A key property of B-cells is that these unique cells of the adaptive immunity branch are the only cells capable of editing epitope-specific tools, i.e., antibodies (Abs), by rapidly superimposing genetic events that end with somatic mutations and class switching to ensure extremely efficient immune responses to a panoply of pathogens, particularly T-dependent (TD) antigens (Ags). Although progress is still being made in understanding the intricate mechanisms of this extremely complex genetic event, this function of B-cells is considered their classical property. Evidence is growing that innate immunity is regaining the importance it probably lost when the much-sought-after function of adaptive immunity and the highly sophisticated role of conventional lymphocytes were finally understood. Intriguingly, while many cell subsets are specific to each branch of immunity, it is now apparent that B-cells have almost equivalent roles in either branch, and their importance as innate immune cells has now been recognised with the discovery of functional and tissue-specific subsets in bone marrow (BM), secondary lymphoid organs and mucosae. B-cells may behave as one-man-bands, using the most appropriate instrument (B-cell receptor, Toll-like receptors, cytokine/chemokine-receptors, etc.) for each situation, but they are also highly plastic in their capacity to produce cytokines. This property was only recently appropriately acknowledged, in comparison to the champion cells for polarising immune responses such as T-cells, dendritic cells (DCs) or macrophages (and now platelets). In hindsight, however, this property could have been deduced from one of the first descriptions of how a lymphocyte was reprogrammed, i.e., after infection with Epstein-Barr virus (EBV) through its unique receptor, CR2/CD21 1 . Paradoxically, the rediscovery of the roles and functions of B-cell subsets in innate immunity provides the “non-classical” (new) part of the story, which itself is far from being complete, since many discrepancies between mouse and human systems plead for novel extensive research efforts to fill in the gaps.

Decades ago, lymphocytes were discovered as the major tools of specific immunity, now defined as adaptive immunity, which has unique characteristics based on clonal differentiation and a repertoire of Ag recognition. A long-term paradigm then arose, which postulated that B-cells underwent terminal differentiation into plasma cells and that the majority were then confined to antibody (Ab) production, while some differentiated B-cells were spared from this terminal pathway to constitute a pool of memory B-cells. Memory as well as activated B-cells were then shown to efficiently present Ags to T-cells and undergo mutual activation/differentiation processes that involve cytokines and several pairs of co-stimulatory molecules, including the well-known CD40/CD40L pair 2 . The constitutive expression of CD40 was ascribed first to B-cells, and then extended to most Ag presenting cells (DC and macrophages); this property helped to decipher the Ag presenting cell (APC) function of activated/memory B-cells and their use alongside Major Histocompatibility or MHC molecules 3 . The co-expression of MHC and of CD40 does not, however, licence immune cells as APCs (e.g. platelets possess such characteristics) 4 .

B-cells were essentially described for their indispensable role in the so-called humoral (Ab-based) immunity, and the machinery supporting their central functions was thought to be essentially located in the germinal centres (GCs) of lymph nodes and the spleen. Years later, it was acknowledged that this functional microanatomy also concerned the mucosal equivalents of lymph nodes, thus supporting mucosal immunity. With further investigation, however, it became evident that B-cells can exert a number of Ab-independent functions, including capturing and concentrating Ags for presentation, producing cytokines, influencing T-cell and DC responses, contributing towards distinct functions during the immune response in vivo, affecting lymphoid tissue structures, and even participating in tissue repair 5 .

Around fifteen years ago, the conceptual barrier between specific and non-specific immunity was breached and novel functions were ascribed to innate immunity. As such, the role of almost every immune cell type was revisited. There were few changes to some areas, e.g., conventional T-cells, but novel and important functions have been ascribed to non-conventional T-cells, NK-cells and NKT-cells. The issue of B-cells was much more intriguing for two main reasons: 1) the discovery/identification of non-conventional B-cells; and 2) the identification of non-self infectious danger sensors or receptors (Rcs), which are thought to characterize innate immunity and are abundantly-expressed on B-cells. Indeed, the concept of non-self-infectious danger was subsequently addressed 6 , Pathogen Associated Molecular Patterns (PAMPs) were described on pathogens, and PAMP-counter ligands, namely Pathogen Recognition Receptors (PRRs), were identified on the sensing and sentinel cells of innate immunity. PRRs have remained extremely conserved throughout the evolution of living organisms, with a key family being the Toll-like Rcs or TLRs. We, and others, have further described that a specific pattern of TLRs characterizes every B-cell subset 7 8 9 10 . B-cells express Rcs for immunoglobulins (Ig) such as FcγRIIb/CD32 or FcεRII (CD23). CD32 mediates inhibitory signals to BcR-activated B-cells, but also authorizes the internalisation of immune complexes. CD23 is essential for B-cell responses and distinguishes various immature and mature B-cell subsets. Complement Receptor 2 (CR2/CD21) is also found on B-cells, where it amplifies BcR signalling through its association with CD19 and its downstream PI3K-dependent signalling pathway 11 . CD21 also constitutes the unique receptor for EBV, a herpes virus responsible for infectious mononucleosis that is present in more than 85% of healthy individuals in its latent form. Reactivation of EBV is frequently observed in chronically HIV-infected patients where it correlates with a higher frequency of circulating transitional-like B-cells 12 . In addition, EBV⁺ B-cell lymphomas are more frequent in HIV-infected patients than in the general population 13 .

The capacity of specific B-cell subsets to traffic throughout the body is essential for sampling pathogens, but also for their APC functions 14 . Accordingly, distinct programs of chemokine receptor expression were also ascribed to the various B-cell subsets 15 . Whereas most B-cells express CXCR4 and CXCR5 throughout the course of B-cell differentiation, their balanced expression is decisive for the emigration of immature B-cells from BM into the spleen where CXCL12, the ligand of CXCR4, is highly expressed in the red pulp 16 . In steady-state conditions, the homing of mature naïve B-cells into follicle anlagen is driven by strong expression of CXCL13, the ligand of CXCR5, within the white pulp 17 , whereas the trafficking of memory B-cells beneath CCL20-expressing epithelia is orchestrated by CCR6 18 . Other chemokine receptors, in association with the highly tissue-specific expression of their ligands, orchestrate B-cell homing into intestinal mucosa (CCR9, CCR10) or the skin (CCR4) in combination with integrins and other adhesion molecules 16 19 20 . BcR or CD40 signals and TLR agonists can tune the expression of chemokine receptors 18 21 22 23 and therefore impair their trafficking. As such, B-cells that were thought to be almost “rude” cells according to the classical T-cell/B-cell paradigm, are actually split into numerous subpopulations that are functionally and spatially distinct.

Even more recent observations indicate that marginal zone (MZ) and B1-like B-cells are engaged in the innate arm of immune defence 24 25 , and that immunoglobulins (Igs) are comprised of two distinct molecular tools: 1) B-cell receptor/”specific” Abs; and 2) instruments of the innate immunity branch, the so-called “polyreactive Igs” that are largely associated with extrafollicular B1-cells 26 . Besides MZ B-cells, recently defined B-cell subsets have also captured the attention, including regulatory B-cells and CD21^lo B-cells. Although B-cell phenotype characteristics and physiological functions need to be better defined, recent studies in mice (and also in humans) have clearly revealed their distinguished roles in chronic infection, autoimmune disease and aging. This confirms that B-cells are not only actors but also key regulators in the immune response. This essay thus aims to revisit B-cell immune physiology and reconsider B-cells alongside more recent discoveries.

TLRs are commonly divided into two subgroups depending on their cellular localisation and respective PAMP ligands. The first group is composed of TLR1, TLR2, TLR4, TLR5, TLR6 and TLRs 10–13, which are expressed on cell surfaces and recognize mainly microbial surface components. The second group is composed of TLR3, TLR7, TLR8 and TLR9, which are preferentially sequestered in the endoplasmic reticulum in resting cells and rapidly traffic to endolysosomes after stimulation by microbial nucleic acids 105 . In contrast to mouse B-cells, human B-cells express neither TLR4 nor CD14, the two canonical ligands for Gram^– bacteria LPS, and are therefore unresponsive to LPS. In healthy donors, circulating naïve and memory B-cells express distinct amounts of TLR1, 2, 6, 7, 9 and 10 and are negative for TLR 3–5 and 8 3 10 106 . However, elegant studies by Cerutti’s group have clearly shown that functional TLR3 is expressed by human tonsillar B-cells, with higher expression in GCs and sub-epithelial regions, but is absent from memory B-cells. In the presence of dsRNA, these mucosal B-cells up-regulate AID expression and initiate class switch recombination and IgG/IgA production in the presence of IL-10 and BAFF 107 . Restricted to a discrete population of CD27^– blood B-cells with intermediate levels of CD19, TLR2 is preferentially expressed by naïve and GC B-cells present in mucosa or exposed to inflammatory stimuli. MZ B-cells, as well as mucosal follicular naïve B-cells, strongly express TLR2/TLR1 and TLR2/TLR6 complexes and thus recognize a panoply of unrelated molecules, including peptidoglycans, diacylated and triacylated lipopeptides, and porins from a broad spectrum of microbes 108 . BcR cross-linking with anti-Ig antibodies or protein A from Staphylococcus aureus sensitizes B-cells to TLR2-active diacylated and triacylated lipopeptides, which promote their proliferation and differentiation into IgM-producing cells 106 109 . Accordingly, TLR2 ligands are effective adjuvants in humans and are currently used in the Haemophilus influenzae type B vaccine. In vitro, TLR2 ligands also induce CCR9, CCR10 and J chain expression on human B-cells and enhance IgA production 23 , suggesting that TLR2 activation might favour intestinal homing. It is not clear yet whether this effect ameliorates the mucosal response or induces tolerance.

In humans, TLR9 expression is restricted to B-cells 10 , plasmacytoid DCs (pDCs) 7 and platelets 110 , whereas TLR7 is more widely distributed among the various populations of APCs. TLR7 and TLR9 are consistently present in all B-cell subsets, with the majority of TLR molecules sequestered in the endoplasmic reticulum. Activation of TLR9 requires its cleavage by endolysosomal proteases 111 , and TLR9 is detectable within small intracellular vesicles of primary B-cells 112 . Memory B-cells contain higher amounts of TLR9 than naïve B-cells 113 , and we found that a fraction of TLR9 molecules can traffic to the plasma membrane 10 . BcR- and CD40-mediated stimulation transiently increases TLR9 expression, which enhances responsiveness to its agonist, CpG DNA, proliferation, chemokine production and APC function 7 114 . Besides BcR and CD40 ligation, Type I IFN produced by pDCs increases TLR7 and MyD88 expression in naïve peripheral human B-cells 115 , and therefore their responsiveness to TLR7 agonists. Mouse TLR10 is not functional because of a retroviral insertion, and TLR11–13 have been lost from the human genome 105 . While TLR10 expression is also restricted to B-cells and pDCs, its role in B-cell physiology remains to be clarified in the absence of recognition of any specific ligand. Human TLR10 is strongly related to TLR1 and 6 and associates with TLR2. While TLR10/TLR2 complexes recruit MyD88 after TLR2 ligand-induced stimulation, they fail to activate typical TLR-induced signalling, including NF-κB 105 . However, it seems likely that TLR10 is functional because TLR10 gene variants have been associated with susceptibility to asthma 105 .

Thus, TLR expression is constitutive in human B-cells, assigning these cells a functional role in sensing infectious danger and participating in innate immunity; this function is well understood in mucosal surfaces but less so in other compartments. There are indications that B-cells display various arrays of TLR molecules that seem to characterize subsets. A question that is beginning to be addressed is whether TLR binding of pathogen-derived material tethers pathogen-derived Ags when the BcR is of low affinity (IgM). This may be the case for MZ B-cells, although this has not yet been ascertained because of the puzzling observation that MZ B-cells are CD21^hi. The consequences of dual B-cell stimulation through a PRR (e.g., a TLR) and BcR deserve further exploration, particularly on Ab responses initiated by a TI Ag.

Recent studies have demonstrated that in addition to the BcR-mediated uptake described for specific lipid Ags 116 117 , B-cells can use an apolipoprotein-mediated pathway of lipid Ag uptake for presentation to innate-like NKT cells. Indeed, invariant NKT-cells (iNKT), a subset of NKT-cells, recognize exogenous and self-lipids as well as glycolipid Ags (LPS-free) presented by CD1d, a non-classical class I molecule. Through their characteristically high expression of CD1d, MZ B-cells may establish cognate interactions with iNKT in the MZ 118 and elicit a stronger lipid-driven NKT-cell stimulation than follicular B-cells 119 . The uptake of lipid Ags requires the expression of the low-density lipoprotein receptor (LDL-R) by B-cells, a BcR-independent uptake pathway 120 that probably favours the production of “polyreactive” (innate), and even self-reactive, Abs. While the BcR-mediated route facilitates the uptake of particulate lipid Ags and their transport to CD1d-containing endocytic vesicles, this pathway likely favours a more lipid-specific Ab response 118 (Figure  4). Although direct stimulation of TLR2, 7 and 9 by their ligands elicits IL-6, IL-10 and IFNβ (TLR9 only) production by MZ B-cells in mice, it does not result in cognate interactions with iNKT 121 . These results show that MZ B-cells may differentially contribute to humoral immunity according to their activation route and the nature of the Ag. Moreover, part of the innate Ab response depends on rapid but transient cognate interactions between MZ B-cells and NKT-cells. Recent studies have shown that impaired CD1d recycling in patients with systemic lupus erythematous, as compared with healthy individuals, causes defective B-cell mediated iNKT stimulation. This work highlights the physiological importance of B-cells in lipid presentation to iNKT during the innate immune response 122 .

B-cells are generally considered positive regulators of the immune response (their major regulator in Ig secretion function being Ig binding through the FcγRIIb itself). However, several regulatory B-cell subsets can negatively regulate these immune responses 123 124 125 . A general feature of these regulatory B-cells is to preferentially produce IL-10 upon appropriate stimulation 126 127 128 , but this is not a unique feature. In particular, B10, a potent regulatory B-cell subset within the rare CD1d^hiCD5⁺ B-cell subset of the spleen, has been shown to regulate acute inflammation and autoimmunity through the production of IL-10 126 127 . Based on the expression of TIM-1 (T-cell Ig domain and Mucin domain protein 1), mouse B10 cells can be further distinguished from B-1a and MZ B-cells 129 . BAFF was recently shown to induce or expand B10 cells in mice 130 .

Data obtained following B-cell depletion by treatment with Rituximab (a chimeric CD20 mAb), Ofatumumab or Ocrelizumab (two fully humanised anti-CD20 mAbs) in patients with autoimmune diseases or having undergone BM transplantation strongly support the existence of regulatory B-cells in humans 131 . However, it is not clear whether they constitute a distinct lineage or a subset, and have sufficient plasticity to change their effector functions into regulatory ones. IL-10-producing regulatory B-cells have been described in healthy individuals, as well as in patients with autoimmune disease 132 133 . In contrast to mice where B10 cells are naïve B-cells, human IL-10-producing B-cells have been characterised either as memory (CD27⁺) or immature transitional (CD38^hiCD24⁺) B-cells. Only a population of human CD19⁺CD24^hiCD38^hi B-cells was shown to exert regulatory functions on CD4⁺ T-cells 134 . Recent data from Iwata et al. partially reconcile these data by showing that both human B10 cells and their precursors belong to a CD24^hiCD27⁺ B-cell population 133 . Thus, human B10 cells, similar to MZ B-cells, differ from mouse B10 cells by expressing a memory phenotype. The distinct organisation of MZ in mouse and human probably facilitates different interactions between MZM, DC and B-cell subsets 135 . In this context, it will be important to better identify BAFF-producing cells in human MZ because BAFF is likely involved in the differentiation or survival of human B10. Indeed, treatment of patients with multiple sclerosis (MS) with Atacicept (TACI-Ig), an antagonist of BAFF and APRIL, led to an unexpected increase in inflammatory activity and condemned the Atacicept trials in MS 136 . These data strongly support the notion that human B10 cells are also dependent on BAFF. While the various human B10 subsets produce IL-10 only after appropriate stimulation, a distinct population of CD20⁺CD27⁺CD43⁺CD11b⁺ B1 cells produce it spontaneously 137 . This population also down regulates the T-cell response. The respective role of these various IL-10-producing cells deserves to be extensively examined.

As previously discussed, ageing is accompanied by compromised immune responses and an increased propensity for autoimmunity. Besides reduced numbers of both hematopoietic stem cells entering into the lymphoid lineages and of MZ B-cells, accumulation of an exhausted B-cell population has recently been reported in aged mice 138 . With no surface CD21, CD23 and CD43, these CD19^hi SIgM⁺ BAFF-R⁺ B-cells differ from MZ, B1 and follicular B-cells. These B-cells further exhibit impaired functional responses to CD40 and BcR ligands, and to BAFF, and produce IgM, IL-4 and IL-10 only in response to TLR9 and TLR7 stimulation. Moreover, these B-cells are prone to induce IL-17 production by activated T-cells at the expense of T follicular helper cell generation, thus thwarting the interactions necessary for affinity maturation and memory cell formation 139 .

In aged female mice and in mice presenting with autoimmune pathologies, an accumulation of CD21^lo/-CD11c⁺ B-cells has also been described 140 . These CD11c⁺CD21^- B-cells, which are also CD5⁺ and CD138⁺, express high levels of CD95, CD80 and CD86. Unresponsive to BcR triggering, these B-cells secrete high levels of anti-chromatin IgG following TLR7 stimulation. Intact TLR7 signalling is required for the development of this population in mice. In humans, this population of CD21^- CD11c⁺ B-cells is CD5^hiCD80^hiCD86^hiCD20^hiCD23^-, but CD27^hi, and does not express any surface immunoglobulin; it therefore resembles plasmablast precursors 140 . A high frequency of circulating CD21^lo B-cells is associated with a propensity to autoimmunity, and the CD21^loCD11c⁺ population is found more frequently in the blood of elderly female patients with autoimmune disorders than in healthy age-matched individuals 11 141 . First described by Ehrhardt et al. as a unique population of memory B-cells expressing mutated BcR but not CD27, the FcRL4⁺ (Fc receptor-like protein 4) population normally resides in epithelial tissue-associated niches 142 143 144 145 , but is expanded in the blood of patients with common variable immunodeficiency 146 . Because FcRL4 disrupts immune synapse formation and blocks antigen-induced BcR signalling, FcRL4⁺ B-cells are unresponsive to BcR ligands, but still responsive to CD40L, TLR9 agonists, IL 2 and IL 10 147 . In viremic HIV-infected patients and patients chronically exposed to plasmodium falciparum, this population of FcRL4⁺CD27^- memory B-cells was found in the blood, where it progressively replaces conventional memory B-cells. In these HIV+ patients, FcRL4⁺ B-cells showed reduced proliferation and differentiation into plasma cells in response to BcR ligands, and also to cytokines and CD40L 148 . In light of this pathogen-induced loss of functions, Moir et al. called these cells “exhausted” memory B-cells. The role of FcRL4 in exhaustion is obvious: downregulation of FcRL4 expression by RNA interference in CD27^– memory B-cells in HIV-infected patients partially restores their capacity to respond to BcR stimulation and to produce HIV-specific antibodies 149 . At least in HIV-infected patients, these exhausted memory B-cells express high levels of PD-1 (Programmed Death 1) 148 . During pathogenic SIV infection in macaques, PD-1 drives the rapid and sustained loss of activated memory B-cells (CD21^loCD27⁺) in the so-called “rapid progressors”. This depletion, which constitutes an early predictor of disease progression, can be reversed by blockade of PD-1. In vivo blockade of PD1 in SIV-infected macaques have enhanced polyclonal and virus-specific Ab responses 150 . Whether similar mechanisms/pathways are responsible for CD21 downregulation in HIV-infected patients, patients with autoimmune diseases and aged individuals remains to be firmly established.

In conclusion, B lymphocytes display hallmarks of three distinct functional categories of immune cells (Table  1): i) they bear a unique B-cell receptor for antigen on their membranes, which characterizes conventional B-cells and adaptive immunity; ii) they express CD40 on their surfaces, a property which is shared by most APC (though this is not a unique feature); iii) they express PAMP-ligands i.e. PRRs, which is considered a landmark of innate immune cells. Indeed, human B-cell subsets express distinct PRRs, including FcRs, complement receptors and TLRs. B-cells, despite being from a unique origin, have diversified into distinct subsets, the survival and functions of which are associated with their microanatomical locations. Accordingly, one can distinguish four main orientations. The first orientation comprises conventional, or follicular, B-cells that populate B-cell follicles in secondary lymphoid organs and mucosa, and correspond to the B2 lymphocytes in the mouse system. They principally use their capacity to bind amino acid-based Ags (although independent of APC and HLA in the conformational presentation) through the BcR and are specialised in the response to TD Ag. Qualified as naïve follicular B-cells before they encounter Ag, they evolve into activated B-cells that populate GCs and into memory or long-lived plasma cells that colonize specific areas around follicles and bone marrow, respectively. The second orientation concerns MZ non-conventional B-cells and B1-like B-cells, that can bind either peptide Ags through their BcR and/or non-peptide Ags such as lipids and polyosides or nucleic acids via a plethora of surface receptors. This subset is specialised for TI-Ab responses and preferentially home in spleen MZ. In contrast to follicular B-cells, which are highly recirculating cells, MZ B-cells have a restricted trafficking pattern. The third orientation is forwards regulatory B-cells (Breg), most of which produce high amounts of IL-10 upon appropriate stimulation. They have been identified both in mouse and humans, and parallel regulatory T-cells. Like MZ B-cells, Breg mainly locate in the spleen MZ where their survival and expansion are dependent on BAFF; In the fourth orientation, there is a novel complex of CD21^lo B-cell subsets that include at least three different populations: activated memory B-cells (CD21^loCD27⁺), plasmablast precursors (CD21^loCD27^hiCD11c⁺CD138⁺) and “exhausted” memory B-cells (CD21^loCD27^loPD1⁺FcRL4⁺). Increased numbers of these B-cell subsets in blood have been preferentially reported in aged mice and humans, and in patients with autoimmune diseases or during chronic viral infection. B-cells belonging to these subsets harbour a non-functional BcR, and are frequently unresponsive to CD40L but strongly sensitive to TLR7 or TLR9 agonists. In healthy individuals, the latter CD21^lo population would preferentially reside in mucosa-associated tissues, immediately beneath the epithelium, likely sampling invasive pathogens. The identification of these new B-cell subsets, whose functions and survival are highly dependent on signals outside of those targeted at BcR, particularly homing, highlights the previously unexpected complexity of the B-cell compartment. It is thus a novel challenge in immunology to decipher the respective influence of innate and adaptive signals on the outcome of the B-cell responses in healthy individuals, and also during aging, autoimmune diseases, and chronic viral infection. The increasingly complex roles of B-cells are not yet fully characterised, nor are the differences between murine models and the human system fully determined. Further characterisation of the remaining mysteries of B-cell compartmentalisation is of key importance to human pathology, both in the fields of autoimmune disorders and onco-haematology.