Introduction

In humans the concentration of fasting plasma glucose is subject to tight homeostatic control: a mechanism directed by both neural and humoral factors. A few genetic variants have been reported to influence plasma glucose levels in the general population including GCK rs1799884 [1, 2], GCKR rs780094 [3] and variants in the MTNR1B locus [4–6]; however, these variants explain only a minor part of the total inter-individual variation in plasma glucose levels.

In a meta-analysis of two genome-wide association studies totalling 5,088 non-diabetic individuals a significant association between the rs563694 A allele in the G6PC2/ABCB11 locus and elevated fasting plasma glucose was reported (p = 4 × 10−7). This finding was substantiated in a large-scale replication study involving 23,524 non-diabetic white individuals (p = 7 × 10−26) [7] showing an effect size of 0.065 mmol/l per A allele. A total of 5,734 individuals from the Danish population-based Inter99 cohort genotyped for rs563694 [7–9] was included in the study. Additional analyses of the locus harbouring rs563694 revealed an even stronger association of rs560887 (p = 3 × 10−10) and rs853789 (p = 1 × 10−9) with fasting plasma glucose levels in the initial sample of 5,088 non-diabetic individuals from Finland and Sardinia [7]. Another study including 9,353 French individuals demonstrated that rs560887 was strongly associated with fasting plasma glucose with an effect size of 0.06 mmol/l per G allele (p = 4 × 10−23) [10].

rs560887 is located within the third intron of the glucose-6-phosphatase catalytic subunit 2 gene (G6PC2) and is in strong linkage disequilibrium (LD) (r 2 > 0.8) with rs563694 located 10.9 kb away from rs560887 downstream of G6PC2 [7], and with rs853789 located in intron 19 of the ATP-binding cassette, sub-family B, member 11 gene (ABCB11) 38.3 kb from rs560887 and 27.4 kb from rs563694 (r 2 > 0.8) [7].

Although the association of the G6PC2/ABCB11 variants with fasting plasma glucose has been demonstrated, more detailed physiological studies are lacking. The aim of the present study was to evaluate the impact of rs560887 on plasma glucose and serum insulin levels following oral or i.v. glucose loads and to measure hepatic glucose production both in the basal state and during hyperinsulinaemia. Furthermore, we examined the effect of the variants on the risk of having impaired fasting glycaemia (IFG), impaired glucose tolerance (IGT) and type 2 diabetes. In addition, potential associations with components of the metabolic syndrome were examined.

Methods

Participants

Details of the study populations are given in Electronic supplementary material (ESM) Table 1. Participants from the population-based Inter99 cohort (ClinicalTrials.gov ID no: NCT00289237) [8, 9] involving 4,407 glucose-tolerant individuals who were characterised by an OGTT were investigated for association between the rs560887 genotype and quantitative metabolic traits. Studies of insulin release after an i.v. glucose load (0.3 g glucose/kg body weight) were performed in 366 individuals with normal fasting glucose levels from a population-based sample of young, healthy Danes recruited at the Research Centre for Prevention and Health, Glostrup, Denmark [11] and in a sample of 196 glucose-tolerant relatives of Danish type 2 diabetic patients recruited at the Steno Diabetes Center, Gentofte, Denmark (SDC) [12]. Additionally, rs560887 was examined in a sample of 53 elderly and 106 young twins with normal glucose tolerance subjected to a 2 h hyperinsulinaemic–euglycaemic clamp (40 mU m−2 min−1) with measures of hepatic glucose production during the fasting state and upon insulin infusion [13].

The case–control studies of impaired glucose regulation involving rs560887 included unrelated individuals with isolated IFG recruited from the Inter99 cohort (n = 486) and at SDC (n = 22), isolated IGT (Inter99 n = 466, SDC n = 64), both IFG and IGT (Inter99 n = 199, SDC n = 16), type 2 diabetes (Inter99 n = 323, SDC n = 1,640), and glucose-tolerant control individuals (Inter99 n = 4,407, SDC n = 506). All control participants had normal fasting glycaemia and were glucose-tolerant following an OGTT. Diabetes, IFG and IGT were defined in accordance with World Health Organization 1999 criteria [14].

All participants from the Inter99 cohort (including patients with treated type 2 diabetes) involving a total of 5,899 middle-aged individuals were evaluated in a case–control study examining the association between the rs560887 genotype and the WHO-defined metabolic syndrome (n = 1,326) [14]. In the latter study, those having no components of the metabolic syndrome were considered as control individuals (n = 1,702). More detailed description of participants, biochemical assays and anthropometrical measurements are given in the ESM.

All participants were of Danish nationality and informed written consent was obtained from all individuals before participation. The studies were approved by the Ethical Committee of Copenhagen and were in accordance with the principles of the Helsinki Declaration II.

Biochemical and anthropometric measures

The plasma glucose and serum insulin concentrations were measured as described in the ESM.

Genotyping

Genotyping of rs560887 was performed using TaqMan allelic discrimination (KBioscience, Hoddesdon, UK). Genotype data were obtained in more than 97% of the DNA samples, and among 459 duplicate pairs no discrepancies were found. All genotype groups obeyed Hardy–Weinberg equilibrium.

Statistical analysis

Logistic regression analysis and Fisher’s exact test were applied to test for differences in genotype distribution or allele frequencies. A general linear model was applied to test quantitative variables for differences between genotype groups assuming an additive model. Assuming an allele frequency of the rs560887 A allele of 30% our study of quantitative traits in the Inter99 sample (n = 4,407) has a power of 58%, 93% and 99% to detect a genotype effect difference of 5%, 8% and 10% of an SD, respectively, considering an α of 5%. In the cohort of young, healthy Danes (n = 366) we have a power of 9%, 17% and 22% to detect a genotype effect difference of 5%, 8% and 10% of an SD. Power for the quantitative traits was estimated using simulations assuming an additive genetic model both for the simulation of the data and for testing the data using a linear model. The empirical variance of the observed traits was used to simulate phenotypes from a normal distribution so that variance across genotypes was drawn from the estimated variance. The statistical analyses were performed using RGui version 2.6.1 (available at www.r-project.org, accessed 28 May 2009). Analyses in the sample of non-diabetic relatives of Danish type 2 diabetic patients were performed applying a mixed model including the family relationship (coded as a family number) as a random factor, the genotype and sex as fixed factors, and age and BMI as covariates using the Statistical Package for Social Sciences (SPSS, Chicago, IL, USA) version 14.0. Analyses of hepatic glucose production in the sample of young and elderly twins were calculated using the SAS system version 8.2 (Cary, NC, USA) applying a mixed model including a random-effect term for twin pair membership and a fixed-effects term for zygosity. p < 0.05 was considered significant.

Results

The impact of rs560887 on OGTT-derived diabetes-related quantitative traits was evaluated in a population-based sample of glucose-tolerant individuals from the Inter99 cohort (n = 4,407) (Table 1). The rs560887 G allele was associated with elevated fasting plasma glucose (p = 2 × 10−14), as well as elevated levels of serum insulin (p = 1 × 10−4), 30 min after an oral glucose load and with increased AUC for serum insulin (p = 0.003) during an OGTT, applying an additive genetic model adjusting for age, sex and BMI. Furthermore, the rs560887 G allele was associated with an increased glucose-stimulated insulin release as assessed by the insulinogenic index (p = 3 × 10−6). The same genotype was not associated with 30 or 120 min plasma glucose levels following an OGTT. The insulin concentrations at each time point during the OGTT (0, 30 and 120 min) were additionally analysed adjusting for age, sex, BMI and the simultaneous or fasting plasma glucose concentration; however, the results remained the same when compared with the insulin data adjusted for age, sex and BMI alone (data not shown).

Table 1 Quantitative metabolic traits in the population-based Inter99 cohort including 4,407 glucose-tolerant middle-aged individuals stratified according to rs560887 genotype
Full size table

In a subsequent analysis of rs560887 in a population-based sample of 366 young, healthy glucose-tolerant Danes based on normal fasting glucose according to the WHO criteria [14], we found a tendency towards increased acute serum insulin response during an i.v. glucose load (p = 0.09) for carriers of the G allele (Table 2). These findings were substantiated in a study of 196 glucose-tolerant relatives of Danish type 2 diabetic patients. Here we observed a significant association of rs560887 G allele carriers with increased acute serum insulin response following an i.v. glucose load (p = 4 × 10−4; Table 2). The results remained unchanged when adjusting the acute serum insulin response for fasting plasma glucose.

Table 2 Acute serum insulin response in young, healthy Danes and non-diabetic relatives of Danish type 2 diabetic patients subjected to an IVGTT
Full size table

To further investigate the effect of rs560887 on hepatic glucose production we examined a group of 53 elderly and 106 young glucose-tolerant twins all subjected to a 2 h hyperinsulinaemic–euglycaemic clamp. Among the elderly twins, G allele carriers had a higher basal hepatic glucose production (p = 0.04), whereas no effect of genotype was seen on hepatic glucose production (p = 0.5) or insulin-stimulated glucose disposal rate (p = 1.0) during the clamp (Table 3). In the group of young twins (Table 3), and when combining data from young and elderly twins, we found no genotype effect on hepatic glucose production.

Table 3 Metabolic rates in young twins and elderly glucose-tolerant twins stratified according to rs560887 genotype
Full size table

The risk of having IFG was evaluated in a case–control study of 508 individuals with IFG and 4,913 glucose-tolerant control individuals. With an OR of 1.26 (95% CI 1.09–1.47) the rs560887 G allele associates with an increased risk of having isolated IFG (p = 0.002) (Table 4). On the contrary, the rs560887 G allele was not associated with the risk of having isolated IGT (OR 0.94, 95% CI 0.82–1.08, p = 0.4) or type 2 diabetes (OR 0.93, 95% CI 0.84–1.04, p = 0.2). However, the rs560887 G allele associates with an increased risk of having combined IFG and IGT (OR 1.40, 95% CI 1.11–1.76, p = 0.005) (Table 4).

Table 4 Genotype distribution and allele frequency of rs560887 among patients with type 2 diabetes, individuals with isolated IFG, isolated IGT or both IFG and IGT, and glucose-tolerant control individuals
Full size table

Potential associations of the rs560887 G allele with components of the WHO-defined metabolic syndrome were evaluated in a case–control study involving all participants from the Inter99 cohort (n = 5,881). We found no association of the rs560887 G allele with the risk of having either the metabolic syndrome (OR 1.01, 95% CI 0.89–1.15, p = 0.9) or any components of the metabolic syndrome (ESM Table 2).

Discussion

In the present study we show that even though the G allele of rs560887 associates with increased fasting plasma glucose, the G allele is not implicated in altered plasma glucose levels during an OGTT. Furthermore, the G allele associates with increased insulin release after both an oral glucose load and an i.v. glucose injection. Interestingly, the G allele shows nominal association with elevated basal hepatic glucose production in elderly twins. Also, we show that the variant associates with IFG, but not with IGT, type 2 diabetes, or any component of the 1999 WHO-defined metabolic syndrome.

The human G6PC2 encodes the islet-specific glucose-6-phosphatase-related peptide (IGRP). IGRP is known to be specifically expressed in pancreatic beta cells in humans [15], and although no catalytic activity has been reported for IGRP [15–19] it may be assumed that IGRP dephosphorylates glucose 6-phosphate to form glucose, opposing the action of glucokinase, the glucose sensor of the beta cell. Therefore, the effect of variants in the G6PC2/ABCB11 locus on fasting plasma glucose may, as suggested by previous reports [7, 10], imply that the rs560887 G allele results in an increased expression or higher functional activity of IGRP, hereby decreasing the glucose flux into the glycolytic pathway directing the pancreatic beta cell to a higher glucostatic set-point and thus upregulating the impact of the concentration of plasma glucose on insulin secretion in the fasting state [7, 10]. These findings and interpretations are in accord with observations in the G6pc2 knockout mice, where lack of G6pc2 results in a decreased fasting blood glucose concentration [20]. Moreover, a French functional study reported that the G allele of rs573225, a variant in high LD with rs560887, associates with higher promoter activity, suggesting a higher expression of G6PC2 and thereby IGRP [21]. If the rs560887 G allele or another variant in high LD increases the IGRP activity in the beta cell it would most likely lead to elevated plasma glucose levels during an OGTT as a result of a relatively decreased glucose-stimulated serum insulin release. However, we found normal plasma glucose levels and elevated serum insulin levels 30 min after oral and i.v. glucose loads. Thus, our data does not support a sole IGRP-mediated effect of rs560887 on plasma glucose levels.

rs560887 is in high LD with rs853787 and rs853789 located within ABCB11. ABCB11 encodes the bile salt export pump (BSEP), which is evolutionarily highly conserved and solely expressed in the liver. BSEP is involved in the ATP-dependent secretion of bile salts in the hepatocytes and the protein has been linked to inherited forms of cholestasis, resulting in diverse hepato-pathological phenotypes [22]. Variation in ABCB11 impairs bile acid secretion [23–25], and studies involving bile acid sequestrants have shown association between interruption of bile acid re-absorption and decreased levels of plasma glucose and improved insulin sensitivity, most probably as a result of lowered serum triacyglycerol levels [26]. Interestingly, although exploratory because of the low number of individuals in the cohort, it seems from the present study of elderly twins that the increase in fasting plasma glucose in carriers of the rs560887 G allele may be explained by an increase in glucose production from the liver. This finding could, however, not be replicated in the group of young twins and may in theory be explained by an age-dependency in the expression of this distinct phenotypic trait. If the elevated basal hepatic glucose production in G allele carriers is a true finding, it might imply that the raised insulin levels after both oral and i.v. glucose challenges are a compensatory beta cell response primed by the increased fasting plasma glucose concentrations. An interesting approach to further evaluate the ABCB11 locus would be to examine rs560887 in relation to concentrations of bile acid; however, these data are not available in the present study. Given the complex pattern of association we cannot exclude the possibility that two independent signals of association exist, which influence both the function of G6PC2 in the beta cell and ABCB11 in the liver. Functional studies and/or further fine mapping of the locus will be necessary to identify the actual causal variant and the exact mechanisms leading to elevated fasting plasma glucose levels.

Interestingly, even though the rs560887 G allele associates strongly with fasting plasma glucose and IFG it is not associated with IGT or overt type 2 diabetes in the present study. A recent large-scale study of loci affecting fasting plasma glucose levels reported a significant protective effect of the rs560887 G allele on type 2 diabetes (OR 0.93, 5% CI 0.89–0.97, p = 0.0017) [6] with a similar effect size to that observed in the present study. Combined, these studies strongly indicate that the rs560887 G allele associates with elevated fasting plasma glucose levels, and a decreased risk of type 2 diabetes, probably via a hyper-response of the beta cell to postprandial elevation in circulating glucose levels.

A progressive relationship between elevated plasma glucose levels and increased cardiovascular risk, which extends below the threshold chosen for diabetes, has been demonstrated [27]. We have previously shown that the GCK rs1799884 A allele associates with both fasting and post-OGTT plasma glucose levels and with different components of the metabolic syndrome [1]. However, we find no association of the rs560887 G allele with any components of the metabolic syndrome. This might be because of the lack of association with elevated glucose levels after an oral glucose load or the low number of individuals in the sub-analyses.

In conclusion, the present study reports evidence that the observed association of the common rs560887 G allele in the G6PC2/ABCB11 locus with increased fasting plasma glucose and IFG may be related to increased hepatic glucose production rather than a pancreatic beta cell dysfunction. Furthermore, even though the rs560887 G allele is strongly associated with elevated fasting plasma glucose and glucose-stimulated serum insulin release, the variant is not involved in the inter-individual variation of glucose levels after an oral glucose load or the risk of having the metabolic syndrome among middle-aged Danes.