B-Cells Promote Intra-Islet CD8⁺ Cytotoxic T-Cell Survival to Enhance Type 1 Diabetes

RIP-TNFα NOD mice (N22) and NOD-μMT^−/− (N20) mice have been described elsewhere (10,12,16). TNFα-μMT^−/− mice were generated by breeding male TNFα-NOD-μMT^+/+ mice to female NOD-μMT^−/− mice, then intracrossing TNFα-μMT^−/+ male mice to female NOD-μMT^−/+ littermates to create TNF-μMT^−/− and TNF-μMT^+/+ mice littermates for all studies described. TNFα-μMT^−/− mice were screened by fluorescence-activated cell sorter (FACS) analysis of peripheral blood for B-cells using a rat anti-CD19 Ab (BD Biosciences).

NOD-mIg.μMT^−/− (18) mice were crossed to TNFα-μMT^+/+ mice, and the resultant heterozygous littermates were intracrossed to generate TNFα-μMT^−/−, TNFα-μMT^+/+, TNFα-mIg.μMT^−/−, and TNFα-mIg.μMT^+/+ mice.

All mice were housed in specific pathogen-free barrier conditions at Cambridge University Central Biomedical Services. All experimental procedures were conducted under U.K. Government Home Office guidelines.

Islet-residing cells were isolated after collagenase digestion of pancreata as previously described (12). Cells were isolated from lymph nodes using standard protocols. For purification of CD4⁺ T-cells, cell suspensions were incubated with rat anti-mouse CD8 (53-6.72) and B220 (RA3-6B2) antibodies (both BD Biosciences), followed by BioMag goat anti-rat IgG magnetic beads (Qiagen). Macrophages and DCs were depleted by incubation of the cell suspension on plastic at 37°C for 30 min. For CD8⁺ T-cell purification, a rat anti-mouse CD4 Ab (GK1.5; BD Biosciences) was used in place of anti-CD8 Ab. In all cases, cell purity was >96% as confirmed by FACS analysis.

For purification of intra-islet immune cells for intracellular staining, pancreatic extracts were incubated with rat anti-CD45 Ab (BD Biosciences) and CD45⁺ cells isolated using a MoFlo cell sorter (Dako Cytomation).

A total of 3 × 10⁵ purified T-cells were cultured with 5 × 10⁴ irradiated islet-infiltrating antigen-presenting cells (APCs) and islet cells. The assay was pulsed with 1 μCi/well [³H]-methyl thymidine (Factor; Amersham Biosciences UK) for the last 6 h of a 4-day culture, and incorporation of radioactivity was detected using a Top Count β-counter (PerkinElmer, Beaconsfield).

For CD4⁺ T-cell depletion studies, mice were injected with 0.5 mg purified anti-CD4 (GK1.5; BioExpress). Control mice received 0.5 mg purified rat IgG (Jackson ImmunoResearch). First injections were given at 26, 28, and 30 days of age and then subsequently at 2-week intervals until the end point was reached (diabetic or 30 weeks old). For bromodeoxyuridine (BrdU) studies, mice were injected with 1 mg of BrdU (BD Biosciences) for three consecutive days before analysis.

All antibodies/kits were obtained from BD Biosciences unless otherwise stated. Single-cell suspensions were incubated with the appropriately flurochrome-labeled antibodies CD3 (2C11), CD4 (GK1.5), CD8 (53-6.7), CD45 (30-F11), and CD44 (IM7), following Fc receptors blockage. Foxp3 was detected using a Foxp3 staining kit (eBioscience). For BrdU detection, a BrdU detection kit was used following the manufacturer's instructions. All samples were acquired on a FACS Calibur flow cytometer (Becton Dickinson) using Cellquest software and analyzed using FlowJo software (Tree star). For apoptosis studies, an Annexin V-PE kit was used following the manufacturer's protocol.

Approximately 1.5 × 10⁶ PLN cells or 1 × 10⁶ islet T-cells were stimulated with α-CD3 and α-CD28 Ab (both BD Pharmingen) at 1 μg/ml. For the final 4 h of a 24-h incubation period, 1 μg/ml of golgi stop (BD Pharmingen) was added. The cells were stained with fluorochrome-labeled anti-CD3, CD8, interferon (IFN)-γ, granzyme B (Caltag Laboratories), or the appropriate isotype control and analyzed by FACS.

Mice were tested once weekly for glycosuria using Diastix reagent strips (Johnson & Johnson). Diabetes was confirmed after two serum glucose readings of 13.9 mmol/l 48 h apart.

Comparisons of disease frequencies were performed using the log-rank test. P values <0.01 were considered significant. For other assays, statistical significance was measured using the Mann-Whitney nonparametric test. P values <0.05 were considered significant. To illustrate differences in islet-infiltrating CD8⁺ T-cells in TNF-μMT^+/+ versus TNF-μMT^−/− mice, a nonparametric Kruskal-Wallis test was performed. The P values were corrected using a permutation test permuting the assignment of mice to groups within experiments to allow for potential differences between replicated experiments. P values <0.001 were considered significant.

To assess the importance of B-cells in type 1 diabetes development under potent inflammatory conditions, TNFα-NOD mice were crossed to B-cell–deficient NOD-μMT^−/− mice, and resultant TNFα-μMT^−/− NOD (hereafter called TNFα-μMT^−/−) mice were monitored for progression to diabetes. TNFα-μMT^−/− mice had significantly decreased disease kinetics (P = 0.0001) compared to wild-type TNFα-μMT^+/+ mice, although the overall penetrance of disease was the same (Fig. 1A). This observation was independent of the sex of the mouse. Nontransgenic littermates were protected from disease as previously reported (data not shown) (10,19). Thus, B-cells are essential for accelerating type 1 diabetes in the presence of a strong proinflammatory environment.

B-cell enhancement of disease kinetics was not linked to the secretion of autoantibodies, as has been previously reported (20). This theory was based on our observation that crossing TNFα-μMT^−/− mice to transgenic NOD-mIg.μMT^−/− mice that harbor B-cells carrying a mutation in the secretory component of the Ig gene, thereby preventing Ig secretion but not membrane-anchored expression of Ig (18), efficiently restored disease acceleration in double transgenic TNFα-mIg.μMT^−/− mice (Fig. 1B).

Isolated CD4⁺ or CD8⁺ T-cells from either TNFα-μMT^+/+ or TNFα-μMT^−/− mice efficiently responded to a broad range of islet antigen presented by intra-islet APCs from either strain with equal efficacy (Fig. 1C). Such responses were antigen specific since T-cells isolated from either strain failed to respond to irradiated splenocytes from TNFα-μMT^+/+ mice. This suggested priming of the T-cell repertoire and functionality of intra-islet DCs was unimpaired by B-cell deficiency. We, therefore, hypothesized that the disease kinetics in TNFα-μMT^−/− mice were linked to in situ suppression of T-cell responses by CD4⁺Foxp3⁺ T regulatory (Treg) cells (21). However, FACS analysis established no significant differences in the percentages of CD4⁺Foxp3⁺ Treg cells between TNFα-μMT^+/+ and TNFα-μMT^−/− mice in either the islets or the PLNs (Fig. 2A). Furthermore, the delayed kinetics of disease in TNFα-μMT^−/− mice did not require functional CD4⁺Foxp3⁺ Treg cells. This finding was based on observations that administration of anti-CD4 depleting Ab to TNFα-μMT^+/+ and TNFα-μMT^−/− mice did not alter disease development in any mouse group with respect to the appropriate isotype control antibody–treated mice (Fig. 2B). These data suggest that immune-mediated β-cell destruction can be controlled independent of the traditional CD4⁺Foxp3⁺ Treg cell suppression.

Previously we demonstrated that ablation of CD8⁺ T-cells protected TNFα-μMT^+/+ mice from developing type 1 diabetes, whereas CD4⁺ T-cells played a redundant role (16), a finding that was supported by our above observations. Therefore, we focused our attention on the CD8⁺ T-cell population in TNFα-μMT^−/− mice and TNFα-μMT^+/+ mice. Flow cytometric analysis of lymphocytes from each mouse strain mice revealed a significant decrease (P = 0.0001) in the percentage of intra-islet CD8⁺ T-cells in TNFα-μMT^−/− mice compared with TNFα-μMT^+/+ mice at 8 weeks of age. This time point corresponds to the transition of activated CD8⁺ T-cells into CTLs (16). In contrast, we saw no differences in the percentage of intra-islet CD4⁺ T-cells in TNFα-μMT^−/− mice compared with TNFα-μMT^+/+ mice (P = 0.042) (Fig. 3A). No variation was observed in T-cell proportions within the PLNs of all mice examined (data not shown). To determine whether B-cell ablation results in a general decrease in intra-islet CD8⁺ T-cell accumulation, we performed time-course studies. We found no significant differences in CD8⁺ T-cell percentages in the islets of TNFα-μMT^−/− mice compared with TNFα-μMT^+/+ mice at 4 and 6 weeks of age (not shown). These data imply that during inflammation, B-cells are central for the accumulation of autoreactive CD8⁺ T-cells in the target tissue at the CTL generation stage of the disease process.

Using MHC class I tetramers–peptide complexes, we determined that B-cell deficiency did not hinder intra-islet accumulation of CD8⁺ T-cells specific for two key diabetogenic antigens: insulin and islet-specific glucose-6-phosphate catalytic subunit–related protein (IGRP; data not shown) (6,22).

We speculated B-cells to be involved in the later postpriming stage of disease. To test the requirement for B-cells at the intra-islet secondary activation/expansion phase, groups of TNFα-μMT^−/− and TNFα-μMT^+/+ mice were injected with the thymidine analogue, bromodeoxyuridine (BrdU), which incorporates into the DNA of proliferating cells. We initially focused on 8-week-old mice, as 8 weeks is the critical age when significant differences in intra-islet CD8⁺ T-cell levels are apparent in the two strains of mice. Flow cytometric analysis of BrdU⁺CD8⁺ T-cells isolated from the islets, PLNs, and control peripheral inguinal lymph nodes from each mouse strain established the greatest percentage of proliferating CD8⁺ T-cells to be located in the islet fraction in both TNFα-μMT^−/− and TNFα-μMT^+/+ mice (Fig. 3B). Interestingly, although TNFα-μMT^−/− mice showed significantly decreased turnover of CD8⁺ T-cells in the lymph nodes compared with TNFα-μMT^+/+ mice, we detected no differences in the percentage of intra-islet BrdU⁺CD8⁺ T-cells between TNFα-μMT^+/+ and TNFα-μMT^−/− mice (mean values 12 and 14%, respectively). Furthermore, we could detect no differences in the level of the activation/memory marker CD44 on intra-islet CD8⁺ T-cells between the two strains of mice (Fig. 3C). Examination of 6-week-old mice revealed an identical pattern of cell turnover (not shown). Thus, environmental influences dictate the necessity for B-cell help in the expansion of CD8⁺ T-cells.

The above evidence that CD8⁺ T-cells are significantly decreased in the islets of TNFα-μMT^−/− mice with respect to TNFα-μMT^+/+ mice, even though activation and in situ proliferation of CD8⁺ T-cells are similar between the two strains of mice, led us to hypothesize that intra-islet B-cells directly or indirectly provide survival signals to activated CD8⁺ T-cells. This hypothesis was supported by evidence that between 7 and 9 weeks of age, the percentage of CD8⁺ T-cells undergoing apoptosis was significantly higher in the islets of TNFα-μMT^−/− compared with TNFα-μMT^+/+ mice (Fig. 4). This observation was restricted to the CD8⁺ T-cell population within the islets, as similar analysis of the PLNs revealed no significant differences in the survival of CD8⁺ T-cells between the two strains of mice. In contrast, we found no significant differences in the survival of intra-islet or PLN-derived CD4⁺ T-cells in TNFα-μMT^+/+ and TNFα-μMT^−/− mice. Thus, in the absence of B-cells, it would therefore appear that CD8⁺ T-cells lack a vital signal required for their prolonged intra-islet survival.

The above data clearly state a role for B-cells in the in situ survival of activated CD8⁺ T-cells. However, we wished to know if this was the only role B-cells played in type 1 diabetes. CD8⁺ T-cell–mediated destruction of cells requires an activated CD8⁺ T-cell to transform to a CTL. We therefore determined whether CD8⁺ T-cells exposed to an inflammatory environment in the presence or absence of B-cells had equal capacity to transform to CTL. Our data suggest that a B-cell role for the transition to CTL is dependent on the environment. IFNγ⁺ and granzyme B⁺ production at the per cell basis in CD8⁺ T-cells was significantly reduced in the PLNs of 7- to 9-week-old TNFα-μMT^−/− mice in comparison to TNFα-μMT^+/+ mice (Fig. 5). Interestingly, in inflamed islets, the percentage of CD8⁺ T-cells that expressed IFNγ or granzyme B was slightly enhanced in TNFα-μMT^−/− mice compared with TNFα-μMT^+/+ mice (Fig. 6A,B). Nevertheless, the absolute number of intra-islet differentiated CTLs in TNFα-μMT^−/− mice was decreased between 10 and 30% with respect to the absolute numbers of CTLs in TNFα-μMT^+/+ mice (Fig. 6C), a finding that is consistent with the increased apoptosis of CD8⁺ T-cells at this time point in TNFα-μMT^−/− mice.

These data show B-cells have distinct roles to play in type 1 diabetes development depending on the environment: In the PLNs, they are essential for CTL generation, not survival of CD8⁺ T-cells; in the islet, alternative mechanisms promote CTL generation. In this latter case, B-cells are central for providing CD8⁺ T-cell survival signals to maintain high levels of anti-islet CTLs needed to promote type 1 diabetes progression.

Understanding the complex interactions of distinct immune cells that invade inflamed target tissues and promote autoimmunity will help decipher the mechanisms that lead to the destruction of the target cell and associated immunopathologies. Here we show for the first time a link between chronic inflammation, B-cells and CD8⁺ T-cells, and the promotion of type 1 diabetes. We show B-cells play dual roles: In the PLNs, they drive CD8⁺ T-cell transition to CTL; in the islet, B-cells promote survival of activated CD8⁺ T-cells at the CTL transition stage, thereby accelerating disease progression.

It is known that B-cells are important in type 1 diabetes development in the NOD mouse at the level of CD4⁺ T-cell priming to islet antigen GAD (8,9,23,24). The need for B-cells at the level of the anti-islet CD8⁺ T-cell, postulated to be the main effectors of β-cell destruction in both NOD mice and humans (5,25,26), is not clear and may depend on the specificity of the CD8⁺ T-cell (27,28). Here we used the TNFα-NOD mouse model of type 1 diabetes, where intra-islet TNFα-driven chronic inflammation resulted in CD8⁺ T-cell–dependent, CD4⁺ T-cell–independent destruction of β cells (16) to determine the importance of the B-cell–CD8⁺ T-cell relationship in the presence of a normal T-cell repertoire. We showed that TNFα-μMT^−/− mice had significantly reduced kinetics of type 1 diabetes development compared with TNFα-μMT^+/+ mice.

There are several possibilities as to why the ablation of B-cells significantly decrease the kinetics of diabetes development even in the presence of chronic inflammation: 1) failed priming of the CD8⁺ T-cell repertoire to islet antigen either directly or indirectly through production of Ig that enables cross-presentation of islet antigens by DCs (29); 2) failed recruitment of CD8⁺ T-cells to islets; 3) enhanced functionality of CD4⁺Foxp3⁺ Treg cells that control autoreactive T-cells (30–32); 4) impaired differentiation of activated CD8⁺ T-cells to CTLs (33); and 5) decreased survival of activated CD8⁺ T-cells/CTLs (34). Our data suggest the first three hypotheses do not account for the phenotype seen in TNFα-μMT^−/− mice. Restoration of B-cells devoid of secretory Ig production reestablished the accelerated kinetics of disease, and islet-antigen primed CD8⁺ T-cells are present in the lymphatic system and respond to stimulation with islet-residing APCs, presenting a vast range of islet peptides. Interestingly, CD4⁺ T-cell priming is unimpaired in our model, contrasting with reports in unmanipulated NOD mice in which CD4⁺ T-cell priming to GAD is defective (11). This disparity likely reflects our experimental system, for which we used TNFα-exposed intra-islet APCs and a vast array of islet antigens. Nevertheless, CD4 depletion experiments showed CD4⁺ T-cells are redundant for disease progression. Surprisingly, CD4⁺ T-cells also do not contribute to suppression of disease in TNFα-μMT^−/− mice, a finding that contrasts to studies by Hu et al., where depletion of B-cells in adult NOD mice delays disease, in part, by enhancing CD4⁺ Treg mechanisms (35). Our data highlight the unique pathway by which autoreactivity is dampened independent of CD4⁺ Treg cells, a finding that has potential implications for autoimmune disease linked to failed CD4⁺ Treg functionality.

The answer to why B-cell ablation delays diabetes development seems to lie within the CD8⁺ T-cell population, with two distinct mechanisms operating depending on the environment investigated. In the PLNs, B-cell deficiency affects the proliferation and differentiation of primed CD8⁺ T-cells but not their survival. This was evident by similarities in both the numbers of and significant decreases in both BrdU⁺ cells, as well as an absence of IFNγ and granzyme B production by PLN-residing CD8⁺ T-cells in TNFα-μMT^−/− mice compared with TNFα-μMT^+/+ mice. However, in inflamed islets, although the ability of CD8⁺ T-cells to proliferate and differentiate to CTL is not dependent on B-cells, their survival is. This latter hypothesis was based on observations that at 8 weeks of age, when transition of intra-islet CD8⁺ T-cells to CTL occurs, there was an increased rate of apoptosis and an ∼30% drop in the absolute numbers of CD8⁺IFNγ⁺ and CD8⁺granzyme B⁺ T-cells in TNFα-μMT^−/− mice compared with TNFα-μMT^+/+ mice. Examination of CD8⁺ T-cell proliferation, survival, and numbers at earlier, pre-CTL differentiation stages (4–6 weeks of age) revealed no difference between the two strains of mice, suggesting this effect of B-cells on CD8⁺ T-cell survival is both stage and environment dependent (not shown). This inability of CD8⁺ T-cells to proliferate and develop into CTLs in the PLN of TNFα-μMT^−/− mice may reflect impaired B-cell–DC cooperation for efficient CTL generation due to an absence in BAFF–TACI signaling between the two cell populations (33). In support of this hypothesis, depletion of B-cells in adult NOD mice results in DCs defective in inducing expansion of and IFNγ production from both CD4⁺ BDC2.5 TcR and insulin-specific CD8⁺ TcR cells (35).

Nevertheless, in the islets, the proliferation and differentiation of CD8⁺ T-cells to CTL are not dependent on B-cells. These intriguing findings likely reflect the uniqueness of the inflamed islet environment created by TNFα, empowering residing DCs to efficiently present islet antigens to CD8⁺ and CD4⁺ T-cells and impart the necessary signals for the former to differentiate to CTLs. Indeed, both absolute numbers and activation status of intra-islet DCs were identical in the two strains of mice (not shown). This finding is not surprising considering the potency of TNFα-associated inflammation to promote DC functionality independent of conventional activation signals (16). Despite these attributes, it is clear that even in inflamed islets activated DCs alone are not sufficient to enhance progression to diabetes; B-cells are necessary to provide direct, or indirect, survival signals to responding CD8⁺ T-cells during their transition phase to CTLs. It is interesting that TNFα-μMT^−/− mice eventually succumb to diabetes, suggesting that in chronically inflamed islets, eventually the need for B-cells is compensated by as yet unknown mechanisms. Future studies resolving the disparity in the roles B-cells play in the PLNs and islets, as well as identification of compensatory pathways that push B-cell–independent assault of β cells, are ongoing. In addition, it will be important to determine whether this B-cell–CD8⁺ T-cell relationship is operational in NOD mouse models that require CD4⁺ T-cells to promote diabetes progression.

Despite that autoantibody production to islet antigens is a strong predictor of future diabetes development in prediabetic individuals (36,37), subscribing a role for B-cells in diabetes development in humans is controversial, principally because of an absence of direct experimental evidence and documentation of a single patient with X-linked agammaglobulinemia developing diabetes (38). Thus, translating data from inbred strains of NOD mice to genetically diverse humans should be viewed with caution. Nevertheless, clinical trials using rituximab—a humanized anti-human CD20 mAb that depletes B-cells and has been shown to ameliorate T-cell–mediated rheumatic arthritis—are ongoing to test its efficacy to prevent/reverse type 1 diabetes (www.diabetestrialnet.org). Our data suggest B-cell depletion in patients with clinical manifestations of diabetes but not overt disease could significantly delay disease progression. This hypothesis has recently been substantiated by studies in NOD mice transgenic for human CD20 on B-cells, where B-cell depletion using anti-CD20 Abs during the insulitis phase resulted in significant delay in disease (35). The emergence of data from the ongoing clinical trials will be imperative for determining the importance of B-cells in human type 1 diabetes development and the validation in translation of the findings from B-cell–deficient NOD mouse models to humans.

This work was supported by the Juvenile Diabetes Research Foundation and the Wellcome Trust (E.A.G.). E.A.G. is a Wellcome Senior Research Fellow in Basic Biomedical Science. F.S.W. is a Wellcome Senior Research Fellow in Clinical Science. P.S. is supported by the Canadian Institutes of Health and Research and is a scientist of the Alberta Heritage Foundation for Medical Research.

We would like to thank Jason Cooper for statistical support, Anna Petrunkina-Harrison for cell sorting expertise, and Philip Spence for critical reading of the manuscript.