Grafts of supplementary thymuses injected with allogeneic pancreatic islets protect nonobese diabetic mice against diabetes

Abstract

In nonobese diabetic (NOD) mice, the autoimmune attack of the β-cells in pancreatic islets is now believed to result from abnormal thymic selection. Accordingly, grafts of thymic epithelium from NOD donors to athymic recipients promote autoimmune islet inflammation in normal strains, and intrathymic islet grafts decrease the incidence of disease in NOD animals. Two competing hypotheses of abnormal thymic selection in diabetic mice have been proposed: deficient negative selection with poor elimination of aggressive organ-specific T cells vs. deficient positive selection of protective T regulatory cells. We have now addressed these alternatives by grafting, into young NOD mice whose own thymus was left intact, newborn NOD thymuses containing allogeneic pancreatic islets. If the NOD defect represented poor negative selection, these animals would develop disease at control rates, as the generation of autoreactive T cells proceeds undisturbed in the autologous thymus. In contrast, if NOD thymuses are defective in the production of T regulatory cells, lower disease incidence is expected in the chimeras, as more protective cells can be produced in the grafted thymus. The results show a reduced incidence of diabetes in the chimeras (24%) as compared with control (72%) NOD mice, throughout adult life. We conclude that amelioration of NOD mice by intrathymic islet grafts is not caused by enhanced negative selection and suggest that autoimmune diabetes in this system is the result of inefficient generation of T regulatory cells in the thymus.

The nonobese diabetic (NOD) mouse is an animal model for human insulin-dependent diabetes mellitus (IDDM) that spontaneously develops a T cell-mediated autoimmune disease characterized by massive infiltrations of pancreatic islets and several other organs (namely the submandibular salivary and thyroid glands) by CD4⁺ and CD8⁺ T cells (1–7). Genetic regions have been identified that are related to the onset of autoimmune insulitis in these mice, where full development of diabetes requires expression of the unique I-A^NOD gene at the MHC (1, 8–11).

Two lines of research have been pursued to disclose the mechanisms responsible for autoimmunity in NOD mice. One has concentrated on the autoantigens expressed by the peripheral targets of autoimmunity and has shown that various β cell-specific antigens are critical in this process (12–14). The alternative approach ascribes an essential role in the onset of the disease to a primary dysfunction of the immune system, particularly to abnormal T cell selection in the thymus. IDDM can be prevented in NOD mice by neonatal thymectomy (15) or by administration of anti-IA mAbs (16), and diabetes can be adoptively transferred to healthy NOD neonates (6) or to (irradiated) adult male NOD recipients (17, 18) by spleen cells from diabetic mice. Neonatal injection of semiallogeneic spleen cells into NOD mice results in protection against insulitis and diabetes, demonstrating that manipulation of the T cell repertoire at birth can modulate the autoimmune process (19). Critical evidence for abnormal thymic selection was reported by Thomas-Vaslin et al. (20) who showed that thymic epithelium (TE), removed from NOD embryos at embryonic day 10 (E10) and grafted into C57BL/6 nude mice, was able to induce insulitis and syalitis in the recipient. These experiments demonstrated that, regardless of putative abnormalities in peripheral tissue antigen expression and or presentation, T cell repertoire selection by the NOD TE is responsible for the induction of autoimmune inflammation in an otherwise normal host. This conclusion is in line with our previous findings using the TE/chimera model in birds and mammals (21–29), strongly suggesting that self-tolerance is established and maintained by the thymic generation of a subset of T cells that have the capacity to negatively regulate the activation of autoreactive T lymphocytes in the periphery. This notion of “dominant” tolerance (27, 30) has been documented in a number of independent experimental systems, notably in the NOD model. Indeed, the adoptive transfer of IDDM to healthy recipients by spleen cells from diabetic NOD mice requires sublethal irradiation of the recipients (17), acute onset of diabetes is induced in young (healthy) NOD mice by thymectomy at weaning (31) or by cyclophosphamide treatment (32, 33), and CD4 T cells from young NOD donors delay disease onset in older animals (34).

Our studies (28, 29) and those of others (see ref. 35, for review) have demonstrated that normal individuals harbor two functionally distinct populations of CD4⁺ T cells, one capable of mediating an immune response against nonself as well as autoimmunity against self-antigens, and another that inhibits the former in a dominant and specific fashion. Regulatory T cells (Tregs) are now widely considered as “key controllers of immunologic self-tolerance” (36), such that deficits in their thymic production, as well as their deletion or inactivation, can break natural self-tolerance, thus leading to the onset of autoimmune diseases. If T cells are depleted of the Treg subpopulation and injected into syngeneic T cell-deficient or T cell-depleted mice, the recipients develop various organ-specific autoimmune diseases including IDDM, thyroiditis, gastritis, and systemic wasting diseases (ref. 36 and references therein). Tregs have been characterized at the molecular and functional levels; they represent some 10% of the mature thymic and peripheral CD4 T cell population and express CD25, RT6.1, low levels of CD45RB/RC, and high levels of CD5 (refs. 37 and 38 and references therein). In vivo and in vitro, Tregs are capable of inhibiting CD4 T cell responses to appropriate presentation of specific antigens and mitogens (28, 29, 39, 40).

The thymic origin of regulatory T cells was originally demonstrated by their functional defect in neonatally thymectomized mice (37, 41) and studied in the TE/chimera model in birds (21–23) and mammals (24, 25, 28, 29). Most recently, the presence of Tregs among mature thymocytes in normal animals was directly demonstrated (see ref. 42 for review). Previous data, showing that intrathymic islet transplantation ameliorates IDDM in biobreeding (BB) rats (43, 44) and in NOD mice (45–48), were essentially interpreted as the elimination of islet reactive effector T cell precursors by the negative selection induced by exposure of thymocytes to β cell antigens. Alternatively, the selection of regulatory T cells capable of suppressing anti-islet autoimmune responses could be the mechanism by which intrathymic islet injection protects from disease. Thus, we decided to investigate the alternative possibility that intrathymic grafts of antigen reinforce dominant tolerance by enhanced production of regulatory T cells.

In the experiments reported here, NOD mice were engrafted at 3 months of age with newborn syngeneic NOD thymuses into which 30–50 pancreatic islets had been injected. More than 70% of the nonengrafted controls developed IDDM from 3 months onward, in contrast with only 24% of the experimental animals. We propose that this result can be accounted for by the production in the grafted thymus of Tregs that are able to inhibit the autoreactive T cells produced both in this organ and in the endogenous thymus of the recipient.

Materials and Methods

Mice.

BALB/c (H-2d) and NOD (H-2 g7) mice, originally obtained from Iffa Credo, were bred in our animal facilities.

Experimental Protocol.

Thymuses of newborn NOD mice were dissected and injected with 30–50 pancreatic islets, isolated from 3-month-old BALB/c mice according to Lacy and Kostianovsky (49). Two injected thymic lobes were grafted into the mesentery of 4- to 16-week-old NOD female mice (Fig. 1).

Figure 1.

Thirty to 50 BALB/c pancreatic islets were inoculated with a micropipette into thymuses from newborn NOD mice. These thymuses were thereafter grafted in the mesentery of female NOD mice.

For histological analyses, mice were killed, the grafted thymuses and pancreas were removed, fixed in Bouins' solution, embedded in paraffin, and stained with hematoxylin/eosin, or processed for immunohistochemical staining for insulin with guinea pig Ab (Vector Laboratories). After further washing, avidin-biotin-peroxidase (Vectastain AB kit, Vector Laboratories) was added, followed by the substrate 3-amino-9-ethylcarbazole (Sigma).

Assessment of Diabetes Development.

Glycemia was scored by strip bands (BM-test glycemie, Roche Molecular Biochemicals) in blood obtained from the retroorbital sinus.

Results

Decreased Incidence of Diabetes in NOD Mice Grafted with NOD Thymus Containing Pancreatic Islets.

Pancreatic islets obtained from normal BALB/c mice were inoculated into the two thymic lobes obtained from newborn NOD mice. These were then grafted in the mesentery of 4- to 6-week-old NOD female mice. Of the NOD female mice, 63 were grafted while 98 female age-matched NOD mice served as controls. To control for the putative role of anesthesia and surgery in the onset of diabetes (50), another 29 NOD female mice were engrafted with newborn NOD thymuses devoid of pancreatic islets. These mice are designated as grafted controls.

The glycemia of ungrafted control, grafted control, and experimental mice was regularly followed for 12 months after the operation. Of the 98 ungrafted control NOD mice (72.44%), 71 became diabetic, mostly around 6 months. Of the 63 experimental NOD mice grafted with thymus plus pancreatic islets (23.80%), 15 developed the disease (Fig. 2). For the grafted controls, disease was scored in 20 of the 29 grafted animals (68.96%).

Figure 2.

Effect of the transplantation of a newborn NOD thymus with BALB/c pancreatic islets on diabetes incidence in female NOD mice. ▩, ungrafted control—unmanipulated NOD mice (n = 98); incidence of the disease at 12 months = 72.44%. ▨, grafted controls—NOD mice grafted with newborn thymuses only (n = 29); incidence of the disease at 12 months = 68.96%. ■, experimental mice—NOD mice grafted with newborn NOD mice thymuses containing BALB/c pancreatic islets (n = 63); incidence of the disease at 12 months = 23.80%.

Histological analysis of the pancreas of these diabetic grafted controls showed infiltrations and destruction of the islets, a similar picture to that found in control NOD mice (Fig. 3 A and B). Histological analysis of the grafted thymuses and pancreas of normoglycemic experimental NOD mice was carried out 10–12 months after surgery (n = 38). Sections of the pancreas were stained with either hematoxylin/eosin or treated for insulin detection. The islets were clean of mononuclear cell infiltration and in healthy condition, as further revealed by immunohistology for insulin-containing cells (Fig. 3 C, F, and G). The grafted thymuses from the same animals were removed from the mesentery and treated for insulin immunolabeling. Numerous insulin-positive cells (5–30 by section) were present, essentially in the thymic cortex (Fig. 3 D and E). In contrast with these findings, histological analysis of the pancreas of experimental NOD mice that developed diabetes revealed numerous infiltrates and destruction of the pancreatic islets. None or few insulin-positive cells were present in the grafted thymus of these experimental mice.

Figure 3.

(A–C) Histology of pancreas sections stained with hematoxylin/eosin. (A) Ungrafted control NOD (diabetic) mice and (B) grafted control (diabetic) mice; both show lymphocytes in infiltrated islets, in contrast with the healthy pancreatic islet from experimental mice, removed 12 months after grafting (C). [Bar = 30 μm (A and C) and 118 μm (B).] (D–G) Immunohistological analysis of the grafted thymus and the pancreas from an experimental NOD mouse. Sections of each organ were stained with antiinsulin Ab, revealing several insulin-containing cells in the grafted thymus (D, E), as well as healthy islets, strongly labeled for insulin in the pancreas (F, G). [Bar = 65 μm (D) and 10 μm (E, higher magnification of the boxed-in area in D). Bar = 65 μm (F) and 25 μm (G, higher magnification of the boxed-in area in F).]

Discussion

The notion that potentially aggressive autoreactive T cells are available in the normal repertoire, but controlled by T cell-mediated protective or “dominant” mechanisms, has emerged during the last decade from a number of experimental or pathological cases (21–25, 27–29, 35, 42, 51, 52). In our previous experiments, it was shown that chick embryos grafted with quail TE were tolerant to grafts of several quail tissues. This tolerance was not a consequence of clonal deletion in the quail thymus, because it was observed in animals where only one-third of the thymic mass contained quail TE, the rest being entirely of chick-recipient origin. Moreover, athymic mice grafted with allogeneic TE were similarly tolerant to donor-type grafts (25, 28, 29). We have interpreted these observations to indicate that TE generates regulatory T cells that are capable of maintaining dominant tolerance to tissues (grafts) of the same genotype, as shown in subsequent experiments.

The present experiments extend this work to the NOD diabetes model, relying on the repeated demonstration that the presence of pancreatic islet-specific antigens in the thymus reduces the onset of IDDM in genetically prone diabetic rodents (see ref. 53 for review). From these latter observations, however, it was unclear whether the presence of “pancreatic antigens” in the thymus simply enhanced removal of autoreactive T cells by negative selection or whether it induced higher levels of regulatory T cell activity. In the model described here, the endogenous thymus of the NOD mouse recipient was left intact whereas a thymus from a newborn NOD individual was engrafted with pancreatic islets and ectopically placed in its peritoneal cavity. This experimental model, therefore, allows to test whether or not the protection afforded by “thymic antigens” is recessive or dominant. Under these conditions, the unmanipulated (autologous) thymus is expected to continue to produce autoreactive T cells as in controls, such that a putative effect of pancreatic islets in the grafted thymus can be explained only by the production of more regulatory T cells.

Our results (concerning the follow-up of experimental and control mice for more than 12 months) reveal a strong protective effect of the ectopic thymus containing pancreatic islets (but not of ectopic thymuses containing no islets). Because this enhanced state of tolerance is “dominant” and prevails in the presence of the autologous (pathogenic) NOD thymus, it is very likely that it is mediated by regulatory T cells.

Several lines of evidence point to the presence of regulatory T cells in the NOD mice (32–34, 54–56) and to their rapid decline with age (57), in a manner related to disease onset (see ref. 58 for review). Because, as we have previously demonstrated, NOD TE suffices to induce islet inflammation if used to reconstitute athymic mice of non-disease-prone strains, it would seem that NOD thymus produces deficient numbers and/or functionally deficient Tregs.

The detailed mechanisms of enhanced Treg generation in thymus containing pancreatic islets remain to be elucidated. Others have previously shown in NOD mice that enhanced tolerance induced by intrathymic islets is antigen-specific (45, 47). We have proposed that CD4 T cells with high avidity for antigens presented by TE (but not on hemopoietic-presenting cells) would be positively selected and activated for regulatory effector functions (28, 29). This suggestion has recently received direct confirmation in a transgenic mouse model in which self-peptide-specific CD4⁺ T cells are regulated by CD4⁺ CD25⁺ regulatory T cells. The authors have shown that interactions with a single self-peptide can induce thymocytes that bear an autoreactive T cell receptor to undergo selection to become CD4⁺ CD25⁺ regulatory T cells (59). Such interaction could explain our present findings. It is interesting to note that, as others have previously shown (53), allogeneic islets are not rejected if grafted intrathymically, and, as shown here, obviously operate in the selection of Tregs. Although a putative MHC restriction of Tregs remains to be clearly established, the competence of allogeneic islets is likely the result of presentation of islet antigens by TE cells, as also suggested by similar results obtained by intrathymic injection of protein antigens instead of islet cells. Experiments are in progress to explore this problem.

The mechanisms mediating the protective activity of Tregs in the periphery also demand further experimentation (60). These cells are likely to operate within the pancreas or in the draining lymph nodes to counteract β cell-specific effector T helper 1 cells that are responsible for initiating the destruction of the islets in IDDM (61–63).

Abbreviations

NOD

nonobese diabetic

TE

thymic epithelium

IDDM

insulin-dependent diabetes mellitus

Tregs

regulatory T cells