Association of Solute Carrier Family 12 (Sodium/Chloride) Member 3 With Diabetic Nephropathy, Identified by Genome-Wide Analyses of Single Nucleotide Polymorphisms

DNA samples were obtained from peripheral blood of the patients with type 2 diabetes who come regularly to the outpatient clinic of Shiga University of Medical Science, Tokyo Women’s Medical University, Juntendo University, Kawasaki Medical School, Iwate Medical University, or Chibanishi General Hospital. All patients provided informed consent before enrolling in this study, and DNA extraction was performed according to standard phenol-chloroform procedures. Diabetic patients were divided into two groups according to the following diagnostic criteria: 1) nephropathy cases, i.e., patients with diabetic retinopathy as well as overt nephropathy indicated by urinary albumin excretion rates (AERs) ≥200 μg/min or urinary albumin-to-creatinine ratio (Alb/Cr) ≥300 mg/g Cr, or under chronic renal replacement therapy; and 2) control subjects, patients with diabetic retinopathy but showing no evidence of renal dysfunction, i.e., AER <20 μg/min or Alb/Cr <30 mg/g Cr. The SNPs for genotyping were randomly selected from our gene-based SNP database (snp.ims.u-tokyo.ac.jp) (16,17). The genotype of each SNP locus was analyzed with invader assay as previously described (14).

Our first screening involved genotyping 94 nephropathy patients and 94 control subjects for 56648 SNP loci. By evaluating the statistical data using a 2 × 2 contingency table, we selected SNPs that showed significant differences in allelic frequencies between the nephropathy and control groups. Then we analyzed in a larger number of subjects to clarify statistical significance. The study protocol was approved by the ethics committees of the Institute of Physical and Chemical Research and each participating institution.

On the basis of the GenBank information pertaining to the genomic sequence containing the SLC12A3 gene (accession no. AC_012181.6), we designed PCR primers to amplify fragments of this DNA. Repetitive elements were excluded from the search by invoking the REPEAT MASKER computer program in the manner described by Bedell et al. (18). PCR experiments and DNA sequencing were carried out as previously described (19). Genotyping of each SNP was performed with invader assay or in some cases with direct sequencing.

Analysis of haplotype structure was carried out by estimating haplotype phasing using Expectation Maximization (EM) algorithm (20) and by constructing haplotype blocks as previously described (21) with the following modifications. First, we clustered neighboring SNPs that were linked absolutely into a single representative SNP to simplify further analyses. Second, we added the constraint that the set of common haplotypes of max(N+1,2 [0.5 N]), where N was the number of SNPs in the block, would cover >90% of the population. We counted the frequency of each haplotype for patients and control subjects independently.

Statistical analysis for association study, for haplotype frequencies and Hardy-Weinberg equilibrium, and for calculation of linkage disequilibrium (LD) coefficients D′ were described previously (22,23). For analyzing clinical data, statistical significance between any two groups was analyzed by one-way ANOVA followed by Scheffe’s test.

Clinical characteristics of each group are summarized in Table 1. First, we genotyped 56,648 SNP loci for 94 nephropathy patients and 94 control subjects. A total of 402 SNP loci that had shown P < 0.01 between the two groups were further analyzed using a larger number of patients. This process excluded most of the 402 SNP loci because of the higher P value (>0.01), but five SNP loci still showed a significant association (P < 0.01). Among them, one SNP locus in the 24th intron of the SLC12A3 gene on chromosome 16q13 is most strongly associated with diabetic nephropathy (G vs. A, χ² = 15.4, P = 0.000087, odds ratio = 2.53, 95% CI 1.57–4.09; Table 2).

Subsequent LD analysis of 40 SNPs around this landmark SNP in the SLC12A3 gene revealed that the LD in this region extended to ∼100 kb upstream and to 50 kb downstream of the landmark SNP site (Fig. 1). Therefore, the critical region for susceptibility to diabetic nephropathy seemed to lie within this 150-kb LD block, which contained three genes (SLC12A3, KIAA0095, and the gene encoding homocysteine-inducible, endoplasmic reticulum stress-inducible, ubiquitin-like domain member 1 [HERPUD1]). We then analyzed 16 SNPs in KIAA0095 and 15 in HERPUD1, as well as five in the gene encoding cholesteryl ester transfer protein for the association with diabetic nephropathy; the third one had been proposed as a candidate for diabetic nephropathy (24). We found no significant association of these additional SNPs with diabetic nephropathy (P > 0.01; data not shown) and concluded that SLC12A3 itself was the most likely candidate for conferring susceptibility to diabetic nephropathy.

To validate further the candidacy of SLC12A3 for diabetic nephropathy, we searched for additional polymorphisms in the gene and identified 65 novel SNPs, making a total of 78 in this gene, and four insertion/deletion polymorphisms (Fig. 2). From among those SNPs, we selected 26 for the Invader assays (one SNP from the 5′ flanking region, one in the first intron, 21 in other introns, and three in exons) and successfully genotyped for a total of 553 nephropathy patients and 317 control subjects. In addition, we analyzed the three exonic SNPs as well as two in the first intron by direct sequencing. Four of these 31 SNPs showed significant associations with diabetic nephropathy; particularly, one SNP in the 23rd exon, which substituted an amino acid (+78, Arg913Gln), was strongly associated (χ² = 18.5, P = 0.00002, odds ratio = 2.53, 95% CI 1.64–3.90; Table 3). The four SNPs were in quasi-complete LD. We then selected 18 SNPs with the minor allelic frequency of >5% and analyzed the haplotype structure. Eleven of the 18 SNP loci, including the one in exon 23, constituted one haplotype block, and 10 common haplotypes of this block accounted for >90% of the study population (Fig. 3). A subsequent investigation of these haplotypes and diabetic nephropathy identified a significant association of haplotype 7. However, this association was less significant than that observed by the Arg913Gln SNP alone. The genotypic distributions of SNPs shown in the tables all were in Hardy-Weinberg equilibrium.

In a genome-wide case-control association study using SNPs as genetic markers, we identified the SLC12A3 gene as a candidate for conferring susceptibility to diabetic nephropathy. Our data also suggested that an amino-acid substitution of Gln for Arg at codon 913 might contribute directly to the reduction of risk of diabetic nephropathy by affecting the function of the gene product, a thiazide-sensitive Na⁺-Cl⁻ co-transporter.

The results presented here provide the evidence that a genome-wide case-control association study using gene-based SNPs as genetic markers is a powerful strategy for identifying genes associated with susceptibility to common diseases. Epidemiological finding had strongly suggested a contribution of genetic factors to the development and progression of diabetic nephropathy (4,5), but worldwide efforts have so far failed to identify any solid evidence to indicate a gene susceptible to the disease. Results even conflicted in many cases, probably because sample sizes were often inappropriate. However, the major cause of failures to obtain solid conclusions is possibly that multiple genetic factors are involved in diabetic nephropathy in a complex manner and the power of each factor is too weak to be identified. Therefore, approaches other than standard candidate-gene analysis or family-based linkage analysis seem to be required to identify genes that are involved in susceptibility to common diseases such as this one. The strategy that we used for this study had two major advantages over previous investigations. We started by analyzing a huge number of loci (56,648 SNPs) on a genome-wide basis, using a high-throughput genotyping system developed in our institute (14). Second, all of the SNPs contributing to this report are the gene-based SNPs in the Japanese population (16,17); therefore, we were able to screen for the candidate genes more efficiently than other SNP databases.

In a large-scale genome-wide association study for common diseases such as in this study, the elimination for type 1 error should be concerned. Hence, we performed correction of multiple-testing error for our whole screening process according to the following calculations. Overall P values (P 1st × P 2nd) × number of test (test 1st + test 2nd) = 0.0095 × 0.000087 × (56,648 + 402) = 0.047. Therefore, we thought that the association of this locus with diabetic nephropathy was statistically significant.

Another possible cause to generate spurious association in case-control studies is related to the power of the study protocol to detect the association. In evaluation of polymorphisms whose contribution is modest to the disease susceptibility, >1,000 subjects might be necessary for each patient and control group if the minor allelic frequencies are relatively small (<10%) like the one in this study. However, we could observe the positive association of this locus with diabetic nephropathy consistently in two independent sets of case-control groups; therefore, we believe that the possibility of false positive is not likely in this case.

The SLC12A3 gene, at chromosome 16q13, encodes a thiazide-sensitive Na⁺-Cl⁻ co-transporter that mediates reabsorption of Na⁺ and of Cl⁻ at the renal distal convoluted tubule; this molecule is the target of thiazide diuretics. Mutations in the SLC12A3 gene are responsible for Gitelman syndrome (15), which is inherited as an autosomal recessive trait characterized by hypokalemia, metabolic alkalosis, hypomagnesemia, hypocalciuria, and volume depletion. Simmons et al. (15) reported 16 mutations of the SLC12A3 gene among 12 patients with this syndrome, and others have identified additional mutations (25–27). Loss of function of this thiazide-sensitive Na⁺-Cl⁻ co-transporter consistently results in fluid and electrolyte abnormalities, characteristics of Gitelman syndrome. Although the substitution of Gln for Arg at codon 913 is among the mutations identified in patients with Gitelman syndrome (25), no individuals in our test population showed clinical Gitelman syndrome even though two of them were homozygous for the substituted allele. Therefore, this substitution is not the cause of Gitelman syndrome but a polymorphism. In fact, higher frequency of the substituted allele in our control subjects (8% in control vs. 3% in nephropathy; χ² = 18.5, P = 0.00002) suggested that the carrier of the substituted allele might be protected from the progression of nephropathy through a chronic and mild decrease in the activity of this ion transporter.

The precise mechanism accounting for the reduced risk of diabetic nephropathy in patients with the minor allele (913Gln) is unclear. It was reported that the loss of function of this thiazide-sensitive Na⁺-Cl⁻ co-transporter significantly reduced arterial blood pressures in sodium-restricted mice (28); therefore, one can hypothesize that substitution of 913Arg to Gln could affect the blood pressure in humans under pathological conditions such as diabetes. Furthermore, thiazide diuretics are widely used for the treatment of patients with hypertension. Because elevated blood pressure is an independent risk factor for development and progression of diabetic nephropathy (29), this substitution might contribute to the reduced susceptibility for renal disease by decreasing arterial blood pressure in type 2 diabetic subjects. Alternatively, this transporter could regulate reabsorption of unknown molecules, and a sustained decrease in its activity might lead to chronic accumulation of nephrotoxic substances in patients with type 2 diabetes. Our cross-sectional data suggest that the former is not the case (data not shown), but prospective randomized trials will be necessary to clarify the precise mechanism by which this genetic polymorphism contributes to susceptibility for diabetic nephropathy.

In summary, we identified the SLC12A3 gene as a candidate for conferring susceptibility to diabetic nephropathy, on the basis of a genome-wide case-control association study using gene-based SNPs. The evidence presented here, when combined with previous reports, suggests that SLC12A3 should be a target for new drugs to aid in the prevention and treatment of diabetic nephropathy.

This work was supported by a grant from the Japanese Millennium Project.

We acknowledge Akio Kanazawa, Shuichi Tsukada, Kana Ohashi, Yoshihiro Kawamura, Atsuyuki Shimazaki, Shuichi Iwashita, and the technical stuff at the SNP Research Center, The Institute of Physical and Chemical Research.