Cellular Expression Requirements for Inhibition of Type 1 Diabetes by a Dominantly Protective Major Histocompatibility Complex Haplotype

NOD/LtDvs mice are maintained at The Jackson Laboratory by brother-sister mating. NOD mice made B-cell deficient by a functionally disrupted Igμ allele (designated NOD.Igμ^null) are maintained at the N10 backcross generation and have been previously described (23). Type 1 diabetes–resistant NOD mice congenic for the NON strain–derived H2^nb1 haplotype are maintained at the N21 backcross generation and have also been described (9). These latter two stocks were crossed to generate a B-cell–deficient NOD stock also homozygous for the H2^nb1 MHC haplotype (designated NOD.H2^nb1.Igμ^null). Development of a NOD stock transgenically expressing the rearranged T-cell receptor (TCR) α (Vα8) and β (Vβ2) chain genes from the β-cell–autoreactive CD8 T-cell clone AI4 (designated NOD.AI4) has also been previously described (24).

Female (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 mice at 4–6 weeks of age were lethally irradiated (1300R from a ¹³⁷Cs source) and reconstituted as previously described (10) with 5 × 10⁶ of the indicated T-cell–depleted bone marrow cells. In some experiments, 5 × 10⁶ T-cell–depleted bone marrow cells were admixed with 5 × 10⁶ purified B-cells as indicated. Splenic B-cells were purified using the previously described magnetic bead system (25). The purity of B-cells was >92%, as determined by flow cytometry described below.

Type 1 diabetes was assessed by weekly monitoring of glycosuric levels with Ames Diastrix (Bayer, Elkhart, NJ), with disease onset defined by values ≥3.

The indicated bone marrow–chimeric mice remaining diabetes free for 26 weeks postreconstitution were assessed for insulitis development. Mean insulitis scores were determined as previously described (26). Briefly, pancreata 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 20 per 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.

Flow cytometry was used to determine the levels of B-cell reconstitution, bone marrow chimerism, and AI4 T-cells in recipient mice. B-cells were identified by costaining with the B220-specific antibody (clone RA3-6B2) and polyclonal anti-mouse immunoglobulin. Origins of dendritic cells were detected with antibodies specific for CD11c (clone HL3), the K^d (NOD type) MHC class I variant (clone SF1-1.1), and the K^b (NOD.H2^nb1 type) MHC class I variant (clone 28-13-3). To determine the frequency of AI4 T-cells, thymocytes and splenocytes were stained with antibodies specific for CD8 (clone 53-6.7), CD4 (clone GK1.5), and the TCR Vα8 component (clone B21.14).

Normally resistant (NOD × NOD.H2^nb1)F1 mice develop type 1 diabetes if irradiated and reconstituted with NOD bone marrow (10). However, when reconstituted with a 1:1 mixture of NOD and NOD.H2^nb1 bone marrow, stable mixed chimerism results, and such F1 hybrids are now protected from type 1 diabetes (10). Subsequent studies indicated that H2^nb1-expressing APCs mediated the deletion or permanent inactivation of autoreactive diabetogenic T-cells that normally develop from precursors in NOD bone marrow (27). An unresolved issue was if H2^nb1 molecules needed to be expressed on all APC subtypes (dendritic cells, macrophages, and B-cells) to mediate the deletion or permanent inactivation of a sufficient array of β-cell–autoreactive T-cell clonotypes to prevent type 1 diabetes development. To do this, we adapted a previously described bone marrow–chimeric system that allows T-cells to mature and function in an environment where expression of H2^nb1 is restricted to dendritic cells/macrophages or B-cells (26).

We first generated (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 mice and tested if type 1 diabetes would develop when reconstituted with NOD.Igμ^null bone marrow and purified NOD B-cells. In this case, all APCs express only the type 1 diabetes–susceptible H2^g7 haplotype, but nonhematopoietically derived cells heterozygously express both H2^g7 and H2^nb1 molecules. As shown in Table 1, >60% of the recipients became diabetic by 26 weeks post–bone marrow reconstitution. These results confirmed the previous observation (10) that expression of H2^nb1 MHC molecules on nonhematopoietic cells is not sufficient to block type 1 diabetes development.

We then tested if H2^nb1 expression solely on dendritic cells/macrophages is capable of suppressing type 1 diabetes. This was done by reconstituting lethally irradiated (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 females with NOD.H2^nb1.Igμ^null bone marrow admixed with purified NOD B-cells. As summarized in Table 1, none of these recipients became diabetic by 26 weeks postengraftment. It should be noted that in these bone marrow chimeras, the positive selection of T-cells could be mediated by either H2^g7 and H2^nb1 MHC molecules (class I: K^d, K^b, or D^b; class II: A^g7, A^nb1, or E^k) expressed on non–hematopoietically derived thymic epithelial cells. However, with the exception of D^b, all other H2^nb1- or H2^g7-specific MHC class I/II molecules were solely expressed by either the dendritic cell/macrophage or B-cell subsets of hematopoietically derived APCs, respectively, available to activate T-cell effector function in the periphery. Therefore, complete type 1 diabetes resistance in these particular chimeras could be due to the lack of K^d- and/or A^g7-expressing dendritic cells/macrophages, which may be required to fully activate β-cell–autoreactive T-cells restricted to these H2^g7-encoded MHC molecules. To rule out this possibility, we reconstituted lethally irradiated (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 recipient mice with syngeneic bone marrow admixed with purified NOD B-cells. In this case, the same H2^g7 MHC molecules available to positively select T-cells in the thymus are also coexpressed with H2^nb1-encoded variants on all dendritic cells/macrophages in the periphery. As shown in Table 1, such recipient mice were still highly protected from type 1 diabetes. These results indicated that heterozygous expression of H2^g7 MHC molecules on dendritic cells/macrophages was not sufficient to break type 1 diabetes resistance conferred by H2^nb1-encoded gene products on the same cells. In addition to protection from overt diabetes, insulitis was also significantly reduced by the presence of H2^nb1-expressing dendritic cells/macrophages (Fig. 1). Therefore, heterozygous expression of H2^nb1-encoded genes solely by the dendritic cell/macrophage lineages of APCs is sufficient to strongly inhibit type 1 diabetes development in NOD background mouse stocks.

We next tested if H2^nb1 expression exclusively by all B-cells was sufficient to suppress type 1 diabetes. This was done by reconstituting lethally irradiated (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 females with NOD.Igμ^null bone marrow admixed with purified NOD.H2^nb1 B-cells. Expression of H2^nb1 MHC molecules solely on B-cells was also capable of retarding type 1 diabetes development and reducing insulitis. However, B-cells did so less efficiently than dendritic cells/macrophages (Table 1 and Fig. 1). It should be noted that the extent to which H2^nb1-expressing B-cells were able to block type 1 diabetes development in the F1 recipients did not correlate with their level of engraftment. This is illustrated by the fact that the proportion of H2^nb1-expressing B-cells among splenic leukocytes in the 2/10 F1 recipients that did develop type 1 diabetes was 5.8 and 18.6%, while the overall group mean repopulation level was 13.2%.

We next determined what proportion of the dendritic cell/macrophage subset of APCs must express H2^nb1 molecules to independently block type 1 diabetes development in an efficient fashion. This was done by reconstituting lethally irradiated (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 females with different proportions of admixed syngeneic and NOD bone marrow. In this case, all B-cells are of NOD origin, and different levels of H2^nb1-expressing dendritic cells/macrophages are present in various groups of recipient mice. The proportion of dendritic cells/macrophages displaying H2^nb1 molecules was determined by analyzing K^b expression on CD11c⁺ cells (Fig. 2). As shown in Fig. 3A, compared with controls reconstituted with NOD bone marrow alone, type 1 diabetes development was most efficiently suppressed in recipients in which H2^nb1 molecules were independently expressed on ≥50% of the dendritic cell/macrophage lineages of APCs. While also developing at levels that statistical analyses indicated to be significantly suppressed compared with controls reconstituted with NOD bone marrow alone, a more modest type 1 diabetes–protective effect was observed in recipients where the proportion of dendritic cells/macrophages independently expressing H2^nb1 molecules did not exceed 50% (Fig. 3A). In those recipients in which the proportion of H2^nb1-expressing dendritic cells/macrophages did not exceed 50%, the reconstitution levels of such APCs did not correlate with type 1 diabetes susceptibility or resistance (Fig. 3B). Furthermore, the finding that the engraftment levels of H2^nb1-positive dendritic cells/macrophages must approach 50% to efficiently inhibit type 1 diabetes development indicates that low-level contamination by these myeloid APCs was unlikely to be a protective factor in the studies that assessed the ability of highly purified B-cells (>92%) expressing the same MHC molecules to block disease onset.

H2^nb1-expressing APCs can mediate the deletion to variable degrees of at least some β-cell–autoreactive T-cell clonotypes normally contributing to type 1 diabetes in NOD mice (13,14). We previously demonstrated that H2^nb1 expression on 15–40% of dendritic cells/macrophages resulted in partial, but incomplete, deletion in both the thymus and spleen of the important diabetogenic CD8 T-cell clonotype AI4 (27). However, as shown earlier, H2^nb1 molecules must be expressed by at least 50% of the dendritic cell/macrophage lineages of APCs to efficiently inhibit type 1 diabetes development. Thus, we reasoned that an evaluation of the level to which AI4 T-cells develop when forced to differentiate under conditions where the proportion of H2^nb1-expressing dendritic cells/macrophages exceeded 50% could provide insight regarding the extent that such APCs mediate strong type 1 diabetes resistance by inducing the deletion of pathogenic effectors versus the activation of other potential regulatory mechanisms.

Lethally irradiated (NOD.Igμ^null × NOD.H2^nb1.Igμ^null)F1 mice were reconstituted with a 2:1 mixture of syngeneic F1 and NOD.AI4 bone marrow. At 8 weeks postreconstitution, 57–73% of the dendritic cells/macrophages in these chimeras heterozygously expressed H2^nb1 molecules (all B-cells were homozygous for H2^g7). Control chimeras consisted of the same type of F1 recipients reconstituted with a 2:1 mixture of standard NOD and NOD.AI4 bone marrow. All APCs in the control chimeras homozygously expressed H2^g7 molecules. AI4 T-cells were identified in the control and test chimeras by coexpression of CD8 and the TCR Vα8 element (Fig. 4A). Compared with their levels in controls, the numbers of CD8 single-positive thymocytes that expressed the AI4 TCR were reduced by >93% in the chimeras in which the proportion of H2^nb1-expressing dendritic cells/macrophages exceeded 50% (Fig. 4B). This carried over to the periphery, since numbers of AI4 T-cells in the spleens of the test chimeras were also reduced by >87% compared with their levels in controls (Fig. 4C). We previously found that some CD4 T-cells are still generated in NOD.AI4 mice but that their development and function is controlled by endogenously derived rather than coexpressed transgenic TCR molecules (24). This probably explains why peripheral levels of CD4 T-cells expressing the AI4 TCR were not influenced by the presence of H2^nb1-expressing APCs (data not shown). The array of autoreactive T-cell responses that aberrantly develop and cause type 1 diabetes in human patients and NOD mice are clearly normally prevented in most other individuals and mouse strains by multiple regulatory mechanisms (28). However, our current results show that the virtually complete deletion of at least AI4, and perhaps other highly pathogenic T-cell clonotypes, represents one contributory mechanism by which engraftment with a sufficiently high threshold level of dendritic cells/macrophages expressing dominantly protective H2^nb1 MHC molecules elicits type 1 diabetes resistance in normally susceptible NOD mice.

In the current study, we have demonstrated that expression of a disease-protective MHC haplotype (H2^nb1) solely on either dendritic cells/macrophages or B-cells is capable of inhibiting insulitis and type 1 diabetes development in NOD mice. However, dendritic cells/macrophages accomplish this more efficiently than B-cells. It was also found that their ability to independently inhibit type 1 diabetes development in NOD recipients was most efficient when bone marrow chimerization resulted in at least ∼50% of dendritic cells/macrophages expressing H2^nb1 molecules, albeit with lower but still significant protection achieved at lower repopulation levels. One likely explanation why dendritic cells/macrophages provide stronger H2^nb1-mediated type 1 diabetes–protective effects than B-cells is that the former myeloid lineages represent the primary type of APCs available in the thymus to mediate negative selection. Indeed, when dendritic cells and macrophages are the only types of APCs to express H2^nb1 molecules and have engrafted at levels that most efficiently inhibit type 1 diabetes development in NOD recipients, autoreactive CD8 T-cells of the AI4 clonotype undergo virtually complete intrathymic deletion. On the other hand, while they do so in a less efficient manner than the dendritic cell/macrophage lineages of APCs, our results indicate that H2^nb1-expressing B-cells are also able to significantly reduce insulitis and inhibit progression to type 1 diabetes in NOD mice most likely through peripheral tolerance induction mechanisms. These collective results suggest that perhaps due to an ability to activate a wider array of immunoregulatory mechanism than when limited to a single lineage, protective MHC haplotypes may most efficiently inhibit type 1 diabetes development when expressed by all types of APCs.

Certain HLA molecules, such as the DQ6 class II variant, also provide dominant type 1 diabetes resistance in humans (rev. in 15). Therefore, bone marrow chimerization with donor cells expressing protective HLA alleles represents an attractive treatment for prevention of type 1 diabetes in humans at risk for this disease or allowing reversal of hyperglycemia by cotransplantation of donor-matched pancreatic islets. However, clinical application of bone marrow chimerization protocols for type 1 diabetes prevention or reversal has been precluded by the risks associated with this procedure. One adverse outcome is the development of graft versus host disease. Another concern is the toxicity caused by high-dose irradiation, which is required to precondition recipients for donor bone marrow engraftment.

Previous studies (20–22) have successfully used preconditioning protocols not including a lethal dose of irradiation to achieve allogeneic bone marrow chimerization as a means to block type 1 diabetes development or reverse the disease in NOD mice when combined with donor-matched islet transplantation. However, in all cases, initially established mixed bone marrow chimerism could not be maintained, as virtually all hematopoietic cells converted to donor type at some point postreconstitution (20–22). While type 1 diabetes can be prevented, full allogeneic bone marrow chimerism also predisposes recipients to an undesirable immunocompromised state (27). We recently reported a new conditioning protocol that includes a low nonlethal irradiation dose (600 R) and injection of CD154 (i.e., CD40L)-blocking and CD8 T-cell–depleting antibodies (27). This protocol induced a stable partial state of allogeneic hematopoietic chimerism (∼50% donor derived) and completely prevented type 1 diabetes development in NOD mice. Most importantly, the generalized immunosuppression observed in full allogeneic bone marrow–chimerized recipients was avoided (27).

Our new preconditioning regimen, which does not require high-dose irradiation, allowed us to achieve encouraging results, but it did not allow us to establish allogeneic hematopoietic chimerism levels of <50% in NOD recipients (27). As a result, we were forced to use a high-dose irradiation preconditioning regimen to determine the minimal level of engraftment that must be achieved by various subtypes of APC-expressing type 1 diabetes–protective MHC molecules to block disease development in NOD recipients. One potentially significant outcome of these analyses is that the ∼50% engraftment threshold at which H2^nb1-expressing APCs efficiently block type 1 diabetes development in NOD recipients is a level achievable through use of our newly developed preconditioning regimen, which does not require a high lethal dose of irradiation.

In conclusion, we have demonstrated in the NOD mouse model that the spectrum of disease-inhibitory effects that can be elicited by a complete type 1 diabetes–protective MHC haplotype are most readily manifest when its gene products are expressed by the dendritic cell and macrophage subsets of APCs rather than B-cells. In addition, our results indicate that when dendritic cells and macrophages are the only APCs expressing the type 1 diabetes–protective H2^nb1 haplotype, at least 50% of them have to do so to achieve a maximal level of disease resistance. These results further enhance our knowledge for designing bone marrow transplantation protocols that may ultimately provide a therapeutic means to prevent or reverse type 1 diabetes in humans.

This work was supported by National Institutes of Health Grants DK46266 and DK51090, Cancer Center Support Grant CA34196, and grants from the Juvenile Diabetes Research Foundation International.