Basics of memory B‐cell responses: lessons from and for the real world

Summary

The production of pathogen‐specific B cells and antibodies underlies protective immunity elicited by most vaccines and many infections. Humoral immunity follows a regulated process by which high‐affinity antibody‐secreting plasma cells and memory B cells are generated. Yet for certain pathogens, protective immunity is inefficiently generated and/or maintained. For example, Dengue virus infections lead to lasting immunity against re‐infection by the same serotype. However, if infected with a different Dengue serotype, the individual is predisposed to more severe disease than if he/she was completely naive. As another example, both natural infections with or vaccination against malaria do not necessarily lead to lasting immunity, as the same individual can be re‐infected many times over the course of a lifetime. In this review, we discuss how these real‐world problems can both instruct and be informed by recent basic studies using model organisms and antigens. An emphasis is placed on protective epitopes and functional distinctions between memory B‐cell subsets in both mice and humans. Using flavivirus and Plasmodium infections as examples, we also speculate on the differences between ineffective B‐cell responses that actually occur in the real world, and perfect‐world responses that would generate lasting immunity.

Introduction

Protective immunity elicited by almost all vaccines and many infections is mediated by pathogen‐specific B cells and antibodies. The generation of such protective cells and antibodies is mediated by a process of clonal expansion and diversification of fates, followed by a contraction to the essential components of durable immunity. For T‐cell‐dependent responses, naive B lymphocytes that recognize cognate antigens rapidly proliferate and differentiate into short‐lived plasma cells or germinal center B cells.1 Short‐lived plasma cells secrete low‐affinity antibodies that are important for the initial defense against pathogens such as West Nile virus,2, 3 and are thought to live for only several days.4 Germinal center B cells progressively affinity mature their B‐cell receptors through multiple rounds of somatic hypermutation and selection.5, 6, 7 Cells carrying affinity‐enhancing mutations can then differentiate into memory B cells or long‐lived plasma cells.8, 9

Long‐lived plasma cells constitutively secrete affinity‐matured antibodies that pre‐exist subsequent infections, thereby providing sterilizing immunity.10, 11 These long‐lived plasma cells can live up to an entire human lifetime in the absence of cellular division.12 Yet, their persistence varies greatly with the specific infection or vaccine for reasons that still remain unclear. Metabolic, but not transcriptional, pathways appear to be the major determinants of plasma cell longevity,13, 14 but how infection‐ or vaccine‐specific properties such as antigen avidity, innate immune activation, and T‐cell help influence these programs remain unknown. Pathogens that evade pre‐existing serum antibodies may be recognized by memory B cells, which make up a second line of defense.15, 16 Memory B cells are quiescent, do not require continued antigenic stimulation for survival, and are more easily reactivated than are their naive counterparts.17, 18 The molecular and cellular details of memory B‐cell development have been recently reviewed,19, 20 and hence, the focus of this article will be on memory B cells during recall responses in the context of real‐world infections.

During a recall response, reactivated memory B cells can differentiate into antibody‐secreting cells and/or germinal center B cells, depending on the responding subpopulation. In mice, memory B‐cell subpopulations were initially distinguished by immunoglobulin isotype expression.21, 22 These studies showed that IgG⁺ memory B cells preferentially generate plasma cells, whereas IgM⁺ cells re‐initiate germinal center reactions. Subsequent studies reported that CD80 and programmed cell death 1 protein ligand 2 (PD‐L2) expression on antigen‐specific cells are functional markers that distinguish memory subsets irrespective of isotype.23 CD80⁺ PD‐L2⁺ cells are poised to differentiate into antibody‐secreting cells, CD80⁻ PD‐L2⁻ cells robustly differentiate into germinal center B cells, and CD80⁻ PD‐L2⁺ cells possess an intermediate ability to differentiate into antibody‐secreting cells and re‐initiate germinal center reactions. These markers mostly, but not fully, correlate with IgM (enriched in the CD80⁻ PD‐L2⁻) or IgG (enriched in the CD80⁺ PD‐L2⁺) usage. Moreover, the specific isotype used by memory B cells correlates with unique transcriptional programs that support function and development.24 It is likely that the isotype is a marker of the innate signals received by memory B cells during ontogeny rather than a causative determinant of a specific transcriptional program.25 In humans, memory B cells are identified based on CD27 expression,26, 27 and as in mice, their fate upon reactivation can be predicted by surface immunoglobulin expression. Similar to mouse memory B‐cell subpopulations, IgG⁺ human memory B cells are poised to differentiate into antibody‐secreting cells, and IgM⁺ memory B cells are predisposed to form germinal center‐like cells ex vivo.28 This correlation is reassuring, because mouse models allow for a deeper mechanistic understanding of cellular behavior, whereas studies in humans bring a real‐world relevance to these findings. Another subset composed of atypical memory B cells has been identified in tonsils.29 These cells express neither CD27 nor CD21 and express high levels of inhibitory receptor FcRL4 or FcRL5.29, 30 These cells tend to be hyporesponsive to B‐cell receptor (BCR) stimulation and are expanded in the blood in chronic infections like malaria and HIV.31, 32 Understanding the recall behavior of each memory B‐cell subpopulation in different infections is important for determining how to establish protective immunity, particularly when faced with the challenge of vaccinating a population that is already partially immune to an endemic pathogen.

In this review, we focus on flavivirus and Plasmodium infections as they pose some unique challenges for generating immunity. As a result, there are interesting lessons to be applied to the basic study of memory B cells. Reciprocally, principles from the cell biology of memory B cells can be potentially applied to vaccination efforts. As examples of the challenges that these globally relevant pathogens pose, infections of flavivirus‐immune individuals by heterologous or heterotypic strains can result in markedly exacerbated symptoms compared with the primary challenge. Malaria, caused by Plasmodium infections, is characterized by the lack of a durable antibody response and requires multiple exposures to develop naturally acquired immunity. For each infection, we will discuss the underlying antibody and memory B‐cell responses, speculate on the ideal memory B‐cell response that considers the challenges faced, and draw conclusions on implications for vaccine design and remaining questions. We fully acknowledge that many aspects of this review are speculative. Yet, we believe it is imperative to apply the fundamentals of memory B‐cell biology to contemporary, problematic infections to better guide vaccine design and future research.

Flavivirus pathogenesis, epidemiology, and immunity

Flaviviruses present a global threat to public health, especially with the recent emergence of epidemic Zika virus (ZIKV). Among many relatives, members of the Flavivirus genus include the human pathogens ZIKV, West Nile virus (WNV), Dengue virus (DENV), and Japanese encephalitis virus (JEV). These viruses are mainly transmitted by mosquitoes, and for WNV and JEV, humans are a dead‐end host. For the epidemic pathogens YFV, DENV, and ZIKV, viral titers in humans can reach sufficient levels that these pathogens can be re‐transmitted by mosquitoes or by direct human–human contact.33, 34 Most infections are asymptomatic or present mild symptoms, such as fever, arthralgia, and myalgia. However, some cases of severe symptoms, such as severe hemorrhagic fever and vascular leakage, have been reported. These severe symptoms have been largely associated with secondary DENV infections.

There are four DENV serotypes, DENV1 to DENV4, all of which co‐circulate in the same geographic regions. Work by Sabin in the 1950s showed that DENV infection by one serotype provided lifelong protection against homotypic infection, but not against heterotypic infections.35 Indeed, heterotypic infections increase the severity of symptoms when infections occur after antibodies generated from the primary challenge waned.36 Hence, primary DENV infection generates a durable, serotype‐specific antibody response that can be harmful to the host upon heterotypic challenge. The increased severity of secondary infections is thought to be mediated by antibody‐dependent enhancement (ADE), a process whereby antibodies that neutralize poorly, either due to epitope specificity or insufficient concentrations, enhance viral uptake through Fcγ receptors on mononuclear phagocytes.37, 38, 39 In addition to the humoral contribution (ADE) to increased disease severity upon heterologous infection, there is also a cellular contribution termed original antigen sin (OAS).

The OAS hypothesis was first described as ‘the imprint established by the original virus infection governs the antibody response thereafter’,40 whereby memory B cells from the primary infection are activated during subsequent infections. When antigenic determinants vary between strains, these memory B cells can often bind only weakly and provide poor protection to the second infection. Yet, by virtue of reduced activation requirements, these ineffective recall responses dominate over primary naive B cells. For DENV, OAS was first described after observations that while serum antibodies had varying degrees of neutralizing activity to all four DENV serotypes after heterotypic DENV infection, potent neutralization only occurred to the primary infecting serotype.41 Similar observations have since been made for memory T cells.42 Given the overlap in the geographical prevalence of many flaviviruses, increases in travel, and a potentially exacerbated immune response to secondary infections, there is a need for understanding recall responses to heterologous flavivirus infections. To understand why non‐protective antibodies dominate secondary responses, we must first discuss the different antibody epitopes in flavivirus infections.

Non‐protective and protective antibodies in response to flavivirus infections

Most flavivirus antibodies target epitopes found on the envelope (E) protein, which mediates cellular attachment and membrane fusion to the host cell.43 The E protein contains three different domains (DI to DIII) with both non‐protective and protective epitopes.44 The primary, non‐protective epitope on the E protein is the DII fusion loop (DII‐fl), which is highly conserved across flavivirus species.45, 46, 47 In humans, this is the immunodominant epitope,46 and explains why antibodies from one flavivirus infection can cross‐react with other flaviviruses that have yet to be encountered.48 In addition, these DII‐fl antibodies are frequently poorly neutralizing. For many other types of infections, the presence of such non‐neutralizing antibodies is at worst harmless and at best protective due to engagement of alternate antibody effector mechanisms such as complement deposition and opsonization. Yet, ADE renders such antibodies pathogenic in acute secondary flavivirus infections.37 How these non‐protective epitopes impact the cellular response in secondary infections is discussed next.

As mentioned above, in humans, antibodies generated during primary infections are mostly directed against DII‐fl and have poor neutralization ability.46 During acute secondary infections with a different DENV serotype, plasmablasts secrete cross‐reactive, E‐specific antibodies with varying degrees of ADE.49, 50 These antibodies carry high levels of mutations, suggesting that they originate from memory B cells. Memory B cells established after primary and secondary infections appear to retain a greater breadth of antigen specificities (E, prM, and complex epitopes) than do the responding plasmablasts (predominately E).51, 52 Hence, only a subset of memory B cells differentiates into antibody‐secreting cells, where the majority of antibodies are poorly neutralizing with high ADE activity. Similar findings were observed following ZIKV infection of DENV‐immune patients.53, 54, 55, 56, 57 Vaccines to protect against flavivirus infections have so far revealed similar problems. A licensed tetravalent Dengue vaccine, CYD‐TDV, can mount a neutralizing response to all four DENV serotypes.58 However, a longitudinal study on the efficacy of CYD‐TDV found an increased risk of severe disease in previously unexposed individuals compared with those who had previously been exposed to at least one DENV serotype through natural infection.59 It is probable that CDY‐TDV promotes ADE upon natural secondary DENV infection, at least in a subset of individuals who were naive before vaccination. These findings demonstrate the importance of engaging B cells with strongly neutralizing specificities while avoiding those that preferentially enhance infection.

While non‐protective epitopes dominate the human antibody response, mouse studies have provided critical insight on protective epitopes that can potentially be targeted selectively. For example, mouse studies have shown that B‐cell responses are critical for protection against flavivirus infections.2 Passive transfer of minute quantities of immune serum from WNV‐convalescent mice into B‐cell‐deficient animals confers protection to homologous virus infection.2 Analysis of mouse monoclonal antibodies has revealed that the most potent neutralizing epitopes are located on the lateral ridge epitope of DIII (DIII‐lr).45, 60, 61, 62, 63 Interestingly, these DIII‐lr‐specific antibodies are immunodominant in mice, whereas in humans, these specificities clearly exist but are overwhelmed by DII‐fl‐reactive cells and antibodies.38, 45, 46, 64 The DIII‐lr epitope is only weakly conserved across flavivirus species, so long‐lived plasma cell DIII‐lr responses provide type‐specific immunity with minimal ability to promote ADE.61, 63 Another protective and highly neutralizing epitope is the recently described E dimer epitope (EDE), present in both DENV and ZIKV.65, 66 Analysis of anti‐EDE antibodies isolated from DENV‐infected humans found that these antibodies potently and broadly neutralized all four DENV serotypes and ZIKV.53, 65, 66 This epitope is present on intact, mature virions but not on recombinant E, indicating that a quaternary structure is required for antibody recognition. Because of the complexity of this epitope, an equally complex strategy will likely be required to selectively generate EDE‐specific antibody responses.

Diversity of memory B cells and their reactivation in secondary challenges

Interestingly, mouse studies on WNV infections and vaccinations have demonstrated a greater diversity of DIII antigenic specificities in memory B cells than in long‐lived plasma cells.16 This fundamental difference in repertoires between memory B cells and constituent serum antibodies has also been found in humans following a range of different vaccines.15 The diversity of WNV‐reactive memory B cells imparts them with the ability to cross‐react to JEV DIII upon secondary challenges.16 These data suggest an ability of DIII‐specific memory B cells, but not pre‐existing long‐lived plasma cells, to recognize DIII‐lr of related flaviviruses upon secondary exposures. This could lead to potent cross‐neutralizing responses if these memory B cells were to be selectively engaged. Several innovative potential solutions are possible. For example, Richner et al. used a modified RNA to express replication‐incompetent ZIKV subviral particles carrying a mutation in the non‐neutralizing DII‐fl epitope.67 Mice challenged with this modified RNA were resistant to subsequent DENV2 challenge, whereas animals immunized with wild‐type ZIKV subviral particles were highly susceptible to heterologous infections. A second potential approach involves the use of DIII subunit vaccines. Multivalent or sequential DIII‐based vaccines could potentially be given to specifically elicit DIII‐lr neutralizing antibodies, either from naive B cells or pre‐existing memory B cells in individuals that are already immune to at least one flavivirus strain. In seropositive individuals, responses to heterologous DIII‐lr epitopes would require clonal expansion and plasma cell differentiation of pre‐existing cross‐reactive DIII‐lr‐specific IgG⁺ memory B cells (Fig. 1). Alternatively, engagement of IgM⁺ memory B cells specific for DIII‐lr of one strain could re‐initiate germinal centers to affinity mature towards DIII‐lr of the new strain (Fig. 1). A real‐life example where cross‐reactive DIII‐lr‐specific antibodies were expanded in secondary infections involved ZIKV infection in DENV1‐experienced individuals.56 DENV1 and ZIKV share two critical amino acids in DIII, and hence memory B cells recognizing this epitope were selectively expanded. Tapping the potential of pre‐existing DIII‐lr memory B cells may guide the best vaccination strategy to elicit potently neutralizing antibodies. Yet, different considerations will need to be employed for vaccines that fail not because of their epitope specificity, but rather due to the lack of durable immunity elicited, as has been observed in malaria.

Figure 1.

Actual and ideal responses to heterologous flavivirus infections. Secondary flavivirus infections in humans can result in pathogenic responses, where the immunodominant epitope is the highly conserved fusion loop on the envelope protein domain II (DII‐fl). Non‐neutralizing DII‐fl antibodies facilitate increased viral load through a process called antibody‐dependent enhancement (ADE). Ideally, the response to heterologous challenge should result in the expansion of protective antibodies and memory B cells. This can either be achieved by recruiting DIII‐lr cross‐reactive IgM⁺CD80⁻PD‐L2⁻ memory B cells into the germinal center (GC) response to affinity mature towards the new heterologous DIII‐lr epitope, or by differentiating pre‐existing cross‐reactive DIII‐lr specific IgM⁺CD80⁺PD‐L2⁺ memory B cells into long‐lived plasma cells (LLPCs).

Durability of plasma cell survival and antibody responses – the Achilles heel of malaria vaccines

Malaria is caused by the Plasmodium parasite, and the species Plasmodium falciparum is the leading cause of death and disease in many developing countries, particularly in Africa.68 Over half of the world's population is at risk for malaria, where over 200 million cases of clinical malaria occur each year.68 In addition, over 400, 000 deaths, primarily in children, occur each year.68 Protection against re‐infection depends on the presence of long‐lasting, neutralizing serum antibodies,69 as passive transfer of immune serum into actively infected individuals results in rapid recovery.70 However, Plasmodium infections themselves do not reliably generate durable antibody responses;71 not only are antibodies inefficiently generated, they are lost in the absence of consistent repeated exposure.72 Given the high infection and mortality rates, especially in sub‐Saharan African children, development of a highly effective vaccine is of utmost importance, but the provision of durable immunity poses a major challenge as described more below.

The most advanced approach is the RTS,S subunit vaccine containing the circumsporozoite protein (CSP). Preliminary studies revealed that vaccination in malaria‐naive adults yielded robust antibody responses in which 86% were protected from subsequent Plasmodium challenge.73 Though CSP‐specific antibody titers gradually waned, these initial findings showed promise for a realistic and scalable vaccine against malaria. However, a study analyzing vaccine efficacy in the field found that only 34% of Plasmodium partially immune Gambian adults given the vaccine were protected from the first round of natural infection.74 Within 6 months after RTS,S vaccination, both antibody titers and epidemiological protection waned to negligible levels.75, 76 Though difficult to conclude with certainty without a side‐by‐side comparison, these data suggest that the ability of Plasmodium‐naive individuals to mount and maintain a protective response after vaccination is higher than that of Plasmodium‐experienced individuals. These findings further suggest that natural Plasmodium exposures may alter the ability of B cells to respond, either directly or indirectly. Alternatively, a response against CSP may intrinsically be short‐lived (discussed below). Several elegant studies have followed the B‐cell response over time in malaria‐endemic areas and have provided intriguing clues as to the source of the dysfunction.

Atypical memory B cells in Plasmodium‐infected individuals

Repeated exposures to Plasmodium over the course of multiple wet seasons lead to a progressive increase in antibodies, but this protective humoral immunity rapidly fades during dry seasons as the mosquito population diminishes.77 There is also a progressive increase in the frequency of atypical memory B cells, balanced by a relative decrease in classical memory B cells.32, 78, 79 These atypical memory B cells might be the key to understanding why antibody responses are so transient following vaccination or repeated natural infections.32 Originally identified in HIV‐infected patients by expression of the inhibitory receptor FcRH4, these cells fail to proliferate in response to BCR ligation.29, 80 Moreover, studies of the mouse age‐related memory B cells, the potential analog to human atypical memory B cells, have shown markedly elevated expression of the transcription factor Zbtb32.81 This factor suppresses the duration of plasma cell survival following memory B‐cell recall responses.82 Hence, it might be that recall responses mounted by atypical memory B cells are inherently transient.

There remain some conflicting findings regarding the biological properties of atypical memory B cells. For example, analysis of a relatively small number of cells suggested that atypical and classical memory B cells were clonally unrelated in Plasmodium‐immune individuals,83 whereas another larger study found these two cell types possess overlapping immunoglobulin repertoires.80 Some studies observed immunoglobulin secretion by atypical memory B cells both in vivo and ex vivo,83 whereas others have found these cells to be completely unresponsive to stimuli.80 A recent study by Pérez‐Mazliah et al. analyzed the MSP1‐specific memory B‐cell compartment following Plasmodium infection and found that atypical memory B cells were short‐lived and had activated B‐cell and plasmablast gene signatures.84 Mouse studies have not resolved these issues, as age‐associated memory B cells can readily respond to Toll‐like receptor stimulation,85, 86 whereas in some studies, human atypical memory B cells do not.87 Clearly, additional studies must be performed to explain the differences in these sets of studies. This in turn will help to define whether these cells should be intentionally avoided in vaccine strategies, or whether they can be safely ignored.

A second finding that may impact the transience of immunity is the emergence of an unusual T follicular helper (Tfh) cell phenotype. Mouse studies have shown that a pro‐inflammatory environment promotes T‐bet expression in pre‐Tfh cells, which skews toward a T helper type 1‐like phenotype.79 Similarly, human circulating Tfh‐like cells from Plasmodium‐immune individuals are poised to secrete interferon‐γ and promote T‐bet expression in atypical memory B cells.79, 88 The high levels of T‐bet expression in atypical memory B cells are associated with decreased BCR signaling and a switch from IgG1 to IgG3 surface expression.88 In some cases, T‐bet expression limits the ability of B cells to differentiate into germinal center B cells, and ultimately results in reduced IgG antibody titers,89 though in other cases it appears that T‐bet expression actually enhances serum IgG2c levels.24, 90 Hence, severe malaria is associated with an inflammatory environment that results in a defective adaptive immune response. However, for those that recover from severe malaria, T helper type 1‐skewed Tfh cells drive the gradual acquisition of T‐bet that is eventually enriched in the atypical memory B‐cell compartment.79, 88 Hence, T‐bet expression may result in impaired recall responses through its effects on atypical memory B cells.

Alternatives in mounting a durable anti‐Plasmodium antibody response

Based on the evidence provided above, there are several aspects to consider to develop a successful vaccine that provides durable immunity to malaria. First, the outcome of the immune response changes greatly with exposures. Developing vaccines by testing on naive Western populations may not be fully predictive of the response in malaria‐endemic areas.87, 91 Along these lines, it may become necessary to either productively engage atypical memory B cells to ‘awaken’ them to produce long‐lived plasma cells, or to avoid them entirely and target the classical memory populations instead (Fig. 2). This latter strategy would depend on the existence of at least some non‐overlapping antigen specificities between atypical and classical memory B cells. How best to awaken atypical memory B cells is far from clear, but engagement of certain pattern‐ or danger‐associated molecular pattern sensors might still be possible. Our own work has demonstrated that Zbtb32, which is highly expressed by age‐associated memory B cells, restrains the duration of antibody production after recall responses to some challenges, such as murine cytomegalovirus and alum‐adjuvanted immunogens (Jash and Bhattacharya, unpublished results). In contrast, Zbtb32 has no role in controlling recall responses to intestinal bacteria, influenza booster vaccinations, or Ribi‐adjuvanted immunogens (Jash and Bhattacharya, unpublished results). Hence, selection of proper adjuvants and innate stimuli can greatly influence the survival program engaged by plasma cells and the duration of immunity.92, 93

Figure 2.

Actual and ideal response to Plasmodium. Plasmodium infections in humans result in expanded, hyporesponsive atypical memory B cells (aMBCs), contracted classical memory B cells (cMBCs), and short‐lived antibody responses. Each additional infection by Plasmodium results in further increases in aMBCs and short‐lived plasma cells (SLPCs). Therefore, multiple infections are needed for immunity. Given this challenge, vaccines need to be carefully designed to induce long‐lived plasma cells (LLPCs) for a robust and durable response. To mount a robust and durable response, either adjuvants that can render aMBCs responsive must be used, or memory B cells that can recognize epitopes that confer durable and protective responses must be selectively activated and expanded.

A second, and potentially more problematic, scenario is that inherent constraints within the human germline repertoire or pathogen evasion mechanisms prevent durable responses to certain Plasmodium antigens. Although the RTS,S vaccine response was more effective in malaria‐naive individuals than in malaria‐experienced subjects, antibody responses still faded quite rapidly relative to a number of more robust vaccines and infections.94, 95 To define the underlying cause of these problems, a recent mouse study demonstrated that blood‐stage parasites actively suppress the response to CSP.96 Other animal studies using different immunization and infection models have demonstrated that both the avidity of the antigen for its target BCR as well as the self‐reactivity of the starting naive B cell can greatly influence the duration of immunity. Studies using model antigens demonstrated that when the starting avidity of the antigen is very high, paradoxically the duration of IgG memory B‐cell persistence is diminished.97 Additionally, high model antigen avidity promotes the expansion of short‐lived plasma cells.98, 99 Yet, in the context of HIV infection, a minimum antigen affinity is essential to effectively recruit pathogen‐specific B cells into the response.100 So, it may be that an antigen avidity ‘sweet spot’ is required to generate the ideal durable response. Adding to this complication is that some foreign antigen‐specific B cells are also somewhat self‐reactive, thereby diminishing their contribution to or durability following the response.101, 102, 103, 104 These problems can begin to be addressed by empirical antigen redesign and experimental animal studies. For example, although blood‐stage infection inhibits responses to CSP, responses to merozoite surface protein 1 remain intact and protect mice from subsequent infection.96, 105

Targeting naive B cells of very rare specificities may be an alternate but effective strategy. For instance, although antibody responses to the four amino acid (NANP) repeats of CSP increase over time post‐infection, this is not due to increases in affinity maturation within the germinal center. Instead, recruitment of naive B cells with higher affinity than responding memory B cells drives these increases, as somatic hypermutations are often not affinity‐enhancing.106 In addition, a surprising study isolated RIFIN‐binding, broadly reactive antibodies that recognize multiple P. falciparum isolates.107 These highly unusual antibodies have inserted the collagen‐binding domain of LAIR1, an immunoglobulin superfamily inhibitory receptor, between the V and DJ segments. Somatic mutations in the LAIR1 domain abolished binding to collagen and increased binding and opsonization of infected erythrocytes. These broadly reactive antibodies were isolated from IgG⁺ memory B cells and found to be highly clonally expanded in the two donors studied. Hence, targeting high‐affinity naive B cells may be a way to avoid atypical memory B cells and produce durable immunity.

There are still many outstanding questions on the association and causation between malaria severity and atypical memory B cells. Longitudinal studies analyzing the repertoire of, and clonality between, the memory B‐cell and plasmablast compartments after each malaria season will be important for understanding the development and contribution of atypical memory B cells. Are atypical memory B cells originally classical memory B cells that have been exhausted or are they a separate fate during germinal center reactions? Given that naturally acquired immunity is associated with an expanded atypical memory B‐cell pool, what is the association between protection and atypical memory B cells? In addition, there appear to be different subsets of atypical memory B cells based on T‐bet expression and/or isotype expression. Do the functions of these different subsets differ in recall response? In this case, malaria is a good infection system to use to study different aspects of atypical memory B‐cell biology.

Conclusion

Decades of research in model organisms have defined the basic and essential elements of cellular and humoral immunity. Increasingly, sophisticated analysis in the field has revealed properties of human responses to problematic infections, which in turn has driven a progression of mouse models that accurately mimic these issues.108 Of course not all aspects of responses are identical between humans and mice, but our viewpoint is that evolution likely preserves the most fundamental aspects of antibody responses. Even in the cases where there are differences between mice and humans, such as in the immunodominant antigens after flavivirus infection, mouse studies have revealed critical neutralizing epitopes that may form the basis for new human vaccines. As increasing numbers of reports define new structure‐based methods to elicit epitope‐specific antibodies, basic studies will help to guide these specificities into the correct cellular compartments to provide robust and durable humoral immunity.

Disclosures

DB is a co‐founder of Sana Biotechnology, Inc. and owns significant stock in Forty Seven Inc. RW declares no conflict of interest.