analysis of slc16a11 variants in 12,811 american indians · 2015-10-16 · 1 1 analysis of...

1

Analysis of SLC16A11 variants in 12,811 American Indians: 1

genotype-obesity interaction for type 2 diabetes and 2

an association with RNASEK expression 3

4

5

SLC16A11 and T2D in American Indians 6

7

8

Michael Traurig,1,6

Robert L. Hanson,1,6

Alejandra Marinelarena,1

Sayuko Kobes,1 Paolo Piaggi,

1 9

Shelley Cole,2 Joanne E. Curran,

3 John Blangero,

3 Harald Göring,

3 Satish Kumar,

3 Robert G. 10

Nelson,1 Barbara V. Howard,

4,5 William C. Knowler,

1 Leslie J. Baier,

1 and Clifton Bogardus

1* 11

12

13 1Phoenix Epidemiology and Clinical Research Branch, National Institute of Diabetes and 14

Digestive and Kidney Diseases, National Institutes of Health, Phoenix, Arizona, 85004, USA. 15 2Texas Biomedical Research Institute, San Antonio, Texas, 78227, USA. 16

3South Texas Diabetes and Obesity Institute, University of Texas Rio Grande Valley School of 17

Medicine , Brownsville, TX, 78520 USA. 18 4Medstar Health Research Institute, Hyattsville, Maryland, 20782, USA. 19

5Georgetown-Howard Universities Center for Clinical and Translational Science. 20

6These authors contributed equally to the manuscript. 21

22

23

24

25

*Corresponding author: 26

Clifton Bogardus 27

Phoenix Epidemiology and Clinical Research Branch 28

NIDDK, NIH 29

455 North 5th

street 30

Phoenix, AZ 85004, USA 31

Tel: (602) 440-6571 Fax: (602) 253-4140 32

Email:[email protected] 33

34

Abstract word count: 197 35

Main text word count: 3865 36

Tables: 3 37

Figures: 5 38

39

of 41 Diabetes

Diabetes Publish Ahead of Print, published online October 20, 2015

2

ABSTRACT 40

Genetic variants in SLC16A11 were recently reported to be associated with type 2 diabetes in 41

Mexican and other Latin American populations. The diabetes risk haplotype had a frequency of 42

50% in Native Americans from Mexico, but was rare in Europeans and Africans. In the current 43

study, we analyzed SLC16A11 in 12,811 North American Indians and found that the diabetes risk 44

haplotype, tagged by the rs75493593 A allele, was nominally associated with type 2 diabetes (P 45

= 0.001, odds ratio = 1.11). However, there was a strong interaction with body mass index (BMI) 46

(P = 5.1 × 10-7

) such that the diabetes association was stronger in leaner individuals. 47

Rs75493593 was also strongly associated with BMI in individuals with type 2 diabetes (P = 3.4 48

× 10-15

) but not in non-diabetic individuals (P = 0.77). Longitudinal analyses suggest that this is 49

due, in part, to an association of the A allele with greater weight loss following diabetes onset (P 50

= 0.02). Analyses of global gene expression data from adipose, skeletal muscle, and whole blood 51

provide evidence that rs75493593 is associated with expression of the nearby RNASEK gene, 52

suggesting that RNASEK expression may mediate the effect of genotype on diabetes. 53

54

55

56

57

58

59

60

61

62

of 41Diabetes

3

INTRODUCTION 63

The SIGMA Type 2 Diabetes Consortium recently reported a haplotype consisting of 5 64

common exonic variants, including rs75493593, in the monocarboxylic acid transporter gene 65

SLC16A11 that was strongly associated with type 2 diabetes in Mexican and other Latin 66

American populations (1). The diabetes risk haplotype had a frequency of 50% in Native 67

Americans from Mexico, but was rare in Europeans and Africans (1) suggesting it may 68

contribute to the higher burden of type 2 diabetes among American Indians. Our previous study 69

of 63 established type 2 diabetes variants in 3,421 individuals from a longitudinal diabetes study 70

in a Southwestern American Indian (SWAI) population showed little evidence of association 71

between rs75493593 and type 2 diabetes (2). To further assess whether SLC16A11 variants are 72

associated with type 2 diabetes in American Indians, we expanded our analysis by genotyping 3 73

tag SNPs (rs75493593, rs148775056, and rs9303212) that capture all variants with a frequency 74

>0.01 across the SLC16A11 locus in our population-based SWAI sample (N = 7,295). Since one 75

of the tag SNPs, rs75493593, captures the previously reported Mexican diabetes risk haplotype, 76

it was further analyzed in two additional samples of American Indians from the Phoenix 77

extension of the Family Investigation of Diabetes (FIND, N = 3,095) and the Strong Heart Study 78

(SHS, N = 2,421) for a combined total of 12,811 North American Indians. Since the SIGMA 79

study found that the SLC16A11 diabetes risk alleles were associated with leaner BMI in 80

individuals with diabetes, we also examined whether SLC16A11 genotypes interact with BMI in 81

their association with diabetes. To date, none of the SLC16A11 missense variants have been 82

shown to alter protein function; therefore, we merged rs75493593 genotypic data with gene 83

expression data to search for potential expression quantitative loci (eQTLs) which could provide 84

an alternative mechanism affecting the risk for type 2 diabetes. 85

of 41 Diabetes

4

RESEARCH DESIGN and METHODS 86

Participants for the type 2 diabetes and BMI analyses 87

Brief descriptions of the study cohorts are provided below and their clinical characteristics are 88

shown in Supplementary Table 1. Each study was approved by an institutional review board and 89

participants gave informed consent. 90

91

SWAI 92

From 1965–2007, SWAI from a community in Arizona participated in biennial health exams 93

that included a 75 g oral glucose tolerance test (OGTT), measures of height and weight to 94

calculate BMI, and a blood draw for DNA isolation (3). Diabetes status was determined 95

according to the 1997 criteria of the American Diabetes Association (fasting plasma glucose ≥7.0 96

mmol/l or 2-hour plasma glucose ≥11.1 mmol/l) (4) or from a previous clinical diagnosis. 97

Fasting glucose, 2 hr glucose, and insulin concentrations were analyzed using data from the 98

participant’s last non-diabetic exam. Estimates of insulin resistance (HOMA-IR) and β-cell 99

function (HOMA-B) were calculated using the homeostatic model (5). The current study includes all 100

individuals (N = 7,710) for whom diabetes status is recorded and DNA is currently available. 101

102

FIND 103

Participants from the Phoenix extension of FIND, a multicenter study designed to identify 104

genes involved in diabetic nephropathy (6), were required to be of ≥50% American Indian 105

heritage (self-reported) and most were urban-dwelling American Indians living near Phoenix, 106

Arizona. Individuals were examined once and diabetes status was determined by 2010 American 107

Diabetes Association criteria based on fasting blood glucose concentrations (≥7.0 mmol/l), 108

of 41Diabetes

5

HbA1c (≥6.5%) (7), or from a prior clinical diagnosis (6). The current study includes all 109

participants (N = 3,106) who were not also studied as part of the SWAI sample. 110

111

SHS 112

Participants are American Indians who underwent health exams at study sites located in 113

Arizona, North and South Dakota, and Oklahoma (8). Diabetes status was determined by 1998 114

World Health Organization criteria based on 2 hr blood glucose ≥11.1 mmol/l, fasting blood 115

glucose ≥7.0 mmol/l, current use of diabetes medication, and/or if an individual was previously 116

diagnosed with diabetes (9). The current study includes all participants (N = 2,451) who were not 117

also studied as part of the SWAI sample. 118

119

Participants for the inpatient metabolic studies 120

Some individuals (N = 561, age = 18–45 years) from the SWAI sample also participated in 121

inpatient metabolic exams in our Clinical Research Center. These individuals were non-diabetic 122

as determined by a 75 g OGTT, and were healthy based on

medical history, physical 123

examination, and routine laboratory tests, and were not taking any medications. The OGTT was 124

performed after an overnight fast and blood was drawn before and at 30, 60, 120, and 180 min 125

after glucose ingestion for determining plasma glucose and insulin concentrations. Insulin action 126

was assessed using the two-step hyperinsulinemic-euglycemic clamp technique to measure 127

insulin-stimulated glucose disappearance (uptake) rates (10). To determine acute insulin 128

response (AIR), a 25 g bolus of glucose was injected intravenously over 3 minutes and blood 129

samples were collected before infusion, and at 3, 4, and 5 min after the injection to determine 130

plasma glucose and insulin concentrations. AIR was calculated as one-half the mean increment in 131

of 41 Diabetes

6

plasma insulin concentrations from 3 to 5 min (10). Percent body fat was estimated by 132

underwater weighing or by total body dual energy X-ray absorptiometry (DPX-1; Lunar 133

Radiation, Madison, WI, USA) (11). A subset of the inpatient subjects (N = 423) also 134

participated in our metabolic chamber studies to determine energy expenditure and substrate 135

(carbohydrate, lipid, and protein) oxidation in the metabolic respiratory chamber (12). 136

137

Genotyping 138

Genotyping was done by allelic discrimination using Taqman genotyping assays (Applied 139

Biosystems, Foster City, CA). All genotypic data included in the analyses met our quality control 140

criteria which requires a successful call rate of ≥95%, a lack of deviation from Hardy-Weinberg 141

equilibrium at P <0.001, and a discrepancy rate of ≤1% in masked duplicates (430, 120, and 88 142

blind duplicates for the SWAI, FIND, and SHS samples respectively). 143

144

Statistical analyses 145

The association between genotypes and type 2 diabetes was determined with logistic 146

regression modeling (additive model), where homozygotes for the major allele (A1/A1), 147

heterozygotes (A1/A2), and homozygotes for the minor allele (A2/A2) were coded to a numeric 148

variable for genotype (0, 1, and 2). In individuals from the SWAI and SHS community-based 149

studies, the model was fit with the generalized estimating procedure to account for sibship. For 150

the FIND sample, the genomic control procedure (13) was used to account for unspecified 151

relationships based on 42 random markers from throughout the genome. Individual estimates of 152

European admixture, which were used as covariates in analyses, were derived from 45 markers 153

with large differences in allele frequency between populations (14) by the method of Hanis et al 154

of 41Diabetes

7

(15). A meta-analysis for the SWAI, FIND, and SHS samples was conducted by the inverse 155

variance method (16). Heterogeneity among the samples was quantified by the I2 measure and 156

statistical significance was tested by Cochran’s Q statistic (17). For longitudinal analyses of 157

diabetes, SWAI individuals who were not initially diabetic were followed from their first 158

examination after age 15 years until development of diabetes or until their last examination. Cox 159

proportional hazard regression was used to estimate the risk (hazard ratio) of developing type 2 160

diabetes. A similar analysis was performed for the Strong Heart Study using data from phases I 161

through III. Linear regression was used to analyze the associations between genotype and 162

continuous variables and the models were fit with the generalized estimating equation procedure 163

to account for sibship. Statistical analyses were performed using the statistical analysis system of 164

the SAS Institute (Cary, North Carolina, USA). 165

166

Gene expression and mediation analyses 167

Genome-wide mRNA levels had previously been measured using the Affymetrix 1.0 Human 168

Exon microarray in adipose tissue (N = 187) and skeletal muscle (N = 196) biopsies from a 169

subset of the SWAI sample (all were non-diabetic) (18). Genome-wide mRNA levels in whole 170

blood were also available from a prior study in a subset of the FIND sample (N = 1,416, 23% 171

with diabetes) who had been genotyped for 42 established type 2 diabetes-susceptibility variants 172

including rs75493593. Whole blood gene expression levels were measured using the Illumina 173

HumanHT-12 v4 Expression Beadchip. A χ2 tail test was used to assess associations between 174

genotype and 15,854 transcripts expressed above background (19), that mapped to unique 175

locations, and that were free of variants located in probe set binding sites according to ReMOAT 176

annotation (20). For variants associated with both type 2 diabetes (P <0.05) and whole blood 177

of 41 Diabetes

8

transcript levels (P <0.005), a formal mediation analysis was used to quantify the extent to which 178

the pattern of associations was consistent with an effect on transcript that mediates the variant-179

diabetes association. Significance of mediation was assessed by the Sobel test (21), and the 180

percentage of the variant effect on diabetes that was potentially mediated by the transcript was 181

calculated (22). Additional details are described in the supplementary text. Since associations 182

between SLC16A11 variants and type 2 diabetes have previously been established in other 183

populations, we considered a P <0.05 as statistically significant for variant/diabetes-related 184

phenotype associations. In the whole transcriptome analyses, which lack a specific a priori 185

hypothesis, we calculated the false discovery rate (23) to account for multiple comparisons 186

187

RESULTS 188

Association analysis of type 2 diabetes using cross-sectional data 189

Previously analyzed whole genome sequence data (40× coverage) from 296 SWAI 190

individuals were used to identify 14 variants with a minor allele frequency ≥0.01 in the genomic 191

region spanning the SLC16A11 locus (Supplementary Figure 1A). Linkage disequilibrium (LD) 192

information (r2 ≥0.85) for the 14 variants could be captured by genotyping 3 tag SNPs, 193

rs75493593, rs148775056 and rs9303212 (Supplementary Figure 1B). Most of the common 194

variants, including all 5 exonic variants that make up the previously reported type 2 diabetes risk 195

haplotype in the SIGMA study, fell into one major LD group tagged by rs75493593. The 3 tag 196

SNPs were genotyped in our population-based SWAI sample (N = 7,295), but none were 197

associated with type 2 diabetes (Supplementary Table 2). Since rs75493593 tagged the diabetes 198

risk haplotype reported in the SIGMA study, this variant was further genotyped in two additional 199

samples of American Indians from the FIND (N = 3,095) and SHS (N = 2,421) studies. Meta-200

of 41Diabetes

9

analysis of the three samples (N = 12,811) detected a nominal association between the 201

rs75493593 A allele and higher risk of type 2 diabetes (P = 0.001, OR = 1.11 [1.04–1.19], Table 202

1) which is directionally consistent with the Mexican study. 203

204

Analyses of rs75493593 genotype by BMI and type 2 diabetes interactions 205

A significant genotype × BMI interaction was observed for the association of rs75493593 206

with type 2 diabetes (combined, Pinteraction = 5.1 × 10-7

, Fig. 1 and Supplementary Table 3). After 207

stratifying the individuals into BMI categories, a modest association was observed between 208

rs75493593 and diabetes for the non-obese groups (in the combined American Indian sample, P 209

= 4.9 × 10-3

and 7.5 × 10-6

for the BMI categories <25 and 25–30 kg/m2, respectively) where the 210

A allele was associated with increased risk for diabetes (OR = 1.41 and 1.35), but among the 211

heavier individuals (BMI 35–40 and 40+ kg/m2), the A allele was associated with decreased risk 212

for diabetes (OR = 0.88 and 0.84). Similar trends were observed in each American Indian sample 213

when analyzed separately (Fig. 1 and Supplementary Table 3). This same phenomenon is 214

reflected in a strong diabetes × genotype interaction for the association of rs75493593 with BMI 215

(in the combined American Indian sample, Pinteraction = 3.4 × 10-10

, Table 2). Although an 216

association between rs75493593 and BMI in the combined sample was observed (6.4 × 10-5

), 217

where individuals with diabetes who carry the risk (A) allele have lower BMIs, this association 218

was due entirely to individuals with type 2 diabetes (with diabetes, P = 3.4 × 10-15

; without 219

diabetes, P = 0.77, Table 2). The distribution of genotype across BMI categories and diabetes 220

status is given in Supplementary Table 4. The association between the risk haplotype and lower 221

BMI in individuals with diabetes was also observed in the Mexican study (P = 5.2 × 10-4

) (1); 222

of 41 Diabetes

10

and, in combination with the present data, rs75493593 is a replicated variant for BMI at genome-223

wide statistical significance in individuals with diabetes (combined, P = 5.2 × 10-16

). 224

225

Longitudinal analysis of diabetes risk as a function of rs75493593 226

To assess whether individuals with the rs75493593 A allele develop diabetes at a leaner BMI 227

or lose more weight after diabetes onset, 4,088 SWAI and 1,073 SHS individuals who were non-228

diabetic at their first exam and had longitudinal follow-up data were analyzed. Among the SWAI 229

individuals (mean age = 22.0 years), 1,498 (37%) developed diabetes during a mean follow-up 230

time of 13.5 years; and, for the SHS participants (mean age = 55.1 years), 222 (21%) developed 231

diabetes during a mean follow-up of 6.6 years. A proportional hazard regression analysis (hazard 232

ratio, HR) showed that rs75493593 did not predict type 2 diabetes in either sample (SWAI, HR = 233

1.04[0.97–1.12], P = 0.31; SHS, HR = 1.00[0.81–1.23], P = 0.97) or when the SWAI and SHS 234

samples were combined (HR = 1.03[0.97–1.11], P = 0.35). However, when grouped by BMI, a 235

nominally significant HR was detected for SWAI individuals with a BMI <25 kg/m2 (HR = 236

1.18[1.01–1.38], P = 0.03, Fig. 2) where the rs75493593 A allele was associated with increased 237

risk of diabetes, a finding consistent with the cross-sectional analyses. The hazard ratio for 238

rs75493593 was similarly elevated in the SHS individuals with BMI <25 kg/m2, and the 239

association was nominally significant in the combined sample (HR = 1.20 [1.03–1.39], P = 0.02, 240

Fig. 2). However, a significant rs75493593 genotype × baseline BMI interaction was not 241

observed in the SWAI, SHS, or combined samples (Pinteraction = 0.94, 0.11, and 0.59, respectively, 242

Fig 2). The longitudinal relationship with change in BMI, as measured by the slope, across non-243

diabetic exams and across exams conducted after diabetes onset was also analyzed. When SWAI 244

participants were non-diabetic, there was no difference in BMI change by genotype (N = 4,480, 245

of 41Diabetes

11

P = 0.32, Fig. 3A). In contrast, after type 2 diabetes onset, there was an association between 246

rs75493593 and BMI slope (N = 1,632, P = 0.02, Fig. 3A), where the A allele was associated 247

with increased weight loss. The decrease in BMI after diabetes onset was 1.2% vs. 0.5% per year 248

for individuals homozygous for the A and C alleles, respectively. There were no significant 249

associations between rs75493593 and weight change in the SHS individuals, although it should 250

be noted that the weight changes in this older population were smaller in magnitude (Fig 3B). 251

252

Association analysis of quantitative type 2 diabetes related-traits 253

To evaluate the physiology underlying the associations with type 2 diabetes, glucose and 254

insulin concentrations along with homeostatic model (HOMA) measures of insulin secretion and 255

resistance were analyzed in SWAI individuals who had available data from their last non-256

diabetic exam (N ≈ 5,200). Nominal evidence for a rs75493593 genotype × BMI interaction was 257

detected for various traits including fasting insulin, HOMA-IR, and HOMA-B (Pinteraction = 0.03 – 258

0.04, Supplementary Table 5); however, these associations are too weak to be conclusive of the 259

underlying physiology. In a subset of non-diabetic SWAI individuals who had undergone 260

metabolic testing in our Clinical Research Center, there was a nominal but consistent association 261

between rs75493593 and basal carbohydrate and lipid oxidation when measured during a 262

hyperinsulinemic-euglycemic clamp (N = 521) and when measured in a human respiratory 263

chamber (N = 423) (Supplementary Table 6). Individuals homozygous for the A allele had higher 264

mean carbohydrate oxidation rates and lower mean lipid oxidation rates compared to the other 2 265

genotypes (P = 0.007 and 0.006, respectively, Fig. 4A,B) resulting in a slightly higher mean 24 266

hr respiratory quotient (P = 0.01). These associations, although weak, are consistent between two 267

independent in vivo measures of carbohydrate and lipid oxidation. 268

of 41 Diabetes

12

Analysis of rs75493593 as an eQTL 269

Based on SIFT prediction, only 1 of the 4 SLC16A11 missense variants, rs13342692 270

(D127G), is predicted to be damaging with a SIFT score (0.05, Supplementary Table 7) just 271

reaching the cutoff for deleterious amino acid substitutions (SIFT score thresholds: tolerated 272

>0.05, damaging ≤0.05 [24]). To search for additional mechanisms whereby these variants could 273

have a functional effect, genotypic data for rs75493593 were merged with previously measured 274

global adipose (N = 187) and skeletal muscle (N = 196) gene expression data to determine 275

whether variants tagged by rs75493593 could function as an expression quantitative trait locus 276

(eQTL). The genome-wide eQTL analyses identified RNASEK which is located approximately 277

27 kb upstream of rs75493593 (SLC16A11) as one of the top eQTL-genes in both tissues which 278

were biopsied from different SWAI individuals (adipose P = 2.1 × 10-4

and skeletal muscle 1.7 × 279

10-8

, Table 3). The microarray mean RNASEK expression levels based on rs75493593 genotypes 280

are shown in figure 5A and B. The association between rs75493593 and adipose RNASEK 281

expression was replicated by qRT-PCR (N = 182, P = 8.0 × 10-4

, Fig. 5C). There were no 282

microarray expression data for SLC16A11 in either adipose or skeletal muscle; however, directly 283

measuring SLC16A11 mRNA levels by qRT-PCR showed no association between rs75493593 284

and expression in adipose tissue (N = 182, P = 0.30, data not shown). The association between 285

rs75493593 and RNASEK expression was further analyzed in whole blood from a subset of the 286

FIND sample (N = 1,416, 23% with diabetes) for whom both genome-wide expression data and 287

genotyping data for 42 established type 2 diabetes variants, including rs75493593, were 288

available. In these individuals, the rs75493593 A allele was also associated with lower RNASEK 289

expression in whole blood (P = 1.6 × 10-5

, Fig. 5D) and increased risk for diabetes (P = 3.0 × 10-

290

3, OR = 1.36, Fig. 5E); and, lower RNASEK expression was associated with an 18% increase per 291

of 41Diabetes

13

standard deviation in the odds of type 2 diabetes (P = 6.0 × 10-3

, OR = 0.82, Fig. 5F). To 292

examine the extent to which the rs75493593-RNASEK eQTL could account for the association 293

between genotype and diabetes, a formal mediation analysis was conducted and significant 294

mediation was observed (P = 0.01). To compare this mediation effect with the potential effect of 295

other type 2 diabetes susceptibility variants in a transcriptomic context, the remaining 41 296

established type 2 diabetes variants genotyped in the 1,416 FIND participants were also 297

analyzed. Six of the 41 variants were associated with type 2 diabetes (P <0.05, data not shown) 298

and these 6 variants along with rs75493593 gave rise to 85 potential eQTL genes (P <0.005 for 299

association between variant and transcript levels, Supplementary Table 8). Only 3 additional 300

eQTLs displayed statistically (P <0.05) significant mediation and none of the 3 displayed a 301

stronger mediation effect than the rs75493593-RNASEK eQTL (Supplementary Table 9). 302

303

DISCUSSION 304

In the SIGMA study, it was noted that among all individuals with type 2 diabetes, those who 305

carry the risk haplotype had lower BMIs than those who do not carry the risk haplotype, 306

suggesting that diabetes onset occurred at a lower BMI among carriers (1). Our study of 307

American Indians identified a significant rs75493593 genotype × BMI interaction, where 308

individuals with a BMI <35 kg/m2 who carry the diabetes risk (A) allele were at increased risk 309

for type 2 diabetes, whereas individuals with higher BMIs who carry the A allele were protected 310

from diabetes. Consistent with the cross-sectional results, for the SWAI and SHS individuals 311

with longitudinal data on BMI and diabetes status, we observed that hazard ratios for type 2 312

diabetes by genotype were higher in the leanest BMI (<25) category and lower in the heaviest 313

BMI category (35+). Since analyses of diabetes incidence, by necessity, exclude prevalent cases 314

of 41 Diabetes

14

at baseline, the hazard ratios from the longitudinal analyses may not be reflective of lifetime 315

diabetes risk. However, the cross-sectional and longitudinal results together, suggest that BMI 316

may have a modulatory effect on the development of type 2 diabetes in individuals who carry the 317

rs75493593 A allele. We also detected a strong rs75493593 genotype × diabetes interaction for 318

BMI in the American Indians where an association between rs75493593 and BMI only occurred 319

in individuals with type 2 diabetes. However, the cross-sectional analyses cannot determine 320

whether individuals were leaner before or after their onset of type 2 diabetes; i.e., are only lean 321

individuals at increased risk for type 2 diabetes if they carry the A allele, or do individuals with 322

the risk allele who develop type 2 diabetes lose more weight as a consequence of the disease. It 323

has previously been shown that American Indians, like individuals in other ethnic groups, often 324

lose weight after they develop diabetes (25). Analysis of the SWAI individuals with longitudinal 325

data suggests that individuals with the risk (A) allele lose weight after developing type 2 326

diabetes. There was no difference in BMI by genotype in individuals who had not developed 327

type 2 diabetes, but a nominal association (P = 0.02) between BMI and genotype was observed 328

in individuals following onset of the disease, where individuals carrying the diabetes risk allele 329

had greater weight loss. It is unclear whether the weight loss is due to a more severe 330

pathophysiology of diabetes or medical treatment. Nevertheless, the association between 331

rs75493593 and post-diabetes weight loss observed in the SWAI sample is very modest 332

suggesting that there may be other genetic or physiological reasons for the strong association 333

between the diabetes risk allele and lower BMI only in the individual with diabetes. Although 334

analysis of in vivo metabolic data available for a subset of SWAI subjects did not provide a 335

conclusive explanation of the physiological link between rs75493593 and type 2 diabetes, a 336

modest but reproducible effect on lipid oxidation rates was observed which is consistent with the 337

of 41Diabetes

15

suggestion that SLC16A11 may be involved in lipid metabolism (1). Our genome-wide eQTL 338

analysis suggests that rs75493593 or a variant in high LD may also affect RNASEK expression 339

which is located approximately 27 kb upstream of SLC16A11, where individuals carrying the 340

type 2 diabetes risk (A) allele have lower RNASEK expression levels compared to individuals 341

homozygous for the C allele. Based on these results observed in 3 different tissues which were 342

biopsied from different individuals, we propose that RNASEK expression may be mediating, at 343

least in part, the association between rs75493593 and type 2 diabetes. Our mediation analyses 344

demonstrate that the pattern of associations is consistent with this hypothesis; however, these 345

results could be confounded by complex pleiotropic relationships among the traits or linkage 346

disequilibrium between causal variants that independently influence RNASEK expression and 347

diabetes risk. Studies in additional populations and molecular mechanistic analyses are required 348

to evaluate this hypothesis more rigorously. Very little is known regarding RNASEK except that 349

it is expressed in most tissues and functions as an endoribonuclease that cleaves ApU and ApG 350

phosphodiester bonds (25,26). Additional studies are needed to determine how RNASEK may 351

play a role in the development of type 2 diabetes. 352

In conclusion, our study identified a modest association between the SLC16A11 variant 353

rs75493593 and type 2 diabetes in 12,811 American Indians, where the effect on diabetes was 354

much more pronounced in non-obese individuals. Analysis of longitudinal measures of BMI 355

suggests that at least some of the difference in BMI is due to greater weight loss in those that 356

carry the A risk allele and develop type 2 diabetes. Rs75493593 was also associated with 357

RNASEK gene expression suggesting that either altered SLC16A11 protein function, RNASEK 358

gene expression, or both may be contributing to the phenotypic associations observed in this 359

study. 360

of 41 Diabetes

16

Author Contributions 361 M.T. and R.L.H. contributed to the study design, researched data, and wrote the manuscript; 362

A.M., S.K., P.P., S.C., J.E.C., J.B., H.G., S.K., R.G.N., and B.H. contributed data; W.C.K. 363

contributed to the study design and contributed data; L.J.B. and C.B. contributed to the study 364

design, researched data, and wrote the manuscript. C.B. is the guarantor of this work and, as 365

such, had full access to all the data in the study and takes responsibility for the integrity of the 366

data and the accuracy of the data analysis. 367

368

Acknowledgments 369 We gratefully acknowledge all of the American Indian volunteers included in this study. This 370

research was supported by the Intramural Research Program of the NIH, NIDDK. The Strong 371

Heart Study was supported by cooperative agreement grants U01-HL41642, U01-HL41652, 372

U01-HL41654, U01-HL65520, and U01-HL65521 from the National Heart, Lung, and Blood 373

Institute. This study utilized the computational resources of the Biowulf system at the National 374

Institutes of Health, Bethesda, MD (http://biowulf.nih.gov). 375

376

Disclaimer 377 The opinions expressed in this paper are those of the author(s) and do not necessarily reflect the 378

views of the Indian Health Service who helped support the Strong Heart Study. 379

380

Duality of Interest 381 No potential conflicts of interest relevant to this article were reported. 382

383

384

385

386

387

388

389

390

391

392

393

394

of 41Diabetes

17

REFERENCES 395

1. SIGMA Type 2 Diabetes Consortium, Williams AL, Jacobs SB, Moreno-Macías

H, Huerta-Chagoya A, Churchhouse C, Márquez-Luna C, García-Ortíz H, Gómez-

Vázquez MJ, Burtt NP, Aguilar-Salinas CA, González-Villalpando C, Florez JC,

Orozco L, Haiman CA, Tusié-Luna T, Altshuler D. Sequence variants in

SLC16A11 are a common risk factor for type 2 diabetes in Mexico. Nature

2014;506:97–101

2. Hanson RL, Rong R, Kobes S, Muller YL, Weil EJ, Curtis JM, Nelson RG, Baier

LJ. The role of established type 2 diabetes-susceptibility genetic variants in a high-

prevalence American Indian population. Diabetes. 2015 Feb 9. [Epub ahead of

print]

3. Knowler WC, Bennett PH, Hamman RF, Miller M. Diabetes incidence and

prevalence in Pima Indians: a 19-fold greater incidence than in Rochester,

Minnesota. Am J Epidemiol 1978;108:497–505

4. American Diabetes Association Expert Committee on the Diagnosis and

Classification of Diabetes Mellitus. Report of the Expert Committee on the

Diagnosis and Classification of Diabetes Mellitus. Diabetes Care 1997;20:1183–

1197

5. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, Turner RC.

Homeostasis model assessment: insulin resistance and beta-cell function from

fasting plasma glucose and insulin concentrations in man. Diabetologia

1985;28:412–419

6. Knowler WC, Coresh J, Elston RC, Freedman BI, Iyengar SK, Kimmel PL, Olson

JM, Plaetke R, Sedor JR, Seldin MF. The Family Investigation of Nephropathy and

Diabetes (FIND): design and methods. J Diabetes Complications 2005;19:1–9

7. American Diabetes Association. Diagnosis and classification of diabetes mellitus.

Diabetes Care 2010;33 (Suppl 1):S62–69

8. Lee ET, Welty TK, Fabsitz R, Cowan LD, Le NA, Oopik AJ, Cucchiara AJ,

Savage PJ, Howard BV. The Strong Heart Study. A study of cardiovascular disease

in American Indians: design and methods. Am J Epidemiol 1990;132:1141–1155

9. Alberti KGMM, Zimmet PZ. Definition diagnosis and classification of diabetes

mellitus and its complication part 1: diagnosis and classification of diabetes

mellitus provisional report of a WHO consultation. Diabet Med 1998;15:539–553

10. Lillioja S, Mott DM, Spraul M, Ferraro R, Foley JE, Ravussin E, Knowler WC,

Bennett PH, Bogardus C. Insulin resistance and insulin secretory dysfunction as

precursors of non-insulin-dependent diabetes mellitus. Prospective studies of Pima

Indians. N Engl J Med 1993;329:1988–1992

11. Norman RA, Tataranni PA, Pratley R, Thompson DB, Hanson RL, Prochazka M,

Baier L, Ehm MG, Sakul H, Foroud T, Garvey WT, Burns D, Knowler WC,

Bennett PH, Bogardus C, Ravussin E. Autosomal genomic scan for loci linked to

obesity and energy metabolism in Pima Indians. Am J Hum Genet 1998;62:659–

668

12. Ravussin E, Lillioja S, Anderson TE, Christin L, Bogardus C. Determinants of 24-

hour energy expenditure in man. Methods and results using a respiratory chamber.

J Clin Invest 1986;78:1568–1578

of 41 Diabetes

18

13. Devlin B, Bacanu SA, Roeder K. Genomic control to the extreme. Nat Genet

2004;35:1129–1130

14. Tian C, Hinds DA, Shigeta R, Adler SG, Lee A, Pahl MV, Silva G, Belmont JW,

Hanson RL, Knowler WC, Gregersen PK, Ballinger DG, Seldin MF. A genome-

wide single nucleotide polymorphism panel for Mexican American admixture

mapping. Am J Hum Genet 2007;80:1014–1023

15. Hanis CL, Chakraborty R, Ferrell RE, Schull WJ. Individual admixture estimates:

disease associations and individual risk of diabetes and gallbladder disease in

Mexican-Americans in Starr County, Texas. Am J Phys Anthropol 1986;70:433–

441

16. Petitti DB. Statistical methods in meta-analysis. In Meta-Analysis, Decision

Analysis and Cost-Effectiveness Analysis: Methods for Quantitative Synthesis in

Medicine. Petitti DB, Ed. Oxford, Oxford University Press, 2000, p.94–118

17. Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in

meta-analysis. Brit Med J 2003;327:557–560

18. Mason CC, Hanson RL, Ossowski V, Bian L, Baier LJ, Krakoff J, Bogardus C.

Bimodal distribution of RNA expression levels in human skeletal muscle tissue.

BMC Genomics 2011;12:98

19. Göring HH, Curran JE, Johnson MP, Dyer TD, Charlesworth J, Cole SA, Jowett

JB, Abraham LJ, Rainwater DL, Comuzzie AG, Mahaney MC, Almasy L,

MacCluer JW, Kissebah AH, Collier GR, Moses EK, Blangero J. Discovery of

expression QTLs using large-scale transcriptional profiling in human lymphocytes.

Nat Genet 2007;39:1208–1216

20. Barbosa-Morais NL, Dunning MJ, Samarajiwa SA, Darot JF, Ritchie ME, Lynch

AG, Tavaré S.A re-annotation pipeline for Illumina BeadArrays: improving the

interpretation of gene expression data. Nucleic Acids Res 2010;38:e17

21. Baron RM, Kenny DA. The moderator-mediator variable distinction in social

psychological research: conceptual, strategic, and statistical considerations. J Pers

Soc Psychol 1986;51:1173–1182

22. Preacher KJ, Kelley K. Effect size measures for mediation models: quantitative

strategies for communicating indirect effects. Psychol Methods 2011;16:93–115

23. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and

powerful approach to multiple testing. J R Stat Soc B 1995;57:289–300

24. Kumar P, Henikoff S, Ng PC. Predicting the effects of coding non-synonymous

variants on protein function using the SIFT algorithm. Nat Protoc 2009;4:1073–

1081

25. Looker HC, Knowler WC, Hanson RL. Changes in BMI and weight before and

after the development of type 2 diabetes. Diabetes Care 2001;24:1917–1922

25. Kiritsi MN, Fragoulis EG, Sideris DC. Essential cysteine residues for human

RNase K catalytic activity. FEBS J 2012;279:1318–1326

26. Economopoulou MA, Fragoulis EG, Sideris DC. Molecular cloning and

characterization of the human RNase kappa, an ortholog of Cc Rnase. Nucleic

Acids Res 2007;35:6389–6398

396 397

398

of 41Diabetes

19

Table footnotes and figure legends 399

400

Table 1―Diabetes association results for rs75493593 in the American Indian samples 401 *Combined = SWAI + FIND + SHS subjects.

†The A allele for rs75493593 is the diabetes risk 402

allele in the Mexican (SIGMA) study. ‡P-value for the SWAI sample was adjusted for age, sex, 403

birth-year, admixture estimate, and American Indian heritage. P-value for the FIND sample was 404

adjusted for age, sex, admixture estimate, and tribal membership; the P-value was then adjusted 405

by genomic control to account for any relatedness or population stratification. P-value for the 406

SHS sample was adjusted for age, sex, admixture estimate, and study center. P-value for the 407

combined analysis was adjusted for age, sex, and study center. §OR, odds ratios, for rs75493593 408

are given per copy of the A allele. ||The proportion of variance (I

2) in the effect estimates 409

attributable to heterogeneity between the 3 populations was 65% (P = 0.06). 410

411

Figure 1―Odds ratios for the associations between rs75493593 and type 2 diabetes when 412

the subjects are stratified by BMI 413 P-values for the SWAI sample were adjusted for age, sex, birth-year, admixture estimate, and 414

American Indian heritage. P-values for the FIND sample were adjusted for age, sex, admixture 415

estimate, and tribal membership; the P-values were then adjusted by genomic control to account 416

for any relatedness or population stratification. P-values for the SHS sample were adjusted for 417

age, sex, admixture estimate, and study center. OR, odds ratios are given per copy of the A 418

allele. Error bars indicate 95% confidence interval (CI). Combined = SWAI + FIND + SHS. P-419

values for BMI by genotype interaction on the risk of having diabetes are P = 5.5 × 10-6

(SWAI), 420

P = 0.03 (FIND), P = 0.19 (SHS) and P = 3.4 × 10-7

(Combined). 421

422

Figure 2―Longitudinal analysis of diabetes risk as a function of rs75493593 and obesity-423

status 424 Hazard ratios for the SWAI, SHS, and combined samples after stratifying by BMI. P-values 425

were adjusted for baseline BMI, baseline age, birth-year, admixture estimate, and American 426

Indian heritage. Hazard ratios and 95% confidence intervals were estimated using the Cox 427

proportional hazards model. Pinteraction is the P-value for BMI by rs75493593 genotype interaction 428

on the risk of developing diabetes. 429

430

Table 2―BMI association results for rs75493593 in the American Indian samples 431

*Combined = SWAI + FIND + SHS.

†P-values for the SWAI sample were adjusted for age, sex, 432

birth-year, admixture estimate, and American Indian heritage. P-values for the FIND sample 433

were adjusted for age, sex, admixture estimate, and tribal membership; the P-values were then 434

adjusted by genomic control to account for any relatedness or population stratification. P-values 435

for the SHS sample were adjusted for age, sex, admixture estimate, and study center. 436

437

Figure 3―Longitudinal BMI pattern when non-diabetic and after diabetes onset in the 438

SWAI and SHS individuals for rs75493593 439 The longitudinal relationship with change in BMI for the (A) SWAI and (B) SHS individuals. P-440

value for pre-diabetes was adjusted for sex, birth-year, baseline age, and American Indian 441

heritage. P-value for post-diabetes was adjusted for sex, birth-year, diabetes duration, and 442

American Indian heritage. 443

444

of 41 Diabetes

20

Figure 4―Association analyses for rs75493593 and carbohydrate and lipid oxidation 445 (A) Association between rs75493593 and carbohydrate intake, carbohydrate oxidation, and 446

carbohydrate balance. (B) Association between rs75493593 and lipid intake, lipid oxidation, and 447

lipid balance. Genotype N = CC, 145; CA, 179; AA, 66. P-values for carbohydrate and lipid 448

intake and carbohydrate and lipid balance were adjusted for age, sex, percent body fat, family 449

membership, and American Indian heritage. P-values for carbohydrate and lipid oxidation were 450

adjusted for age, sex, percent body fat, family membership, American Indian heritage, and 451

energy balance. 452

453

Table 3―Top 10 eQTL genes for rs75493593 in adipose tissue and skeletal muscle 454 Transcript cluster start and stop locations are from build 37. 455

456

Figure 5―Rs75493593-RNASEK eQTL results for the SWAI and FIND subjects, and 457

diabetes prevalence based on rs75493593 genotypes and RNASEK expression for the FIND 458

subjects. 459 (A) Exon microarray mean RNASEK expression levels based on rs75493593 genotypes for 460

adipose tissue. (B) Exon microarray mean RNASEK expression levels based on rs75493593 461

genotypes for skeletal muscle. (C) RT-PCR association results for rs75493593 and mean 462

RNASEK gene expression in adipose tissue. RT-PCR assays were done using the standard curve 463

method. RNASEK mRNA expression levels were normalized with beta actin. P-values were 464

adjusted for age, sex, and American Indian heritage. (D) Mean RNASEK expression in whole 465

blood based on rs75493593 genotype. Beta is the mean difference in RNASEK expression and is 466

shown in standard deviation (SD) units per copy of the A allele. P-value adjusted for age, sex, 467

and American Indian heritage (E) Diabetes prevalence based on rs75493593 genotype for the 468

1,416 American Indians from the FIND study. Odds ratio is expressed per copy of the A allele. 469

P-value adjusted for age, sex, and American Indian heritage. (F) Diabetes prevalence based on 470

RNASEK expression quartiles. N ≈ 353 for each quartile. Odds ratio is expressed per standard 471

deviation (SD) of expression. P-value adjusted for age, sex, American Indian heritage, and 472

genotype. 473

474

of 41Diabetes

21

Table 1―Diabetes association results for rs75493593 in the American Indian samples

Genotype N

for diabetic subjects

Genotype N

for non-diabetic subjects

Subjects*

Allele

A† freq

C/C C/A A/A C/C C/A A/A P‡ OR (95% CI)

§,||

SWAI 0.39 840 1142 406 1879 2292 736 0.38 1.04 (0.95–1.14)

FIND 0.44 265 392 209 763 1036 430 0.008 1.27 (1.10–1.45)

SHS 0.40 369 490 202 566 618 176 0.04 1.15 (1.01–1.30)

Combined 0.41 1474 2024 817 3208 3946 1342 0.001 1.11 (1.04–1.19)

of 41 Diabetes

22

Table 2―BMI association results for rs75493593 in the American Indian samples

BMI association based on diabetes status

Genotype × diabetes interaction Genotype N Mean BMI ± SD

Subjects* N B Pinteraction

† C/C C/A A/A C/C C/A A/A B P

†

SWAI 6269 -0.054 1.1 × 10-10

All 2295 2958 1016 34.5 ± 8.7 34.7 ± 8.6 33.6 ± 7.8 -0.018 5.0 × 10-5

With diabetes 803 1099 388 36.9 ± 8.7 35.5 ± 8.8 33.3 ± 8.0 -0.051 4.2 × 10-14

Without diabetes 1492 1859 628 33.3 ± 8.4 34.2 ± 8.5 33.7 ± 7.7 0.001 0.84

FIND 3088 -0.031 0.02

All 1027 1423 638 31.7 ± 7.6 31.5 ± 7.2 31.1 ± 7.0 -0.0007 0.91

With diabetes 265 391 209 35.2 ± 7.2 34.4 ± 7.8 33.3 ± 7.3 -0.029 5.5 × 10-3

Without diabetes 762 1032 429 30.4 ± 7.3 30.4 ± 6.7 30.1 ± 6.6 0.004 0.55

SHS 2411 -0.013 0.23

All 933 1101 377 30.5 ± 6.1 30.7 ± 6.2 30.1 ± 5.7 -0.012 0.03

With diabetes 367 484 202 32.5 ± 6.4 32.2 ± 6.7 31.0 ± 5.7 -0.02 0.01

Without diabetes 566 617 175 29.2 ± 5.6 29.4 ± 5.6 29.1 ± 5.6 -0.012 0.13

Combined 11768 -0.037 3.4 × 10-10

All 4255 5482 2031 33.0 ± 8.1 33.0 ± 8.0 32.2 ± 7.4 -0.012 6.4 × 10-5

With diabetes 1435 1974 799 35.4 ± 8.1 34.5 ± 8.2 32.7 ± 7.4 -0.037 3.4 × 10-15

Without diabetes 2820 3508 1232 31.7 ± 7.8 32.2 ± 7.8 31.8 ± 7.3 -0.001 0.77

of 41Diabetes

23

Table 3―Top 10 eQTL genes for rs75493593 in adipose tissue and skeletal muscle

Adipose Skeletal muscle

eQTL gene Chr Transcript

cluster ID

Transcript

cluster start

Transcript

cluster stop P eQTL gene Chr

Transcript

cluster ID

Transcript

cluster start

Transcript

cluster stop P

FAM53C 5 2830698 137673257 137685632 7.7 × 10-5

RNASEK 17 3708186 6915756 6917837 1.7 × 10-8

RNASEK 17 3708186 6915756 6917837 2.1 × 10-4

IGSF8 1 2439975 160061138 160119332 1.7 × 10-5

SIRPA 20 3873629 1858178 1926239 9.0 × 10-4

SPR 2 2488584 73114532 73119282 8.6 × 10-5

RSPH10B 7 3037100 5965177 6010420 9.9 × 10-4

ARFGEF1 8 3139035 68087455 68255892 1.3 × 10-4

DUPD1 10 3295235 76796453 76833383 1.2 × 10-3

ENC1 5 2862696 73923235 73937249 2.6 × 10-4

TNNT3 11 3317117 1940940 1959931 1.4 × 10-3

NXNL1 19 3854477 17547267 17571725 2.7 × 10-4

KCNA6 12 3401953 4918308 4923944 1.4 × 10-3

C1orf222 1 4053322 1853396 1859457 2.9 × 10-4

CCRL2 3 4047185 46442652 46454468 1.4 × 10-3

PUS1 12 3438581 132413765 132430421 3.9 × 10-4

TMEM92 17 3726298 48306764 48358846 1.5 × 10-3

L3MBTL3 6 2925510 130339608 130462583 4.1 × 10-4

FHL2 2 2568687 105969445 106134573 1.7 × 10-3

C19orf36 19 3816225 2084101 2099580 7.7 × 10-5

of 41 Diabetes

254x338mm (300 x 300 DPI)

of 41Diabetes

254x338mm (300 x 300 DPI)

of 41 Diabetes

254x338mm (300 x 300 DPI)

of 41Diabetes

254x338mm (300 x 300 DPI)

of 41 Diabetes

254x338mm (300 x 300 DPI)

of 41Diabetes

SUPPLEMENTARY TEXT

Mediation analyses

Analyses of the extent to which gene expression (X) may mediate the association between

genotype (G) and diabetes (D) were conducted in the FIND samples in which gene expression

had been measured in peripheral blood (N =1,416). These analyses were conducted by the Sobel

method (21) in which the following regression models were fit:

D = cG + ∑1 (1)

X = aG + ∑2 (2)

D = bX + c′G + ∑3 (3)

where ∑i represent the intercept and covariates of interest for each equation. Equations 1 and 3

were fit by logistic regression, while equation 2 was fit by linear regression. The mediation effect

was taken as ab, and its standard error (21) as:

The standard errors for a and b (SEa, SEb) were adjusted by the genomic control procedure to

account for population stratification as described in the text. The percentage mediation, or the

percentage of the genotypic effect on diabetes potentially explained by its effect on expression,

was calculated as (22):

100 1 .

The primary mediation analyses were conducted for the SLC16A11 rs75493593 variant and

RNASEK expression, as there is consistent evidence across multiple tissues that this is an eQTL

(see text). However, the theoretical distribution of the mediation test is a complex function of

the individual associations and to interpret these results in a genomic context, we also analyzed

the additional 41 established type 2 diabetes genes typed in the FIND participants. In addition to

of 41 Diabetes

rs75493593, 6 of the variants, rs10906115 (CDC123), rs2237892 (KCNQ1), rs231362 (KCNQ1),

rs8050136 (FTO), rs4607517 (GCK), and rs3923113 (GRB14), were nominally associated (P

<0.05) with diabetes in this sample. These 7 SNPs were then tested for association with all

15,854 transcripts genome-wide to identify potential eQTLs. EQTLs identified with nominal P

<0.005 were further subjected to mediation analysis. This threshold was chosen to represent

associations of comparable magnitude to the rs75493593-RNASEK eQTL and not to identify

robust eQTL associations. To correct for multiple comparisons across all 110,978 SNP-transcript

pairs, we also calculated the false discovery rate (23). Among the 85 potential eQTLs identified

with nominal P <0.005, only 4 had P <0.05 for mediation and the strongest result was for the

rs75493593-RNASEK eQTL. These results demonstrate that the pattern of associations is

consistent with partial mediation of the rs75493593-diabetes association by a genotypic effect on

RNASEK transcription, and they show that the pattern of mediation is stronger than expected for

eQTL identified across the transcriptome with other diabetes associated variants of comparable

effect.

of 41Diabetes

rs17203120 (

0.4

3)

rs75075014 (

0.4

3)

rs145952323 (

0.4

1)

rs7225198 (

0.0

04)

rs75493593 (

0.4

3)

rs148775056 (

0.1

1)

rs75418188 (

0.4

1)

rs13342232 (

0.4

1)

rs13342692 (

0.4

2)

rs11776786 (

0.4

1)

rs75636181 (

0.0

04)

rs2292351 (

0.4

3)

rs77151854 (

0.0

04)

rs77086571 (

0.4

2)

rs9303212 (

0.2

0)

rs74577409 (

0.4

3)

rs140910714 (

0.0

04)

rs8078000 (

0.0

09)

rs114815410 (

0.0

02)

rs9674813 (

0.2

0)

3’ region (1 kb) 5’ region (1 kb)

A

B

3’ region (1 kb) 5’ region (1 kb)

* * * * *

Supplementary Figure 1

(A) Schematic showing the 20 variants located in the SLC16A11 region identified in SWAIs by

whole-genome sequencing. Numbers in parentheses indicate minor allele frequencies. (B)

Linkage disequilibrium (r2) plot for the variants with frequencies >0.01. The 3 tag SNPs (r2

≥0.85) that were genotyped in the SWAI sample are underlined. Asterisks (*) indicate the

variants for the 5 SNP haplotype described in the SIGMA type 2 diabetes study1.

of 41 Diabetes

*BMI restricted to subjects ≥age of 15 years, N = 6,646.

Supplementary Table 1―Descriptions for the cohorts

Study samples N N with diabetes (%) Males/Females Age (years) BMI (kg/m2)

Southwestern American Indians in

Arizona (SWAI) 7710 2549 (33.1) 3426/4284 33.9 ± 16.4 34.5 ± 8.6

*

Family Investigation of Nephropathy

and Diabetes (FIND) 3106 868 (27.9) 1543/1563 36.6 ± 12.9 31.5 ± 7.3

Strong Heart Study (SHS) 2451 1074 (43.8) 1042/1409 56.2 ± 8.1 30.5 ± 6.1

of 41Diabetes

1

*Representative variant for the diabetes risk haplotype described in the Mexican (SIGMA) study.

†P-values adjusted for age, sex, birth

-year, admixture estimate, and American Indian heritage. Odds ratios are given per copy of the A2 allele.

Supplementary Table 2―Diabetes association results for the 3 tag SNPs spanning the SLC16A11 locus for the SWAI sample

Genotype N

for individuals with diabetes

Genotype N

for individuals without diabetes

SNP A1/A2 A2 freq A1/A1 A1/A2 A2/A2 A1/A1 A1/A2 A2/A2 P† OR (95% CI)

rs75493593* C/A 0.39 840 1142 406 1879 2292 736 0.38 1.04 (0.95–1.14)

rs148775056 C/T 0.13 1811 587 54 3855 1042 80 0.41 1.05 (0.93–1.20)

rs9303212 C/A 0.16 1737 654 50 3442 1383 141 0.47 0.96 (0.85–1.08)

of 41 Diabetes

*Combined = SWAI + FIND + SHS.

†P-values for the SWAI sample were adjusted for age, sex, birth-year, admixture estimate, and American Indian heritage. P-values for the

FIND sample were adjusted for age, sex, admixture estimate, and tribal membership; the P-values were then adjusted by genomic control to account for any relatedness or

population stratification. P-values for the SHS sample were adjusted for age, sex, admixture estimate, and study center. ‡OR, odds ratios are given per copy of the A allele.

Supplementary Table 3―Diabetes association for rs75493593 stratified by BMI category

Diabetes association stratified by BMI category

Genotype × BMI interaction Diabetes prevalence per genotype

Samples* N B Pinteraction

† BMI (kg/m

2) N C/C C/A A/A B P

† OR (95% CI)

‡

SWAI 6269 -0.89 5.5 × 10-6

<25 702 0.16 0.26 0.38 0.42 0.05 1.53 (1.01–2.31)

25–30 1327 0.28 0.39 0.42 0.29 4.2 × 10-3

1.34 (1.10–1.63)

30–35 1575 0.35 0.38 0.4 0.16 0.07 1.17 (0.99–1.40)

35–40 1258 0.43 0.37 0.34 -0.26 7.4 × 10-3

0.77 (0.64–0.93)

40+ 1407 0.45 0.41 0.35 -0.19 0.04 0.83 (0.69–0.99)

FIND 3088 -0.66 0.04 <25 522 0.08 0.12 0.17 0.39 0.12 1.47 (0.91–2.38)

25–30 953 0.16 0.23 0.27 0.53 1.2 × 10-4

1.69 (1.29–2.21)

30–35 778 0.3 0.27 0.32 0.1 0.48 1.11 (0.84–1.47)

35–40 454 0.44 0.37 0.53 0.19 0.28 1.20 (0.86–1.68)

40+ 381 0.46 0.51 0.5 0.002 0.99 1.00 (0.70–1.43)

SHS 2411 -0.47 0.19 <25 406 0.25 0.24 0.43 0.26 0.17 1.30 (0.89–1.88)

25–30 818 0.32 0.4 0.48 0.15 0.22 1.16 (0.92–1.46)

30–35 700 0.43 0.49 0.6 0.32 6.7 × 10-3

1.38 (1.09–1.74)

35–40 326 0.56 0.58 0.64 0.008 0.96 1.01 (0.70–1.45)

40+ 161 0.69 0.58 0.64 -0.48 0.08 0.62 (0.36–1.07)

Combined 11768 -0.77 5.1 × 10-7

<25 1630 0.16 0.21 0.32 0.35 4.9 × 10-3

1.41 (1.11–1.80)

25–30 3098 0.26 0.34 0.38 0.31 7.5 × 10-6

1.35 (1.18–1.54)

30–35 3053 0.36 0.38 0.42 0.19 2.2 × 10-3

1.22 (1.07–1.38)

35–40 2038 0.45 0.4 0.44 -0.11 0.11 0.88 (0.76–1.03)

40+ 1949 0.47 0.44 0.41 -0.17 0.03 0.84 (0.72–0.98)

of 41Diabetes

Supplementary Table 4―Distribution of participants according to diabetes status and rs75493593 genotype

stratified by BMI category

Genotype N (mean BMI )

Participants and diabetes status BMI (kg/m2) C/C C/A A/A Freq (A)

SWAI

Without diabetes <25 212 (22.2) 240 (22.3) 78 (22.4) 0.37

25-30 368 (27.6) 360 (27.6) 135 (27.6) 0.37

30-35 360 (32.3) 470 (32.5) 160 (32.5) 0.40

35-40 258 (37.2) 396 (37.3) 122 (37.0) 0.41

40+ 294 (46.2) 394 (46.5) 133 (45.1) 0.40

With diabetes <25 40 (23.1) 84 (22.9) 48 (22.8) 0.52

25-30 141 (27.8) 225 (27.7) 98 (27.4) 0.45

30-35* 194 (32.6) 285 (32.5) 106 (32.1) 0.42

35-40 191 (37.4) 229 (37.2) 63 (37.4) 0.37

40+ 237 (47.5) 276 (47.3) 73 (46.4) 0.36

FIND

Without diabetes <25* 167 (22.6) 211 (22.4) 85 (22.4) 0.41

25-30 260 (27.4) 332 (27.4) 156 (27.7) 0.43

30-35 177 (32.1) 261 (32.1) 114 (32.0) 0.44

35-40 78 (37.4) 148 (37.1) 39 (37.3) 0.43

40+ 80 (46.1) 80 (45.4) 35 (45.5) 0.38

With diabetes <25 14 (22.8) 28 (23.0) 17 (22.6) 0.53

25-30 49 (27.4) 97 (27.7) 59 (27.3) 0.52

30-35 74 (32.5) 98 (32.3) 54 (32.0) 0.46

35-40 60 (37.0) 85 (37.3) 44 (37.2) 0.46

40+ 68 (44.7) 83 (45.8) 35 (45.3) 0.41

SHS

Without diabetes <25 128 (22.4) 133 (22.5) 35 (22.2) 0.34

25-30 213 (27.7) 221 (27.5) 71 (27.3) 0.36

30-35 213 (27.7) 221 (27.5) 71 (27.3) 0.35

35-40* 56 (37.3) 65 (36.9) 17 (36.5) 0.36

40+ 20 (44.7) 31 (43.0) 8 (44.5) 0.40

With diabetes <25* 42 (23.5) 42 (22.7) 26 (22.7) 0.43

25-30 99 (27.9) 148 (27.8) 66 (27.6) 0.45

30-35 111 (32.4) 163 (32.4) 66 (32.2) 0.43

35-40 70 (37.2) 88 (37.1) 30 (36.8) 0.39

40+ 45 (44.1) 43 (46.2) 14 (44.0) 0.35

Freq (A) is the frequency of the A allele. *P <0.05 for difference in BMI.

of 41 Diabetes

Supplementary Table 5―Association results for rs75493593 with quantitative type 2 diabetes-related traits in American Indians (N = 5,200–5,600) from the SWAI sample

Stratified by BMI

Genotype × BMI interaction N based on genotype Means ± SD based on genotype

Trait N B Pinteraction* BMI C/C C/A A/A C/C C/A A/A B P

*

Fasting glucose (mg/dl) 1921 -0.40 0.63 <25 230 252 85 87.2 ± 9.6 87.4 ± 8.2 89.1 ± 8.6 0.78 0.14

25-30 402 414 168 90.8 ± 8.9 91.8 ± 9.0 91.3 ± 7.9 0.12 0.74

30-35 501 662 224 94.4 ± 9.7 94.7 ± 10.2 94.8 ± 10.9 0.32 0.42

35-40 383 511 164 95.6 ± 10.9 96.2 ± 9.7 97.3 ± 9.9 0.56 0.21

40+ 405 567 179 98.7 ± 10.9 98.7 ± 10.5 98.7 ± 9.8 -0.19 0.67

2 hr glucose (mg/dl) 1921 -1.39 0.59 <25 230 252 85 95.8 ± 28.1 96.7 ± 25.1 98.1 ± 31.7 -0.77 0.66

25-30 402 414 168 105.7 ± 30.6 106.0 ± 31.0 103.8 ± 32.4 -1.60 0.25

30-35 501 662 224 116.8 ± 32.2 118.8 ± 31.5 116.9 ± 31.2 -0.01 0.99

35-40 383 511 164 122.5 ± 30.4 119.2 ± 31.4 125.2 ± 35.7 -0.13 0.93

40+ 405 567 179 129.2 ± 32.7 123.9 ± 30.1 126.2 ± 30.9 -2.85 0.03

Fasting insulin† 1919 0.15 0.03 <25 230 251 85 -1.07 ± 0.80 -1.14 ± 0.33 -1.16 ± 1.0 -0.10 0.09

25-30 401 414 167 -0.55 ± 0.85 -0.53 ± 0.75 -0.64 ± 0.77 -0.07 0.05

30-35 501 661 223 -0.10 ± 0.83 -0.02 ± 0.78 -0.02 ± 0.78 0.03 0.30

35-40 383 508 164 0.31 ± 0.80 0.35 ± 0.69 0.36 ± 0.76 0.03 0.41

40+ 404 563 178 0.62 ± 0.77 0.68 ± 0.80 0.69 ± 0.66 0.01 0.70

2 hr insulin† 1883 0.14 0.07 <25 225 247 83 -0.67 ± 0.91 -0.65 ± 0.85 -0.83 ± 1.22 -0.10 0.10

25-30 397 402 165 -0.31 ± 0.96 -0.34 ± 1.04 -0.37 ± 0.99 -0.07 0.12

30-35 487 654 217 0.08 ± 0.97 0.19 ± 0.92 0.21 ± 0.95 0.04 0.27

35-40 378 506 164 0.39 ± 0.87 0.30 ± 0.92 0.39 ± 0.87 -0.02 0.57

40+ 396 559 174 0.50 ± 0.84 0.53 ± 0.77 0.56 ± 0.77 -0.005 0.89

HOMA-IR† 1919 0.13 0.04 <25 230 251 85 -1.19 ± 0.71 -1.24 ± 0.74 -1.24 ± 0.88 -0.07 0.17

25-30 401 414 167 -0.69 ± 0.76 -0.67 ± 0.68 -0.76 ± 0.69 -0.06 0.05

30-35 501 661 223 -0.26 ± 0.75 -0.20 ± 0.70 -0.20 ± 0.72 0.03 0.33

35-40 383 508 164 0.10 ± 0.72 0.14 ± 0.63 0.16 ± 0.68 0.03 0.30

40+ 404 563 178 0.40 ± 0.70 0.45 ± 0.72 0.47 ± 0.60 0.01 0.63

HOMA-B† 1914 0.12 0.04 <25 226 251 85 -0.26 ± 0.78 -0.32 ± 0.74 -0.40 ± 0.85 -0.11 0.03

25-30 400 413 167 0.005 ± 0.71 -0.03 ± 0.60 -0.09 ± 0.66 -0.06 0.04

30-35 501 660 223 0.23 ± 0.65 0.29 ± 0.63 0.31 ± 0.60 0.03 0.25

35-40 383 508 164 0.54 ± 0.69 0.52 ± 0.58 0.50 ± 0.62 -0.008 0.77

40+ 404 563 178 0.68 ± 0.62 0.71 ± 0.63 0.70 ± 0.55 -0.003 0.90 *P-values were adjusted for age, sex, admixture estimate, and American Indian heritage.

†Insulin measurements were logarithmically transformed and standardized within different

insulin assays using regression equations. Beta values are given per copy of the A allele.

of 41Diabetes

*P-values were adjusted for age, sex, family membership, and American Indian heritage. †P-values were adjusted for age, sex, family membership, percent body fat, and American Indian heritage. ‡P-value was adjusted for age, sex, family membership, percent body fat, American Indian heritage, 30 min glucose, and glucose uptake rate

during low dose insulin stimulation. §P-value was adjusted for age, sex, family membership, percent body fat, American Indian heritage, and glucose uptake rate during

low dose insulin stimulation. ||P-value was adjusted for age, sex, fat mass, fat free mass, family membership, American Indian heritage, and movement. ¶P-values were adjusted for age, sex, family membership, percent body fat, American Indian heritage, and energy balance. #Analysis was restricted to individuals with normal glucose

tolerance: N (males/females) = 143 (97/46), 168 (103/65), and 69 (40/29) for C/C, C/A, and A/A, respectively. Mean ages (yrs) = 26.6 ± 5.8, 26.1 ± 5.8, and 28.8 ± 6.6

for C/C, C/A, and A/A, respectively. **Beta values are given per copy of the A allele. EMBS, estimated metabolic body size.

Supplementary Table 6―Association of rs75493593 with obesity and type 2 diabetes related-traits in non-diabetic American Indians

Means ± SD

Metabolic trait C/C C/A A/A B**

P

Non-diabetic individuals analyzed for metabolic traits

N (males/females) 193 (122/71) 234 (133/101) 94 (51/43) – –

Mean age (yrs) 27.1 ± 6.1 25.9 ± 5.8 27.9 ± 6.4 – –

Adiposity

Percent body fat* 32.0 ± 8.7 32.6 ± 8.7 32.2 ± 8.4 -0.007 0.99

Waist/thigh ratio† 1.64 ± 0.15 1.63 ± 0.18 1.62 ± 0.17 -0.007 0.40

Oral glucose tolerance test (OGTT)

Fasting glucose (mg/dl)† 88.0 ± 9.5 89.1 ± 10.0 89.6 ± 10.9 0.72 0.23

30 min glucose (mg/dl)† 148.4 ± 27.0 142.0 ± 23.7 146.0 ± 26.5 -2.06 0.22

2 hr glucose (mg/dl)† 121.9 ± 29.9 121.6 ± 29.7 123.3 ± 32.8 0.024 0.99

Fasting insulin (μU/ml)† 1.55 ± 0.21 1.55 ± 0.19 1.53 ± 0.23 -0.005 0.62

30 min insulin (μU/ml)‡,#

2.35 ± 0.26 2.31 ± 0.27 2.28 ± 0.24 -0.027 0.08

2 hr insulin (μU/ml)† 2.17 ± 0.33 2.17 ± 0.34 2.14 ± 0.36 -0.015 0.43

Intravenous glucose tolerance test (IVGTT)

Acute insulin response (μU/ml)§,#

2.32 ± 0.27 2.30 ± 0.29 2.25 ± 0.26 -0.018 0.33

Hyperinsulinemic-euglycemic clamp

Basal glucose output (mg ∙ kgEMBS-1

∙ min-1

)† 1.89 ± 0.22 1.92 ± 0.25 1.91 ± 0.22 0.015 0.26

Basal carbohydrate oxidation (mg ∙ kgEMBS-1

∙ min-1

)† 1.39 ± 0.47 1.40 ± 0.46 1.50 ± 0.44 0.049 0.08

Basal lipid oxidation (mg ∙ kgEMBS-1

∙ min-1

)† 0.73 ± 0.25 0.72 ± 0.25 0.67 ± 0.24 -0.027 0.07

Glucose uptake rate (mg ∙ kgEMBS-1

∙ min-1

)log10† 0.55 ± 0.12 0.56 ± 0.12 0.57 ± 0.13 0.009 0.14

Carbohydrate oxidation (mg ∙ kgEMBS-1

∙ min-1

)† 2.07 ± 0.60 2.10 ± 0.54 2.17 ± 0.52 0.03 0.37

Lipid oxidation (mg ∙ kgEMBS-1

∙ min-1

)† 0.44 ± 0.28 0.41 ± 0.25 0.37 ± 0.27 -0.022 0.19

Subset of individuals who participated in the metabolic chamber studies

N (males/females) 145 (92/53) 179 (103/76) 66 (39/27) – –

Mean age (yrs) 28.0 ± 6.3 26.8 ± 6.1 29.1 ± 6.8 – –

Percent body fat* 32.2 ± 8.8 33.4 ± 7.8 32.0 ± 7.7 -0.11 0.82

Energy balance (kcal/day)† -87.17 ± 191 -90.21 ± 183 -57.41 ± 196 10.8 0.39

24 hr energy expenditure (kcal/day)|| 2359 ± 400 2357 ± 404 2309 ± 384 -0.83 0.93

24 hr respiratory quotient¶ 0.85 ± 0.02 0.85 ± 0.02 0.86 ± 0.03 0.004 0.01

Carbohydrate intake (kcal/day)† 1135 ± 174 1131 ± 171 1130 ± 160 1.35 0.86

Lipid intake (kcal/day)† 681 ± 104 680 ± 102 676 ± 98 0.66 0.88

Protein intake (kcal/day)† 454 ± 69 453 ± 68 450 ± 66 0.40 0.89

Carbohydrate oxidation (kcal/day)¶ 1061 ± 230 1068 ± 230 1145 ± 209 38.4 0.007

Lipid oxidation (kcal/day)¶ 985 ± 302 961 ± 272 857 ± 330 -41.8 0.006

Protein oxidation (kcal/day)¶ 285 ± 134 300 ± 120 293 ± 91 8.1 0.30

Carbohydrate balance (kcal/day)† 75 ± 192 63 ± 179 -15 ± 178 -35.8 0.008

Lipid balance (kcal/day)† -303 ± 258 -281 ± 228 -178 ± 286 50.2 0.005

Protein balance (kcal/day)† 169 ± 124 153 ± 107 160 ± 104 -6.8 0.39

of 41 Diabetes

Supplementary Table 7―SIFT scores for the 4 SLC16A11 missense variants

Using orthologues

in the protein alignment

Using homologues

in the protein alignment

Variant Amino acid change Prediction SIFT score Prediction SIFT score

rs117767867 V113I tolerated 0.69 tolerated 0.72

rs13342692 D127G damaging 0.05 tolerated 0.10

rs75418188 G340S tolerated 0.42 tolerated 0.24

rs75493593 P443T tolerated 0.59 tolerated 0.68

of 41Diabetes

Supplementary Table 8―Potential eQTLs identified in analyses of whole blood genome-wide expression data in 1,416 American Indians

from the FIND study

eQTL variant information eQTL gene information

Variant Gene Chr Pos* Probe Gene Chr Pos

* Distance

† Beta

‡ P

§ FDR

#

rs10906115 CDC123 10 12.31 ILMN_1795561 CAMK1D 10 12.87 0.55 0.25 9.2 × 10-9

0.0010

rs10906115 CDC123 10 12.31 ILMN_1705477 CAMK1D 10 12.87 0.55 0.21 8.1 × 10-8

0.0045

rs10906115 CDC123 10 12.31 ILMN_1906187 LOC283070 10 12.88 0.56 0.21 1.8 × 10-6

0.0541

rs2237892 KCNQ1 11 2.84 ILMN_1718565 CDKN1C 11 2.91 0.07 0.20 2.0 × 10-6

0.0541

rs10906115 CDC123 10 12.31 ILMN_1751561 CAMK1D 10 12.87 0.56 0.18 2.4 × 10-6

0.0541

rs75493593 SLC16A11 17 6.95 ILMN_1715698 RNASEK 17 6.92 0.03 -0.17 1.6 × 10-5

0.2797

rs10906115 CDC123 10 12.31 ILMN_1721977 ARD1A X 152.20 ― 0.16 1.8 × 10-5

0.2797

rs8050136 FTO 16 53.82 ILMN_1699772 RRAGD 6 90.08 ― 0.22 3.5 × 10-5

0.4872

rs10906115 CDC123 10 12.31 ILMN_1711314 NUDT5 10 12.21 0.10 0.19 4.3 × 10-5

0.5275

rs10906115 CDC123 10 12.31 ILMN_1784977 DOHH 19 3.49 ― 0.15 1.0 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_1665859 RAB27A 15 55.50 ― -0.15 1.4 × 10-4

0.9999

rs231362 KCNQ1 11 2.69 ILMN_2123743 FCER1G 1 161.19 ― -0.18 1.5 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1725666 GTF2H3 12 124.15 ― -0.20 1.8 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_2209163 CHD6 20 40.03 ― -0.20 2.1 × 10-4

0.9999

rs231362 KCNQ1 11 2.69 ILMN_1743621 C17orf69 17 43.72 ― -0.17 2.1 × 10-4

0.9999

rs3923113 GRB14 2 165.50 ILMN_1741300 ZNF407 18 72.78 ― 0.30 2.5 × 10-4

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1738450 GAGE1 X 49.36 ― 0.15 2.7 × 10-4

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1681008 CGRRF1 14 55.00 ― -0.14 3.0 × 10-4

0.9999

rs4607517 GCK 7 44.24 ILMN_1689461 SP7 12 53.72 ― -0.16 3.4 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1801246 IFITM1 11 0.32 ― 0.19 3.5 × 10-4

0.9999

rs75493593 SLC16A11 17 6.95 ILMN_1747244 CCNG2 4 78.09 ― 0.14 4.2 × 10-4

0.9999

rs4607517 GCK 7 44.24 ILMN_1677165 SF3B1 2 198.28 ― -0.16 4.4 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1722981 TLR5 1 223.28 ― 0.19 4.4 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1715068 AQP9 15 58.48 ― 0.19 4.7 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_1694888 TAF2 8 120.74 ― -0.13 5.0 × 10-4

0.9999

rs231362 KCNQ1 11 2.69 ILMN_1746386 HERV-HHHLA1 8 133.08 ― 0.15 5.5 × 10-4

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_2383516 WDR7 18 54.70 ― -0.13 6.0 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1810725 FAM129A 1 184.76 ― 0.19 6.3 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_1803698 STX16 20 57.23 ― -0.13 7.1 × 10-4

0.9999

rs4607517 GCK 7 44.24 ILMN_2287276 FAM177A1 14 35.52 ― 0.14 7.5 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_2336595 ACSS2 20 33.52 ― 0.20 7.7 × 10-4

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_2352724 PIGN 18 59.71 ― -0.13 7.8 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_1693310 ITFG1 16 47.20 ― -0.13 7.8 × 10-4

0.9999

rs75493593 SLC16A11 17 6.95 ILMN_1654497 ATAD1 10 89.51 ― 0.13 7.8 × 10-4

0.9999

rs4607517 GCK 7 44.24 ILMN_1810652 LMBRD2 5 36.11 ― -0.14 8.0 × 10-4

0.9999

rs4607517 GCK 7 44.24 ILMN_1703622 PPIB 15 64.45 ― -0.14 8.0 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_2398865 VPS13C 15 62.17 ― 0.13 8.1 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1703471 ATF6 1 161.93 ― 0.19 8.5 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1673352 IFITM2 11 0.31 ― 0.18 8.6 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_2278653 ZNF493 19 21.59 ― 0.13 8.7 × 10-4

0.9999

rs75493593 SLC16A11 17 6.95 ILMN_1756920 ADAM15 1 155.04 ― -0.13 8.7 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1803564 YIPF1 1 54.32 ― 0.18 8.8 × 10-4

0.9999

rs8050136 FTO 16 53.82 ILMN_1675756 KCNJ15 21 39.67 ― 0.19 8.8 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_1905413 BU674282 5 81.88 ― 0.13 9.2 × 10-4

0.9999

of 41 Diabetes

Of the 42 established type 2 diabetes-susceptibility variants, 7 variants, rs10906115 (CDC123), rs2237892 (KCNQ1), rs231362 (KCNQ1),

rs75493593 (SLC16A11), rs8050136 (FTO), rs4607517 (GCK), and rs3923113 (GRB14) were nominally significant with type 2 diabetes in the

FIND subjects (N = 1,416, data not shown). These 7 variants were selected for the eQTL analyses. The analyses were conducted using whole

blood genome-wide expression data from the FIND subjects. Associations between each of the 7 variants and gene expression was examined for

transcripts (N = 15,854) that were expressed above background as assessed by a χ2 tail test, that mapped to unique locations, and that were free

of variants located in probe set binding sites according to ReMOAT annotation. *Positions (Mb) are from build 37 and for the transcripts are

taken as the mid-point. †The distance between the variant and transcript if located on the same chromosome.

‡Beta is the effect size and is shown

in SD units per diabetes risk allele. §P-values were calculated with genomic control; results were adjusted for age, sex, and America Indian

heritage. #Benjamini-Hochberg false discovery rate calculated across all 110,978 SNP-transcript pairs.

rs10906115 CDC123 10 12.31 ILMN_1715908 NCAPH2 22 50.96 ― 0.13 9.4 × 10-4

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1736894 HNRNPA1L2 13 53.21 ― 0.13 9.5 × 10-4

0.9999

rs4607517 GCK 7 44.24 ILMN_1771118 SLCO1B1 12 21.39 ― -0.15 9.5 × 10-4

0.9999

rs10906115 CDC123 10 12.31 ILMN_1693227 ZC3H7A 16 11.84 ― -0.13 1.0 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1778177 ZNF207 17 30.70 ― -0.13 1.1 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1803984 MAK 6 10.76 ― 0.19 1.2 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1812474 TFG 3 100.47 ― -0.13 1.2 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1667970 SUMO2 X 114.95 ― 0.14 1.2 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1694182 C21orf84 21 44.88 ― -0.15 1.3 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1768867 AP3B1 5 77.30 ― -0.13 1.3 × 10-3

0.9999

rs231362 KCNQ1 11 2.69 ILMN_1765525 WBSCR16 7 74.49 ― 0.17 1.3 × 10-3

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1654792 C8orf15 8 10.98 ― 0.13 1.3 × 10-3

0.9999

rs75493593 SLC16A11 17 6.95 ILMN_1694890 HAUS7 X 152.72 ― -0.13 1.4 × 10-3

0.9999

rs231362 KCNQ1 11 2.69 ILMN_1779515 CSAD 12 53.55 ― 0.15 1.4 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1669440 ZNF326 1 90.47 ― -0.15 1.5 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1809191 RAP1A 1 112.26 ― -0.14 1.5 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1672728 PRO0461 16 2.76 ― -0.13 1.5 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1810559 RHOQ 2 46.81 ― 0.19 1.5 × 10-3

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1859109 LOC100128002 12 131.78 ― 0.13 1.7 × 10-3

0.9999

rs75493593 SLC16A11 17 6.95 ILMN_2342638 ASGR2 17 7.01 0.06 0.15 1.7 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1691106 CDV3 3 133.29 ― -0.18 1.7 × 10-3

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_2073010 TMEM203 9 140.10 ― -0.13 1.8 × 10-3

0.9999

rs75493593 SLC16A11 17 6.95 ILMN_1753773 ANAPC11 17 79.86 72.91 -0.13 2.0 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1701911 C5orf20 5 134.78 ― -0.14 2.1 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1651433 DCK 4 71.90 ― -0.13 2.2 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1683273 SNAPC5 15 66.79 ― 0.13 2.4 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1771120 TMEM45B 11 129.73 ― 0.18 2.6 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1771815 C20orf43 20 55.09 ― 0.18 2.7 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1684771 PGRMC1 X 118.38 ― -0.12 2.7 × 10-3

0.9999

rs231362 KCNQ1 11 2.69 ILMN_1660436 HSPA1A 6 31.80 ― -0.15 2.7 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1845157 TTC15 2 3.49 ― 0.14 2.8 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1694194 SCAF1 19 50.16 ― -0.14 3.0 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1763359 PEG10 7 94.30 50.06 -0.15 3.1 × 10-3

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1735712 KRT1 12 53.07 ― -0.12 3.2 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_2223350 C13orf1 13 50.49 ― -0.14 3.4 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1663422 RGL4 22 24.04 ― 0.18 3.5 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_2235283 MAPK1 22 22.12 ― 0.18 3.7 × 10-3

0.9999

rs8050136 FTO 16 53.82 ILMN_1653292 PFKFB4 3 48.56 ― 0.18 3.8 × 10-3

0.9999

rs2237892 KCNQ1 11 2.84 ILMN_1688630 RECK 9 36.12 ― -0.13 4.1 × 10-3

0.9999

rs4607517 GCK 7 44.24 ILMN_1834802 BM980639 7 77.63 33.39 -0.14 4.3 × 10-3

0.9999

rs10906115 CDC123 10 12.31 ILMN_1798164 PHF3 6 64.42 ― -0.12 5.0 × 10-3

0.9999

of 41Diabetes

Supplementary Table 9―Whole blood transcripts with nominally significant (P <0.05) mediation of variant-diabetes associations

Association analysis

eQTL variant eQTL gene Variant–

Diabetes

Variant–

Transcript levels

Transcript–

Diabetes Mediation

Variant Chr Gene Chr OR* P

|| Beta

† P

|| OR

‡ P

¶ PM

§ P

#

rs75493593 17 RNASEK 17 1.36 0.003 -0.17 1.6 × 10-5

0.82 0.006 9.9 0.01

rs75493593 17 ADAM15 1 1.36 0.003 -0.13 8.7 × 10-4

0.82 0.007 7.6 0.02

rs8060136 16 GTF2H3 12 1.52 0.003 -0.20 1.7 × 10-4

0.86 0.04 8.1 0.04

rs10906115 10 PRO0461 16 1.25 0.03 -0.13 1.5 × 10-3

0.86 0.04 9.0 0.05

For variants associated with both type 2 diabetes (P <0.05) and whole blood transcript levels (P <0.005), a formal mediation analysis

was used to quantify the extent to which the pattern of associations was consistent with an effect on transcript that mediates the

variant-diabetes association. The analyses were conducted using whole blood genome-wide expression data from the FIND

participants (N = 1,416). *Odds ratio for association between variant and type 2 diabetes expressed per copy of the diabetes risk allele.

†Effect of variant on transcript expressed in SD units per copy of diabetes risk allele.

‡Odds ratio for diabetes per SD increase in

transcript levels. §PM (percentage mediation) is the percentage of the variant effect on diabetes that is mediated by the variant effect

on transcript levels. ||P-value adjusted for age, sex, and American Indian heritage.

¶P-value adjusted for age, sex, American Indian

heritage, and genotype. #P-value for the mediation effect (path) was calculated by the Sobel method adjusted for age, sex, and

American Indian heritage. This P-value is one-sided. All P-values were calculated with genomic control.

of 41 Diabetes

analysis of slc16a11 variants in 12,811 american indians · 2015-10-16 · 1 1 analysis of...

Documents