B Lymphocytes Are Critical Antigen-Presenting Cells for the Initiation of T Cell-Mediated Autoimmune Diabetes in Nonobese Diabetic Mice

In both humans and nonobese diabetic (NOD)² mice, insulin-dependent diabetes mellitus (IDDM) results from autoimmune destruction of pancreatic β cells by T lymphocytes (1, 2). The primary diabetes susceptibility (Idd) genes in both humans and NOD mice are provided by multiple alleles within specific MHC haplotypes (3, 4, 5). However, the development of IDDM in both genera also requires contributions from multiple non-MHC Idd genes (3, 4, 5). In NOD mice, it appears that interactions between Idd susceptibility genes both within and outside of the unusual H2^g7 MHC haplotype (K^d, I-A^g7, I-E^null, D^b) diminishes the ability of hematopoietically derived APC to activate various immunoregulatory functions that would normally block the development and/or function of T lymphocytes autoreactive against pancreatic β cells (6).

Subpopulations of hematopoietically derived APC include B lymphocytes, macrophages, and dendritic cells (DC). To partially address which of these APC subpopulations could exert immunotolerogenic defects underlying the development of IDDM, we produced a stock of NOD mice made deficient in B lymphocytes by congenic transfer of an Igμ gene functionally disrupted by homologous recombination (formally designated as Igh6^tmiCgn and, here, as Igμ^null) (7). It had been reported that B lymphocytes have a greater capacity to induce various immunotolerogenic functions than other APC populations (8, 9, 10). Thus, it was anticipated that the elimination of B lymphocytes would not alter, or might even accelerate, IDDM development in NOD mice. However, we were surprised to find that B lymphocyte-deficient NOD.Igμ^null mice are IDDM resistant (7). Similar results were obtained by two other groups that produced B lymphocyte-deficient NOD mice by either congenic transfer of an Igμ^null allele or treatment with a μ chain-specific Ab (11, 12). The finding that IDDM is inhibited rather than accelerated in NOD.Igμ^null mice indicated that B lymphocytes played a newly identified diabetogenic role that is distinct, but not mutually exclusive, from the APC-controlled tolerogenic defects underlying the original development of β cell-autoreactive T cells in NOD mice.

This newly identified diabetogenic role for B lymphocytes in NOD mice could be as APC with an unique ability to process and present certain β cell Ags to autoreactive T cells that have been generated as a consequence of the tolerogenic defects described above. Alternatively, B lymphocytes may contribute to IDDM in NOD mice through their ability to secrete autoantibodies that bind to pancreatic β cells and subsequently trigger autoreactive T cells through an Ab-dependent cell-mediated cytotoxicity response. The present study was conducted to determine whether B lymphocytes contribute to the development of T cell-mediated autoimmune IDDM in NOD mice through either of these two mechanisms.

NOD/Lt mice are maintained at The Jackson Laboratory (Bar Harbor, ME) by brother-sister mating. Currently, IDDM develops in 90% of female and 63% of male NOD/Lt mice by one year of age. Derivation of a “speed congenic” N7 backcross stock of B lymphocyte-deficient NOD.Igμ^null mice fixed to homozygosity for linkage markers delineating all previously identified Idd loci of NOD origin has been described previously (7). From >900 NOD.Igμ^null mice generated to date, only 4 female breeders of >30 wk of age have spontaneously developed IDDM. The previously described stock of T and B lymphocyte-deficient NOD-scid (official designation NOD-Prkdc^scid) mice is maintained at the N11 backcross generation (13, 14). Similarly, the previously described congenic stock of MHC class I and CD8⁺ T cell-deficient NOD.β2m^null (official designation NOD.β2m^tm1Unc) mice is also maintained at the N11 backcross generation (15). These latter two strains served as progenitors for the previously described stock of NOD mice homozygous for both the scid and β2m^null mutations (designated NOD-scid.β2m^null) (16, 17). Mice housed at The Jackson Laboratory were maintained under specific pathogen-free conditions and allowed free access to food (National Institutes of Health diet 31A, Purina, Richmond, IN) and acidified drinking water. In addition, all stocks of scid mice were treated for 3 days/week with trimethoprim-sulfamethoxazole (Sulfatrim, Barre National, Baltimore, MD) in the drinking water. Some experiments used NOD/Lt and NOD.Igμ^null mice housed in the animal facility of the Department of Microbiology and Immunology at the University of North Carolina at Chapel Hill. Mice in this facility were maintained under viral pathogen-free conditions and allowed access to National Institutes of Health diet 31A (Purina) and acidified drinking water.

The indicated mice were monitored for development of glycosuria with Ames Diastix (kindly supplied by Miles Diagnostics, Elkhart, IN). Glycosuric values of ≥3 were considered diagnostic of diabetes onset. Pancreases from mice assessed for insulitis development were fixed in Bouin’s solution and sectioned at three nonoverlapping levels. Granulated β cells were stained with aldehyde fuchsin and leukocytes with a hematoxylin and eosin counterstain. Islets (at least 25/mouse) were individually scored as follows: 0, no lesions; 1, peri-insular leukocytic aggregates, usually periductal infiltrates; 2, <25% islet destruction; 3, >25% islet destruction; and 4, complete islet destruction. An insulitis score for each mouse was obtained by dividing the total score for each pancreas by the number of islets examined. Data are presented as mean insulitis score (MIS) ± SEM for the indicated experimental group.

The cloning and preparation of the 65-kDa isoform of the candidate murine β cell autoantigen glutamic acid decarboxylase (GAD) has been described previously (18). Briefly, the cDNA was engineered to encode six histidine residues at the COOH terminus of the protein. Recombinant murine GAD65 was generated in SF21 cells using a Baculovirus expression system and purified using a Ni²⁺-conjugated resin (Invitrogen, San Diego, CA). The GAD65 was further purified by preparative SDS-PAGE. The protein was then electroeluted and extensively dialyzed against PBS. An SF21 cell extract purified in an identical manner to the recombinant GAD65 protein is not antigenic to T cells. Keyhole limpet hemocyanin (KLH) (Sigma, St. Louis, MO) was used as a control Ag.

The indicated number of NOD or NOD.Igμ^null mice were immunized in a hind foot pad with a 50-μl emulsion of CFA containing 20 μg of GAD65. After 10 days, T cells were purified by the previously described panning technique (14) from pooled draining lymph nodes (LN) of mice in each group. Triplicate aliquots of 2.5 × 10⁵ T cells were seeded into flat-bottom 96-well microtiter plates in a final volume of 200 μl of the previously described culture medium (19) containing the indicated concentration of GAD, plus 2.5 × 10⁵ irradiated (2000 rad) splenic leukocytes from NOD or NOD.Igμ^null mice as a source of APC. Following a 72-h incubation at 37°C in a 95% air/5% CO₂-humidified atmosphere, the cultures were pulsed with 1 μCi/well of [³H]thymidine for an additional 16 h. The cultures were then harvested, and [³H]thymidine incorporation was determined using an LKB Betaplate 1205 system (LKB Instruments, Gaithersburg, MD). Data are presented as mean cpm ± SEM.

In other experiments, the indicated mice were immunized in a hind foot pad with a 50-μl emulsion of CFA containing 20 μg of GAD65 or KLH. At 10 days after priming, single-cell suspensions were prepared from pooled draining LN of two to three mice in each experimental group. Triplicate aliquots of 5 × 10⁵ LN cells were seeded into flat-bottom 96-well microtiter plates in a final volume of 200 μl of medium with or without 10 μg/ml of GAD or KLH plus 5 × 10⁵ irradiated (2000 rad) NOD or NOD.Igμ^null splenocytes as an additional source of APC. Following a 72-h incubation, the cultures were pulsed with 1 μCi/well of [³H]thymidine for an additional 16 h. Data are presented as mean Δcpm ± SEM (calculated from the mean Ag stimulated minus unstimulated responses).

GAD65-specific T cell clones were established by culturing 5 × 10⁶ splenocytes from 4-wk-old unimmunized NOD female mice in a 24-well plate in 1.5 ml of RPMI 1640, 5 × 10⁻⁵ M 2-ME, 1 mM sodium pyruvate, 1× nonessential amino acids, 1 mM glutamine, 1.0% NOD serum, and 10 μg/ml of intact murine GAD65 for 7 days. T cells (1 × 10⁶) harvested on a Lympholyte M gradient (Cedarlane Laboratories, Hornby, Ontario, Canada) were cultured with 5 × 10⁶ irradiated (3000 rad) NOD splenocytes in 1.5 ml of the above medium and 10 μg/ml GAD65 in a 24-well plate. After 3 days, the cultures were supplemented with medium containing 20 U/ml murine IL-2 (PharMingen, San Diego, CA) and maintained for an additional 3 days, at which time CD4⁺ T cells were purified using magnetic bead separation (Miltenyi Biotec, Auburn, CA). GAD65-specific CD4⁺ T cell clones were established via limiting dilution.

To assess the capacity of APC from NOD vs NOD.Igμ^null mice to activate GAD65-specific T cell clones, 2.5 × 10⁴ cells from the GAD65-specific 6E12 T cell clone were cultured with 2.5 × 10⁵ irradiated (3000 rad) splenocytes from NOD or NOD.Igμ^null mice in 0.1 ml of the above medium containing 10% FBS with or without 10 μg/ml of GAD65 for 72 h in a 96-well plate. T cell proliferation was assessed by measuring the amount of [³H]thymidine incorporation following a 16-h pulse (1 μCi/well). Data are presented as mean Δcpm ± SEM of triplicate cultures.

B lymphocytes were purified from NOD splenic leukocyte preparations using a streptavidin-conjugated magnetic bead system (Miltenyi Biotec) to deplete T cells and macrophages/granulocytes that had been prestained with biotinylated mAbs specific for CD3 (145-2C11) or Mac-1 (M1/70), respectively. Subsequent FACS analysis using an FITC-conjugated goat polyclonal antiserum specific for mouse Ig (Southern Biotechnology Associates, Birmingham, AL) indicated that >95% of the resulting cell preparation consisted of B lymphocytes.

Total Ig was precipitated from pooled serum of overtly diabetic NOD females by 40% saturation with (NH₄)₂SO₄. The precipitated Ig was then dialyzed extensively against PBS and quantified by ELISA as described previously (16). Female NOD.Igμ^null mice were injected i.p. with 70 μg of this Ig preparation twice weekly between 8 and 20 wk of age and simultaneously monitored for diabetes development. Controls consisted of NOD.Igμ^null injected on the same schedule with the PBS vehicle alone. At 20 wk of age, circulating serum Ig levels in both groups were determined by ELISA. Insulitis development was also assessed at this time.

In an initial experiment, a group of 6-wk-old NOD.Igμ^null female mice were injected i.v. with 3 × 10⁶ purified B lymphocytes from standard NOD donors. Controls consisted of NOD-scid females injected with the same preparation of B lymphocytes. At the indicated timepoints, the proportion of donor B lymphocytes among PBL in the NOD.Igμ^null and NOD-scid recipients were assessed by FACS using the FITC-conjugated polyclonal antiserum specific for mouse Ig described above.

In subsequent experiments, 3 × 10⁶ purified B lymphocytes from NOD or NOD.β2m^null female donors, respectively, were injected i.v. into 6-wk-old NOD-scid and NOD-scid.β2m^null female mice. At the indicated timepoints, the proportion of donor B lymphocytes among PBL in the NOD-scid and NOD-scid.β2m^null recipients was assessed by FACS. Total PBL counts were also determined to calculate the total number of donor B lymphocytes per ml of peripheral blood in the NOD-scid and NOD-scid.β2m^null recipients. After stable B lymphocyte repopulation was achieved, a subset of the NOD-scid and NOD-scid.β2m^null recipients were injected i.v. with 1 × 10⁷ T cells purified from the spleens of female NOD.Igμ^null donors as described previously (14). At weekly intervals thereafter, total numbers of donor B lymphocytes per ml of peripheral blood were compared in NOD-scid and NOD-scid.β2m^null recipients that had or had not been subsequently injected with NOD.Igμ^null T cells.

Female NOD.Igμ^null mice were lethally irradiated (1200 rad) at 4 wk of age and then reconstituted as described previously (20) with 5 × 10⁶ T cell-depleted syngeneic bone marrow cells that had been mixed with 3 × 10⁶ purified NOD B lymphocytes. Control chimeras consisted of NOD.Igμ^null females reconstituted with syngeneic bone marrow only. Bone marrow chimeras were then monitored at 21 wk postreconstitution for diabetes development. Pancreases from mice that remained normoglycemic through 21 wk postreconstitution were assessed for insulitis development. In addition, upon diabetes onset or at 21 wk postreconstitution, splenic leukocytes from the bone marrow chimeras were typed by FACS for the presence of B lymphocytes as described above and for CD4⁺ and CD8⁺ T lymphocytes with the mAbs GK1.5 and 53-6.72, respectively. Furthermore, at 8 wk postreconstitution, a subset of NOD.Igμ^null mice reconstituted with syngeneic bone marrow in the presence or absence of NOD B lymphocytes were assessed for presence of GAD-primed T cell responses as described above.

We hypothesized that, if the production of autoantibodies represents the primary mechanism by which B lymphocytes contribute to the development of IDDM, disease resistance would be abrogated in NOD.Igμ^null mice infused with Ig isolated from overtly diabetic NOD donors. Two groups of NOD.Igμ^null females were injected twice weekly from 8 to 20 wk of age with either 70 μg of Ig isolated from overtly diabetic NOD donors or with the PBS vehicle control. By 20 wk of age, ELISA measurements confirmed the presence of circulating Ig in sera of the experimental (69.2 ± 4.9 μg/ml), but not the control group (0.9 ± 0.2 μg/ml), of NOD.Igμ^null mice (Table I). However, IDDM failed to develop in any of the Ig-reconstituted (0/8) or control (0/8) NOD.Igμ^null female mice. Furthermore, the level of insulitis in Ig-reconstituted NOD.Ig^null mice (MIS = 0.85 ± 0.40) was not significantly different from the low levels observed in PBS-treated controls (MIS = 1.29 ± 0.36). Thus, the levels of Ig reconstitution achieved in this experiment failed to accelerate autoimmune destruction of pancreatic β cells in NOD.Ig^null mice. This indicates that the production of autoantibodies is unlikely to represent the primary mechanism by which B lymphocytes contribute to the development of T cell-mediated autoimmune IDDM in standard NOD mice.

Given that Ig infusions did not abrogate IDDM resistance or promote insulitis development in NOD.Igμ^null mice, we hypothesized that the pathogenic role for B lymphocytes in this disease may be as APC with a unique ability to process and present certain pancreatic β cell Ags to autoreactive T cells. As an initial test of this hypothesis, we determined whether NOD and NOD.Igμ^null mice differed in ability to generate T cell responses to the candidate pancreatic β cell autoantigen GAD. While clearly many different β cell proteins are recognized by autoreactive T cells in IDDM, we chose GAD as a model autoantigen for our studies based on reports that it is among the earliest targets of diabetogenic T cells in NOD mice (18, 21). NOD.Igμ^null mice were characterized by an absence of spontaneous T cell responses to GAD (data not shown). However, it was possible that GAD-reactive T cells were still present but not efficiently activated in NOD.Igμ^null mice because of a quantitative loss in APC resulting from the absence of B lymphocytes. We addressed this issue through use of an in vivo priming protocol that would amplify any GAD-reactive T cells present in either standard NOD or NOD.Igμ^null mice. As expected, T cells from GAD65-primed NOD mice proliferated when restimulated with this Ag in vitro in the presence of standard B lymphocyte-positive NOD APC and, to a lesser extent, in the presence of B lymphocyte-deficient APC from NOD.Igμ^null mice (Fig. 1). In contrast, T cells from GAD65-primed NOD.Igμ^null mice failed to respond upon antigenic restimulation in vitro in the presence of either B lymphocyte intact or deficient APC. These results indicated that the initial in vivo priming of GAD-reactive T cell responses in NOD mice requires the presence of Ag-presenting B lymphocytes. However, other APC populations such as macrophages and DC appear to be able to process and present GAD65, albeit less efficiently than B lymphocytes, to NOD T cells previously primed against this Ag.

While B lymphocytes most likely make direct contributions as APC to the initiation of GAD-reactive T cell responses in NOD mice, we recognized that they could also indirectly regulate such reactions by affecting the development of other leukocyte populations. We had previously found that both T cells and macrophages, which represent an important subpopulation of APC, develop normally in B lymphocyte deficient NOD.Igμ^null mice (7). However, it was not known whether DC, which represent another important APC subpopulation, also develop normally in NOD.Igμ^null mice. FACS analysis demonstrated the proportion of splenic DC, defined by coexpression of the cell surface markers Mac-1 and that detected by the previously described mAb 33D1 (22), were equivalent in standard NOD (0.91 ± 0.22%, n = 3) and B lymphocyte-deficient NOD.Igμ^null mice (1.16 ± 0.33%, n = 3). Thus, these results support the conclusion that B lymphocytes directly serve as necessary APC for the initiation of GAD-reactive T cell responses in NOD mice.

The results described above indicated that once activated in a B lymphocyte-dependent fashion, the response of GAD-reactive T cells from standard NOD mice can be maintained by other APC subpopulations. Assuming that this conclusion was correct, we reasoned that B lymphocyte-deficient APC should stimulate a response by the GAD-reactive T cell clone 6E12, which was originally propagated from a standard NOD mouse. As shown in Figure 2, B lymphocyte-deficient APC from NOD.Igμ^null mice were able to process and present antigenic peptides from intact GAD65 to the 6E12 T cell clone, albeit at lower levels than B lymphocyte-intact APC from standard NOD mice. Similar results were obtained using a peptide comprising amino acids 216 to 235 from GAD65, which is the antigenic target of the 6E12 T cell clone (R. M. T., personal observation). Thus, while B lymphocytes are necessary APC for initiating GAD-autoreactive T cell responses in NOD mice, once such effectors are triggered their activities can be maintained by other APC subpopulations.

The failure of NOD.Igμ^null mice to generate a T cell response to GAD suggested the diabetogenic role of B lymphocytes may be as APC with a preferential ability to process and present certain MHC class II-restricted β cell autoantigens. However, IDDM resistance in NOD.Igμ^null mice could also be explained if the absence of B lymphocytes results in a generic inability to generate a T cell response to all, rather than a specific subset, MHC class II-restricted Ags. In mouse strains other than NOD, KLH has been reported to be an MHC class II-restricted Ag that does not require B lymphocytes for presentation to T cells (23). T cells within LN from both standard NOD and NOD.Igμ^null mice primed with KLH responded equivalently when restimulated with this Ag in the presence of B lymphocyte-deficient APC (Fig. 3). However, the recall response of T cells within LN from both KLH-primed NOD and NOD.Igμ^null mice was greater in the presence of B lymphocyte-intact than B lymphocyte-deficient APC. Thus, B lymphocytes are not required for the initial priming of KLH-reactive T cells in NOD mice but do contribute to the amplification of such responses. These data demonstrate that the absence of B lymphocytes in the NOD.Igμ^null stock has only eliminated their ability to generate T cell responses to certain MHC class II-restricted Ags.

The fact that NOD.Igμ^null mice cannot generate a T cell response against GAD, but can do so to KLH, suggests that the diabetogenic role of B lymphocytes is as APC with a preferential ability to process and present certain MHC class II-restricted β cell autoantigens. As an initial test of this hypothesis, we wished to determine whether NOD.Igμ^null mice repopulated with B lymphocytes were rendered IDDM susceptible and whether this was associated with a restored ability to generate a T cell response to GAD. However, purified NOD B lymphocytes only transiently repopulated unmanipulated NOD.Igμ^null recipients (Fig. 4). These same B lymphocytes permanently repopulated NOD-scid recipients, which lack endogenous T as well as B lymphocytes. Thus, we hypothesized that a population of MHC class I-restricted cytotoxic T cells present in NOD.Igμ^null, but not NOD-scid, recipients mediates an immunologic rejection of transplanted B lymphocytes. We were able to test this hypothesis because of the availability of both standard NOD and NOD-scid mice made deficient in MHC class I expression by the presence of a β2m^null allele (15, 16, 17). Standard MHC class I-positive NOD B lymphocytes previously engrafted into NOD-scid recipients were rapidly eliminated following the subsequent infusion of T cells from NOD.Igμ^null donors (Fig. 5). In contrast, infusion of these same T cells did not mediate the elimination of MHC class I-negative B lymphocytes from NOD.β2m^null donors that had been previously engrafted into NOD-scid.β2m^null recipients. Thus, NOD.Igμ^null mice harbor a population of T cells that can mediate a MHC class I-restricted cytotoxic response against standard NOD B lymphocytes. This is most likely due to the fact that during the course of their differentiation, T cells in NOD.Igμ^null mice fail to encounter B lymphocytes and thus are not rendered tolerant to them. It should also be noted that even following in vivo priming, NOD.Igμ^null mice failed to generate T cell responses to Ig isolated from standard NOD donors (data not shown). Thus, Ig does not appear to represent a B lymphocyte protein to which NOD.Igμ^null T cells fail to establish tolerance. This was not unexpected, since NOD.Igμ^null T cells reject transplanted NOD B lymphocytes in an MHC class I-restricted fashion, while any T cell response engendered against soluble Ig would most likely be MHC class II restricted.

We hypothesized that T cells from NOD.Igμ^null mice could be rendered tolerant to B lymphocytes if forced to mature in their presence. If correct, such an approach would make it possible to determine whether restoring the presence of B lymphocytes reconstitutes an APC function that abrogates IDDM resistance in NOD.Igμ^null mice. To address these issues, NOD.Igμ^null mice were lethally irradiated to ablate pre-existing immunologic effectors and then reconstituted with purified NOD B lymphocytes admixed with NOD.Igμ^null bone marrow as a source of T cell precursors. Controls consisted of lethally irradiated NOD.Igμ^null mice reconstituted with syngeneic bone marrow only, as well as NOD-scid recipients repopulated with the purified B lymphocytes alone.

As expected, spleens from NOD.Igμ^null mice reconstituted with syngeneic bone marrow alone contained both CD4⁺ and CD8⁺ T cells but no B lymphocytes (Table II). IDDM developed in only 1/15 of NOD.Igμ^null mice reconstituted with syngeneic marrow alone. Insulitis development was also quite limited (MIS = 1.32 ± 0.33) in NOD.Igμ^null mice that remained free of overt IDDM following reconstitution with syngeneic bone marrow alone, indicating minimal activation of diabetogenic effectors. In contrast, permanent B lymphocyte as well as CD4⁺ and CD8⁺ T cell repopulation was detected in spleens of NOD.Igμ^null mice reconstituted with syngeneic marrow admixed with purified NOD B lymphocytes. This indicated that T cells derived from NOD.Igμ^null marrow are rendered tolerant to B lymphocytes when forced to mature in their presence. IDDM developed in 15/23 of NOD.Igμ^null mice characterized by chimeric restoration of both T and B lymphocyte populations. In addition, significant levels of insulitis were present (MIS = 2.06 ± 0.41) in the few T and B lymphocyte-repopulated NOD.Igμ^null mice that remained free of overt IDDM. IDDM failed to develop in any (0/5) NOD-scid females repopulated with the purified B lymphocytes alone. GAD-reactive T cell responses were also restored in NOD.Igμ^null mice that had been rendered diabetes susceptible following reconstitution with syngeneic bone marrow plus NOD B lymphocytes (Fig. 6). Such GAD-reactive T cell responses remained absent in diabetes-resistant NOD.Igμ^null mice that had been reconstituted with syngeneic marrow only. Thus, the presence of B lymphocytes appears to restore IDDM susceptibility to NOD.Igμ^null mice by reconstituting a normally absent APC function.

We and others have previously demonstrated that NOD mice made deficient in B lymphocytes by congenic transfer of an Igμ^null allele or treatment with a μ chain-specific Ab are rendered resistant to T cell-mediated autoimmune IDDM (7, 11, 12). Our NOD.Igμ^null congenic stock is fixed to homozygosity for linkage markers delineating all previously identified Idd loci of NOD origin. However, as suggested by another group (24), it was possible that the inhibition of IDDM in our NOD.Igμ^null congenic stock did not result from the absence of B lymphocytes, but rather to the presence of a previously unidentified Idd resistance gene that was inadvertently cotransferred from the original donor strain of the Igμ^null allele. The current study indicates that the presence of a previously unidentified Idd resistance allele cannot account for the inhibition of IDDM development in our NOD.Igμ^null stock, since their disease susceptibility is restored upon repopulation with NOD B lymphocytes.

Unlike reconstitution with B lymphocytes, infusion of Ig isolated from overtly diabetic NOD donors did not abrogate IDDM resistance or enhance insulitis development in NOD.Igμ^null mice. This suggested that autoantibody production does not represent the primary diabetogenic role of B lymphocytes in NOD mice. It should be noted that the reconstituted levels of circulating Ig achieved in NOD.Igμ^null mice (∼70 μg/ml) were approximately 10-fold less than that of standard NOD mice (∼700–1000 μg/ml). Thus, we cannot exclude the possibility that higher levels of Ig reconstitution might have a diabetogenic effect in NOD.Igμ^null mice. However, findings from a recent report suggest that if autoantibodies do play a primary diabetogenic role, the levels of Ig reconstitution we achieved in NOD.Igμ^null mice should have been sufficient to trigger an observable pathogenic effect (25). This study found that the Sjogren’s syndrome-like pathology that characterizes standard NOD mice does not develop in the same B lymphocyte-deficient NOD.Igμ^null stock used in this study. Interestingly, a Sjogren’s pathology was restored in NOD.Igμ^null mice that received a single 100 μg injection of Ig isolated from either standard NOD mice or human patients with this disease. Given these results, it seems likely that if autoantibodies also play a primary role in IDDM development, their pathogenic effect would have been observed in NOD.Igμ^null mice injected twice weekly from 8 to 20 wk of age with 70 μg of Ig isolated from standard NOD donors. Thus, while contributions from autoantibody production cannot be completely excluded, our data support the conclusion that the primary diabetogenic role for B lymphocytes in NOD mice is as a subpopulation of APC essential to the initiation of certain β cell-autoreactive T cell responses.

B lymphocytes represent an essential subpopulation of APC for generating MHC class II-restricted T cell responses to certain Ags (23). It has been reported that an MHC class II-restricted T cell response against the putative β cell Ag GAD may have a critical early role in triggering a cascade of other autoreactive T cell responses, which ultimately leads to the development of IDDM in NOD mice (18, 21, 26). Thus, GAD was used as a model to test the hypothesis that the diabetogenic role for B lymphocytes in NOD mice is as a subpopulation of APC with a preferential ability to process and present certain MHC class II-restricted β cell Ags to autoreactive T cells. Unlike standard NOD, B lymphocyte-deficient NOD.Igμ^null mice were found to be incapable of generating primed T cell responses against GAD. However, NOD.Igμ^null mice could still generate primed T cell responses against the control Ag KLH. Thus, B lymphocytes represent a critical subpopulation of APC for triggering certain, but not all, T cell responses in NOD mice. The fact that GAD-specific T cell responses are among those eliminated in the NOD.Igμ^null stock indicates that the diabetogenic role of B lymphocytes in standard NOD mice is as a subpopulation of APC with a preferential ability to initiate T cell reactivity against certain key β cell autoantigens. Support for this conclusion was provided by the finding that GAD-reactive T cell responses were restored in NOD.Igμ^null mice that had been rendered IDDM susceptible by B lymphocyte repopulation. The fact that they are among the earliest cells to infiltrate the pancreatic islets of NOD mice (27, 28) also supports our conclusion that B lymphocytes represent a critical subpopulation of APC for initiating the development of T cell-mediated autoimmune IDDM.

Interestingly, only the initiation of GAD-reactive T cell responses in NOD mice requires B lymphocytes as APC. Once such GAD-reactive T cells have been initially activated in NOD mice, our data indicate that their responses can be maintained by APC other than B lymphocytes. It is also possible that B lymphocytes are necessary APC only at the initiative phases of other MHC class II-restricted β cell-autoreactive T cell responses. If preactivated β cell-autoreactive T cells do not require B lymphocytes as APC, this could explain the finding that T cells from overtly diabetic NOD donors transferred disease to young prediabetic recipients depleted of B lymphocytes by treatment with a μ chain-specific Ab (29).

One factor that complicated our ability to assess the diabetogenic function of B lymphocytes was the rejection of these cells upon transfer into unmanipulated NOD.Igμ^null recipients. This rejection of B lymphocytes in NOD.Igμ^null recipients was mediated by an MHC class I-restricted cytotoxic T cell response. The presence of such effectors appears to result from the fact that during their differentiation, T cells in NOD.Igμ^null mice fail to encounter B lymphocytes and thus are not rendered tolerant to them. This was demonstrated by our finding that T cells differentiating from precursors in NOD.Igμ^null bone marrow were rendered tolerant to B lymphocytes when forced to mature in their presence. Full IDDM susceptibility was restored to NOD.Igμ^null mice that had been permanently repopulated with both T and B lymphocytes by this chimerization technique.

The fact that T cells from unmanipulated NOD.Igμ^null are not tolerant to B lymphocytes does not result from the generalized autoimmune proclivity of the NOD strain (30). We base this conclusion on the finding that T cells developing from NOD.Igμ^null marrow could be rendered tolerant to B lymphocytes when forced to mature in their presence. This indicates that the induction of tolerance to B lymphocytes is a normal feature in development of the T cell repertoire. The thymic site of T cell differentiation contains very few B lymphocytes. Thus, it seems most likely that T cells that are potentially autoreactive against syngeneic B lymphocytes are deleted or inactivated shortly after their emigration from the thymus. If this supposition is correct, it would also suggest that an immediate encounter with their cognate Ag results in tolerogenic responses, rather than the functional activation of T cells that have recently emigrated from the thymus. Conversely, it has been reported that the small numbers of B lymphocytes that do reside within the thymus may play an important role in the normal induction of T cell tolerance to endogenous Ags (31, 32). Thus, the normal induction of T cell tolerance to B lymphocytes could occur intrathymically. Interestingly, NOD mice are reported to be characterized by increased numbers of intrathymic B lymphocytes (33). However, regardless of whether NOD B lymphocytes induce tolerance to themselves intrathymically or in the periphery, they are clearly unable to block the development or function of pancreatic β cell-autoreactive T cells.

In conclusion, this study indicates that B lymphocytes play a diabetogenic role in NOD mice as APC essential to the initiation of effector T cell responses against certain pancreatic β cell autoantigens such as GAD. It remains to be determined whether the initiation of T cell responses to any candidate β cell autoantigens other than GAD also requires B lymphocytes as APC. However, regardless of their identity or number, our study indicates that any β cell Ags which initially require B lymphocytes as APC will be among the most pathogenically relevant targets of MHC class II-restricted diabetogenic T cell responses.

We thank Don S. Varnum and Ellis A. Johnson for their skilled technical assistance.