BB Rat Thymocytes Cultured in the Presence of Islets Lose Their Ability to Transfer Autoimmune Diabetes

BBDR/Wor/Brm (RT1^u, ART2a⁺) and histocompatible athymic WAG-rnu/rnu (RT1^u) rats were purchased from Biomedical Research Models (Worcester, MA). All rats were certified by the vendor to be free of the Sendai virus, pneumonia virus of mice, sialodacryoadenitis virus, rat corona virus, Kilham rat virus, H1 (Toolan’s virus), GD7 (mouse poliovirus), Reo-3, Mycoplasma pulmonis, lymphocytic choriomeningitis virus, mouse adenovirus, Hantaan virus, Encephalitozöon cuniculi, and pinworm. Rats were housed in a viral antibody–free facility, and monthly testing of sentinel rats was used to assure the absence of infection in experimental animals. All rats were maintained in accordance with the guidelines of the Institutional Animal Care and Use Committee of the University of Massachusetts Medical School and recommendations in the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, National Research Council, National Academy of Sciences, 1996).

Pancreatic islets for use in co-culture with thymocytes were isolated by collagenase digestion, handpicked, and counted as described (12). Islets were usually added to ATOC either immediately (“fresh islets”) or after freezing at −80°C in the absence of cryopreservatives both to expose intracellular antigens and to destroy passenger antigen-presenting cells. In a preliminary experiment, the viability of a sample of frozen islets was tested by the method of trypan blue exclusion and was 0%. In one experiment, islets were exposed to streptozotocin (STZ) (“STZ islets”) to deplete insulin stores before addition to ATOC. STZ treatment of islets was conducted in vivo. Because of disruption of islet architecture, native islets could not be isolated from STZ-treated animals by conventional collagenase digestion. To obtain STZ islets, islets were isolated from normal BBDR rats and placed into semipermeable hollow fibers with a molecular pore size of 80 kDa (Amicon, Bedford, MA). Islets were introduced through a 25-gauge needle into the fibers, which were then heat-sealed and implanted into the peritoneal cavity of athymic WAG-rnu/rnu rats. Recipients were then treated with STZ (85 mg/kg; Upjohn, Kalamazoo, MI) in phosphate citrate buffer, pH 4.7. Successful STZ treatment was defined as a plasma glucose concentration >17.8 mmol on 2 successive days (Beckman Glucose Analyzer II; Beckman, Fullerton, CA). Cells were recovered by flushing fibers three times with 10 ml complete ATOC medium (see below). Aliquots of these cells were either used immediately in ATOC or stored at −80°C until assayed for insulin content using a commercial enzyme-linked immunosorbent assay (ELISA) (ALPCO, Windham, NH). The amount of insulin present in STZ islets used in co-culture ranged from 3.9 to 4.7 μg per thymus. For comparison, the amount of insulin present in an identical number of otherwise untreated islets harvested after 3 days of culture ranged from 42.8 to 51.3 μg.

Thyroid tissue was recovered aseptically from BBDR donors by dissection, rinsed with sterile HEPES-RPMI, and digested in 100 PZ units of Clostridium histolyticum collagenase (Serva Biochemicals, Heidelberg, Germany) and 7.5 mg trypsin (Sigma, St. Louis, MO) in 10 ml Hanks’ balanced salt solution supplemented with 0.1% fetal bovine serum. Thyroids were added in batches of 8–10 to the digestion solution and incubated for 20 min at 37°C and then triturated 10 times. The cycles of incubation and trituration were repeated a total of four times. The resulting cell suspension was filtered through a 70-μm nylon cell strainer to remove debris and washed three times in ATOC complete medium. The cells were counted, and cell pellets were stored at −80°C until use. The average yield of thyroid cells was 1.0 × 10⁶ per donor.

ATOCs were established from 8- to 10-week-old BBDR rat thymi as described (11). Briefly, thymi were cut into 1-mm³ fragments and cultured on filters placed on top of Gelfoam sponges (Upjohn, Kalamazoo, MI) in Dulbecco’s modified Eagle’s high-glucose medium supplemented with 20% fetal bovine serum (Hyclone, Logan, VT). Cultures were incubated at 37°C for 3–14 days in an atmosphere of 7% CO₂. Cultures maintained for longer than 7 days were fed by adding 5 ml media to the cultures on day 7. ATOCs were harvested using 10 ml collagenase solution per 6–8 ATOC filters (11).

In certain experiments, thymic fragments in the ATOC were co-cultured with syngeneic islets or thyrocytes from BBDR rat donors. Islet or thyroid cells were deposited directly on the surface of each individual thymus fragment. Cells were applied to each fragment twice at an interval of several minutes to optimize cell contact (0.25 μl per application). A total of 1,500–1,800 islets or 3 × 10⁶ thyrocytes per thymus were applied. This number of islets is comparable to the number of intrathymic islets reported to prevent spontaneous BBDP rat diabetes in vivo (10). Typically, to generate cells for adoptive transfer, three thymi would be processed at one time for each treatment group, and the cells generated at the end of the culture were pooled. No attempt was made to remove islets or thyrocytes from ATOC at the conclusion of culture.

Cells harvested from ATOC were washed three times in HEPES-buffered RPMI, counted, and transfused intravenously via the tail vein into WAG-rnu/rnu recipients in a final volume of 1 ml. The number of cells transferred to each recipient is given in the results for each experiment and in all cases represented the yield of 0.25–0.75 thymi. Recipient rats were injected intraperitoneally with anti-ART2a monoclonal antibody (mAb) (1 ml of 2× culture supernatant, five times weekly) and poly I:C. The rat anti-rat ART2a mAb was produced by culture of the DS4.23 hybridoma and was tested to document the absence of Mycoplasma contamination. Poly I:C was purchased from Sigma, dissolved in Dulbecco’s phosphate-buffered saline (1 mg/ml), sterile filtered, and stored at −20°C until used. The dose of poly I:C in early experiments (Fig. 1) was 0.5 mg/100 g given three times weekly, but it was subsequently learned that lower doses of poly I:C could be used. In later experiments (Fig. 2), the dose of poly I:C was 0.025 mg/100 g three times weekly.

Flow microfluorometry was used to quantify the expression of surface markers on cells freshly harvested from ATOC and on recipient spleen and lymph node cells as described (6). Antibodies to the αβ T-cell receptor (αβ TcR, clone R73), interleukin (IL)-2 receptor α-chain (IL-2R, clone OX-39), CD4 (clone OX-35), CD8α chain (clone OX-8), transferrin receptor (clone OX-26), CD134 (clone OX-40), CD45RC (clone OX22), appropriate isotype control antibodies (mouse IgG1, mouse IgG2a, and rat IgG2b), and R-phycoerythrin (PE)- or CyChrome-conjugated streptavidin were purchased from Pharmingen (San Diego, CA). Anti-ART2a (clone DS4.23) was purified using Protein G Plus agarose (Calbiochem, La Jolla, CA). Anti-ART2a was biotin-conjugated using BIOTIN-X-NHS (Calbiochem). Antibodies were used either directly conjugated with fluorochromes (fluorescein isothyacyanate [FITC], PE, or CyChrome) or were used as biotin conjugates followed by PE- or CyChrome streptavidin. Samples were fixed in a final concentration of 1% paraformaldehyde in phosphate-buffered saline and analyzed using a FACScan. A minimum of 5,000 viable cells in each sample was analyzed. The lymphocyte fraction was gated on the basis of forward and side scatter.

Pancreas and thyroid samples for histology were fixed in Bouin’s solution or in 10% buffered formalin, embedded in paraffin, sectioned, and stained with hematoxylin and eosin. Insulitis was scored as follows: 0, no infiltration; 1+, peri-islet infiltration only; 2+, infiltration of some but not all islets; and 3+, infiltration of most islets. Thyroiditis was scored as follows: 0, no infiltration; 1+, minimal focal infiltration; 2+, moderate infiltration; and 3+, widespread dense infiltration. Scoring was performed by a qualified pathologist (M.C.A.) unaware of the treatment status of the animals.

Elispot assays to enumerate the frequency of cells secreting γ-interferon (IFN-γ) and IL-4 were performed as described (13). For IL-4, the capture antibody was OX81 (anti–IL-4; Pharmingen), and the detecting antibody was biotin-conjugated B11–3 (anti–IL-4). For IFN-γ, the capture antibody was DB.1 (anti–IFN-γ), and the detection reagent was biotin-conjugated polyclonal rabbit anti–IFN-γ (Torrey Pines, La Jolla, CA). Antibodies were used at a concentration of 1 μg/well. Serial dilutions of cells were activated for 2 days in 96-well plates with concanavalin A (5 μg/ml) in HEPES-RPMI supplemented with 100 IU/ml penicillin, 0.1 mg/ml streptomycin, 2 mmol/l glutamine, 5 × 10⁻⁵ mol/l 2-mercaptoethanol (2-ME), and 10% fetal bovine serum (HyClone). Cells were then lysed with distilled water, incubated serially with the detecting antibody followed by streptavidin-alkaline phosphatase, overlaid with 0.6% agarose (Sigma), and incubated overnight at 37°C. The number of spots present in each well was counted using a Nikon E600 microscope, and results are expressed as the number of spots per 1 or 2 × 10⁵ cells.

The concentrations of IFN-γ and IL-4 present in ATOC supernatants were measured using commercial kits (BioSource International, Camarillo, CA). In some experiments, suspensions of spleen and lymph node cells recovered from WAG-rnu/rnu recipients of cells recovered from ATOC were activated in vitro with concanavalin A (5 μg/ml) for 40 h, and the concentrations of IFN-γ and IL-4 present in the culture supernatants were measured in the same way.

Average latency to onset of diabetes is presented as the median. Statistical comparisons of diabetes-free survival in adoptive recipients were performed using the method of Kaplan and Meier (14). The equality of nondiabetic survival distributions for animals in different treatment groups was tested using the log-rank statistic (15). Parametric data are shown as arithmetic means ± 1 SD. Groups of three or more means were compared using either univariate general linear model analysis or one-way analysis of variance and the Scheffe procedure for a posteriori contrasts (16). P values <0.05 were considered statistically significant.

Thymocytes from adult BB rats can adoptively transfer autoimmune diabetes to WAG-rnu/rnu recipients (6). It has been also reported that the development of BB rat T-cells is recapitulated in adult thymus organ cultures (11). Based on these observations, we first tested the hypothesis that cells capable of the adoptive transfer of diabetes would be present in long-term ATOC. As shown in Fig. 1A, we confirmed that transfer of fresh thymocytes present in one-fourth of an adult BBDR rat thymus induces diabetes in ∼80% of recipients with a median time to diabetes onset of 40 days. In addition, we observed that cells derived from the culture of one-fourth to one-half of an adult thymus are also capable of transferring diabetes, but the efficiency of diabetes transfer declined with extended culture (Fig. 1A). The reduced efficiency of transfer as a function of the duration of culture could be attributable to either the change in phenotype that occurs in ATOC over time or to the exponential decline in the number of viable cells present in ATOC (11). To investigate the effect of cell dose independent of changes in phenotype, we compared the ability of cells derived from the culture of one-fourth and three-fourths of an adult thymus to transfer diabetes. As shown in Fig. 1B, cells derived from one-fourth thymus cultured for 13–14 days were unable to transfer diabetes, whereas cells derived from three-fourths thymus cultured for 14 days transferred diabetes to ∼70% of recipients with a median latency to onset of 72 days.

As expected (11), cells recovered from BBDR rat ATOC exhibited a mature phenotype by day 6 and maintained that phenotype thereafter. The phenotype of cells recovered from a representative 14-day culture was as follows: CD4⁺, 53%; CD8⁺, 31%; CD4⁺ CD8⁺, 11%; αβTCR^hi, 95%; and TCR⁺ IL-2Rα⁺, 24%. In addition, 38% expressed de novo the ART2a⁺ rat maturational T-cell alloantigen, which is normally absent in the thymus. The CD4:CD8 ratio of cells recovered after 6 or more days of culture approximated the 2:1 ratio characteristic of rat peripheral lymphoid tissues.

Having determined that cells capable of the adoptive transfer of diabetes are present in cultured BBDR rat thymi, we next tested the hypothesis that exposure of cultured thymi to an appropriate antigen would reduce the diabetogenicity of recovered cell populations. We used ATOC carried out for 4–7 days based on our observation (Fig. 1) that cells cultured for 3–7 days transfer diabetes to about half of recipients.

As shown in Fig. 2A, we first observed that culture of BBDR rat thymocytes in the presence of frozen thyrocytes had no effect on the ability of the cultured cells to transfer diabetes. About two-thirds of recipients given thymocytes cultured in the presence or absence of thyroid tissue became diabetic with a median latency to disease onset of 42 days in both cases. In contrast, the frequency of diabetes in recipients of thymocytes cultured for 4–5 days in the presence of frozen islet tissue was dramatically reduced. Only 1 of 17 recipients of such cells developed disease.

In a second experiment, we repeated this analysis using cells cultured for a slightly longer period of time (7 days). As in the previous experiment, the starting inoculum in the ATOC was one-half thymus, but because of the increased duration of culture, the actual number of cells surviving to the time of transfer was smaller. We also compared the efficacy of frozen, fresh, and STZ-exposed islets to alter the diabetogenicity of cultured BBDR rat thymocytes and performed histological analyses. As shown in Fig. 2B, we confirmed that the addition of thyrocytes to BBDR rat ATOC does not affect the ability of these cells to transfer diabetes. We also observed that the addition of any of the three forms of islet tissue to the cultures was associated with a reduction in the ability of cultured thymocytes to transfer diabetes. The three islet–co-culture groups were statistically similar to each other, and when considered as a group, they were significantly different from both the control and thyrocyte co-culture groups (P < 0.001 in both cases).

Histological analyses were performed on the pancreata of nondiabetic recipient rats at the conclusion of the experiment. The frequency of insulitis of any intensity (1+, 2+, or 3+) in nondiabetic recipients of thymocytes cultured in the presence of fresh (7 of 11), frozen (2 of 5), and STZ-treated (4 of 5) islets was statistically similar. Overall, some degree of insulitis was present in 62% of these 21 pancreata. In the case of pancreata from nondiabetic recipients of cells from control thymocyte cultures, four of five (80%) displayed insulitis. Very few recipients of cells from cultures performed in the presence of thyrocytes remained nondiabetic; the two pancreata that were available from such animals did not show insulitis. The average insulitis score for all animals in the experiment was 1.8 ± 0.9, and there were no statistically significant differences among groups with respect to the intensity of insulitis.

Histological analyses were also performed on 46 thyroid specimens obtained from both diabetic and nondiabetic recipients of cells recovered from ATOC. Overall, the frequency of thyroiditis was low (28%, 13 of 46). The frequency of thyroiditis in recipients of cells from control ATOC was 25% (3 of 12), and the mean intensity score was 0.6. The frequency of thyroiditis in recipients of cells from ATOC performed in the presence of any type of islet tissue was 33% (8 of 24), with a mean intensity score of 0.6. The frequency of thyroiditis in recipients of cells from ATOC performed in the presence of isolated thyrocytes was 20% (2 of 10), with a mean intensity score of 0.4. There were no statistically significant differences among these groups with respect to either the frequency or intensity of thyroiditis. Because nondiabetic recipients were on average observed for a longer period of time than diabetic recipients before recovery of thyroid specimens, this subgroup was reanalyzed separately, but again, there were no differences among groups with respect to the frequency or intensity of thyroiditis.

In an effort to determine the mechanism underlying these observations, we measured the in vitro effect of co-culture with frozen islets or thyrocytes on the phenotype of BBDR rat thymocytes and their production of IL-4 and IFN-γ after 5 days of ATOC. As shown in Table 1, there were no statistically significant differences in the percentage of αβTcR⁺ cells expressing both CD4 and CD8, CD4 alone, CD8 alone, ART2a, transferrin receptor, or the IL-2 receptor α-chain (IL-2Rα, CD25). We also assessed the production of IFN-γ and IL-4 by these same cells. After 5 days of culture, we measured the frequency of cytokine-secreting cells using a sensitive elispot assay. We found no significant differences in the frequencies of IL-4– or IFN-γ–producing cells between the three treatment groups.

We next analyzed WAG-rnu/rnu recipients of cultured BBDR rat thymocytes with respect to the in vivo phenotype of peripheral lymphoid cells and their production of IL-4 and IFN-γ. These animals had received thymocytes that had been cultured for 5 days in the presence or absence of either islets or thyrocytes. Lymph node and spleen cells were recovered either at the time of diabetes onset or at the conclusion of the experiment.

As shown in Table 2, there were no statistically significant differences in the percentage of CD4 single positive, CD8 single positive, or IL-2Rα⁺ (CD25⁺) cells present in nondiabetic recipients at the conclusion of the experiment. The percentage of CD4⁺CD8⁺ double positive cells in all recipients was comparably low in all groups, and the percentages of CD45RC⁺ and OX40⁺ spleen and lymph node cells were statistically similar in all treatment groups (data not shown).

We also assessed the production of IFN-γ and IL-4 cells obtained from both diabetic and nondiabetic recipients in two ways. We measured the production of IFN-γ in spleen and lymph node cells by ELISA. We measured the frequency of IFN-γ– and IL-4–secreting spleen cells using a sensitive elispot assay. As shown in Table 2, there were no detectable differences in cytokine production or in the frequency of cytokine-secreting cells, either as a function of the type of cells transfused into the recipients or the presence or absence of diabetes.

These data document that, like fresh BBDR rat thymocytes (6), cells recovered from BBDR rat ATOC retain the ability to adoptively transfer diabetes and thyroiditis to histocompatible athymic recipients. In one respect, the finding is not surprising because the phenotype of T-cells generated in BBDR rat ATOC is similar to that of mature BBDR rat peripheral T-cells (11). In both the native BBDR rat and the athymic WAG-rnu/rnu recipient, the presence of these T-cells leads to diabetes, provided that the immune system is activated (e.g., by poly I:C) and populations of regulatory T-cells expressing the ART2 antigen are absent (3). In another respect, however, the observation is surprising in that the number of viable cells present in adult rat ATOC declines exponentially (11), and after 14 days of culture, only 1–2% of the cells originally present are viable (11). Nonetheless, diabetogenic cells remain recoverable throughout this time period.

These characteristics of BBDR ATOC—a stable but small population of diabetogenic cells that survive “intrathymically” in culture—make the ATOC a potentially powerful tool for the identification of autoreactive cell populations and analysis of the mechanisms underlying their development. In particular, over the interval between 3 and 7 days of culture, diabetogenic cells survived, were readily recoverable, and actually appeared to comprise a progressively larger percentage of surviving cells. As a first step toward exploiting this tool, we demonstrated that the diabetogenic potential of cells present in BBDR rat ATOC can be modified in vitro. The data show clearly that co-culture of thymic fragments in the presence of syngeneic islets, but not thyrocytes, almost completely prevents the adoptive transfer of diabetes. The observations are consistent with the hypothesis that thymocytes with diabetogenic potential differentiate in culture and that the presence of an appropriate antigen can modify the course of this process and reduce diabetogenicity.

A large number of islet constituents have been proposed as the candidate autoantigen responsible for the tissue-specific autoimmune process that leads to type 1 diabetes (1). Our observations suggest that defective intrathymic processing of one or more islet components may contribute to the development of autoimmunity (17). They are also consistent with previous observations that the prophylactic implantation of islet allografts intrathymically reduces the frequency of subsequent diabetes in young BBDP rats (7,8) and NOD mice (9).

We observed that fresh, frozen, and STZ-exposed islets were comparably effective in reducing the diabetogenicity of thymocytes present in BBDR rat ATOC. The effectiveness of fresh islets is consistent with a role for both insulin and islet surface antigens in this process. Fresh islets also supplied “passenger” antigen-presenting cells (APCs) to the co-cultures, and it could be argued that the presence of these cells was also important. The data generated using a frozen islet preparation devoid of viable cells argue against a role for metabolically active APCs in this process, but the possibility of indirect presentation of islet antigens by thymic APCs has not been excluded.

Our experiments using STZ-exposed islets were intended to investigate further the role of one of the principal candidate autoantigens, insulin. Extensive literature suggests that, despite its presence in the circulation and expression in the thymus, defects in expression, recognition, processing, or affinity intrathymically may, in susceptible animals, lead to loss of tolerance to insulin and the β-cells that produce it (1,18,19,20,21,22,23). Analyses of anti-insulin autoantibodies, insulin-reactive T-cells, and intrathymic insulin expression have implicated this protein as a potentially important autoantigen in both NOD mice and humans with type 1 diabetes (1,18,19,20,21,22,23,24,25,26,27). The potential importance of insulin as an autoantigen has also received support from studies demonstrating that insulin administration can prevent diabetes in NOD mice and possibly in humans (1,24,25). In our experiments, exposure to STZ reduced the insulin content of the islets used in co-culture by >90%. Nonetheless, STZ-exposed islets retained the ability to modify the cells generated in ATOC and to reduce their diabetogenicity. Because of the small residual amount of insulin remaining in these co-cultures, we cannot definitively exclude a role for insulin in reducing thymocyte diabetogenicity. The presence of even a small amount of insulin could in theory modulate the diabetogenicity of T-cells. We would point out, however, that other data do suggest that insulin may not be as strong a candidate autoantigen in the BB rat as in NOD mice or humans. It appears that parenteral insulin prevents diabetes in BBDR rats only at doses that produce persistent hypoglycemia and a state of “β-cell rest,” and even in protected animals, autoreactive T-cells are still present (28). Similarly, neither oral insulin (29) nor intrathymic insulin B-chain peptide 9-23 (30) prevents diabetes in BB rats. In contrast, the intrathymic injection of 200 μg insulin β-chain ameliorates (31), and the systemic (including intrathymic) expression of proinsulin II (19) completely prevents, the expression of diabetes in the NOD mouse.

It has been reported that the in vivo implantation of intrathymic islets into young BBDR rats does not preclude the subsequent induction of diabetes in these animals by poly I:C and anti-ART2 mAb (10). This in vivo observation would seem to predict that the in vitro intrathymic modification of diabetogenicity should not be possible. The reason for the apparent discrepancy between the two data sets is not clear. Speculatively, it could be argued that the presentation of antigen in ATOC, dispersed widely across thymic fragments, might have been more efficient than that achieved by the method of bilateral injection into thymic lobes in vivo. Alternatively, it can be pointed out that the in vivo study was not performed in a viral antibody–free environment (10). Viral infection—e.g., with Kilham rat virus (3)—could have prevented or circumvented the effect of the intrathymic islet antigen in the face of intercurrent regulatory cell depletion.

In the present study, we observed that co-culture in the presence of islets markedly reduced the ability of cultured BBDR thymocytes to transfer diabetes but eliminated insulitis in only ∼40% of the animals that remained normoglycemic. Why ∼60% of animals had nonprogressive insulitis cannot be determined from the available data, but the simplest explanation is that, because of the geometry of layering islets onto thymic fragments, a limited number of thymocytes in the ATOC was not adequately exposed to islet antigen. Speculatively, this limited cell population retained the ability to recognize and home to islets but was incapable of, or prevented from, completing the process of islet destruction. We have previously documented in studies of BB rat T-cells that the number of cells injected is a critical determinant of diabetogenicity (32). Alternatively, co-culture with islets may have affected the generation of a population of cells required to transform nonprogressive (or benign) insulitis to destructive (or malignant) insulitis (33). In the BDC2.5 TCR transgenic mouse, it was clearly shown that signals provided by IL-18, IL-12, and tumor necrosis factor-α are required for the development of diabetes in animals with otherwise nonprogressive insulitis (34).

We performed several analyses to attempt to define the mechanism by which islet co-culture altered the behavior of BBDR rat ATOC, examining both the cells generated in culture and the cells recovered from diabetic and nondiabetic recipients. One candidate mechanism studied was immune deviation, but we observed no differences in phenotype or the production of IFN-γ or IL-4 that were consistent with this mechanism. Other potential mechanisms include deletion and the generation of anergy. There was no detectable difference in the number of cells generated by BBDR rat ATOC in the presence or absence of islets, but the fact that islet co-culture prevented diabetes but not insulitis is compatible with either a partial reduction in the number of diabetogenic cells or the induction of anergy in some but not all of those cells. A final possible mechanism not addressed in the present study is positive selection of islet-reactive ART2^– regulatory cells. Such cells could in theory home to the pancreas and, in the form of “protective insulitis,” prevent diabetes in the adoptive recipients of cells from ATOC.

The data on lymphocytic thyroiditis generated in these experiments are intriguing but inconclusive. Co-culture of BBDR thymocytes with thyroid cells did not reduce the frequency of thyroiditis in adoptive recipients, but overall the frequency of thyroiditis was low (∼30%). Because lymphocytic thyroiditis in BB rats rarely, if ever, progresses to overt hypothyroidism (2,3), it is impossible to know from these limited data if anti-thyroid autoreactivity was modified.

In summary, we have demonstrated that it is possible to modify in vitro the generation of diabetogenic thymocytes that are genetically capable of inducing autoimmune diabetes. Given the expanding literature documenting the role of intrathymic self-antigen expression in the generation of self-tolerance (35,36), we propose that defective or deficient intrathymic expression of a relevant autoantigen present in pancreatic islets in BBDR rats is crucial to the development of autoimmune diabetes.

This study was supported in part by Center Grant DK32520; grants DK49106, DK25306, and DK36024 from the National Institutes of Health; and a Career Development Award from the Juvenile Diabetes Foundation International (B.J.W.).

We thank Kelly Lake and Michael Bates for technical assistance and Dr. Eugene Handler for reading the manuscript.