It remains to be seen whether this mechanism could also convert autoreactive bNAb precursors into mature bNAbs with reduced binding self-epitopes. Using the 2F5 Ab as an example, it is conceivable that minor structural differences in the nominal ELDKWA epitope present in HIV-1 Env and KYNU might permit gl2F5 to undergo affinity maturation to produce a mature 2F5-like bNAb with high affinity for Env and low affinity for KYNU. Presumably, such a bNAb no longer would be subject to strict tolerance controls.

7.3. Are vaccine strategies that relax or “break” tolerance unacceptably risky?

Immunological tolerance is important for mitigating the risk of autoimmune disease, so vaccination regimens that intentionally relax or “break” tolerance controls must be rigorously tested to assure safety. Encouragingly, there is evidence that such strategies will not necessarily result in autoimmune pathology. First, as mentioned above, autoreactive B cells can be recruited into GCs to undergo V(D)J SHM and affinity maturation that enhances BCR specificity for foreign antigen while reducing avidity for even closely related self-antigen.^75,77,158 Thus, vaccination regimens that reliably elicit enhanced BCR discrimination between foreign and self-antigen would presumably have low risks. Second, Abs with high avidity for self-antigens do not necessarily induce pathology. In HIV-1 patients receiving passive immunotherapy, infusion of the autoreactive bNAbs 4E10 or 2F5 was well tolerated.¹⁵⁹ Moreover, although 4E10 exhibited some anti-coagulant activity and modestly increased the activated partial thromboplastin time, the risk of thrombotic complications was deemed low.^30,159 Similar results were achieved in mice and macaques, in which passive infusion of 2F5 or elicitation of 2F5-like Abs by MPER-liposome vaccination did not inhibit KYNU activity nor produce other obvious side effects.^13,157 Therefore, while we reiterate the need to thoroughly evaluate the safety risks of any vaccination regimen designed to activate F+S-reactive B cells, existing evidence indicates that such strategies are not a priori disqualified.

8. CONCLUSION

The maintenance of self-tolerance depends in part on purging B cells with self-reactive BCRs from B-cell populations that can be activated by antigen. Some pathogens, such as C. jejuni, HIV-1, and perhaps influenza and others, have evolved so that structures critical to their replication and transmission resemble self-antigens of their hosts. This mimicry affords pathogens the opportunity to evade immune responses and increase the probability of transmission. For hosts, molecular mimicry turns immunological tolerance into a double-edged sword that both mitigates the risk of autoimmunity but also reduces the frequencies of the B cells potentially best fit to respond and protect against certain infections or vaccinations. This evolutionary “arms race” may explain why traditional vaccination strategies have failed to elicit reliable and broadly protective Abs against select pathogens. Fortunately, recent studies have identified possible avenues for surmounting the obstacles that tolerance imposes on immunization. First, tolerance controls can be pharmacologically modulated to increase the abundance of poly/autoreactive B cells with the potential for protective breadth.⁴¹ Second, in at least some instances, tolerance to certain self-antigens can be broken to gain access to the forbidden Ab repertoire, without inducing autoimmune disease.^13,157,159 Third, autoreactive BCRs can be “redeemed” in GCs, where SHM and antigen-driven selection can reduce their affinity for self-antigen while increasing their affinity for even highly similar foreign antigen.^75,77,158 We expect that future studies in this field will focus on these areas, particularly with regard to strategies for transiently relaxing immunological tolerance in conjunction with vaccination, which may be able to provide broad, durable protection against otherwise elusive pathogens.

Abbreviations

Ab

antibody

BCR

B-cell receptor

bNAb

broadly neutralizing antibody

DEL

duck egg lysozyme

dsDNA

double-stranded DNA

Env

envelope protein

F-reactive

foreign-reactive

F+S-reactive

foreign- and self-reactive

GBS

Guillain-Barré syndrome

GC

germinal center

gl

germline

HA

influenza hemagglutinin

HC

heavy chain

HCDR3

heavy chain third complementarity-determining region

HEL3x

triply mutated hen egg lysozyme

imm/T1

immature/transitional-1

KLH

keyhole limpet hemocyanin

KYNU

kynureninase

LC

light chain

LPS

lipopolysaccharide

MFI

mean fluorescence intensity

MPER

membrane-proximal external region

nNAb

non-broadly neutralizing antibody

OVA

ovalbumin

Poly/autoreactive

poly- or autoreactive

SF3B3

splicing factor 3b subunit 3

SHM

somatic hypermutation

SLE

systemic lupus erythematosus

S-reactive

self-reactive

RA

rheumatoid arthritis

rAbs

recombinant antibodies