Research article The Journal of Clinical Investigation      http://www.jci.org  � Variations in the G6PC2/ABCB11 genomic region are associated with fasting glucose levels Wei-Min Chen,1,2 Michael R. Erdos,3 Anne U. Jackson,4 Richa Saxena,5 Serena Sanna,4,6 Kristi D. Silver,7 Nicholas J. Timpson,8 Torben Hansen,9 Marco Orrù,6 Maria Grazia Piras,6 Lori L. Bonnycastle,3 Cristen J. Willer,4 Valeriya Lyssenko,10 Haiqing Shen,7 Johanna Kuusisto,11 Shah Ebrahim,12 Natascia Sestu,13 William L. Duren,4 Maria Cristina Spada,6 Heather M. Stringham,4 Laura J. Scott,4 Nazario Olla,6 Amy J. Swift,3 Samer Najjar,13 Braxton D. Mitchell,7 Debbie A. Lawlor,8 George Davey Smith,8 Yoav Ben-Shlomo,14 Gitte Andersen,9 Knut Borch-Johnsen,9,15,16 Torben Jørgensen,15 Jouko Saramies,17 Timo T. Valle,18 Thomas A. Buchanan,19,20 Alan R. Shuldiner,7 Edward Lakatta,13 Richard N. Bergman,20 Manuela Uda,6 Jaakko Tuomilehto,18,21 Oluf Pedersen,9,16 Antonio Cao,6 Leif Groop,10 Karen L. Mohlke,22 Markku Laakso,11 David Schlessinger,13 Francis S. Collins,3 David Altshuler,5 Gonçalo R. Abecasis,4 Michael Boehnke,4 Angelo Scuteri,23,24 and Richard M. Watanabe20,25 1Department of Public Health Sciences and 2Center for Public Health Genomics, University of Virginia, Charlottesville, Virginia, USA. 3Genome Technology Branch, National Human Genome Research Institute, Bethesda, Maryland, USA. 4Center for Statistical Genetics and Department of Biostatistics, University of Michigan, Ann Arbor, Michigan, USA. 5Program in Medical and Population Genetics, Broad Institute of MIT and Harvard, Cambridge, Massachusetts, USA. 6Istituto di Neurogenetica e Neurofarmacologia, Consiglio Nazionale delle Ricerche, Cagliari, Italy. 7Division of Endocrinology, Diabetes and Nutrition, University of Maryland School of Medicine, Baltimore, Maryland, USA. 8MRC Centre for Causal Analyses in Translational Epidemiology, Department of Social Medicine, University of Bristol, Bristol, United Kingdom. 9Steno Diabetes Center, Gentofte, Denmark. 10Department of Clinical Sciences, Diabetes and Endocrinology, Lund University, University Hospital Malmö, Malmö, Sweden. 11Department of Medicine, University of Kuopio and Kuopio University Hospital, Kuopio, Finland. 12Department of Epidemiology and Population Health, Non-communicable Disease Epidemiology Unit, London School of Hygiene and Tropical Medicine, University of London, London, United Kingdom. 13Gerontology Research Center, National Institute on Aging, Baltimore, Maryland, USA. 14Social Medicine Department, University of Bristol, Bristol, United Kingdom. 15Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark. 16Faculty of Health Sciences, University of Aarhus, Aarhus, Denmark. 17Savitaipale Health Center, Savitaipale, Finland. 18Diabetes Unit, Department of Health Promotion and Chronic Disease Prevention, National Public Health Institute, and Department of Public Health, University of Helsinki, Helsinki, Finland. 19Department of Medicine, Division of Endocrinology, and 20Department of Physiology and Biophysics, Keck School of Medicine, University of Southern California, Los Angeles, California, USA. 21South Ostrobothnia Central Hospital, Senäjoki, Finland. 22Department of Genetics, University of North Carolina, Chapel Hill, North Carolina, USA. 23Laboratory of Cardiovascular Science, National Institute on Aging, NIH, Baltimore, Maryland, USA. 24Unità Operativa Geriatria, Istituto Nazionale Ricovero E Cura Anziari, Rome, Italy. 25Department of Preventive Medicine, Keck School of Medicine, University of Southern California, Los Angeles, California, USA. Introduction Glucose is the major source of energy in humans, with levels in  vivo determined by a balance of glucose absorption via the gut,  production primarily by the liver, and utilization by both insulin- sensitive and insulin-insensitive tissues (1, 2). Homeostatic control  of glucose levels involves complex interactions between humoral  and neural mechanisms that work in concert to regulate tightly  the balance between production and utilization to maintain a nor- Nonstandard abbreviations used: ABCB11, ATP-binding cassette, subfamily  B (MDR/TAP), member 11; BWHHS, British Women’s Heart and Health Study;  DGI, Diabetes Genetics Initiative; FUSION, Finland–United States Investigation of  Non–Insulin-Dependent Diabetes Mellitus Genetics; G6PC2, glucose-6-phosphatase  catalytic subunit 2; GWA, genome-wide association; LD, linkage disequilibrium;  METSIM, METabolic Syndrome in Men; T2DM, type 2 diabetes mellitus. Conflict of interest: The authors have declared that no conflict of interest exists. Citation for this article: J. Clin. Invest. doi:10.1172/JCI34566. Identifying the genetic variants that regulate fasting glucose concentrations may further our understanding of the pathogenesis of diabetes. We therefore investigated the association of fasting glucose levels with SNPs in 2 genome- wide scans including a total of 5,088 nondiabetic individuals from Finland and Sardinia. We found a significant association between the SNP rs563694 and fasting glucose concentrations (P = 3.5 × 10–7). This association was fur- ther investigated in an additional 18,436 nondiabetic individuals of mixed European descent from 7 different stud- ies. The combined P value for association in these follow-up samples was 6.9 × 10–26, and combining results from all studies resulted in an overall P value for association of 6.4 × 10–33. Across these studies, fasting glucose concentra- tions increased 0.01–0.16 mM with each copy of the major allele, accounting for approximately 1% of the total varia- tion in fasting glucose. The rs563694 SNP is located between the genes glucose-6-phosphatase catalytic subunit 2 (G6PC2) and ATP-binding cassette, subfamily B (MDR/TAP), member 11 (ABCB11). Our results in combination with data reported in the literature suggest that G6PC2, a glucose-6-phosphatase almost exclusively expressed in pancreatic islet cells, may underlie variation in fasting glucose, though it is possible that ABCB11, which is expressed primarily in liver, may also contribute to such variation. research article � The Journal of Clinical Investigation      http://www.jci.org mal fasting glucose. Elevations in blood glucose are diagnostic of  diabetes. Type 2 diabetes mellitus (T2DM) afflicts more than 171  million worldwide and is a leading cause of kidney failure, blind- ness, and lower limb amputations (3–5). Even more modest eleva- tions in glucose concentration (so-called prediabetes) are associated  with cardiovascular disease and accelerated atherosclerosis (6). In  individuals progressing toward future T2DM, the fasting glucose  concentration appears to change only modestly over time until the  advent of β cell dysfunction, at which point the glucose concen- tration increases rapidly (7, 8). Many studies have shown that the  lowering of glucose levels in individuals with diabetes can prevent  or delay diabetes-related complications, providing further evidence  for the damaging effects of chronic glucose elevations. Both genetic and environmental factors contribute to the patho- physiology of T2DM (9–11). The contributions of environmental  exposures to T2DM risk are best illustrated by results from the Dia- betes Prevention Program (11) and the Finnish Diabetes Prevention  Study (12), in which T2DM incidence was significantly reduced  by intensive lifestyle modification. However, the contribution of  genetic factors to T2DM risk is not as well understood. Recent  genome-wide association (GWA) studies have identified 16 novel  T2DM susceptibility loci (13–18), generating new insights into the  genetic architecture underlying T2DM. In contrast to disease sta- tus, even less is known about genetic variation that alters specific  T2DM-related quantitative traits such as glucose and insulin con- centrations. As seen for T2DM, identification of genetic variants  associated with T2DM-related quantitative traits is likely to require  large sample sizes due to relatively small gene effect sizes. Fasting  glucose concentrations have been shown to be heritable, with nar- row-sense heritability estimates ranging from 25% to 40% (19–24).  Given the central role of glucose concentration in the pathogenesis  and diagnosis of T2DM and its complications, GWA for glucose  concentrations provides an excellent opportunity to identify genes  underlying variation in glucose concentrations that may also repre- sent additional T2DM susceptibility loci. An example of this comes  from the studies by Weedon et al., who showed by metaanalysis and  large cohorts that variation in the glucokinase gene was associated  with both fasting glucose and birth weight (25). GWA studies for T2DM and adiposity were completed by the  groups undertaking the Finland–United States Investigation of  Non–Insulin-Dependent Diabetes Mellitus Genetics (FUSION)  (14, 26, 27) and the SardiNIA Study of Aging (24, 28), respec- tively. Both studies assessed fasting glucose in their respective  cohorts, allowing GWAs for fasting glucose in each study and  combination of these results in a metaanalysis. The strongest  signals from the fasting glucose GWA metaanalysis were from  variants near genes for ATP-binding cassette, subfamily B (MDR/ TAP), member 11 (ABCB11) and glucose-6-phosphatase catalytic  subunit 2 (G6PC2). This association was replicated in a series of  7 studies involving a total of 18,436 individuals (13, 29–35), sug- gesting for what we believe is the first time that variation in one  of these genes may play a role in the regulation of fasting glucose  concentrations in humans. Results Subject demographics and clinical characteristics for the FUSION  and SardiNIA samples are summarized in Table 1. Because treat- ment for T2DM affects fasting glucose concentrations, all analyses  in this report were restricted to nondiabetic subjects. Initial review  of association results from both the FUSION stage 1 and SardiNIA  GWA scans of a combined total of 5,088 nondiabetic individuals  focused on SNPs that were genotyped in the SardiNIA study and  imputed in the FUSION study. Among these, rs563694 exhibited  the strongest evidence for association in both samples (SardiNIA,  P = 7.6 × 10–5; FUSION stage 1, P = 8.0 × 10–4; Table 2), with a  metaanalysis P value of 3.5 × 10–7. Given the strength of this initial  association, our follow-up efforts focused on rs563694. Additional  independent associations from our fasting glucose GWA study are  presented in Supplemental Table 1 (supplemental material avail- able online with this article; doi:10.1172/JCI34566DS1). Analyses were repeated once imputation was completed in both  the FUSION stage 1 and SardiNIA samples. SNP rs563694 and  other SNPs in strong linkage disequilibrium (LD; defined as r2 > 0.8   in the FUSION samples) constituted the 17 strongest association  results in the combined FUSION/SardiNIA GWA for fasting glu- cose metaanalysis (Figure 1). In fact, 22 SNPs associated with fast- ing plasma glucose with P ≤ 1 × 10–4 were located within a 63.9-kb   region on chromosome 2 (Supplemental Table 2). These SNPs  were located in an extended region of LD that spans 2 biologically  plausible candidate genes for glucoregulation (Figure 1). The first  is  G6PC2,  also  known  as  islet-specific  glucose-6-phosphatase– related protein (IGRP). G6PC2 is part of a larger family of enzymes  involved in hydrolysis of glucose-6-phosphate in the gluconeogen- ic and glycogenolytic pathways (36, 37). The second is ABCB11,  a member of the MDR/TAP subfamily of ATP-binding cassette  transporters involved in multidrug resistance (38, 39). Table � Subject demographics and clinical characteristics for individuals with rs563694 genotype data Study Phenotyped Geographic Study age BMI Fasting subjects origin (years) (kg/m2) glucose (mM) FUSION stage 1 1,233 Finland 63.0 (13.7) 26.6 (5.0) 5.36 (0.72) FUSION stage 2 655 Finland 61.0 (12.3) 26.3 (4.9) 5.48 (0.50) FUSION additional spouses/offspring 522 Finland 39.1 (12.2) 26.0 (6.4) 5.11 (0.78) SardiNIA 3,855 Sardinia, Italy 41.3 (27.1) 24.7 (6.3) 4.72 (0.77) DGI 1,411 Finland and Sweden 58.7 (15.4) 26.7 (4.78) 5.28 (0.70) Amish 1,655 USA 49.0 (23.7) 26.7 (6.6) 4.90 (0.58) METSIM 4,386 Finland 59.0 (10.0) 26.4 (4.5) 5.60 (0.70) Caerphilly 1,063 United Kingdom 56.7 (4.4) 26.2 (3.5) 4.80 (0.86) BWHHS 3,532 United Kingdom 68.5 (5.9) 27.6 (5.0) 5.80 (0.87) Inter99 5,734 Denmark 46.1 (7.9) 26.3 (4.5) 5.54 (0.80) Values are reported as median (interquartile range). research article The Journal of Clinical Investigation      http://www.jci.org  � Among all genotyped or imputed SNPs in this region, rs560887,  which was genotyped in FUSION stage 1, and imputed and fol- lowed up by genotyping in SardiNIA, showed the strongest overall  evidence for association (SardiNIA, P = 4.4 × 10–8; FUSION stage 1,   P = 1.7 × 10–3; Supplemental Table 2), with a metaanalysis P value  of 2.8 × 10–10. In addition, rs853789 and rs853787, both located in  intron 19 of ABCB11 and in perfect LD with each other (D′ = 1.0,   r2  =  1.0),  showed  strong  evidence  for  association  with  fasting  glucose concentrations with metaanalysis P values of 1.4 × 10–9  and 1.0 × 10–9, respectively (Supplemental Table 2). rs853789 is  located 38.3 kb from rs560887 and 27.4 kb from rs563694 and is  in strong LD with both SNPs (D′ = 0.98, r2 = 0.81 with rs560887;  and D′ = 0.98, r2 = 0.95 with rs563694). rs560887 is 10.9 kb from  rs563694, is in high LD with rs563694 (D′ = 0.99, r2 = 0.84), and  is located in intron 3 of G6PC2. In contrast, rs563694 lies between  G6PC2 and ABCB11 and is in extended LD with ABCB11. In both  the SardiNIA and FUSION stage 1 samples, each copy of the A  allele for rs563694 was associated with small increases in fast- ing glucose (0.064 mM for SardiNIA and 0.051 mM for FUSION  stage 1; Table 2) that are clinically insignificant and accounted  for approximately 1% of the variance in fasting glucose. Similar  effect sizes were observed for rs560887 (0.089 mM for SardiNIA  and 0.052 mM for FUSION stage 1). We assessed the potential contribution of population stratifica- tion by computing the genomic control parameter (40) indepen- dently for both studies. The genomic control values were 1.01 for  both FUSION and SardiNIA, suggesting that population stratifi- cation and/or unmodeled relatedness did not contribute signifi- cantly to our observed association. Analyses that included BMI  as a covariate did not significantly alter the association between  rs563694 and fasting glucose in the FUSION stage 1 (P = 8.1 × 10–4  without BMI versus 9.1 × 10–4 with BMI) and SardiNIA samples  (7.6 × 10–5 without BMI versus 3.8 × 10–5 with BMI) independently  or jointly (P = 3.5 × 10–7 without BMI versus 1.8 × 10–7 with BMI),  suggesting the association was not a consequence of adiposity,  which is known to induce insulin resistance and increase glucose  concentrations (2). The association between rs563694 and fasting  glucose also remained significant after individual adjustment for  each of the 10 SNPs shown to be associated with T2DM in our  recent GWA studies (Supplemental Table 3) (13–15) or when all  10 SNPs were included jointly in the model (P = 6.8 × 10–4 versus  P = 8.0 × 10–4 for FUSION stage 1 samples and 5.1 × 10–5 versus  7.6 × 10–5 for SardiNIA samples). The 10 SNPs shown to be associ- ated with T2DM were themselves not significantly associated with  fasting glucose concentrations in the FUSION stage 1 or SardiNIA  samples (Supplemental Table 4). FUSION investigators genotyped rs563694 in 655 stage 2 sam- ples and 522 additional spouses and offspring of T2DM patients  included  in  stage  1;  this  SNP  continued  to  show  evidence  for  association with fasting glucose (FUSION stage 2, P = 2.0 × 10–3;  FUSION stage 1 families, P = 1.9 × 10–5; see Table 2). The meta- analysis that combined results from the FUSION stage 1 and 2  and SardiNIA studies resulted in a P value of 5.3 × 10–9, surpassing  standard thresholds for genome-wide significance. We also examined the association between rs563694 and fast- ing glucose in 6 follow-up samples (Table 2). The characteristics  of these samples are summarized in Table 1. Association between  rs563694 and fasting glucose was confirmed in the Amish study  (P = 4.1 × 10–5); the METabolic Syndrome In Men study (METSIM;   P = 1.3 × 10–10), the Caerphilly study (2.6 × 10–7), the British Wom- en’s Heart and Health Study (BWHHS; P = 1.2 × 10–3), and Inter99  (P = 8.2 × 10–8; Table 2), with fasting glucose concentrations increas- ing with each copy of the A allele in all studies. While evidence for  association in the Diabetes Genetics Initiative (DGI) study was  not statistically significant (P = 0.19; Table 2), the results show a  trend in the same direction as observed in the other samples. When  the results from all follow-up studies were combined in a meta- analysis of 24,046 samples, there was strong evidence for associa- tion between rs563694 and fasting glucose in both the follow-up  samples (n = 18,435, P = 6.9 × 10–26; Table 2) and in all GWA and  follow-up samples combined (n = 24,046, P = 6.4 × 10–33; Table 2).   In contrast, rs563694 did not show evidence for association with  T2DM in the FUSION stage 1 study (P = 0.22), the DGI GWA sam- ples (P = 0.78), or the METSIM study (P = 0.09). Table � Association between rs563694 and fasting glucose in nondiabetic individuals Frequency Mean fasting glucose (mM) (SD) Effect (SE) Effect (SE) Study n C allele CC AC AA mM Standardized P value GWA Samples FUSION stage 1 1,233 0.34 5.26 (0.48) 5.31 (0.48) 5.33 (0.47) 0.051 (0.019) 0.143 (0.043) 8.0 × 10–4 SardiNIA 3,855 0.46 4.88 (0.67) 4.95 (0.62) 5.00 (0.59) 0.064 (0.018) 0.118 (0.030) 7.6 × 10–5 3.5 × 10–7A FUSION 1 familiesB 1,755 0.34 5.20 (0.49) 5.28 (0.50) 5.31 (0.54) 0.065 (0.018) 0.155 (0.036) 1.9 × 10–5 Follow-up samples FUSION stage 2 655 0.36 5.28 (0.43) 5.44 (0.35) 5.46 (0.36) 0.068 (0.021) 0.180 (0.058) 2.0 × 10–3 DGI 1,411 0.34 5.24 (0.50) 5.28 (0.51) 5.29 (0.49) 0.022 (0.021) 0.053 (0.039) 0.19 Amish 1,655 0.24 4.90 (0.47) 4.89 (0.51) 5.03 (0.53) 0.090 (0.022) 0.175 (0.042) 4.1 × 10–5 METSIM 4,386 0.32 5.55 (0.49) 5.64 (0.50) 5.71 (0.49) 0.074 (0.011) 0.145 (0.023) 1.3 × 10–10 Caerphilly 1,063 0.36 4.69 (0.91) 4.87 (0.99) 5.00 (1.19) 0.155 (0.047) 0.214 (0.041) 2.6 × 10–7 BWHHS 3,532 0.34 6.01 (1.69) 6.09 (1.81) 6.06 (1.49) 0.006 (0.042) 0.079 (0.025) 1.2 × 10–3 Inter99 5,734 0.36 5.46 (0.85) 5.52 (0.87) 5.58 (0.70) 0.057 (0.015) 0.135 (0.019) 8.2 × 10–8 6.3 × 10–28C 6.1 × 10–35D Glucose values are reported unadjusted for covariates. A GWA metaanalysis P value. BThese data represent the 522 additional spouses and offspring com- bined with their FUSION stage 1 family members. C Follow-up metaanalysis P value. D Overall metaanalysis P value. research article � The Journal of Clinical Investigation      http://www.jci.org Figure  2  shows  the  results  of  a  metaanalysis  based  upon  the  effect size observed in each of the 8 studies. Overall, fasting glucose  concentrations increased 0.065 mM (95% CI: 0.053–0.077 mM)   with each copy of the major allele. Discussion We took advantage of GWA studies originally performed to identify  susceptibility genes for T2DM (FUSION) and aging-related traits  (SardiNIA) to also identify genes underlying variation in fasting  glucose concentration. Both FUSION and SardiNIA initially identi- fied rs563694 as being associated with fasting glucose levels. Given  that both studies were performed in relatively homogeneous popu- lations of mixed European descent, it is unlikely that population  stratification accounted for the initial association. The estimated  genomic control (40) values for FUSION stage 1 and SardiNIA were  both 1.01, providing further evidence against the contribution of  population stratification to the observed association. In  the  SardiNIA  sample,  we  genotyped  SNPs  rs560887  and  rs853789 to validate the results based on imputation. The dis- crepancy rate per allele between the imputed and typed genotypes  at these 2 SNPs was 1.4% and 2.4%, respectively, and the associa- tion result with the actual genotypes was stronger than with the  imputed genotypes: P = 9.0 × 10–10 and 2.6 × 10–8, respectively Adiposity may induce insulin resistance and thus alter glucose  concentrations (2) independent of the effects of the SNP on glu- cose concentrations per se. However, the association remained sig- nificant even when we included BMI as a covariate in the analysis,  suggesting adiposity is not a major contributor to the observed  association.  Similarly,  in  the  follow-up  studies,  the  results  did  not change whether BMI was included or excluded as a covariate.  Some known sex-specific effects, such as differences in fat distri- bution, could also confound our results. We found no sex-specific  effect modification in the FUSION and SardiNIA samples. Also, it  should be noted that we observed evidence for association between  rs563694 in the METSIM and Caerphilly samples that only includ- ed men and in the BWHHS that comprised women only. Thus,  the lack of a sex-specific effect in FUSION and SardiNIA is sup- ported by the independent associations observed in these samples.  Subsequent analyses of the GWA data revealed rs560887 as having  the strongest evidence for association with fasting glucose in this  region and suggested, based on the SNP location, that G6PC2 plays  a role in glucoregulation. However, 2 additional SNPs in strong LD  with rs560887 located in the adjacent ABCB11 also showed similar  evidence for association with fasting glucose. In the 7 follow-up studies, rs563694 continued to show associa- tion with fasting glucose, although marginal evidence for hetero- geneity among studies was noted (Q = 14.6; P = 0.02; I2 = 59.0%;  95% CI: 5.6–82.2%) (37). For example, the DGI samples did not  exhibit a significant association and the BWHHS samples, despite  being among the largest follow-up samples, showed only modest  evidence for association (Table 2). These 2 studies yielded similar  effect size estimates (0.053 for DGI and 0.079 for BWHHS) that  were smaller than in the other studies (Table 2). Differences in  both populations and sample ascertainment could be contribut- ing to the observed heterogeneity. When these 2 studies are not  considered, the heterogeneity estimate is reduced (Q = 3.7; P = 0.45;  I2 = 0%; 95% CI = 0–77.6%) However, despite the variability in effect  size, the direction of the effect was the same in all studies. There are 2 biologically plausible candidate genes in the region  identified  by  our  association  analyses  that  may  affect  glucose  levels. Although rs560887, which is located in intron 3 of G6PC2  just 26 bp proximal to exon 4, showed the strongest evidence for  association in the GWA studies, SNPs in LD with rs560887 and  rs563694 that show similar levels of association with fasting glu- cose concentrations were located in intron 19 of ABCB11. ABCB11  is involved in ATP-dependent secretion of bile salts and is almost  exclusively expressed in the liver. Mutations in ABCB11 have been  shown  to  be  associated  with  intrahepatic  cholestasis  (OMIM  603201) (38) and drug-induced hepatotoxicity (39, 41). In anti- lipid drug trials, bile acid sequestrants have been shown to lower  glucose concentrations and improve insulin sensitivity, presum- ably through reduction of triglyceride levels (42). Based upon these  observations, if ABCB11 were contributing significantly to varia- tion in fasting glucose, one might expect to also see associations  Figure � Fasting glucose association in the FUSION and SardiNIA GWA metaanalysis. Top panel shows evidence for association with fasting glucose under an additive genetic model for the com- bined FUSION stage 1 and SardiNIA metaanal- ysis. The –log(P value) for the test of associa- tion is plotted against genomic position (NCBI build 35) for all genotyped (red circles, SardiNIA; blue circles, FUSION) and imputed (gray circles) SNPs in the SardiNIA data. SNPs typed in both samples are indicated by red circles with blue centers. SNPs rs560887 and rs853789 were later typed in the SardiNIA samples, and the actual genotypes resulted in even stronger asso- ciation than shown here. Bottom panel shows the LD pattern (r2) around G6PC2 and ABCB11 for the CEPH (Utah residents with ancestry from northern and western Europe) sample from the HapMap. The scale at the bottom shows the magnitude of LD in 15 colors, ranging from blue for low LD to red for high LD. research article The Journal of Clinical Investigation      http://www.jci.org  � with lipids or insulin sensitivity. However, rs560887, rs563694,  rs853789, and rs853787 were not associated with lipid measure- ments in a metaanalysis of FUSION stage 1 and SardiNIA samples  (P > 0.16). Also, none of these SNPs were associated with minimal  model-derived insulin sensitivity in FUSION samples (P > 0.30).  Thus, our data do not support a role for ABCB11 in glucoregula- tion, and other evidence directly linking ABCB11 to regulation of  glucose concentrations is scarce. In contrast, G6PC2, the β cell–specific isoform of glucose-6-phos- phatase is a highly relevant candidate gene for glucoregulation. The  mouse homolog G6pc2 has been previously implicated as an auto- antigen in the NOD mouse model of type 1 diabetes (43). Wang et  al. recently generated G6pc2-null mice and noted that at 16 weeks  of age, fasting glucose concentrations had decreased approximately  13% in both male and female G6pc2-null mice when compared with  wild-type mice (44). This modest decrease in glucose concentration  was observed despite the absence of any differences in body weight,  fasting insulin, or fasting glucagon concentrations. The character- istics of these G6pc2-null mice closely paralleled our observations  that rs560887 and rs563694 were associated with modest chang- es in fasting glucose but not in BMI or fasting insulin, which are  consistent with the hypothesis that presence of a C allele results in  lower G6PC2 expression and therefore lower glucose concentrations.  Interestingly, G6pc2 mRNA levels appear to increase with increasing  glucose concentration in isolated mouse islets (36). Molecular cloning of G6pc2 identified 2 splice forms that differ  by the presence or absence of exon 4 in BALB/C and ob/ob mice  and in insulinoma tissue (45). The longer cDNA including exon 4  has approximately 50% homology with glucose-6-phosphatase cat- alytic subunit (G6pc) across a variety of species including humans  and is membrane bound in the endoplasmic reticulum (46). The  corresponding G6PC2 splice forms have been observed in human  pancreas (47). rs560887 is located in intron 3, just 26 bp proximal  to exon 4, raising the possibility that this variant may play a role in  whether the full-length transcript is formed. G6PC  hydrolyzes  glucose-6-phosphate  to  form  glucose  and  release a phosphate group. Despite its similarity to G6PC, G6PC2  is reported to have little to no hydrolase activity in humans (36, 37,  45, 46). In normal and genetically obese mice, the splice form lack- ing exon 4 appears to be the most predominant observed in islets  (45) and lacks sequences that may be critical for hydrolytic activity  (45, 48), suggesting the full-length form of G6pc2 may have impli- cations for activity of G6pc2 and its potential role in glucoregula- tion. Greater hydrolase activity has been reported in cell lines over- expressing the full-length form of G6PC2 (36). Also, in islets from  streptozotocin-treated mice, glucose cycling, an indicator of G6pc2  activity, was approximately 3-fold higher compared with islets from  untreated mice (49), and even greater increases were observed in  islets from ob/ob mice (50, 51). The conversion of glucose to glu- cose-6-phosphate is the critical step in stimulus-secretion coupling  for  insulin  secretion.  Variation  in  G6PC2  may  increase  glucose  cycling in β cells, resulting in altered generation of ATP, which  would have implications for insulin secretion. In addition, G6PC2- induced alterations in β cell glucose metabolism would also have  downstream effects on phosphoinositide 3-kinase activity, which  regulates pancreas duodenum homeobox-1 (PDX1) binding to the  insulin gene and subsequent insulin gene transcription (52). The possible role for G6PC2 in altering glucose concentrations  raises the question of whether this gene also confers susceptibil- ity to T2DM. We observed no association between fasting glucose  and rs563694 and rs560887 in individuals with T2DM from the  FUSION, DGI, and METSIM studies (P > 0.50). However, the anal- ysis of fasting glucose concentration in individuals with T2DM  is confounded by diabetes pathology, treatment, and differential  response to therapy. Therefore the lack of association with fast- ing glucose in individuals with T2DM does not preclude G6PC2  as  contributing  to  susceptibility  to  T2DM.  Similarly,  when  we  tested  these  SNPs  for  association  with  T2DM  in  the  FUSION,  DGI, and METSIM samples, we observed no evidence for associa- tion (P > 0.08). Further, the modest effect on glucose concentra- tions observed in our analysis of nondiabetic individuals suggests  we may lack sufficient power to detect association with T2DM.  Whereas the cumulative evidence would suggest that G6PC2 may  regulate fasting glucose concentrations and does not contribute  significantly  to  susceptibility  to  T2DM,  larger  studies  may  be  required to elucidate the role of this gene in T2DM susceptibility. Variation in the promoter region of glucokinase (GCK, rs1799884)  has been shown to be associated with fasting glucose and impaired  insulin secretion (53–55) and may play a role in altering birth weight  (56). These initial findings were confirmed in a comprehensive meta- analysis performed by Weedon et al., demonstrating that rs1799884  was associated with fasting glucose (meta P = 1.0 × 10–9) and that the  presence of a maternal A allele for rs1799884 was associated with  increased birth weight of the child (P = 0.02) (25). GCK, an enzyme  that works counter to G6PC2, converts glucose to glucose-6-phos- phate,  forming  the  critical  step  in  secretion-stimulus  coupling  in pancreatic β cells. In addition, the recent GWA study from the  DGI identified variation in glucokinase regulatory protein (GCKR)  (rs780094) to be associated with triglyceride levels (13). GCKR is an  allosteric regulator of GCK in both liver and pancreatic islets whose  inhibitory  effect  is  enhanced  by  fructose-6-phosphate  and  sup- pressed by fructose-1-phosphate (57). We found modest evidence  Figure � Effect size and 95% CI for rs563694 are shown for the 8 studies. The overall metaanalysis across these studies yielded an effect size of 0.065 mM (95% CI: 0.053, 0.077 mM). research article � The Journal of Clinical Investigation      http://www.jci.org for association between fasting glucose and rs1799884 (FUSION  stage 1, P = 1.6 × 10–2; SardiNIA, P = 2.0 × 10–3; meta P = 1.1 × 10–4)   and no evidence for association between fasting glucose and rs780094  (FUSION stage 1, P = 0.44; SardiNIA, P = 0.11; meta P = 0.077).   While these results provide evidence for association between varia- tion in GCK and fasting glucose but not between GCKR and fast- ing glucose in our studies, we cannot exclude the possibility that a  complex interaction among GCK, GCKR, and G6PC2 may regulate  fasting glucose levels. This will require further study. In conclusion, we used GWA to identify variation in both ABCB11  and  G6PC2 as genes that potentially contribute to variation in  fasting glucose concentrations in nondiabetic subjects of mixed  European descent. There is more literature with data supporting  a role for G6PC2, but in the absence of functional data, we cannot  discount the possibility that ABCB11 may also contribute signifi- cantly to variation in fasting glucose concentration. Heritability  for fasting glucose has been estimated to be 25%–40% (19–24),   yet the variants we identified account for approximately 1% of  the variance in fasting glucose, indicating that the majority of the  variability in fasting glucose remains unexplained. The remaining  variability is likely due to the effects of additional common genetic  variants of modest effect, less common genetic variants of mod- erate effect, and a variety of gene-gene and gene-environmental  interaction effects. It should also be noted that the magnitude of  the effect observed in our study is consistent with other reports of  quantitative trait associations (58–60). Additional studies, likely with larger sample sizes, will be required  to identify additional genetic variants contributing to variation  in fasting glucose. The variants identified in our study are not  likely to be functional, but in LD with the functional variant(s).  Additional fine mapping, sequencing, and functional studies will  be required to define the molecular mechanisms underlying our  observed association. Methods The FUSION and SardiNIA study samples and GWA genotyping have been  described in detail (14, 24, 26–28). Here, we briefly review the study cohorts  and genotyping methods. We also describe briefly each of the 7 follow-up  samples. Subject demographics and basic clinical characteristics for indi- viduals genotyped for rs563694 for each sample are described below and  summarized in Table 1. All protocols were approved by the institutional  review boards or research ethics committees at the respective institutions,  and informed consent was obtained from all subjects. FUSION GWA study. The goal of the FUSION study is to identify genet- ic variants that predispose to T2DM or that determine the variability in  T2DM-related quantitative traits. The study began as an affected sibling- pair family study (26, 27), later augmented by large numbers of cases and  controls for association analysis (14). The FUSION GWA study was per- formed using a 2-stage case-control design (14). Cases and controls were  approximately frequency matched on 5-year age category, sex, and birth  province. All stage 1 DNA samples were genotyped using the Illumina  HumanHap300 BeadChip version 1.0, resulting in data on 315,635 SNPs  that passed quality control filters (14). Genotype data for an additional  2.09 million SNPs were estimated using an imputation procedure (61).  The genotype imputation method uses stretches of chromosome shared  between  individuals  genotyped  at  relatively  low  density  in  our  studies  and individuals genotyped in greater density by the International Hap- Map Consortium (61) to estimate the missing genotypes. Comparison of  imputed and measured genotypes yielded estimated error rates of 1.46%  (Illumina) to 2.14% (Affymetrix) per allele with an average concordance of  98.5%, consistent with expectations from HapMap data (61). SNPs show- ing promising association with fasting plasma glucose in the stage 1 sam- ples were genotyped in the stage 2 DNA samples by homogeneous MassEX- TEND reaction using the MassARRAY System (Sequenom) (14). Because  treatment for T2DM affects fasting glucose concentrations, all analyses  in this report were restricted to nondiabetic subjects. Diabetes status was  confirmed by WHO criteria (62) or confirmation of treatment for diabe- tes by medical record review. Fasting plasma glucose concentrations were  available for 1,233 stage 1 and 655 stage 2 samples. Additional FUSION  samples included nondiabetic spouses or offspring from FUSION stage  1 families; fasting plasma glucose data were available for 578 individuals.  These 578 samples were genotyped using the Applied Biosystems Taq- Man allelic discrimination assays (63) and yielded 522 samples with both  genotype and fasting glucose data. These samples were integrated into the  FUSION stage 1 samples and independently analyzed to assess whether the  additional family members improved the evidence for association. We have  denoted this analysis FUSION 1 families. SardiNIA GWA study. The SardiNIA study is a longitudinal study of aging- related quantitative traits and comprises a cohort of 6,148 individuals 14  years or older recruited from 4 towns in the Lanusei Valley in Sardinia.  Data from 4,350 individuals with fasting serum glucose measurements  from this cohort were used for the GWA study; 3,331 were genotyped using  the Affymetrix 10K SNP Mapping Array, and an additional 1,412 were  genotyped using the Affymetrix 500K SNP Mapping Array (28). 356,359  SNPs passed quality control and were tested for association with fasting  serum glucose. We first used the genotyped SNPs in the 1,412 individuals  to estimate genotypes for all the polymorphic SNPs genotyped by the Hap- Map Consortium. Taking advantage of the relatedness among individuals  in the SardiNIA sample, we then conducted a second round of computa- tional analysis to impute genotypes for analysis in the 2,938 individuals  not genotyped with the 500K SNP Array. In this second round, we identi- fied large stretches of chromosome shared within each family and proba- bilistically “filled-in” genotypes within each stretch whenever 1 or more  of its carriers was genotyped with the 500K Array Set (64, 65). For these  analyses, 37 non-Sardinians and 281 of their family members (n = 318)   and 177 individuals with known diabetes were excluded from the analysis,  resulting in a final sample size of 3,855. Follow-up samples. The initial association identified in the metaanalysis  of the FUSION and SardiNIA GWA studies was also tested in a series of  follow-up samples (Table 1), 1 from FUSION described above and 6 others,  which are described briefly below. DGI.  The  DGI  case-control  GWA  sample  consists  of  1,464  cases  with  T2DM and 1,467 normoglycemic controls from Finland and Sweden and  has been previously described in detail (13). Fasting glucose measurements  were available for 1,455 nondiabetic control subjects (1,305 unrelated sub- jects and 150 siblings). Among these, fasting plasma glucose was measured in  537 subjects and fasting whole blood glucose was measured in 918 subjects.  Whole-blood glucose concentrations were converted to equivalent plasma  values using a conversion factor of 1.13 (66). All samples were genotyped  using the Affymetrix GeneChip Human Mapping 500K Array set; results of  GWA of 389,878 SNPs with fasting glucose levels (including SNP rs563694)  are publicly available at www.broad.mit.edu/diabetes/scandinavs/index. html. 1,411 individuals were available with both rs563694 genotype and  fasting glucose data. Old Order Amish subjects. The Old Order Amish study participants report- ed here were 1,655 nondiabetic subjects from Lancaster, Pennsylvania,  USA, for whom fasting plasma glucose measurements were available. These  subjects were enrolled in ongoing family studies of complex diseases and  traits (29–31). Genotyping for rs563694 was performed using the TaqMan  allelic discrimination assay (63). research article The Journal of Clinical Investigation      http://www.jci.org  � METSIM study.  Subjects  were  selected  from  the  ongoing  METSIM  study, which includes 7,000 men, aged 50 to 70 years, randomly selected  from the population of the town of Kuopio, Eastern Finland, Finland  (population 95,000). The present analysis is based on the first 4,386 non- diabetic subjects examined for METSIM with available fasting plasma  glucose values. Genotyping was performed using the TaqMan allelic dis- crimination assay (63). Caerphilly study. The Caerphilly study is a cohort study of white, Euro- pean men (n = 1,069; 97.4% born in the United Kingdom), aged 45–59 years  at entry in 1979–1983 (32), recruited from the town of Caerphilly, United  Kingdom, and 5 adjacent villages. Men were selected using the electoral role  and general practitioner records. DNA and fasting plasma glucose measure- ments used in this study relate to the first phase of data collection. BWHHS. The BWHHS consists of female participants, aged 60 to 79 years  and recruited between April 1999 and March 2001. Initially, 4,286 women  were randomly selected from 23 British towns and were interviewed and  clinically examined. They also completed medical questionnaires (33). Genotyping for the Caerphilly study and BWHHS was performed by  KBioscience using their f luorescence-based competitive allele-specific  PCR (KASPar) technology. The Inter99 Study. rs563694 was genotyped in 5,734 Danes for whom fast- ing plasma glucose values were available. This sample comprises part of the  population-based Inter99 sample of middle-aged people sampled at Research  Centre for Prevention and Health (Glostrup, Denmark; refs. 34, 35). Geno- typing was performed using TaqMan allelic discrimination (KBioscience). Statistics. Association between fasting glucose and genotypes in the FUSION  and SardiNIA studies was carried out using a regression framework in which  regression coefficients were estimated in the context of a variance compo- nent model to account for relatedness among individuals (65). For FUSION  samples, plasma glucose concentration was adjusted for sex, age, age2, birth  province, and study group. Analyses were carried out in nondiabetic individ- uals excluding those known to be taking medications that directly affect glu- cose concentration. Similarly, SardiNIA serum glucose values were adjusted  for sex, age, and age2. Because diabetes-based exclusions were based only on  medical records and SardiNIA only measured fasting serum glucose, a small  number of undiagnosed new-onset diabetes cases may have been included  in the analysis. For both studies, analyses were repeated including BMI as  an additional covariate to assess whether adiposity significantly contributed  to the evidence for association. Covariate-adjusted trait values were trans- formed to approximate univariate normality by applying an inverse normal  scores transformation; the scores were ranked, ranks were transformed into  quantiles, and quantiles were converted to normal deviates. A weighted z score–based fixed effects metaanalysis method was used to  combine results from the FUSION and SardiNIA studies. In brief, for each  SNP, a reference allele was identified and a z statistic summarizing the mag- nitude of the P value for association and direction of effect was generated for  each study. An overall z statistic was then computed as a weighted average  of the individual statistics, and a corresponding P value for that statistic was  computed. The weights were proportional to the square root of the num- ber of individuals in each study and scaled such that the squared weights  summed to 1. For the metaanalysis of the effect size, the inverse variance was  used as weights for each study. For the FUSION 1 families (FUSION stage 1  plus additional FUSION spouses and offspring) a regression-based analysis  under a variance components framework was used to appropriately account  for relationships among individuals (65). Because we did not have birth prov- ince information for the additional spouses and offspring, these analyses  were carried out adjusting for age, age2, sex, and study group only. Given the different sampling schemes, statistical analyses for the follow- up samples varied by study. The Old Order Amish samples consisted of  large Amish pedigrees, so the evidence for association between genotype  and  fasting  plasma  glucose  was  evaluated  using  variance  components  analysis implemented in SOLAR to adjust for the relatedness of study sub- jects (67, 68). Plasma glucose levels were natural logarithm transformed for  analysis, and covariates included sex, age, and age2. For the DGI study, glu- cose values were converted to z scores separately by sex, and tests for associ- ation were carried out using a regression framework with age and log(BMI)  included as covariates; genomic control was applied to account for related- ness (13). For the METSIM study, analyses were carried out identically as  in FUSION, with the exception that birth province was not included as a  covariate. For the Caerphilly and BWHHS studies, association was assessed  using a regression framework with age, age2, and BMI as covariates. For  the Inter99 study, association was assessed using a regression framework  with age and sex as covariates. Individuals with known diabetes at the time  of examination were excluded from the analyses. Results from all follow- up studies were combined in a metaanalysis as described above. Finally, a  metaanalysis that combined results from all GWA and follow-up studies  was performed as described above. Acknowledgments We would like to thank the many research volunteers who generous- ly participated in the various studies represented in this study. For  the FUSION study, we also thank Peter S. Chines, Narisu Narisu,  Andrew G. Sprau, and Li Qin for informatics and genotyping sup- port and the Center for Inherited Disease Research for the FUSION  GWA genotyping. For the SardiNIA study, we thank the mayors of  Lanusei, Ilbono, Arzana, and Elini, the head of local Public Health  Unit ASL4, and the residents of the towns for their volunteerism  and cooperation. In addition, we are grateful to the mayor and the  administration in Lanusei for providing and furnishing the clinic  site. We thank the team of physicians — Maria Grazia Pilia, Danilo  Fois, Liana Ferreli, Marcello Argiolas, Francesco Loi, and Pietro  Figus — and the nurses Paola Loi, Monica Lai, and Anna Cau, who  carried out the physical examinations and made the observations. We thank the former Medical Research Council (MRC) Epide- miology Unit (South Wales) who undertook the Caerphilly study.  The Department of Social Medicine, University of Bristol, now acts  as custodian for the Caerphilly database. We are grateful to all of  the men who participated in this study. For the BWHHS, we thank  all of the general practitioners and their staff who supported data  collection and the women who participated in the study. For the Amish studies, we thank members of the Amish com- munity for the generous donation of time to participate in these  studies and our field nurses, Amish liaisons, and clinic staff for  their extraordinary efforts. We also acknowledge Sandy Ott and  John Shelton for genotyping of Amish DNA samples. Support for this study was provided by the following: Ameri- can Diabetes Association (ADA) (1-05-RA-140 to R.M. Watanabe;  7-04-RA-111 to A.R. Shuldiner; and postdoctoral fellowships to  C.J.  Willer  and  H.M.  Stringham);  and  NIH  grants  (DK069922  and U54 DA021519 to R.M. Watanabe; DK062370 to M. Boehn- ke;  DK072193  to  K.L.  Mohlke;  DK062418  to  W-M.  Chen;  R01  DK54361, U01 HL72515, and R01 AG18728 to A.R. Shuldiner;  R01 HL69313 to B.D. Mitchell; and R01 DK068495 to K.D. Sil- ver). D.A. Lawlor is funded by a UK Department of Health career  scientist award, and N. Timpson is funded by a studentship from  the MRC of the United Kingdom. The  Inter99  Study  was  supported  by  the  European  Union  (EUGENE2, LSHM-CT-2004-512013); the Lundbeck Founda- tion Centre of Applied Medical Genomics in Personalized Dis- ease Prediction, Prevention and Care; the FOOD Study Group/ research article � The Journal of Clinical Investigation      http://www.jci.org the  Danish  Ministry  of  Food,  Agriculture  and  Fisheries  and  Ministry of Family and Consumer Affairs (2101-05-0044); and  the Danish Medical Research Council. This research was supported in part by the intramural Research  Program of the NIH, National Institute on Aging, and the NIDDK.  Additional support came from contract N01-AG-1-2109 from the  NIA intramural research program for the SardiNIA (ProgeNIA)  team;  National  Human  Genome  Research  Institute  intramural  project number 1 Z01 HG000024 (to F.S. Collins); University of  Maryland General Clinical Research Center (M01 RR 16500); Johns  Hopkins University General Clinical Research Center (M01 RR  000052); the NIDDK Clinical Nutrition Research Unit of Maryland  (P30 DK072488); and the Department of Veterans Affairs and Veter- ans Affairs Medical Center Baltimore Geriatric Research, Education  and Clinical Center (GRECC). The BWHHS receives core funding  from the United Kingdom Department of Health policy research  program. The DNA extraction and genotyping for BWHHS were  funded by the British Heart Foundation. The Caerphilly study was  funded by the MRC of the United Kingdom. Funding for the Caer- philly DNA Bank was from an MRC grant (G9824960). The United  Kingdom MRC supports work undertaken in the Centre for Causal  Analyses in Translational Epidemiology. The views expressed in this paper are those of the authors and  not necessarily those of any funding body or others whose support  is acknowledged. Those providing funding had no role in study  design, data collection and analysis, decision to publish, or prepara- tion of the manuscript. Received  for  publication  November  26,  2007,  and  accepted  in  revised form April 23, 2008. Address correspondence to: Angelo Scuteri, Unità Operativa Geria- tria, Istituto Nazionale Ricovero E Cura Anziari, Rome, Italy. Phone:  39-3334564136; Fax: 39-06-30362896; E-mail: angeloelefante@  interfree.it. Or to: Richard M. Watanabe, Keck School of Medicine  of  USC,  Department  of  Preventive  Medicine,  1540  Alcazar  St.,  CHP-220, Los Angeles, California 90089-9011, USA. Phone: (323)  442-2053; Fax: (323) 442-2349; E-mail: rwatanab@usc.edu. Wei-Min Chen and Michael R. Erdos are co–first authors.   1. Reaven, G.M. 1988. Role of insulin resistance in  human disease. Diabetes. 37:1595–1607.   2. DeFronzo,  R.A.  1987.  The  triumvirate:  B-cell,  muscle, liver. A collusion responsible for NIDDM.  Diabetes. 37:667–687.   3. National Diabetes Data Group. 1979. Classification  and diagnosis of diabetes mellitus and other catego- ries of glucose intolerance. Diabetes. 28:1039–1057.   4. [No authors listed]. 1985. Diabetes Mellitus: Report  of a WHO Study Group. World Health Organ. Tech. Rep. Ser. 727:1–113.   5. The  Expert  Committee  on  the  Diagnosis  and  Classification of Diabetes Mellitus. 1997. Report  of  the  expert  committee  on  the  diagnosis  and  classification of diabetes mellitus. Diabetes Care.  20:1183–1197.   6. DeFronzo, R.A., and Ferrannini, E. 1991. Insulin  resistance: A multifaceted syndrome responsible  for NIDDM, obesity, hypertension, dyslipidemia,  and atherosclerotic cardiovascular disease. Diabetes Care. 14:173–194.   7. Xiang, A.H., et al. 2006. Coordinate changes in plas- ma glucose and pancreatic β-cell function in Latino  women at high risk for type 2 diabetes.  Diabetes.   55:1074–1079.   8. Mason,  C.C.,  Hanson,  R.L.,  and  Knowler,  W.C.  2007. Progression to type 2 diabetes characterized  by moderate then rapid glucose increases. Diabetes.  56:2054–2061.   9. Rich, S.S. 1990. Mapping genes in diabetes. Diabetes.   39:1315–1319.   10. Ghosh, S., and Schork, N.J. 1996. Genetic analysis  of NIDDM: the study of quantitative traits. Diabetes.   45:1–14.   11. Diabetes  Prevention  Program  Research  Group.  2002. Reduction in the incidence of type 2 diabetes  with lifestyle intervention or metformin. N. Engl. J. Med. 346:393–403.   12. Tuomilehto,  J.,  et  al.  2001.  Prevention  of  type  2  diabetes  mellitus  by  changes  in  lifestyle  among  subjects with impaired glucose tolerance. N. Engl. J. Med. 344:1343–1350.   13. Diabetes Genetics Initiative of Broad Institute of  Harvard and MIT, et al. 2007. Genome-wide asso- ciation analysis identifies loci for type 2 diabetes  and triglyceride levels. Science. 316:1331–1336.   14. Scott, L.J., et al. 2007. A genome-wide association  study of type 2 diabetes in Finns detects multiple  susceptibility variants. Science. 316:1341–1345.   15. Zeggini, E., et al. 2007. Replication of genome-wide  association signals in U.K. samples reveals risk loci  for type 2 diabetes. Science. 316:1336–1341.   16. Sladek, R., et al. 2007. A genome-wide association  study identified novel risk loci for type 2 diabetes.  Nature. 445:881–885.   17. Steinthorsdottir,  V.,  et  al.  2007.  A  variant  in  CDKAL1 influences insulin response and risk of  type 2 diabetes. Nat. Genet. 39:770–775.   18. Zeggini, E., et al. 2008. Meta-analysis of genome- wide association data and large-scale replication  identifies additional susceptibility loci for type 2  diabetes. Nat. Genet. 40:638–645.   19. Beaty,  T.H.,  and  Fajans,  S.S.  1982.  Estimating  genetic and non-genetic components of variance  for fasting glucose levels in pedigrees ascertained  through  non-insulin  dependent  diabetes.  Ann. Hum. Genet. 46:355–362.   20. Boehnke,  M.,  Moll,  P.P.,  Kottke,  B.A.,  and  Weid- man,  W.H.  1987.  Partitioning  the  variability  of  fasting plasma glucose levels in pedigrees. Am. J. Epidemiol. 125:679–689.   21. Sakul, H., et al. 1997. Familiality of physical and  metabolic characteristics that predict the develop- ment of non-insulin-dependent diabetes mellitus  in Pima Indians. Am. J. Hum. Genet. 60:651–656.   22. Watanabe, R.M., et al. 1999. Familiality of quantita- tive metabolic traits in Finnish families with non- insulin-dependent diabetes mellitus. Hum. Hered.  49:159–168.   23. Henkin,  L.,  et  al.  2003.  Genetic  epidemiology  of  insulin resistance and visceral adiposity. The IRAS  family study design and methods. Ann. Epidemiol.  13:211–217.   24. Pilia, G., et al. 2006. Heritability of cardiovascular  and  personality  traits  in  6,148  Sardinians.  PLoS Genet. 2:e132.   25. Weedon, M.N., et al. 2006. A common haplotype of  the glucokinase gene alters fasting glucose and birth  weight: Association in six studies and population- genetics analyses. Am. J. Hum. Genet. 79:991–1001.   26. Valle, T., et al. 1998. Mapping genes for NIDDM.  Design of the Finland-United States Investigation  of  NIDDM  Genetics  (FUSION)  Study.  Diabetes Care. 21:949–958.   27. Silander,  K.,  et  al.  2004.  A  large  set  of  Finnish  affected sibling pair families with type 2 diabetes  suggests susceptibility loci on chromosomes 6, 11,  and 14. Diabetes. 53:821–829.   28. Scuteri, A., et al. 2007. Genome-wide association  scan shows genetic variants in the FTO gene are  associated with obesity-related traits. PLoS Genet.  3:1200–1210.   29. Hsueh,  W.-C.,  et  al.  2001.  Genome-wide  scan  of  obesity in the Old Order Amish. J. Clin. Endocrinol. Metab. 86:1199–1205.   30. Sorkin,  J.,  et  al.  2005.  Exploring  the  genetics  of  longevity in the Old Order Amish. Mech. Ageing Dev.  126:347–350.   31. Post, W., et al. 2007. Associations between genetic  variants in the NOS1AP (CAPON) gene and cardiac  repolarization in the Old Order Amish. Hum. Hered.  64:214–219.   32. The Caerphilly and Speedwell Collaborative Group.  1984. Caerphilly and Speedwell collaborative heart  disease  studies.  J. Epidemiol. Community Health.  38:259–262.   33. Lawlor, D., Bedford, C., Taylor, M., and Ebrahim, S.  2003. Geographical variation in cardiovascular dis- ease, risk factors, and their control in older women:  British Women’s Heart and Health Study. J. Epide- miol. Community Health. 57:134–140.   34. Jørgensen, M.E., et al. 2003. Obesity and central fat  pattern among Greenland Inuit and a general pop- ulation of Denmark (Inter99): relationship to met- abolic risk factors. Int. J. Obes. Relat. Metab. Disord.   27:1507–1515.   35. Glümer, C., Jørgensen, T., Borch-Johnsen, K., and  Inter99  study.  2003.  Prevalences  of  diabetes  and  impaired glucose regulation in a Danish population:  the Inter99 study. Diabetes Care. 26:2335–2340.   36. Petrolonis, A.J., et al. 2004. Enzymatic character- ization of the pancreatic islet-specific glucose-6- phosphatase-related protein (IGRP). J. Biol. Chem.  279:13976–13983.   37. Shieh, J.-J., Pan, C.-J., Mansfield, B.C., and Chou,  J.Y. 2005. In islet-specific glucose-6-phosphatase- related  protein,  the  beta  cell  antigenic  sequence  that is targeted in diabetes is not responsible for  the loss of phosphohydrolase activity. Diabetologia.  48:1851–1859.   38. van Mil, S.W.C., et al. 2004. Benign recurrent intra- hepatic cholestasis type 2 is caused by mutations in  ABCB11. Gastroenterology. 127:379–384.   39. Lang, C., et al. 2007. Mutations and polymorphisms  in  the  bile  salt  export  pump  and  the  multidrug  resistance protein 3 assocaited with drug-induced  liver injury. Pharmacogenet. Genomics. 17:47–60.   40. Devlin, B., and Roeder, K. 1999. Genomic control  for association studies. Biometrics. 55:997–1004.   41. Funk, C., Ponelle, C., Scheuermann, G., and Pantze,   M. 2001. Cholestatic potential of troglitazone as  a  possible  factor  contributing  to  troglitazone- induced hepatotoxicity: in vivo and in vitro interac- research article The Journal of Clinical Investigation      http://www.jci.org  � tion at the canalicular bile salt export pump (Bsep)  in the rat. Mol. Pharmacol. 59:627–635.   42. Staels, B., and Kuipers, F. 2007. Bile acid seques- trants and the treatment of type 2 diabetes mellitus.   Drugs. 67:1383–1392.   43. Mukherjee, R., Wagar, D., Stephens, T.A., Lee-Chan,  E., and Singh, B. 2005. Identification of CD4+ T  cell-specific  epitopes  of  islet-specific  glucose-6- phosphatase  catalytic  subunit-related  protein:  a  novel  beta  cell  autoantigen  in  type  1  diabetes.   J. Immunol. 174:5306–5315.   44. Wang, Y., et al. 2007. Deletion of the gene encoding  the  islet-specific  glucose-6-phosphatase  catalytic  subunit-related protein autoantigen results in a mild  metabolic phenotype. Diabetologia. 50:774–778.   45. Arden, S.D., et al. 1999. Molecular cloning of a pan- creatic islet-specific glucose-6-phosphatase catali- ytic subunit-related protein. Diabetes. 48:531–542.   46. Shieh, J.-J., Pan, C.-J., Mansfield, B.C., and Chou,  J.Y. 2004. The islet-specific glucose-6-phosphatase- related protein, implicated in diabetes, is a glyco- protein embedded in the endoplasmic reticulum  membrane. FEBS Lett. 562:160–164.   47. Dogra,  R.S.,  et  al.  2006.  Alternative  splicing  of  G6PC2, the gene coding for the islet-specific glu- cose-6-phosphatase catalytic subunit-related pro- tein  (IGRP),  results  in  differential  expression  in  human thymus and spleen compared with pancreas.   Diabetologia. 49:953–957.   48. Pan, C.-J., Lei, K.-J., Annabi, B., Hemrika, W., and  Chou, J.Y. 1998. Transmembrane topology of glu- cose-6-phosphatase. J. Biol. Chem. 273:6144–6148.   49. Khan, A., et al. 1990. Glucose cycling in islets from  healthy and diabetic rats. Diabetes. 39:456–459.   50. Khan, A., et al. 1989. Evidence for the presence of  glucose  cycling  in  pancreatic  islets  of  the  ob/ob  mouse. J. Biol. Chem. 264:9732–9733.   51. Khan, A., et al. 1990. Glucose cycling is markedly  enhanced in pancreatic islets of obese hyperglycemic   mice. Endocrinology. 126:2413–2416.   52. Vaulont,  S.,  Vasseur-Cognet,  M.,  and  Kahn,  A.  2000.  Glucose  regulation  of  gene  transcription.   J. Biol. Chem. 275:31555–31558.   53. Stone, L.M., Kahn, S.E., Deeb, S.S., Fujimoto, W.Y.,  and  Porte,  D.,  Jr.  1994.  Glucokinase  gene  varia- tions in Japanese-Americans with a family history  of NIDDM. Diabetes Care. 17:1480–1483.   54. Stone, L.M., Kahn, S.E., Fujimoto, W.Y., Deeb, S.S.,  and Porte, D., Jr. 1996. A variation at position -30  of the β-cell glucokinase gene promoter is associ- ated with reduced β-cell function in middle-aged  Japanese-American men. Diabetes. 45:422–428.   55. Rose, C.S., et al. 2005. A -30G>A polymorphism of  the beta-cell-specific glucokinase promoter associ- ates with hyperglycemia in the general population  of whites. Diabetes. 54:3026–3031.   56. Weedon, M.N., et al. 2005. Genetic regulation of  birth  weight  and  fasting  glucose  by  a  common  polymophism in the islet promoter of the glucoki- nase gene. Diabetes. 54:576–581.   57. Malaisse,  W.J.,  Malaisse-Lagae,  F.,  Davies,  D.R.,  Vandercammen, A., and Van Schaftingen, E. 1990.  Regulation  of  glucokinase  by  a  fructose-1-phos- phate-sensitive protein in pancreatic islets. Eur. J. Biochem. 190:539–545.   58. Frayling, T.M., et al. 2007. A common variant in the  FTO gene is associated with body mass index and  predisposes to childhood and adult obesity. Science.  316:889–894.   59. Sanna,  S.,  et  al.  2008.  Common  variants  in  the  GDF5-UQCC region are associated with variation  in human height. Nat. Genet. 40:198–203.   60. Willer, C.J., et al. 2008. Newly identified loci that  influence lipid concentrations and risk of coronary  artery disease. Nat. Genet. 40:161–169.   61. Li, Y., Willer, C.J., Ding, J., Scheet, P., and Abecasis,  G.R. 2008. Markov model for rapid haplotyping  and genotype imputation in genome wide studies.  Nat. Genet. In press.   62. [Anonymous].  1999.  Definition,  diagnosis  and  classification of diabetes mellitus and its compli- cations.  Report  of  a  WHO  Consultation.  WHO.  Geneva, Switzerland. www.diabetes.com.au/pdf/ who_report.pdf.   63. Livak, K.J. 1999. Allelic discrimination using fluo- rogenic  probes  and  the  5′  nuclease  assay.  Genet. Anal. 14:143–149.   64. Burdick,  J.T.,  Chen,  W.M.,  Abecasis,  G.R.,  and  Cheung, V.G. 2006. In silico method for inferring  genotypes in pedigrees. Nat. Genet. 38:1002–1004.   65. Chen, W.-M., and Abecasis, G.R. 2007. Family based  association tests for genome wide association scans.   Am. J. Hum. Genet. 81:913–926.   66. D’Orazio, P., et al. 2005. Approved IFCC recom- mendations on reporting results for blood glucose  (abbreviated). Clin. Chem. 51:1573–1576.   67. Blangero, J., and Almasy, L. 1997. Multipoint oligo- genic linkage analysis of quantitative traits. Genet. Epidemiol. 14:959–964.   68. Almasy, L., and Blangero, J. 1998. Multipoint quan- titative-trait linkage analysis in general pedigrees.  Am. J. Hum. Genet. 62:1198–1211.