Identification of Novel Candidate Genes for Type 2 Diabetes From a Genome-Wide Association Scan in the Old Order Amish Evidence for Replication From Diabetes-Related Quantitative Traits and From Independent Populations Evadnie Rampersaud, 1 Coleen M. Damcott, 1 Mao Fu, 1 Haiqing Shen, 1 Patrick McArdle, 1 Xiaolian Shi, 1 John Shelton, 1 Jing Yin, 1 Yen-Pei C. Chang, 1 Sandra H. Ott, 1 Li Zhang, 1 Yiju Zhao, 1 Braxton D. Mitchell, 1 Jeffery O’Connell, 1 and Alan R. Shuldiner 1,2 OBJECTIVE—We sought to identify type 2 diabetes suscepti- bility genes through a genome-wide association scan (GWAS) in the Amish. RESEARCH DESIGN AND METHODS—DNA from 124 type 2 diabetic case subjects and 295 control subjects with normal glucose tolerance were genotyped on the Affymetrix 100K single nucleotide polymorphism (SNP) array. A total of 82,485 SNPs were tested for association with type 2 diabetes. Type 2 diabetes– associated SNPs were further prioritized by the following: 1) associations with 5 oral glucose tolerance test (OGTT) traits in 427 nondiabetic Amish subjects, and 2) in silico replication from three independent 100L SNP GWASs (Framingham Heart Study Caucasians, Pima Indians, and Mexican Americans) and a 500K GWAS in Scandinavians. RESULTS— The strongest association (P � 1.07 � 10�5) was for rs2237457, which is located in growth factor receptor– bound protein 10 (Grb10), an adaptor protein that regulate insulin receptor signaling. rs2237457 was also strongly associated with OGTT glucose area under the curve in nondiabetic subjects (P � 0.001). Of the 1,093 SNPs associated with type 2 diabetes at P � 0.01, 67 SNPs demonstrated associations with at least one OGTT trait in nondiabetic individuals; 80 SNPs were nominally associ- ated with type 2 diabetes in one of the three independent 100K GWASs, 3 SNPs (rs2540317 in MFSD9, rs10515353 on chromo- some 5, and rs2242400 in BCAT1 were associated with type 2 diabetes in more than one population), and 11 SNPs were nominally associated with type 2 diabetes in Scandinavians. One type 2 diabetes–associated SNP (rs3845971, located in FHIT) showed replication with OGTT traits and also in another population. CONCLUSIONS—Our GWAS of type 2 diabetes identified sev- eral gene variants associated with type 2 diabetes, some of which are worthy of further study. Diabetes 56:3053–3062, 2007 T ype 2 diabetes, a complex disease that is char- acterized by insulin resistance and impaired �-cell function, represents a serious global pub- lic health problem, with more than 100 million people affected worldwide. While the primary molecular defects in type 2 diabetes remain largely unknown, it is clear that both genetic and environmental risk factors (including diet and physical inactivity) play critical roles. More than 20 genome-wide linkage scans of type 2 diabe- tes have been published, with evidence for linkage re- ported to a number of loci, including regions on chromosomes 1, 3, 8, 10, 12, 14, and 20 (1– 8). Of the numerous candidate genes studied for their functional role in pancreatic �-cell function, insulin action, or energy metabolism, as well as positional candidate genes identi- fied under linkage peaks, very few have variants that are consistently associated with type 2 diabetes. Indeed, com- mon variants in only a few genes (PPAR�, KCNJ11, CALPN10, TCF7L2, and HNF4A) have been replicated in multiple populations (9). The Old Order Amish are a closed founder population who emigrated from Switzerland in the early 1700s. They are a well-suited population for carrying out genetic studies since they live a relatively homogeneous lifestyle and maintain extensive family history records. The Amish Family Diabetes Study (AFDS) was initiated in 1995 with the goal of identifying the genetic determinants of type 2 diabetes (10). The sibling relative risk (�s) of type 2 diabetes in the Amish is 3.28 (95% CI 1.58 – 6.80), similar to that observed in other Caucasian populations. Genome- wide linkage analysis of type 2 diabetes and impaired glucose tolerance conducted in AFDS pedigrees (6) re- vealed regions on chromosomes 1q and 14q, both of which have been implicated in linkage scans from other popula- tions (1–5,7). Specific variants in several well-replicated type 2 diabetes susceptibility genes are associated with type 2 diabetes in the Amish, including TCF7L2 rs7903146 From the 1Division of Endocrinology, Diabetes and Nutrition, University of Maryland, Baltimore, Maryland; and the 2Geriatric Research and Education Clinical Center, Baltimore Veterans Administration Medical Center, Balti- more, Maryland. Address correspondence and reprint requests to Alan R. Shuldiner, MD, Division of Endocrinology, Diabetes and Nutrition, University of Maryland School of Medicine, 660 West Redwood St., Room 494, Baltimore, MD 21201. E-mail: ashuldin@medicine.umaryland.edu. Received for publication 3 April 2007 and accepted in revised form 5 September 2007. Published ahead of print at http://diabetes.diabetesjournals.org on 10 Sep- tember 2007. DOI: 10.2337/db07-0457. Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/db07-0457. AFDS, Amish Family Diabetes Study; DGI, Broad-Lund-Novartis Diabetes Genome Initiative; GAUC, glucose area under the curve; GWAS, genome-wide association scan; HOMA-IR, homeostasis model assessment of insulin resis- tance; IAUC, insulin area under the curve; ISI, insulinogenic index; LD, linkage disequilibrium; NGT, normal glucose tolerant; OGTT, oral glucose tolerance test; SNP, single nucleotide polymorphism. © 2007 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ORIGINAL ARTICLE DIABETES, VOL. 56, DECEMBER 2007 3053 (odds ratio 1.60, P � 0.008) (11) and HNF4A rs2425640 (1.60, P � 0.03) (12). These findings suggest that the common type 2 diabetes gene variants in the Amish will likely be relevant to more outbred Caucasian populations. Increased knowledge of common variation in the human genome learned as part of the HapMap initiative, coupled with advances in technologies, make possible the geno- typing of thousands of single nucleotide polymorphisms (SNPs) in genome-wide association scans (GWAS). This is a powerful approach for identifying novel susceptibility genes for complex diseases (13,14). Recently, four GWAS studies of type 2 diabetes have identified variants at several novel loci, including SLC30A8, IGF2BP2, CK- DAL1, CDKN2A/CKDN2B, and HHEX/IDE, that show strong replicated association with type 2 diabetes (15–18). In this article, we report results from a GWAS of type 2 diabetes in the Amish using the Affymetrix 100K SNP genotyping platform. We further characterize our findings using diabetes-related quantitative traits measured in non- diabetic Amish individuals. Lastly, we interpret the results of this scan in the context of three recently completed 100K GWAS studies for type 2 diabetes, as part of the Type 2 Diabetes 100K GWAS Consortium, along with a publicly available 500K GWAS of type 2 diabetes recently per- formed in a Scandinavian population. RESEARCH DESIGN AND METHODS Study population and phenotype assessment. Individuals with type 2 diabetes were identified from the AFDS. Details of the AFDS have been previously described (10). Phenotypic characterization of participants in- cluded medical and family history, anthropometry, and a 3-h 75-g oral glucose tolerance test (OGTT) with insulin levels. We based our primary analyses on 124 type 2 diabetic case and 295 normal glucose tolerant (NGT) control subjects. Type 2 diabetes was defined by fasting plasma glucose level (�7 mmol/l), 2-h OGTT plasma glucose level (�11.1 mmol/l), random plasma glucose level (�11.1 mmol/l), the use of insulin or prescription oral glucose- lowering agents, or a diagnosis of diabetes documented by a physician. To minimize potentially misclassifying subjects with type 1 diabetes as having type 2 diabetes, case subjects with age of diagnosis �35 years were excluded. NGT control subjects were aged �38 years at the time of study and were selected based on fasting plasma glucose level (�6.1 mmol/l) and 2-h OGTT plasma glucose level (�7.8 mmol/l). We performed secondary quantitative analyses of our mostly highly asso- ciated signals (P � 0.01) in a set of 427 nondiabetic Amish study participants, 132 of whom had impaired glucose tolerance and 295 of whom were part of the NGT control group used in our primary analysis. We estimated the mean levels of two OGTT-derived quantitative glucose traits (fasting glucose and glucose area under the curve [GAUC]) and three insulin traits (insulinogenic index [ISI], insulin area under the curve [IAUC], and homeostasis model assessment of insulin resistance [HOMA-IR]) according to the SNP genotypes in these individuals. Total GAUC and IAUC were calculated based on measurements at 0, 30, 60, 90, 120, 150, and 180 min using the trapezoidal method. The ISI was calculated as (insulin at 30 min � fasting insulin)/ (glucose at 30 min � fasting glucose). HOMA-IR was calculated as fasting insulin (mU/l) � fasting glucose (mmol/l)/22.5. Table 1 describes the charac- teristics of this sample. The study protocol was approved by the institutional review board at the University of Maryland School of Medicine, and informed consent was obtained from each study participant. Genotyping. Genomic DNA from leukocytes were genotyped using the Affymetrix GeneChip Mapping 100K array set, which consists of two microar- ray chips (XbaI and HindIII) (Affymetrix, Santa Clara, CA). Total genomic DNA (250 ng) was digested with XbaI or HindIII restriction enzymes and processed according to the Affymetrix protocol. The GeneChip Genotyping Analysis Software (GTYPE 4.0) was used to generate dynamic modeling algorithm– derived genotypes that were reanalyzed with the BRLMM (Bayes- ian RLMM) genotype calling algorithm (confidence threshold of 0.33) to improve the proportion of heterozygote calls (19). As an initial quality-control measure, BRLMM-generated chip files with call rates �90% for both enzymes across all SNPs were excluded. The resulting median call rate across all of the remaining 419 case-control samples was 97.5% (97.6% for XbaI and 97.4% for HindIII). We further removed individual SNPs with genotype call rates �90%, monomorphic SNPs and SNPs with minor allele frequency � 5%, and those deviating from Hardy-Weinberg equilibrium in control subjects (P � 0.001). The number of monomorphic and low-frequency SNPs (n � 26,816) in the Amish was not appreciably different from that observed in more outbred Caucasians of the HapMap CEU sample. For this report, we focused our analysis on the 82,485 autosomal SNPs that passed our quality-control standards. The concordance rate for 11 quality-control samples that were run twice on the Affymetrix GeneChip mapping panel was 97.5%. We also calculated a cross-platform concordance rate of 98% for 419 samples in which 61 SNPs were genotyped using the Affymetrix GeneChip Mapping 100K panel and an independent Illumina 1536-plex GoldenGate assay. Supplementary Table 3 (available in an online appendix at http://dx.doi.org/10.2337/db07-0457) sum- marizes the quality checks and informativeness of the data. Association testing and SNP prioritization scheme. Our GWAS analysis and SNP prioritization scheme is shown in Fig. 1. We selected the SNPs most highly associated with type 2 diabetes in our Amish case-control dataset based on P value rankings (P value cutoff �0.01) and then used two complementary approaches to further prioritize them. In one approach, we evaluated the most highly type 2 diabetes–associated SNPs for association with diabetes-related quantitative traits in an expanded set of 427 nondiabetic Amish subjects, 295 of whom were NGT control subjects from the primary type 2 diabetes association analysis (internal consistency). In a parallel approach, we as- TABLE 1 Description of sample characteristics for type 2 diabetes GWAS in the Amish Characteristics Type 2 diabetic case and NGT control subject dataset Type 2 diabetic case subjects NGT control subjects n 124 295 Male subjects (%) 33 52 Age (years) 51.3 � 10.5 64.4 � 12.9 BMI (kg/m2) 29.3 � 5.8 27.4 � 4.7 Characteristics OGTT-derived quantitative trait dataset All Men Women n 427 200 227 Age (years) 51.9 � 11.9 52.2 � 11.9 51.7 � 11.9 BMI (kg/m2) 27.7 � 5.0 26.4 � 4.0 28.9 � 5.5 Fasting glucose (mmol/l) 5.09 � 0.45 5.12 � 0.47 5.07 � 0.43 GAUC (mmol � l�1 � h�1) 379.2 � 68.0 362.6 � 65.3 393.9 � 67.1 IAUC (mU� l�1 � h�1) 137.9 � 88.9 108.5 � 60.6 164.5 � 101.3 HOMA-IR (mU per mmol/l2) 2.6 � 1.7 2.6 � 2.2 2.6 � 1.1 ISI (units/g) 0.9 � 1.5 0.85 � 1.9 0.93 � 1.0 Data are means � SD. All were nondiabetic subjects. A TYPE 2 DIABETES GWAS IN THE AMISH 3054 DIABETES, VOL. 56, DECEMBER 2007 sessed replication of the most highly associated type 2 diabetes–associated SNPs in the Amish in four independent GWASs from other populations (external replication). Type 2 diabetes association analysis. We performed case-control associa- tion analysis using a variance component approach as implemented in SOLAR software (20). Using a liability threshold model, we modeled the probability that the individual was a case or control subject as a function of the individual’s age, sex, and genotype, conditional on the correlations in pheno- type among relative pairs. Statistical testing was performed using a likelihood ratio test, in which we compared the likelihood of the data under a model in which the genotype effect was estimated with the likelihood of a nested model in which the genotype effect was constrained to be zero. Odds ratios (ORs) were computed from variance components models. We chose to report the additive model as our primary analysis. Supplementary analyses using a dominant or recessive model did not yield any SNP showing genome-wide significance. Of 82,485 SNPs, 611 had P � 0.01 under a dominant model and 569 had P � 0.01 under the recessive model. Our complete dataset with results from all models is available online (available at http://www.medschool.umary land.edu/amishstudies/index.asp). Pairwise linkage disequilibrium (LD) corre- lation statistics (r 2) were computed using the Helixtree software, version 5.0.2 (GoldenHelix, Bozeman, MT). Quantitative trait analysis. For quantitative trait analyses performed in nondiabetic Amish subjects, we used the measured genotype approach, in which we estimated the likelihood of an additive genetic model given the pedigree structure (21). Before analysis, all insulin traits (IAUC, ISI, and HOMA-IR) were transformed by their natural logarithm to reduce skewness. Parameter estimates were obtained by maximum likelihood methods, and the significance of association was tested by the likelihood ratio test. Within each model, we simultaneously estimated the effects of age and sex. These analyses were performed using the SOLAR program (20). Power calculations. Power calculations, based on the genetic power calcu- lator of Purcell et al. (22), indicated that our sample would provide 80% power to detect a diabetes susceptibility allele having a genotype relative risk of 1.8 (for allele frequency of 30%, 124 case and 295 control subjects, 8% population prevalence of diabetes, assuming a multiplicative model) and 80% power to detect a quantitative trait loci accounting for 4% or higher of the trait variance for a continuously distributed phenotype (427 subjects). In silico replication samples. We considered whether our best type 2 diabetes association signals (P value cutoff �0.01) replicated in at least one of three distinct populations (Framingham Caucasians, Mexican Americans, and Pima Indians), each with different study designs but performed using the same Affymetrix 100K genotyping platform. Descriptions of each of the Type 2 Diabetes 100K GWAS Consortium study populations are provided in accom- panying articles (23–25) and in supplementary Table 4. We directly checked whether any of the 1,093 SNPs with the best type 2 diabetes association signals (P � 0.01) in the Amish were also associated with type 2 diabetes based on generalized estimating equations and family-based association tests in the Framingham Heart Study, Fisher’s exact allelic association test in the Mexican-American study, and case-control and sib-based association tests in the Pima Indian study. We also utilized publicly available prereleased data (March 2007) from a type 2 diabetes GWAS carried out in a Scandinavian cohort of 1,464 type 2 diabetic case and 1,467 matched control subjects and genotyped using the Affymetrix 500K platform by the Broad-Lund-Novartis Diabetes Genetics Initiative (DGI) (available at http://www.broad.mit.edu/ diabetes/) (18). We specifically checked replication of 295 of 1,093 of our most highly type 2 diabetes–associated SNPs that were present on both 100K and 500K Affymetrix genotyping arrays. Since LD structure may differ across populations, and to limit multiple comparisons, we defined replication only if the same SNP was associated with type 2 diabetes at P � 0.05 with an OR in the same direction (i.e., reflective of the same allelic risk). RESULTS Following quality-control and Hardy-Weinberg equilibrium checks, 82,485 informative SNPs were included in our analyses. The median physical inter-SNP distance was 11.3 kb, and the average distance between SNPs was 29 kb. Under the additive model, a total of 1,093 SNPs, some of which were in LD, were associated (P � 0.01) with type 2 diabetes (Fig. 2). The 50 most strongly type 2 diabetes– associated SNPs (i.e., lowest P values) are shown in Table 2. The complete dataset is available online (available at FIG. 1. Schematic diagram of analysis and SNP prioritization approach for a 100K type 2 diabetes GWAS in the Amish. FASG, fasting glucose during an OGTT; FHS, Framingham Heart Study. E. RAMPERSAUD AND ASSOCIATES DIABETES, VOL. 56, DECEMBER 2007 3055 http://www.medschool.umaryland.edu/amishstudies/ afds.asp). No SNP was associated with type 2 diabetes at a conservative Bonferroni-corrected level. The strongest association (P � 1.07 � 10�5) was for rs2237457 on chromosome 7, which is located in intron 4 of growth factor receptor– bound protein 10 (Grb10), an adaptor protein known to regulate signaling of insulin and IGF receptors (26 –28). In addition to Grb10, 15 SNPs were associated with type 2 diabetes at P � 1 � 10�4 (Fig. 2 and Table 2). These SNPs are located in or near MSH6 (chromosome 2), PRKG2 (chromosome 4), COL13A1 (chromosome 10), MTHFSD (chromosome 16), and SPECC1 (chromosome 17), none of which are obvious candidate genes for type 2 diabetes. Adjustment for BMI did not have a large impact on the strength of the associations of these SNPs with type 2 diabetes (Table 2). As a measure of internal consistency, we tested whether the 1,093 SNPS associated with type 2 diabetes (P � 0.01) were also associated with OGTT-derived quantitative traits in nondiabetic individuals. In these analyses, we consid- ered two OGTT glucose traits (fasting glucose and GAUC) and three OGTT insulin traits (IAUC, HOMA-IR, and ISI), with P � 0.01 as our threshold for significance. Thirty- eight nonredundant (r2 � 0.80) type 2 diabetes–associated SNPs were also associated with at least one glucose trait and showed the same allelic association as that for diabe- tes (i.e., the diabetes risk allele was also associated with higher glucose levels), while 29 nonredundant type 2 diabetes–associated SNPs were also associated with at least one insulin-related trait (Fig. 1; Table 3). Of the top 16 SNPs associated with type 2 diabetes at P � 1 � 10�4, rs2237457 in Grb10 was the only one also associated with an OGTT trait (P � 0.001 for GAUC). Two perfectly correlated (r2 � 1) type 2 diabetes–associated SNPs in ADAMTS1 (chromosome 5) (P � 0.004 – 0.005) were asso- ciated with one glucose trait (P � 0.006 for GAUC) and one insulin trait (P � 0.007 for IAUC). We next sought to determine which of our 1,093 most highly type 2 diabetes–associated SNPs were also asso- ciated with type 2 diabetes in any of three independent populations for which the same 100K Affymetrix plat- form was used or in the DGI Scandinavian population for which the 500K Affymetrix platform was used. We identified 80 nonredundant SNPs for which the same risk allele was also associated with type 2 diabetes in one of the three studies from the Type 2 Diabetes 100K GWAS Consortium (P � 0.05) and 11 nonredundant SNPs that showed consistent association in the DGI sample (P � 0.05) (Fig. 1; supplementary Table 3). In total, three SNPs demonstrated associations in the Amish as well as in two independent populations. The T-allele for rs2540317 in MFSD9 on chromosome 2 was associated with decreased risk of type 2 diabetes in the Amish (OR 0.72, P � 0.007) and showed nominal asso- ciation in the Pima Indian dataset (case-control OR 0.67, P � 0.042; sib-based OR 0.50, P � 0.043; and summary OR 0.63, P � 0.016) and also in the Mexican-American sample (case-control OR 0.75, P � 0.047). The G-allele in rs10515353 on chromosome 5 was associated with de- creased risk of type 2 diabetes in the Amish (OR 0.61, P � 0.005) and also with decreased type 2 diabetes risk in Mexican-American (0.69, P � 0.035) and DGI (0.79, P � 0.007) samples. The T-allele in rs2242400 in BCAT1 on chromosome 10 was associated with decreased risk of type 2 diabetes in the Amish (0.71, P � 0.004) and also in the Pima Indian dataset (sib-based OR 0.66, P � 0.019; summary OR 0.78, P � 0.034) and the Mexican-American dataset (OR 0.67, P � 0.009); borderline association was also seen in the DGI sample (0.86, P � 0.051). The direction of effect was the same for all studies. FIG. 2. SNP association P values (<0.01) across all autosomal chromosomes. A TYPE 2 DIABETES GWAS IN THE AMISH 3056 DIABETES, VOL. 56, DECEMBER 2007 Table 4 highlights our most consistent overall findings. We present 21 type 2 diabetes–associated SNPs in the Amish (P � 0.005) that also demonstrated either 1) association with a diabetes-related quantitative trait (P � 0.005) in the Amish or 2) in silico replication of type 2 diabetes association in one independent population (P � 0.005). Of interest, the T-allele in rs3845971 in FHIT was associated with increased risk of type 2 diabetes in the Amish (OR 1.42, P � 0.004) and also in Mexican Americans (1.46, P � 0.004) and with increased GAUC (P � 4.0 � 10�4) in nondiabetic Amish subjects. DISCUSSION In this article, we described the results of a GWAS of type 2 diabetes of 82,485 SNPs in the Old Order Amish, a genetically closed founder population with a homoge- neous lifestyle. We reasoned that this population is likely TABLE 2 Fifty SNPs most highly associated with type 2 diabetes from Amish GWAS SNP Chromo- some Position* Gene† Alleles 1/2 Strand Case subjects (n) Control subjects (n) Allele 2 case subjects Allele 2 control subjects Type 2 diabetes P‡ OR§ Type 2 diabetes P (BMI)� rs2237457 7 50693638 Grb10 A/G � 124 293 0.53 0.68 1.07 � 10�5 0.61 1.46 � 10�5 rs980720 4 82272087 PRKG2 A/G � 122 287 0.80 0.90 1.25 � 10�5 0.52 4.87 � 10�5 rs10509199 10 65295820 G/T � 124 295 0.31 0.46 1.68 � 10�5 0.62 4.46 � 10�5 rs1373147 17 20147237 SPECC1 A/T � 124 293 0.47 0.60 2.79 � 10�5 0.63 5.26 � 10�6 rs4082516 10 71374662 COL13A1 C/G � 122 293 0.77 0.88 3.19 � 10�5 0.53 1.27 � 10�5 rs10509195 10 65193372 A/C � 118 281 0.42 0.28 3.30 � 10�5 1.62 7.1 � 10�5 rs1395931 2 123535443 A/G 124 293 0.68 0.58 6.36 � 10�5 1.57 2.90 � 10�5 rs3136279 2 47871272 MSH6 G/T � 124 294 0.20 0.11 6.81 � 10�5 1.80 9.2 � 10�5 rs1446732 2 134374210 G/T � 121 295 0.51 0.37 7.38 � 10�5 1.55 0.0001 rs10485249 6 70161908 C/G 124 295 0.82 0.92 7.96 � 10�5 0.54 0.0004 rs2703813 17 20055907 SPECC1 A/G � 123 285 0.51 0.64 8.12 � 10�5 0.64 2.36 � 10�5 rs1916412 10 65340801 C/T � 124 293 0.62 0.49 8.35 � 10�5 1.54 0.0002 rs1916411 10 65340839 C/G � 123 292 0.38 0.51 8.48 � 10�5 0.65 0.0002 rs3751797 16 85124993 MTHFSD A/T � 123 285 0.76 0.64 8.74 � 10�5 1.60 0.0005 rs930621 2 134418548 C/T 122 290 0.54 0.41 8.74 � 10�5 1.55 0.0001 rs10509201 10 65342248 A/G � 123 293 0.62 0.49 8.80 � 10�5 1.53 0.0002 rs2158473 17 20079682 SPECC1 C/T 119 292 0.51 0.39 0.0001 1.54 3.00 � 10�5 rs2502497 6 75399083 G/T 116 279 0.05 0.10 0.0001 0.43 0.0002 rs430123 5 106109632 A/G � 114 272 0.16 0.27 0.0002 0.61 0.0002 rs10504553 8 75038957 TCEB1 A/G � 115 275 0.77 0.89 0.0002 0.56 0.0003 rs9287428 2 133912538 FLJ34870 C/T 124 287 0.83 0.71 0.0002 1.62 0.0002 rs10507601 13 55047379 A/G 123 294 0.04 0.12 0.0002 0.44 0.0006 rs994952 6 78319674 A/G 120 287 0.52 0.42 0.0002 1.52 0.0002 rs7604549 2 133923233 FLJ34870 A/G � 123 294 0.26 0.40 0.0002 0.65 0.0002 rs2737245 8 116727757 TRPS1 A/C � 113 276 0.58 0.71 0.0002 0.67 0.0003 rs1513287 3 114754898 SIDT1 A/G 120 289 0.58 0.44 0.0002 1.49 0.0003 rs7817780 8 119486147 SAMD12 C/T � 120 294 0.91 0.80 0.0002 1.85 0.0002 rs297765 20 4435111 A/G 124 295 0.75 0.63 0.0002 1.55 0.0004 rs4410442 3 112434500 A/G � 123 294 0.48 0.34 0.0002 1.52 0.0008 rs1351916 2 139479918 A/T � 122 288 0.37 0.56 0.0002 0.67 0.0006 rs1507666 2 123532455 A/T � 119 285 0.44 0.56 0.0002 0.66 0.0002 rs10498934 6 83516566 C/T � 124 293 0.57 0.69 0.0002 0.66 0.0004 rs721729 3 174671967 A/G � 119 285 0.43 0.56 0.0002 0.65 0.0002 rs1492908 3 159869201 GFM1 A/G 122 291 0.20 0.29 0.0002 0.62 0.0004 rs10484725 6 78325880 A/G 124 294 0.19 0.11 0.0002 1.75 0.0002 rs10487440 7 125763796 G/T � 121 284 0.17 0.23 0.0002 0.59 0.0002 rs10498881 6 72999327 RIMS1 C/G � 121 289 0.88 0.95 0.0002 0.48 0.0019 rs9317821 13 69008405 C/T � 123 290 0.52 0.64 0.0002 0.67 0.0008 rs10487442 7 125850492 A/G 124 295 0.83 0.76 0.0003 1.68 0.0003 rs723397 5 107375942 FBXL17 C/T 123 295 0.24 0.36 0.0003 0.65 0.0014 rs6897150 5 165833376 A/G � 123 294 0.26 0.14 0.0003 1.60 0.0003 rs1458405 6 78321198 G/T � 112 270 0.51 0.40 0.0003 1.54 0.0002 rs10521205 17 12809867 KIAA0672 A/G � 122 292 0.16 0.26 0.0003 0.60 0.0007 rs1023738 3 174685924 C/T � 123 294 0.44 0.55 0.0003 0.67 0.0002 rs1367313 3 159848232 GFM1 C/T � 124 295 0.79 0.71 0.0003 1.57 0.0005 rs10521204 17 12809385 KIAA0672 A/C 124 293 0.83 0.74 0.0003 1.65 0.0007 rs2034531 4 137364060 A/G � 123 289 0.80 0.88 0.0003 0.59 0.0006 rs1282090 3 112779804 CD96 A/G 124 295 0.82 0.68 0.0004 1.58 0.0011 rs9328099 6 2382544 A/G � 122 294 0.74 0.84 0.0004 0.63 0.0004 rs10497567 2 181111709 KIAA1604 A/G 124 295 0.93 0.85 0.0004 1.98 0.0005 *Genome build 36.1. †Genic region that contains associated SNPs. ‡P values derived using variance components analysis under an additive genetic model, adjusted for age, sex, and family structure. §OR calculated from a liability threshold model in SOLAR and estimated as allele 2 versus allele 1. �P values derived using variance components analysis under an additive genetic model, adjusted for age, sex, BMI, and family structure. Our complete dataset with results from all models are available online (available at http://www.medschool.umaryland.edu/ amishstudies/index.asp). E. RAMPERSAUD AND ASSOCIATES DIABETES, VOL. 56, DECEMBER 2007 3057 TABLE 3 SNPs associated with type 2 diabetes (P � 0.01) and at least one OGTT-derived trait (P � 0.01) in nondiabetic Amish subjects SNP Chromo- some Position* Gene† Strand Alleles 1/2 Frequency allele 2 Type 2 diabetes OGTT trait analysis P‡ OR§ Trait Mean 11� Mean 12� Mean 22� P‡ rs667222 1 55090696 DHCR24 � A/G 0.75 0.005 0.71 GAUC 212.21 199.85 192.58 0.008 rs6588186 1 66319597 PDE4B � A/G 0.78 0.008 1.45 GAUC 192.09 185.94 201.07 0.002 rs570021 1 71217978 PTGER3 A/T 0.27 0.007 0.70 IAUC 550.64 613.59 690.13 0.007 rs6424414 1 71228058 PTGER3 C/T 0.27 0.007 0.70 IAUC 548.85 614.86 683.35 0.007 rs211706 1 75802445 SLC44A5 � A/G 0.08 0.002 0.52 Fasting glucose 2.83 2.74 2.68 0.006 rs10493580 1 76277500 LOC729766 C/G 0.15 0.003 0.60 GAUC 199.25 186.85 189.20 0.005 rs1030414 1 81241030 � C/T 0.07 0.005 0.51 GAUC 197.22 182.88 154.78 0.003 rs2257963 1 105672501 � C/T 0.61 0.009 0.74 GAUC 208.52 195.36 190.49 0.009 rs1516150 1 105758114 A/G 0.39 0.010 1.36 GAUC 190.76 195.70 208.75 0.009 rs2576216 1 215202099 ESRRG A/G 0.16 0.006 0.64 GAUC 200.81 187.83 190.21 0.002 rs2818781 1 215203396 ESRRG C/T 0.84 0.003 1.61 GAUC 168.17 189.21 201.24 2.2 � 10�4 rs2576212 1 215204637 ESRRG � A/G 0.84 0.005 1.58 GAUC 178.70 188.25 200.71 0.001 rs1343747 1 240794106 A/C 0.23 0.002 1.43 HOMA-IR 2.22 2.53 2.72 0.002 IAUC 555.29 646.59 694.29 0.002 rs10495824 2 34604049 � C/T 0.76 0.008 1.41 IAUC 677.25 635.10 545.83 0.002 rs10490049 2 40426854 SLC8A1 � A/C 0.17 0.002 1.54 Fasting glucose 2.79 2.88 2.89 0.001 GAUC 192.09 204.49 216.33 0.001 rs897097 2 55539162 A/G 0.61 0.001 1.46 HOMA-IR 2.09 2.37 2.49 0.010 rs2540317 2 102716239 MFSD9 � C/T 0.78 0.007 0.72 GAUC 216.62 199.07 189.80 3.4 � 10�4 rs4324336 2 105593460 � A/T 0.78 0.005 1.44 Fasting glucose 2.78 2.76 2.85 0.002 rs2321201 2 134136406 � C/T 0.37 0.006 1.35 IAUC 539.03 627.77 628.08 0.006 rs10510530 3 22459609 A/T 0.86 0.003 0.65 Fasting glucose 2.90 2.86 2.79 0.007 rs3845971 3 59975712 FHIT C/T 0.72 0.004 1.42 GAUC 185.46 188.48 203.71 1.0 � 10�4 rs1373340 3 66984977 � A/C 0.86 0.007 0.67 HOMA-IR 1.89 2.14 2.43 0.006 IAUC 529.61 519.62 610.68 0.009 rs2587015 3 147740690 PLSCR1 � C/T 0.76 0.007 1.40 IAUC 751.72 609.81 547.07 0.001 rs2587014 3 147741153 PLSCR1 A/G 0.24 0.009 0.72 IAUC 548.12 610.88 743.82 0.001 rs2587012 3 147741360 PLSCR1 � G/T 0.24 0.009 0.72 IAUC 549.20 618.96 758.40 4.0 � 10�4 rs837678 3 191168575 LEPREL1 � C/T 0.23 0.007 1.39 Fasting glucose 2.79 2.82 3.02 0.008 rs10517351 4 35352269 � A/G 0.83 0.010 0.71 ISI 0.41 0.47 0.59 0.010 rs4128879 4 162098442 � A/G 0.85 0.007 0.67 ISI 1.46 0.62 0.52 0.009 rs10521005 5 33638471 ADAMTS1 � C/T 0.46 0.004 0.74 IAUC 659.01 568.57 550.30 0.007 rs9292501 5 33639167 ADAMTS1 A/G 0.45 0.005 0.75 GAUC 201.46 197.18 186.88 0.006 IAUC 655.02 570.18 550.12 0.008 rs9292502 5 33639516 ADAMTS1 A/G 0.46 0.003 0.74 GAUC 201.78 197.24 186.94 0.006 IAUC 661.14 571.39 551.03 0.006 rs709668 5 96199942 � C/T 0.28 0.009 0.73 GAUC 201.47 191.03 183.01 0.002 rs950664 5 103381835 C/T 0.62 0.009 1.35 GAUC 188.26 190.43 205.91 1.4 � 10�4 rs990133 5 108073268 C/T 0.86 0.004 1.58 GAUC 169.38 189.40 197.90 0.006 rs7720835 5 117269389 C/T 0.20 0.006 1.41 HOMA-IR 2.26 2.48 3.25 0.002 rs4365869 5 117282887 � A/T 0.79 0.009 0.72 HOMA-IR 3.24 2.46 2.27 0.003 rs1423003 5 147427736 SPINK5 � A/G 0.45 0.004 1.38 HOMA-IR 2.51 2.32 2.12 0.008 rs1862446 5 147460749 SPINK5 A/C 0.45 0.006 1.37 HOMA-IR 2.50 2.32 2.12 0.009 rs1422930 5 167330171 ODZ2 C/T 0.09 0.003 0.52 HOMA-IR 2.27 2.72 1.45 0.002 rs7770797 6 6566961 LY86 � C/T 0.43 0.005 0.73 IAUC 536.16 579.90 649.95 0.010 rs10484908 6 108410287 C/G 0.11 0.009 1.53 GAUC 192.64 207.68 212.03 0.001 rs10237701 7 8860569 C/T 0.45 0.009 1.33 ISI 0.67 0.49 0.51 0.008 rs10487563 7 48905713 � C/T 0.23 0.002 0.68 ISI 0.49 0.60 0.80 1.4 � 10�4 rs2237457 7 50693638 Grb10 � A/G 0.63 1.1 � 10�5 0.61 GAUC 214.90 197.81 191.93 0.001 rs10499761 7 56161137 LOC650200 � A/G 0.75 0.007 1.42 GAUC 178.50 192.32 201.72 0.001 rs3753107 7 91467087 AKAP9 � C/T 0.77 0.006 0.72 HOMA-IR 2.88 2.50 2.27 0.007 rs10488510 7 91498161 AKAP9 G/T 0.23 0.006 1.39 HOMA-IR 2.29 2.54 2.76 0.010 rs647055 7 114159262 A/G 0.25 0.007 1.41 IAUC 551.83 620.21 780.97 3.8 � 10�4 ISI 0.49 0.65 0.68 0.001 rs10488284 7 119742580 KCND2 � A/G 0.15 0.001 0.60 HOMA-IR 2.49 2.14 2.53 0.003 rs192392 8 53094411 A/C 0.54 0.004 0.74 IAUC 650.81 600.14 540.14 0.007 rs2450148 8 53112873 � C/T 0.47 0.010 1.31 IAUC 537.54 580.77 669.94 0.002 rs10504133 8 53152586 C/T 0.51 0.002 0.72 IAUC 659.06 595.05 525.29 0.001 rs7001645 8 105192526 RIMS2 C/G 0.44 0.007 1.34 ISI 0.63 0.53 0.46 0.008 rs9297357 8 106211509 � A/G 0.23 0.006 0.68 GAUC 201.81 185.25 199.13 0.001 rs10505229 8 115908561 C/T 0.18 0.002 1.50 GAUC 193.84 200.43 230.74 0.004 Continued on facing page A TYPE 2 DIABETES GWAS IN THE AMISH 3058 DIABETES, VOL. 56, DECEMBER 2007 to carry a subset of the same common type 2 diabetes susceptibility variants as those found in the general pop- ulation and that these variants might be easier to identify. GWAS studies are prone to false-positives due to the very large number of statistical tests that must be per- formed. We were restricted by our relatively modest sample size and also computationally in our attempts to define a genome-wide significance level for which fol- low-up was justified (i.e., variance components tests were not feasible for the many replications needed for case- control permuted family datasets in the Amish). Thus, we relied heavily on a prioritization of SNPs worthy of fol- low-up by testing for 1) internal consistency of type 2 diabetes–associated SNPs with OGTT-derived quantitative traits in nondiabetic Amish individuals, 2) external repli- cation of type 2 diabetes associations in three independent non-Amish 100K SNP GWAS studies, and 3) external replication in a 500K SNP GWAS of type 2 diabetes in a large population of Scandinavians. We found that no single SNP replicated consistently and in the same direction across all GWAS studies, nor were all SNPs associated with type 2 diabetes also associated with quantitative traits in nondiabetic indi- viduals (supplementary Table 4). This is not particularly surprising since we expect that an appreciable number of type 2 diabetes–associated SNPs will be false-posi- tives. Furthermore, a true susceptibility gene in one population might not be readily discernible in other populations due to inadequate sample sizes as well as differences in genetic background, LD, and environmen- tal exposures. Similarly, a true susceptibility gene for type 2 diabetes might not show association with diabe- tes-related quantitative traits in nondiabetic individuals, especially since our OGTT-derived traits are only surro- gates for gold-standard measures of insulin sensitivity and insulin secretion. Nevertheless, we were able to identify a number of candidate genes and loci that showed evidence for association with type 2 diabetes in TABLE 3 Continued SNP Chromo- some Position* Gene† Strand Alleles 1/2 Frequency allele 2 Type 2 diabetes OGTT trait analysis P‡ OR§ Trait Mean 11� Mean 12� Mean 22� P‡ rs10511574 9 11941204 A/C 0.09 0.002 1.82 GAUC 193.05 209.84 170.35 0.007 rs10511777 9 26773463 � A/G 0.94 0.006 0.53 Fasting glucose 3.19 2.92 2.80 0.002 rs10491665 9 29401568 A/C 0.09 0.006 0.58 GAUC 198.29 185.73 171.34 0.007 rs2804498 10 33660713 NRP1 � A/G 0.40 0.001 1.42 Fasting glucose 2.78 2.82 2.89 0.003 rs768676 10 44022687 A/T 0.89 0.002 0.61 Fasting glucose 3.00 2.89 2.80 3.9 � 10�4 rs1111803 10 72426094 C/T 0.55 0.008 0.75 GAUC 208.11 193.50 191.12 0.004 rs2437871 10 90268360 C10orf5 A/C 0.54 0.003 0.72 ISI 0.47 0.51 0.68 0.001 rs10509589 10 91804145 � A/C 0.11 0.004 1.61 HOMA-IR 2.34 2.74 2.51 0.010 rs1887979 10 110133240 A/T 0.41 0.009 0.74 GAUC 201.35 193.00 187.75 0.010 rs10500651 11 5530156 G/T 0.73 0.006 1.42 HOMA-IR 2.88 2.46 2.27 0.005 rs7119814 11 107941856 EXPH5 C/T 0.14 0.007 1.50 ISI 0.58 0.46 0.29 0.008 rs10506173 12 39431860 CNTN1 A/G 0.95 0.007 0.55 Fasting glucose 2.97 2.91 2.81 0.007 rs312272 12 39534200 CNTN1 C/T 0.69 0.007 1.37 Fasting glucose 2.77 2.78 2.86 0.002 rs192852 12 39537359 CNTN1 � C/T 0.32 0.009 0.74 Fasting glucose 2.86 2.77 2.77 0.001 rs2289522 12 39616798 CNTN1 C/T 0.55 0.001 1.46 Fasting glucose 2.76 2.81 2.86 0.004 rs3794247 12 39646028 CNTN1 A/C 0.57 0.002 1.42 Fasting glucose 2.76 2.81 2.86 0.003 rs10521210 17 12966928 C/T 0.65 0.009 1.34 GAUC 184.59 193.63 199.65 0.009 rs9915220 17 13630703 A/G 0.95 0.001 2.72 Fasting glucose 2.69 2.73 2.83 0.009 rs530205 18 42639148 C/T 0.61 0.009 0.76 IAUC 642.24 605.90 538.52 0.008 rs2953271 18 47696031 � A/T 0.38 0.002 0.69 GAUC 200.87 195.67 183.77 0.004 rs10502971 18 49141959 DCC A/G 0.59 0.009 1.34 GAUC 186.36 193.47 203.52 0.001 rs615696 18 49946910 MBD2 C/T 0.79 0.006 0.69 Fasting glucose 2.89 2.84 2.78 0.010 rs739453 19 40771776 A/G 0.60 0.005 0.73 IAUC 685.23 586.92 536.51 0.002 rs3745718 19 53406965 CARD8 � A/G 0.40 0.010 0.75 GAUC 202.60 193.05 186.48 0.003 rs297765 20 4435111 A/G 0.67 2.1 � 10�4 1.55 Fasting glucose 2.74 2.80 2.85 0.004 GAUC 186.37 191.30 203.12 0.001 rs2255140 21 31975858 SFRS15 C/T 0.87 0.009 1.57 ISI 0.37 0.48 0.60 0.004 SNPs with P � 0.01 for type 2 diabetes associations were tested for consistency in a sample of nondiabetic individuals (295 of whom overlapped with the type 2 diabetes association dataset). Direction of association for glucose traits was required to be higher for diabetes risk allele. Neighboring SNPs in bold are in high LD (r2 � 0.80). *Genic region that contains associated SNPs. ‡P values derived using the additive genetic model, adjusted for age, sex, and family structure. The complete dataset including results for dominant and recessive models are available online (available at http://www.medschool.umaryland.edu/amishstudies/index.asp). §OR calculated from a liability threshold model for allele 2 versus allele 1. �Mean values for traits are presented by genotype, with alleles shown in alphabetical order as “1/2.” All insulin traits (IAUC, HOMA-IR, and ISI) were natural log transformed prior to analysis. E. RAMPERSAUD AND ASSOCIATES DIABETES, VOL. 56, DECEMBER 2007 3059 T A B L E 4 S N P s as so ci at ed w it h ty p e 2 d ia b et es in th e A m is h (P � 0. 00 5) an d p ro vi d in g ev id en ce fo r ei th er in te rn al co n si st en cy (a ss o ci at io n w it h an O G T T -d er iv ed qu an ti ta ti ve tr ai t in th e A m is h ) o r ex te rn al re p li ca ti o n (a ss o ci at ed w it h ty p e 2 d ia b et es in an in d ep en d en t p o p u la ti o n ) S N P C h ro m o - so m e P o si ti o n * G en e† S tr an d A ss o ci at ed w it h ty p e 2 d ia b et es in A m is h (P � 0. 00 5) In te rn al co n si st en cy w it h O G T T -d er iv ed qu an ti ta ti ve tr ai ts in A m is h (P � 0. 00 5) E x te rn al re p li ca ti o n w it h ty p e 2 d ia b et es in in d ep en d en t p o p u la ti o n s (P � 0. 00 5) A ll el e 1/ 2 A ll el e 2 ca se su b je ct s A ll el e 2 co n tr o l su b je ct s P O R ‡ T ra it P P o p u la ti o n P § O R ‡ rs 68 91 57 1 18 30 74 13 5 C 1 o r f2 4 A /T 0. 69 0. 80 0. 00 1 0. 67 P im a In d ia n s 0. 00 2 0. 65 rs 28 18 78 1 1 21 52 03 39 6 E S R R G C /T 0. 89 0. 81 0. 00 3 1. 61 G A U C 0. 00 2 rs 13 43 74 7 1 24 07 94 10 6 — A /C 0. 30 0. 21 0. 00 2 1. 43 H O M A -I R 0. 00 2 IA U C 0. 00 2 rs 10 49 00 49 2 40 42 68 54 S L C 8 A 1 � A /C 0. 23 0. 14 0. 00 2 1. 54 F as ti n g gl u co se 0. 00 1 G A U C 0. 00 1 rs 38 45 97 1 3 59 97 57 12 F H I T C /T 0. 79 0. 68 0. 00 4 1. 42 G A U C 4. 0 � 10 � 4 M ex ic an A m er ic an s 0. 00 4 1. 46 rs 93 12 11 3 4 63 64 37 92 — C /G 0. 75 0. 64 0. 00 4 1. 41 F H S 0. 01 3, 0. 00 3 1. 66 rs 37 75 74 5 4 71 14 76 63 C S N 3 � A /C 0. 26 0. 39 0. 00 2 0. 70 M ex ic an A m er ic an s 0. 00 3 0. 68 rs 14 22 93 0 5 16 73 30 17 1 O D Z 2 C /T 0. 04 0. 11 0. 00 3 0. 52 H O M A -I R 0. 00 2 rs 93 21 74 3 6 14 00 05 65 0 — � A /C 0. 42 0. 32 0. 00 3 1. 42 F H S 0. 00 2, 0. 02 4 1. 60 rs 10 48 75 63 7 48 90 57 13 — � C /T 0. 15 0. 27 0. 00 2 0. 68 IS I 0. 00 1 rs 22 37 45 7 7 50 69 36 38 G r b 1 0 � A /G 0. 53 0. 68 1. 1 � 10 � 5 0. 61 G A U C 0. 00 1 rs 10 48 82 84 7 11 97 42 58 0 K C N D 2 � A /G 0. 12 0. 17 0. 00 1 0. 60 H O M A -I R 0. 00 3 rs 10 50 41 33 8 53 15 25 86 — C /T 0. 43 0. 54 0. 00 2 0. 72 IA U C 0. 00 1 rs 10 50 52 29 8 11 59 08 56 1 — C /T 0. 27 0. 15 0. 00 2 1. 50 G A U C 0. 00 4 rs 28 04 49 8 10 33 66 07 13 N R P 1 � A /G 0. 49 0. 36 0. 00 1 1. 42 F as ti n g gl u co se 0. 00 3 rs 76 86 76 10 44 02 26 87 — A /T 0. 85 0. 91 0. 00 2 0. 61 F as ti n g gl u co se 0. 00 4 rs 24 37 87 1 10 90 26 83 60 C 1 0 o r f5 9 A /C 0. 47 0. 57 0. 00 3 0. 72 IS I 0. 00 1 rs 22 89 52 2 12 39 61 67 98 C N T N 1 C /T 0. 64 0. 52 0. 00 1 1. 46 F as ti n g gl u co se 0. 00 4 rs 79 86 01 0 13 91 51 72 60 G P C 5 A /G 0. 44 0. 34 0. 00 2 1. 41 F H S 0. 00 4, 0. 04 2 1. 59 rs 29 53 27 1 18 47 69 60 31 — � A /T 0. 27 0. 43 0. 00 2 0. 69 G A U C 0. 00 4 rs 29 77 65 20 44 35 11 1 — A /G 0. 75 0. 63 2. 1 � 10 � 4 1. 55 F as ti n g gl u co se 0. 00 4 * G en o m e b u il d 36 .1 .† G en ic re gi o n th at co n ta in s as so ci at ed S N P s. ‡O R ca lc u la te d fr o m a li ab il it y th re sh o ld m o d el in S O L A R an d es ti m at ed as al le le 2 ve rs u s al le le 1. §C as e- co n tr o l ge n er al es ti m at in g eq u at io n an d fa m il y- b as ed as so ci at io n te st P va lu es gi ve n fo r F ra m in gh am H ea rt S tu d y (F H S ) d at a; su m m ar y P va lu e gi ve n fo r P im a In d ia n s. A ll in su li n tr ai ts (H O M A -I R ,I A U C , an d IS I) w er e n at u ra l lo g tr an sf o rm ed p ri o r to an al ys is . A TYPE 2 DIABETES GWAS IN THE AMISH 3060 DIABETES, VOL. 56, DECEMBER 2007 more than one population and/or were also associated with OGTT-derived quantitative traits. These results are intriguing but must be interpreted with caution. None of these loci fall within previously identified linkage re- gions for type 2 diabetes (chromosomes 1 and 14) in the Amish. Our strongest type 2 diabetes association signal in the Amish was observed on chromosome 7 in a functionally relevant type 2 diabetes candidate gene, Grb10. Grb10 encodes growth factor– binding protein 10 and has been shown to bind to activated insulin receptor and act as a negative regulator of insulin action and glucose uptake (26 –28). Overexpression of Grb10 in mice causes postna- tal growth retardation and insulin resistance (29). Our 100K GWAS contained a total of 12 SNPs in Grb10, 6 of which were associated with type 2 diabetes (P � 0.05) and were in partial LD with each other (r2 � 0.16 – 0.78). Rs2237457, located in intron 4, provided the lowest P value for association (OR 0.61 for the G- vs. A-allele, P � 1.07 � 10�5). This SNP was also strongly associated with OGTT GAUC in nondiabetic Amish individuals (P � 0.001). Rs2237457 was not associated with type 2 diabetes in the other three populations in which this SNP was genotyped or in the 500K SNP Scandinavian type 2 diabetes GWAS; however, three SNPs (rs2190496, rs2237478, and rs7805310) in Grb10 that were genotyped in the Scandina- vian cohort were associated with type 2 diabetes (P � 0.029, P � 0.01, and P � 0.004, respectively) and are in partial LD with rs2237457 (r2 � 0.12– 0.49 in HapMap CEU). Lack of replication could suggest a false-positive or that variation in Grb10 is a true positive specific to the Amish due to a founder effect or context-dependent phe- notypic expression of the variant due to genetic back- ground or environmental influences. Alternatively, this variant could be in LD with a functional variant, and extended LD in the Amish enabled a type 2 diabetes association to be detected in this population and not the others. In a recent report by Di Paola et al. (30), the A-allele of rs4947710, a synonymous coding SNP in Grb10, was associated with decreased risk of type 2 diabetes in a relatively homogeneous population of Italian Caucasians (P � 0.0001). This SNP was not part of the 100K SNP panel nor was our most highly type 2 diabetes–associated SNP (rs2237457) genotyped in the Italian sample. We found that rs2237457 and rs4947710 are not in LD (r2 � 0) in HapMap CEU samples. However, rs10486757, another Grb10 SNP associated with type 2 diabetes in the Amish (P � 0.024), is in LD with rs4947710 (r2 � 0.64 in HapMap CEU). Further investigation of Grb10 is currently underway. Our GWAS and replication strategy have several limita- tions. First, the relatively small sample size limits our ability to detect gene variants of modest effect size. Second, we recognize that the definition of external repli- cation of our top SNPs across three independent 100K studies of type 2 diabetes might represent a skewed distribution of the overall results since replication was limited to our 1,000 most highly type 2 diabetes–associ- ated SNPs. This approach was used to facilitate compari- sons across populations and also to limit the number of false-positive replications due to multiple comparisons. To the extent that we attempted to pursue signals that repre- sent the “lowest hanging fruit,” we believe that the ap- proach we have taken is reasonable. A formal meta- analysis of the entire set of data from all four 100K studies is currently underway. Third, our replication approach was focused at the level of the SNP in order to avoid additional multiple comparisons. However, it is possible that we did not identify significantly associated SNPs in other populations that were in LD with our top SNPs. This is particularly relevant for our comparisons with the Scandinavian 500K GWAS, for which only 27% of the SNPs identified in the Amish with P � 0.01 were identified in the 500K SHP panel. The likelihood that we missed common variants impor- tant to type 2 diabetes is high due to the relatively sparse density of the 100K SNP panel (mean intermarker dis- tance � 29 kb) compared with other denser GWAS SNP panels. For example, SNPs in well-replicated genes (SLC30A8, IGF2BP2, CKDAL1, CDKN2A/CKDN2B, and HHEX/IDE) found in four recently published type 2 dia- betes GWAS studies (15–18), as well as previously known type 2 diabetes–associated variants in TCF7L2, KCNJ11, HNF4A, or CAPN10 (9), were not adequately covered on the 100K genotyping panel (i.e., r2 � 0.8 between the SNP of interest and SNPs on the 100K panel). As a positive control, we previously demonstrated that TCF7L2 SNP rs7903146 and the HNF4A promoter SNP rs2425640, nei- ther of which is present on the 100K panel, were associ- ated with type 2 diabetes and impaired glucose tolerance in the Amish Family Diabetes Study (OR 1.57, P � 0.008; 1.60, P � 0.04, respectively) (12,25). Interestingly, rs10509645 in HHEX on the 100K panel (r2 � 0.7 with rs7923837 found previously to be strongly associated with type 2 diabetes in other GWAS studies) was significantly associated with type 2 diabetes in the Amish (OR 1.30 for the G-allele; P � 0.02). Rs9300039 on chromosome 11, shown to be associated with type 2 diabetes in the other GWAS studies (17), was present on the 100K panel but was not significantly associated with type 2 diabetes in the Amish (OR 1.09 for the C-allele; P � 0.67). In summary, we presented results from our initial ex- amination of a GWAS of type 2 diabetes in the Amish. Although we did not identify any genes associated with type 2 diabetes that reached genome-wide significance, we report a number of genes and loci that are worthy of further study based on replication in other studies or on quantitative trait loci consistency. This report (and the three companion articles) provides a valuable resource for other investigators to utilize in the search for the patho- genic variants for type 2 diabetes. ACKNOWLEDGMENTS This work was supported by the following National Insti- tutes of Health Research Grants: Training Grant in Cardiac and Vascular Cell Biology (T32 HL072751), R01 DK54261, the University of Maryland General Clinical Research Center (M01 RR16500), Hopkins Bayview General Clinical Research Center (M01 RR02719), the Maryland Clinical Nutrition Research Unit (P30 DK072488), and the Balti- more Veterans Administration Geriatric Research and Education Clinical Center. We extend our thanks to our collaborators in the Type 2 Diabetes 100K GWAS Consortium for sharing prepublica- tion results from the Starr County Health Studies, Fra- mingham Heart Study, and the Pima Indians as well as to the Broad-Lund-Novartis Diabetes Genomic Initiative for access to results from their GWAS. We also thank Soren Snitker for helpful comments and Adam Naj for help with formatting figures for this manu- script. Lastly, we gratefully acknowledge our Amish liai- E. RAMPERSAUD AND ASSOCIATES DIABETES, VOL. 56, DECEMBER 2007 3061 sons and research staff and the extraordinary cooperation and support of the Amish community, without whom these studies would not be possible. REFERENCES 1. Elbein SC, Hoffman MD, Teng K, Leppert MF, Hasstedt SJ: A genome-wide search for type 2 diabetes susceptibility genes in Utah Caucasians. Diabetes 48:1175–1182, 1999 2. Wiltshire S, Hattersley AT, Hitman GA, Walker M, Levy JC, Sampson M, O’Rahilly S, Frayling TM, Bell JI, Lathrop GM, Bennett A, Dhillon R, Fletcher C, Groves CJ, Jones E, Prestwich P, Simecek N, Rao PV, Wishart M, Foxon R, Howell S, Smedley D, Cardon LR, Menzel S, McCarthy MI: A genomewide scan for loci predisposing to type 2 diabetes in a U.K. population (the Diabetes UK Warren 2 Repository): analysis of 573 pedigrees provides independent replication of a susceptibility locus on chromosome 1q. Am J Hum Genet 69:553–569, 2001 3. Du W, Sun H, Wang H, Qiang B, Shen Y, Yao Z, Gu J, Xiong M, Huang W, Chen Z, Zuo J, Hua X, Gao W, Sun Q, Fang F: Confirmation of susceptibility gene loci on chromosome 1 in northern China Han families with type 2 diabetes. Chin Med J (Engl) 114:876 – 878, 2001 4. Meigs JB, Panhuysen CI, Myers RH, Wilson PW, Cupples LA: A genome- wide scan for loci linked to plasma levels of glucose and HbA1c in a community-based sample of Caucasian pedigrees: the Framingham Off- spring Study. Diabetes 51:833– 840, 2002 5. Hanson RL, Ehm MG, Pettitt DJ, Prochazka M, Thompson DB, Timberlake D, Foroud T, Kobes S, Baier L, Burns DK, Almasy L, Blangero J, Garvey WT, Bennett PH, Knowler WC: An autosomal genomic scan for loci linked to type II diabetes mellitus and body-mass index in Pima Indians. Am J Hum Genet 63:1130 –1138, 1998 6. Hsueh WC, St. Jean PL, Mitchell BD, Pollin TI, Knowler WC, Ehm MG, Bell CJ, Sakul H, Wagner MJ, Burns DK, Shuldiner AR: Genome-wide and fine-mapping linkage studies of type 2 diabetes and glucose traits in the Old Order Amish: evidence for a new diabetes locus on chromosome 14q11 and confirmation of a locus on chromosome 1q21– q24. Diabetes 52:550 – 557, 2003 7. Ng MC, So WY, Cox NJ, Lam VK, Cockram CS, Critchley JA, Bell GI, Chan JC: Genome-wide scan for type 2 diabetes loci in Hong Kong Chinese and confirmation of a susceptibility locus on chromosome 1q21– q25. Diabetes 53:1609 –1613, 2004 8. McCarthy MI: Growing evidence for diabetes susceptibility genes from genome scan data. Curr Diab Rep 3:159 –167, 2003 9. Owen KR, McCarthy MI: Genetics of type 2 diabetes. Curr Opin Genet Dev 17:239 –244, 2007 10. Hsueh W-C, Mitchell BD, Aburomia R, Pollin T, Sakul H, Ehm MG, Michelsen BK, Wagner MJ, St. Jean PL, Knowler WC, Burns DK, Bell CJ, Shuldiner AR: Diabetes in the old order Amish: characterization and heritability analysis of the Amish Family Diabetes Study. Diabetes Care 23:595– 601, 2000 11. Damcott CM, Pollin TI, Reinhart LJ, Ott SH, Shen H, Silver KD, Mitchell BD, Shuldiner AR: Polymorphisms in the transcription factor 7-like 2 (TCF7L2) gene are associated with type 2 diabetes in the Amish: replica- tion and evidence for a role in both insulin secretion and insulin resistance. Diabetes 55:2654 –2659, 2006 12. Damcott CM, Hoppman N, Ott SH, Reinhart LJ, Wang J, Pollin TI, O’Connell JR, Mitchell BD, Shuldiner AR: Polymorphisms in both promot- ers of hepatocyte nuclear factor 4-� are associated with type 2 diabetes in the Amish. Diabetes 53:3337–3341, 2004 13. Klein RJ, Zeiss C, Chew EY, Tsai JY, Sackler RS, Haynes C, Henning AK, San Giovanni JP, Mane SM, Mayne ST, Bracken MB, Ferris FL, Ott J, Barnstable C, Hoh J: Complement factor H polymorphism in age-related macular degeneration. Science 308:385–389, 2005 14. Gibbs JR, Singleton A: Application of genome-wide single nucleotide polymorphism typing: simple association and beyond. PLoS Genet 2:e150, 2006 15. Zeggini E, Weedon MN, Lindgren CM, Frayling TM, Elliott KS, Lango H, Timpson NJ, Perry JR, Rayner NW, Freathy RM, Barrett JC, Shields B, Morris AP, Ellard S, Groves CJ, Harries LW, Marchini JL, Owen KR, Knight B, Cardon LR, Walker M, Hitman GA, Morris AD, Doney AS, the Wellcome Trust Case Control Consortium (WTCCC), McCarthy MI, Hattersley AT: Replication of genome-wide association signals in UK samples reveals risk loci for type 2 diabetes. Science 316:1336 –1341, 2007 16. Steinthorsdottir V, Thorleifsson G, Reynisdottir I, Benediktsson R, Jonsdottir T, Walters GB, Styrkarsdottir U, Gretarsdottir S, Emilsson V, Ghosh S, Baker A, Snorradottir S, Bjarnason H, Ng MC, Hansen T, Bagger Y, Wilensky RL, Reilly MP, Adeyemo A, Chen Y, Zhou J, Gudnason V, Chen G, Huang H, Lashley K, Doumatey A, So WY, Ma RC, Andersen G, Borch-Johnsen K, Jorgensen T, van Vliet-Ostaptchouk JV, Hofker MH, Wijmenga C, Christiansen C, Rader DJ, Rotimi C, Gurney M, Chan JC, Pedersen O, Sigurdsson G, Gulcher JR, Thorsteinsdottir U, Kong A, Stefansson K: A variant in CDKAL1 influences insulin response and risk of type 2 diabetes. Nat Genet 39:770 –775, 2007 17. Scott LJ, Mohlke KL, Bonnycastle LL, Willer CJ, Li Y, Duren WL, Erdos MR, Stringham HM, Chines PS, Jackson AU, Prokunina-Olsson L, Ding CJ, Swift AJ, Narisu N, Hu T, Pruim R, Xiao R, Li XY, Conneely KN, Riebow NL, Sprau AG, Tong M, White PP, Hetrick KN, Barnhart MW, Bark CW, Goldstein JL, Watkins L, Xiang F, Saramies J, Buchanan TA, Watanabe RM, Valle TT, Kinnunen L, Abecasis GR, Pugh EW, Doheny KF, Bergman RN, Tuomilehto J, Collins FS, Boehnke M: A genome-wide association study of type 2 diabetes in Finns detects multiple susceptibility variants. Science 316:1341–1345, 2007 18. Saxena R, Voight BF, Lyssenko V, Burtt NP, de Bakker PI, Chen H, Roix JJ, Kathiresan S, Hirschhorn JN, Daly MJ, Hughes TE, Groop L, Altshuler D, Almgren P, Florez JC, Meyer J, Ardlie K, Bengtsson BK, Isomaa B, Lettre G, Lindblad U, Lyon HN, Melander O, Newton-Cheh C, Nilsson P, Orho- Melander M, Rastam L, Speliotes EK, Taskinen MR, Tuomi T, Guiducci C, Berglund A, Carlson J, Gianniny L, Hackett R, Hall L, Holmkvist J, Laurila E, Sjogren M, Sterner M, Surti A, Svensson M, Svensson M, Tewhey R, Blumenstiel B, Parkin M, Defelice M, Barry R, Brodeur W, Camarata J, Chia N, Fava M, Gibbons J, Handsaker B, Healy C, Nguyen K, Gates C, Sougnez C, Gage D, Nizzari M, Gabriel SB, Chirn GW, Ma Q, Parikh H, Richardson D, Ricke D, Purcell S: Genome-wide association analysis identifies loci for type 2 diabetes and triglyceride levels. Science 316:1331–1336, 2007 19. BRLMM: an improved genotype calling method for the GeneChip Human Mapping 500K Array Set [article online], 2006. Available from http:// www.affymetrix.com/support/technical/whitepapers/brlmm_whitepaper. pdf. Accessed 5 September 2007 20. Almasy L, Blangero J: Multipoint quantitative-trait linkage analysis in general pedigrees. Am J Hum Genet 62:1198 –1211, 1998 21. Boerwinkle E, Chakraborty R, Sing CF: The use of measured genotype information in the analysis of quantitative phenotypes in man. I. Models and analytical methods. Ann Intern Med 50:181–194, 1986 22. Purcell S, Cherny SS, Sham PC: Genetic Power Calculator: design of linkage and association genetic mapping studies of complex traits. Bioin- formatics 19:149 –150, 2003 23. Florez JC, Manning AK, Dupuis J, McAteer J, Irenze K, Gianniny L, Mirel DB, Fox CS, Cupples LA, Meigs JB: A 100K genome-wide association scan for diabetes and related traits in the Framingham Heart Study: replication and integration with other genome-wide datasets. Diabetes 56:3063–3074, 2007 24. Hanson RL, Bogardus C, Duggan D, Kobes S, Knowlton M, Infante AM, Marovich L, Benitez D, Baier LJ, Knowler WC: A search for variants associated with young-onset type 2 diabetes in American Indians in a 100K genotyping array. Diabetes 56:3045–3052, 2007 25. Hayes MG, Pluzhnikov A, Miyake K, Sun Y, Ng MCY, Roe CA, Below JE, Nicolae RI, Konkashbaev A, Bell GI, Cox NJ, Hanis CL: Identification of type 2 diabetes genes in Mexican Americans through genome-wide asso- ciation studies. Diabetes 56:3033–3042, 2007 26. Deng Y, Bhattacharya S, Swamy OR, Tandon R, Wang Y, Janda R, Riedel H: Growth factor receptor-binding protein 10 (Grb10) as a partner of phos- phatidylinositol 3-kinase in metabolic insulin action. J Biol Chem 278: 39311–39322, 2003 27. Langlais P, Dong LQ, Ramos FJ, Hu D, Li Y, Quon MJ, Liu F: Negative regulation of insulin-stimulated mitogen-activated protein kinase signaling by Grb10. Mol Endocrinol 18:350 –358, 2004 28. Mounier C, Lavoie L, Dumas V, Mohammad-Ali K, Wu J, Nantel A, Bergeron JJ, Thomas DY, Posner BI: Specific inhibition by hGrb10zeta of insulin- induced glycogen synthase activation: evidence for a novel signaling pathway. Mol Cell Endocrinol 173:15–27, 2001 29. Shiura H, Miyoshi N, Konishi A, Wakisaka-Saito N, Suzuki R, Muguruma K, Kohda T, Wakana S, Yokoyama M, Ishino F, Kaneko-Ishino T: Meg1/Grb10 overexpression causes postnatal growth retardation and insulin resistance via negative modulation of the IGF1R and IR cascades. Biochem Biophys Res Commun 329:909 –916, 2005 30. Di Paola R, Ciociola E, Boonyasrisawat W, Nolan D, Duffy J, Miscio G, Cisternino C, Fini G, Tassi V, Doria A, Trischitta V: Association of hGrb10 genetic variations with type 2 diabetes in Caucasian subjects. Diabetes Care 29:1181–1183, 2006 A TYPE 2 DIABETES GWAS IN THE AMISH 3062 DIABETES, VOL. 56, DECEMBER 2007