A Polymorphism Within the G6PC2 Gene Is Associated with Fasting Plasma Glucose Levels

See allHide authors and affiliations

Science  23 May 2008:
Vol. 320, Issue 5879, pp. 1085-1088
DOI: 10.1126/science.1156849


Several studies have shown that healthy individuals with fasting plasma glucose (FPG) levels at the high end of the normal range have an increased risk of mortality. To identify genetic determinants that contribute to interindividual variation in FPG, we tested 392,935 single-nucleotide polymorphisms (SNPs) in 654 normoglycemic participants for association with FPG, and we replicated the most strongly associated SNP (rs560887, P = 4 × 10–7) in 9353 participants. SNP rs560887 maps to intron 3 of the G6PC2 gene, which encodes glucose-6-phosphatase catalytic subunit–related protein (also known as IGRP), a protein selectively expressed in pancreatic islets. This SNP was associated with FPG (linear regression coefficient β = –0.06 millimoles per liter per A allele, combined P = 4 × 10–23) and with pancreatic β cell function (Homa-B model, combined P = 3 × 10–13) in three populations; however, it was not associated with type 2 diabetes risk. We speculate that G6PC2 regulates FPG by modulating the set point for glucose-stimulated insulin secretion in pancreatic β cells.

Recent innovations in genotyping technology have led to the identification of genetic variants associated with increased risk of type 2 diabetes (T2D) (14). However, the genetic factors contributing to interindividual variation in blood glucose levels in the general population are largely unknown. In addition to affecting metabolic health, these variants may also affect coronary heart disease (CHD) risk, for which a linear correlation between FPG and CHD mortality has been described for both diabetic and nondiabetic individuals (5, 6). Here, we describe the results of a genome-wide association (GWA) study in which we aimed to identify genetic variants involved in glucose homeostasis in the general population.

We analyzed FPG as a quantitative trait in 654 normoglycemic (NG) nonobese individuals by using Illumina Infinium Human1 (Illumina Incorporated, San Diego, CA) and Hap300 BeadArray (Illumina Incorporated, San Diego, CA) (table S1) (1, 7). The strongest association signal was observed with SNP rs560887 [P = 4 × 10–7 adjusted for age, gender, and body mass index (BMI) under the additive model]. This SNP is part of a 17-kb linkage disequilibrium (LD) block on chromosome 2 that encompasses the gene and the 3′ flanking region of the islet-specific glucose-6-phosphatase–related protein [IGRP, also known by the gene name G6PC2 (glucose-6-phosphatase catalytic unit 2)], a glycoprotein embedded in the endoplasmic reticulum (ER) membrane (8). Six SNPs within this LD block were genotyped in our GWA analysis. These included a nonsynonymous SNP (rs492594-L219V), which showed modest association with FPG [P = 0.04 for rs492594-L219V, which is in low LD with rs560887 at r2 = 0.24 in the Centre d'Etude du Polymorphisme Humaín (CEPH) families with Northern and Western European ancestry (CEU), according to HapMap data], and a SNP located in the 3′ flanking region of G6PC2 (rs563694), which showed stronger association (P = 2 × 10–5, r2 = 0.76 with rs560887) (table S2). We used linear regression analysis to adjust for the effect of rs560887 on all SNPs located within 100 kbp upstream and 100 kbp downstream that showed various levels of association with FPG (P < 0.05) and found no SNP that remained associated at significance level of 5% (table S2). Imputation of genotypes using the LD structure from HapMap data (CEU population) did not identify any SNP showing potentially stronger association with FPG (Fig. 1). On the basis of these findings, we tested the association of rs560887 with FPG by using a complementary subset (table S1) of the French DESIR (data from the Epidemiological Study on the Insulin Resistance Syndrome) population (DESIR NG stage 2, N = 3419) and replicated our original association with FPG (linear regression coefficient β = –0.06 mmol/l per A allele, P = 3 × 10–8) (Table 1).

Fig. 1.

Genomic context and association with FPG of the G6PC2 gene variant rs560887 on chromosome 2q. (Top) The genomic context of rs560887 that maps to intron 3 of G6PC2. (Middle) The association magnitude [-log10(P value)] with fasting plasma glucose of the SNPs from the Illumina 317 mapping array (Tag-SNPs) (dots) and the HapMap CEU data imputed SNPs (plus signs), not directly genotyped around (±100 kb) rs560887. (Bottom) The LD structure (defined by pairwise r2) between SNPs in the CEU population from HapMap phase II using the Haploview software.

Table 1.

Association of rs560887 with FPG levels in normoglycemic individuals from a GWA study and replication analyses. Association was tested in the additive model, adjusted for age, sex, and BMI. All individuals were normoglycemic (FPG < 6.1 mmol/l). Per A allele effect size was estimated from the regression coefficient β. Overall meta-analysis includes replication populations (DESIR NG, NFBC86, and obese children). DESIR controls are nonobese NG individuals.

View this table:

IGRP has been proposed to modulate the glycolytic pathway and eventually glucose-stimulated insulin secretion by dephosphorylating glucose-6-phosphate generated by glucokinase, the β cell glucose sensor (9). G6pc2 knockout mice display decreased FPG and normal insulin sensitivity (10). In the DESIR NG cohort, the rs560887 A allele [minor allele frequency (MAF) = 0.30] associated with decreased FPG is also associated with a decreased fraction of erythrocyte glycosylated hemoglobin (HbA1c) (β = –0.04, P = 8 × 10–5). This reflects a better long-term glucose control, with increased basal insulin secretion as assessed by the homeostasis model index of pancreatic β cell function Homa-B (β = 5.46, P = 2×10–5), a measure derived from physiological data of FPG and fasting insulin levels (7).

To validate these findings in different European populations, we genotyped young NG individuals from the Northern Finland Birth Cohort 86 (7) (NFBC86, N = 5073) and confirmed the association between rs560887 and decreased FPG (β = –0.07 mmol/l per A allele, P = 6 × 10–16) and increased Homa-B index (β = 3.79; P = 6 × 10–10). In contrast, no association was found between rs560887 and fasting insulin in either French (P = 0.5) or Finnish (P = 0.1) cohorts. Similarly, neither cohort showed strong association between rs560887 and BMI (DESIR P = 0.1, NFBC86 P = 0.04). Adjustment for BMI did not affect the association with FPG, suggesting that the observed effect on FPG is independent of obesity-induced insulin resistance. To test this hypothesis, we studied 861 severely obese NG French children (mean age 11 ± 3 years) (7). In this cohort, we still observed association between rs560887 and decreased FPG (β = –0.07 mmol/l per A allele, P = 0.004) and increased Homa-B index (β = 9.13, P = 0.01). As in the French DESIR and Finish cohorts, rs560887 did not associate with fasting insulin (P = 0.30) or BMI (P = 0.60) in obese children.

rs560887 is located in the third intron of G6PC2, 26 base pairs from its boundary with the fourth exon. Splice site modeling shows that the rs560887 A allele may have a lower splicing acceptor score than the G allele ( (11); however, further investigation is required to define the functional consequences of the observed association signal, which may be caused by rs560887 or an unidentified SNP in high LD with rs560887, and G6PC2 splicing or gene expression.

Longitudinal analysis of the DESIR cohort showed that carriers of the rs560887 A allele had a decreased risk of developing mild hyperglycemia (defined as a FPG ≥ 6.1 mmol/l) over a 9-year follow-up period [hazard ratio = 0.83, 95% confidence interval (CI) 0.72 to 0.96, P = 0.009]. Surprisingly, rs560887, or any SNP within its 17-kb LD block, did not associate with T2D risk in our GWA study (1) or other published T2D case-control studies ( or To confirm this, we analyzed 2972 additional T2D cases (table S1). With the 4073 DESIR NG individuals as controls, we observed no significant association with T2D risk (P = 0.6, odds ratio (OR) = 0.98 95% CI 0.91 to 1.05). We note that our case-control analysis achieved statistical power of at least 99.95% to detect a significant effect (MAF = 0.30, OR ≥ 1.20, N = 7045). Previous reports have shown that the SNP most strongly associated with T2D identified so far (rs7903146, located in TCF7L2) does not associate with FPG in European general populations (12, 13), suggesting that the genetic determinants that regulate FPG in physiological states may be different from those that increase T2D risk. IGRP is homologous to two other ER proteins of the glucose-6-phosphatase family that catalyze the conversion of glucose-6-phosphate into glucose, a key enzymatic step for both gluconeogenesis and glycogenolysis (8). In contrast to the hepatocyte-specific member (G6pc1) that is highly expressed in gluconeogenic tissues but not in pancreas (fig. S1, top), the expression of G6pc2 is restricted to islets and pancreatic β cell line MIN6 (fig. S1, middle), as shown in humans (14). Because expression of G6pc2 mRNA is restricted to β cells in the mouse, as indicated in previous studies (14) and as confirmed in our own experiments (fig. S1), rs560887 may play an important role in the early pathogenesis of this disease without contributing to overall T2D risk. For example, the rs560887 common allele (G) might alter the glucostat set point and up-regulate FPG, but the development of overt T2D may require additional compromise of the pancreatic β cell function.

Whether IGRP has the same catalytic activity as the hepatic glucose-6-phosphatase encoded by the gene G6PC1 is a matter of debate (14). When IGRP dephosphorylates glucose-6-phosphate, it opposes the action of the β cell glucose sensor glucokinase and hence moderates the glycolytic pathway and eventually glucose-stimulated insulin secretion (9). We hypothesize that G6PC2 is part of the same glucose phosphorylation pathway that includes the glucokinase gene (GCK), responsible for the monogenic form of early onset diabetes, MODY2 (15), and familial hypoglycemia-hyperinsulinemia phenotypes (16). This pathway also includes the glucokinase regulatory protein gene (GCKR). Thus, IGRP may antagonize the glucokinase activity in β cells, in a similar manner to the GCKR action as a potent inhibitor of glucokinase activity in hepatocytes (17). Recently, SNPs in both GCK (rs1799884–-30G/A) (18) and GCKR (rs1260326-P446L) (2, 19, 20) were found to modulate FPG in European population-based cohorts. To study the interaction of rs560887 with rs1260326-P446L and rs1799884–-30G/A, we analyzed their combined effect in the French DESIR cohort and observed an additive effect (interaction P = 0.56) of the three SNPs on FPG (P = 5 × 10–21) (Fig. 2). Individuals carrying more than four alleles associated with low FPG (10.7% of the DESIR population) showed a mean 0.24 mmol/l (4.5%) decrease of FPG compared with subjects carrying only one allele (Fig. 2).

Fig. 2.

Combined effects of G6PC2 rs560887, GCKR rs1260326-P446L, and GCK rs1799884-30G variants on FPG levels. Data are presented as mean [CI 95%], and the P value is for the β coefficient in the linear regression model (including age, sex, and BMI as covariates) of FPG levels on the number of alleles associated with low FPG (G6PC2 rs560887A allele frequency = 0.30; GCKR rs1260326-P446 allele frequency = 0.44; GCK rs1799884-30G allele frequency = 0.82).

G6PC2 rs560887 is associated with significant reductions in both FPG and glycated hemoglobin A1C (HbA1c). There is strong evidence suggesting that even small changes in blood glucose can have an impact on cardiovascular morbidity and mortality (21). Meta-analyses have shown that there is a linear relationship, with no threshold, between FPG and CHD risk (6), with individuals whose FPG is lower than 5.5 mmol/l having the lowest mortality rate (5, 22). In contrast, individuals whose FPG lies between 5.6 and 6 mmol/l, although still considered NG, showed a 30% increased risk of heart failure and a 60% increased risk of nephropathy (23). In Asians, a slight reduction in FPG from 5.5 mmol/l to 5 mmol/l is associated with a 25% reduction in CHD risk (24). Even tight blood glucose control, reflected by HbA1c between 5.1 and 5.6% (normal value < 6.1%), is associated with a 30% higher mortality compared with a lower HbA1c (25). In this context, Marz et al. previously reported that the glucokinase rs1799884/-30 A allele was associated with T2D and increased risk for coronary artery disease in both diabetics and nondiabetics (26). We therefore conclude that the three SNPs identified in our study, which additively modulate FPG by 0.24 mM in the general population, are likely to have a nonnegligible impact on human health.

We found a strong association between SNP rs560887, located in the G6PC2 gene, and FPG, and this association was replicated in three independent populations (N = 9353). The magnitude of the rs560887 effect on FPG is similar in each cohort [explaining ∼1% of FPG variance in each population (heterogeneity test P = 0.95)]. Each copy of the A allele results in a 0.06 mmol/l decrease (overall meta-analysis P = 4 × 10–23). Our findings underscore the role of the glucose phosphorylation pathway in glucose homeostasis in the general population.

Supporting Online Material

Materials and Methods

Figs. S1 and S2

Tables S1 to S3


References and Notes

View Abstract

Navigate This Article