K R Owen

Meta-Analysis and a Large Association Study Confirm a Role for Calpain-10 Variation in Type 2 Diabetes Susceptibility

To the Editor: Variation in the calpain-10 gene (CAPN10 [MIM 605286]) was recently linked and associated with type 2 diabetes mellitus (T2DM) susceptibility (Horikawa et al. 2000). The initial linkage of T2DM to chromosome 2 was found in a population of Mexican Americans from Starr County, Texas (Hanis et al. 1996). Specific combinations of three intronic variants, designated “SNP-43,” “SNP-19,” and “SNP-63,” that capture most of the haplotype diversity at CAPN10 were associated with a three-fold increased risk of T2DM in this population and could account for the observed linkage (Horikawa et al. 2000). Subsequent association and linkage studies of these three polymorphisms in other populations have produced conflicting results, with association being observed in some populations (Baier et al. 2000 [Pima Indian]; Cassell et al. 2002 [South Indian]; Garant et al. 2002 [African American]; Malecki et al. 2002 [Polish]; Orho-Melander et al. 2002 [Finnish/Botnia]), but not others (Evans et al. 2001 [British]; Hegele et al. 2001 [Oji-Cree Indians]; Tsai et al. 2001 [Samoan]; Xiang et al. 2001 [Chinese]; Daimon et al. 2002 [Japanese]; Elbein et al. 2002 [whites from Utah]; Fingerlin et al. 2002 [Finnish]; Rasmussen et al. 2002 [Danish and Swedish]; Horikawa et al. 2003 [Japanese]). We previously reported that another variant, SNP-44 (designated “CAPN10-g4841T→C”; minor allele frequency 16%), located in intron 3 and 11 bp from SNP-43, was independently associated with T2DM in whites from the United Kingdom (Evans et al. 2001). Further studies have provided tentative support for a role of SNP-44 in T2DM and related traits: associations with polycystic ovary syndrome (Gonzalez et al. 2002) and with measures of oral glucose tolerance (Wang et al. 2002; Tschritter et al. 2003) have been reported. Functional studies suggest that SNP-44 is located in an enhancer element and might affect CAPN10 expression (Horikawa et al. 2000). Also, in the U.K., German, Japanese, and South Indian populations, SNP-44 is in perfect linkage disequilibrium (r2=1) with a missense mutation Thr504Ala (SNP-110) and two polymorphisms in the 5′-UTR (SNP-134 and SNP-135) (Evans et al. 2001; Cassell et al. 2002; Y. Horikawa and P. E. Schwarz, unpublished data). To assess the association of SNP-44 with T2DM more comprehensively, we performed a meta-analysis of all published SNP-44/T2DM association study data. To identify all relevant published studies, we searched PubMed using the keywords “calpain 10,” “diabetes,” “44,” “SNP 44,” “CAPN10,” and “type 2,” in different combinations. When necessary, authors were contacted to obtain exact genotype numbers, so that precise odds ratios (ORs) from each study could be calculated. Our search identified 10 published case/control studies, consisting of 3,303 subjects. The studies were spread across a number of ethnic groups: British (three studies, Evans et al. 2001); Chinese (Wang et al. 2002); Japanese (Daimon et al. 2002; Horikawa et al. 2003); Finnish/Botnia (two studies, Orho-Melander et al. 2002); South Indian (Cassell et al. 2002); and Mexican American (Horikawa et al. 2000). The frequency of the T2DM-associated SNP-44 C allele (allele 2) ranged from 6% in Mexican Americans to 25% in the Botnia I control population. There was no evidence for OR heterogeneity (Q test P=.27), and, although these studies are only a small sample from the many existing T2DM genetic resources, a funnel-plot analysis (Egger et al. 1997) suggested an absence of publication bias (P=.44). A Mantel-Haenszel meta-analysis of these studies showed that the C allele was associated with increased risk of T2DM (OR 1.17 [1.02–1.34], P=.02). Three transmission/disequilibrium tests (TDT) had been performed (Evans et al. 2001; Cassell et al. 2002; Orho-Melander et al. 2002). The combined TDT results demonstrated that the C allele was significantly overtransmitted (117 transmitted vs. 77 not transmitted, P=.004) from heterozygous parents to diabetic offspring. Although this result cannot be considered independent replication, as proband data was included in the case/control meta-analysis from two of the TDT studies (Evans et al. 2001; Cassell et al. 2002), it provides evidence that the association is not due to population stratification. Of the 10 studies in the meta-analysis, only 1 reported a significant (P<.05) association (Evans et al. 2001). However, these studies were small and the mean power to detect an OR of 1.17 at P<.05 was ∼11% (range 5%–14%). In the context of genetic association studies, which test many polymorphisms in numerous candidate genes, a P value of .02 can only be considered evidence suggestive of a real association. We therefore genotyped SNP-44 in an additional 4,213 subjects: 3,274 white European subjects from four case/control studies (one British, two German, and one Czech); 691 Japanese subjects from two case/control studies; and 248 Mexican (mestizo) subjects from Mexico City and Orizaba City from one case/control study. Overall, this provided 2,056 subjects with T2DM and 2,157 controls, and a power of ∼80% to detect an OR of 1.17. Clinical details of the study subjects are presented in table 1; further details are available as supplementary information from the authors. All studies were approved by the relevant ethics committee, and all subjects gave their informed consent. Table 1 Clinical Characteristics of Subjects in Study[Note] When all the studies were combined, there was no evidence for between-studies OR heterogeneity (Q test P=.23); a Mantel-Haenszel fixed-effects model was therefore used for subsequent analysis. Meta-analysis of the new studies gave an OR for the SNP-44 C allele of 1.18 (1.04–1.34), P=.01 (fig. 1). A combined meta-analysis of all previously published data and our new data gave an OR of 1.17 (1.07–1.29), P=.0007. All study populations were in Hardy-Weinberg equilibrium except the T2DM cohort of Horikawa et al. 2003 (P=.005) and the control population of the third Japanese study (P=.02). Although these deviations may be due to random fluctuation and multiple-hypothesis testing, they contributed a large amount to heterogeneity (27% of the Q statistic); excluding these studies, the SNP-44 C allele OR for the new studies was 1.23 (1.07–1.40), P=.003; the overall OR was 1.19 (1.08–1.31), P=.0005. This OR is of similar magnitude to that of E23K (Gloyn et al. 2003; Love-Gregory et al. 2003; Nielsen et al. 2003) and Pro12Ala (Altshuler et al. 2000), the other common variants confirmed as T2DM-susceptibility polymorphisms. An OR of 1.17 is low and may help explain why there is little evidence for linkage of the CAPN10 region to T2DM in most populations. The haplotypes responsible for the CAPN10 linkage seen in the Mexican American population were associated with a higher T2DM OR (∼3.0) and were more likely to be detected by linkage analysis (Horikawa et al. 2000). These haplotypes are less common in other populations. Figure 1 Mantel-Haenszel OR meta-analysis plot (fixed effects) for SNP-44 association with T2DM. Point estimates and 95% CLs for each previously published, new, and combined case/control study. SNP-44 is in perfect linkage disequilibrium (r2=1) with the missense mutation, Thr504Ala, and two SNPs (SNP-134 and SNP-135) in the 5′-UTR and therefore may not be the causal variant. Further haplotype and functional analyses are required to confirm which of these polymorphisms contribute to T2DM susceptibility. In conclusion, our results have confirmed that a CAPN10 haplotype defined by the SNP-44 polymorphism predisposes to T2DM. Meta-analyses of published genetic associations, combined with large replication studies, are a powerful approach to detecting susceptibility variants in common disease.

An association analysis of the HLA gene region in latent autoimmune diabetes in adults

Aims/hypothesisPathophysiological similarities between latent autoimmune diabetes in adults (LADA) and type 1 diabetes indicate an overlap in genetic susceptibility. HLA-DRB1 and HLA-DQB1 are major susceptibility genes for type 1 diabetes but studies of these genes in LADA have been limited. Our aim was to define patterns of HLA-encoded susceptibility/protection in a large, well characterised LADA cohort, and to establish association with disease and age at diagnosis.Materials and methodsPatients with LADA (n = 387, including 211 patients from the UK Prospective Diabetes Study) and non-diabetic control subjects (n = 327) were of British/Irish European origin. The HLA-DRB1 and -DQB1 genes were genotyped by sequence-specific PCR.ResultsAs in type 1 diabetes mellitus, DRB1*0301_DQB1*0201 (odds ratio [OR] = 3.08, 95% CI 2.32–4.12, p = 1.2 × 10−16) and DRB1*0401_DQB1*0302 (OR = 2.57, 95% CI 1.80–3.73, p = 4.5 × 10−8) were the main susceptibility haplotypes in LADA, and DRB1*1501_DQB1*0602 was protective (OR = 0.21, 95% CI 0.13–0.34, p = 4.2 × 10−13). Differential susceptibility was conferred by DR4 subtypes: DRB1*0401 was predisposing (OR = 1.79, 95% CI 1.35–2.38, p = 2.7 × 10−5) whereas DRB1*0403 was protective (OR = 0.37, 95% CI 0.13–0.97, p = 0.033). The highest-risk genotypes were DRB1*0301/DRB1*0401 and DQB1*0201/DQB1*0302 (OR = 5.14, 95% CI 2.68–10.69, p = 1.3 × 10−8; and OR = 6.88, 95% CI 3.54–14.68, p = 1.2 × 10−11, respectively). These genotypes and those containing DRB1*0401 and DQB1*0302 associated with a younger age at diagnosis in LADA, whereas genotypes containing DRB1*1501 and DQB1*0602 associated with an older age at diagnosis.Conclusions/interpretationPatterns of susceptibility at the HLA-DRB1 and HLA-DQB1 loci in LADA are similar to those reported for type 1 diabetes, supporting the hypothesis that autoimmune diabetes occurring in adults is an age-related extension of the pathophysiological process presenting as childhood-onset type 1 diabetes.

Large-scale studies of the association between variation at the TNF/LTA locus and susceptibility to type 2 diabetes

Aims/hypothesisThe proinflammatory cytokine TNF-α has been implicated in the pathogenesis of insulin resistance and type 2 diabetes, and variation in the gene encoding TNF-α (TNF) has shown inconsistent associations with susceptibility to both conditions. Additionally, the coding non-synonymous variant T60N in the neighbouring LTA gene has been reported to be associated with type 2 diabetes. The present study aimed to obtain a robust assessment of the role of variation in the tightly linked TNF/LTA region in diabetes susceptibility by genotyping TNF and LTA variants in large case-control resources.Materials and methodsThe G-308A and G-238A TNF promoter variants and the LTA T60N polymorphism were genotyped in two UK case samples that were ascertained for positive family history and/or early onset of type 2 diabetes (combined n=858) and in 1,257 ethnically matched controls.ResultsThere were no significant associations between the T60N, G-308A or G-238A genotype and type 2 diabetes in the combined analysis (exact Cochran–Mantel–Haenszel statistic for ordered genotypes for T60N, p=0.69; for G-308A, p=0.51; for G-238A, p=0.16).Conclusions/interpretationThe present study, one of the largest association analyses yet reported at this locus, provides no evidence that the specific TNF or LTA variants examined influence susceptibility to type 2 diabetes. More comprehensive studies of the TNF/LTA locus in substantially larger sample sets are required to establish whether genome sequence variation at this locus truly influences susceptibility to type 2 diabetes.

Prevalence of monogenic diabetes in young adults: a community-based, cross-sectional study in Oxfordshire, UK

To the Editor: Recently, Shields and colleagues used cases ascertained from diagnostic genetic-testing referrals to estimate a minimum population prevalence for monogenic beta cell disorders (i.e. MODY) in the UK population [1]. They estimated a minimum prevalence of 108 cases of MODY per million, with half due to mutations in HNF1AMODY. The study included all cases of MODY diagnosed through the UK testing laboratory over a 12 year period, but had the limitations of not being community-based and only included individuals who had been referred for molecular testing through secondary care. The identification of monogenic diabetes allows targeted management and screening of at-risk relatives, so accurate prevalence figures are necessary for service provision. Despite this, information in this area is limited. This is largely because of the high costs of molecular testing, meaning that large-scale re-sequencing efforts to establish prevalence have not been done. Two studies performed prior to the availability of molecular testing estimated the prevalence of MODY in Germany at 0.14% and 1.8% of diabetes cases, respectively [2, 3]. A paediatric survey from Germany and Austria of 40,757 individuals diagnosed with diabetes before 20 years of age identified that 0.65% had MODY confirmed by genetic testing [4]. The only previous population-based study (from Norway) reported a prevalence of 63 cases of HNF1A-MODYper million but did not investigate other genes [5]. We aimed to estimate the prevalence of diabetes subtypes in young adults diagnosed with diabetes up to age 45 years by a combination of community survey and subsequent molecular investigation [6]. The study was performed in 12 general practitioner (GP) surgeries in Oxfordshire, UK, with 118,927 individuals registered. Between March 2005 and June 2006, practice computer databases were used to identify individuals diagnosed with diabetes. The study was approved by the Oxfordshire Local Research Ethics Committee. Individuals participating in the study gave written consent. J. Kropff :M. I. McCarthy :K. R. Owen (*) Oxford Centre for Diabetes, Endocrinology and Metabolism (OCDEM), University of Oxford, Churchill Hospital, Oxford OX3 7LJ, UK e-mail: katharine.owen@drl.ox.ac.uk

Examining the relationships between the Pro12Ala variant in PPARG and Type 2 diabetes-related traits in UK samples.

Aims  The Pro12Ala polymorphism in the PPARG gene alters amino acid sequence and has shown consistent association with susceptibility to Type 2 diabetes in several populations. The present study makes use of large, well‐characterized case‐control resources to enhance understanding of this susceptibility effect by examining related traits, such as body mass index (BMI), waist–hip ratio and age at diagnosis.

Analysis of the contribution to type 2 diabetes susceptibility of sequence variation in the gene encoding stearoyl-CoA desaturase, a key regulator of lipid and carbohydrate metabolism

Aims/hypothesisStearoyl-CoA desaturase (SCD) is emerging as a key regulator of lipid and carbohydrate metabolism. Scd-null mice display a beneficial metabolic phenotype characterised by resistance to obesity, diabetes and hyperlipidaemia. The human homologue, SCD, maps to a region of chromosome 10 linked to type 2 diabetes, and SCD activity correlates with insulin sensitivity. Given this strong positional and biological candidacy, the present study sought to establish whether sequence variation in SCD influences susceptibility to type 2 diabetes and related traits.MethodsThe SCD gene was resequenced in 23 diabetic subjects. Six variants within coding and adjacent sequence, including a non-synonymous SNP in exon 5 (M224L), were selected for genotyping in a primary set of 608 diabetic subjects and 600 control subjects.ResultsThere was no association (at the allele, genotype or haplotype level) with type 2 diabetes, although genotype frequencies at the +14301 A>C SNP in the 3′ untranslated region showed borderline association (p~0.06) when evidence for linkage was taken into account. However, replication studies (350 young-onset diabetic patients; 747 controls) failed to confirm any relationship with diabetes for this variant. No significant associations were seen for diabetes-related traits including BMI and waist-to-hip ratio.Conclusions/interpretationThe present study, the first reported analysis of this gene, indicates that the SCD variants typed do not explain chromosome-10-encoded susceptibility to type 2 diabetes. Although this study provided no evidence that SCD sequence variation influences diabetes susceptibility or related traits, SCD remains a major target for pharmaceutical and/or environmental manipulation.

Apolipoprotein M can discriminate HNF1A-MODY from Type 1 diabetes

Missed diagnosis of maturity‐onset diabetes of the young (MODY) has led to an interest in biomarkers that enable efficient prioritization of patients for definitive molecular testing. Apolipoprotein M (apoM) was suggested as a biomarker for hepatocyte nuclear factor 1 alpha (HNF1A)‐MODY because of its reduced expression in Hnf1a–/– mice. However, subsequent human studies examining apoM as a biomarker have yielded conflicting results. We aimed to evaluate apoM as a biomarker for HNF1A‐MODY using a highly specific and sensitive ELISA.

Sequencing PDX1 (insulin promoter factor 1) in 1788 UK individuals found 5% had a low frequency coding variant, but these variants are not associated with Type 2 diabetes.

Diabet. Med. 28, 681–684 (2011)

Role of the D76N polymorphism of insulin promoter factor-1 in predisposing to Type 2 diabetes

To the Editor: Insulin promoter factor-1 (IPF-1, PDX-1) is a transcription factor and homeodomain protein required for development of the pancreas and initiation and maintenance of the insulin-secreting phenotype of the beta cell [1]. Reduced levels of IPF-1/PDX-1 are associated with pancreatic agenesis and hyperglycaemia in rodent models. A dominant negative mutation in man causes pancreatic agenesis [2] and MODY [3]. Therefore IPF-1/PDX-1 is an excellent candidate gene for Type 2 diabetes. Coding variants in IPF-1/PDX-1 were found in 3 to 5% of young-onset familial Type 2 diabetes compared to 0.5 to 1% of control subjects with normoglycaemia in the UK [4], France [5] and Sweden [6]. However, these variants were rare in lateonset Type 2 diabetes [7, 8, 9]. In vitro studies of the naturally occurring IPF-1/PDX-1 mutations have shown diminished activity in most [4, 5, 6] but not all studies [7]. Physiological assessment of non-diabetic carriers of IPF-1/PDX-1 mutations showed higher glucose levels during an OGTT than those found in control subjects [4, 6] and a lower insulin response to a glucose load [5]. Most studies in Type 2 diabetes have been small (~200 cases) and the variants identified individually are rare and not found in all populations. In most European populations studied, only one polymorphism, D76N (CAG→AAG), causing an amino acid change of aspartic acid to asparagine, has been reported in both diabetic subjects and control subjects. It was also the most frequent variant found in the initial UK and French studies [4, 5]. The aim of our study was to assess whether D76N predisposed to young-onset familial Type 2 diabetes in a large UK case-control study. All subjects were white UK citizens. The Type 2 diabetic subjects had been diagnosed with diabetes by World Health Organization criteria or were on treatment for diabetes. Genetic and autoimmune subtypes were excluded. The Type 2 diabetic subjects (n=860) were recruited from three sources: (i) youngonset (≥18 and ≤45 years at diagnosis) subjects (the youngonset Type 2 diabetes cohort described in [10]); (ii) probands of parent–offspring trios; and (iii) probands of affected sib-pairs. All subjects were therefore selected either for early-onset or for familial diabetes consistent with an increased genetic susceptibility. Population control subjects (n=1275) were recruited from two sources: (i) non-diabetic parents from a consecutive birth cohort (the Exeter Family Study) and (ii) a population control group of blood donors without known diabetes from the European Collection of Cell Cultures. Genotyping was either performed on the ABI 7000 Taqman (Applied Biosystems, Warrington, UK) or by a previously described RFLP method [4]. Taqman reactions were carried out in a total volume of 25 μl containing 12.5 μl Taqman universal PCR master mix, 0.5 μl of probes 1 and 2, and 1.125 μl of forward and reverse primers (F primer CAGCCCCCCGGACATCT, R primer CGGGAGGTGTGGTGAAGGT, probe 1 CTCGCCGACGACC, probe 2 CTCGCCGACAACC). The default thermal cycling cycle was used for all experiments followed by an allelic discrimination step. Data were analysed and scored with the Prism 7000 Sequence Detection System (Applied Biosystems). Samples that could not be scored after three repeats were excluded from the final analysis. All samples scored as containing the rare allele were confirmed by direct sequencing. Using a control frequency of 1% for D76N (as in [4]), we had 80% power at p=0.05 to detect an Odds Ratio (OR) of 2.15. Statistical analysis was performed using Stats direct v2.2. Chi square analysis was used to calculate the OR and 95% CI. Recent studies have shown that the evidence for or against genetic associations in common phenotypes needs to be assessed over many studies. We therefore performed a meta-analysis of the current plus previously published studies [4, 5, 6, 7, 8, 9] examining the D76N variant in Type 2 diabetes, and calculated a Mantel-Haenszel pooled odds ratio. Using STATXACT with calculation of EXACT estimations did not make any difference to the result, suggesting there were no significant departures from asymptotic expectation. One study [8], where the rare allele was not detected in diabetic or control subjects, was not included in the meta-analysis, as it is impossible to calculate an odds ratio in this situation. All genotypes were in Hardy-Weinberg equilibrium. The rate of failed and/or unscoreable samples was 3%. Table 1 shows the genotype frequency in the case and control subjects, as well as the case subjects diagnosed at 45 years of age or earlier. There were no rare homozygotes (AA). Of case and control subjects 1.7% and 1.6% respectively were heterozygous (GA) for the rare allele, giving an OR of 1.1 (CI: 0.52 to 2.1, p=0.86). In the subjects diagnosed at 45 years or earlier, 2.5% were heterozygous, giving an OR of 1.5 (CI: 0.7 to 3.2, p=0.26). Meta-analysis showed that a Mantel-Haenszel pooled estimate of the OR was 1.7 (CI: 1.0 to 2.8, p=0.07). When only study populations with a mean age of onset of 45 years or earlier were included [4, 5, 6], the Mantel-Haenszel pooled OR was 2.5 (CI: 1.4 to 4.6, p=0.004). We compared the phenotype of the subjects with the D76N variant to wild-type, but no difference was observed in the age of onset, parental history of diabetes, BMI or current insulin treatment. Diabetologia (2004) 47:957–958

Atypical phenotype associated with reported GCK exon 10 deletions: Clinical judgement is needed alongside appropriate genetic investigations.

Maturity‐onset diabetes of the young (MODY) caused by heterozygous mutations in the glucokinase (GCK) gene typically presents with lifelong, stable, mild fasting hyperglycaemia. With the exception of pregnancy, patients with GCK‐MODY usually do not require pharmacological therapy. We report two unrelated patients whose initial genetic test results indicated a deletion of GCK exon 10, but whose clinical phenotypes were not typical of GCK‐MODY.