B cells and their cytokine activities implications in human diseases.

B cells are the only cell type that can give rise to antibody-producing cells, and the only cell type whose selective depletion can, today, lead to an improvement of a wide range of immune-mediated inflammatory diseases, including disorders not primarily driven by autoantibodies. Here, I discuss this paradoxical observation, and propose that the capacity of B cells to act as cytokine-producing cells explains how they can control monocyte activity and subsequently disease pathogenesis. Together with current data on the effect of anti-CD20 B cell-depleting reagents in the clinic, this novel knowledge on B cell heterogeneity opens the way for novel safer and more efficient strategies to target B cells. The forthcoming identification of disease-relevant B cell subsets is awaited to permit their monitoring and specific targeting in a personalized medicine approach.

Introduction

B cells are the only cell type that can give rise to antibody-secreting cells (ASC). Antibody production is a key feature of the immune system. The acquired durable production of specific antibodies is the basis for the protection afforded by most successful vaccines, and antibody deficiencies result in an increased susceptibility to infections [1]. However, antibodies can provoke severe immunopathology when reacting against self. Not surprisingly, B cell depletion therapy (BCDT) is a highly successful treatment option for diseases caused by autoantibodies such as idiopathic autoimmune thrombocytopenia, anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis, or neuromyelitis optica. More surprisingly, BCDT can also ameliorate autoimmune diseases considered as T cell-mediated including multiple sclerosis (MS), rheumatoid arthritis (RA), and type 1 diabetes. The discovery of B cells as drivers of these diseases represents a major paradigm shift for our understanding of their pathogenesis, which still lacks a mechanistic explanation. Meanwhile, investigations in the corresponding experimental models have highlighted the importance of the antibody-independent functions of B cells, demonstrating that they controlled the course of these diseases via the presentation of antigen and the production of cytokines. This review builds on these observations to discuss the roles of cytokine-producing B cells in MS and RA. In both disorders, it appears that B cells can contribute multiple and sometimes antagonistic functions via cytokine production. This novel knowledge on B cell heterogeneity might explain why BCDT led to an amelioration of disease in a large number of MS and RA patients, while a few case studies reported an aggravation of pathology in treated patients. Notably, this novel knowledge opens the perspective for improved B cell targeting in immune-mediated disorders, once the disease-relevant pathogenic and protective B cell subsets will have been identified.

Cytokine-producing B cells and immune regulation in MS

MS is a chronic inflammatory and demyelinating disorder of the central nervous system (CNS) characterized by multifocal demyelination, glial scar formation, substantial disabling axonal damage, and the accumulation of immune cells at lesion sites. It is a leading cause of neurological disability among young adults (20–40 years) affecting 2.5 million people worldwide. Clinically MS can take several forms, with new symptoms and signs occurring either in discrete attacks (relapsing-remitting MS; RR-MS) or in a continued accumulation over time (primary or secondary progressive forms; PP-MS or SP-MS). The most compelling clinical observation illustrating the pathogenic involvement of B cells in MS has been that the selective depletion of B cells with anti-CD20 such as rituximab, ocrelizumab, or ofatumumab led to major decrease in disease activity and to the improvement of disease course [2], [3], [4]. The key role of B cells in MS pathogenesis asks for an identification of the responsible pathogenic mechanisms.

Abnormal B cell activation is an early event in MS, noticeable in the vast majority of patients already at disease onset by the presence in the cerebrospinal fluid (CSF) of oligoclonal bands (OCB), which result from local antibody production [5]. OCB are the only immunological biomarker with diagnostic and prognostic value in MS, and their detection is still the criterion used to discriminatively diagnose MS in many clinics [6]. OCB reactivity is directed at least in a large part towards ubiquitous intracellular autoantigens [7]. Similar autoantibodies have been found in patients with inflammatory bowel disease who have no sign of CNS inflammation [8]. These antibodies are therefore not specific for inflammatory disorders of the CNS, and are most likely produced in response to tissue damage. Indeed, some antibodies recognize antigens at the border of demyelinating lesions but do not react against the healthy CNS [9]. In some patients, plasma cells produced antibodies reacting against native myelin antigens [9], [10], and antibody deposits have been observed in association with complement, inside macrophages containing myelin debris, in areas of active demyelination [11], [12]. Do such antibodies contribute to lesion progression across the majority of patients? Antibodies against myelin oligodendrocyte glycoprotein (MOG), which can aggravate experimental forms of MS [13], are often evoked as the prototypical example of myelin-reactive pathogenic antibody in MS. However, they have only been found in rare patients who usually present with distinctive clinical features [14], [15]. The role of humoral immunity in MS pathogenesis therefore remains controversial, all the most that BCDT inhibits new disease relapses without reducing intrathecal antibody levels, arguing towards the importance of the antibody-independent functions of B cells in MS [3], [16], [17].

B cells can profoundly influence experimental autoimmune diseases through the production of cytokines. In experimental autoimmune encephalomyelitis (EAE), the primary animal model for MS, B cells drive disease via both the presentation of antigen to CD4 T cells and the production of interleukin (IL)-6, which exacerbates unwanted TH17 responses [18], [19]. BCDT ameliorates ongoing EAE progression by removing IL-6-secreting B cells [18]. This pathogenic B cell function might develop and operate during cognate T-B cell dialog with primed T cells stimulating IL-6 expression in B cells through CD40 engagement [18], and in return B cell-derived IL-6 promoting T helper cell proliferation [20] as well as TH17 polarization [18]. B cells from MS patients produced higher amounts of IL-6, compared to healthy donors (HD), suggesting that this pathogenic axis might be at play in MS [18]. Supporting this notion, BCDT was associated with a reduction of TH17 responses in vitro, and the B cells re-emerging subsequently to rituximab treatment displayed a normalized cytokine expression [18]. B cells from MS patients also produced more lymphotoxin, granulocyte macrophage colony stimulating factor (GM-CSF), and tumor necrosis factor TNF-α compared to HD [18], [21], [22], [23]. These deregulations, which involved metabolic pathways around miR-132 and the histone deacetylase sirtuin 1 [22], might also contribute to B cell-mediated pathogenesis. Lymphotoxin production by B cells might fuel the accumulation of T cells in the inflamed CNS through the local stimulation of CCL19 and CXCL13 expression [24]. CXCL13 is expressed in macrophages in active MS lesions [25], and BCDT leads to a reduced expression of CXCL13 and CCL19 in the CNS [26], as well as to a rapid decline of intrathecal T cell number [27]. CXCL13 might contribute to the infiltration of T cells at lesion sites since its receptor CXCR5 was found on 20 to 30% of blood CD4 T cells in MS patients [25]. CCL19 has recently been shown to inhibit re-myelination by influencing myeloid cell polarization in vitro [28]. Thus, B cell-derived lymphotoxin might influence immune cell recruitment and repair process in the CNS by modulating the function of monocytes/myeloid cells locally. As a result of their excessive GM-CSF production, B cells from MS patients exacerbated the pro-inflammatory activity of monocyte-derived macrophages [23], which was on the opposite side attenuated by BCDT [23]. The emerging role of cytokine-producing B cells as key drivers of T cell and monocyte activation in MS provides an explanation for the capacity of BCDT to rapidly ameliorate MS, and to be beneficial in diseases not driven by autoantibodies. The identification of a dedicated pro-inflammatory B cell subset would provide a novel opportunity for treating MS. There is some evidence that such cells may exist. The B cells expressing GM-CSF also co-expressed IL-6 and TNF-α at high levels in MS patients [23]. They carried a CD27+ CD24hiCD25hiCD49dhiCD38lo phenotype, and thus corresponded to a subset of memory B cells [23]. Further precision of their phenotype might highlight novel targets to undertake their specific removal, while leaving the remaining of the B cell compartment intact.

B cells have protective functions through the provision of anti-inflammatory cytokines in autoimmune diseases, [29], [30]. For instance, B cells controlled the remission from EAE through the production of IL-10 and IL-35 [31], [32]. A deficit in B cell-mediated regulation was associated with exacerbated encephalitogenic T cell responses, and the augmented accumulation of inflammatory monocytes in the CNS. The signals instructing this regulatory B cell function involved the BCR, CD40, TLR, and IL-21 [20], [31], [33], which also facilitated plasma cell differentiation, and indeed, plasmocytes were the main source of B cell-derived IL-10 and IL-35 in this disease [32]. Human B cells also produce IL-10 after activation [34], [35], and depending on the stimulation protocol, distinct B cell subsets emerged as the major IL-10 producer in vitro. Thus, naïve or memory B cells were the main IL-10 producers, respectively, upon CD40 or TLR9 engagement [34], [35]. B cells from MS patients displayed a reduced IL-10 expression selectively after stimulation via CD40 [34]. The phenotype of the B cells actually producing IL-10 and possibly providing beneficial regulatory function in vivo in human however remains an unresolved issue. Are antibody-secreting cells the major IL-10- and IL-35-producing B cell subsets in vivo in human as described in mice? Human B cells produced IL-10 in the course of their differentiation into plasma cells in vitro [36], and human B cells co-expressing the transcription factor IRF-4 at high levels, a feature characteristic of plasma cells, and EBi3, the inducible IL-35 subunit, have been observed in human tonsils [37]. Along the same line, IL-10-producing B cells carried a plasmablast-like phenotype in graft-versus-host-disease [38]. The existence of human regulatory plasmocytes could explain why the treatment of MS patients with atacicept, a drug that neutralized BAFF as well as APRIL and interfered with plasma cell survival, led to a rapid disease exacerbation and an interruption of the corresponding clinical trial [39]. In sum, MS progression might involve a deficit in B cell-mediated immune regulation, in addition to an increased pro-inflammatory activity of these cells. This disequilibrium might involve the transcription factor STAT6, which is expressed in its active form at higher levels in B cells from MS patients compared to HD, and which stimulates GM-CSF as well as inhibits IL-10 expression [23]. Highlighting this antagonistic regulation between the pro- and anti-inflammatory functions of B cells, distinct B cells expressed GM-CSF and IL-10 after in vitro stimulation, supporting the notion that these activities were controlled by antagonistic transcription networks [23].

The modulation of cytokine expression in B cells might be of therapeutic relevance in MS. Hygienic conditions associated with an increased IL-10 expression in B cells, such as infection with helminth parasites [40], have been correlated with an improved disease course [41]. Several drugs used to treat MS increase IL-10 production by B cells. Glatiramer acetate is a randomized copolymer consisting of four amino acids, which was approved by the Food and Drug Administration in 1996 as a therapy for MS for its capacity to decrease the progression of disability and the relapse rate [42]. It restored IL-10 production by B cells from MS patients [43], and also stimulated IL-10 production by mouse B cells [44]. In mice, B cells were essential for the protective function of glatiramer acetate against EAE. Indeed, glatiramer acetate augmented the protective function of B cells in adoptive transfer experiments, and, conversely, it failed to attenuate EAE in B cell-deficient mice [44]. Mitoxantrone increased IL-10 production by B cells from MS patients, at least in part by shifting the balance between B cell subsets due to a reduction in the frequency memory B cells [34]. Fingolimod, a drug that inhibited the sphingosine-1-phosphate receptor (S1P), led to an increased expression of IL-10 and a reduced production of TNF-α by B cells [45], [46]. The drugs currently used to improve MS might thus mediate their beneficial effect at least in part by promoting the protective activities of B cells and reducing their pro-inflammatory functions. Could more efficient therapies by obtained with approaches optimized to reach this goal?

Cytokine-producing B cells and immune regulation in rheumatoid arthritis

Rheumatoid arthritis (RA) is among the most frequent autoimmune disorders, affecting around 1% of the adult population [47]. It manifests with synovial inflammation and hyperplasia, as well as bone and cartilage destruction, and additionally variable extra-articular signs [47]. The clinical onset of RA is preceded by a long pre-disease phase that can last for several years, and is noticeable by the presence of abnormally high levels of antibodies against post-translationally modified (citrullinated, carbamylated, and acetylated) proteins [48], [49]. The levels of anti-citrullinated peptide antibodies (ACPA) rise sharply just before disease onset, reflecting a surge in B cell activity [48]. ACPA are present in 80% of individuals with established RA, and are associated with a more severe disease progression compared to seronegative patients [50]. Some of these antibodies are produced by plasma cells located in the synovium [51], where numerous B cells (especially memory B cells) and plasmocytes are present [52], [53]. In the synovial membrane, B cells are predominantly located in aggregates also containing T cells, while plasmocytes are more widely dispersed [54]. Consistent with a B cell involvement in this disease, BCDT led to a sustained improvement of RA in treated patients [55], [56], reducing inflammation, and improving structural joint damage by decreasing bone resorption as well as increasing bone formation [57], [58], [59]. As observed in MS, the depletion of B cells led to a reduction in local and systemic levels of CXCL13 in RA [60], [61]. Distinct patients responded differently to BCDT. The outcome of the treatment has been inversely correlated with the presence of a type I IFN signature in synovial tissue [62], and with several B cell parameters. Thus, the depletion of B cells and plasmocytes from the synovial tissue correlated with the clinical improvement [47], and patients with elevated amounts of ACPA or rheumatoid factor antibodies at baseline usually displayed a better response than seronegative individuals [63], [64], [65]. ACPA aggravated tissue injury in experimental arthritis [66], and patients showing disease amelioration after BCDT often displayed a diminution of autoantibody titers [67]. The reduction of autoantibodies might thus contribute to disease improvement after BCDT. However, several observations suggested a more complex mode of action. First, disease improvement often preceded the decline of autoantibody titers after BCDT. Second, treatment with atacicept reduced RF and ACPA levels [68] but did not ameliorate RA [69]. Third, some seronegative patients displayed an improvement of disease after BCDT [70]. This lack of correlation between changes in autoantibody levels and clinical response after BCDT [67], [71] indicates that B cells also perpetuate RA by antibody-independent mechanisms such as the presentation of antigen to T cells, and the production of cytokines.

B cells in the synovial fluid of RA patients expressed RANKL, an essential cytokine for osteoclast differentiation, survival, and function [72], [73]. The RANKL-blocking antibody denosumab improves bone turnover, and has been approved for the treatment of patients with osteoporosis [74]. In the synovium a large proportion of memory CD20 + B cells carried RANKL on their surface [75], and such RANKL-expressing cells were adjacent to RANK-positive cells with features of monocytes and osteoclast precursors [75]. RANKL-expressing B cells promoted the differentiation of monocytes into osteoclasts in vitro, suggesting that they might drive osteoclastogenesis and thus bone damage in RA [75]. In keeping with this notion, in mouse, RANKL expression by B cells directly contributed to the increase in osteoclasts and bone loss after ovariectomy [76]. The characterization of human RANKL-expressing B cells highlighted several unique features. First, switched memory B cells (CD27 + IgD-) were particularly prone to express RANKL upon activation compared to other B cell subsets, and a higher frequency of blood memory B cells up-regulated RANKL after activation in RA patients compared to HD [75]. The signals controlling RANKL up-regulation in B cells are starting to emerge. IFN-γ strongly increased the expression of RANKL in B cells, while decreasing the expression of its decoy receptor osteoprotegerin, thus tipping the B cells towards a more osteoclastogenic profile [77]. IFN-γ also induced CXCR3 on B cells, and all RANKL + B cells were CXCR3 + in the synovial fluid of RA patients [77]. Collectively, these observations suggest that IFN-γ is involved in the generation of RANKL + CXCR3 + memory B cells in RA patients. A possible source of such IFN-γ might be citrulline-peptide-reactive memory TH1 cells, which are more abundant in RA patients compared to HD [78] and accumulate into affected joints [79]. CXCR3 plays a role in RA pathogenesis because preliminary data indicate that the chemokine inhibitor MDX-1100, which neutralizes the CXCR3 ligand CXCL10/IP-10, has a beneficial effect in RA [80]. TNF-α might also play a role in this pathogenic axis. B cells from RA patients produced higher amounts of TNF-α than cells from HD upon activation via TLR9 and CD40 [81]. TNF-α promotes the formation of osteoclasts, and increases the expression of RANKL in the presence of IL-1β, suggesting that B cell-derived TNF-α may contribute to bone resorption in RA [54]. Independent studies found that RANKL was expressed preferentially on memory B cells expressing Fc receptor like 4 (FcRL4), a low affinity receptor for IgA [82], in the joints of RA patients [83]. These cells carried antibodies of IgG and IgA isotypes, and were in part reacting against citrullinated peptides [84]. RANKL is primarily expressed by activated memory B cells not engaged in the plasma cell differentiation pathway. They express CD11c, CD80, CD86, CD95 and TNF-α at high levels [77], [83], [85], but BLIMP-1 and XBP1, which are two transcription factors associated with plasma cell differentiation, at low levels compared to FcRL4-negative B cells [84]. Accordingly, the plasma cell inducing-cytokine IL-21 inhibited RANKL expression in B cells [77]. Nonetheless, FcRL4-expressing B cells can give rise to antibody-producing plasma cells upon appropriate activation [85]. RANKL + B cells might therefore give rise to ACPA-producing plasmocytes in the synovium of RA patients. Although the relationship between the various subsets of RANKL-expressing memory B cells has not been fully clarified, their elimination might contribute to the beneficial effect of BCDT on bone homeostasis. A decrease in the synovial levels of RANKL has been described in one study based on immunohistological studies [59], but was not observed in an independent investigation based on mRNA quantification [62].

Studies in experimental arthritis have identified additional mechanisms of B cell-mediated pathogenesis involving cytokine production. B cells exacerbated proteoglycan-induced arthritis through the provision of IFN-γ [86]. Mice with an IFN-γ deficiency restricted to B cells were almost completely resistant to the disease [86], and displayed markedly reduced autoreactive T cell as well as humoral responses [86]. B cell-derived IFN-γ is therefore a key checkpoint for both T and B cell autoimmune responses. Its mode of action has not been elucidated, but it is intriguing that a deficiency in Icosl in B cells resulted in a similar reduction of both autoimmune T and B cell responses [87], suggesting that B cell-derived IFN-γ and ICOSL intervened in the same pathway. ICOSL and ICOS are important for the cognate interaction between B cells and T cells, and the subsequent differentiation of T follicular helper (TFH) cells [88]. B cells are essential antigen-presenting cells (APC) in proteoglycan-induced arthritis. Mice lacking both CD80 and CD86 in B cells are resistant to the disease, even though they produce normal amounts of anti-proteoglycan antibodies [89]. Antibody production is therefore insufficient to provoke disease, and B cell-mediated antibody-independent functions are required. IFN-γ production by B cells might thus promote autoimmunity by acting on T cells. Notably, increased IFN-γ signalling in T cells can lead to elevated accumulation of TFH cells, and subsequently uncontrolled autoimmunity [90]. IFN-γ producing-B cells might thus promote the accumulation of TFH cells above the threshold required for arthritis development.

Through the production of cytokines, B cells can play diverse roles, and even counteract RA progression. IL-10 production by B cells is protective in collagen-induced [91] and antigen-induced arthritis [92], by inhibiting pathogenic T cell responses of TH1 and TH17 types as well as autoantibody production [91]. This IL-10-mediated protective B cell function might be deficient in RA patients. After activation via TLR9 for 24 h and additionally with phorbol myristate acetate and ionomycin for the last 4 h, blood B cells from patients with RA for less than 5 years displayed a reduced frequency of IL-10 producers compared to HD [93]. The proportion of IL-10-expressing B cells was inversely correlated with disease activity, suggesting a possible protective role [93]. In this culture setting, IL-10 was primarily produced by transitional CD24hiCD38hi and memory CD24hiCD27+ B cells [93]. In a distinct study, fewer CD24 + CD27 + memory B cells (lacking CD38 expression) expressed IL-10 in RA patients compared to HD after activation via TLR9 and CD40 [81]. This was associated with the reduced activation of p38 and ERK, which promoted IL-10 expression in B cells [81], [94]. Interestingly, addition of IL-21 to these cultures, a known inducer of plasma cell differentiation [95], increased the amount of IL-10-expressing B cells 4-fold for HD and 10-fold for RA patients, thus erasing the difference between RA patients and HD [81]. This finding highlights the difficulty in drawing final conclusion on cytokine production by B cells in patients using in vitro culture systems whose relatedness to what is actually happening in the patient is uncertain. This is even more so that IL-21 plasma levels, and IL-21-producing TFH cells are elevated in RA compared to HD [96]. Nonetheless, such type of assay enables the detection of aberrations in signalling pathways in RA B cells. It appears that aberrations in the cytokine network might underlie some of these signalling defects. Indeed, treatment by TNF-α blockade for 18 months resulted in an increased expression of IL-10 in B cells from RA patients stimulated via TLR9 and CD40 [93]. Notably, the fact that IL-21 also stimulated antibody production, further highlighted that the signals most efficient at inducing IL-10 production were those also triggering plasma cell differentiation [97], [98].

Conclusion

Our knowledge of B lymphocytes has profoundly changed over the last 15 years. They have emerged as key drivers of pathogenesis not only in inflammatory disorders known to be caused by autoantibodies, but also in diseases such as MS and RA thought to be mediated primarily by T cells acting as monocyte-activating cells. Most intriguingly, T cell-targeted therapies such as anti-CD4, anti-CD5, or alemtuzumab, have produced only limited effects in these diseases [47]. Meanwhile, it has been discovered that B cells could directly influence monocytes and T cells through the production of cytokines. Thus, B cells can exist in the form of multiple cytokine-producing subsets with either pro- or anti-inflammatory functions. This raises the question of the possibility of having more efficient B cell-targeted therapies by depleting only disease-driving subsets. Of note, cytokine production by B cells also affects their antibody production [30]. There is strong evidence that current anti-CD20 antibodies do not provide the most efficient approach to target B cells. For instance, B cell depletion in the synovium is not always complete in RA patients after rituximab treatment [52], [99], [100], and the extent of depletion has been associated with the clinical response [101]. Notably, a more profound depletion of all CD20-expressing B cells might lead to a too strong lowering of host defence mechanisms against infections. Thus, a clinical trial assessing the efficacy of B cell depletion with ocrelizumab in RA was stopped due to the increased risk of infection [102]. There is therefore a need for novel approaches that more efficiently deplete disease-driving subsets of B cells and plasmocytes, while sparring protective cells that either confer anti-microbial immunity or mediate anti-inflammatory functions. The discovery of markers allowing the identification of these various B cell subsets, to undertake their monitoring and their selective targeting, is highly awaited.

Conflict of interest disclosure

The author declares no commercial or financial conflict of interest.