Assessment of the Role of Common Genetic Variation in the Transient Neonatal Diabetes Mellitus (TNDM) Region in Type 2 Diabetes : A Comparative Genomic and Tagging Single Nucleotide Polymorphism Approach

Type 2 diabetes is a polygenic disorder, and progress in unraveling its underlying molecular genetics has so far been modest. However, two types of study have yielded valuable insights. First, >30 genome-wide scans for linkage to type 2 diabetes have now been performed, and these have yielded some chromosome regions that have been replicated in more than one scan (1). Second, there is increasing evidence that common variation in the genes associated with rare monogenic conditions is important for disease susceptibility in common polygenic disorders (1). We have shown that heterozygous activating mutations in KCNJ11, which encodes the Kir6.2 subunit of the ATP-sensitive K⁺ channel, are a common cause of permanent neonatal diabetes (2), while the common variant in this gene (E23K), with an allele frequency of ∼0.35, has been reproducibly associated with type 2 diabetes (3–6). This overlap between mono- and polygenic disease susceptibility is not restricted to neonatal diabetes genes but is also seen in the PPARG, HNF4A, and GCK genes (7,8,9,10).

A chromosomal region between 6q22 and 6q25 is associated with type 2 diabetes in several studies (11,12,13,14,15,16). In one study (13), evidence for linkage of diabetes to chromosome 6q was observed from paternally derived genes. Among the genes in this region are the two genes of which one or both are paternally overexpressed in 80% of cases of transient neonatal diabetes mellitus (TNDM) (17,18). These genes are the transcription factor zinc finger protein associated with cell cycle control and apoptosis (ZAC; also known as pleomorphic adenoma gene 1 [PLAGL1]), and HYMAI, which encodes an untranslated mRNA; both are located at chr6q24 (17,19). The primary defect in TNDM due to chromosome 6q abnormalities is β-cell dysfunction (20). This presents around birth with marked hyperglycemia, which then remits during infancy, allowing discontinuation of insulin treatment, only to relapse in late childhood or early adulthood (20). The role of common genetic variation in the TNDM region is not known but could potentially disrupt normal β-cell function and predispose to type 2 diabetes.

Since the TNDM locus lies within a chromosomal region associated with type 2 diabetes in several studies, it represents an excellent candidate gene for type 2 diabetes susceptibility. Therefore, we performed a large-scale case-control (n = 3,594) association study of the region. As the locus is imprinted, we also examined the parent of origin of transmitted alleles in a large family study (n = 1,654). Details of the cohorts studied are given in Table 1. Inclusion of other known types of diabetes in the type 2 diabetes probands were previously minimized through clinical, immunological, and genetic testing, as previously described (21).

Single nucleotide polymorphisms (SNPs) were extracted from GoldenPath (http://genome.ucsc.edu) and were also sought by resequencing the nine exons of ZAC, its splice junctions, putative promoter region, and noncoding regions of high homology (>75% homology over >100 bp) between human and mouse in 47 individuals. This gave >99% power to detect alleles of >5% frequency. In total, 26 SNPs were identified; this represented an average coverage of one SNP per 3.1 kb of genomic sequence (Fig. 1). The linkage disequilibrium pattern across the region was assessed using both r² and D′ (Fig. 2). We identified tag SNPs (tSNPs) using both pairwise and aggressive tagging approaches (22). The multimarker approach did not improve the tSNP efficiency; thus, SNPs were chosen based on pairwise r² statistics. We successfully typed 15 tSNPs in 1,988 case and 1,606 control subjects and in a type 2 diabetes family-association study (1,654 individuals from 536 nuclear families). The 15 tSNPs captured 88% of SNPs in our study with an r² > 0.7 and 92% of the 26 SNPs with an r² > 0.5. The results of our tSNP analysis for the case-control study are shown in Table 2 and supplementary Table 1 (online appendix [available at http://diabetes.diabetesjournals.org]). There was no association of any of the tSNPs with type 2 diabetes, nor were there significant deviations from expected transmission rates in our family-based study (Table 3 and supplementary Table 2). Only one of the SNPs (rs2076684) reached nominal significance (P = 0.05) in the family study. There was no association of any of the SNPs with diabetes when the family-based study was combined with the case-control study (data not shown).

Given the imprinted status of the TNDM locus, we used our family data to test the hypothesis that variants were overtransmitted by parent of origin. There was no evidence for a parent-of-origin effect (data not shown).

This study is well designed to have sufficient power to detect moderate diabetes susceptibility polymorphisms. The case-control component of our study was designed to have >80% power to detect ORs of 1.14 (minor allele frequency [MAF] 0.1 [0.50−1.24]) at P < 0.05. The coverage of the ZAC locus and surrounding sequence was good, with an average density of one SNP per 3 kb, and we have captured the majority of common variants with a mean r² > 0.7. Of the 26 SNPs in our study, 14 are also in HapMap 2. These 14 SNPs captured 58% (42 of 73) of the 73 HapMap SNPs (>5% MAF in Caucasians) with an r² > 0.7 and 86% (63 of 73) with an r² > 0.5. A comparison of our data with HapMap 2 is shown in supplementary Fig. 2. As with any large association study, we cannot exclude the possibility that multiple rare alleles (<5%) at this locus are associated with type 2 diabetes. We are aware that the control subjects used are a younger, population-based resource rather than age matched to our cases. As ∼5% of the control subjects will go on to develop diabetes, this would slightly reduce the power of our study compared with the use of the same number of age-matched control subjects. However, using the same case-control and family-based samples, associations have recently been shown between the KCNJ11, the K23 allele (3), the HNF4A P2 promoter haplotype (9), and the PPARG P12 allele (23) and type 2 diabetes, with ORs consistent with other large type 2 diabetes case-control studies and meta-analyses of multiple studies. Conversely, the large negative studies we have performed in HNF1A (24) also have not been found to be positive in similar-sized studies (25), suggesting that our results are representative.

We have the largest set, internationally, of type 2 diabetic nuclear families to assess parent-of-origin effects. These resources are difficult to obtain due to the late age of onset of most patients with type 2 diabetes (26). Even though our cohort is relatively large, we acknowledge that our parent-of-origin study was of limited power to rule out a moderate susceptibility variant, but we had power to detect ORs of 1.4 and 1.68 for MAF of 0.5 and 0.1, respectively.

In conclusion, using a comparative genomic and tSNP approach, we have comprehensively assessed the role of common genetic variation in susceptibility to type 2 diabetes. We have found no evidence in a large-scale study in a U.K. Caucasian population that common genetic variation at this locus influences type 2 diabetes susceptibility. TNDM is unusual among genetic disorders in that the disease is caused by an overabundance of a protein, rather than its absence, and occurs chiefly by epigenetic rather than genetic change. This study suggests that it is unlikely that a possible epigenetic difference at the TNDM locus is inherited from this locus. However, the degree of epigenetic variation at the TNDM locus in critical tissues for glucose homeostasis has not been investigated and thus cannot be excluded.