Common Polymorphisms in the USF1 Gene Are Not Associated With Type 2 Diabetes in French Caucasians

American Diabetes Association 2003 criteria (4) for the classification of subjects as diabetic or normoglycemic were applied. The type 2 diabetes case-control study was carried out with a cohort of 744 unrelated type 2 diabetic subjects (age at diagnosis 47 ± 12 years, BMI 27.0 ± 3.5 kg/m², men/women 55/45%) and 731 unrelated normoglycemic subjects (age at examination 54 ± 11 years, BMI 24.6 ± 3.9 kg/m², men/women 41/59%). The type 2 diabetic case subjects were composed of 372 probands from type 2 diabetic families, recruited by P.F.’s Centre National de la Recherche Scientifique, Institute Pasteur Unit in Lille, and 372 singleton patients with a family history of type 2 diabetes recruited at the Endocrinology-Diabetology Department of the Corbeil-Essonne Hospital. The control subjects included 361 normoglycemic husbands or wives from type 2 diabetic families and 370 normoglycemic subjects from the SUVIMAX (Supplementation en Vitamines et Mineraux Antioxydants) prospective population-based cohort study (5). Subjects were all of French Caucasian ancestry. A number of quantitative clinical phenotypes (including plasma levels of glucose, triglycerides, total cholesterol, and apolipoproteins A1 and B) were assayed in >95% of our normoglycemic cohort. Informed consent was obtained from all subjects, and the study was approved by the local ethics committees.

Pajukanta et al. (3) identified 17 common (minor allele frequency >5%) SNPs in the USF1 gene by sequencing the entire 6.7-kb USF1 genomic sequence together with 2 kb upstream of the 5′ end of the gene in 31 FCHL probands. SNP location, dbSNP rs numbers, and linkage disequilibrium (LD) data are presented in Table 2 and Supplementary Table 3 of that report. SNP pairs stated as being in complete LD (Δ² = 1) were deemed redundant.

Genotyping was performed with the Sequenom MassARRAY system (7). Every SNP genotyped in our case-control cohort was also genotyped in the subset of 55 French “strict-lean” type 2 diabetic families that produced a linkage signal at 1q (8). The case cohort contains 53 samples that are also present in the strict-lean family set. The genotyping concordance rate of these duplicate samples allowed us to estimate the accuracy of our genotyping.

Power calculations were performed with the program of Purcell et al. (9) (available at http://statgen.iop.kcl.ac.uk/gpc/). Assuming a disease prevalence of 0.1, a genotype relative risk of 1.25, and an allele frequency of 0.73 (the frequency of the major allele of rs3737787, previously associated with FCHL [3]), the case-control sample afforded an estimated 83% power at P < 0.05. Comparisons of SNP allele and haplotype frequencies in case and control groups were performed using the χ² test. Quantitative traits were log transformed and corrected for age, sex, and BMI, as appropriate. Testing SNPs for association with quantitative traits was carried out with the ANOVA test, using a codominant model. Pairwise SNP LD was calculated with the GOLD software package (10) from the haplotype counts output by PHASE (11). The Haplotype Trend Regression program (12) was used to test inferred haplotypes for association with quantitative traits.

The USF1 gene consists of 11 exons that span only 6.7 kb of genomic DNA. Pajukanta et al. (3) previously identified 17 common (minor allele frequency >5%) noncoding SNPs in the USF1 gene by sequencing the entire USF1 genomic sequence gene in 31 FCHL probands. From the LD information provided in that report, we identified a set of nine nonredundant SNPs. After removing rs2073657, which failed to successfully genotype using either the Sequenom (7) or Taqman methods (13), the remaining eight SNPs (Table 1) were genotyped in our type 2 diabetic case-control cohort. Since rs2073657 is in complete LD (Δ² = 1) with the typed SNP rs2516837 in the Finnish population (3), it is highly probable that the loss of rs2073657 had very little or no effect on the proportion of common variation captured in the French Caucasian population. The genotyping success rate for the eight USF1 SNPs genotyped was 93%, and the concordance rate of duplicate genotypes was >99%. The genotype distribution was in accordance with Hardy-Weinberg equilibrium for all SNPs (data not shown). The pairwise LD values, quantified by the metrics D′ and Δ², indicate that there is a high level of LD across the USF1 gene in the French population (Fig. 1), corroborating the Finnish LD data (3). The average D′ and Δ² values were 1 and 0.26, respectively. Three SNP pairs, including rs2073658 and rs3737787, previously associated with FCHL (3) exhibited complete LD (Δ² = 1).

The results of evaluating each of the USF1 SNPs for association with type 2 diabetes is presented in Table 2. None of the SNPs, including two SNPs previously associated with FCHL (rs2073658 and rs3737787), showed evidence of association with type 2 diabetes. There were no significant differences in SNP allele or haplotype frequencies between men and women in the case-control cohort (data not shown). In addition, none of the USF1 SNPs were associated with BMI, homeostasis model assessment of insulin resistance, or with plasma levels of glucose, triglycerides, cholesterol, or apolipoproteins A1 or B in the normoglycemic subjects (online appendix Table 1 [available at http://diabetes.diabetesjournals.org]). These findings are in agreement with those of Putt et al. (14), who also reported no associations between USF1 SNPs and anthropometric or plasma lipid phenotypes in a study of >800 European Caucasian men. No associations with quantitative phenotypes were uncovered by stratifying for sex (data not shown). Finally, and perhaps not surprisingly, none of the USF1 SNPs were associated with the evidence for linkage of type 2 diabetes to 1q21-q24 (data not shown).

The USF1 haplotype frequency distribution of the type 2 diabetic case and control subjects is shown in Table 3. A total of four common haplotypes were identified, accounting for >99% of chromosomes in both the case and control subjects. It is clear that there was no significant difference in the USF1 haplotype distribution of the case and control subjects. In addition, none of the four common USF1 haplotypes were associated with any of the quantitative phenotypes in our normoglycemic subjects (data not shown).

In conclusion, we report here that we were unable to find any evidence to support the hypothesis that genetic variation in the USF1 gene contributes to type 2 diabetes susceptibility in the French Caucasian population. The data presented do not rule out the possibility that distant USF1 regulatory variants lying outside the USF1 gene region surveyed by Pajukanta et al. (3) could make a contribution to type 2 diabetes susceptibility. The case-control sample size afforded >80% power to detect an association with an allele of a frequency equivalent to that of the two SNPs that were previously associated with FCHL (3), assuming a genotype relative risk of 1.25. Nevertheless, a more highly powered study involving thousands of subjects may yet uncover a role for USF1 variants in type 2 diabetes susceptibility.

This work is supported by a Wellcome Trust University Award to F.G. (GR065414MF) and grants from Diabetes U.K. (BDA: RD01/0002308), the Association Française des Diabétologues de Langue Française (ALFEDIAM), the European Union–funded GIFT Grant, the Direction de la Recherche Clinique/Assistance Publique-Hopitaux de Paris, and the Program Hospitalier de Recherche Clinique (AOM 96088).

We thank Cliona Boyle and Amina Bibi for technical support and two anonymous referees for critically appraising draft versions of the manuscript.