Association of Variants in the Sterol Regulatory Element-Binding Factor 1 (SREBF1) Gene With Type 2 Diabetes, Glycemia, and Insulin Resistance
A Study of 15,734 Danish Subjects
- Niels Grarup1,
- Kirstine L. Stender-Petersen1,
- Ehm A. Andersson1,
- Torben Jørgensen2,
- Knut Borch-Johnsen1,2,3,
- Annelli Sandbæk4,
- Torsten Lauritzen4,
- Ole Schmitz5,6,
- Torben Hansen1 and
- Oluf Pedersen1,3
- 1Steno Diabetes Center, Copenhagen, Denmark
- 2Research Centre for Prevention and Health, Glostrup University Hospital, Glostrup, Denmark
- 3Faculty of Health Sciences, University of Aarhus, Aarhus, Denmark
- 4Department of General Practice, University of Aarhus, Aarhus, Denmark
- 5Department of Endocrinology and Diabetes, Aarhus University Hospital, Aarhus, Denmark
- 6Department of Clinical Pharmacology, University of Aarhus, Aarhus, Denmark
- Address correspondence and reprint requests to Niels Grarup, Steno Diabetes Center, Niels Steensens Vej 1, NLC2.14, 2820 Gentofte Denmark. E-mail:
OBJECTIVE—We evaluated the association of variants in the sterol regulatory element-binding factor 1 gene (SREBF1) with type 2 diabetes. Due to the previous inconclusive quantitative trait associations, we also did studies of intermediate quantitative phenotypes.
RESEARCH DESIGN AND METHODS—We genotyped four variants in SREBF1 in the population-based Inter99 cohort (n = 6,070), the Danish ADDITION study (n = 8,662), and in additional type 2 diabetic patients (n = 1,002). The case-control studies involved 2,980 type 2 diabetic patients and 4,522 glucose-tolerant subjects.
RESULTS—The minor alleles of rs2297508, rs11868035, and rs1889018 (linkage disequilibrium R2 = 0.6–0.8) associated with a modestly increased risk of type 2 diabetes (rs2297508: OR 1.17 [95% CI 1.05–1.30], P = 0.003), which was confirmed in meta-analyses of all published studies (rs2297508 G-allele: 1.08 [1.03–1.14] per allele, P = 0.001). The diabetes-associated alleles also associated strongly with a higher plasma glucose at 30 and 120 min and serum insulin at 120 min during an oral glucose tolerance test (all P < 0.006) and the minor allele of rs1889018 with a surrogate measure of insulin sensitivity (P = 0.03). Furthermore, the diabetes-associated alleles associated with a modestly increased A1C level in the population-based Inter99 of middle-aged subjects and in the ADDITION study of high-risk individuals (P = 0.006 and P = 0.008, respectively).
CONCLUSIONS—We associate sequence variation in SREBF1 with a modestly increased predisposition to type 2 diabetes. In the general population, the diabetes-associated alleles are discreetly associated with hyperglycemia presumably due to decreased insulin sensitivity. Because sterol regulatory element–binding protein-1c is a mediator of insulin action, the findings are consistent with the presence of a yet undefined subtle loss-of-function SREBF1 variant.
- GWAS, genome-wide association studies
- LD, linkage disequlibrium
- OGTT, oral glucose tolerance test
- SREBP, sterol regulatory element–binding protein
The sterol regulatory element-binding factor (SREBF1) gene encodes the transcription factors sterol regulatory element–binding protein (SREBP)-1a and -1c by differential transcription start sites (1). SREBP-1a and -1c, and the third family member SREBP-2, are implicated in regulation of cholesterol and fatty acid synthesis (rev. in 2). SREBP-1c is, in humans, expressed in most tissues including liver, adipose tissue, and skeletal muscle, while SREBP-1a is expressed mainly in the spleen and intestine (3). SREBP-1c is a mediator of insulin action in liver, adipose tissue, and skeletal muscle (4) with the ability to activate lipogenic genes, glucokinase, and hexokinase in these metabolic tissues (5,6). SREBP-1c thereby induces both glucose utilization and lipid metabolism, suggesting that a low level of SREBP-1c is a contributing factor in the pathogenesis of insulin resistance and type 2 diabetes. In the adipose tissue and skeletal muscle of type 2 diabetic patients, expression of SREBP-1c is indeed decreased (7). This is consistent with studies of SREBP-1c–specific knock-out mice showing a mild hyperglycemic phenotype (8).
In contrast, since SREBP-1c promotes fatty acid synthesis and lipogenesis, SREBP-1c overexpression could moreover be a factor responsible for insulin resistance through overaccumulation of lipids also leading to lipotoxicity (9). Interestingly, increased expression of SREBP-1c in the liver has been observed in animal models of obesity and type 2 diabetes (10).
Genome-linkage scans have linked the 17p11 region comprising SREBF1 to type 2 diabetes (11), and case-control studies including 1,000–2,000 participants have consistently associated SREBF1 with type 2 diabetes, although with different variants in linkage disequilibrium (12–15). Recent genome-wide association studies (GWAS) did not, however, report SREBF1 as a type 2 diabetes locus (16–20). Furthermore, associations with obesity (12), circulating total and LDL cholesterol (13), HDL cholesterol (15), and plasma glucose levels (14) have been reported, yet none of these associations has been conclusively replicated (12,13,15).
In the present study we evaluated the association between SREBF1 variants and type 2 diabetes. Due to the prior inconsistent quantitative trait associations, we additionally aimed to establish a quantitative metabolic phenotype in statistically well-powered cohorts of middle-aged Danes. Given the biological evidence, we primarily hypothesized that SREBF1 variants influenced peripheral insulin action.
RESEARCH DESIGN AND METHODS
Further details and phenotypic characteristics are given in the online appendix (available at http://dx.doi.org/10.2337/db07-1534). Participants from the population-based Inter99 cohort (clinical trial reg. no. NCT00289237, clinicaltrials.gov) (21) involving 5,970 middle-aged subjects who were characterized by an oral glucose tolerance test (OGTT) as having normal glucose tolerance (n = 4,522), impaired fasting glycemia (n = 503), impaired glucose tolerance (n = 693), or screen-detected and treatment-naïve type 2 diabetes (n = 252) were investigated for associations between genotype and quantitative metabolic traits. Patients with treated type 2 diabetes (n = 100) were not included.
The case-control studies included all type 2 diabetic case subjects and all glucose-tolerant control subjects from the Inter99 cohort (ncases = 352, ncontrols = 4,522) and the Danish ADDITION study (ncases = 1,626), as well as samples recruited from the outpatient clinic at Steno Diabetes Center (ncases = 1,002). All control subjects had normal fasting glycemia and were glucose tolerant following an OGTT. Diabetes was diagnosed according to the World Health Organization 1999 criteria (23).
Informed written consent was obtained from all participants. The studies were conducted in accordance with the Declaration of Helsinki II and were approved by the local ethics committees of Copenhagen and Aarhus.
Biochemical and anthropometric measures.
Biochemical and anthropometric measures are described in the online appendix.
Selection of gene variants.
Tag single nucleotide polymorphisms were selected based on the HapMap CEU population (www.hapmap.org, release 21a) using the Carlson greedy algorithm (24) capturing all variants in SREBF1 including 1,000 bp up- and downstream with a minor allele frequency >4% with R2 = 0.8. In the analysis, we force-included three variants in SREBF1 (rs11868035, rs2297508, and rs1889018) that have been reported to associate with metabolic traits (12–15).
Genotyping of four selected variants in SREBF1 (rs4925118, rs1889018, rs2297508, and rs11868035) was performed by TaqMan allelic discrimination (KBiosciences, Hoddesdon, U.K.). All genotyping success rates were above 97% with a mismatch rate below 0.25% in 968 duplicate samples. The distributions of genotypes for all variants were in Hardy-Weinberg equilibrium (all P > 0.05) (online appendix Table 2).
Linkage disequilibrium (LD) between markers was evaluated using Haploview version 4.0 (http://www.broad.mit.edu/mpg/haploview/). All other analyses were performed using RGui, version 2.2.4 (http://www.r-project.org). In the association studies of type 2 diabetes, logistic regression was used to examine differences in genotypes assuming an additive model with adjustment for sex, age, and BMI. Meta-analyses of the present and previously published studies and tests of homogeneity between studies were performed using the Mantel-Haenszel method applying a generalized linear model. In all meta-analyses, imputed data from the Wellcome Trust Case-Control Consortium (19) were included, and, for the rs1889018 variant, data were further obtained from the Diabetes Genetics Initiative (18) by use of a perfect proxy (R2 = 1) (rs9899634) based on LD in the HapMap database (www.hapmap.org). No perfect proxies for rs2297508 and rs11868035 were available. A general linear model was used for testing quantitative traits in relation to genotype, adjusting for the effect of sex, age, and BMI, when appropriate. Quantitative traits were checked for normality of the residuals and, if appropriate, logarithmically transformed. Haplotype frequencies were estimated using the EM-algorithm and association, and effect sizes of each haplotype were estimated by modeling the haplotype-phenotype interaction (25). Haplotypes with a frequency >1% were included. A P value <0.05 was considered significant.
Initially we analyzed the pattern of LD between the four genotyped variants in SREBF1 and observed relatively high LD between rs1889018, rs2297508, and rs11868035 (R2 = 0.6–0.8) (Fig. 1). The four SREBF1 variants were assessed for potential associations with type 2 diabetes in a case-control study involving 2,980 type 2 diabetic patients and 4,522 glucose-tolerant control subjects (Table 1). The rs2297508 G-allele associated with increased susceptibility to type 2 diabetes with an odds ratio (OR) of 1.17 per risk allele (95% CI 1.05–1.30, P = 0.003) when adjusting for the impact of age, sex, and BMI. The minor alleles of rs1889018 and rs11868035 showed similar associations with type 2 diabetes (Table 1). We performed meta-analyses of the unadjusted association with type 2 diabetes including the present study, all previously published studies (12–15), and online data from two GWASs (18,19). The minor alleles of rs2297508, rs1889018, and rs11868035 all associated with a modest increase in type 2 diabetes susceptibility (OR 1.06–1.08, P < 0.01) (Fig. 2). Tests of between-study homogeneity showed no heterogeneity for rs1889018 and rs11868035 variants but some heterogeneity for rs2297508 (P = 0.02).
To evaluate the metabolic phenotype of the type 2 diabetes susceptibility allele carriers, the four SREBF1 variants were investigated in the population-based Inter99 sample involving 5,970 middle-aged treatment-naïve individuals (Table 2). At the population level, the G-allele of rs2297508 associated with a slight increase in fasting plasma glucose level (P = 0.02) and with a slightly increased plasma glucose at 30 and 120 min during an OGTT (P = 0.006 and P = 0.001, respectively). In addition, we observed a strong association with a higher 120-min serum insulin during an OGTT (P = 0.0002) and with a slightly increased A1C level (P = 0.006) (Table 2). The minor alleles of the rs1889018 and rs11868035 variants showed similar associations with glycemia (online appendix Table 2).
A surrogate measure of insulin resistance was reported as the BIGTT-insulin sensitivity index (BIGTT-Si) (26) and as the homeostasis model assessment index of insulin resistance. We observed a decreased insulin sensitivity assessed by BIGTT-Si in carriers of the minor rs1889018 C-allele (P = 0.03) (online appendix Table 2), and this association was strengthened after adjustment for the level of insulin release (P = 0.006).
Further analyses of metabolic phenotypes were performed in the ADDITION screening cohort consisting of 8,662 subjects at high risk of type 2 diabetes. We replicated the association of the G-allele of rs2297508 with a 0.45% increase in A1C per allele (P = 0.008) (Table 2). In the ADDITION study we also found nominal associations with decreased BMI in rs2297508 minor G-allele carriers (Table 2) and increased fasting serum cholesterol level for carriers of the minor allele of rs11868035 (online appendix Table 3). The rs4925118 variant did not associate with examined metabolic traits (Table 1 and online appendix Tables 2–3). Haplotype association analyses did not further add to single variant analyses (online appendix Table 5).
In the present study we report associations of sequence variation in SREBF1 with a modest increase in type 2 diabetes risk. In the well-characterized population-based Inter99 sample of middle-aged treatment-naïve individuals, we report strong associations for the diabetes risk-allele carriers with higher plasma glucose and serum insulin after an oral glucose load as well as with a slight increase in A1C. The latter was supported by association with A1C in the Danish ADDITION study. Also, variants in SREBF1 associate with BIGTT-Si, which is a well-documented surrogate measure of insulin sensitivity (26).
Because the transcription factor SREBP-1c is a mediator of insulin action in skeletal muscle, adipose tissue, and liver, it is conceivable that a polymorphism conferring a subtle loss of function may affect the expression of SREBP-1c and could contribute to the insulin resistance phenotype. The observed effects are evident in the postprandial state. Enzymes involved in glucose metabolism are highly regulated by SREBP-1c, and if downregulated due to a mild SREBP-1c dysfunction, glucose metabolism would likely be impaired in the peripheral tissues, potentially leading to postprandial hyperglycemia and insulin resistance.
Prior association studies have shown an impact of the rs2297508 variant (12,15), the rs11868035 variant (13,14), and rs2236513, rs6502618, rs1889018 variants (14) in SREBF1 on risk of type 2 diabetes. All of these variants are in substantial LD (HapMap: R2 = 0.67–0.95). In the present report we replicate associations of rs2297508, rs11868035, and rs1889018 with type 2 diabetes. As none of the recent GWASs (16–20) reported SREBF1 as a type 2 diabetes locus, we engaged in meta-analyses of the present data and previously published studies (12–15) as well as all online available data from GWASs (18,19). In combined analyses we showed discreet increases in type 2 diabetes risk for all three variants. However, it should be noted that meta-analysis, which in principle might be expected to provide conclusive answers, may be compromised by heterogeneity of ethnicity and outcome phenotypes in addition to publication and ascertainment bias. Moreover, the present meta-analyses include imputed data and data based on perfect LD with another marker, yet data from three other GWASs (16,17,20) were not available. In any case, in the meta-analyses we were not able to adjust for the effect of confounding factors such as age, sex, and BMI, which further weakens the analysis; in fact, in the present report the association with type 2 diabetes was abolished when not adjusting for the effect of age and sex. Therefore, cautious interpretations of these meta-analyses are crucial.
Previous reports regarding associations between SREBF1 variants and quantitative metabolism have been inconclusive. One study has indicated an association of the rs11868035 variant with increased fasting total and LDL cholesterol levels (13). In contrast, two studies did not show any association of the rs2297508 variant with fasting cholesterol levels (12,15). We found nominal associations of the rs11868035 variant with increased fasting serum cholesterol levels and the rs2297508 and rs1889018 variants with decreased BMI in the ADDITION study of subjects at high risk for type 2 diabetes; however, these associations were not observed in the population-based Inter99 study. These ambiguities may be due to the diversity of the populations, e.g., accentuation of associations in the ADDITION cohort due to the high-risk selection procedure, yet could more likely be a result of statistical type I errors. Also, associations with obesity previously have been conflicting (12,13,15,18), and the present study does not support previous associations to obesity-related traits. Moreover, a recent study found borderline significant associations of the minor alleles of the rs1889018 and rs2236513 variants with higher plasma glucose level at fasting and 2 h after an oral glucose challenge (14). We substantiate these findings in the current study. Because published studies have not investigated the exact same variants in SREBF1, these somewhat inconsistent reports may be explained by population-specific differences in LD pattern, but could also be caused by discrepancies in the examined populations, e.g., environmental modifying effects, or by statistical type I or II errors.
Thus far the causal variant giving rise to the described associations has not been identified. The present results are not able to elucidate this, as none of the variants are generally more associated than others. None of the investigated variants are obvious functional candidates, yet a causal variant in LD with the associated variants may influence regulation or function of SREBP-1c. However, since HapMap data demonstrate that SREBF1 is located in an extended haploblock spanning almost 300 kb, a putative causal variant may be situated at some distance.
Although we present data supporting an association with type 2 diabetes and glycemia, we recognize that the present results may be falsely positive due to the fact that no correction for multiple testing was applied. Yet, we argue that based on previous reports a SREBF1 effect on risk of type 2 diabetes was the primary hypothesis of this study, and as the association with glycemia, evaluated by A1C, was supported by association in an independent cohort, no correction is strictly needed.
In conclusion, we associate variants in SREBF1 with a slight increase in type 2 diabetes risk. Of novelty, we present data suggesting an association with glycemia in the general population of middle-aged people, possibly due to a decreased SREBP-1c function in individuals carrying the as yet undefined causal variant.
The study was supported by grants from the Lundbeck Foundation Centre of Applied Medical Genomics for Personalized Disease Prediction, Prevention and Care, the Danish Health Research Council, The European Union (HEPADIP Grant LSHM-CT-2005-018734), the Danish Council for Strategic Research (DanORC Grant 2101-06-0005), the Faculty of Health Sciences of Aarhus University, the Danish Clinical Intervention Research Academy, and the Danish Diabetes Association. The Danish ADDITION trial was supported by the National Health Services in the counties of Copenhagen, Aarhus, Ringkøbing, Ribe, and South Jutland in Denmark, the Danish Research Foundation for General Practice, the Danish Centre for Evaluation and Health Technology Assessment, the Diabetes Fund of the National Board of Health, The Danish Medical Research Council, The Aarhus University Research Foundation, and the Novo Nordisk Foundation. Furthermore, the ADDITION trial has been given unrestricted grants from Novo Nordisk, Novo Nordisk Scandinavia, ASTRA Denmark, Pfizer Denmark, GlaxoSmithKline Pharma Denmark, SERVIER Denmark, and HemoCue Denmark.
The authors thank A. Forman, I.-L. Wantzin, and M. Stendal for technical assistance and G. Lademann for secretarial support.
Published ahead of print at http://diabetes.diabetesjournals.org on 11 January 2008. DOI: 10.2337/db07-1534.
N.G. and K.L.S.-P. contributed equally to this work.
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.
- Received October 27, 2007.
- Accepted January 7, 2008.