Genetic Variation and Association Analyses of the Nuclear Respiratory Factor 1 (nRF1) Gene in Chinese Patients With Type 2 Diabetes

A total of 1,027 unrelated subjects, including type 2 diabetic patients (n = 596) and nondiabetic control subjects (n = 431), were enrolled in the present study. All subjects were Han Chinese and randomly selected from four Top Grade hospitals in Beijing city and Hebei province, China. Nondiabetic control subjects were >45 years old, had normal fasting plasma glucose levels (5.0 ± 0.3 mmol/l), and had no relatives with type 2 diabetes. All type 2 diabetic patients were diagnosed according to the World Health Organization criteria (WHO 1998). Most of the patients had medication with oral hypoglycemic agents (OHAs; 30.9%), insulin alone (53.6%), or insulin in combination with OHA (13.3%), whereas a few patients (2.2%) received only diet therapy. Genomic DNA from the peripheral blood was extracted by using a salting-out protocol. Clinical characteristics of nondiabetic control subjects and type 2 diabetic patients are summarized in Table 1. The study was performed after obtaining the informed consent from all subjects and was approved by the local ethics committee.

The sequences of the putative promoter region, exon-intron boundaries, and all exons, including the 5′- and 3′-untranslated regions (UTRs) (∼1 kb) of NRF1 gene were PCR amplified in 20 type 2 diabetic patients and 20 nondiabetic control subjects. Human NRF1 mRNA NM_005011.2 was used as the contig sequence for defining the variants identified. PCR products were then sequenced using an optimized direct sequencing analysis protocol (Dye Terminator Cycle Sequencing Ready Reaction kit; ABI model 377 genetic analyzer; Perkin-Elmer, Foster City, CA). Primer specifics and optimized PCR conditions are available upon request.

Using tagging with an r² cutoff of 0.8 and/or minor allele frequency (MAF) of >5% and our own sequencing data, 9 of 16 SNPs (6 novel and 10 known) in the NRF1 gene were selected for genotyping experiments with PCR–restriction fragment–length polymorphism (PCR-RFLP) technique. Negative controls (water blanks) were included in each plate. For genotyping quality control, the case and control subjects were distributed randomly across the plates. Successful genotype calls were >97%, and 20% of samples were randomly genotyped twice for duplication accuracy, which was calculated to be 99%. Description of PCR-RFLP genotyping experiments is available in supplemental Table 1, which is detailed in the online appendix (available at http://dx.doi.org/10.2337/db07-0008).

Genotype distributions for all studied SNPs in nondiabetic control subjects were tested for Hardy-Weinberg equilibrium (HWE). The minor allele was too rare for most polymorphisms to give enough power. Therefore, additive and dominant models were used to detect the allelic association. In the additive model, odds ratio (OR) value was expressed per difference in number of minor alleles. In the dominant model, ORs were shown as heterozygotes and homozygotes with minor alleles compared with homozygotes with major alleles. ORs and 95% CIs corresponding to P values without adjustment for age, sex, and BMI were obtained using χ² tests. Among nondiabetic control subjects, the association between genotypes and quantitative traits was tested using Kruskal-Wallis analysis for traits with non-normal distribution or, alternatively, ANOVA for normally distributed traits. Retrospective statistical powers were calculated using software PowerSampleSize (PS version 2.1.31, available at http://biostat.mc.vanderbilt.edu/twiki/bin/view/Main/PowerSampleSize).

All statistical analyses were performed using SPSS statistical package (version 10.0) and/or SAS statistical package (version 9.1.2). We obtained estimates of pairwise linkage disequilibrium values using DnaSP 4.0 software package (Universitat de Barcelona, Barcelona, Spain) and/or Haploview software (version 3.32). Haplotype blocks were determined according to the CI method using the criteria of Gabriel et al. (12). Haplotype estimation was carried out using PHASE program (version 2.1). Haplotype frequency was determined by means of the algorithms implemented in the PHASE program (13). A P value of <0.05 was considered to be statistically significant. Pc values were calculated for multiple testing using Bonferroni's inequality method (14) and defined as P values (single test) × number of tests.

We performed systematic variation screening of the NRF1 gene in type 2 diabetic patients and nondiabetic control subjects (n = 20/group) and identified 16 SNPs (Fig. 1). Three novel SNPs were found, one (−46127T>C) in the 5′ region and two (+97884G>A and +98560A>G) in the 3′-UTR of the gene. A silent substitution encoding 47Ser (+141G>T) in exon 2 was recorded in dbSNP (rs1882094). Additionally, 12 SNPs were found to be intronic: three novel (+51889G>A, +52989A>C, and +97617T>G) and nine recorded, including rs753947 (−44339G>A), rs6969098 (−24833A>G), rs6951103 (+14255G>T), rs11769146 (+51910G>A), rs4731617 (+52196T>C), rs7782640 (+54099A>G), rs1962038 (+54253C>G), rs2293629 (+60107T>C), and rs2896423 (+60387C>T).

By the pairwise linkage disequilibrium analyses, two linkage disequilibrium blocks in the NRF1 gene (r² >0.8) were found that were constructed with SNPs rs753947-rs6969098 and rs11769146-rs1962038-rs2293629, respectively. Nine SNPs were selected for genotyping experiments: −46127T>C, −rs6969098, rs1882094, rs11769146, rs4731617, +52989A>C, rs2896423, +97884G>A, and +98560A>G. These nine genotyped SNPs were named P1–P9 for convenience in analyses and description. ID numbers of SNPs recorded in dbSNP are represented in Fig. 1 and Table 2.

Genotype distributions of all studied SNPs P1–P9 in nondiabetic control subjects were in HWE (P > 0.05). We assessed the association of each SNP with type 2 diabetes using additive and dominant models (Table 2). Two novel SNPs, P1 (−46127T>C) and P9 (+98560A>G), were significantly associated with type 2 diabetes (P = 0.001, OR 0.620 [95% CI 0.462–0.832] and P = 0.002, 3.199 [1.466–6.981]; Pc = 0.018 and 0.036 with correction for multiple comparisons in dominant models, respectively). The associations of SNPs, including P2 (rs6969098), P3 (rs1882094), P5 (rs4731617), and P7 (rs2896423), with type 2 diabetes were also found to be significant when the P values were not corrected for multiple comparisons. Of these four SNPs, P2 and P7 were linked to a decreased susceptibility for type 2 diabetes using both additive (P = 0.008, 0.793 [0.666–0.945]; P = 0.033, 0.671 [0.460–0.978]) and dominant (P = 0.011, 0.682 [0.507–0.916]; P = 0.041, 0.667 [0.451–0.985]) models. However, P3 was associated with an increased susceptibility risk for type 2 diabetes when applying a dominant model (P = 0.016, 1.355 [1.057–1.738]), whereas P5 (rs4731617) showed a decreased risk of type 2 diabetes when testing with an additive model (P = 0.027, 0.778 [0.626–0.968]). No significant association of P4 (rs11769146), P6 (novel +52989A>C), and P8 (novel +97884G>A) with type 2 diabetes was found.

We further examined the degree of linkage disequilibrium of SNPs P1–P9 and performed a haplotype association analysis. Table 3 displays the estimates of pairwise linkage disequilibrium (r² and D′ value) of the studied SNPs in the NRF1 gene. P9 was excluded from further analyses of haplotypes because of its low MAF (∼0.02). Four common haplotypes with >5% frequency accounted for 71.1% of the observed haplotypes in type 2 diabetic patients (Table 4). Interestingly, in these four common haplotypes, the alleles from SNPs P1-P4-P5-P6-P7 were the same. Therefore, these four common haplotypes were simply defined as A-G-G, G-G-G, G-T-G, and A-G-A, and the alleles were from SNPs P2, P3, and P8. Furthermore, the common haplotype H2 was found to be significantly associated with a decreased risk of type 2 diabetes (P = 0.004; Pc = 0.016, OR 0.706 [95% CI 0.557–0.895]).

To predict the association between haplotypes and phenotypes, we conducted the comparison analyses in nondiabetic control subjects, because most of type 2 diabetic patients included in the present study had medical treatments, which may affect the real parameters. We found that the carriers with GG genotype of SNP P3 had lower fasting plasma glucose levels than those with other genotypes (4.95 vs. 5.11, P = 0.0002, Pc = 0.001). Moreover, the subjects carrying haplotype H2 had lower fasting triglyceride levels when compared with the subjects with other haplotypes (1.43 vs. 1.61 mmol/l, P = 0.002, Pc = 0.010; Table 5). There were no strong associations between the diabetes-related phenotypes and other genotypes or haplotypes when adjusting for multiple comparisons (data not shown).

In the present study, we identified 16 SNPs, including 6 novel variants, in the NRF1 gene and found that 2 of the novel SNPs, P1 (−46127T>C) and P9 (+98560A>G), were significantly associated with type 2 diabetes in a Han Chinese population. Additionally, we demonstrated that P3 (rs1882094), a synonymous polymorphism located at codon 47 of exon 2, was related to fasting plasma glucose levels in nondiabetic control subjects. In this SNP, the carriers with genotype GG had lower fasting plasma glucose levels compared with subjects carrying genotypes GT and TT. The common haplotype, H2, mainly composed of SNPs P2, P3, and P8, was associated with lower levels of fasting triglyceride.

Interestingly, SNPs P1 and P2 are located at the putative promoter region of the NRF1 gene, whereas P8 and P9 are at the 3′-UTR. The noncoding portion of genomes may have a crucial role in gene expression (15), whereas UTRs are involved in many posttranscriptional regulatory pathways related to mRNA localization, stability, and translation efficiency. Initiation of protein synthesis could be influenced by sequence elements in both 5′- and 3′-UTRs (16). Diabetic dyslipidemia is a common clinical feature of type 2 diabetes, and patients have high levels of fasting triglycerides and hypertriglyceridemia (17,18). In the present study, haplotype H2 revealed the decreased association with the fasting triglyceride levels in nondiabetic subjects. It suggests that NRF1 genetic polymorphisms may be involved in regulating dyslipidemia and may subsequently contribute to the development of type 2 diabetes.

Many factors may contribute to variable results of genetic association study, major ones being sample size and ethnic stratification. In the present study, the sample size was adequately powerful (>80%) to detect differences between case and control subjects because type I error probability for a two sided-test (α) was defined as 0.05 and the OR was estimated as 1.5. However, the sample size of this study may not have been sufficiently powerful to detect modest effects of the NRF1 genetic polymorphisms. Further studies should replicate this study using a larger cohort of Han Chinese or Caucasian subjects.

Recently, 13 SNPs were identified in the NRF1 gene and 6 of these were studied for genetic association in a Korean cohort (11). In the present study, we identified 16 SNPs of the NRF1 gene in a Han Chinese population. Of the 16 identified SNPs, g.+141G>T, g.+97884G>A, and g.+98560A>G were represented in both the Han Chinese and Korean populations, whereas the remaining 13 SNPs identified in the present study were not reported in the Korean study (11). The differences in the NRF1 genetic polymorphisms between these two populations may be due to population specificities and/or the different range of exon-intron boundaries and different number of samples for direct-sequencing. However, SNP rs1882094 (+141G>T, P3 in the present study) was included in both the Korean and the present studies. In the Korean study, this polymorphism, composed of a common haplotype with another SNP +54529A>G, was found to be loosely associated with type 2 diabetes (11). Data from the present study are not only consistent with the report from the Korean study but also provide evidence of a link between this polymorphism and fasting glucose levels in nondiabetic control subjects. Furthermore, we found that the haplotype H2 constructed by this and other SNPs are associated with fasting triglyceride levels.

Type 2 diabetes is a multifactorial disease that involves insulin resistance and impaired glucose-induced insulin release (1). In most patients, type 2 diabetes is caused by alterations of several genes, each of which has a partial and additive effect (19). PGC-1, NRF1, and PPARγ interact to influence the biogenesis and maintenance of normal function of mitochondria (2,6). To better understand the mechanisms of abnormal mitochondrial oxidative phosphorylation involved in the pathogenesis of type 2 diabetes, our further studies will involve a multiplex-gene analysis of NRF1, PGC-1, and PPARγ genes.

In conclusion, the present study provides evidence that SNPs in the NRF1 gene are associated with type 2 diabetes in a Han Chinese population. Our findings strengthen a previous genetic report from a Korean population and suggest that the susceptibility of the NRF1 genetic polymorphisms may be involved in abnormalities in triglyceride metabolism.

This study has received grants from Natural Sciences Foundation of China (no. 30170886), the Research Fund for the Doctoral Training Program from the Ministry of Education in China (no. 20020023008), the Science Fund for Creative Research Groups (no. 30421003), and the Grant for Collaborative Diabetes Research between China and Europe from European Foundation for the Study of Diabetes.

We thank all subjects for participating in this study and the clinical doctors from four Top Grade hospitals in Beijing and Hebei province, China, for their excellent collection of samples. We are also grateful to Prof. Dongfeng Gu and Prof. Zhenglai Wu for revision of the English language in the manuscript.