Introduction

The gene KCNQ1 encodes the pore-forming α subunit of a voltage-gated potassium channel and was initially identified as a type 2 diabetes susceptibility gene in genome-wide association studies (GWAS) [1, 2]. Three single-nucleotide polymorphisms (SNPs)—rs2237895, rs2237897 and rs2283228—within KCNQ1 have been found to be strongly associated with type 2 diabetes in both East Asian and European populations. More recently, a study of Japanese patients with type 2 diabetes had suggested that, interestingly, KCNQ1 may also confer susceptibility to diabetic nephropathy, as evidenced by the presence of macroalbuminuria [3]. In that report, results of the initial study showed an association with proteinuria, while replication studies showed a similar trend but individually failed to reach statistical significance. In view of this initial report by Ohshige et al [3], we have attempted to replicate this association with diabetic nephropathy in another East Asian population by studying 752 Chinese patients with type 2 diabetes. In addition, we also investigated whether these SNPs modulated albuminuria as a continuous trait (log e albumin/creatinine ratio [ACR]).

Methods

Chinese patient population

A total of 752 Chinese patients with type 2 diabetes from the Singapore Diabetes Cohort Study were included in this study. Briefly, the Singapore Diabetes Cohort Study was made up of consenting patients who had been previously diagnosed with type 2 diabetes at primary care facilities of the National Healthcare Group Polyclinics in Singapore. Information on demographics, lifestyle factors and medical family history were obtained by questionnaire. Patients’ physical measurements and blood and spot urine specimens were taken, and their medical records were reviewed to obtain information on their metabolic control and the presence of comorbidities and complications. The research protocol was approved by the National Healthcare Group Domain-Specific Review Board.

Renal traits

Mid-stream random spot urine samples were collected and transported chilled to the laboratory, centrifuged and stored at −80°C before analysis. Albuminuria was determined on the basis of the ACR (mg/mmol) using commercially available kits for albumin and creatinine measurements (Exocell, Philadelphia, PA, USA). Patients with ACR <3.4 mg/mmol were considered normoalbuminuric, while those with 3.4 ≤ ACR < 33.9 mg/mmol and ACR ≥33.9 mg/mmol were microalbuminuric and macroalbuminuric, respectively. Renal function (expressed as estimated GFR [eGFR]) was estimated using the simplified Modification of Diet in Renal Disease equation where eGFR (ml min−1 1.73 m−2) = 186.3 × (plasma creatinine in mg/dl)−1.154 × (age in years)−0.203 × (0.742 for women) × (1.21 if participant is black) [4].

Japanese replication study

For the purpose of replication and meta-analyses, we studied type 2 diabetic patients from the initial case–control study reported by Ohshige et al [3]. Cases (n = 754) comprised patients with diabetic retinopathy and a urinary albumin excretion rate ≥200 μg/min or an ACR ≥33.9 mg/mmol, while controls (n = 558) had retinopathy with a urinary albumin excretion rate <20 μg/min or ACR <3.4 mg/mmol.

Genotyping

Three SNPs in intron 15 of KCNQ1, rs2237895 (A>C), rs2237897 (C>T) and rs2283228 (A>C), were genotyped using a multiplex PCR invader assay [5] and TaqMan SNP genotyping assay (Applied Biosystems, Foster City, CA, USA). All SNPs had a more than 90% genotyping success rate.

Statistical analysis

All statistical analyses were carried out using Stata V.10 (http://www.stata.com), assuming a two-sided test with a 5% level of significance. Genotype and allele frequencies were tested for Hardy–Weinberg equilibrium proportions using the χ2 test. Differences in characteristics between sexes were compared using the χ2 test for categorical variables. For continuous variables that were normally distributed, ANOVA was used to compare differences in mean between sex. Otherwise, the Kruskal–Wallis test was implemented and the medians were compared. Multivariate analyses on the association of individual KCNQ1 SNPs with log e ACR were performed using linear regression. Multivariate logistic regression was used to determine the associations between KCNQ1 SNPs and the presence of micro- and macroalbuminuria assuming dominant, recessive and additive genetic models. Correction for multiple hypothesis testing was not straightforward, particularly as the renal traits or genetic models were not entirely independent. Nevertheless, to minimise type 1 error, Bonferroni correction was performed based on albuminuria analysed as categorical (micro- and macroalbuminuria) and continuous (log e ACR) traits under three genetic models (3 × 3 = 9 comparisons). An uncorrected p value <0.006 would therefore be required to declare significance.

Meta-analysis was performed for each SNP under all three genetic models. Log e OR was used as the measure of effect size, and the standard error of the log e OR as a measure of the study’s precision. The I 2 index was used to measure the extent of between-study heterogeneity. Since the I 2 index, like other heterogeneity indices, has little power in detecting heterogeneity with a small number of studies, a liberal type I error of 10% was used for detecting heterogeneity. If the studies were found to be heterogeneous, a random-effect meta-analysis method was performed. Otherwise, fixed-effect meta-analysis was performed.

Results

The clinical characteristics of the 752 type 2 diabetic patients analysed in this study are presented according to sex (electronic supplementary material [ESM] Table 1) and nephropathy status (ESM Table 2). Intersex differences were observed for WHR, diastolic blood pressure, and total and HDL-cholesterol. Diabetes duration, HbA1c, blood pressure and eGFR were associated with nephropathy status. The genotype distributions of all three KCNQ1 SNPs were in Hardy–Weinberg equilibrium regardless of whether the patients were analysed as a whole or categorised by sex or nephropathy status (ESM Tables 1 and 2). Moderate linkage disequilibrium was observed between the three SNPs in our Chinese patients, and this pattern was similar to that observed in the Japanese study based on pilot data from the 1000 Genomes Project (ESM Table 3).

SNP rs2283228 was strongly associated with macroalbuminuria in multivariate analyses under the recessive (uncorrected p = 1.492 × 10−5, corrected p = 1.343 × 10−4) and additive (uncorrected p = 8.167 × 10−5, corrected p = 7.350 × 10−4) models (Table 1). The dominant model revealed associations of only borderline significance, which did not withstand correction for multiple hypothesis testing (Table 1). SNP rs2237897, which was associated with macroalbuminuria in the earlier Japanese study [3], was also associated in our study, but the significance was borderline (uncorrected p = 0.025 under the additive model, corrected p = NS).

Table 1 Association between KCNQ1 SNPs and macroalbuminuria in multivariate analyses based on recessive, dominant and additive models
Full size table

None of the SNPs were associated with microalbuminuria (ACR = 3.4 to <33.9 mg/mmol) regardless of the genetic model used (ESM Table 4). A trend of increasing OR was, however, observed when patients with low (ACR = 3.4 to <7.5 mg/mmol) or high (ACR = 7.5 to <33.9 mg/mmol) microalbuminuria and finally macroalbuminuria (ACR ≥33.9 mg/mmol) were compared. For instance, the OR for rs2283228 increased from low microalbuminuria (OR 1.19, 95% CI 0.57, 2.46) to high microalbuminuria (OR 1.88, 95% CI 0.98, 3.61) to macroalbuminuria (OR 6.00, 95% CI 2.68, 13.41) under the recessive model (Table 1 and ESM Table 4). This increasing trend was also observed for rs2283228 under the additive and dominant models.

With respect to the analysis of albuminuria as a continuous trait, SNP rs2283228 was significantly associated with log e ACR under the recessive model (p = 0.004, corrected p = 0.036) even after adjustment for eGFR (Table 2). The other significant covariates included diabetes duration and HbA1c. This SNP also showed borderline association with log e ACR under the additive model (p = 0.028, corrected p = NS) (Table 2). Patients who were ‘CC’ recessive had the highest ACR reading (median = 2.0 mg/mmol, interquartile range [IQR] = 0.6–14.7 mg/mmol) compared with ‘AA’ (1.6 mg/mmol, IQR = 0.7–4.2 mg/mmol) and ‘AC’ (1.5 mg/g, IQR = 0.7–4.9 mg/mmol) genotypes (p = 0.0331).

Table 2 Association of log e ACR with KCNQ1 SNPs in Chinese patients with type 2 diabetes under recessive, dominant and additive models
Full size table

To investigate whether rs2283228 could be replicated in another East Asian population, this SNP was genotyped in a case–control study of type 2 diabetic Japanese patients previously included in the initial study by Ohshige et al [3] (see the Methods section). In multivariate analyses, rs2283228 was significantly associated with macroalbuminuria (ESM Table 5). In meta-analyses combining data from Chinese and Japanese patients, macroalbuminuria was significantly associated with rs2283228 under the dominant model with a pooled OR of 1.50 (95%CI 1.15, 41.96) (meta-analysis p value = 0.003) (ESM Fig. 1). A similar but non-significant trend was observed for this SNP under the recessive and additive models (ESM Fig. 1). Macroalbuminuria was significantly associated with rs2237897 under the dominant (meta-analysis p value = 0.002), recessive (meta-analysis p value = 0.031) and additive (meta-analysis p value = 0.001) models (ESM Fig. 2).

Discussion

Our study has provided evidence to support the hypothesis that genetic variation in KCNQ1 can significantly affect albuminuria in Chinese patients with type 2 diabetes. This result is consistent with the earlier report by Ohshige et al [3], which showed that KCNQ1 SNPs were associated with proteinuria in Japanese patients with type 2 diabetes. In the Japanese population, rs2237897 was the SNP most strongly associated with proteinuria. While we did find some evidence of replication of this SNP in our Chinese patients, our study also revealed a novel positive association with rs2283228, which we subsequently replicated in the Japanese population. Meta-analyses of the Chinese and Japanese datasets revealed both SNPs to be positively associated with macroalbuminuria. As with the original GWAS findings linking KCNQ1 to type 2 diabetes, the effect of these genetic variants would seem to be more relevant in East Asian populations as opposed to certain other human populations (e.g. those of European descent), as these KCNQ1 SNPs may be substantially less polymorphic.

A major limitation of our study is the modest number of Chinese patients with macroalbuminuria. Thus, although the associations were significant even after correction for multiple hypothesis testing, one might wish to be cautious in drawing conclusions. Encouragingly, however, we were able to successfully follow up on our positive findings in an independent collection of Japanese patients. It was also notable that, as the severity of diabetic nephropathy increased—namely, as it progressed from low to high microalbuminuria and finally macroalbuminuria—a rise in the magnitude of association (OR) was observed. This, in our opinion, is at least consistent with the notion that the association with macroalbuminuria was not by chance. Another limitation is that the ACR readings were based on a single urine sample collected at entry into the Chinese study. This could potentially affect the precision of the disease classification and inadvertently made it more difficult to detect minor genetic effects. On the other hand, this ACR reading correlated highly with multiple urinary dipstick measurements obtained in visits to the clinic preceding their entry into our patient population. Any disease misclassification that might have inadvertently occurred did not appear to have a large negative effect on our study, as highly significant associations were readily detected.

It is worth reiterating that KCNQ1 was initially identified as a gene predisposing to type 2 diabetes. Yet, its implication in renal traits is biologically plausible. For instance, KCNQ1 (potassium voltage-gated channel, KQT-like subfamily, member 1) molecularly assembles with KCNE1 (potassium voltage-gated channel, Isk-related family, member 1) to form a potassium ion channel that has been localised to the brush border of the proximal tubule. This potassium channel facilitates the maintenance of an ion potential gradient, which serves to drive Na+ secretion in the nephron [6]. Given that the SNPs are intronic, it is not clear how they would act to affect the molecular biology of KCNQ1. Potentially, this may affect its gene expression, but a survey of the current literature did not shed light on the potential functional roles of these KCNQ1 SNPs. Unfortunately, this lack of functional information is commonplace for many of the GWAS associated SNPs. Future studies aimed at delineating this biological aspect will certainly aid our understanding of the role of these disease genes.