Factors in B cell competition and immunodominance

Abstract

The majority of all vaccines work by inducing protective antibody responses. The mechanisms by which the B cells responsible for producing protective antibodies are elicited to respond are not well understood. Interclonal B cell competition to complex antigens, particularly in germinal centers, has emerged as an important hurdle in designing effective vaccines. This review will focus on recent advances in understanding the roles of B cell precursor frequency, BCR affinity for antigen, antigen avidity, and other factors that can substantially alter the outcomes of B cell responses to complex antigens. Understanding the interdependence of these fundamental factors that affect B cell responses can inform current vaccine design efforts for pathogens with complex proteins as candidate immunogens such as HIV, influenza, and coronaviruses.

Introduction

Affinity matured antibodies are thought to be the major mechanism by which most vaccines work^1,2. Neutralizing antibody mediated immunity can provide sterilizing immunity, a special level of protection that can only be elicited by antibodies in which infection is never initiated in the host. Germinal centers (GCs) are the microanatomical engines of antibody affinity maturation. Following vaccination, antigen activated B cells start proliferating, interact with CD4 T cells, and then migrate to the center of B cell follicles in which they establish the GC microstructure^3–5. Mature GCs consist of two distinct zones, dark zones (DZ) and light zones (LZ), which are delineated by ligands^6,7 CXCL12 and CXCL13 in concert with their corresponding chemokine receptors CXCR4 and CXCR5^5,6. B cells preferentially undergo division in the DZ^5,8. AID-mediated somatic hypermutation is thought to occur in the DZ as mutations are associated with proliferation and AID is known to act in G1 phase⁹. However, G1 phase GC B cells are readily detectable in both DZ and LZ¹⁰ and there has been no direct in vivo visualization of which anatomical compartment AID-mediated SHM occurs in. There is some evidence provided by Rajewsky and colleagues¹¹ that mutation can occur in lack of a DZ, as ablation of FOXO1 in B cells abrogated DZ formation as well as class switch recombination but SHM remained intact. Nevertheless, after mutation and proliferation, GC B cells then migrate to the LZ, where they encounter, and uptake antigen displayed as immunocomplexes on the surface of follicular dendritic cells (FDCs)^12–15. Antigen uptake and antigen processing by the GC B cell is followed by encounter with T follicular helper cells (T_FH) within the GC, which provide T cell help to the GC B cells via surface molecules such as CD40L or secreted cytokines such as IL-4 or IL-21^16,17. The amount of peptide major histocompatibility complex (p:MHC) presented by individual mutated GC B cells to T_FH is thought to be reflective of higher affinity B cells that are generated. B cells with higher p:MHC receive more T_FH help signals and are conditioned to return to the DZ and proliferate⁵. These iterative rounds of Darwinian selection and proliferation, shuttling between the DZ and LZ, is coined the “cyclic re-entry” model of GC B cell selection¹⁸.

The Darwinian nature of GCs gives mechanistic explanation to affinity maturation of antibodies following immunization or infection, a phenomenon which has been known for more than 50 years¹⁹. Some affinity matured antibodies can reach > 10,000-fold increases²⁰ (RKA and SC unpublished data) in affinity in the weeks following immunization. The three outputs of GC reactions are death by apoptosis²¹, memory B cells (B_mem)²², and plasma cells^23,24. One type of plasma cell that exits from the GC are long lived plasma cells (LLPCs) that reside in the bone marrow and produce protective antibody. B_mem and LLPCs can provide protection over a lifetime^25,26. While there is still great debate as to the mechanistic cues within the GC that govern the decision for each GC B cell to proceed toward one of these three fates, there is evidence that affinity for antigen plays a role. High affinity B cells in the GC are thought to exit toward LLPC fate while lower affinity B cells are thought to be preferentially selected toward memory B cell fate^27–30, while B cells that are of insufficiently low affinity or fail to express enough BCR die by apoposis³¹.

Many of these aspects of GC biology have been elegantly elucidated by experiments with haptens, simple small chemical compounds complexed at high density on a carrier protein³². How does the GC biological engine work in response to complex protein antigens? Particularly antigens from pathogens for which there are no current vaccines, such as HIV Env trimer? What variables are critical to consider when designing vaccines for protein targets? Immunodominance is the natural focusing of an immune response toward a specific number of B cell or T cell clones at the expense of expansion of other epitope specific B or T cells. Immunodominance of B cells specific for non-neutralizing epitopes of viral proteins has recently proven to be a major hurdle in vaccine design to complex viral proteins ^{27–29,33–37}. Why are off-target B cell responses often immunodominant? What strategies can be employed to overcome this immunodominance in vaccine design? This review will cover what is known from current and past in vivo models, as well as recent advances in the understanding of the roles of precursor frequency, antigen affinity, antigen avidity, and other parameters that can affect immunodominance of B cell responses. Further fundamental biological studies to elucidate the mysteries of GC biology will likely help current translational science approaches to develop effective vaccines.

Precursor frequency

Specific knowledge as to how many B cells exist with an epitope specificity against vaccine targets, otherwise called ‘precursor B cells’, is relatively sparse. There exist broadly neutralizing antibodies (bnAbs) against HIV ^38–43, which are directed toward multiple conserved sites on the HIV Env trimer. VRC01-class bnAbs, which target the CD4 binding site of HIV Env, develop in some HIV⁺ patients^39,41,43. It was recently discovered that precursor B cells for VRC01-class bnAbs are present in 96% of humans^44–46. This finding helps give hope that a bnAb-type HIV vaccine is indeed possible. However, in the same studies it was discovered that human VRC01-class naive B cells are relatively rare, existing at a frequency of 1 in 300,000 B cells in the human B cell repertoire. The precursor frequency of B cells with relatively high affinity (< 3μM K_D) is even more rare, at 1 in 10⁶ B cells⁴⁵. While those human B cell findings facilitated advancement of a candidate HIV vaccine immunogen to clinical trial, those findings also raised questions regarding the meaning of abundance of epitope-specific B cells in the naive repertoire, which may relate to the phenomenon of immunodominance. What is the role of precursor frequency in B cell responses? And the natural corollary: If precursor frequency is limiting for desired B cell responses then what strategies can be employed to overcome this limitation?

B cell receptors (BCRs), and T cell receptors (TCRs), are incredibly diverse. BCR rearrangement during B cell development can yield an extraordinary level of diversity^47–49. While an impressive diversity of BCR sequences exist in vivo, experimental models have been sparse that account for how many unique B cells exist for a given epitope, and at what precursor frequency each clonotype exists. It is not true that all epitope-specificities are equally abundant in the repertoire⁵⁰. B cell specific for ‘easy’ to recognize epitopes, such as the hapten NP, are much more abundant than B cells to difficult epitopes such as VRC01-class precursor B cells which bind the CD4 binding site of HIV Env^35,50,51.

Immunodominance is a known phenomenon for antigen-specific CD8 and CD4 T cell responses. T cell immunodominance is largely a function of the abundance of epitope-specific T cells in naive repertoire. There is data available on the importance of precursor frequency on in vivo T cell responses. The precursor frequencies for many different T cell epitopes have been determined and range anywhere between 1 in 10⁵ to 1 in 10⁶ T cells⁵². However, until recently, far less was known about the importance of B cell precursor frequencies in any given immune response. The focus of this review will be on B cell precursors.

B cell biologists have relied predominantly on hapten models such as NP^53–55 or TNP/DNP⁵⁶, to study antigen specific B cell responses in vivo. While powerful, hapten models have limitations in modelling vaccine immune responses to complex protein antigens. Compared to immunization with complex antigens, hapten models are often highly clonally restricted. For example, the anti-NP response in C57BL/6 mice is dominated >90% by B cells utilizing V_H1–72 and Vλ1 genes^53,57. Hapten models also usually have high precursor frequencies of B cells specific for haptens, such as 1 in 4000 B cells for NP ⁵⁰, reflective of the ‘easy’ nature of those epitopes, making it challenging to use hapten models to study how rare precursor frequency B cell responses proceed. Other antigen models, such as fluorescent proteins PE or APC⁵⁸, have precursor B cell frequencies of ~1 in 5000 or ~1 in 25000 for the whole protein (not per epitope). The limited availability of animal models for studying immunodominance and B cell precursor frequencies led to the development of a B cell transfer model using mice containing germline reverted forms of VRC01 (VRC01^gHL)⁵⁹. Wildtype mice do not possess an IgH V gene capable of encoding a VRC01-class Ab. Additionally, the immunization studies made use of engineered outer domain germline targeting (eOD-GT) immunogens for HIV Env, which are protein antigens specifically designed to have affinity for germline versions of VRC01-class bnAbs⁴⁶, currently in human clinical trial⁶⁰ (eOD-GT8 60mer). VRC01-class bnAbs, like most bnAbs, lack measurable binding to normal versions of HIV Env when the bnAb in reverted to an inferred unmutated sequence (germline reverted)⁶¹, suggesting bnAb-class B cells normally undergo a long pathway of affinity maturation in GCs⁴³.

To explore B cell immunodominance in vivo, we employed the VRC01^gHL transfer model system with eOD-GT proteins of various affinities and precursor frequencies. We found that the precursor frequency of epitope-specific B cells was critical for determining whether the B cells successfully competed in GCs following immunization. When precursor frequency was high (1 in 10³ B cells), immunogens with any affinity in the tested range (0.5–40μM) could successfully prime VRC01^gHL B cells to proliferate and enter GCs⁵⁹. In contrast, when precursor frequency was reduced to the human physiological range (1 in 10⁵ to 10⁶ B cells ^45,62), antigen affinity became a separating factor⁵⁹. This data predicts that the first-in-class human clinical trial of eOD-GT8 60mer will be successful and that VRC01-class B cells will be primed in a manner dependent on precursor frequency and affinity. Overall, the data showed that differences in B cell precursor frequency could result in ~1000-fold changes in the ability to initiate an epitope-specific GC B cell response⁵⁹, highlighting the importance of B cell precursor frequency in the process of immunodominance.

This effect of precursor frequency on B cell immunodominance also persisted throughout the course of the GC response, as only the high affinity antigen primed mice were able to undergo substantial clonal bursting in GCs when starting from physiological rare precursor frequencies⁵⁹. This precursor frequency immunodominance effect also applied to the development of memory B cells over time⁵⁹. Similar findings were observed in a study of a different VRC01-class BCR knock-in B cell model, using different germline-targeting immunogens, reiterating the importance of precursor frequency on B cell responses⁶³.

The effect of precursor frequency on B cell responses has been seen in other models of immunization as well. A thorough study by Jenkins and colleagues made use of in vivo B cell dilution experiments using non-transgenic CD45.2⁺ donor B cells into CD45.1⁺ recipients followed by immunization with the fluorescent protein APC⁵⁸. While the B cells transferred were polyclonal and the precursor frequency could only be measured for whole protein and not specific epitopes, the results indicated that the resultant B cell responses following immunization were dependent on precursor frequency⁵⁸. Precursor frequency has also been observed to affect memory B cell formation in a B cell transfer model utilizing B cells specific for the hapten NP¹²⁵. The importance of precursor frequency has also been observed in an elegant study by Lingwood and colleagues¹²⁵ which made use of human knock-in B cell mice expressing IgHV1–69 to study broadly neutralizing anti-influenza B cell responses. The authors showed that these B cell responses were dependent on precursor frequency through B cell transfers as well as diluting the IgHV1–69 allele through crossing to various mouse strains⁶⁴.

Taken together, the data from multiple independent model systems indicate the importance of precursor frequency in B cell responses to any given antigen (Fig 1A). This is relevant for vaccine development when designing vaccines focused on specific epitopes. Precursor frequency measurements should be taken into account when designing vaccine immunogens.

Figure 1: Interdependence of precursor frequency, antigen affinity, and avidity in B cell outcomes.

Models of B cell priming dynamics across ranges of precursor frequency, antigen affinity, and antigen avidity. (A) Model of B cell priming outcomes when precursor frequency is high or low. When antigen-specific B cells are frequent (1 in 10³ B cells) many B cells can easily be recruited to GCs across a wide affinity and avidity range. However, when precursor frequency is low, high affinity multimeric immunogens may be required to prime B cells to successfully compete in GCs. (B) Model of B cell priming dynamics across a range of high to low affinities. (C) Model of B cell priming dynamics across a range of theoretical avidities. A 60mer antigen primed B cells to GCs under conditions when a monomeric antigen did not⁵⁹. (D) Graphical models of interdependence of all 3 parameters on successful priming of B cells.

Antigen Affinity

There is mounting evidence that antigen affinity is one key parameter in determining fate of B cells following immunization (Fig 1A, D). Early studies utilizing NP, as well as studies with hen egg lysozyme (HEL) protein, suggested antigen affinity was a critical parameter in activating B cells. NP specific BCR transgenic B cells are frequently used to study in vivo B cell responses and have a starting affinity of 2μM⁶⁴ for the NP hapten. Early studies utilizing B-1–8-hi or B-1–8-lo transgenic B cells (BCR affinities of 0.2μM or 8μM Kd) found that, when B-1–8 B cell precursor frequency was saturating, there was no distinguishable difference in GC responses mounted by each respective B cell⁶⁵. However, upon transfer of B cells to new hosts, antigen affinity became a separating factor in the magnitude of specific GC responses, although the mutational pattern was conserved⁶⁵. Still, the precursor frequencies were quite high even in that particular B cell transfer model. When tested in a T-cell independent model of B cell activation, antigen affinity still played a role in B cell activation but to a lesser extent⁶⁶. Studies of protein antigens such as hen egg lysozyme (HEL) gave a somewhat more conflicting finding. GC B cell responses were not substantially altered across a range of affinities examined. However, the affinities tested were high, ranging between 100 pM to 100 nM^67,68. It is also worth noting that this observation is in the context of high antigen avidity as these studies utilized HEL that was directly conjugated to sheep red blood cells (SRBCs). Avidity will be discussed in greater detail later in this review. Overall, this finding may indicate that above a certain high affinity threshold that antigen affinity may provide no further benefit for given B cells in GCs. An ultra-high, supraphysiological affinity naive B cell may not be favorably selected to compete within GCs. As speculation, two non-mutually exclusive hypotheses are that this may activate some intrinsic tolerance mechanism that exists within GCs that limits the ceiling of affinity maturation, or ultra-high affinity circulating antibody occludes access to specific epitopes. The VRC01^gHL transfer model was designed to study affinities in the human physiological ranging from 0.5–40μM by utilizing different eOD-GT variant immunogens⁵⁹. In the VRC01^gHL transfer model, priming with an ultra-high affinity (30pM) supraphysiological immunogen as a positive control failed to cause any appreciable outgrowth of VRC01^gHL cells in GCs (unpublished data RKA and SC).

Affinity-dependent selection of B cell responses has also been observed in gut associated lymphoid tissue. A recent study by Shulman and colleagues⁶⁹ investigated parameters affecting B cell selection for entry into GCs in Peyer’s patches (PP)⁷⁰. Utilizing elegant whole organ imaging, it was observed that, while antigen affinity was dispensable for activating and recruiting B cells to the subepithelial dome (SED), affinity was a separating factor for recruiting B cells into GCs⁶⁹. This is evidence that affinity dependence of B cell recruitment to GCs is also conserved in gut associated lymphoid tissues

Low affinity B cells (high micromolar range) in GCs have been noted in multiple studies^71–74, indicating that B cells certainly have the capacity to be activated by very low affinity antigens (10–100uM). However, in these studies, precursor frequency was not specifically tested in parallel to antigen affinity, and thus key information on the competitive landscape was lacking. In the VRC01^gHL B cell transfer model, VRC01-class responses were observed to a 40uM affinity antigen, but only if the B cell precursor frequency was high (~1 in 1,000 to 10,000)⁵⁹. A recent study by Victora and colleagues that investigated the capacity of memory B cells to re-enter GCs upon booster immunizations also noted a correlation between antigen affinity and the B cell clones that re-entered secondary GCs⁶³. While the authors only noted a small re-seeding of secondary GCs by GC derived memory B cells, they did correlate this with higher affinity memory B cells being recalled into the GC⁶³. This is consistent with other reports of precursor frequency and affinity being important for priming B cells to GCs^45,50.

A puzzling observation in multiple experimental models has been that a substantial fraction of GC B cells do not measurably bind antigen by flow cytometry^59,75,76 or antibody binding assays^71,73,77. One interpretation of this is that these B cells are simply too low affinity to be detected by conventional means. Another hypothesis, non-mutually exclusive, is that these “negative” B cells are actually responding to a form of modified or degraded antigen⁷¹. This phenomenon has been coined “dark antigen”⁷¹. The simplest example is a proteolytic breakdown product. Many proteins are rapidly degraded in vivo, at least partially^78,79. Substantial B cell responses to “dark antigen” is plausible, as GC B cells evolved to combat bacterial or viral proteins that are likely produced in widely varying quality and forms in the host. The best evidence that this “negative” population of GC B cells is antigen-specific (not bystander activation) comes from a clever experiment done by Kelsoe and colleagues⁷¹. Separate mice were immunized with two different complex antigens (anthrax PA and flu HA) and the “negative” GC B cells were sorted from each group. If the “negative” B cell response was non-specific bystander B cell activation, or a B cell response to autoantigen, the recovery of the BCR repertoires would be similar. However, to the contrary, the BCR repertoires recovered from these two different immunizations yielded substantially different BCR usage, indicating the “negative” GC B cell response is antigen-specific⁷¹. Whether or not the cells were low affinity B cells for intact antigen or B cells specific to “dark antigen” was not directly determined. “Grey antigen” would be epitopes that are modified but still at least partially present. For example, modified amino acids, glycan changes, or proteolysis that leaves a partial epitope. In those cases, part of the epitope is intact on the full antigen, but not enough to be detectable by most binding assays. To add to the complexity of the interpretation, the day 8 clones in the antigen non-binding populations that seem to dominate the reaction are not dominant on day 16⁷¹. The ability to detect low affinity GC B cells is limited by multiple factors. One is the fact that GC B cell express lower levels of surface BCR than naive B cells or memory B cells^37,80. Taken together these observations offer no definitive answer, but hint that at least these two explanations are indeed possible. Both “dark antigen” and low affinity B cells could make up substantial fractions of any given GC (Fig 2).

Figure 2: Estimated specificities of B cells in a representative GC.

A model diagraming potential composition of a representative GC response over time. See main text for details.

What is a likely composition of a GC, as it relates to antigen affinity and antigen-binding? In early GCs, antigen-binding cells are 26–31% for anthrax PA or flu HA⁷¹, 5% for HIV Env trimer³⁷, and 6–36% for eOD-GT8. Antigen-binding frequencies are higher later in GCs, with 50–55% for anthrax PA or flu HA⁷¹, and 25–50% for Env trimer³⁷. Early GC B cell binding to Env trimer is an outlier most likely because the biotinylated probe largely covers the Env trimer base, and the base is by far the most immunodominant site^34,37,81. Thus, on average, ~25% of B cells in early GCs exhibit detectable binding to antigen, and that fraction increases to ~50% in late GCs or secondary responses (Fig 2). That change over time implies a commensurate reduction in the fraction of cells that originally had undetectable affinity, given the known phenomenon of affinity maturation in GCs (Fig 2).

The GC B cell population consists of more subsets beyond antigen probe-binding B cells, very low affinity B cells, and dark antigen specific B cells. One subpopulation of GC B cells is, of course, “background” GC B cells, specific to antigens present in the animal that are unrelated to the immunization (Fig 2). This fraction will vary greatly depending on circumstances and anatomical site (e.g. sterile LN vs Peyer’s Patches). For example, it may be expected that GC background is substantially higher in non-human primates than in specific-pathogen-free mice due to housing conditions. The magnitude of this subpopulation may be especially large for late GC time points, as the vaccine-specific GC response fades. An additional subpopulation of GC B cells express non-functional BCRs (Fig 2) due to recent random SHMs introducing stop codons or structure-disrupting mutations. An elegant study by Nussenzweig and colleagues³¹ showed that many B cells in GCs that undergo apoptosis in the DZ die due to mutations creating non-functional BCRs. B cells are known to require tonic signaling through the BCR to survive^82,83. 5–17% of GC B cells had completely nonfunctional BCRs³¹. An additional 5% were severely structurally compromised, with that potentially being an underestimate. That provides a current best estimate of 10%−25% of GC B cells having nonstructural BCRs due to SHM at any given moment in time (Fig 2), essentially all of which die within 6 hours, and those are immediately replaced by new SHM damaged B cells³¹. Taken together, the GC consists of multiple subpopulations of B cells that do not have measurable affinity to the intact antigen used for immunization (Fig 2). Understanding these subpopulations is relevant for understanding the role of affinity in B cell immunodominance to complex antigens.

In summation, antigen affinity has been investigated as an important parameter in B cell responses in several independent experimental model systems and this parameter has stimulated great debate as to what affinity is necessary to prime successful B cell responses. However, only until recently has affinity been investigated in the context of other variables such as precursor frequency and avidity as well as complex vaccine relevant antigens. When accounting for these variables, this may help explain some previously observed phenomena (e.g. ultra-low affinity B cells in GCs). This is discussed further in the next section.

Interrelationship of Affinity and Precursor Frequency

We hypothesized that the role of antigen affinity was dependent on the precursor frequency of B cells in the existing naive B cell repertoire for a given epitope. As described above, studying broadly neutralizing antibody B cell biology utilizing the VRC01^gHL B cell transfer model, we immunized mice containing different precursor frequencies ranging from 1 in 10³ B cells to 1 in 10⁶ B cells, and immunizing with immunogens of different affinities ranging from 0.5 and 40μM K_D. Those parameters were selected based on numbers measured directly for human B cells^45,62. When the epitope specific naive B cell precursor frequency was high in recipient mice, both high and low affinity B cells were readily detected in GCs. Only a small advantage was observed with the high affinity condition. However, when the epitope specific naïve B cell precursor frequency was reduced to the physiologically relevant 1 in 10⁶ B cells, antigen affinity became a major distinguishing factor for priming VRC01^gHL B cells⁵⁹. At this rare precursor frequency of 1 in 10⁶, the high affinity resulted in >100-fold increased representation within the GC compartment compared low or medium affinity conditions⁵⁹. Another study confirmed this interdependent relationship between precursor frequency and affinity in eliciting GC B cell responses against different antigens and VRC01-class B cells⁶³. Separately, using a germline-targeting for N332-supersite on HIV Env trimers showed that high affinity antigen designs could very effectively elicit expansion of desired bnAb precursor B cells under rare precursor frequency conditions²⁰. Thus, both affinity and B cell precursor frequency can substantially affect B cell immunodominance and outcomes in vivo (Fig 1A, D).

Understanding how truly naïve human precursor will respond in vivo is of particular interest for the sake of understanding the role of affinity in B cell responses. This is also important for improving predictive power of current animal models to inform vaccine design efforts. Inferred germline (iGL) mouse models are very useful in studying lineage specific antibody responses to candidate HIV immunogens. However, they do suffer from limitations. To address this challenge, we developed three new knock-in mice⁸⁴ utilizing authentic naïve BCR sequences recovered from HIV-negative healthy human donors^45,62. These B cells had specific affinities for the candidate HIV immunogen eOD-GT8 which spanned the human physiological affinity range (125nM→16μM K_D) and were considered VRC01-cass based on their usage of VH1–2*02 allele and 5 amino acid L-CDR3. Three different light chains were used (VK3–20, VK1–33, VK 1–5) which are representative of three of the major light chains used by VRC01-class bnAbs⁸⁵. These new mice are referred to as HuGL mice (human germline naïve B cell derived). In order to model human epitope-specific B cell precursor frequency conditions, we tested the ability of these HuGL B cells possessing authentic human naïve B cell BCRs to respond to antigen in vivo at precursor frequencies down to 1 in 10⁶ B cells⁸⁴. B cells with high (125nM K_D HuGL18^HL) or medium (1.3μM K_D HuGL17^HL) affinity BCRs were primed, recruited to GCs, accrued substantial SHM, and formed memory B cells⁸⁴. Precursor frequency and affinity interdependently influenced these responses⁸⁴. Taken together, these experiments utilizing authentic naive human VRC01-class BCRs⁸⁴ validate key aspects of the eOD-GT8 60mer germline-targeting design, highlight the importance of precursor frequency and affinity, and validate a central tenet of the germline-targeting approach to vaccine design, which is also more broadly applicable to reverse vaccinology 2.0 approaches.

Antigen Avidity

Antigen avidity has long been known to affect antibody responses following immunization or infection in that higher avidity antigens or viruses substantially improve the magnitude of antibody responses⁸⁶. In vitro studies have revealed that BCRs can engage with antigen in a monovalent fashion and induce activation, but is inefficient at inducing antigen presentation^87,88. This is in agreement with earlier studies which observed substantially more activation of B cells when treated with anti-BCR antibody compared to monovalent fragments of antibodies⁸⁹. However, there is still substantial debate about this. Other studies have not seen activation of B cells utilizing B cells specific for different proteins or haptens under monomeric conditions^90,91. Structurally, there are approximately ~200,000 BCRs on the surface of any given resting B cell⁹². It is logical that B cells can discriminate between not only different affinity immunogens but different avidity immunogens. There is some evidence to support this. In vitro studies of BCR revealed that B cells utilized mechanical energy and contract via myosin IIa to acquire and internalize antigen only when it was both high affinity and high avidity⁹³ and that B cells preferentially use mechanical force to extract antigen⁹⁴. Studies of the development of anergic B cells utilizing HEL specific MD4 mice crossed to mice that expressed their cognate antigen hen egg lysozyme (HEL) in either soluble or surface bound format found that surface bound HEL was far more potent at inducing an anergic state or deletion of B cells⁹⁵, indicating antigen avidity is a critical parameter that B cells consider during development.

In the context of immunization, antigen avidity was essential for activation of VRC01^gHL B cells in vivo, particularly at physiologically relevant precursor frequencies⁵⁹. In an elegant study by Irvine and colleagues, nanoparticle antigens were rapidly shuttled to, and concentrated within, GCs in a manner dependent on mannose binding lectin while monomeric antigen was inefficient at doing this⁹⁶. Avidity has also proven to be beneficial in B cell responses to vaccine designs for respiratory syncytial virus (RSV). A study by King and colleagues⁹⁷ recently reported the success of a nanoparticle of F glycoprotein trimers containing 20 trimers which could induce neutralizing antibodies in both mice and monkeys. This nanoparticle induced neutralizing antibody responses that were ~10 fold higher than trimeric RSV F protein. Wyatt and colleagues have also used avidity to improve antibody responses against HIV Env by coupling engineered stabilized trimers to liposomes⁹⁸. This approach significantly improved B cell responses and has the added advantage of occluding the HIV Env trimer base, which is a major site of off-target antibody responses^34,35. Avidity can also be co-opted to be used as a threshold for selecting B cells containing broadly reactive potential for activation. Graham and colleagues⁹⁹ developed a nanoparticle influenza vaccine candidate for influenza that consisted of a mosaic of 8 different HA strains displayed on the same nanoparticle. The authors hypothesized that this mosaic of proteins on the same nanoparticle would preferentially select for B cells that had broad neutralization capacity and would effectively use antigen avidity as a threshold for selection⁹⁹. This strategy worked in mice as the authors identified broadly neutralizing mouse antibody to influenza⁹⁹, providing proof-of-concept that substantial effects of antigen avidity on B cell responses can be co-opted through clever design to focus epitope-specific B cell responses.

While harnessing the effects of valency on B cells can be powerful, individual custom design of multivalent immunogens is a major hurdle. Therefore, having an immunization strategy that is easily adaptable to many proteins and having a clear path to use in humans is highly desired. One promising approach is the utilization of phosphoserine tagged immunogens. Phosphoserine tagged (pSer) monomeric proteins directly bind to alum adjuvant when mixed together (pSer:alum), creating an avidity effect and possible slow-delivery effect. 100-fold increases in VRC01-class GC B cell and memory B cell responses were generated¹⁰⁰. B cells were able to directly uptake antigen-coated alum particles. Increased avidity of pSer:alum nanoparticles is one of the mechanisms by which this new vaccine platform works to potentiate humoral immunity. However, pSer:alum actually positively influences B cell responses via multiple additional mechanisms of action. One mechanism of pSer:alum is the slow delivery of immunogen, which potently affects GCs¹⁰⁰. This is discussed further in the next section. Additionally, the pSer tag forces a specific orientation of the antigen on the alum microparticle, creating a masking effect on undesirable epitopes as well as increasing avidity. Epitope shielding as a strategy to combat immunodominant off-target B cell responses is discussed in the final section. Altogether, antigen avidity is clearly a critical factor in promoting B cell responses following vaccination (Fig 1C). Contributions of avidity are essential for activating B cells when precursor frequency is low (Fig 1A-C). Only with an ultra-high picomolar affinity could a monomeric protein prime rare VRC01-class B cells⁵⁹. More broadly, all three variables—affinity, precursor frequency, and valency—are interdependent and should be taken into account in understanding successful B cell responses (Fig 1 D), and for designing vaccines capable of eliciting such responses. Taken together these studies in multiple different experimental models reiterate the importance of antigen avidity on successful B cell responses following vaccination. Further studies into how antigen avidity affects B cell responses to complex antigens in vivo, as well as a more complete understanding of how BCRs interact with antigen, will aide in future vaccine design efforts.

T cell help

T cell help is another critical factor required for GC B cell responses. It is known that the strength of T cell help can regulate entry of individual B cells to GCs¹⁰¹ as well as the speed of cell cycle of individual GC B cells after positive selection within the GC¹⁰². T cell help is absolutely required for GCs^17,103. In the absence of T_FH cells, via deletion of BCL6 in CD4 T cells, GC B cells are lost^104–106 . Conversely, artificially increased T cell help can favor expansion of B cells in GCs¹⁰. A study by Kelsoe and colleagues¹⁰⁷ has raised questions as to how much of an increase in p:MHC is required for favorable expansion of given B cells in a GC. B cells with MHCII haploinsufficiency (which equated to a ~50% loss of surface MHCII) had reduced ability to enter GCs compared to WT B cells¹⁰⁷, consistent with earlier findings that early T cell help is a critical checkpoint for early GC entry¹⁰¹. However, the authors found that within GCs there was no competitive advantage of WT B cells over MHCII haplo-insufficient B cells. These findings suggest that a two-fold increase in p:MHC complexes is not enough of a difference in antigen presentation for T_FH cells to discriminate between two different GC B cells¹⁰⁷.

T_FH help has been positively associated with reduced B cell immunodominance in multiple studies of HIV Env immunized non-human primates^36,37,108. A recent study by Pulendran and colleagues utilized the classic advantage of hapten models for studying T cell help¹⁰⁹ and B cell immunodominance. The authors immunized animals with either hapten-protein conjugates in adjuvant alone, or with in combination with excess carrier protein¹⁰⁹. Increased doses of carrier protein increased antigen competition, and diminished the hapten specific GC B cell and antibody response¹⁰⁹. To elegantly and cleverly attribute this to B cell competition for T cell help, as opposed to simply increasing B cell responses to carrier protein, the authors utilized the immune epitope database (IEDB) tool^110,111 to predict MHCII peptides and develop mutant carrier proteins which had impaired potential for being a source of p:MHC¹⁰⁹. Critically, such small modifications to the carrier proteins are unlikely to reduce or substantially change the magnitude of the off-target B cell responses. Addition of wild type carrier proteins suppressed hapten specific GC responses, while addition of mutant carrier proteins did not, strongly suggesting that this epitope-specific effect was due to CD4 T cell help¹⁰⁹. However, whether T_FH cell help can be modulated to favor selection of rare precursor B cells in vivo remains to be determined.

Additional factors affecting B cell competition

Clearly there are a number of factors that affect antigen-specific B cell responses in vivo (Fig 3). Further work with models that allow for testing of individual parameters in isolating, quantitatively recapitulate human physiological conditions, and utilize biomedically relevant complex antigens will likely reveal how these different parameters interact and which are most important in vivo. Recent studies have found that the kinetics of B cell access to antigen can have profound effects on the B cell composition of the GC response^34,37,112. In a mouse model utilizing osmotic pumps to slowly immunize with HIV Env trimers, the B cell response was redirected away from non-neutralizing V3 loop-specific responses which is only present on degraded trimers³⁴. A similar phenomenon was observed when osmotic pumps or continuous dosing were used as slow delivery immunization strategies to deliver HIV Env trimers in rhesus monkeys, resulting in substantially higher HIV neutralizing antibody titers³⁷.

Figure 3: Immunodominance factors involved in development of protective B cell responses.

Precursor frequency, BCR affinity, and antigen avidity are interdependent variables in determining B cell success in GCs following immunization. Additional parameters such as method of immunization, T cell helper epitopes, and adjuvant properties can also be important factors influencing outcomes of immunodominance.

Circulating antibody also plays a role in modulating B cell responses post immunization^113–116 . Perhaps the best example of this is influenza virus strain specific responses. Preexposure to previous variants of the flu yields circulating antibody that can inhibit B cell responses to epitopes on new variants of the flu, a term coined ‘original antigenic suppression’^117,118. The role of circulating antibodies on GC responses has been assessed by Toellner and colleagues¹¹⁹. High affinity monoclonal antibodies injected into immunized mice limited access to antigen and focused higher affinity GC B cell responses¹¹⁹. This antibody feedback mechanism on GCs could also serve to synchronize affinity maturation across multiple GCs which are anatomically isolated islands as well as partially explain how GCs terminate over time¹¹⁹. A separate study by Heyman and colleagues¹²⁰ found that epitope-specific antibodies only suppressed responses toward this epitope, suggesting epitope masking is a major mechanism by which circulating antibody can regulate GC responses¹²⁰.

Strategies to overcome immunodominance

B cell immunodominance is not an immutable phenomenon. Structural design approaches are being widely employed to silence immunodominant non-neutralizing epitopes in HIV and flu vaccine protein designs, in an effort to enhance B cell responses to subdominant, difficult neutralizing epitopes. While stabilizing mutations have been used to present HIV Env timers in conformationally correct manners, other approaches have also been utilized to re-order immunodominant epitopes. One approach has been the addition of glycans (called “glycan masking”) to exposed protein surfaces that are non-neutralizing. This approach has succeeded in reducing access to different epitopes in the design of HIV eOD-GT8⁴⁴. It has also had success in enhancing desired VRC01-class B cell responses during immunization¹²¹. Another approach has been developed by Kim and colleagues¹²² called “protect, modify, deprotect” in which the desired epitope is bound to an antibody, the immunogen is then pegylated to reduce immunogenicity of all other sites, and then the antibody is removed to leave the desired epitope intact. Another cleaver strategy that has been used on influenza antigens is modifying the immunization strategy utilizing different variants of the immunogen¹²³. Angeletti and colleagues conducted immunization studies with variants of influenza HA head and stem protein to understand if modifying the composition of the immunizing antigen can favor development of B cell responses toward the subdominant stem. Several other strategies exist that are capable of modulating B cell immunodominance, such as slow delivery of immunogen³⁷, pSer tagging of proteins¹⁰⁰, as well as mosaic nanoparticle formulation⁹⁹ that were described in detail in previous sections. Another strategy has been to utilize intrinsic tolerance mechanisms of the GC¹²⁴ to favor expansion of subdominant B cells by injecting soluble antigen consisting of only the immunodominant epitope⁷⁶, thus re-ordering B cell immunodominance in an ongoing reaction.

Open questions

There are many open questions pertaining to B cell immunodominance. What is the role of T_FH frequency in competitive fitness of GC B cells? What is the role of antigen valency in activating different B cells to compete in GCs? How do intermediate avidity immunogens affect competition of individual B cell precursors within GCs to complex antigens? What is the composition of the ‘dark antigen’ component of the GC? Can pharmacological means be used to “tune” the GC reaction through modulation of T_FH help or otherwise to be more permissive to outgrowth of desired bnAb precursors? How can circulating antibody to prior immunization steps be overcome to elicit desire recall responses? Future work by the field will have to answer many of these questions and more. The available data accumulated over the past few years clearly shows the interdependent importance of precursor frequency, affinity, and avidity in determining B cell outcomes (Fig 1), but precise relationships remain to be defined, particularly for valency. The availability of new animal model platforms that take into account salient quantitative features of B cell responses to complex antigens and provide definitive “yes-no” outcomes (e.g. GC competitive fitness, memory B cell formation, SHMs accrued) for prospective immunization strategies will be critical. Accounting for known precursor frequencies and affinities to those found in the prospective human vaccine pool will empower the field to answer these questions in relevant ways.