An Interval Tightly Linked to but Distinct From the H2 Complex Controls Both Overt Diabetes (Idd16) and Chronic Experimental Autoimmune Thyroiditis (Ceat1) in Nonobese Diabetic Mice

Inbred mice (NOD, CBA/J, and C57BL/6) were housed in our animal facility under specific pathogen-free conditions and in keeping with the European Union legislation on animal care. NOD.H2^k congenic mice were previously described (18). They were intercrossed with CBA/J mice to produce F1 and F2 generations.

NOD mice congenic for the portion of chromosome 17 derived from the C57BL/6 (B6) strain surrounding the H2 complex were bred by iterative backcrossing with NOD parents and by genotypic selection of heterozygous mice at each generation. The R0 homozygous line was established by appropriate sister-brother matings at the 17th generation of backrossing. Recombinant congenic lines were derived from the R0 line by backcrossing R0 heterozygous mice with NOD parents. The H2-K gene of R114 mice was of the NOD type, as indicated by typing the D17Mit28 microsatellite, which is located in the promoter of H2-K. The H2-K gene product was also verified by immunofluorescence labeling with monoclonal antibodies specific for H2-K^d (K9.18) or for H2-K^b (5F1). Both antibodies were generous gifts from Dr. François Lemmonier (Institut Pasteur, Paris).

Development of spontaneous diabetes in females was followed weekly by testing urinary levels of glucose with Glukotest (Roche Diagnostics, Basel). Mice were classified as diabetic after producing two consecutive tests ≥3+. Incidences of diabetes were compared with the log-rank test.

Chronic EAT was induced in 6- to 8-week-old mice of both sexes as described previously (21). Briefly, porcine thyroglobulin (100 μg, purchased from Sigma, St. Louis, MO) was emulsified in 100 μl of complete Freund’s adjuvant (Difco, Detroit, MI) and injected subcutaneously at four different points. Mice were boosted 2 weeks later with the same dose of antigen in incomplete Freund’s adjuvant. Then, 10 weeks later, mice were killed and thyroid glands were removed, fixed in formaldehyde (4%), and embedded in paraffin. Three serial sections at four noncontiguous levels were stained with hematoxylin and eosin and examined for mononuclear cell infiltration. As described in the results section, two types of infiltrates, including clusters and foci, were observed. For each type of infiltrate, counts were summed over the four levels. Immunization of mice and scoring of thyroiditis were carried out before, and therefore blindly relative to, genotyping.

Microsatellite markers were drawn from the Massachusetts Institute of Technology (MIT) database (22) (found at www-genome.mit.wi.edu). Some of the markers used for the genome screen of the (NOD.H2^kxCBA/J)F2 cross had to be tested for polymorphism between NOD and CBA/J in addition to those previously described (23). The PCR was carried out following standard conditions as described (24). PCR products were analyzed by electrophoresis on 5–6% agarose gels. At each marker, the NOD and CBA/J alleles were designated as N and C, respectively. Orders and distances were those of the Mouse Genome Database (MGD) (25) (found at www.informatics.jax.org). Previously unordered markers in the Idd16 region were ordered by typing recombinant haplotypes and by minimization of the number of recombination events. We found two major differences with the MGD map. First, we observed no recombination between D17Mit113 and D17Mit114. These markers are 4.5 cM distant in MGD. Second, we mapped D17Mit260 distal to D17Mit114, unlike MGD.

The list of microsatellites markers that were genotyped in (NOD.H2^kx CBA/J)F2 mice are listed in the online appendix, available at http://diabetes.diabetesjournals.org.

Initial examination of the data set indicated that thyroiditis phenotypes, including foci and clusters, did not follow a normal distribution, even after logarithmic or Arcsine transformation. Data were therefore analyzed with nonparametric methods, using the “Nonparametrics and distributions” module of the Statistica package (Statsoft, Tulsa, OK). Correlation between clusters and foci was tested with the Spearman-ρ and Kendall-τ coefficients. Because the two types of infiltrate were partially correlated, they were analyzed separately and together.

Qualitative phenotypes were considered first. For clusters, mice with at least one cluster were considered as affected. Likewise, for foci, mice with at least one focus were considered as affected. The homogeneity of frequencies of affected and nonaffected mice of each genotype, NN, NC, or CC, was tested with 3×2 tables. Exact probabilities of the Pearson χ² distribution were calculated with the StatXact software (Cytel Software, Cambridge, MA).

A semi-quantitative analysis was also carried out on clusters, foci, and a global score using the Kruskal-Wallis test. The global score was computed by summing the numbers of foci and clusters after assigning an equivalence of five clusters to one focus. Analyses made no assumption on the mode of inheritance of loci. In Table 3, genetic models, including dominance or recessivity, were tested with the Mann-Whitney U test after pooling genotypic groups appropriately. Only P values < 5% were reported. They were not corrected for the number of tests done. The threshold P value for calling linkage in a genome-wide scan of an F2 population was 5.2 × 10⁻⁵, as previously recommended (26).

An NOD.H2^b congenic strain, here also designated as R0, was bred by iterative backcrossing. It carries a 27-cM segment derived from the C57BL/6 (B6) strain centered on the H2^b complex, whose approximate size is 2 cM (Table 1). At the 17th generation of backcross, two recombinant congenic lines were selected. The R2 line retains the entire H2 and the distal part of the R0 interval (∼14 cM). Its proximal boundary is located between the D17Mit199 and D17Mit16 markers (Table 1). The latter marker maps 0.3 cM proximal to H2-K, the most centromeric gene of the H2 complex. The R114 line yields a mirror image of R2. The breakpoint is also between D17Mit199 and D17Mit16, but the DNA is of B6 origin in the proximal part of the R0 interval (∼14 cM). In this line, therefore, the H2 derives from the NOD (g7) haplotype. As shown in Fig. 1A, both lines are equally and strongly protected against diabetes (P < 1 × 10⁻⁵ for comparison of either R2 or R114 with NOD by the log-rank test). Such protection was expected for R2 mice because they carry the protective B6 MHC alleles. In addition, it was previously shown that the H2^b haplotype confers nearly dominant protection against diabetes, with almost complete penetrance in the heterozygous state (27,28). As shown in Fig. 1B, the protection afforded by the R114 interval is partially dominant.

Histopathological analysis showed that R114 mice develop an infiltration of Langerhans islets. However, it is milder than that of NOD mice at 12 weeks of age (Fig. 2). Eventually, the majority of islets of R114 mice become infiltrated at 24 and 38 weeks of age, sharply contrasting with the almost complete absence of clinical diabetes. Comparison with 24-week-old NOD mice should be made with the important reservation that only the rare NOD mice not yet diabetic can be analyzed. Therefore, R114 mice differ from previously described NOD.H2^b congenic mice (28) and also from R2 mice (data not shown) in that they develop insulitis. However, this inflammatory infiltration evolves at a slower rate than in NOD mice, unlike what is observed in NOD mice congenic at the Idd5 or Idd9 loci (29–31).

In an independent line of experiments, we sought to map non-MHC genes associated with susceptibility of NOD mice to chronic EAT. Progeny mice of F1 and F2 intercrosses between NOD.H2^k and CBA/J parents were induced with thyroglobulin plus adjuvant, and histopathology of their thyroid glands was assessed 10 weeks later, at a time when inflammation is no longer visible in conventional H2^k strains.

As previously observed (18), all CBA/J mice had recovered from thyroiditis, whereas 90.3% of NOD.H2^k mice still showed severe lesions (Fig. 3A). These lesions consisted of dense foci containing numerous (>100) mononuclear cells and disrupting the normal architecture of the thyroid gland (Fig. 3B). They were associated with partial or complete destruction of the thyroid vesicles. In F1 and F2 mice, the number and severity of the infiltrates were variable. Mild infiltrates were observed in addition to destructive foci. They consisted of clusters of mononuclear cells (<20) found in the interstitium between the thyroid vesicles and not associated with damage of the thyroid epithelium (Fig. 3C). These clusters were not seen in CBA/J or B10.BR mice 10 weeks after immunization. Among F1 mice, 34.7% (17 of 49) developed clusters, whereas a smaller proportion (12.2%, 6 of 49) had foci. This suggested a role of one or more partially dominant genes. Among F2 mice, 61.2% (112 of 183) had clusters and 28.4% (52 of 183) had foci. Although there was a significant correlation between the two types of infiltrates in these F2 mice (r = 0.56, P = 5 × 10⁻¹⁴), their segregation was also partially independent: 67 of the 112 mice with clusters (59.8%) displayed only clusters, and 7 of the 52 mice with foci (13.5%) had only foci (Fig. 3E). This suggested that the mechanisms underlying the formation of clusters and of foci were not necessarily linked. The two types of infiltrates were therefore analyzed both separately and together, using a global score and nonparametric methods.

F2 mice (n = 183) were genotyped for 81 microsatellite markers spanning the mouse genome. The frequency of mice affected for each phenotype was determined in each genotypic group. Three loci were detected (Table 2), of which only one, on chromosome 17, yielded a genome-wide significant association, defining the Ceat1 (for chronic experimental autoimmune thyroiditis 1) locus. Four markers covering a distance of 38 cM on this chromosome influenced both clusters and foci. The strongest association was with the D17Mit114 marker, which, as seen above, is proximal to the H2 complex. The other two loci influenced either clusters (on chr. 5) or foci (on chr. 7). Distribution of genotype frequencies suggested a dominant mode of action for Ceat1 and for the putative locus on chromosome 5 and a recessive inheritance for that on chromosome 7. A semi-quantitative analysis was also performed (Table 3) and confirmed the models as fitted just above.

Congenic and recombinant congenic NOD mice used for mapping Idd16 were tested for their responsiveness to induction of chronic thyroiditis. As shown in Table 4, R114 mice were fully protected from chronic EAT. This definitively demonstrated the existence of Ceat1 and established its tight linkage to Idd16.

The MHC extends over several megabases and is one of the most gene-rich regions of the genome (32). A large proportion of MHC genes are expressed in the immune system, but only a fraction of them have a known function. Distinguishing between the contribution of these numerous candidate genes will be a considerable endeavor, especially in humans. In the mouse, breeding of congenic strains and derivation of recombinant congenics still represent the most valid approach available to a precise understanding of MHC gene contribution to complex phenotypes and diseases (33).

Based on the analysis of the R114 recombinant congenic strain, our work brings definitive evidence for the existence of at least one MHC-linked locus other than class I and class II genes in diabetes predisposition. Affording almost complete protection, this locus is one of the strongest non–H2 Idd loci. If B6 and CTS strains share the same alleles proximal to H2-K, then it is very likely that the Idd16 MHC-linked resistance allele defined by CTS outcross (14) is also present in the B6 genome and has been positioned in our study. Of note, our present data do not rule out the existence of other loci, either proximal to H2-K (16) or adjacent or distal to H2-D (14,15). Idd16 therefore maps proximally to the H2 complex in a region whose distal boundary is limited by the D17Mit16 marker, and it is at least 0.3 cM proximal to the H2-K gene. The physical distance that separates the R114 interval from the H2 is not known precisely. Using available information from the ongoing Mouse Genome Sequencing program (found at www.ncbi.nlm.nih.gov/genome/seq/MmHome.html), it appears that a segment of 250 kb upstream of the H2-K gene has been sequenced and that it does not include the D17Mit16 marker. The Idd16 locus is therefore unambiguously distinct from the proper H2 complex.

Rather unexpectedly, whereas our genome scan was designed to identify MHC-unlinked genes for chronic EAT by crossing parent mice having a common MHC haplotype, the only genome-wide significant association was observed with the H2-linked Ceat1 locus. There was milder evidence for two other loci, which are nevertheless worthy of consideration because they influence clusters and foci differentially, suggesting that the two types of infiltrates might arise from different molecular mechanisms.

Importantly, the existence of the Ceat1 locus was demonstrated by the study of congenic mice. Because B10.BR mice do not develop chronic EAT (see Fig. 3A), Ceat1 can be mapped to the portion of NOD.H2^k chromosome of NOD origin, that is proximal to the crossover point between NOD and B10.BR and that was mapped between D17Mit260 and D17Mit100. Ceat1 is therefore currently included in a ∼8-cM interval extending from D17Mit114 to D17Mit100, the latter marker being excluded.

Tight linkage of Ceat1 and Idd16 is consistent with the occurrence of both types of autoimmune manifestations in NOD mice and provides a genetic basis for their association. In humans also, the association of autoimmune thyroiditis and type 1 diabetes is commonly observed (34). Whether Ceat1 and Idd16 represent two manifestations of a single gene is now open to investigation, which notably will require the breeding of new mouse congenic lines. However, both loci influence long-term evolution and pathogenicity of inflammatory infiltrates. The effect of the NOD haplotype at the Idd16 locus is to amplify a preexisting infiltration and to promote its development into an islet-aggressive infiltrate. In this regard, the Ceat1 locus seems to have a similar effect on thyroidal inflammatory infiltrate, as it requires a preliminary step that is provided by immunization with thyroglobulin. Chronic EAT in NOD mice has been associated with decreased T-cell-mediated response to thyroglobulin, production of γ-interferon, and switch of antibodies to a Th1-dependent isotype (18). How Ceat1 affects these cellular and molecular parameters will be important to determine and may provide useful clues for its molecular identification. The study of Idd16/Ceat1 may also be relevant to a better understanding of the behavior of inflammatory cell infiltrates in other immunopathological situations, notably viral and parasitic infections, graft rejection, and tumor-host interaction.

This work was supported by funds from INSERM and a grant of the Juvenile Diabetes Research Foundation International.

We are most grateful to Isabelle Cisse for expert assistance with mouse care, Olivier Babin for help with histopathology, Dr. François Lemmonier for the generous gift of antibodies, and Dr. Matthieu Levi-Strauss for careful reading of the manuscript.