A 212-kb Region on Chromosome 6q25 Containing the TAB2 Gene Is Associated With Susceptibility to Type 1 Diabetes

We typed DNA from 478 families, the majority having two affected siblings with type 1 diabetes. The 256 American families were from the Human Biological Data Interchange (Philadelphia, PA), whereas the remaining 222 English families were from the British Diabetes Association’s Warren Repository (London, U.K.). All families were Caucasian, with the proband diagnosed with type 1 diabetes and having an age of onset of ≤18 years. Differences were not detected in the frequencies of DNA polymorphisms in the families from the Human Biological Data Interchange and British Diabetes Association’s Warren Repository, and data were pooled for subsequent data analyses.

SNP genotyping was done using a three-step primer extension assay (17). First, a defined region was amplified using 30 cycles of PCR consisting of 1 min at 94°C (denaturation), 1 min at 55°C (annealing), and 1 min at 72°C (extension). In the next step, deoxynucleoside triphosphates and primers remaining from the PCR were removed by enzymatic digestion using shrimp alkaline phosphatase and exonuclease I (both from Amersham Pharmacia, Piscataway, NJ). In the final step, a single base extension (SBE) reaction using an oligonucleotide primer that resides directly adjacent to the nucleotide in question was used. The SBE reaction contained the extension primer, all four dideoxy deoxynucleoside triphosphates (Amersham Pharmacia), Taq polymerase (PGC Scientifics, Gaithersburg, MD), and the purified PCR product. The SBE assay was done using 49 cycles of denaturation at 96°C for 30 s, annealing at 55°C for 30 s, and extension at 60°C for 30 s. After the SBE primer was extended with the nucleotide analog, the resulting products were analyzed using ion-pair, reverse-phase high-performance liquid chromatography with the Wave System (Transgenomic, Omaha, NE). Locations of SNPs and flanking sequences can be found in the Human Genome Database (http://www.ncbi.nlm.nih.gov/). PCR and SBE primers are shown in Table 1. Markers 47, 57, and 244 were insertions/deletions of two or three base pairs and were detected on the Wave System by size. All markers tested contained only two alleles. Nucleotide sequence analysis was conducted at our core sequencing facility at Baylor College of Medicine.

Haplotypes were determined by visual inspection of alleles in the parents and offspring, assuming a lack of recombination between the closely linked loci. The transmission disequilibrium test (TDT) was used to access linkage disequilibrium (LD) between common marker alleles and disease (18). The TDT test statistic is a χ² (1 degree of freedom) statistic that tests deviation of transmission from the expected 50% transmission of an allele from heterozygous parents to offspring. However, in cases where linkage is already known, the analysis of multiple affected children invalidates the χ² statistic with respect to LD. To address this problem, we analyzed the TDT statistic in the probands only. Markers were also examined in unaffected siblings, but these were randomly distributed, and the data are not shown. Unaffected siblings were used to show that the TDT statistic was not due to a general segregation distortion to all siblings. To determine familial association, we also used the Tsp statistic, which is an extension of TDT that allows for multiple affected children (19). Tsp examines whether a heterozygous parent transmits the same or different marker alleles to each of their affected children (sibpairs) and provides a valid χ² test for LD in the presence of linkage. Tsp is generally a more powerful statistic than TDT using the proband only (19). The D′ value for LD was also examined (20). Values range from 1 (complete disequilibrium) to 0 (complete equilibrium) to −1 (alleles never found on the same haplotype).

We have taken a candidate gene approach and selected TAB2 because it is located on chromosome 6q25 (IDDM5), which is involved in NF-κB regulation (12). Figure 1 shows a 300-kb segment of genomic DNA containing the physical map of TAB2, which is encoded by seven exons, spanning 94 kb. Two additional gene sequences were identified within this stretch of DNA: LOC340152, a predicted gene with unknown function that is located adjacent to the 3′ end of TAB2, and SMT3, which has homology to SUMO (small ubiquitin-related modifier) (21) and is located within intron 6 of TAB2 (Fig. 1). The nucleotide sequence of the coding regions of each gene (13 exons total) were determined in 12 unrelated type 1 diabetic individuals. In addition, randomly selected regions located within introns and flanking regions were tested for sequence variation using the Wave System. In total, 18 DNA polymorphisms were identified, confirmed by nucleotide sequence analysis, and typed in our families (Table 2). All but three of the DNA polymorphisms (Table 2) were independently found by the Human Genome Project.

The 14 markers between positions 87 and 244 were in strong LD (D′ > 0.92) and were present on eight haplotypes; however, four haplotypes were relatively rare (haplotypes E–H) (Table 3). In contrast, markers at positions 47, 57, 269, and 300 were randomly associated with all markers tested between 87 and 244 (D′ < 0.16). The 87–244 markers thus form a haplotype block, with the outer boundaries of the block currently defined by markers 57 and 269 (a 212-kb stretch of DNA). We used these haplotypes to examine linkage of the TAB2 region with IDDM5. In our affected sibpair families having heterozygous parents, the affected sibpairs shared one and zero haplotypes in 312 and 260 cases, respectively (54.6%, χ² = 4.7, logarithm of odds = 1.03). Previously, in a study of 767 families (11), a logarithm of odds score of 1.96 was obtained for IDDM5. These results indicate that the TAB2 region defined here and the IDDM5 region previously defined are contributing to the overall diabetes susceptibility in families, even though the linkage is relatively weak.

In contrast, in our familial association studies, 10 of the 18 markers were associated with diabetes susceptibility as determined by both TDT analysis of the probands (TDT, P = 0.05–0.006) (Table 2) and Tsp analysis of multiple affected siblings (Tsp, P = 1.0 × 10⁻² to 4.0 × 10⁻⁴) (Table 2). Because of LD, many of these associations are not independent, as the markers on haplotypes A and C are positively associated and the markers on haplotype D are negatively associated with type 1 diabetes susceptibility (comparison of Tables 2 and 3). The frequencies of haplotypes E–H are too rare to evaluate the significance of diabetes associations. Normally when typing multiple markers and/or haplotypes, a statistical correction for multiple tests is warranted. However, in this study, strong LD exists within the TAB2 gene (D′ > 0.92), and the majority of the markers and haplotypes were associated with the disease. Thus, P values were not corrected for multiple comparisons in this study because any correction would be very small and not very meaningful.

No SNPs leading to amino acid substitutions were detected in the coding region of the TAB2 gene in 24 alleles that we sequenced or are described in the Human Genome Project (http://www.ncbi.nlm.nih.gov/). We did detect polymorphisms in the TAB2 5′ (244-GGC⁺¹, six copies of the nucleotide triplet versus five copies) and 3′ untranslated regions (152-C) (Table 2). It remains to be determined if the described polymorphisms in the TAB2 5′ and 3′ untranslated regions are involved in mRNA regulation or if other polymorphisms in introns or flanking regions modulate TAB2 gene expression. In addition to TAB2, we identified LOC340152, which is a predicted gene with currently unknown function, and SMT3, which is homologous to SUMO (21) (Fig. 1). Nucleotide sequence analysis of these genes identified one DNA polymorphism in exon 2 of LOC340152 (position 87) and one in the SUMO gene (position 161) (Fig. 1 and Table 2). The SNPs in LOC340152 and SMT3 were associated with diabetes susceptibility (Table 2) and resulted in the predicted amino acid substitutions of Pro to Leu and Met to Val, respectively. Whether these associations are secondary to those in LD with TAB2 or if they are biologically significant will require studies that genetics alone cannot answer.

We report a strong familial association of genetic markers with type 1 diabetes susceptibility within our candidate TAB2 gene on chromosome 6q25 (strongest association, P = 4.0 × 10⁻⁴). Diabetes association maps within a 212-kb region. Although, TAB2 was selected as our candidate gene and covers 94 kb of the 212 kb, we have not formally proven that TAB2 is biologically relevant to the pathogenesis of the type 1 diabetes susceptibility. Indeed, a proline-to-leucine amino substitution in LOC340152 and a methionine-to-valine substitution in SUMO were also associated with the diabetes susceptibility. It is possible that more than one type 1 diabetes susceptibility gene lies within this region. Indeed, both TAB2 and SMT3 (SUMO) may be involved in NF-κB activation (12,21). Furthermore, TAB2 is normally membrane bound and migrates to the cytoplasm after cytokine stimulation (12). This process is consistent with a sumoylation step (21), and TAB2 indeed appears to have the requisite sumoylation sequence (12). Therefore, both the TAB2 and SUMO genes may influence NF-κB activation and apoptosis in pancreatic β-cells, either directly by abnormal expression or indirectly through sumoylation with a variant SUMO polymorphism.

This work was supported in part by grants from the Juvenile Diabetes Foundation and the Harry B. and Aileen B. Gordon Foundation.

We thank Dr. Stan Lilleberg of Transgenomic for assistance in setting up the SBE assay.