Marginal-Zone B-Cells of Nonobese Diabetic Mice Expand With Diabetes Onset, Invade the Pancreatic Lymph Nodes, and Present Autoantigen to Diabetogenic T-Cells

C57BL/6, BALB/c, and CBA mice were obtained from the Animal Resource Center (Perth, Western Australia, Australia). NOD/Lt (NOD) mice were obtained from the Walter and Elisa Hall Institute (Melbourne, Australia). Mice were monitored for diabetes twice weekly from 10 weeks of age onwards, as described (25). The incidence of diabetes in NOD mice in our facility is ∼80% by 40 weeks of age. All experimental procedures were conducted using 6- to 10-week-old mice, unless otherwise specified. All animal experiments were performed with the approval of the St. Vincent's Campus Animal Experimentation Committee.

Lymphocytes were isolated from spleen, peripheral lymph nodes (PLNs), and pancreata of various strains of mice using standard techniques. Primary biotin–fluorescein isothiocyanate–, phycoethrin-, PerCP-, and APC-labeled monoclonal rat antibodies against mouse cell-surface antigens B220/CD45R (RA-6B2), CD4 (GK1.5), CD8a (53-6-7), IgM (11/41), CD21/CD35 (7G6), CD23/Fc RII (B3B4), CD1d/CD1.1/Ly-3B (1B1), CD9 (KMC8), CD44 (Pgp-1, Ly-24) (IM7), Vβ4 T-cell receptor (KT4), I-A^K (AB^K) (10-3.6), CD86 (B7-2) (GL1), CD80 (B7-1) (16-10A1), as well as secondary reagents, were purchased from BD Biosciences (San Jose, CA). Flow cytometric analysis was conducted on a FACScalibur flow cytometer (BD Biosciences). Splenic B-cell subpopulations were identified based on the expression pattern of IgM, B220, CD21, and CD23, as described by Loder et al. (8). Thus, FoBs are defined as CD23^hi, IgM⁺, and CD21^low cells; MZBs and MZB-like cells are defined as CD23^low, IgM^hi CD21^hi, CD1d^hi, and CD9^hi cells.

Freshly isolated splenic tissue was snap frozen and later analyzed by immunohistochemistry using anti-mouse CD1d-biotin (BD Biosciences) and rat anti-mouse Moma-1 (Serotec/Australia Laboratory Services). Primary antibody labeling was revealed with horseradish peroxidase–linked anti-rat IgG and alkaline phosphatase–linked streptavidin. Chromogenic substrate reagents diaminobenzidine (Sigma) and NBT/BCIP (Sigma) were used to develop horseradish peroxidase and alkaline phosphatase, respectively.

Expression of the sphingosine-1-phosphate (S1P) receptors S1P1 and SIP3 on fluorescence-activated cell sorter (FACS)-purified FoB and MZBs was determined by quantitative real-time PCR as described (26). Primers for S1P1, S1P3, and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) are listed as follows: S1P1 forward 5′GCGCTCAGAGACTTCGTCTT3′, reverse 5′ACCAGCTCACTCGCAAAGTT3′; S1P3 forward 5′CCTTGCAGAACGAGAGCCTA3′, reverse 5′TTCCCGGAGAGTGTCATTTC3′; and GAPDH forward 5′CTCATGACCACAGTCCATGC3′. All values were normalized to GAPDH mRNA levels.

Responses to S1P were determined essentially as described (27). Briefly, 5 × 10⁵ splenocytes were seeded in the upper chamber of transwell plates with a 3-μm filter (Corning Costar, Lindfield, NSW, Australia). Transmigration of B-cell subpopulations in response to increasing concentrations of recombinant S1P (Sigma) (0.1–10 nmol/l) was assessed after 3 h by phenotypic analysis (FACS) of cell populations in the lower chamber.

Splenic B-cells were purified from 5-week-old NOD mice using a MACS Pan-B-cell isolation kit (Miltenyi Biotec, Sydney, Australia). FoB and MZB populations were sorted to 90% purity on a FACSAria cell sorter (BD Biosciences) based upon the expression of IgM, B220, CD21, and CD23. For in vitro stimulation assays, cells were seeded at 1 × 10⁵ per well into round-bottom microtitre plates in 100 μl medium (RPMI-1650; Life Technologies/Invitrogen; 10% heat-inactivated FCS and 1:100 penicillin/streptomycin; Gibco Life Technologies; and 50 μmol/l 2-mercaptoethanol; Merck) and cultured in triplicate with either the F(ab)2 fragment of goat anti-murine IgM (μ-chain specific, 20 μg/ml; Jackson Immunoresearch), interleukin-4 (100 ng/ml; Gibco/Invitrogen), anti-mouse CD40 (HM40-3) (1 μg/ml; BD Biosciences), lipopolysaccharide (500 ng/ml; Gibco/Invitrogen), or bacterial DNA (CpG dinucleotide) (3 μg/ml; Gibco/Invitrogen) for 48 h. Cells were pulsed for the last 24 h with ³[H]-thymidine (1 μCi/well; Amersham Biosciences), harvested, and then assayed for ³[H]-thymidine incorporation (cpm).

For in vivo T-cell priming, female NOD mice were immunized on the hind flank with 50 μg of the insulin peptide B9:23 SHLVEALYLVCGERG (Mimotopes, Clayton, Australia) in Freunds’ complete adjuvant. Seven days later, T-cells from the local lymph nodes were isolated by a MACS Pan-T-cell isolation kit (Miltenyi Biotec). Splenic B-cells, FoBs, or MZBs were obtained as above. For the assay, 2.5 × 105 T-cells were cultured with either 2.5 × 105 total B-cells, MZBs, or FoBs primed with either 200 μg B9:23 peptide or 200 μg purified porcine insulin (I5523; Sigma). Cultures were pulsed at 72 h for the last 16 h with ³[H]-thymidine (1μCi/well; Amersham Biosciences), harvested, and then assayed for ³[H]-thymidine incorporation (cpm). Preliminary experiments established this to be the optimal protocol to assess APC activity.

To determine whether the increase in MZB-like cells in the PLN related to migration or proliferation locally, 6- and 16-week-old female NOD mice were pulsed with 5-bromo-2-deoxyuridine (BrdU) (2 mg/kg) for 3 consecutive days and assessed for BrdU incorporation on day 4 by a FACScalibur flow cytometer (BD Biosciences). B-cell subpopulations were identified by FACS, as described above.

Statistical significance was determined by calculating P values using a Mann-Whitney and t test on Instat software (GraphPad Softward, San Diego, CA).

NOD mice have a phenotypically altered B-cell compartment favoring an increased frequency of MZBs (10,11). In young NOD mice, this relates to a ∼70% decrease in the absolute numbers of FoBs (n ≥ 5) (P < 0.01), and a small, but significant, increase in MZB numbers of over that seen in C57BL/6 and BALB/c mice (increased ∼1.4-fold, P < 0.05, n ≥ 5 and ∼2-fold, P < 0.01, n ≥ 5, respectively) (Fig. 1A–C). The marked decrease in FoB numbers contributes to marked lymphopenia, evidenced by the reduced number of total splenocytes in female NOD mice (Fig. 1D). The relative decrease in FoB numbers combined with the minor increase in MZB numbers manifested as a marked alteration in the FoB-to-MZB ratio in NOD mice; this ratio was decreased from ∼10:1 (FoB:MZB) in C57BL/6 and BALB/c mice to ∼2:1 in NOD mice (n ≥ 5) (P < 0.01) (Fig. 1C).

An additional unusual feature of the NOD MZB population is that their numbers increase in parallel with insulitis (11). However, the relationship between their expansion and the onset of overt diabetes, which occurs from ∼13 weeks of age and onwards in NOD mice, has not been investigated. Accordingly, we conducted a longitudinal study of splenic MZB numbers in a cohort of female NOD mice (n = 70) from 5 to 23 weeks of age to determine how the expansion in MZB numbers correlated with diabetes development. In our control cohort of female NOD mice (n = 30), frank diabetes manifested at ∼13 weeks of age, reaching an incidence of ∼70% by 30 weeks of age (Fig. 2A). Analysis of MZB numbers revealed that the absolute number of MZBs dramatically increased approximately threefold (n ≥ 6) (P < 0.001) from a steady state of ∼5 × 10⁶ at 11 weeks of age to ∼15 × 10⁶ at 12 weeks of age (Fig. 2B). In contrast, no changes were observed in MZBs in C57BL/6 mice over this same period. The MZB expansion was accompanied by an approximately threefold increase in FoB numbers (n ≥ 6) (P < 0.01) from ∼10 × 10⁶ at 11 weeks of age to ∼27 × 10⁶ at 12 weeks of age (Fig. 2C). Thus, the lymphopenic state of young NOD mice (28) is due to a marked reduction in the numbers of FoBs, which corrects at ∼12 weeks of age due to an expansion of both MZBs and FoBs, just before the occurrence of frank diabetes. The change in MZB number at this time may relate to an underlying alteration in the dynamic-controlling B-cell development, as there was a parallel increase in the absolute numbers of T2MZ transitional B-cell precursors in the spleen. Thus, T2MZ increased approximately fourfold (n ≥ 6) (P < 0.001), from ∼2 × 10⁶ to ∼8.3 × 10⁶ at 12 weeks of age (Fig. 2E). Therefore, we show that MZB numbers do not remain static in NOD mice but show an age-dependent increase. Interestingly, this expansion correlates with the onset of frank diabetes in female NOD mice.

The increased accumulation of MZBs in the spleen of female NOD mice was clearly apparent by histological analysis, evidenced as an enlarged region of CD1d-bright cells colocalizing with MOMA-1⁺ macrophages surrounding the follicles (Fig. 3A). MZB localization at the splenic marginal sinus is controlled by chemotactic cues provided by the lysophospholipid S1P and integrin-dependent interactions (27,29). Since expression of α1β2 and β2 integrins have been shown to be normal in NOD MZBs (10), we focused on examining the S1P pathway. We found that S1P1 and S1P3, the S1P receptors responsible for MZB retention (27), were significantly (n = 3) (P < 0.001) increased on MZBs from NOD versus C57BL/6 mice (Fig. 3B). Examination of the chemotactic response of MZBs to S1P revealed that NOD MZBs were approximately twofold more sensitive than MZBs from C57BL/6 mice to S1P at 1 nmol/l (n = 3) (P < 0.001) and 10 nmol/l (n = 3) (P < 0.01) (Fig. 3C). Pretreatment of NOD mice with FTY720 before harvesting the B-cells, to downregulate S1P1 expression (27), did not negatively impact the heightened response to S1P, suggesting that MZB chemotaxis was independent of S1P1, as has been previously reported (27). In contrast, NOD FoBs expressed normal levels of S1P1 and S1P3 (Fig. 3B) and, as expected (28), were not responsive to S1P (Fig. 3D). Together, these data indicate that NOD MZBs are hyperresponsive to chemotactic cues provided by S1P, due to an increase in S1P3 expression. The increased sensitivity of NOD MZBs to S1P may support their unusually magnified accumulation at the splenic marginal sinus.

We next examined MZB responses from NOD mice to mitogens and costimulatory ligands to determine whether their heightened chemotactic responses reflected a generalized hyperresponsiveness or was selective for particular pathways. MZB and FoB populations were isolated to 90% purity by a FACSorter, as depicted in Fig. 4A. As reported previously (16), we found that MZBs from either NOD or C57BL/6 mice responded poorly to B-cell receptor cross-linking (i.e., ≤500 cpm) (Fig. 4B), a phenotype ascribed to an increased sensitivity of MZBs to activation-induced apoptosis following B-cell receptor stimulation (16,17). However, we found that NOD MZBs were approximately twofold (P < 0.001, n = 5) more responsive to CD40 ligation than MZBs from C57BL/6 mice, a pattern that was heightened with the addition of interleukin-4 (Fig. 4C and D). That is, NOD MZBs were approximately fourfold (n = 5) (P < 0.0001) more responsive to CD40 plus interleukin-4 stimulation than MZBs from C57BL/6 mice. Moreover, NOD MZBs were ∼10-fold (n = 5) (P < 0.001) more responsive to toll-like receptor (TLR) 9 stimulation than MZBs from C57BL/6 mice (Fig. 4E). In contrast, NOD and C57BL/6 MZBs were equally responsive to TLR4 stimulation (Fig. 4F).

We found that NOD FoBs proliferated strongly to B-cell receptor stimulation, as did FoBs from C57BL/6 mice (Fig. 4B). Similar to NOD MZBs, NOD FoBs were hyperresponsive to CD40 ligation in the presence, or absence, of interleukin-4, showing an approximately two- and threefold (n = 5) (P < 0.01) increase over the proliferative response of C57BL/6 FoBs, respectively (Fig. 4C and D). In contrast to MZBs, NOD and C57BL/6 FoBs showed equivalent proliferative responses to TLR9 and TLR4 stimulation (Fig. 4E–F). Thus, NOD FoBs do not exhibit the same pattern of hyperresponsiveness toward TLR9 stimulation as NOD MZBs. Therefore, NOD MZBs exhibit a cell-intrinsic pattern of hyperresponsiveness to selected B-cell mitogens, namely TLR9 ligation and CD40 ligation compared with MZBs from C57BL/6 mice.

One potential mechanism by which B-cells contribute to diabetes pathogenesis is as APCs, presenting autoantigens to self-reactive T-cells (5–7). We were curious as to whether MZBs were able to act as APCs, given that they were hyperresponsive to TLR ligation and CD40 ligation, stimuli important for the maturation and activation of dendritic cells (30,31). Unfractionated splenic NOD B-cells were able to present the insulin peptide B9:23 and efficiently drive T-cell proliferation (n = 5) (P < 0.001) (Fig. 5A). To examine the APC capacity of B-cell subsets, MZBs and FoBs were purified as in Fig. 4A. We found that MZBs were able to efficiently present the B9:23 peptide to self-reactive T-cells and elicit their proliferation (n = 5) (P < 0.001) (Fig. 5B). Indeed, MZBs could present B9:23 with similar efficiency (n = 5) (P < 0.01) to that of FoBs (Fig. 5B and C).

Having shown that MZBs could present insulin peptides to self-reactive T-cells, we next examined whether MZBs were able to capture and process intact insulin as an autoantigen. As has been shown previously (32), for both NOD and C57BL/6 B-cells, we also found that unfractionated B-cells could process and present insulin to diabetogenic T-cells (Fig. 5D). Significantly, we found that both MZBs (n = 5) (P < 0.01) and FoBs (n = 5) (P < 0.05) were able to effectively activate self-reactive T-cells (Fig. 5E and F), indicating that MZBs in particular are capable of capturing and processing intact autoantigens so as to present them as peptide–major histocompatibilty complex (MHC) complexes to self-reactive T-cells. Neither the MZB nor FoB populations alone were able to recapitulate the efficacy of unfractionated splenic B-cells as APC (Fig. 5A–F), suggesting that both subpopulations contribute to the APC activity of unfractionated B-cells.

The PLN is a critical site for the presentation of diabetes autoantigens to self-reactive T-cells (33–35). Since we demonstrated that splenic NOD MZBs can capture and present autoantigen to self-reactive T-cells, it was important that we examined whether MZBs were present in PLNs of NOD mice. A precedent for extrasplenic migration of MZBs has been seen in patients with Sjogren's syndrome and also BAFF transgenic and NZB/W F1 mice (22). We could detect an increased frequency of CD23^low, IgM^hi, and CD21^hi cells, a phenotype reminiscent of splenic MZBs, in the PLN of 16-week-old NOD mice (Fig. 6 vs. Fig. 1A). These MZB-like cells were not detected at the same frequency in the inguinal lymph node (ILN), mesenteric lymph node (MLN), the peripheral blood, or the peritoneal cavity of either NOD or C57BL/6 mice. These MZB-like cells were further characterized as being CD1d^hi and CD9^hi (data not shown), additional markers delineating MZBs in the spleen (36,37).

NOD B-cells present autoantigen to diabetogenic CD4⁺ T-cells (5–7), and we demonstrated that NOD MZBs exhibit APC activity. Significantly, we found that the MZB-like cells present in the PLNs were enriched for the expression of MHC class II, CD80, and CD86 at 16 vs. 6 weeks of age (Fig. 7A). Concordantly, the frequency of antigen-activated diabetogenic Vβ4⁺ CD4⁺ T-cells in the PLNs, as evidenced by high CD44 expression, was increased from ∼12.75% in 6-week-old NOD mice to ∼37.5% in 16-week-old NOD mice (n = 5) (P < 0.001) (Fig. 7B). Thus, the accumulation of MZB-like cells in the PLNs with a phenotype indicative of APC activity correlates with an increase in the frequency of antigen-experienced diabetogenic CD4⁺ T-cells. Together with our in vitro studies (Fig. 5), these data support the concept that MZBs may be capable of presenting autoantigen in vivo.

Intriguingly, a CD23^low, but IgM^hi, CD21^hi MZB-like population was also observed in the pancreas of NOD but not C57BL/6 mice (Fig. 6 vs. Fig. 1A). Using a cutoff of 1% determined from analysis of pancreata from C57BL/6 mice, we found that ∼60% (n = 31 of 48) (P < 0.001) of female NOD mice between 4 and 16 weeks of age exhibited MZBs in the pancreas (MZB mean percentage ± SE = 3.89 ± 0.58%; n = 31). The increased number of MZB-like cells in the PLNs of NOD mice, and their localization to the pancreas, suggests that this does not represent random migration but rather was related to the pathophysiology of diabetes.

A longitudinal study to enumerate MZB-like cell numbers in the PLN over time revealed a marked ∼15-fold increase (n ≥ 6) (P < 0.005) in the absolute numbers of MZBs in NOD, but not C57BL/6 mice, at ∼16 weeks of age (Fig. 8A). In contrast, neither the frequency nor number of MZBs in the ILN and MLN of NOD mice changed over this same period (Fig. 6 and Fig. 8A). To determine whether the increase in MZB-like cells in the PLN related to migration or proliferation locally, we analyzed MZB-like cell BrdU incorporation rates at 6 and 16 weeks of age. As shown in Fig. 8B, a substantial increase in the proliferation rates for MZBs of the spleen and MZB-like cells of the PLN was observed at 16 vs. 6 weeks of age in female NOD mice (n ≥ 5 per group) (P < 0.001), correlating with the increase in absolute numbers of MZB-like cells in the PLN (Fig. 8A). Thus, local proliferation of MZB-like cells in the PLN contributes to their expansion at 16 weeks of age.

Using the numbers of MZBs in the PLN of age-matched C57BL/6 mice as a cutoff, it could be seen that the frequency of NOD mice with a number of MZB-like cells in the PLN greater than that observed for C57BL/6 also increased with age, reaching ∼80% (n > 7) (P < 0.01) at 16 weeks of age (Fig. 8C). Of further interest is the similarity between the frequency of NOD mice harboring MZB-like cells in the PLN at 16 weeks of age and the incidence of type 1 diabetes in NOD mice.

One important role for B-cells in the pathogenesis of type 1 diabetes is as an APC (1), capturing and presenting autoantigen to self-reactive CD4⁺ T-cells driving their activation and expansion (5–7). To date, the role of MZBs versus FoBs as APCs has not been considered in this context. Here, we here provide the first evidence that in the NOD model of diabetes, MZB have the potential to break tolerance to autoantigens by virtue of their capacity to present insulin-derived peptide-MHCs to self-reactive T-cells. We found that MZBs were neither the only B-cell subtype with this function, nor the most efficacious, at least in vitro, where MZBs and FoBs exhibit a similar capacity to activate diabetogenic T-cells. However, the contribution of different B-cell subsets as important APCs in vivo will be determined by their relative ratios at specific anatomical locations and also by their activation status.

In this regard, the extrasplenic localization of MZB-like cells may be particularly significant. The expansion and aberrant migration of MZBs is associated with autoimmunity in BAFF transgenic mice (20), which develop pathologies reminiscent of Sjogren's syndrome and systemic lupus eyrthematosus (22). In BAFF transgenic mice, MZBs migrate to disease-related sites, including lymph nodes but also the salivary gland, the target organ in Sjogren's syndrome (38). In human subjects with Sjogren's syndrome, CD27⁺ memory B-cells, analogous to rodent MZBs, are present in the peripheral blood and colonize the salivary glands (38). With respect to diabetes pathogenesis, the PLN is recognized as a site critical for the presentation of autoantigen to self-reactive T-cells (33–35). We show that MZB-like cells accumulate at this site in increasing numbers in an age-dependent manner, suggesting that is not a random event but rather is closely linked to the conversion to overt diabetes in the NOD model.

Regarding the activation status of NOD MZBs, we demonstrate that NOD MZBs exhibit a selective hyperresponsiveness to S1P, TLR9, and CD40 ligation as well as enhanced expression of costimulatory molecules. The molecular cause for some of these changes may relate to enhanced nuclear factor (NF)-κB signaling. NOD B-cells exhibit an exaggerated basal level of NF-κB activation (32), which may sensitize them to NF-κB–responsive signaling pathways such as CD40 and TLR9 to increase S1P receptor CD80 and CD86 expression. Of note, BAFF activates NF-κB (22), and B-cells from autoimmune-prone BAFF transgenic mice also accumulate at the marginal sinus (20) and are enriched for APC activity (39), which is supportive of a link between activation of NF-κB and B-cell APC function in autoimmunity.

Signals received through TLRs and CD40 are important for the maturation of professional APCs such as dendritic cells, favoring their acquisition of a phenotype that supports the induction of antigen-specific T-cell–dependent immunity (30,31,40). We demonstrate that NOD MZBs can process and present autoantigen. The MZB-like cells of the PLN exhibit a number of features indicative of in vivo activation and enhanced APC potential. This includes their increased proliferative status, a classic hallmark of lymphocyte activation, and also their high expression levels of costimulatory molecules. Significantly, the presence of these activated MZB-like cells correlated with an increased frequency of activated diabetogenic CD4⁺ T-cells in the PLN, consistent with their demonstrated APC potential.

The natural ability of MZBs to rapidly respond to innate signals (24), as well as present antigen to T-cells (18), conceptually places MZBs within a unique position in the immune system as cells being able to connect innate with adaptive immune responses. This may have relevance for the etiology of type 1 diabetes because though an innate trigger for diabetes has not been convincingly shown, it is highly speculated that such an event precipitates the activation of self-reactive T-cells and subsequent β-cell destruction (41,42). The specialized role of MZBs in immunity, coupled with the altered features of the NOD MZB compartment we now present, marks MZBs as candidate players in the series of events leading toward autoimmunity. We argue that MZBs may be important in diabetes pathogenesis by virtue of their capacity to respond to, and thus integrate, innate-derived inflammatory signals with the capture and presentation of autoantigens, enabling MZBs to direct effector T-cell responses against β-cell autoantigens.

E.M. is supported by a National Health and Medical Research Council Dora Lush Fellowship. This work has been supported by grant 5-2005-1132 from the Juvenile Diabetes Research Federation and a New South Wales BioFirst Award (to S.T.G.).

We would like to thank Dr. Pablo Silveira for helpful comments and critical reading of this manuscript. We also gratefully acknowledge the expert technical assistance of Michael Pickering and Eric Schmied.