Common variants at the GCK, GCKR, G6PC2–ABCB11 and MTNR1B loci are associated with fasting glucose in two Asian populations
- First Online:
To test fasting glucose association at four loci recently identified or verified by genome-wide association (GWA) studies of European populations, we performed a replication study in two Asian populations.
We genotyped five common variants previously reported in Europeans: rs1799884 (GCK), rs780094 (GCKR), rs560887 (G6PC2–ABCB11) and both rs1387153 and rs10830963 (MTNR1B) in the general Japanese (n = 4,813) and Sri Lankan (n = 2,319) populations. To identify novel variants, we further examined genetic associations near each locus by using GWA scan data on 776 non-diabetic Japanese samples.
Fasting glucose association was replicated for the five single nucleotide polymorphisms (SNPs) at p < 0.05 (one-tailed test) in South Asians (Sri Lankan) as well as in East Asians (Japanese). In fine-mapping by GWA scan data, we identified in the G6PC2–ABCB11 region a novel SNP, rs3755157, with significant association in Japanese (p = 2.6 × 10-8) and Sri Lankan (p = 0.001) populations. The strength of association was more prominent at rs3755157 than that of the original SNP rs560887, with allelic heterogeneity detected between the SNPs. On analysing the cumulative effect of associated SNPs, we found the per-allele gradients (β = 0.055 and 0.069 mmol/l in Japanese and Sri Lankans, respectively) to be almost equivalent to those reported in Europeans.
Fasting glucose association at four tested loci was proven to be replicable across ethnic groups. Despite this overall consistency, ethnic diversity in the pattern and strength of linkage disequilibrium certainly exists and can help to appreciably reduce potential causal variants after GWA studies.
KeywordsAsians Association study Ethnicity Fasting plasma glucose Polymorphisms
Utah residents with northern and western European ancestry from the Centre d’Etude du Polymorphisme Humain collection
Fasting plasma glucose
- GWA studies
Genome-wide association studies
Japanese in Tokyo
Risk allele frequency
Single nucleotide polymorphism
Fasting plasma glucose (FPG) levels are associated with the future risk of type 2 diabetes and cardiovascular diseases [1, 2] and are tightly regulated despite considerable variation in food intake . It has been reported that genetic effects explain 54.8% of the variance of glucose levels in a European population . Recent progress in complex-trait genetics has allowed the identification of loci regulating FPG levels [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20].
Several loci influencing FPG levels have been identified or verified by genome-wide association (GWA) studies of Europeans; these include glucokinase (GCK) [5, 6, 7], glucokinase regulatory protein (GCKR) [8, 9, 10, 13], glucose-6-phosphatase catalytic subunit 2 (G6PC2), the ATP-binding cassette, subfamily B (MDR/TAP), member 11 (ABCB11) [14, 15, 16, 17], and melatonin receptor 1B (MTNR1B) [16, 17, 18].
All the associations were originally identified in populations of European ancestry. While some studies have shown reproducible associations [9, 11, 12, 19, 20], it remains to be further defined to what degree loci discovered in Europeans will show an association in populations of different ancestries. In addition, to localise the variant(s) responsible for an association signal, we need to generate a comprehensive list of potential causal variants in the regions of interest, i.e. to conduct fine-mapping after GWA studies. As discussed elsewhere , this fine-mapping will be challenging and genetic information from populations of different ancestries is expected to be useful [21, 22, 23, 24].
Apart from assessing the previously identified variants in two Asian populations, Japanese of East Asian ancestry and Sri Lankan of South Asian ancestry, we also explored index single nucleotide polymorphism (SNP) markers, which either tag the SNPs attaining a locus-wise significance level in the GWA scan of Japanese or were previously reported in Europeans. This was done to advance the fine-mapping of the associated loci [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20].
A replication study of the previously identified variants was performed in the general Japanese and Sri Lankan populations (Electronic supplementary material [ESM] Table 1, ESM Study samples for continuous traits), using 5,456 Japanese samples (including 4,813 non-diabetic participants) consecutively enrolled in a population-based setting as described elsewhere  and 3,012 Sri Lankan samples (including 2,319 non-diabetic participants) who had participated in the baseline survey of the Ragama Health Study  in Sri Lanka. Complementary to this replication study, we organised genetic studies of FPG levels as part of an ongoing GWA scan for cardiometabolic disorders among the Japanese population (ESM Study samples for continuous traits). We used 776 population-based, non-diabetic Japanese samples for preliminary screening of association with FPG levels. Then, the association signals were examined in the general populations mentioned above. In addition to quantitative trait analysis, type 2 diabetes associations were tested for index SNPs at G6PC2–ABCB11 in a Japanese case–control study panel comprising 5,629 cases and 6,406 controls as previously reported , and in a Sri Lankan case–control study panel (ESM Study samples for type 2 diabetes case–control studies). All participants from these different studies provided written informed consent and the local Ethics Committees approved the protocols.
Type 2 diabetes was diagnosed according to the WHO criteria as described in ESM Study samples for type 2 diabetes case–control studies.
SNP genotyping and quality control
In the replication study, samples were genotyped using the TaqMan assay (Life Technologies Japan, Tokyo, Japan) for five SNPs from four gene loci previously identified in European-descent populations [5, 6, 7, 8, 9, 10, 13, 14, 15, 16, 17, 18]. These included GCK (rs1799884), GCKR (rs780094), G6PC2–ABCB11 (rs560887) and MTNR1B (rs1387153 and rs10830963).
In the GWA scans, genotyping was performed with a bead array (Infinium HumanHap550; Illumina, San Diego, CA, USA) as described elsewhere  (ESM Fig. 1, ESM SNP genotyping, ESM Quality control of the GWA scan data). After the GWA scan, three additional SNPs in the G6PC2–ABCB11 region, rs483234, rs3755157 and rs853778, were genotyped with the TaqMan assay for follow-up.
SNP association analysis
SNPs were tested for association with FPG levels by using linear regression analysis in the additive genotype model (ESM SNP-based association analysis). A p value of <0.05 was considered statistically significant. For an association to be considered significant, it had to involve the same risk allele as that reported in Europeans and was accordingly assessed with a one-tailed test. To assess the proportion of variance for FPG that could be explained by a SNP, we calculated the coefficient of determination R2. The per-allele gradients, which correspond to the increase in FPG levels by additional ‘high FPG’ alleles of associated SNPs, were calculated in the linear regression model (including age, sex and BMI as covariates) as previously reported [14, 18] (ESM Evaluation of cumulative effect of multiple loci on FPG). We used PLINK (http://pngu.mgh.harvard.edu/∼purcell/plink/), the R software (version 2.8.1; www.r-project.org) and the rmeta package (http://cran.r-project.org) for association test and meta-analysis (websites accessed 15 October 2009).
In the G6PC2–ABCB11 region, we selected SNPs attaining a locus-wise significance level (p < 0.002 by Bonferroni’s correction for 23 SNPs genotyped in the relevant region) or reported in European studies, inferring the haplotypes using PLINK  and PHASE  software (http://depts.washington.edu/ventures/UW_Technology/Express_Licenses/PHASEv2.php). We then tested which haplotypes were strongly associated with the trait. In parallel, haplotypes were inferred from the genotype data of the SNPs in HapMap (www.hapmap.org) Utah residents with northern and western European ancestry from the Centre d’Etude du Polymorphisme Humain collection (CEU) and Japanese in Tokyo (JPT) categories using HaploView software (www.broad.mit.edu/mpg/haploview/)  and in South Asians from the Human Genome Diversity Panel (http://hagsc.org/hgdp/files.html) . Haplotype-tagging SNPs were selected and characterised in the large study panels.
Stepwise regression analysis for testing of independent associations
To test the most likely explanation for the signal of association among the index SNPs and their genotyped correlates, we performed stepwise linear regression analysis for FPG levels by forward selection (ESM Index SNPs showing an independent association). If two SNPs simultaneously included in the model each attained significance (p < 0.05), they could have independent associations. Further, when two haplotype classes that are distant in the phylogeny have an opposite effect and are tagged by two SNPs showing independent associations, the haplotype classes are presumably linked to different causative variants, thus implying allelic heterogeneity (ESM Haplotype explaining index association).
Cross-population filtering of causal variants
Association with FPG and metabolic traits at four loci
Association of SNPs with fasting plasma glucose level
Japanese panel (n = 4,813)
Sri Lankan panel (n = 2,319)
Allele frequency d
Reported in European studies
0.032 (0.012, 0.052)
0.074 (0.035, 0.113)
2.1 × 10−4
0.067 (0.045, 0.090)
0.103 (0.044, 0.162)
6.4 × 10−4
0.050 (−0.005, 0.105)
0.064 (0.056, 0.072)
0.075 (0.049, 0.101)
1.1 × 10−8
0.076 (0.028, 0.123)
0.062 (0.048, 0.076)
0.058 (0.038, 0.078)
9.7 × 10−9
0.036 (0.005, 0.068)
0.07 (0.05, 0.08)
0.056 (0.036, 0.075)
2.9 × 10−8
0.064 (0.033, 0.094)
3.6 × 10−5
0.072 (0.062, 0.082)
Tested in additionf
0.046 (0.026, 0.065)
3.9 × 10−6
0.043 (0.012, 0.074)
0.057 (0.037, 0.078)
2.6 × 10−8
0.069 (0.027, 0.111)
0.044 (0.024, 0.064)
1.5 × 10−5
0.026 (−0.006, 0.058)
Besides FPG levels, we analysed the relationship of SNPs with lipid traits (ESM Tables 3 and 4). Notably, rs780094 (GCKR) significantly and consistently modulated triacylglycerol levels in both ethnic groups (p = 2.2 × 10-10 in Japanese, p = 1.4 × 10-4 in Sri Lankan populations), where glucose-increasing alleles were associated with lower triacylglycerol levels as previously reported [8, 9, 11, 13]. Furthermore, glucose-increasing alleles at rs1799884 (GCK) and rs10830963 (MTNR1B) were significantly associated with reduced beta cell function (HOMA-B; p = 0.037 for rs1799884, p = 2.6 × 10-4 for rs10830963 in the Sri Lankan population), with no appreciable effect on fasting insulin or insulin sensitivity (ESM Table 5).
Refinement of genetic association in the G6PC2–ABCB11 region
In fine-mapping with Japanese GWA scan data, we identified in the G6PC2–ABCB11 region a novel associated SNP rs3755157, which was proven to be independent of the SNPs previously reported by GWA studies in Europeans [14, 15, 16, 17].
Fasting glucose association according to haplotypes in the G6PC2–ABCB11 region
Sri Lankan panelc
2.3 × 10−6
9.6 × 10−3
2.8 × 10−7
9.8 × 10−4
2.5 × 10−3
8.1 × 10−3
We then performed a stepwise linear regression (for FPG levels) to test whether one of the four haplotype-tagging SNPs was necessary and sufficient to explain the association signal (ESM Tables 13–17, ESM Fig. 4, ESM Index SNP showing an independent association). The FPG association remained significant (p < 0.05) when two haplotype-tagging SNPs, rs3755157 and rs560887, were included in the regression model (ESM Table 13). This independent association had gone unnoticed among more significant associations of SNPs that were in strong LD with a leading SNP, rs560887, among Europeans [14, 15]. Thus the presence of a novel SNP, rs3755157, and of allelic heterogeneity (ESM Fig. 5) has become evident in the G6PC2–ABCB11 region for the first time, as a result of comparing the GWA scan data between European and Japanese populations.
In the G6PC2–ABCB11 region, G6PC2 and ABCB11 are both biologically plausible candidate genes [15, 32, 33]. During fine-mapping, we attempted to partition the LD block into intervals, each containing SNPs strongly correlated with an index SNP, in the hope that correlation coefficients r2 would reflect phylogenic closeness once the index SNPs were selected from a reasonably dense set of SNP markers. This partitioning approach helped to prioritise the target interval for fine-mapping, thereby reducing the potential candidate variants to manageable proportions. For the G6PC2–ABCB11 region, the target intervals were estimated to be 14 kb (in the ABCB11 gene) for rs3755157 and 14 kb (in the G6PC2 gene and between the genes) for rs560887 when the LD threshold was set at r2 ≥ 0.6 in the HapMap data (Fig. 1, ESM Table 18, ESM Narrowing target intervals in fine-mapping).
Concordance of association for FPG levels and type 2 diabetes
Association of FPG-altering SNPs with type 2 diabetes in the G6PC2–ABCB11 region
rs560887 (FPG-increasing allele: G)
rs483234 (FPG-increasing allele: A)
rs3755157 (FPG-increasing allele: T)
Frequency, cases (n = 5,629)
Frequency, controls (n = 6,406)
OR (95% CI)
p value for trend
Sri Lankan (SL)b
Frequency, cases (n = 599)
Frequency, controls (n = 515)
OR (95% CI)
p value for trend
JPN and SL combined
OR (95% CI)
OR (95% CI)
The present study has proven that common variant loci influencing FPG levels are reproducible in two populations of Asian descent, Japanese (East Asians) and Sri Lankan (South Asians). To our knowledge, this is the first study investigating the genetic associations with FPG and related metabolic traits at four candidate loci, GCK, GCKR, G6PC2–ABCB11 and MTNR1B, in South Asians, who are known to have high prevalence of type 2 diabetes . The combined impact of associated SNPs is almost equivalent across the ethnic groups despite some cross-population diversity in the effect size of individual loci (ESM Figs 2 and 3, ESM Table 2). Other novel aspects of the present study include a fine-mapping approach using Japanese GWA scan data and consistent associations of FPG and type 2 diabetes in the G6PC2–ABCB11 region.
According to genome-wide patterns of SNPs examined in the Human Genomic Diversity Panel , much of sub-Saharan Africa, Europe, South and Central Asia (including Sri Lanka), and East Asia appear to be homogeneous and individuals from these populations can be distinguished from each other. Although limited in the number of examples, our study has provided evidence supporting the importance of human genetic diversity in complex disease studies. For instance, beside replicating FPG association at four candidate loci in two Asian populations, our data also clarified the genetic architecture of the G6PC2–ABCB11 region with regard to ethnic diversity. Using the GWA scan data, we found a novel SNP, rs3755157, to be a leading SNP among Japanese and independent of a leading SNP, rs560887, in Europeans (ESM Table 13–15).
As a fine-mapping approach, we listed HapMap SNPs having the strongest r2 (in the range of r2 ≥ 0.6) with each of the index SNPs in the G6PC2–ABCB11 region (Fig. 1, ESM Table 18). We performed cross-population filtering, which appreciably decreased the number of potential causal variants from 79 to 8 in the G6PC2–ABCB11 region (ESM Narrowing target intervals in fine-mapping). The novel SNP rs3755157 and its correlated SNPs are located in the 3′-side (introns) of the ABCB11 gene. While four different mRNAs, two alternatively spliced variants and two unspliced forms, are known to be transcribed from the ABCB11, it is possible that the potential causal variant(s) will influence the selective production of any of the 3′-side mRNA variants or the alteration of mRNA expression. Thus, closer inspection of subsets of SNP haplotypes shared by multiple ethnicities may allow us to appreciably narrow the field of potential causal variants before starting in-depth resequencing and functional follow-up studies, as demonstrated for the G6PC2–ABCB11 region. During the preparation of our manuscript, replication of the G6PC2 association was also reported in a Chinese population , where four SNPs were selected from the HapMap database so as to tag common variations near the G6PC2 gene. Although the index SNP (rs560887) originally detected in Europeans was not tested, three (of four) SNPs appeared to show significant association in the Chinese population, in agreement with our findings in Japanese.
Our data also verified concordance of association for FPG levels and type 2 diabetes risk by using a systematic study design; i.e. unbiased estimates with stratification of general populations plus large-scale case–control studies involving 12,035 Japanese and 1,114 Sri Lankan samples (Table 3, ESM Fig. 6). It has been debated whether the genetic determinants regulating FPG levels in physiological states differ from those increasing type 2 diabetes risk. Some studies report that carriers of glucose-increasing alleles at three loci (MTNR1B, GCK and GCKR) show a higher risk of type 2 diabetes [8, 17, 18], although there is no significant association between G6PC2–ABCB11 variants and type 2 diabetes in populations of European descent [14, 17, 35]. In this context, our data not only supported the concordant association of G6PC2–ABCB11 variants for FPG and type 2 diabetes in two Asian populations, but also indicated that genetic determinants regulating FPG levels could, at least in part, differ from those increasing type 2 diabetes risk (ESM Fig. 6, ESM Consistent association of fasting glucose and type 2 diabetes in the G6PC2–ABCB11 region). It is likely that genetic susceptibility for FPG levels increases type 2 diabetes risk in the population at large, but that some diabetic patients will develop the disease independently of a predisposition to elevated FPG levels.
In summary, despite the overall reproducibility of FPG association across the populations, ethnic diversity in allele frequencies led to the discovery of allelic heterogeneity in the G6PC2–ABCB11 region. The diversity in the LD pattern also helped to reduce the probable causative variants in the corresponding region. The prevalence of the phenomena described here in human complex trait genetics is another research area warranting investigation. For applicable cases, the use of ethnic diversity in genetic studies can constitute an efficient approach subsequent to GWA scan.
This work was supported by the Program for Promotion of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation Organization (NIBIO) and the Manpei Suzuki Diabetes Foundation. Support also came from the Ministry of Education, Culture, Sports, Science and Technology of Japan. We acknowledge the outstanding contributions of the International Medical Center of Japan (IMCJ) employees, who provided technical and infrastructural support for this work. Above all, we thank the patients and study participants who made this work possible and who give it value. We also thank all the people who continuously support the Hospital-based Cohort Study at IMCJ, the Amagasaki Study and the Kyushu University Fukuoka Cohort Study in Japan, and the Ragama Health Study in Sri Lanka. We also thank C. Makibayashi and the many physicians of the Amagasaki Medical Association, as well as M. Makaya, T. Mizoue, H. Janaka de Silva, U. Ranawaka and other staff at the University of Kelaniya for their help with collection of DNA samples and accompanying clinical information. The DNA samples of type 2 diabetes cases used for this research were partly provided by the Leading Project for Personalized Medicine in the Ministry of Education, Culture, Sports, Science and Technology, Japan. The GWA study conducted by NIBIO GWA Study Group was organised to clarify the pathogenesis of diabetes and associated metabolic disorders as well as cardiovascular complications. The collaborating institutions that constitute the NIBIO GWA Study Group are: International Medical Center of Japan; Kyushu University; Osaka University; Nagoya University; Kinki University; Shimane University; Tohoku University; the Institute for Adult Diseases, Asahi Life Foundation; Chubu Rosai Hospital; Amagasaki Health Medical Foundation; collaborating groups in the Amagasaki Medical Association; and collaborating groups in the Kyushu region.
Duality of interest
The authors declare that there is no duality of interest associated with this manuscript.