Epidemiological studies suggested a relationship between blood glucose levels and risk for cardiovascular disease and all-cause mortality in both diabetic and non-diabetic subjects [2
]. A systematic overview and meta-analysis of non-diabetic cohort studies revealed a graded relationship between glucose levels and cardiovascular risk, extending below the diabetic threshold [6
]. A U-shaped relation between glucose levels and mortality was reported [7
]; that is, both high (diabetes or impaired glucose levels) and low [<70 or 80 mg/dl (3.9 or 4.4 mmol/l)] FPG levels were associated with increased mortality risk from cardiovascular disease and all causes. The DECODE study [8
] also reported increased mortality risk with both high and very low glucose concentrations. These observations highlight the importance of investigating not only the genetic determinants of T2D but also those involved in regulation of FPG levels in non-diabetic individuals.
Two recent GWAS reported an association of small effect size (~1%) between the G6PC2
/rs560887 SNP and reduced FPG levels in the general European population [13
]. The first GWAS [13
] evaluated 392,935 SNPs in a discovery sample of 654 ‘normoglycemic’ French subjects and was able to replicate the strongest association observed with FPG levels (G6PC2
/rs560887; p = 4 × 10–7
adjusted for age, gender and BMI under the additive model) in additional French and Finnish samples. The second GWAS [14
] investigated the SNPs from 2 genome-wide scans in a total of 5,088 ‘non-diabetic’ individuals from Finland and Sardinia (Italy) and found a strong association between FPG levels and a SNP (rs563694) that resides between G6PC2
(p = 3.5 × 10–7
). The same study [14
] was able to confirm the significant association of rs563694 in 6 of 7 follow-up samples of mixed European descent. Subsequent genotyping/analyses of the G6PC2
region by the second GWAS [14
] also revealed that G6PC2
/rs560887 [in high linkage disequilibrium (LD) with rs563694; D’
= 0.99, r2
= 0.84] showed the strongest association with fasting glucose (meta p = 2.8 × 10–10
). Interestingly, rs560887 did not show evidence of association with T2D in either GWAS [13
], leading the authors to hypothesize that either the genetic determinants of FPG levels in physiological states are different than those associated with T2D risk or the effect of G6PC2
variation on T2D susceptibility is too small to be easily detected.
Our results in ‘non-diabetic’ NHWs support the findings of recent GWAS with a significant association (adjusted p = 0.002 under the additive model) of G6PC2/rs560887 with FPG levels in the same direction (minor allele A associated with reduced FPG levels). Consistent with GWAS, the effect size of rs560887 was small and accounted for only ~1.4% of inter-individual variation in FPG levels. Although it was not statistically significant, we observed a similar trend of association in ‘non-diabetic’ Hispanics (adjusted p = 0.076 under the dominant model), which was more pronounced and modestly significant when analyzed only in ‘normoglycemic’ subjects after excluding non-diabetic individuals with impaired fasting glucose levels (adjusted p = 0.036 under the dominant model). It is important to note that we had less power to detect the association in Hispanics due to the smaller sample size (410 Hispanics vs. 623 NHWs) and lower MAF of rs560887 (17.5% in Hispanics vs. 33% in NHWs). In addition, there may be LD structure differences between the two ethnic groups; such that the true causative genetic variant(s) may be in high LD with rs560887 in NHWs but in low LD in Hispanics. Consistent with recent GWAS and mouse studies, we did not detect significant association with BMI, fasting insulin levels, or lipid measurements in either population, suggesting that the observed association with FPG levels is independent of these traits. GWAS also concluded that the effect of the G6PC2 variation on FPG levels might be independent from adiposity-induced insulin resistance.
The rs560887 SNP is located close to the acceptor splice-site in the third intron of G6PC2
and predicted to affect pre-mRNA splicing by alternative usage of exon 4 [13
]. Due to tissue-specific expression of various G6PC2
splice forms [21
], the pancreatic RNA samples from individuals with different rs560887 genotypes (which are not readily available) would be necessary to directly test the latter hypothesis.
In conclusion, our results independently confirm the robust association of G6PC2
/rs560887 (or another tightly linked, yet undiscovered variant) with FPG levels in non-diabetic NHWs and further support the relevance of the glucose phosphorylation pathway to blood glucose homeostasis in the general population. Although additional GWAS and meta-analysis increasingly support the association of G6PC2
/rs560887 with FPG levels in individuals of European descent [22
], our results further extend this observation to the Hispanic population by providing evidence for association that warrants replication in larger samples due to lower MAF of rs560887 in this population. The small effect sizes of G6PC2
/rs560887 and other relevant genetic variants in GCKR
] suggest that the majority of genetic factors that regulate FPG levels remain to be identified. The absence (or very low MAF) of this SNP in our black African sample supports the presence of additional common variants (in G6PC2
and/or other genes) that contribute to blood glucose regulation in individuals from various ethnic origins. In fact, the rs560887 variant is also rare among Asians and a recent study demonstrated that another G6PC2
variant common in Chinese was associated with FPG levels in this population [23
]. Resequencing of G6PC2
in selected individuals with upper and lower range of FPG levels may help to determine the extent to which common and rare variation of G6PC2
explains the variation in FPG levels and whether rs560887 is the most important G6PC2-
linked genetic determinant of glucoregulation.