Introduction

Recently, GWA studies revealed novel single nucleotide polymorphisms (SNPs) strongly associated with type 2 diabetes (T2D) [1–5]. In a French population, the well known rs7903146 TCF7L2 polymorphism ranked first for its effect on T2D prevalence followed by four new loci: SLC30A8, HHEX, LOC387761 and EXT2 [1]. Subsequently, GWA studies in Finnish, English, Icelandic and Danish cohorts emphasized the role of CDKAL1, CDKN2A/2B and IGFBP2 on T2D and confirmed the effect of TCF7L2, SLC30A8 and HHEX [2–5]. In the French GWA scan, additional SNPs located in MMP26, LDLR, KCTD12, CAMTA1, NGN, CXCR4 and LOC646279 were also among the 15 first signals in joint stage I and stage II analyses, but their current status is still uncertain. Although large-scale case-control studies are very sensitive in detecting genetic effects, they are not good models for evaluating the true contribution of disease-associated genes in unselected general populations [6].

The objective of our study was to assess whether the risk alleles or risk genotypes (N = 22) of 14 novel loci from GWA analyses modulate quantitative traits related to glucose homeostasis (fasting glucose, fasting insulin, HOMA-B and HOMA-IR) in the 9-year prospective D.E.S.I.R. cohort (N = 4,707), a French general middle-aged population [7, 8]. As this cohort is limited in statistical power to analyze genetic variants with modest effects on T2D risk, we also examined their association with hyperglycemia (HG; individuals with T2D or impaired fasting glucose) incidence, as we recently did for TCF7L2 [9].

Materials and methods

Study population

Clinical characteristics of D.E.S.I.R. participants are reported in Supplementary Table 1. A French cohort of 2,576 men and 2,636 women from a general population (aged between 30 and 65 years at inclusion) participated in the D.E.S.I.R. longitudinal study and were clinically and biologically evaluated at inclusion and at 3, 6 and 9-year visits [8]. Participants were recruited from volunteers insured by the French social security system, which offers periodic health examinations free of charge. They came from 10 health examination centers in the western-central part of France. All participants signed an informed consent. The protocol was approved by the Ethics Committee for the Protection of Subjects for Biomedical Research of Bicêtre Hospital. Among the 5,212 subjects of the D.E.S.I.R. cohort, 3,786 individuals were followed during the entire study.

Because the ethnic origin could not be legally documented at the beginning of the D.E.S.I.R. study, we estimated the proportion of subjects having non-European ancestry from a subgroup of 654 subjects selected in the D.E.S.I.R. cohort, as previously described [1]. This subgroup was genotyped for 328 SNPs which were spaced by at least 5 Mb and highly differentiated among individuals from different continents (Fst > 0.2 based on the Perlegen dataset) [29]. Analysis using the STRUCTURE software identified only two individuals of non-European ancestry in a total of 654 individuals. From this analysis, the proportion of subjects having non-European ancestry was estimated to 0.30% in the D.E.S.I.R. cohort. Additionally, all individuals born outside France were excluded from our analysis.

Three classes of glycemic status were defined according to the 1997 American Diabetes Association criteria [30]: NG, defined as fasting glucose < 6.1 mmol/l; IFG, defined as fasting plasma glucose (FPG) between 6.1 and 6.99 mmol/l; and type 2 diabetes, defined as fasting plasma glucose ≥ 7.0 mmol/l and/or treatment by glucose lowering agents. Hyperglycemia (HG) was defined by IFG or T2D.

The study of baseline glucose homeostasis was on the 4,283 NG participants; the study of incident glucose status was on 186 incident T2D participants and 508 incident HG participants.

Measurements

Venous blood samples were collected in the morning after subjects had fasted for 12 h. Fasting plasma glucose was assayed by the glucose oxidase method applied to fluoro-oxalated plasma, using a Technicon RA1000 (Bayer, Puteaux, France) or a Kone Automate (Evry, France). To adjust for differences between and within laboratories over the four triennial examinations, glucose concentrations were standardized by age and sex with respect to a reference population of 211,427 individuals examined in the IRSA Health Examination Centres and assayed in the IRSA laboratory. Fasting serum insulin was measured by an enzymo-immunoassay with IMX (Abbott, Rungis, France) (48). Insulin secretion was assessed by calculating the HOMA-B index, defined as (fasting insulin × 20) / (fasting glucose − 35), and peripheral insulin resistance was estimated by HOMA-IR, defined as (fasting insulin × fasting glucose) / 22.5. Other biological and metabolic parameters were assessed as previously described [8].

SNP genotyping

All participants were genotyped using an allelic discrimination assay-by-design TaqMan method on ABI 7900 (Applied Biosystems). There was a 97–99% genotyping success rate (Table 1). The genotyping error rate was assessed by randomly re-genotyping 384 participants. No difference was found with the first genotyping results. The genotypic distributions of all polymorphisms were in Hardy–Weinberg equilibrium.

Statistical analysis

Polymorphism effects on quantitative traits at baseline were calculated by linear regression models adjusted for age, gender and BMI. All metabolic traits were log-transformed for linear regression analysis in order to normalize their distribution. Hazard ratios for HG and T2D incidence were assessed using the Cox model, adjusted for age, gender and BMI. For each study on quantitative traits or T2D/HG incidence, a conservative Bonferroni correction (multiplication by the number of SNPs) was applied to the P values for multiple comparisons. All P values are two-sided. SPSS (version 14.0.2) and R statistics (version 2.5.1) software were used for general statistics.

Results

Metabolic effects on glucose homeostasis

Clinical characteristics at baseline of the 4,707 participants of the D.E.S.I.R. (Data from an Epidemiological Study on the Insulin Resistance syndrome) study are reported in Supplementary Table 1. Physiological effects of the studied SNPs on quantitative phenotypes related to glucose homeostasis were analyzed in all participants normoglycemic (NG) at baseline (N = 4,283). The genotypic distribution of each SNP at baseline is presented in Table 1. Fasting plasma glucose levels were higher in participants carrying risk alleles of SLC30A8 (rs13266634 P = 0.0003), NGN3 (rs10823406 P = 0.01) and MMP26 (rs2499953 P = 0.04) with no impact on fasting insulin (Table 2). Conversely, an association with lower fasting plasma insulin level was found in subjects carrying risk alleles of CDKAL1 (rs7756992 P = 0.003, rs10946398 P = 0.04 and rs7754840 P = 0.05) with no detectable effect on fasting glucose. In individuals carrying one of the studied risk alleles (rs564398) of CDKN2A/2B, trends for associations were observed with higher fasting glucose (P = 0.07) and lower fasting insulin (P = 0.06).

Table 1 Genotypic distribution of the 22 SNPs studied according to glycemic status at baseline in 4,707 individuals: the D.E.S.I.R Study
Full size table
Table 2 Effects of the 22 SNPs on fasting glucose and fasting insulin in 4,283 individuals normoglycemic at baseline: the D.E.S.I.R. Study
Full size table

In a second analysis, effects on HOMA indices were also assessed as markers of insulin secretion (HOMA-B) and insulin resistance (HOMA-IR) (Table 3). We found lower HOMA-B value in individuals with the SLC30A8 risk allele (rs13266634 P = 0.0005) and a trend towards such a decrease in individuals with the HHEX risk alleles (rs7923837 P = 0.06 and rs1111875 P = 0.07). Furthermore, lower HOMA-IR value (P = 0.04) were detected in participants carrying one of the CDKN2A/2B (rs564398) risk alleles. The SLC30A8 genetic variant was the only SNP remaining significant after conservative Bonferroni corrections for its effects on fasting glucose and HOMA-B.

Table 3 Effects of the 22 SNPs on HOMA indices in 4,283 individuals' normoglycemic at baseline: the D.E.S.I.R. Study
Full size table

Association with type 2 diabetes and hyperglycemia incidence

The genotypic distribution of incident T2D and HG cases as well as individuals remaining NG after 9 years of follow-up is presented in Table 4; in order to increase the statistical power, participants with impaired fasting glucose (IFG) or T2D were studied together as a single group of HG subjects [9, 10]. For each SNP, the best fitting genetic model was selected from our previous GWA results in French individuals [1]. In this regard, every polymorphism was analyzed using an additive genetic model except for EXT2 (dominant model) and KCTD12 (recessive model). For CDKAL1, CDKN2A/2B, IGFBP2, HHEX and EXT2, we studied more than one SNP because they have all been reported as associated with T2D in previous studies [1–5].

Table 4 Genotypic distribution of the 22 SNPs according to incident T2D (N = 186) or HG (N = 508): the D.E.S.I.R. Study
Full size table

Only trends or nominal associations with T2D incidence were found with hazard ratios (HR) of 2.03 [1.00–4.11] (rs2499953 P = 0.05) for MMP26, 1.33 [1.02–1.73] (rs10823406 P = 0.03) for NGN3 and 1.25 [0.99–1.58] (rs13266634 P = 0.06) for SLC30A8 (Fig. 1). Two other genetic variants were also associated with trends towards a higher incidence of HG with HRs of 1.15 [0.98–1.35] (rs10811661 P = 0.08) for CDKN2A/2B and 1.13 [0.99–1.29] (rs1111875 P = 0.06) for HHEX (Fig. 1). As expected, no result remained significant after conservative Bonferroni correction.

Fig. 1
figure 1

Hazards ratios for hyperglycemia (HG) and type 2 diabetes (T2D) incidence in a French general population. Hazard ratios (HR) were assessed by Cox survival analysis (adjusted for BMI, age, and gender) during the 9 years of follow-up. All genetic models were additive except for EXT2 (d: dominant) and KCTD12 (r: recessive)

Full size image

Discussion

In the present study, we confirm that individuals from a general population of European origin carrying CDKAL1 or SLC30A8 T2D risk alleles had lower fasting plasma insulin level and lower basal insulin secretion, respectively, compared to non-carriers [5, 11, 12]. However, higher glycemia and T2D incidence were only detected in carriers of the SLC30A8 risk SNP. These data are in agreement with the known function of the zinc transporter ZnT8 (SLC30A8 protein) which is specifically expressed in pancreatic endocrine cells and may participate in insulin secretion [13, 14]. CDKAL1 is also highly expressed in human islets [2] and shares homology with a known inhibitor of CDK5 activation which is implicated in beta-cell function [15]. Our results further support a physiological impact of these genes on the beta-cell function.

Interestingly, the T2D risk alleles of NGN3 or MMP26 were found to be associated with a higher glycemia and T2D incidence in the D.E.S.I.R. cohort, confirming what was previously found in the French case/control GWA study [1]. Despite the fact that no obvious signal was found in these loci by other GWA scans [2–5, 16–19], further meta-analyses and systematic GWA phase 2 studies will be necessary to fully evaluate their contribution, if any, to T2D. These two genes are indeed good biological candidates: the transcription factor NGN3 is essential for the differentiation of pancreatic progenitor cells into endocrine cells [20] and MMP26 is also a target of the Wnt signaling pathway [21].

More modestly, HHEX and CDKN2A/2B SNPs showed a tendency to be associated with HG as well as with lower insulin secretion and lower insulin levels, respectively, as found previously [22, 23]. HHEX, known to be a target of the Wnt signaling pathway, is essential for pancreatic development [24]. The nominal effect of CDKN2A/2B rs564398 on lower insulin resistance estimated by HOMA-IR may be an artifact due to its association with both lower insulin level and greater glucose level. It was suggested that CDKN2A could be a possible biological candidate for T2D [25], as its over-expression in rodents causes a decrease in islet proliferation [26]. More recently, a new large antisense non-coding RNA named ANRIL was characterized [27] and constitutes another good positional candidate.

Surprisingly, we were unable to find any effect of IGFBP2, EXT2, LOC646279, KCTD12, LDLR, CAMTA1, LOC38776 or CXCR4 SNPs. This may be due to their modest effects on T2D incidence (or absence of true contribution) in this French non-selected general population and/or may be explained by limitations in the statistical power of the D.E.S.I.R study. Further investigations in larger prospective cohorts are therefore necessary to address this issue.

It was suggested that decreased insulin secretion, but not insulin resistance determines future glucose intolerance in non-obese subjects [28]. Similarly our data suggest that the best hits identified by the GWA studies modulate beta-cell function but not insulin resistance, at least in the D.E.S.I.R. cohort, mostly composed of non-obese participants. Additional analyses of these genetic variants in obese cohorts may therefore put new functional defect forward.