Figures
Abstract
Background
Potassium voltage-gated channel, KQT-like subfamily, member 1 (KCNQ1) is thought to be an important candidate gene of diabetes. Several single nucleotide polymorphisms (SNPs) in a 40-kb linkage disequilibrium (LD) block in its intron 15 have been identified to be associated with diabetes in East Asian populations in recent genome-wide association studies. The aim of this study was to investigate whether KCNQ1 polymorphisms influence the levels of the metabolic phenotypes in general Chinese populations.
Methodology/Principal Findings
We investigated the associations of two SNPs (rs2237892 and rs2237895) in the aforementioned 40-kb LD block, a missense variant rs12720449 (P448R) in exon 10, and a synonymous variant rs1057128 (S546S) in exon 13 with metabolic phenotypes in a Uyghur population (n = 478) and replicated these associations in a Han population (n = 2,485). We found that rs2237892-T allele was significantly associated with decreased triglyceride levels (pcombined = 0.001). The minor G allele of the rs12720449, with sharp difference of the allelic frequency between European and East Asian populations (0.2% versus 14%, respectively), was associated with a lower triglyceride levels than G allele in Uyghur subjects (p = 0.004), in Han subjects (p = 0.052), and in subjects of meta-analysis (pcombined = 0.001). Moreover, the minor A allele of the rs1057128 was also associated with decreased triglyceride levels in meta-analysis (pcombined = 0.010).
Conclusions
To the best of our knowledge, this is the first report associating a missense mutation of KCNQ1, rs12720449, with triglyceride levels. Rs2237892, representing the 40-kb LD block, is also associated with triglyceride levels in Han population. Further studies are required to replicate these findings in other East Asian populations.
Citation: Chen X-d, Yang Y-j, Li S-y, Peng Q-q, Zheng L-j, Jin L, et al. (2012) Several Polymorphisms of KCNQ1 Gene Are Associated with Plasma Lipid Levels in General Chinese Populations. PLoS ONE 7(3): e34229. https://doi.org/10.1371/journal.pone.0034229
Editor: Alejandro Lucia, Universidad Europea de Madrid, Spain
Received: September 15, 2011; Accepted: February 24, 2012; Published: March 29, 2012
Copyright: © 2012 Chen et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants-in-aids from the National Natural Science Foundation (30625016, 30890034, 30971237, 31171216 and 81100551) and a grant from Major National Science and Technology Program (2008ZX10002-016). The authors thank the members of the survey teams and the participants for their contribution to this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
The potassium voltage-gated channel, KQT-like subfamily, member 1 (KCNQ1), which encodes the pore-forming voltage-gated K+ channel subunit KvLQT1 and is widely expressed in myocardial tissue, plays a key role in the repolarization of the cardiac action potential and in the transport of water and salt in epithelial tissues [1]–[3]. Loss-of-function and gain-of-function mutations in this gene have been associated with cardiac long QT syndrome, Lange-Nielsen cardioauditory syndrome, atrial fibrillation, and congenital deafness [2]–[6].
KCNQ1 is expressed in the inner ear, stomach, intestine, liver, kidney, and pancreatic islets [5]. In the pancreas, KCNQ1 contributes to the regulation of insulin secretion, and a blockade of the KvLQT1 channel may lead to increased insulin secretion [7], [8]. Recently, two independent genome-wide association studies (GWAS) revealed that several SNPs (including rs2237892, rs2237895, rs2283228, and rs2237897) within a 40-kb linkage disequilibrium (LD) block in intron 15 of KCNQ1 were consistently associated with type 2 diabetes and the impairment of insulin secretion in East Asian populations [9], [10]. For these SNPs, the associations with type 2 diabetes have been replicated, predominantly not only in East Asian ethnic groups, including Chinese [11], [12] and Singaporean populations [13], but also in Euro-Caucasian populations from Denmark [10], [14] and Sweden [15]. Moreover, rs2237892 and rs2237895 have also been associated with metabolic phenotypes, such as glucose and body mass index (BMI) levels, in East Asian populations [13], [16].
Because KCNQ1 is critical for the regulation of insulin secretion [5] and insulin is important for the regulation of metabolic phenotypes, we hypothesized that genetic polymorphisms in the KCNQ1 gene may underlie differences in metabolic phenotypes, such as triglyceride (TG) and total cholesterol (TC). In addition to SNPs located in the 40-kb LD region, the pathogenesis effects of another two important SNPs in the exon regions of the KCNQ1 gene, rs1057128 and rs12720449, need to be explored. Synonymous variant rs1057128 (S546S), located in exon 13 of the KCNQ1 gene, has been found to be associated with various cardiac arrhythmias, such as Long-QT syndrome, atrial flutter, and atrial fibrillation, in European and Chinese individuals [17], [18]. Non-synonymous variant rs12720449 (P448R), located in exon 10 of KCNQ1, was also found to be associated with Long-QT syndrome [19], [20]. The minor allele frequency (MAF) of this SNP is distinct between Europeans and East Asians (0.2% and 5.8%, respectively), according to the NCBI SNP databank (http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs=12720449). No prior studies have investigated the association between these two SNPs (rs1057128, rs12720449) and metabolic phenotypes.
The aim of the present study was to investigate the effects of rs1057128, rs12720449, and two SNPs (rs2237892 and rs2237895) within the aforementioned 40-kb LD block on several metabolic phenotypes. We first observed the associations of these four SNPs with BMI, waist-to-hip ratio (WHR), systolic blood pressure (SBP), diastolic blood pressure (DBP), glucose (Glu), triglyceride (TG), and total cholesterol (TC) in a Uyghur population, an ethnic group with European and East Asian ancestry. Then we observed the associations in a Han population, the ethnic group that accounts for 94% of the Chinese populations, with a larger sample size.
Results
The genotypic and allelic frequencies of the studied SNPs
The genomic characteristics of rs1057128, rs2237892, rs2237895, and rs12720449 were summarized in Table 1. All SNPs were consistent with Hardy–Weinberg expectations in both populations (p>0.01). The MAFs of rs12720449 were different for the Han and Uyghur populations in the present study (0.14 and 0.05, respectively). Pair-wise linkage analysis with the four SNPs showed weak LDs (data not shown). To the best of our knowledge, this is the first study investigating SNPs in KCNQ1 in a Uyghur population. Because the sample size for the Uyghur population was relatively small, we compared the metabolic phenotypes of rare allele carriers (CG/GG for rs12720449) with those of non-carriers in the subsequent analysis. The analysis was performed as previously described in the original GWAS [21].
The association between the four SNPs in KCNQ1 and plasma TG levels
We presented the association results between the four SNPs in KCNQ1 and plasma TG levels in Table 2. For SNP rs12720449, significantly lower TG levels were observed in CG/GG carriers (0.94±0.35) than in CC carriers (1.08±0.35) (p = 0.004) in the Uyghur population. A preliminary significant association was also observed in the Han population, and the mean TG levels of GG, CG, and CC carriers were 0.87±0.30, 1.03±0.92, and 1.08±0.96, respectively (p = 0.052). The minor alleles of rs1057128 (A) and rs2237892 (T) were significantly associated with decreased TG levels in Han subjects (p = 0.011 for both alleles).
To investigate allelic associations with TG levels, we performed a meta-analysis across Uyghur and Han datasets (Table 2). We found that the minor alleles of rs12720449 (G), rs1057128 (A), and rs2237892 (T) decreased TG levels at a multiple correction threshold level of p<0.0125 (p = 0.001, 0.010 and 0.001, respectively). We also evaluated these associations by including all four SNPs in one model, and the associations between rs1057128, rs2237892, rs12720449 and TG levels remained significant (p = 0.007, 0.001, and 0.001, respectively).
The association between the four SNPs and the principal components of the metabolic phenotypes
To further validate our results, we conducted a principal component analysis in the Han population using nine metabolic-related traits: BMI, WHR, SBP, DBP, Glu, HDL, LDL, TG, and TC. First, we identified two principal components (PC) of these traits to generate canonical variables (PC1 and PC2) (Table 3). Next, we performed a classic single-trait approach to test the associations between SNPs and the canonical variables. As shown in Table 3, PC2 is strongly correlated with TG (with an eigenvalue of 0.98). All four SNPs were significantly associated with PC2 at a threshold level of p<0.05 (p = 0.017, 0.020, 0.043, and 0.050), but these association did not reach the multiple correction threshold level of p<0.0125.
The association between SNPs in KCNQ1 and other metabolic phenotypes excluding TG
We investigated the relationships of the four SNPs in KCNQ1 with other metabolic-related parameters, including BMI, WHR, SBP, DBP, Glu, HDL, LDL, and TC. We found a significant association between rs1057128 and SBP in the Han population at the significant level of 0.0125. In the Uyghur population, we found a significant association between rs2237892 and WHR at the significant level of 0.05 (Table 4). We also performed a meta-analysis across the two datasets for the Uyghur and Han populations, but we found that none of associations reached the multiple correction threshold level of p<0.0125.
Discussion
In this population-based association study, we replicated the association between rs2237892, a SNP in intron 15 of KCNQ1 gene, and TG levels. We also identified a novel association between rs1057128 (S546S), which is located in exon 13 of the KCNQ1 gene, and TG levels. Most importantly, we identified a missense mutation, rs12720449 (P448R), located in exon 10 of KCNQ1, associated with TG levels.
Intron 15 of the KCNQ1 gene spans approximately 70 kb on chromosome 11p15.5. An extremely long LD block, covering approximately 40 kb, was observed in a Japanese population, and several SNPs (including rs2237892, rs2237895, rs2283228, and rs2237897) located in this LD block were found to be associated with diabetes in Japanese individuals [9], [10]. However, the LD block in this 40-kb region was weak in other ethnically distinct East Asian groups, such as Singaporean and Han Chinese populations [13], and the associations of these SNPs with diabetes were not consistently replicated in these East Asian populations [11], [13], [16], [22]. Moreover, the associations of the SNPs in this region with other metabolic phenotypes, such as lipid levels, assessed in East Asian populations were inconsistent [23]. In the present study, we did not observe an association between glucose levels and the two SNPs, rs2237892 and rs2237895, within this region. However, we detected associations between rs2237892 and TG levels, rs2237895 and LDL levels, and rs2237895 and TC levels. In the Japanese population, the diabetes causal variant may be in LD with this 40-kb region. In other East Asian ethnic groups, such as Han Chinese or Singaporean populations, different variants affecting lipid metabolic regulation may be in LD with rs2237892 or rs2237895 [13]. Further studies will need to be performed to identify these functional variants.
In the present study, rs1057128 was associated with TG levels in the Han population. This SNP was originally identified in association with various cardiac arrhythmias [17], [18]. Although rs1057128 (S546S) is a synonymous variant, there is evidence indicating that synonymous mutations can affect the thermodynamic stability of mRNA secondary structures or affect splicing through phenomena such as exon skipping; therefore, synonymous mutations may not be neutral in evolution [24]. Additionally, a small domain between residues 589 and 620 in the KCNQ1 C terminus may function as an assembly domain for KCNQ1 subunits [25]. Without this domain, KCNQ1 C termini fail to assemble, and functional potassium channels are not produced. Rs1057128 is located in the KCNQ1 C terminus close to this domain and may therefore affect the functional protein through this mechanism.
The most important finding of the present study was the association between rs12720449 in KCNQ1 and TG levels. rs12720449 (P448R), a missense mutation located in exon 10, is a functional variant, and the G allele of rs12720449 can increase channel current by two-fold [19]. The MAFs of this variant are noticeably distinct among different ethnic groups, with ≈15% in East Asians [20], [26], [27], ≈0.2% in Europeans, and ≈6% in Uyghurs, an ethnic group with European and East Asian ancestry. rs12720449 was originally identified as a polymorphism associated with cardiac arrhythmia in people of European descent and was not generally considered to be associated with polygenic diseases, such as dyslipidemia, possibly as a result of its low frequency (0.2%) in European populations [28], [29]. In the present study, this variant was found to be associated with TG levels in both an East Asian population (Han) and an admixed population (Uyghur). However, the underlying mechanism remains unclear, and further studies are needed to elucidate the roles of rs12720449 in evolutionary biology and complex phenotypes.
We also identified several associations between KCNQ1 polymorphisms and other metabolic phenotypes at a threshold of p<0.05 (e.g., rs12720449 and LDL, rs2237892 and WHR, rs2237895 and LDL, and rs2237895 and TC), but we did not detect any association with glucose levels in either Uyghur or Han populations. These variants may result in an increased expression of KCNQ1 and a subsequent increase in insulin secretion in the pancreatic β cells [8]. Increased insulin inhibits glucose production and stimulates lipid synthesis, which may selectively increase circulating lipid levels without increasing blood glucose levels [30]. In light of the present findings, KCNQ1 may have different effects on the complex lipid and glucose phenotypes.
There are some advantages and disadvantages in our study. The major advantage of the present study is the population-based approach, which overcomes the potential selection bias associated with hospital-based designs. However, this study also has some limitations: first, only four important SNPs in the KCNQ1 gene were examined for possible association with metabolic-related parameters, although other associated variants may exist; second, no functional experiments were performed to validate our results; third, the sample size of the present study (N = 2963) was relatively small, which limited the statistic power for identifying associations between genetic variants and phenotypes with low prevalence rate (i.e., the fact that only 20 subjects had type 2 diabetes prohibited us to measure the relationship of these variants with diabetes risk).
In summary, we found that several polymorphisms in the KCNQ1 gene were associated with metabolic phenotypes, especially TG levels, in Han Chinese and/or in Uyghur populations. Further studies are needed to replicate our findings in other East Asian populations.
Materials and Methods
Subjects
The Uyghur subjects in this study included 478 individuals (273 men and 205 women with a mean age of 52.95±11.28 years) randomly recruited from March to May 2005 and April 2006 from four villages in Tulupan District, Xinjiang Uyghur Autonomous Region, China. The 2,485 Han subjects in this study were randomly selected from the baseline samples of the Taizhou Longitudinal Study (TZL). TZL, which was initiated in July 2007, was an open-ended prospective study with very broad research aims. The design and baseline characteristics of this study have been previously described [31].
Subjects who reported receiving medications for the treatment of metabolic-related diseases were excluded. Subjects who reported having tumors, autoimmune diseases, or hematological diseases were also excluded. Written informed consent was obtained from each participant, and all protocols were approved by the Human Ethics Committee of Fudan University.
Data collection
A standardized interview was conducted by trained personnel, and detailed information was collected concerning medical history and lifestyle characteristics, such as smoking and alcohol consumption. All participants received a physical examination and blood tests at a local hospital after overnight fasting. A standardized mercury sphygmomanometer was used to measure SBP and DBP, and these measurements were performed by two cardiologists. Body weight and height were measured with subjects wearing only light indoor clothing and without shoes. BMI was calculated by dividing weight (kg) by height squared (m2). The waist circumference was measured midway between the caudal point of the costal arch, as palpated laterally, and the iliac crest. The hip circumference was measured at the symphysis-trochanter femoris level. WHR was calculated by dividing the waist circumference by the hip circumference. The blood specimens were drawn after overnight fasting, immediately subjected to centrifugation, and analyzed within 8 h for Glu, TG, and TC. The distribution of the clinical parameters is listed in Table 5.
Genotyping
Blood samples were collected into EDTA-containing receptacles, and genomic DNA was extracted using a standard method. Sample DNA (10 ng) was amplified by PCR according to the manufacturer's instructions. Genotyping of the four SNPs in KCNQ1, rs1057128(G>A), rs2237892(C>T), rs2237895(A>C), and rs12720449(C>G) was performed using the TaqMan SNP genotyping assay (Applied Biosystems, Foster City, CA, USA). The genotyping success rates and concordance rates at all four loci were >99% based on 12% duplicate samples (n = 380).
Statistical analysis
Deviation from Hardy–Weinberg expectation for the variants was tested using the chi-square statistic. Haplotype inference was conducted using Haploview software [32]. We treated the metabolic phenotypes (BMI, WHR, SBP, DBP, FBG, LDL, HDL, TG, and TC) as quantitative traits. Because these traits exhibited a skewed distribution, we normalized these traits using the Box-Cox method, which can be used to automatically identify a suitable power transformation for the data. To estimate the association of the variants with the metabolic parameters, we applied a general linear regression model under an additive model, adjusting for sex, age, smoking, and drinking, with the normalized BMI, WHR, SBP, DBP, FBG, LDL, HDL, TG, and TC. Each SNP was separately entered into the model as the explanatory variable. A pre-specified threshold of p<0.0125 was used for multiple correction significance (corresponding to p<0.05 after adjusting for four loci). We also performed a principal component analysis on the nine variables representing metabolic phenotypes in the Han population. The association of the variants in KCNQ1 with changes in the extracted principal components was estimated using a general linear regression model. All analyses were performed using SAS statistical software (release 8.2, SAS Institute Inc, Cary, NC, USA).
Acknowledgments
We thank the members of the survey teams and the participants for their contributions to this study.
Author Contributions
Conceived and designed the experiments: XfW LJ YjY. Performed the experiments: XdC SyL LjZ. Analyzed the data: XdC QqP XfW. Contributed reagents/materials/analysis tools: YjY XfW LJ. Wrote the paper: XdC XfW.
References
- 1. Barhanin J, Lesage F, Guillemare E, Fink M, Lazdunski M, et al. (1996) KvLQT1 and IsK (minK) proteins associate to form the IKs cardiac potassium current. Nature 384: 78.
- 2. Wang Q, Curran ME, Splawski I, Burn TC, Millholland JM, et al. (1996) Positional cloning of a novel potassium channel gene: KVLQT1 mutations cause cardiac arrhythmias. Nature Genetics 12: 17–23.
- 3. Neyroud N, Tesson F, Denjoy I, Leibovici M, Donger C, et al. (1997) A novel mutation in the potassium channel gene KVLQT1 causes the Jervell and Lange-Nielsen cardioauditory syndrome. Nature Genetics 15: 186–189.
- 4. Chen YH, Xu SJ, Bendahhou S, Wang XL, Wang Y, et al. (2003) KCNQ1 gain-of-function mutation in familial atrial fibrillation. Science 299: 251.
- 5. Lee MP, Ravenel JD, Hu RJ, Lustig LR, Tomaselli G, et al. (2000) Targeted disruption of the Kvlqt1 gene causes deafness and gastric hyperplasia in mice. Journal of Clinical Investigation 106: 1447–1455.
- 6. Casimiro MC, Knollmann BC, Ebert SN, Vary JC, Greene AE, et al. (2001) Targeted disruption of the KCNQ1 gene produces a mouse model of Jervell and Lange–Nielsen Syndrome. Proceedings of the National Academy of Sciences 98: 2526.
- 7. Demolombe S, Franco D, de Boer P, Kuperschmidt S, Roden D, et al. (2001) Differential expression of KvLQT1 and its regulator IsK in mouse epithelia. American Journal of Physiology-Cell Physiology 280: C359.
- 8. Ullrich S, Su J, Ranta F, Wittekindt OH, Ris F, et al. (2005) Effects of I Ks channel inhibitors in insulin-secreting INS-1 cells. Pflügers Archiv European Journal of Physiology 451: 428–436.
- 9. Yasuda K, Miyake K, Horikawa Y, Hara K, Osawa H, et al. (2008) Variants in KCNQ1 are associated with susceptibility to type 2 diabetes mellitus. Nature genetics 40: 1092–1097.
- 10. Unoki H, Takahashi A, Kawaguchi T, Hara K, Horikoshi M, et al. (2008) SNPs in KCNQ1 are associated with susceptibility to type 2 diabetes in East Asian and European populations. Nature genetics 40: 1098–1102.
- 11. Liu Y, Zhou DZ, Zhang D, Chen Z, Zhao T, et al. (2009) Variants in KCNQ1 are associated with susceptibility to type 2 diabetes in the population of mainland China. Diabetologia 52: 1315–1321.
- 12. Hu C, Wang C, Zhang R, Ma X, Wang J, et al. (2009) Variations in KCNQ1 are associated with type 2 diabetes and beta cell function in a Chinese population. Diabetologia 52: 1322–1325.
- 13. Tan JT, Nurbaya S, Gardner D, Ye S, Tai ES, et al. (2009) Genetic Variation in KCNQ1 Associates With Fasting Glucose and β-Cell Function. Diabetes 58: 1445.
- 14. Holmkvist J, Banasik K, Andersen G, Unoki H, Jensen TS, et al. (2009) The type 2 diabetes associated minor allele of rs2237895 KCNQ1 associates with reduced insulin release following an oral glucose load. PLoS One 4: e5872.
- 15. Jonsson A, Isomaa B, Tuomi T, Taneera J, Salehi A, et al. (2009) A variant in the KCNQ1 gene predicts future type 2 diabetes and mediates impaired insulin secretion. Diabetes 58: 2409.
- 16. Qi Q, Li H, Loos RJF, Liu C, Wu Y, et al. (2009) Common variants in KCNQ1 are associated with type 2 diabetes and impaired fasting glucose in a Chinese Han population. Human molecular genetics 18: 3508.
- 17. Aydin A, B Hring S, Dahm S, Guenther UP, Uhlmann R, et al. (2005) Single nucleotide polymorphism map of five long-QT genes. Journal of molecular medicine 83: 159–165.
- 18. Zhang Y, Chang B, Hu S, Wang D, Fang Q, et al. (2008) Single Nucleotide Polymorphisms and Haplotype of Four Genes Encoding Cardiac Ion Channels in Chinese and their Association with Arrhythmia. Annals of Noninvasive Electrocardiology 13: 180–190.
- 19. Westenskow P, Splawski I, Timothy KW, Keating MT, Sanguinetti MC (2004) Compound mutations: a common cause of severe long-QT syndrome. Circulation 109: 1834.
- 20. Yang P, Kanki H, Drolet B, Yang T, Wei J, et al. (2002) Allelic variants in long-QT disease genes in patients with drug-associated torsades de pointes. Circulation 105: 1943–1948.
- 21. Richards JB, Rivadeneira F, Inouye M, Pastinen TM, Soranzo N, et al. (2008) Bone mineral density, osteoporosis, and osteoporotic fractures: a genome-wide association study. The Lancet 371: 1505–1512.
- 22. Chen Z, Zhang X, Ma G, Qian Q, Yao Y (2010) Association study of four variants in KCNQ1 with type 2 diabetes mellitus and premature coronary artery disease in a Chinese population. Molecular biology reports 37: 207–212.
- 23. Chen Z, Yin Q, Ma G, Qian Q (2010) KCNQ1 gene polymorphisms are associated with lipid parameters in a Chinese Han population. Cardiovascular Diabetology 9: 35.
- 24. Carlini DB, Chen Y, Stephan W (2001) The relationship between third-codon position nucleotide content, codon bias, mRNA secondary structure and gene expression in the drosophilid alcohol dehydrogenase genes Adh and Adhr. Genetics 159: 623.
- 25. Schmitt N, Schwarz M, Peretz A, Abitbol I, Attali B, et al. (2000) A recessive C-terminal Jervell and Lange-Nielsen mutation of the KCNQ1 channel impairs subunit assembly. The EMBO journal 19: 332–340.
- 26. Liang L, Du ZD, Cai LL, Wu JX, Zheng T, et al. (2003) A novel KCNQ1 mutation in Chinese with congenital long QT syndrome]. Zhonghua er ke za zhi. Chinese journal of pediatrics 41: 724.
- 27. Sharma D, Glatter KA, Timofeyev V, Tuteja D, Zhang Z, et al. (2004) Characterization of a KCNQ1/KVLQT1 polymorphism in Asian families with LQT2: implications for genetic testing. Journal of molecular and cellular cardiology 37: 79–89.
- 28. Kathiresan S, Willer CJ, Peloso GM, Demissie S, Musunuru K, et al. (2008) Common variants at 30 loci contribute to polygenic dyslipidemia. Nature genetics 41: 56–65.
- 29. Teslovich TM, Musunuru K, Smith AV, Edmondson AC, Stylianou IM, et al. (2010) Biological, clinical and population relevance of 95 loci for blood lipids. Nature 466: 707–713.
- 30. Chavez JA, Summers SA (2010) Lipid oversupply, selective insulin resistance, and lipotoxicity: molecular mechanisms. Biochimica et Biophysica Acta (BBA)-Molecular and Cell Biology of Lipids 1801: 252–265.
- 31. Wang X, Lu M, Qian J, Yang Y, Li S, et al. (2009) Rationales, design and recruitment of the Taizhou Longitudinal Study. BMC Public Health 9: 223.
- 32. Johnson GCL, Esposito L, Barratt BJ, Smith AN, Heward J, et al. (2001) Haplotype tagging for the identification of common disease genes. Nature genetics 29: 233–237.