Virus-Induced Autoimmune Diabetes in the LEW.1WR1 Rat Requires Iddm14 and a Genetic Locus Proximal to the Major Histocompatibility Complex

LEW.1WR1 (RT1A^uB/D^uC^a, ART2a) rats were obtained from Biomedical Research Models (Worcester, MA). WF.ART2 rats (RT1u/u, ART2a) were developed by us and obtained from the colony at the University of Massachusetts Medical School; they differ from ordinary WF animals in that they express the BB origin “a” rather than the “b” allotype of the ART2 T-cell alloantigen (25,31). WF.ART2a rats do not develop diabetes either spontaneously or in response to KRV infection (25). WF.ART2 animals are referred to as WF in this report. (LEW.1WR1 × WF) F2 rats were bred by us, housed in viral antibody–free conditions, confirmed monthly to be serologically free of rat pathogens (31), and maintained according to guidelines of the institutional animal care and use committees of the University of Massachusetts, Biochemical Research Models, and the Guide for the Care and Use of Laboratory Animals (32).

Diabetes was induced using KRV (University of Massachusetts) propagated in normal rat kidney cells (ATCC CRL-6509) as described (25). Poly I:C (Sigma, St. Louis, MO) was dissolved in Dulbecco's PBS, sterile filtered, and stored at −20°C until used. Contaminating endotoxin concentration was <50 units/mg (Charles River Endosafe, Charleston, SC). Rats 21–25 days of age of either sex were injected intraperitoneally with poly I:C (1 μg/g body wt) on days −3, −2, and −1 and with KRV (10⁷ plaque-forming units) on day 0. We have previously observed that this short course of poly I:C does not itself elicit diabetes but does increase the frequency of diabetes in KRV-treated LEW.1WR1 rats from ∼40 to 100% (R.S. Tirabassi, D.L.Gr., E.P.B., J.H. Leif, B.A. Woda, Z. Lui, D.A. Winans, D.L.Gu., J.P.M., unpublished data). LEW.1WR1 rats injected chronically with poly I:C (1 μg/g three times per week) become diabetic within 2 weeks (22); we used this protocol to elicit diabetes in rats intended for certain gene expression studies. Animals were screened three times weekly for glycosuria (Tes-Tape; Eli Lilly, Indianapolis, IN) for 40 days. Diabetes was diagnosed on the basis of plasma glucose concentrations >250 mg/dl (OneTouch Ultra Glucometer; LifeScan, Milpitas, CA).

To quantify insulitis, pancreata were harvested and fixed in 10% formalin; 10-μ sections were stained with hematoxylin and eosin and scored by light microscopy on a scale of 0 to 4+ as described (31).

DNA samples were prepared from tail snips or liver samples and genotyped using microsatellite markers as described (25,31). LEW.1W and LEW.1A genomic DNA samples were the gift of Dr. Dirk Wedekind (Hanover, Germany). Genotypes for 133 evenly distributed microsatellite markers were collected in the Map Manager QT program and exported for further analysis to Windows Quantitative Trait Locus (QTL) Cartographer, version 2.5 (33). Severity of insulitis and days to onset (nondiabetic rats being scored as 41 days) were used as the quantitative subphenotypic parameters of diabetes in this QTL analysis. Composite interval mapping was performed using model 6 to identify QTL peaks with a significant likelihood ratio test (LRT) statistic. Permutation analyses to establish thresholds of significance were performed using Map Manager QT, set for 1,000 permutations of the data (34).

High-fidelity Taq polymerase (Platinum PCR Supermix High Fidelity; Invitrogen) was used to amplify genomic DNA or cDNA, and these products were used as sequencing templates. Primers for amplification were designed using the BN rat genome sequence (available at genome.ucsc.edu) and Primer 3 software (available at http://frodo.wi.mit.edu/primer3).

In addition to full sequences, single nucleotide polymorphisms (SNPs) were typed on chromosome 20 using the phototyping technique (35). Two alternate allele-specific SNP primers and one common opposite-strand primer were designed for each SNP using Primer 3 to have a T_melt of 58–60°C, where the allele-specific primer has a 3′ nucleotide that matches one of two SNP alleles. SNP haplotypes were assembled in an Excel database and used to delineate the breakpoints of the recombinant chromosome 20 in LEW.1WR1 rats. Only SNPs that distinguish the two LEW.1WR1 parental strains, LEW.1A and LEW.1W (26), are shown in this report.

cDNA was prepared from rat spleen and mesenteric lymph node RNA using the Omniscript RT Kit 200 (Qiagen). Primers for rat UBD (diubiquitin), TNF (tumor necrosis factor), IFN-γ (interferon-y), and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) were designed from rat genomic sequences as above (online appendix Table 1). Quantitative real-time PCR was performed using SYBR Green PCR mix on an ABI 7900 sequence detector.

Life tables were analyzed using the method of Kaplan and Meier and log-rank statistics (36). Parametric data are given as arithmetic means ± SD. Two-group comparisons used unpaired t tests, and comparisons of three or more groups used ANOVA. Real-time quantitative PCR data were analyzed with the Mann-Whitney test and ANOVA. Two-tailed P values <0.05 are considered statistically significant. Genetic interaction between two QTL was assessed using one-way ANOVA adding in each factor (QTL) sequentially.

In a preliminary study, 97% of LEW.1WR1 rats (29 of 30) developed diabetes after a 3-day low-dose pretreatment with poly I:C followed by 10⁷ plaque-forming units of KRV. None of six rats became diabetic after this brief course of poly I:C alone. We then tested (LEW.1WR1 × WF) F1 rats using the same protocol, and 44% (26/59) became diabetic with an average latency of 15.5 days.

To generate a linkage map of Iddm loci, we then created a cohort of 182 (LEW.1WR1 × WF) F2 rats and, again, treated them with the same induction protocol. Of these, 57 became diabetic (31%; 27% of females and 36% of males, P = 0.22). Average latency to diabetes among sick animals was statistically similar in male and female rats (16.4 ± 2.7 and 16.2 ± 2.9 days, respectively; P = not significant).

Pancreata obtained from 175 (LEW.1WR1 × WF) F2 rats were scored for insulitis. Among diabetic rats, mean histological score was 3.7 ± 0.5 with no scores <3+. Among nondiabetic rats, the mean insulitis score was 0.6 ± 0.8 (range 0 to 2+; P < 0.0001 vs. diabetic rats); only 19.5% (25 of 128) had insulitis with scores of 1+ to 2+.

We predicted that Iddm14 and/or Iddm20 would be important for diabetes induction by KRV following poly I:C in LEW.1WR1 rats, just as they are required for KRV-induced BBDR diabetes (25). Therefore, we prepared genomic DNA samples from 182 (LEW.1WR1 × WF) F2 animals and genotyped them for markers that identify the Iddm14 and Iddm20 intervals on chromosomes 4 and 17, respectively. Iddm14 (as marked by its canonical microsatellite marker D4Arb9 [37]) was highly significantly linked to diabetes in this cohort. All 57 diabetic F2 rats bore at least one Iddm14 allele from the LEW.1WR1 parent (P = 1.5 × 10⁻¹¹, Fisher 2 × 3 exact probability test). There was no linkage to the Iddm20 QTL on chromosome 17 (25), which segregated randomly in this cohort with respect to diabetes in both the total F2 (P = 0.493) and in those F2 rats bearing at least one copy of the susceptible Iddm14^L allele (P = 0.434).

As we observed for the (BBDR × WF) F2 cross (25), not all (LEW.1WR1× WF) F2 rats with diabetogenic Iddm14 alleles became sick during the 40-day observation period. Only 41% (36 of 91) of heterozygotes and 57% (21 of 37) of Iddm14^L/L homozygous rats were diabetic, indicating that another gene (that is not Iddm20 [25] on chromosome 17) is required for diabetes in this strain combination (Table 1). We therefore conducted a full genome scan of F2 to find this QTL.

Permutation analysis of the dataset revealed that “suggestive” linkage for a QTL required an LRT ≥10.5; “significant” linkage, an LRT ≥17.5; and “highly significant” linkage, an LRT ≥25.7. Our first QTL, Iddm14, has a remarkably high LRT of ∼49 (Fig. 1), and it is therefore a highly significant predictor of diabetes in this strain combination. A second QTL was found on chromosome 20 (Fig. 1), and we provisionally designate it Iddm37. It is a significant predictor of diabetes and severity of insulitis in this F2 cohort (LRT = 19.5 and LRT = 29, respectively).

The Iddm14 locus was significantly linked to increased diabetes incidence only when rats carried at least one LEW.1WR1-derived allele of D20Rat47, which, at ∼2.8 Mb from the top of chromosome 20, serves as the closest-linked marker for Iddm37. Among 128 rats bearing a susceptible allele of Iddm14, there was a significant increase in diabetes frequency in Iddm37^L/L (46%), or Iddm37^L/W (63%) animals compared with Iddm37^W/W rats (19%, P < 0.0001, χ² test) (Table 2). Thus, while diabetes was absolutely dependent on the presence of at least one LEW.1WR1-derived allele of Iddm14, diabetes penetrance was significantly enhanced in those rats that also bore LEW.1WR1 alleles at Iddm37 (Fig. 2).

No other QTLs were significantly linked to diabetes incidence, latency, or insulitis in the genome scan of the whole F2 cohort (Fig. 1). However, among the diabetic animals, we noted a significant linkage of latency to Iddm24 (Fig. 3), a diabetes latency QTL originally discovered in a cross between WF and BBDP rats (38). In both crosses, Iddm24 was marked by microsatellite markers at 110–120 Mb distal to the centromere. Among diabetic rats, WF-derived alleles of Iddm24 were protective when homozygous and prolonged latency from 14.5 ± 2 days in Iddm24^L/L rats to 17.6 ± 2 days in Iddm24^W/W rats (P = 0.0056).

Iddm37 is located in the extended RT1C region of the MHC, which is known to differ among LEW.1W, LEW.1A, WF, and LEW.1WR1 rats (39). Because LEW.1WR1 rats bear a recombinant MHC haplotype, it remained to be determined which of three independent MHC-linked haplotypes (LEW.1A, LEW.1W, or the LEW background) contributed the diabetes-promoting allele that we identified as the Iddm37 QTL. The original mapping of LEW.1WR1 suggested that the RT1C region in LEW.1WR1 rats was donated by LEW.1A. To confirm this, we conducted an extensive SNP analysis of the chromosome 20 region containing Iddm37. SNPs representing alleles from LEW.1WR1 rats and their three parental strains were identified. The strain distribution pattern of SNP alleles for genes in the RT1C region indicates that the majority of LEW.1WR1 alleles are of LEW.1A origin, consistent with the original mapping (Table 3).

The Iddm37 region includes the gene for diubiquitin (UBD), a candidate gene of interest because it resides in a region that is homologous to mouse and human diabetes-promoting intervals (40,41). We therefore sequenced the complete UBD gene in the strains we studied. This resulted in the identification of two major coding region haplotypes of UBD that we designate the “A” haplotype (LEW.1A, LEW.1WR1) and the “L” haplotype (LEW, WF, LEW.1W) (Fig. 4). The two UBD haplotypes contain extensive SNPs in the two exons and single UBD intron (Table 3), and there is a simple-sequence length polymorphism upstream of the 5′ end of UBD. However, only two UBD SNPs result in nonsynonymous changes. One substitution of histidine for arginine in the NH₂-terminal domain at amino acid 9 (SNP no. 18) (Table 3) is conservative in that other mammalian UBD sequences include either amino acid at this residue (online appendix Table 2). Interestingly, the LEW.1WR1 and LEW.1A strains have another SNP (no. 22) (Table 3) that encodes an amino acid change in the COOH-terminal of UBD protein (changing an asparagine to a histidine at residue 145 [N145H]). This residue is not seen in any other mammalian UBD sequence available for comparison (online appendix Table 2). Indeed, the peptide sequence containing amino acid 145N in the rat UBD-W haplotypes is conserved in rat, mouse, C. elegans, and S. cerevisiae; human and porcine UBD have glycine in at this position.

In a pilot study designed to quantify UBD expression during the course of diabetes onset, we took advantage of the fact that diabetes can be induced in 100% of LEW.1WR1 rats by chronic treatment with poly I:C alone (22). LEW.1WR1 rats were injected with poly I:C (1 μg/g three times weekly), and, as expected (22), 100% (12/12) developed hyperglycemia between days 12 and 14. As would be predicted by the fact that they do not have an RT1u class II haplotype (28), no identically treated LEW control rats developed diabetes. In a separate experiment, LEW.1WR1 and LEW rats were either untreated or treated with poly I:C three times weekly as described above. Spleens and pancreatic lymph nodes (PLNs) were collected from one to six LEW.1WR1 and LEW rats on day 0 (before any poly I:C) and on days 3, 5, 8, 10, and 12 after the first injection of poly I:C. The level of UBD mRNA as a percentage of a control gene product, GAPDH, was assessed by quantitative real-time RT-PCR.

UBD mRNA expression in PLN cells and spleen was higher in LEW.1WR1 rats than in LEW rats at every time point studied (Fig. 5 A). In the case of spleen cells, UBD mRNA levels were lower overall than in PLN cells, but they increased progressively during treatment with poly I:C. In the PLN cells, levels did not increase appreciably during the course of poly I:C injections. Our observation that UBD mRNA expression was higher in LEW.1WR1 than in LEW rats at every time point including on day 0 suggests that there is a genetic predisposition to higher UBD levels in LEW.1WR1 rat lymphoid tissues.

To lend support to this hypothesis, we next excluded the possibility that the substantially lower UBD mRNA expression in LEW rat tissues was due to the absence of IFN-γ or TNF-α, known inducers of UBD (42). Simultaneous testing of the cDNA from the same rat spleen samples showed comparable levels of IFN-γ and TNF-α transcripts in both LEW and LEW.1WR1 samples (Fig. 5 B). Thus, the allelic coding region polymorphisms seen in the LEW.1WR1 and LEW rat UBD genes, and not differences in UBD-inducing cytokines, are associated with statistically significant differences in UBD gene transcript levels. In addition, TNF-α is thus unlikely to be a strong candidate gene for Iddm37.

These data shed new light on the genetics of autoimmune disease in the rat models of human type 1 diabetes. They further strengthen the already strong case for Iddm14 as a susceptibility locus. In addition, they identify a new locus, Iddm37, which plays a role in susceptibility to disease that is triggered by the diabetogenic KRV parvovirus in LEW.1WR1 rats.

Insulitis and diabetes both segregate with Iddm14 in the LEW.1WR1 × WF intercross, as they are already known to do in BBDR × WF crosses (25). This observation makes Iddm14 one of the most consistent and powerful diabetes susceptibility loci observed in analyses in which WF or F344 rat strains are used as the nondiabetic partner strain (43,44). Strictly speaking, of course, we cannot claim that the specific susceptibility gene or genes associated with the Iddm14 region in BB and LEW.1WR1 rats are identical until they have been cloned, but the peak marker of this susceptibility locus in both strains is the same. Candidates for Iddm14 include the T-cell receptor β (Tcrb-V) chain variable region gene family, especially TcrbV13. Interestingly, LEW.1WR1 rats have the same Tcrbv13 allele as BBDR rats (43).

The QTL most clearly responsible for enhancing diabetes penetrance among rats that had both permissive Iddm14 alleles and insulitis in the KRV-exposed BBDR × WF progeny was Iddm20 on chromosome 17 (25); however, this locus appears not to play a role in the susceptibility of LEW.1WR1 × WF progeny. The chromosome 17 Iddm20 QTL is relatively weak, and studies in WF.chr17-DR congenic rats have found that Iddm20 is recessive in the first three backcross generations (J.P.M., E.P.B., unpublished data). LEW.1WR1 × WF progeny require a locus on chromosome 20 for full diabetes penetrance in Iddm14-permissive animals. The frequency of diabetes in Iddm37 heterozygous (LEW.1WR1 × WF) F2 progeny exceeded that of both Iddm37 L/L homozygotes and Iddm37 W/W homozygotes in this intercross (Table 2). We therefore expect that Iddm37 will be genetically dominant in congenic rats bearing LEW.1WR1-derived alleles at Iddm14 and Iddm37.

Iddm37 represents a new and interesting QTL in rats because it is syntenic with known diabetes-associated regions in both mouse and human genomes. In mice, three H2-linked QTLs were observed in a study of NOD.B6 congenics (45), including Idd24, which maps between Lta (35.3 MB) and D17Mit105 (41.4 MB) on chromosome 17. The mouse Idd24 QTL has not yet been confirmed, but the interval contains the mouse UBD gene at 37.3 Mb.

Iddm37 is also syntenic with the high-risk locus identified in a scan of HLA-linked SNPs inherited by descent (IBD) in siblings of diabetic children in the DAISY study (41). By eliminating genes that have shared alleles in the linkage disequilibrium–limited HLA region, and would thus have no differential effect on diabetes in this cohort, only two genes remained as candidates in the supported interval, UBD and MAS1L. Studies have revealed that siblings who have inherited shared UBD/MAS1L alleles IBD with the diabetic proband have a 65% likelihood of developing type 1 diabetes; the likelihood in HLA-matched but not IBD-matched siblings was only 15% (40,41). Of two candidate loci, only UBD is syntenic with the Iddm37 QTL in rats, as rat MAS1L is on chromosome 1.

UBD is a member of the family of ubiquitin-like proteins. It has several unusual properties including two ubiquitin-like domains. In most genera, these domains are conserved and direct repeats of one another. In mammals, they are homologous but nonidentical. UBD has functions that include directing proteins to the proteasome or to the aggresome (46) using E1 (47), E2, and E3 molecules that are different from the usual ubiquitin pathway members. In addition, UBD is inducible in the setting of inflammation. IFN-γ and TNF-α are cytokines known to increase the expression of UBD (42). The finding that UBD gene expression is upregulated in LEW.1WR1 rats after stimulation with poly I:C, and not in LEW rats, was somewhat unexpected because LEW rats are known to be highly sensitive to the cytokine-inducing effect of poly I:C (48). Direct measurement of TNF-α and IFN-γ transcripts showed that these cytokines were present in both of the strains we studied, but only in LEW.1WR1 was UBD upregulated. UBD upregulation in LEW.1WR1 rats may therefore reflect a different response to the stimulant at the level of the UBD gene. It is formally possible that upregulation of UBD in LEW.1WR1 rats exposed to poly I:C is a result rather than a cause of the inflammation that will ensue in the diabetes-prone strains, but the significantly higher levels of UBD transcripts present in LEW.1WR1 tissues before poly I:C injection and before hyperglycemia argue against this possibility.

A more intriguing hypothesis is that the UBD allele from LEW.1WR1 is causally associated with the failure of tolerance leading to autoimmune diabetes in this rat in response to environmental stimulants such as Toll-like receptor ligation or viral infection. UBD plays an important role in dendritic cell maturation (49,50) and thus has a potential role in generating autoreactive T-cells that recognize self-peptides. It may also control key signaling molecules in T-cells (51). In future studies, it will be of interest to analyze dendritic cell and T-cell functionality in rat strains that carry each of the UBD alleles that we have identified. Similarly, it will be of interest to determine diabetes susceptibility in other RT1u rat strains as a function of UBD genotype.

In conclusion, we have formally confirmed the role of Iddm14 in diabetes susceptibility in a third rat strain, with LEW.1WR1 joining the previously described BBDP and BBDR strains. In addition, we describe a new locus, Iddm37, that is important for KRV-triggered diabetes in the LEW.1WR1 rat, which can now be used to model an important human syntenic QTL identified in the DAISY cohort. Together with other recent reports of new genetic loci in both spontaneous (19) and induced (25) diabetes, our data highlight the growing relevance of rat models to the study of human type 1 diabetes.

This work was supported in part by grants R43DK-60374 (to D.L.Gu.), DK49106 (to D.L.Gr., J.P.M., and E.P.B.), DK25306 (to J.P.M.), U01 AI073871 (to D.L.Gr.), and center grant DK32520 from the National Institutes of Health and by grants 7-08-RA-106 (to J.P.M.) and 7-06-RA-14 (to E.P.B.) from the American Diabetes Association.

No potential conflicts of interest relevant to this article were reported.

The designation Iddm37 has been submitted to rat genome database (RGD) and assigned RGD ID 2305926.

We acknowledge the excellent technical support of Jean Leif and Elaine Norowski on this project.