Introduction

To date, B lymphocytes have been widely considered as part of the pathogenic processes leading to type 1 diabetes, where their effects are mediated directly through antigen presentation to T cells and indirectly through long-term production of circulating islet autoantibodies (attributable to terminally differentiated B lymphocytes, termed plasma cells) [1]. Evidence giving a decisive answer on the role of B lymphocytes in the pathogenesis of type 1 diabetes in humans, however, has not yet been presented. In view of the major role assigned to B lymphocytes in type 1 diabetes it is of critical importance to know their precise role and function in this disease. But what do we know for sure (Table 1)?

Table 1 Gaps in the knowledge related to B lymphocytes and the pathogenesis of type 1 diabetes
Full size table

B lymphocyte function in general and in mice

Autoantibodies occur in many autoimmune diseases and their pathogenic roles in these diseases have been proven, with B lymphocyte-targeted therapy being successfully used as treatment, at least in some [2]. In the NOD mouse model of type 1 diabetes, B lymphocytes play a critical diabetogenic role [3, 4]. In NOD mice, therefore, it is ‘game, set and match’: it is a B cell disease as much as a T cell and beta cell disease, even though the ability of B lymphocytes to produce islet autoantibodies has been proven not to be prerequisite for diabetes to occur in NOD mice. Yet, in other mouse models of type 1 diabetes, B lymphocytes are dispensable [5].

In human disease, the function of B lymphocytes is less obvious and the desire to directly link murine pathogenesis of type 1 diabetes to that of humans (in spite of the profound differences in the immune systems between the two species) has led to various misconceptions and false expectations [6, 7, 8]. As an example of how this can be misleading, maternal transmission of antibodies transfers diabetes from seropositive mothers to offspring in NOD mice, whereas in humans, if anything, it protects offspring from developing type 1 diabetes [9, 10]. A further anomaly that argues against an essential role for B lymphocytes in clinical type 1 diabetes is the report of an individual with X-linked agammaglobulinaemia and complete B lymphocyte deficiency who still developed type 1 diabetes. Although this reflects a single case, as a clinical observation it shows that both islet autoantibodies and antigen-presenting B lymphocytes are dispensable in the human pathogenesis of type 1 diabetes [11].

In this edition of Diabetologia, Willcox et al investigated pancreatic lymph nodes in recent-onset type 1 diabetes, as these are critical sites for the initial interaction between islet autoantigens and autoreactive lymphocytes [12]. This pioneering study showed a decrease in secondary follicle and germinal centre frequencies in donors with recent-onset type 1 diabetes compared with control donors and donors with longstanding type 1 diabetes. This phenomenon was unrelated to the presence of insulitis or the age at diagnosis of type 1 diabetes. Primary follicles in individuals with recent-onset type 1 diabetes proved structurally less distinct and more diffuse, with poor differential localisation of the B and T cell subsets. Type 1 diabetic donors also had more primary follicles that were lacking follicular dendritic cell networks than donors without diabetes or individuals with longstanding disease. This finding may seem surprising if a role for B lymphocytes is considered to be critical in this disease. In their study, Willcox et al interpreted the data assuming a pathogenic role for B lymphocytes in type 1 diabetes (‘fire’) [12]. However, fewer follicular dendritic cells would imply less antigen presentation to B lymphocytes as these cell types are co-dependent, possibly pointing to impaired B lymphocyte activation. Nonetheless, the authors appreciate the large void to be filled concerning the understanding of the B lymphocytic role in type 1 diabetes.

B lymphocytes in type 1 diabetes

B lymphocytes can be detected in severe cases of insulitis. ‘B cell insulitis’, however, is perhaps a misnomer since B lymphocytes are always outnumbered by T cells (especially CD8+ cytotoxic T cells). B lymphocytes are reported to be present in small numbers in early insulitis but are recruited to islets as beta cell death progresses, suggesting that they are a consequence, rather than cause, of insulitis [13]. B lymphocytes take up and process antigens in a concentration-dependent manner and are more than a thousand times less efficient than dendritic cells at presenting antigens if they do not happen to express the B cell receptor for a particular islet autoantigen [14]. For B lymphocytes to get involved, they must first be activated by T cells that are primed prior to seroconversion. In contrast, B lymphocytes cannot prime naive T cells, implying that autoantibodies are always a consequence of an earlier loss of immune tolerance to beta cells that is T cell mediated. It seems unlikely, therefore, that B lymphocytes initiate the autoimmune process but the door for a pathogenic contribution remains ajar.

B lymphocytes and dendritic cells may also differ in the islet peptides they present. For example, as processed and presented by dendritic cells, the HLA-DR peptidome differs from that presented by B lymphocytes, specifically for preproinsulin (PPI) and islet antigen-2 autoantibody (IA-2), even when B lymphocytes are engineered to selectively take up these antigen [15, 16]. This may point to a role for B lymphocytes in the later stages of type 1 diabetes-associated autoimmunity. Through epitope spreading, B lymphocytes might increase the rate and range of islet autoimmunity and increase the risk for the development of type 1 diabetes, implying a secondary, rather than primary, role in the pathogenesis of this disease.

Islet autoantibodies and their possible function

What about the main secreted products of terminally differentiated B lymphocytes, the antibodies? At the time of clinical diagnosis of type 1 diabetes in humans, more than 90% of individuals present with at least one of the type 1 diabetes-associated islet autoantibodies [17]. Despite their presence, no direct diabetogenic role has yet been assigned to these autoantibodies, in spite of their outstanding function as predictive and diagnostic biomarkers (‘smoke’, rather than ‘fire’?) [18]. With a 3% prevalence of diabetes-associated islet autoantibodies in the general population vs a 0.3% prevalence of type 1 diabetes, 90% of seropositive individuals remain healthy [11, 19]. The relatively low proportion of seronegative diagnoses of type 1 diabetes may, in part, be a consequence of overrating the role of islet autoantibodies in the disease, thereby disqualifying bona fide type 1 diabetes as type 2 diabetes just because of seronegativity. As a consequence, type 1 diabetes may be misdiagnosed by false negative islet autoimmunity, underscoring the urgency of defining the exact role of antibodies in type 1 diabetes. Of note, several islet autoantigens have been reported that are not recognised by autoantibodies, e.g. islet-specific glucose-6-phosphatase catalytic subunit-related protein (IGRP) [20].

Curiously, all currently known B lymphocyte epitopes occur intracellularly in beta cells, precluding antibody-dependent cell-mediated cytotoxicity (ADCC), and thus eliminating these antibodies as directly cytolytic (and diabetogenic) against beta cells. Another potential issue with regard to islet autoantibodies is the inverse correlation between islet autoantibody concentrations and T cell proliferation in type 1 diabetes that has been observed in some studies, suggesting that individuals with high antibody titres may have low T cell proliferation and vice versa [21, 22]. Given the unambiguous, central role of T cells in the destruction of the insulin-producing beta cells [23,24,25,26,27,28], autoantibodies might even function as surrogates of attempted immune regulation (‘ice water’). This provocative concept is corroborated by the finding of an overlap between B and T cell epitopes that are associated with anti-inflammatory IL-10 production [29]. Another intriguing observation further supports the concept that impaired B lymphocyte regulation, rather than pathogenic B lymphocytes, influences type 1 diabetes: notably, a lack of anti-idiotypic antibodies to glutamate decarboxylase, rather than the presence of the corresponding autoantibodies, defines type 1 diabetes [30].

B lymphocytes and autoantibodies in immunotherapy

Immunotherapies targeting (auto)antibodies, such as plasmapheresis and intravenous immunoglobulin replacement (IVIG), as a primary treatment for type 1 diabetes, have had negligible effects on disease [31, 32]. Autoantibody titres rarely correlate with clinical remission in immune intervention studies, suggesting that islet autoantibodies play a minor role, at best, in relapse or remission of type 1 diabetes [33]. In contrast, B lymphocyte-targeting therapy, using rituximab, did show a clinical effect, delaying loss of beta cell function over a period of 8 months, albeit in a small group of individuals with type 1 diabetes (primarily adolescents) [34]. Post hoc studies showed that rituximab delayed the fall in C-peptide in type 1 diabetes but did not appear to fundamentally alter the underlying pathophysiology of the disease, nor did it induce immune tolerance [35]. These data may indicate that B lymphocytes are pathogenic but could also be explained by a general immunosuppressive role of rituximab that is not limited to B lymphocytes. Another possibility is that, through the decline in B lymphocytes, a niche was created for repopulation of regulatory T cells (Tregs) or regulatory B lymphocytes [36]. Indeed, extensive phenotypic analysis of blood cells revealed transitional B lymphocytes as the single correlate of clinical efficacy of rituximab [37]. This latter finding could point to an immunoregulatory role for B lymphocytes in the pathogenesis of type 1 diabetes (‘smoke’), rather than painting them as a pathogenic immune cell (‘fire’), which is in accordance with the finding that type 1 diabetes may develop in spite of B lymphocyte deficiency.

Conclusion

In spite of many years of study, and notwithstanding the value of islet autoantibodies to predict development of type 1 diabetes, compelling evidence favouring B lymphocytes or islet autoantibodies in the immunopathogenesis that leads to beta cell destruction is still lacking. Given the function of B lymphocytes and antibodies in immunity and other autoimmune diseases, this is a striking notion that begs for validation or dismissal. Opportunities exist to settle the controversy around the role of B lymphocytes and islet autoantibodies. For instance, studying islet antigen-specific B lymphocytes will elucidate any role for B lymphocytes in autoantigen presentation. The study by Willcox et al in this issue of Diabetologia underscores differences in germinal centres between individuals with recent-onset type 1 diabetes compared with individuals without diabetes or, indeed, longstanding type 1 diabetes, and points to the need for a better understanding of the role of B lymphocytes in the pathogenesis of type 1 diabetes. Since preclinical animal models and humans differ in this regard, novel models of disease are necessary for proper translation into immunotherapeutic strategies. The report by Willcox and colleagues underscores the value of immunohistochemical studies on the tissues of donors with diabetes, such as those from the Network for Pancreatic Organ Donors with Diabetes (nPOD; www.jdrfnpod.org) that were used in this study. Studies like these may be useful in guiding the diabetes research community in the design of experiments that unambiguously define whether B lymphocytes function as fire, smoke or, perhaps, ice water in the underlying disease process that causes type 1 diabetes.