The Idd4 Locus Displays Sex-Specific Epistatic Effects on Type 1 Diabetes Susceptibility in Nonobese Diabetic Mice

All mice used in this study were maintained in a specific pathogen-free facility at the Hospital for Sick Children. Spontaneous diabetes incidence at age 6 months in NOD/Jsd animals is 83% in females and 35% in males. All procedures performed on these mice followed the guidelines of the institutional animal care committee.

Genomic DNA was prepared from tail snips with a DNA Easy kit (Qiagen) for use in PCR amplification. All microsatellite markers used in this study were amplified as follows: 10 cycles: 30 s at 94°C, 30 s at 50°C, and 1 min at 72°C followed by 35 cycles: 30 s at 94°C, 30 s at 55°C, and 1 min at 72°C, in 1.5 mmol/l MgCl₂. The amplification products were visualized following electrophoresis through 2% NuSieve (American Bioanalytical) and 2% agarose (Life Technologies) gels.

Generation of the NOR.NOD-Idd4 (R1-R4) congenic strains was previously described (25). A novel NOR.NOD-Idd4 (R5) was generated using the same approach. Previously described NOR.NOD-Idd4 mice (24) were backcrossed to NOR, their pups were intercrossed, and the resulting progeny genotyped for D11Mit74, D11Mit340, D11Mit230, D11Mit135, D11Mit217, D11Mit310, D11Mit164, D11Mit157, D11Mit177, D11Mit4, D11Mit368, D11Mit30, D11Mit90, D11Mit364, D11Mit219, D11Mit322, and D11Bhm149. NOR.NOD-Idd4 (R5) subcongenic strain was generated by backcrossing informative recombinant to NOR and intercrossing the progeny to fix NOD-derived subcongenic interval. Two novel Idd4 subcongenic strains on the NOD genetic background (NOR.NOD-Idd4 [R1] and NOD.NOR-Idd4 [R2]) were generated from previously reported NOD.NOR-Idd4 mice (24) using a reciprocal strategy.

Cyclophosphamide treatment and diabetes assessment were performed as previously described (24). The difference in diabetes incidence between parentals and congenic strains was assessed using Fisher’s exact test in statistical package SPSS for PC version 8.01 (SPSS, Chicago, IL). Analysis of differences in CY-T1D incidence and latency among strains was determined with Kaplan-Meier life tables and computation of mean time to type 1 diabetes onset ± SD.

Novel microsatellite markers were generated as previously described (25). To identify informative markers, NOD, NOR, DBA/2J, and C57BL/6, DNA was amplified for 35 cycles with the following conditions: 30 s at 94°C, 30 s at 55°C, and 1 min at 72°C in 1.5 mmol/l MgCl₂. The amplification products were electrophoresed through either 2% NuSieve (American Bioanalytical) and 2% agarose (Life Technologies) or 8% acrylamide gels. Novel microsatellite markers used in this study are presented in online appendix Table 1 (available at http://diabetes.diabetesjournals.org).

The coding and regulatory regions of the Nos2 gene were compared between NOD and NOR strains by direct sequencing. cDNA prepared from lipopolysaccharide- and γ-interferon–activated bone marrow–derived macrophages/dendritic cells (25) was used as a PCR template under the following conditions: 30 s at 94°C, 30 s at 65°C, and 1 min at 72°C in 1.5 mmol/l MgSO₄ for 30 cycles, using high-fidelity Taq polymerase (Invitrogen, Carlsbad, CA). The regulatory regions of the Nos2 gene were sequenced from genomic DNA under the same conditions. Sequencing primers, designed with Primer Express 1.5 software (Applied Biosystems, Foster City, CA), are listed in Table 1. PCR products were sequenced on both strands at the Centre for Applied Genomics (Hospital for Sick Children, Toronto, ON, Canada). Sequence files were compared with Lasergene software (DNASTAR, Madison, WI).

Single nucleotide polymorphisms (SNPs) between NOD/LtJ, DBA/2J, and C57BL/6J strains were examined from the Perlegen (available at http://mouse.perlegen.com/mouse) and Broad Institute (available at http://www.broad.mit.edu/mouse) databases. Perlegen reported that SNPs were positioned on the National Center for Biotechnology Information sequence build 34 and Broad Institute SNPs on build 33. SNPs from the two sources were extracted using Pearl script; rows that contained an N (unresolved nucleotide) within these datasets were removed from the analysis. The remaining filtered data, 85 SNPs from the Broad Institute and 639 SNPs from Perlegen, were used for comparison between strains.

To test for sex-by-strain interaction effects, we performed a logistic regression analysis in which the log odds were modeled as a linear function of the covariates, as previously described (24,26). A significant sex-specific effect of a locus was assumed if the ratio of the odds of diabetes in the parental strain and the odds of diabetes in the congenic strain were different for the two sexes. We fitted a logistic regression model with three terms: a main effect for sex, a main effect for strain, and a sex-by-strain interaction term. We also fitted a second model without the interaction term that assumed that the log odds ratio was the same in both sexes. The test statistic for the significance of the sex-dependent effect was twice the difference in log likelihoods between these two models, with a χ² distribution with a single degree of freedom if the null hypothesis of no interaction is true. The P values presented were obtained using this χ² distribution.

We previously mapped Idd4 to a telomeric portion of chromosome 11 where the NOR genome is C57BLKS/J (BKs)-derived (24). To guide our mapping efforts, we first mapped the telomeric boundary of BKs-derived DNA on NOR chromosome 11. Multiple novel microsatellite markers were tested for the polymorphism between NOD and NOR strains (online appendix Table 1) and placed this boundary between BKs- and NOD-derived DNA between the D11Jyh1081 and D11Mit33 spaced 0.9 Mb apart. However, the telomeric BKs-derived segment on NOR chromosome 11 was not identical to either of the reported DBA/2 and C57BL/6 progenitor strains of NOR (27), suggesting an additional genomic contribution from an unknown strain (online appendix Table 1). Our results are consistent with a recent SNP mapping study of the BKs strain, which suggested that ∼9% of its genome differed from both DBA/2 and C57BL/6 (28).

Refinement of the telomeric boundary of Idd4 confirmed that inducible nitric oxide synthase (Nos2) resides within this region (78646511 bp). Multiple studies have demonstrated that nitric oxide can damage pancreatic β-islets (rev. in 29), suggesting that activity/expression of NOS2 may contribute to type 1 diabetes susceptibility. To investigate this possibility, we sequenced coding and regulatory regions of Nos2 in NOD and NOR strains. The regulatory regions were selected based upon the PipMaker algorithm for phylogenetic conservation (30), applied 10 kb upstream of the Nos2 transcription start site gene using mouse and human Ensembl sequence. This analysis revealed four phylogenetically conserved elements: 1) 1,800-bp region immediately upstream of the gene previously identified as a promoter region conferring lipopolysaccharide and interferon responsiveness (31), 2) a 600-bp region located 4,000 bp upstream, 3) a 150-bp region 7,000 bp upstream, and 4) a 1,000-bp region 8,000 bp upstream. The sequence of the minimal promoter region revealed no variations between NOD and NOR strains. Similarly, no differences in sequence were found in either coding or the predicted regulatory regions (data not shown). However, these data could not exclude the presence of polymorphisms between NOD and NOR in other unidentified regulatory regions. Genetic analysis of informative congenic strains was then used to exclude Nos2 as a candidate gene (see below).

Previously, we demonstrated that NOR.NOD-Idd4 mice are susceptible to CY-T1D (24) and refined the Idd4.1 locus to a 1.4-Mb interval using congenic strains (25). The latter study was performed in male mice due to the higher penetrance of the type 1 diabetes susceptibility in this sex. We generated an additional congenic strain (NOR-R5) with a recombination point between D11Mit90 and D11Mit364. Fine mapping of this strain using novel microsatellite markers placed the recombination event within 200 kb telomeric to the boundary of Idd4.1 between markers D11Gul2700 and D11Gul2721 (data not shown). When male and female NOR-R5 mice were assessed for CY-T1D and compared with the parental NOR strain, a striking sex difference in disease susceptibility was observed. In contrast to all other Idd4 recombinants on the NOR background, where CY-T1D susceptibility was equivalent in both sexes compared with NOR parental mice (25; Fig. 1), cyclophosphamide treatment of NOR-R5 mice had a distinct effects depending on the sex of the mice. NOR-R5 males progressed to diabetes (32% type 1 diabetes, n = 60; P = 0.011 vs. NOR; 11% type 1 diabetes, n = 36), whereas females were CY-T1D resistant (P = 0.76 vs. NOR). Analysis of NOR-R5 also indicated that our previously reported strong effect of NOD-derived Idd4.1 locus in males (25) was insufficient to confer CY-T1D in females. Taken together with the observation that NOR-R3 females carrying a longer NOD-derived segment were susceptible to CY-T1D (P = 0.011 vs. NOR), these data suggested that NOD Idd4-mediated CY-T1D susceptibility in females required at least two genes, Idd4.1 and Idd4.2. Enhanced susceptibility to CY-T1D in NOR-R1 and NOR-R3 congenic females was not associated with a change in time to disease onset (Kaplan-Meier analysis data not shown; mean time to onset ± SD in days: 21.00 ± 5.44 to 24.91 ± 3.78). Based upon these NOR recombinant congenic strains, Idd4.2 was located in a 5.6-Mb interval, between D11Gul2700 and D11Mit322.

Having observed that NOD.NOR-Idd4 females, but not males, were protected from CY-T1D (24), we generated two Idd4 subcongenic strains on the NOD background and analyzed them for susceptibility to CY-T1D (Fig. 2). Compared with NOD mice, NOD-R1 females were not protected from disease (P = 0.18), whereas NOD-R2 animals were CY-T1D resistant (P = 0.001). Decreased incidence of CY-T1D in NOD-R2 congenic females was not associated with a change in time to disease onset (Kaplan-Meier analysis data not shown; mean time to onset ± SD in days: 19.00 ± 5.82 to 20.05 ± 6.20). Since NOD-R1 were not distinguishable from NOD females, the region required to confer CY-T1D protection resided telomeric to D11Mit219 consistent with a second locus, Idd4.2.

To refine the position of Idd4.2, we utilized the Celera and Ensembl databases of mouse genome sequence to define a novel set of microsatellite markers (see research design and methods). PCR amplicons containing the microsatellite repeats were tested for polymorphism between NOD and NOR strains (online appendix Table 1), and informative markers were used to generate a high-resolution map of the recombination boundaries in the relevant congenic strains. Analysis of the NOD-R1 strain with new markers between D11Mit219 and D11Mit322 positioned the centromeric boundary of Idd4.2 within a 6-kb interval defined by D11Gul4 and D11Gul5 (Table 2). The telomeric boundary, identified by genotyping the NOR.NOD-R3 strain, was placed in a ∼500-kb interval between the D11Gul33 and D11Gul80. These results refined Idd4.2 to a 0.92-Mb region including the boundary regions containing 29 genes (Table 3). To date, further refinement of the 500-kb Idd4.2 telomeric boundary has been limited by the absence of polymorphic markers between NOD and NOR strains in this region.

Having refined the Idd4.2 locus to 0.92 Mb, we then considered the 29 genes in the interval (Table 3). Of these, 14 are olfactory receptors located within the 0.7-Mb telomeric portion of the locus. Due to the restricted tissue expression pattern of these genes (olfactory epithelium), they were viewed as low priority as candidate genes for Idd4.2. To better appreciate the genomic structure of the locus, we compared the NOD, C57BL/6, and DBA/2 strain SNPs from the Perlegen and Broad Institute databases (online appendix Tables 2 and 3). A summary of these comparisons is provided in Table 4.

The marker distribution across Idd4.2 is different between the Broad Institute and Perlegen datasets because the latter lacked SNPs within the ∼0.7-Mb telomeric region of the locus (online appendix Table 2). Indeed, the majority of Broad Institute markers within Idd4.2 that were not polymorphic between NOD, DBA/2, and B6 were clustered within this 0.7-Mb region. The paucity of informative markers in this portion of the locus limited further refinement of the Idd4.2 telomeric boundary. In the centromeric portion of the locus, the majority of the markers was identical between NOD and B6 and differed from the DBA/2 strain. The genomic diversity at Idd4.2 is consistent with our previous analyses of NOD genomic sequence of the Idd4.1 locus revealing remarkable similarity between NOD and B6 strains (25). Thus, NOD and B6 appear to share an extended ancestral haplotype spanning Idd4.1, the Idd4.1–Idd4.2 interlocus region, and the centromeric portion of the Idd4.2. Broad Institute SNP data for C57BLKS/J, the other progenitor of NOR mice (27), showed that all Idd4.2 SNPs were identical between C57BLKS/J and DBA/2 (online appendix Table 3), indicating that, as for Idd4.1, NOR-derived protection from CY-T1D at Idd4.1 and Idd4.2 is mediated by DBA/2-derived alleles.

The Idd4 locus displays sex-specific behavior and consists of at least two loci: Idd4.1 and Idd4.2 (25; this study). Given the sex-dependent phenotype of NOR-R5 mice, the NOD Idd4.1 locus alone conferred CY-T1D in males but not in females. We compared NOR versus NOR-R5 and NOR versus NOR.NOD-Idd4 strains to determine whether the effects of the genotype differed by sex. A χ² likelihood ratio test was done to determine whether an interaction term was needed to explain these effects. Comparison of NOR-R5 and NOR mice showed a significant sex-by-strain interaction effect (Fig. 3; P = 0.038), supporting the sex-specific behavior of alleles at the Idd4.1 locus. In the NOR-R4 and NOR-R5 congenic strains, neither NOD Idd4.1 nor Idd4.2 alone was sufficient to confer CY-T1D risk in females (Fig. 1). When NOD alleles at Idd4.1 and Idd4.2 were combined in the NOR-R3 strain, females did acquire significant susceptibility to CY-T1D. These findings are consistent with an epistatic interaction between NOD alleles at Idd4.1 and Idd4.2 that together confer risk for CY-T1D development in females. While NOD Idd4.1 alone increased CY-T1D susceptibility in NOR congenic males, comparison between cohorts of male NOR-R5 (19 type 1 diabetic, 41 non–type 1 diabetic) and NOR.NOD-Idd4 (24 type 1 diabetic, 23 non–type 1 diabetic) mice showed a difference (two-sided P = 0.0423), consistent with modest additive effects of NOD alleles at Idd4.1 and Idd4.2.

We present a high-resolution analysis of the Idd4 diabetes susceptibility locus and document features of the complex genetic architecture of the region. Idd4 was found to consist of two regions, Idd4.1 and Idd4.2 (25; this study). Strikingly, both Idd4.1 and Idd4.2 displayed sex-specific effects on CY-T1D susceptibility. NOD-derived Idd4.1 was sufficient to confer enhanced CY-T1D frequency in NOR congenic males, whereas both NOD-derived Idd4.1 and Idd4.2 were required in females, consistent with a sex-specific epistatic interaction between these loci. Theses results complement our earlier report that the Idd4 locus behaved in a sex-specific manner (24). In contrast to most other autoimmune diseases, type 1 diabetes is often cited as an organ-specific autoimmune disease without a female sex bias (rev. in 32). While parity between sexes is generally observed in higher-incidence populations, lower-incidence groups including African Americans and Mexicans show a male bias, while Asians display a female bias (33). Furthermore, there is evidence that the peak age at onset occurs earlier in girls than in boys (34). A recent study examined association between an SNP in the estrogen-responsive IL6 promoter (174G/C) and age at onset in boys and girls. Homozygosity for the 174C genotype was significantly less frequent in females diagnosed after 10 years of age than those diagnosed at younger ages, while no 174G/C genotype difference was observed in males stratified for age at onset (35). Thus, in the context of evidence that many autoimmune disease including type 1 diabetes are influenced by sex, the Idd4 locus in the NOD model provides a system to examine genetic mechanisms that underlie the sexual dimorphism in type 1 diabetes.

Idd4 was first identified in a genome-wide linkage analysis of (NOD × B10.^H-2g7) × NOD animals, as a 30-cM region on chromosome 11 linked to early-onset type 1 diabetes (36). Linkage of the Idd4 locus to spontaneous type 1 diabetes was later confirmed in NOD.B6 congenic animals, where the strongest diabetes resistance mapped to a 20-Mb interval between the D11Mit30 and D11Mit41 markers (37). Both the Idd4.1 and Idd4.2 loci described in our study reside within this interval. Grattan et al. (37) also proposed existence of the two loci within Idd4. However, the resolution provided by these NOD.B6-Idd4 congenic strains limit determination of the precise positional relationship between their Idd4 loci and those we have reported (25; this study).

Which genes within Idd4.2 might regulate susceptibility to CY-T1D and how could they explain the sex-dependent behavior of this locus? The 14 olfactory receptor genes in the Idd4.2 region (Table 3) displayed expression restricted to the olfactory epithelium, making them low priority candidates for type 1 diabetes susceptibility. Among the remaining genes two purinergic receptors, P2rx1 and P2rx5, and an integrin α-E (Itgae) have links to immune functions.

P2RX1 and P2RX5 belong to a family of nonselective cation channel proteins activated by extracellular ATP. Interestingly, P2rx1-deficient males, but not females, displayed reduced fertility compared with the wild-type counterparts, suggesting that function of this receptor in some tissues depends on sex (38). P2RX5 expression is highest in brain and the immune system, implicating a possible role in immune responses (39). Another member of the P2X receptor family (P2RX7) has been directly implicated in response to infection, as macrophages from P2rx7-deficient mice were unable to kill intracellular bacteria (40). P2rx7 deficiency also inhibited secretion of the proinflammatory cytokine interleukin-1 resulting in greater resistance to collagen-induced arthritis in a mouse model (41). P2X receptors form channels as multimers of several subunits, and both homomeric and heteromeric receptors have been described following ectopic expression in cell lines (rev. in 42). One speculation is that either P2RX1 or P2RX5 (or both) may form heteromeric channels with P2RX7 contributing to the immune phenotypes described above. Much functional analysis is needed to appreciate the contribution of P2X channels to immune function, but current evidence makes these two genes interesting candidates for Idd4.2.

Itgae encodes an integrin cell surface adhesion protein (CD103) thought to play an important role in mucosal immunity. CD103 is expressed on >90% of intestinal intraepithelial lymphocytes but on <2% of circulating peripheral blood T-cells (43). Recent studies (44–46) have suggested that coexpression of CD103 on CD8⁺ T-cells marks a potent cytotoxic subset and that some CD103⁺CD8⁺ T-cells have immune regulatory functions. Consistent with its role in mucosal immunity, Itgae-deficient mice had reduced numbers of gut and vaginal mucosal T-cells (47) and developed a cutaneous inflammatory disorder (48). Interestingly, the majority population of dendritic cells in the lung is CD103⁺, and their numbers are increased in mice with induced airway hyperresponsiveness, suggesting a potential role of these cells in asthma pathogenesis (49). Thus, CD103 may play multiple roles in regulating and effecting the immune response making it an attractive candidate as the type 1 diabetes susceptibility gene.

We present a high-resolution map of the type 1 diabetes susceptibility Idd4 locus on mouse chromosome 11 and provide strong evidence for two loci Idd4.1 and Idd4.2. Both intervals displayed dramatic sex-dependent effects. We previously confined Idd4.1 to a 1.4-Mb interval containing 52 genes (25) and showed that the locus conveyed its strongest effect in males. In the present study, Idd4.2 was refined to 0.92 Mb containing 29 genes. NOD alleles at Idd4.1 and Idd4.2 were necessary for CY-T1D susceptibility in females, and both NOD-derived Idd4.1 and Idd4.2 were required for the induction of CY-T1D, consistent with epistatic interaction between them. Efforts are underway to conduct DNA sequence, gene expression, and functional analyses of candidate genes at Idd4.1 and Idd4.2 to further elucidate the mechanisms involved in the genetic control of sex-dependent type 1 diabetes susceptibility by Idd4.

This study was funded by grants to J.S.D. from the Canadian Institutes of Health Research (Institute of Genetics) and from Genome Canada through the Ontario Genomics Institute. E.A.I. was supported by studentships from the Hospital for Sick Children Research Training Centre, the Banting and Best Diabetes Centre Novo-Nordisk Studentship, and the Ontario Graduate Scholarship.

The authors thank Drs. S. Scherer and R. McInnes for critical comments on the manuscript.