Purified Allogeneic Hematopoietic Stem Cell Transplantation Blocks Diabetes Pathogenesis in NOD Mice

Our NOD (H-2^g7, Thy-1.2) mouse colony was the gift of Y. Mullen (Los Angeles, CA) and M. Hattori (Boston, MA). AKR/J mice (H-2^k, Thy-1.1) and NOD.Thy-1.1 mice (NOD.NON-Thy-1^a/J) and SWR (H-2^q) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred in our facility. All mice were kept in filter-top cages in temperature-controlled, light-cycled rooms.

Purified HSCs were obtained by modification of the methods described by Spangrude et al. (10). BM cells were positively selected for c-Kit using the MACS system (Miltenyi Biotech, Auburn, CA). The c-Kit-enriched fraction was stained with fluorescein isothiocyanate (FITC)-conjugated Thy1.1 (19xE5), TX red (TR)-conjugated Sca-1 (E13–161), allophycocyanin (APC)-conjugated c-Kit (2B8), and a mixture of phycoerythrin (PE)-conjugated lineage-specific monoclonal antibodies (mAbs) as follows: B220 (6B2), CD3 (145–2C11), CD5 (53–7.8), CD4 (GK-1.5), CD8 (53–6.7), GR-1 (8C5), Mac-1 (M1/70), and TER119 (TER119). The conjugated mAbs were produced in our laboratory, except for CD3 mAb (145–2C11) and CD8 mAb (53–6.7), which were obtained from Pharmingen (San Diego, CA). Propidium iodide staining (1 mg/ml) was used to exclude dead cells. Cells were sorted on a dual laser FACS (Becton Dickinson, Mountain View, CA) made available through the FACS shared-user group at Stanford University. After sorting for FITC^lo, TR^hi, APC^hi, and PE^−/lo, the Thy-1^loLin^−/loSca⁺c-Kit⁺ cells were checked by FACS reanalysis and determined to be >99% pure. (14) BM from NOD.SCID mice (Thy1.2⁺) was harvested and processed as described above with the difference that CD38^bright instead of Thy1.1^lo expression was used for selection of HSCs (14).

Eight- to 12-week-old NOD mice were prepared for transplant with radiation alone or in combination with the rabbit anti-serum ASGM1 (α-ASGM1; WAKO Chemicals, Dallas, TX) plus α-CD4 mAb (produced from hybridoma GK-1.5). A Phillips Unit Irradiator (250 kv, 15 milliamp) delivered a total of 950 cGy (in two doses 3–4 h apart) on day 0 relative to transplant. The α-ASGM1 was administered at 50 μg intravenously on day −7 and by intraperitoneal injection on day −1. α-CD4 mAb (100 μg intravenously) was given on days −3, −2, and −1. HSCs or whole bone marrow were introduced intravenously by tail vein. Each HSC inoculum was tested by injection of 100 or 200 cells into lethally irradiated syngeneic hosts, which were killed for day +12 spleen colony formation unit assay.

Blood chimerism was assessed by FACS analysis at ∼6–8 weeks posttransplantation and subsequently 2 months later to assess graft stability. Donor versus host cells were differentiated by mAbs (Pharmingen) specific for the donor major histocompatibility complex (MHC) class I (H-2^k) or CD45 allele. NOD mice are CD45.1 (mAb A20), and AKR mice are CD45.2 (mAb 104). Double staining for lineage-specific markers included B-cells (B220, RA3–6B2), monocytes (Mac-1, M1/70.15), or granulocytes (Gr-1, 8C5). T-cell chimerism was assessed by Thy-1.1 (AKR) and Thy-1.2 (NOD; 53–2.1) allelic markers, or CD3 versus CD45.1 staining. FACS analysis was performed by a modified FACS II system equipped with logarithmic amplifiers. Data are presented as contour plots on log-10 scales of increasing intensity.

In selected mice, peripheral blood lymphocytes were analyzed by FACS for Vβ subsets. Biotin-conjugated Vβ3⁺ (KJ25), Vβ6⁺ (RR4–7), or Vβ8⁺ (KJ16) mAbs were used with second-step streptavidin-FITC labeling and then double-stained with PE-conjugated α-CD4 (H129–19) or α-CD3 (145–2C11). A mouse Vβ screening panel containing FITC-conjugated mAbs was also used to stain splenocytes from chimeric mice. All mAbs were obtained from Pharmingen except for α-Mac-1, α-H-2^b, and α-Vβ8⁺, which were obtained from Caltag (Burlingame, CA).

Selected HSC chimeric mice were killed 4 months posttransplantation, and their pancreata were snap-frozen in OCT compound (Tissue-Tek, Torrance, CA). Tissue sections (0.4 μm) were fixed in cold acetone, blocked, and stained with biotinylated mAbs against CD3, Thy 1.2, Thy-1.1 (Pharmingen), and insulin (Linco, St. Charles, MO) incubated with secondary antibody and streptavidin-horseradish peroxidase, visualized with 3-amino-9-ethylcarbazole, and counterstained with hematoxylin.

Diabetic NOD mice (blood glucose levels >450 mg/dl) received 4 units of insulin (Humulin Ultralente; Lilly, Indianapolis, IN) every other day until the time of islet transplant. Diabetic mice were prepared for combined HSC and islet transplant with 700 or 950 cGy plus the α-ASGM1 and α-CD4 mAbs as described above. On day 0, recipients received 10,000 AKR HSCs, and on day 1, islet transplants into the portal vein were performed as previously described (15). Briefly, islets were isolated by collagenase digestion (Collagenase P; Boehringer Mannheim, Mannheim, Germany) and Ficoll gradient separation. Islets (700–800) were handpicked under a dissecting microscope and infused into the livers of recipient mice by intraportal vein injection. Mice were anesthetized by intraperitoneal injection of pentobarbital 0.05 mg/g body wt. Blood glucose values were obtained every 2–3 days for the first week posttransplantation and weekly thereafter.

Time to diabetes onset among groups was compared using the log-rank test with the program GraphPad Prism version 2.01 (GraphPad Software, San Diego, CA). P < 0.05 was considered statistically significant.

Similar to our previous experience in nonautoimmune strains (9), transplants of AKR HSCs in NOD mice demonstrated high levels of engraftment resistance as compared with unfractionated BM. Death between days 10 and 30 posttransplantation or absence of long-term donor blood chimerism is indicative of hematopoietic graft failure. As shown in Fig. 1A, ∼50% of NOD mice died before day 30 when prepared with lethal radiation (950 cGy) and rescued with 10,000 AKR HSCs, and only one of six surviving mice had detectable donor chimerism. In contrast, >90% of NOD mice transplanted with 1–2 × 10⁷ AKR BM (an inoculum containing 5,000–10,000 HSCs, respectively) survived and demonstrated complete donor chimerism. Stable engraftment of purified AKR HSCs was achieved, however, by addition to the radiation regimen of antibodies directed against natural killer cell determinants (α-ASGM1) and CD4⁺ T-cells (CD4). Such treatment resulted in >90% survival, and phenotypic analysis of their blood revealed that 16 of the 18 surviving mice were chimeras.

Also similar to our previous observations in nonautoimmune strains (9), the pattern of donor chimerism differed in NOD mice engrafted with allogeneic HSCs versus BM. Figure 1B and C shows that AKR HSC-engrafted mice were 100% donor type in all white blood cell lineages except T-cells. Significant numbers of residual host T-cells were identified in the peripheral blood of these chimeric mice as compared with those engrafted AKR BM that demonstrated 100% donor T-cell chimerism (P < 0.004). This pattern of persistence of donor T-cells despite complete conversion of other immune cell subsets to donor type (9) is of particular significance in NOD mice because residual NOD T-cells may perpetuate an anti-islet response.

Recipient NOD mice were evaluated posttransplantation for evidence of hyperglycemia. Diabetic mice had blood glucose levels >500 mg/dl, and all nondiabetic mice had levels <125 mg/dl. In our NOD colony, the incidence of diabetes is roughly 90% in female mice and 84% in male mice by ∼6 months of age. For these studies, NOD mice were 8–12 weeks old and thus had an expected onset of hyperglycemia beginning at ∼3–4 months posttransplantation. Data in Fig. 2 demonstrate that diabetes development was blocked in all mice engrafted with AKR HSCs, because none of such treated mice developed hyperglycemia with a follow-up time of >6 months posttransplant. Diabetes development was also blocked in all mice engrafted with AKR BM (data not shown).

For differentiating between the effects of the preparative regimens versus the establishment of allogeneic hematopoietic chimerism on blocking diabetes development, groups of mice received NOD BM or purified NOD HSCs. Congenic NOD.Thy-1.1 mice served as donors for HSCs because currently available mAbs for the Thy-1.1 allelic marker (but not Thy-1.2) permit positive selection and isolation of mouse HSCs. Syngeneic wild-type NOD mice (Thy-1.2) served as BM donors. Figure 1A shows that, as expected, ≥90% of mice that received NOD BM or HSCs survived the HCT procedure. In the HSC group, donor chimerism was verified by Thy-1.1 staining, and similar to what was observed for the allografts, HSC recipient mice were partial T-cell chimeras (data not shown). Diabetes outcome for these groups is shown in Fig. 2. All NOD mice prepared with radiation plus antibody treatment and rescued with NOD BM developed hyperglycemia by ∼4 months posttransplantation. Similarly, 80% of mice prepared with radiation and rescued with purified NOD.Thy-1.1 HSCs developed hyperglycemia within 6 months posttransplantation. Of note, the age of diabetes onset in these syngeneic and congenic recipients was not significantly different compared with unmanipulated NOD mice (P > 0.42). These data demonstrate that neither the preparative regimen nor the transplantation of highly purified NOD HSCs blocked progression to diabetes.

We next addressed the issue of whether the NOD T-cells that persisted after the conditioning regimen were capable of causing islet destruction. To study this question, we used NOD.SCID mice as donors. Because HSCs from NOD.SCID mice cannot give rise to T- or B-cells, such grafts are incapable of generating pathogenic T-cells. Thus, diabetes could occur only if the residual host cells destroyed the islets. The NOD.SCID donors used in these experiments expressed the Thy-1.2 allele and not Thy-1.1. Therefore, purified HSCs were isolated on the basis of CD38 marker selection (14). This method of purification was tested in control studies wherein as few as 250 Lin^−/loSca-1⁺c-Kit⁺CD38⁺ selected NOD HSCs rescued lethally irradiated NOD mice (data not shown). Figure 3A demonstrates that NOD.SCID HSC engrafted mice still developed diabetes within 6 months posttransplantation despite very low numbers of endogenous T-cells. As compared with unmanipulated NOD mice, the age at which NOD.SCID HSC transplanted mice developed diabetes was delayed, but only by ∼2 months. NOD.SCID engrafted mice had persistently reduced absolute counts of CD3⁺ cells in their peripheral blood as compared with unmanipulated mice. Figure 3B shows that at 2 months posttransplantation, the CD3⁺ cells were significantly reduced (P < 0.001). Thus, even very low numbers of residual NOD T-cells are capable of mediating diabetes pathogenesis. Taken together with the allogeneic and congenic HCT data, these studies demonstrate that the allogeneic hematopoietic grafts alter the capacity of residual host cells to mediate autoimmune pathogenesis, whereas congenic hematopoietic grafts do not.

To further assess the effect of allogeneic HCT on disease pathogenesis, we killed a subset of AKR HSC chimeras for immunohistochemical analysis of their pancreata. Lymphocyte infiltration into islets (insulitis) is first detectable in NOD mice in our colony at ∼3–4 weeks of age. By 6 weeks of age, many islets show evidence of extensive insulitis (Fig. 4). Thus, at the time of HCT (8–12 weeks), the pathologic process was already well under way in recipient mice. Table 1 summarizes the histologic analysis of pancreata from AKR HSC chimeras at 4 months posttransplantation, and the results were compared with a pancreas from a 6-week-old control NOD. Islets were scored for severity of insulitis according to a grading system described by Wicker et al. (16). Figure 4 shows representative staining of the 6-week-old control NOD mouse versus an AKR HSC chimera. Islets of the control mouse demonstrated heavy T-cell infiltration as compared with the chimera, which had only mild peri-islet involvement and was without evidence of penetration into the islet parenchyma. Although this peri-islet infiltrate was positive for both donor and host Thy-1 alleles, the Thy-1.2 stain was primarily noncellular, whereas the Thy-1.1 marker correlated with the cellular staining observed with CD3. These data demonstrate that engraftment of MHC-mismatched allogeneic HSCs results in attenuation of the pathologic lesions.

NOD mice express a single MHC class II molecule (I-A^g7), which has been shown to bind peptide poorly, leading to the hypothesis that during development, autoreactive T-cells escape negative selection (17). AKR mice express two non-disease-associated MHC class II alleles (I-A^k and I-E^k). Thus, one way that AKR HSC grafts may alter host immune responses is by mediating T-cell-negative selection. To test this hypothesis, we made use of differences in superantigen-mediated deletion of informative Vβ subsets between the AKR and NOD strains (reviewed in 18). AKR mice delete Vβ6 and other Vβ families, whereas NOD mice delete Vβ3⁺ cells (19). Thus, as shown in Table 2, in control AKR mice, Vβ6⁺CD4⁺ cells constitute <0.1% of blood CD4⁺ cells, whereas Vβ3⁺CD4⁺ cells are well represented. In contrast, NOD mice have ∼8% circulating Vβ6⁺CD4⁺ cells but Vβ3⁺CD4⁺ cells comprise <0.5%. HSC chimeras were evaluated 8–10 weeks posttransplantation for Vβ subset analysis (Table 2). At that time, donor and host T-cells composed ∼70% and ∼30% of blood CD4⁺ cells, respectively. Donor T-cells were distinguished from host by staining for Thy-1.1 (AKR) and Thy-1.2 (NOD). Vβ8 served as the positive control and was present in chimeric mice at levels comparable to unmanipulated mice. In contrast, both Vβ6⁺CD4⁺ and Vβ3⁺CD4⁺ cells were virtually absent from the blood of HSC chimeras. The near complete absence of these subsets reflected deletion of both AKR- and NOD-derived cells. These findings suggest that, assuming residual host NOD cells were from a radiation-resistant post-thymic population, the elimination of these subsets demonstrates that the HSC graft can mediate deletion of both developing T-cells and mature peripheral T-cells.

It was possible that superantigen-mediated deletion was not the only cause for the low Vβ3⁺ and Vβ6⁺ levels observed in HSC chimeras. An alternative explanation was that HSC transplantation resulted in regeneration of a highly skewed or incomplete T-cell repertoire. Thus, using a Vβ staining panel, a more extensive analysis was performed on splenocytes from a subset of AKR HSC chimeras at ∼10 weeks posttransplantation. As shown in Fig. 5, very low levels of T-cells stained for Vβ3, -5, -6, -7, -9, -11, -12, and -17a. These low levels were consistent with β deletions previously reported for the AKR and NOD strains (18–20). Thus, reconstitution of the T-cell repertoire was not indiscriminately skewed, but rather repertoire regeneration was predictable based on the premise that both donor and host elements mediate negative selection of superantigen reactive Vβ subsets.

We recently showed that allogeneic HSC grafts can induce tolerance to donor-matched organs (21) in wild-type mice. Here we asked whether overtly diabetic NOD mice that had undergone autoimmune islet destruction could be cured by cotransplantation of HSCs plus donor-matched islets. In a pilot experiment, diabetic NOD mice (blood sugars >450 mg/dl) were prepared for transplant with lethal radiation plus the α-ASGM1 and α-CD4 antibodies. These mice received 10,000 AKR HSCs, and 1 day later, 600–800 AKR islets were infused into their livers via the portal vein. As shown in Fig. 6A, these animals demonstrated reversal of their diabetes and remained normoglycemic for an extended follow-up time of >140 days. Histology of the livers revealed the presence of intact islets within the liver sinusoids (data not shown). It should be noted that fully diabetic NOD mice are fragile, and many died as a result of the dual transplant procedure. Control diabetic NOD mice (n = 5) were treated with the antibodies but were not engrafted with HSC or islets. The latter regimen did not result in normoglycemia, and mice either died or were killed because of morbidity associated with diabetes.

In attempts to reduce the morbidity of the HSC plus islet transplant procedure, we tested a nonmyeloablative regimen using sublethal radiation (700 cGy). The transplant schema was otherwise identical to the lethal radiation regimen described above. Three of four diabetic recipients prepared in this way developed multilineage blood cell chimerism. Figure 6B is a representative FACS analysis from one of these recipients. The pattern of blood chimerism differed markedly from mice that received 950 cGy (Fig. 1B), because in addition to mixed T-cell chimerism, recipients were mixed chimeras in the other white cell lineages. Although all mice had initial reversal of their hyperglycemia, three of four mice rejected the AKR islets within 3 months. Two of these mice that reverted to hyperglycemia were documented multilineage chimeras. Control mice prepared with the nonmyeloablative regimen and transplanted with AKR HSC plus third-party SWR (H-2^q) islets rejected the islet grafts within 25 days.

Our studies show that engraftment of purified MHC-mismatched HSCs in prediabetic NOD mice uniformly blocks islet destruction despite active insulitis at the time of transplantation and persistence of host T-cells. By contrast, syngeneic or congenic recipients prepared with the same regimens were not protected from disease. Donor- and host-specific superantigens mediated T-cell deletion, and both developing and mature T-cells were effectively eliminated. In pilot studies, when overtly diabetic NOD mice were converted to near full AKR donor chimeras, they were tolerant to AKR islet grafts and were cured of their diabetes.

Consistent with our previous report in non-autoimmune-prone mice (9), we noted here in the AKR to NOD transplants that engraftment of HSCs differs significantly from BM. HSCs were more strongly resisted than BM, and BM transplantation resulted in full donor chimerism, whereas host T-cells persisted in mice that received HCT. One reason that BM engraftment is superior is explained in part by the observations that BM contains non-HSC populations that can facilitate HSC engraftment (9,12,22,23). We have characterized two distinct “facilitator” populations in mouse BM. Both express the CD8 molecule; one is a classical CD8⁺TCRα/β⁺ T-cell, and the other lacks expression of both CD3 and the α/β TCR (12). We observed that transplantation of HSC plus facilitator cells leads to full donor chimerism and thus hypothesized that one mechanism of facilitation is to eliminate radioresistant host cells that act as barriers to HSC engraftment.

An important difference between HSC transplants into NOD versus wild-type mice was the need to modify the preparative regimen. In nonautoimmune recipients, HSC engraftment can be achieved with lethal radiation plus α-ASGM1 only, whereas engraftment of AKR HSC into NOD mice required the addition of α-CD4 mAbs (9). NOD mice are known to have infiltration by autoreactive CD4⁺ T-cells in multiple organs, including islets (24,25). Furthermore, activated T-cells are highly radiation-resistant (26). Thus, the rationale for adding α-CD4⁺ mAbs to the NOD preparative regimen was to eliminate both autoreactive cells and radioresistant cells that could confer engraftment resistance.

It is interesting that although addition of α-CD4 mAbs to the preparative regimen permitted HSC engraftment, significant levels of residual NOD T-cells (a mixture of CD4⁺ and CD8⁺ cells) remained. These NOD T-cells were unlikely to have arisen de novo, because other host-derived blood lineages (i.e., granulocytes, monocytes) were not present, indicating the lack of active NOD-derived hematopoiesis. Theoretically, these persistent NOD cells could be autoreactive and perpetuate an anti-islet response. Thus, it was of particular interest that not only were the allogeneic HSC chimeras protected from progression to hyperglycemia but also histologic analysis of their pancreata revealed minimal islet lesions that were primarily confined to the perivascular/periductal areas. Furthermore, the majority of lymphocytes in these infiltrates were of donor, not host, origin. Studies from other laboratories have shown that allogeneic BMT in young NOD mice can prevent diabetes (1–3,27,28). In those reports, either donor/host chimerism was not evaluated extensively (1,3,27) or, in the studies in which chimerism was measured, the recipients were complete or near complete donor chimeras (2,28). In contrast to those reports, we show that complete replacement of the endogenous T-cell repertoire is not required to block diabetes pathogenesis.

Because replacement of the host immune system is not required to abrogate autoimmunity, how, then, does HCT mediate disease protection? We considered both the effect of the preparative regimen and the influence of the graft. Studies by van Bekkum et al. (29,30) and others (31,32) showed that for certain autoimmune diseases, preparation of rodents with lethal radiation plus syngeneic or “pseudoautologous” HCT conferred some disease protection, suggesting that the preparative regimen(s) can play a major curative role. In contrast, our data show that neither syngeneic BM nor congenic HCT prevented diabetes using regimens identical to allografted mice. It should be noted that these regimens resulted in dramatic reduction of endogenous T-cells. In light of the fact that human clinical trials are under way to treat severe autoimmune diseases with autologous HCT, it is important to use preclinical models to address certain fundamental and yet unanswered questions. One central issue is to differentiate whether regimen-resistant host immune cells versus the cells transferred within an autologous graft perpetuate autoimmune reactivity. To study this point, we used NOD.SCID mice as donors because NOD.SCID mice cannot generate T- or B-cells. Our data showed that myeloablated NOD.SCID HSC recipients develop hyperglycemia, indicating that regimen-resistant host cells can mediate islet destruction. Thus, we favor the conclusion that elements in the allograft play a major role in altering pathogenic responses.

Our studies here and elsewhere (21) on superantigen-mediated Vβ deletion showed that in HSC chimeras, both donor and host elements mediate nonrandom negative T-cell selection. AKR cells as well as residual NOD T-cells were deleted, suggesting that HSC-mediated negative selection occurred during both the intrathymic and the postthymic phases. In previous studies, we demonstrated that HSC grafts can also dictate positive T-cell selection (21). Because of the central role of MHC class II molecules in T-cell selection, the strong association of the NOD I-A^g7 with diabetes susceptibility, and the protective effects that alteration, replacement, or addition of a “nonsusceptible” class II molecule has on NOD disease, it is logical to conclude that the MHC class II⁺ cells are the critical protective elements that arise from an allogeneic HSC graft. Indeed, it has been shown that turnover of donor-type antigen-presenting cells (including dendritic cells) is relatively rapid post-HCT (33,34). We have noted in the thymus of long-term HSC chimeras that most class II⁺ cells were donor type by 16 weeks (21). To test further the importance of class II⁺ cells, we are now using C57BL/6.H-2^g7 mice as HSC donors to determine whether MHC-matched HSC grafts can confer similar disease protection. We must emphasize that our data showed that allogeneic HSC grafts caused superantigen-mediated negative selection of both donor and host T-cell populations. However, superantigen-mediated deletion is a surrogate assay, and we have not ruled out the possibility that HSC grafts have eliminated anti-islet-specific cells. Thus, we are currently studying HCT into T-cell receptor transgenic NOD mice (BDC2.5) that have high levels of anti-islet-specific T-cells that can be monitored posttransplantation. We are also examining whether other mechanisms, such as induction of regulatory cells or alterations in T-cell functional activity (i.e., induction of anergy), are responsible for altering the diabetes pathogenesis in HSC allografted mice.

Finally, to address the problem of overt diabetes, we tested whether HSC transplantation could induce immune tolerance to donor-matched islet grafts. The morbidity of the procedure in hyperglycemic NOD mice limited the experiments to pilot status. Nevertheless, in mice that tolerated the HCT and surgical intervention, long-term islet allograft survival was observed. We then tested a less morbid regimen that resulted in multilineage chimerism. Unfortunately, this approach did not allow long-term islet allograft survival. These latter data are to our knowledge the first demonstration that establishment of stable multilineage partial chimerism in NOD mice is insufficient to induce tolerance to donor-type islets. Previous reports using BM grafts to induce islet allograft tolerance occurred in recipients that were >95% donor type (35,36). The implications of our studies are that establishment of mixed hematopoietic chimerism with low-intensity regimens may not be sufficient to induce allograft tolerance in diabetic humans. We are currently exploring alternative nonmyeloablative regimens that may permit islet allograft survival. Furthermore, we are examining in partial NOD chimeras whether the lack of tolerance to allogeneic islets is specific to islets alone or extends to other organs.

In conclusion, we show for the first time that engraftment of purified allogeneic HSCs blocks an ongoing autoimmune response. Our data further suggest that allogeneic is superior to autologous HCT in disrupting autoimmune pathogenesis. We believe that transplantation of purified hematopoietic cell populations will significantly reduce the morbidity of clinical allogeneic HCT and make possible the use of this approach for the treatment of human autoimmune disease.

This work was supported by the Burroughs Wellcome Fund (J.A.S.), NIH RO1 AI49331-01, NIH PPG PO1 DK53005, a Stanford University Deans Fellowship (G.F.B.), and an Amgen Oncology Fellowship (G.F.B).

We thank T. Knaak for excellent assistance with FACS sorting and L. Hildalgo and B. Lavarro for breeding of mice.