Mapping Non–Class II H2-Linked Loci for Type 1 Diabetes in Nonobese Diabetic Mice

There is substantial evidence that class II loci do not fully explain the genetic predisposition to type 1 diabetes as linked to major histocompatibility complex (MHC) (1). In human type 1 diabetes, the identification of the additional loci remains a challenge owing to strong linkage disequilibria across the HLA complex (2,3). Conversely, in the mouse, congenic strains provide a powerful tool to overcome some of the problems arising from allelic associations (4). Using the nonobese diabetic (NOD) mouse strain, evidence for at least one non–class II MHC-linked locus, termed Idd16, was obtained (5,6). The Idd16 locus was further defined with the breeding of a congenic line, R114, harboring a 12-cM interval from C57BL/6 mice, proximal to and separated from the H2 complex (7). NOD.B6-Idd16 mice were markedly protected against diabetes and also against chronic autoimmune thyroiditis after immunization with thyroglobulin in Freund’s complete adjuvant.

In the present study, we have refined the Idd16 interval using a first series of five subcongenic lines derived from R114. Four lines, including R76, R14, R115, and R12, shared the distal end of the R114 interval and retained segments of variable extent from this end (Table 1). As shown in Fig. 1A–C, the three lines with the largest segments, R76, R14, and R115, were protected against spontaneous diabetes, although the nearly complete protection afforded by the original R114 interval (with a 5% final incidence of diabetes) was not maintained. Mice of the R115 strain, corresponding to the smallest protective interval, had an incidence of diabetes of 27.3% compared with 77.8% in NOD mice (P = 2 × 10⁻⁵). Diabetes incidence in R14 (52.7%) and R76 (51.8%) mice was similarly decreased compared with NOD mice (P = 5 × 10⁻⁴ and P < 1 × 10⁻⁶, respectively) and was not significantly different from that in R115 mice. In contrast, R12 mice, which harbored the smallest interval, were not protected at all against diabetes (Fig. 1D). The fifth line, R156, retained the proximal part of the R114 interval and was of the B6 type between the D17Mit113 and D17Mit260 markers (Table 1). The R156 mice had a slightly but significantly decreased final incidence of diabetes (64.15%) compared with NOD mice (P = 0.03) (Fig. 1E).

Taken together, the data indicated that the protection afforded by the original R114 region was caused by at least two loci. The most protective one, which we propose to designate as Idd16, could be mapped between D17Mit100 on the centromeric side, i.e., the proximal boundary of the R115 interval, and D17Mit101 on the telomeric end, i.e., the proximal boundary of the R12 interval. The physical size of the Idd16 interval is currently estimated to be 3.1 Mb, and its distal boundary is situated at 4.75 Mb, centromerically to the H2-K gene, the most proximal locus of the H2 complex. It is located toward the proximal end of the intervals described in previous studies (5,6). The gene mapped by Hattori et al. (6) in the G3-B line could be the same, since the diabetes-resistant strain used in their study, B10.A(R209), had a non-H2 genetic background (C57BL/10) that is closely related to the C57BL/6 used in our study. This is also assuming that the H2^r209 haplotype in this strain did not extend into the Idd16 interval. In contrast, as previously discussed (7), the gene mapped by Igekami et al. (5) could well be different, as the partner strain used in their study (CTS) appears to show specific allelic variations compared with NOD (8).

The second locus, defined by the R156 interval and which we propose to designate as Idd23, provided a weaker protection and mapped to a still large chromosomal segment of ∼20 Mb. It could explain the higher incidence of diabetes in the G3-B compared with the G3-A line reported by Hattori et al. (6), since G3-B differed from G3A in that it lost a centromeric segment from the donor strain overlapping with the R156 interval.

The two loci, Idd16 and Idd23, appeared to differ by their mode of action. The protection afforded by the B6 allele of Idd23 was dominant as (NODxR156) F1 mice had an incidence of diabetes similar to that of homozygous R156 mice and different from that of NOD mice (P = 0.01) (Fig. 1E). In contrast, (NODxR14) F1 mice developed diabetes similarly to NOD mice and differently than the R14 homozygous mice (P = 0.003) (Fig. 1B), suggesting a recessive mode of action for Idd16.

Using a second series of subcongenic lines derived from the R1 congenic strain, we have also investigated the region distal to the H2 complex. One line, R289, retained a large interval for which the maximal size is currently estimated at 12.1 Mb (Table 1). The proximal breakpoint was localized in the H2 complex, between the Lta and H2-Q4 genes. Therefore, in this line, the proximal class I and the class II genes were of the NOD type, while the distal end of the H2 region was of the B6 type. The distal boundary of the interval mapped between the D17Mit36 and D17Mit66 markers. The B6 interval harbored by the second line, R300, was included in that of R289 and shared its distal end with it. The proximal breakpoint was not precisely located in a 5.2-Mb segment, between a microsatellite telomeric to the H2-Q gene cluster and the D17Mit105 marker. As shown in Fig. 1F, the R289 line had an incidence of diabetes of 53.6% and was therefore mildly but significantly protected (P = 0.044). In contrast, R300 mice developed diabetes similarly to the control mice. We propose to designate this locus as Idd24. It currently maps between the Lta locus and the D17Mit105 marker. It should be nonetheless confirmed by a second follow up of diabetes and the breeding of subcongenic strains.

Altogether, we were able to identify three distinct intervals with a role in type 1 diabetes, in addition to the class II genes, in a 35-Mb chromosomal segment surrounding the H2 complex. The use of congenic mice harboring defined segments was essential to demonstrate the existence of the different loci, given the strength of the class II loci in type 1 diabetes predisposition. Previous studies have not reported a type 1 diabetes locus more distant from the H2 complex on chromosome 17 (9,10). We also failed to detect such a locus in an (NODxB6)xNOD backcross (unpublished data). Therefore, it is unlikely that an additional major locus on this chromosome influences diabetes predisposition by itself, at least in the strain combination investigated.

As previously reported, the R114 interval profoundly influenced the development of inflammatory infiltrates directed at self-antigens, including in the pancreata and the thyroid glands of NOD mice (7). To investigate the cellular response against an exogenous antigen, mice were immunized with ovalbumin emulsified in complete Freund’s adjuvant and were challenged 7 days later in the footpads with the antigen or with PBS in incomplete Freund’s adjuvant. As shown in Fig. 2, specific swelling was significantly more intense and more prolonged in NOD than in B6 mice (P < 1 × 10⁻⁴). However, R114 mice responded exactly as NOD mice, strongly suggesting that the immunoregulatory pathways involved in the response to foreign antigens and against self-antigens are not identical and that the R114 interval might specifically alter the control of anti–self-reactivity. Consistent with this hypothesis was the resistance of R114 mice to cyclophosphamide (CYP)-induced diabetes. CYP is a DNA alkylating agent with complex immunoregulatory properties (11–13). Following treatment with a high dose of CYP, the course of diabetes is markedly accelerated in NOD mice (14). As seen in Fig. 3, the final incidence of diabetes in R114 mice treated with this protocol was significantly decreased (P = 0.0028). Taken together, the data indicate that the R114 interval affects an important immunoregulatory pathway controlling the development of type 1 diabetes (15) and potentially the response against other self-antigens as well. The study of R114 mice might also provide useful insight into the immunopharmacologic properties of CYP.

The maximal size of the Idd16 interval, currently 3.1 Mb, was estimated using the extreme position of the recombination breakpoints, i.e., D17Mit100 and D17Mit101. It will be essential to refine their actual location by typing the neighbor polymorphisms to reduce the interval that currently contains 37 annotated genes (Ensembl database, build 32). Among the candidates, the tightly linked MAPK13 and MAPK14 genes encode the p38δ and p38α mitogen-activated protein kinases (MAPKs), respectively. The p38 MAPK pathway exerts a major influence on activation of inflammatory T-cells and notably on production of interferon-γ (16). In addition, treatment of NOD mice with an inhibitor of p38 MAPK prevents the onset of diabetes (17). However, the effects of p38 MAPK are likely to be quite complex, as STAT1, which mediates signaling of interferon-γ, is phosphorylated by p38 MAPK (18). This transcriptional activator plays an important role in the development of regulatory T-cells, and its germline inactivation increases susceptibility to autoimmunity (19).

The human synteny equivalent to Idd16 is tightly linked to the HLA complex and maps just centromerically to the class II region. A previous study reported an association of the D6S291 marker, situated at 270 kb from the human MAPK14 gene, with type 1 diabetes in the Scandinavian population (20) but not in other populations (21,22). However, a low density of markers was used in these investigations, and this large region shows a low level of linkage disequilibrium, contrary to the HLA region itself (21–23). Our findings might therefore justify a reinvestigation of this region in different populations using a denser map of markers.

The finding of several Idd loci linked to the H2 complex is consistent with the high density of genes expressed in the immune system in this region and contributes to the explanation of its unique strength in type 1 diabetes predisposition. Importantly, H2-congenic mice are widely used in functional studies, notably aiming at understanding the pathogenesis of type 1 diabetes. Our study emphasizes the importance of accurately defining the congenic intervals to improve the correlation between the cellular and molecular changes observed and type 1 diabetes pathogenesis.