Journal of Human Hypertension (2012) 26, 553–561 & 2012 Macmillan Publishers Limited All rights reserved 0950-9240/12 www.nature.com/jhh ORIGINAL ARTICLE Association of intronic single-nucleotide polymorphisms in the EMILIN1 with essential hypertension in a Chinese population

VMS Oh1, B-M Chua1, C-K Heng2, S-B Yeo1, O-S Yim2 and EPH Yap3 1Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore; 2Department of Paediatrics, National University of Singapore, Singapore and 3Defence Medical & Environmental Research Institute, DSO National Labs, Singapore

Studies in mice suggest that the microfibril types with EH. Three SNPs in introns 1 and 5 (rs2289360, interfacer-1 gene (EMILIN1), the gene encoding elastin rs2011616 and rs7424556) were in strong pair-wise microfibril interfacer-1 , contributes to the patho- linkage disequilibrium (r 240.89). All three SNPs were genesis of essential hypertension (EH) in humans. significantly associated with hypertension. Genotypic EMILIN1 in part maintains elastic fibres in vessel walls, frequencies at the three SNPs differed significantly and hence peripheral arterial compliance. In a case– between cases and only those controls without FHH. control study, we assessed 942 non-obese non-diabetic Healthy controls with FHH should be excluded to Chinese, comprising 467 patients with EH and 475 increase the odds of detecting association. All the G normotensive control subjects (166 without, and 309 alleles of rs2289360 (odds ratio ¼ 1.69, P ¼ 0.010), with, family history of hypertension in first-degree rs2011616 (odds ratio ¼ 1.52, P ¼ 0.038) and rs7424556 relatives (FHH)). Hypertension in first-degree relatives (odds ratio ¼ 1.59, P ¼ 0.023) were high-risk alleles in the occurred in 88%, 65% and 0% of cases, all controls and recessive genetic model. We observed significant over- controls without FHH, respectively. We scanned for all haplotypic association with EH (empirical P ¼ 0.0072); single-nucleotide polymorphisms (SNPs) and geno- GGG is a risk haplotype (P ¼ 0.043). The overall results typed them in the EMILIN1 gene using high-resolution support EMILIN1 as a candidate gene for human EH. melt-curve analysis. No exonic variants were detected. Journal of Human Hypertension (2012) 26, 553–561; We assessed the association of SNPs and their haplo- doi:10.1038/jhh.2011.68; published online 14 July 2011

Keywords: blood pressure; elastin microfibril interfacer 1; genetic polymorphism; haplotype; essential hypertension; Chinese

Introduction elastic of blood vessels.4 The association of EH with elastin haplo-insuffi- Essential hypertension (EH), which affects about 1 ciency suggests that elastin defects in the 25% of adults aged 30-69 years in Singapore, has a elastic extracellular matrix are causative factors for multifactorial origin from an interaction of suscept- hypertension.5 ibility with environmental influences. Many The elastin microfibril interfacer-1 gene (EMILIN1), allelic risk factors have been validated in various located on 2p23.3-p23.2 and contain- populations, showing that the genetic risk is ing eight exons spanning about 8.0 kb, regulates heterogeneous. A reduction in vascular compliance elastogenesis.6 Its expressed protein stabilises elas- might contribute to the raised arterial blood pressure tin and related microfibrils. EMILIN1-deficient mice (BP) in EH. Studies of the pathogenesis of EH have show abnormal anchorage of endothelial and therefore examined defects in vascular smooth 7 2,3 smooth muscle cells to elastic lamellae in the aorta, muscle and endothelial cells, and also in the decreased vascular compliance and arterial hyper- tension.8 The structure of the elastic extracellular Correspondence: Professor VMS Oh, Department of Medicine, matrix therefore affects vessel compliance and the Yong Loo Lin School of Medicine, National University of BP. The roles of EMILIN1 protein are mediated Singapore, Room W10-55, Tower Block, 1E Kent Ridge Road, through the inhibition of transforming growth Singapore 119228, Singapore. 9–11 E-mail: [email protected] factor-b; the circulating levels of transforming Received 15 September 2010; revised 31 May 2011; accepted 13 growth factor-b1 mRNA and transforming growth June 2011; published online 14 July 2011 factor-b1 protein are higher in hypertensive than in Association of EMILIN1 gene with hypertension VMS Oh et al 554 normotensive persons.12 The inactivation of a least 5 min, recorded using the Omron HEM-705 CP single allele of transforming growth factor-b1 digital BP monitor (Omron Corp., Tokyo, Japan),16 gene rescues EMILIN1À/À animals from high blood matching the protocol for the British Genetics pressure.8,13,14 of Hypertension study.18 The digital BP monitor However, EMILIN1 has not been widely studied in had been validated by the British Hypertension human hypertension. A recent case–control study Society.18 suggested that excluding control subjects with We excluded individuals with body mass index familial EH might improve the detection of associa- X30; those who had received blood transfusion or tion of gene variants with EH.15 We therefore aimed organ transplantation; subjects with anaemia, to screen EMILIN1 for coding and non-coding single- chronic kidney disease, diabetes mellitus or cancer; nucleotide polymorphisms (SNPs), comparing un- and subjects who had experienced acute coronary or related hypertensive cases against normotensive cerebrovascular events in the preceding 6 months.16 control subjects both with and without familial Clinical chemistry measurements were made on hypertension. fasting venous blood. Genomic DNA was extracted We found significant genotypic association be- from the whole blood using the standard phenol- tween three intronic SNPs of human EMILIN1 and chloroform method.19 the hypertensive phenotype, and provide evidence of linkage disequilibrium between EMILIN1 SNPs Variant screening and genotyping and hypertension. Eight putative SNPs of the EMILIN1 gene listed in dbSNP (Build 125, April 2006) (http://www. ncbi.nlm.nih) were screened for, using high-resolu- Methods tion double-stranded DNA (dsDNA) melt-curve Study population analysis of introns 1 and 5, and exons 3 and 4. The subjects were consecutively recruited in the Amplicons from PCR were mixed with a 10 Â National University of Singapore Investigation of dsDNA binding saturating dye and LCGreen Plus, Genetics of Hypertension study,16 which was ap- and scanned for sequence variation on a Light- proved by the Ethics Committee (now the National Scanner (both from Idaho Technology, Salt Lake Healthcare Group Institutional Review Board) of the City, UT, USA). Melting of dsDNA was detected by a National University Hospital, Singapore. Informed reduction in fluorescence when the sample was and written consent was obtained from each gradually heated. participant. We studied altogether 942 Chinese Genotyping of the polymorphic variants was carried out using the melt-curve analysis of a single persons, consisting of 467 non-related hypertensive 0 patients, 309 normotensive controls with a family 3 -blocked unlabelled gene-specific oligonucleotide 20,21 history of hypertension in first-degree relatives probe. Primer and probe sequences are given in (FHH) and 166 normotensive controls without FHH. Supplementary Table 1 on the journal’s website. The 467 hypertensive cases comprised 442 pa- Asymmetrical PCR was completed and the single- tients with definite hypertension patients and 25 stranded DNA (ssDNA) amplicon-probe duplex was with borderline hypertension, as defined below. melted in the presence of LCGreen Plus dye to detect changes in fluorescence. Both PCR-restriction frag- ment length polymorphism and direct sequencing on randomly selected samples verified the geno- Protocols and measurements typed results. Patients were accepted as ethnic Chinese if all four of their grandparents were registered as ‘of Chinese race’ and spoke a Chinese dialect. We attempted Statistical analysis to recruit normotensive controls aged above A sample size calculation had shown that, for a SNP 55 years to limit the number who might later show with an allele frequency of 0.25 to attain a statistical hypertension. significance level of Po0.05, we needed 436 sub- Definite hypertension was defined as systolic jects in each group if the allele’s effect (odds ratio) and diastolic blood pressures X140 mm Hg and was 1.5.22 We determined allele and genotype 90 mm Hg, respectively; or mean 24-h ambulatory frequencies by counting, and assessed Hardy-Wein- BP X135/85 mm Hg.16 The criteria corresponded to berg equilibrium in each study group by means the threshold BP values for the topmost 5% of the of Pearson’s goodness-of-fit w2-test. We compared BP distribution, in a survey of 4723 persons biometric variables between cases and controls randomly recruited from the Singapore popula- using the w2-test for gender, smoking and alcohol tion.17 Borderline hypertension was defined as intake, and one-way analysis of variance for the either systolic BP of 135–139 mm Hg or diastolic other continuous variables (SPSS for Windows, BP of 85–89 mm Hg. Normotension was defined as a Rel.16.0.1.2007, SPSS Inc., Chicago, IL, USA). BP of 134/84 mm Hg or lower. As a preliminary evaluation had shown that the Systolic and diastolic BP values were taken as the 442 ‘definite hypertension’ and 25 ‘borderline mean of three measurements made at intervals of at hypertension’ patients were indistinguishable, they

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 555 were analyzed en bloc as hypertensive cases. Thus, all the haplotypes for their overall association with we compared the differences in genotypic and hypertension using the Zclump statistical test (Medical allelic frequency between the controls and all 467 statistics and Bioinformatics- Leiden University Med- cases using the w2-test. The pattern of pair-wise ical Center, http://www.msbi.nl/dnn/Genetics/Tests linkage disequilibrium between the markers was forgeneticassociation/tabid/147/Default.aspx).25 expressed in terms of r 2. Applying the case–control design and the reces- Results sive genetic model, the genetic association of each SNP with hypertension was individually tested by Subject characteristics logistic regression with the following as covariates: The demographic and clinical characteristics of gender, age, body mass index, serum fasting glucose, all the 942 individuals (467 cases and 475 controls) serum high density lipoprotein cholesterol and are shown in Table 1. The mean BP of the 467 serum triglycerides. For logistic regression we used hypertensive patients was 159/99 mm Hg, and that a web-based SNP analysis tool (Biostatistics and of the 475 normotensive control subjects and the 166 Bioinformatics Unit, Catalan Institute of Oncology, control subjects without FHH was 114/74 mm Hg http://www.bioinfo.iconcologia.net/index.php?mod (Table 1, Supplementary Table 2). The proportion of ule ¼ snpstats).23 Using SPSS for Windows, we also subjects with FHH was 88.1% in the 467 cases, evaluated the effect of EMILIN1 gene polymorphism 65.1% in all 475 control subjects and 0% in the 166 on the BP of all the subjects by analysis of controls without FHH. On average, the hypertensive covariance, with adjustment for the above six subjects had significantly higher body mass index covariates plus smoking and drinking. A P-value of and mean on-treatment BP than the control subjects o0.05 was considered statistically significant. taken together. Haplotypes, consisting of rs2289360, rs2011616 On average, the cases were slightly older than the and rs742455, were developed using the software controls (Table 1), despite attempts to enrol older programme PHASE version 2.1 (M Stephens, NJ controls. However, mean age was similar in the Smith, P Donnelly, U of Washington, Seattle, WA, cases and the controls without FHH (49.4 and 49.1 USA).24 This software uses an algorithm to estimate years, respectively (Supplementary Table 2)). Aver- haplotype frequencies and obtain the best haplotype age age at the onset of hypertension was 42.7±11.3 configuration of each multi-locus genotype. Haplo- years in the cases. No significant demographic type distributions were compared between hyperten- differences were observed between the two groups sive cases and the controls by the w2-test. We tested of controls, except that the controls with FHH were

Table 1 Demographics characteristics of 942 Singaporean Chinese persons, consisting of 467 hypertensive cases and 475 normotensive control subjects

Variables Cases Controls Cases vs controls (n ¼ 467) (n ¼ 475) P

Gender (men/women) 268/199 157/318 o0.001 Age (years) when studied (years) 49.4±11.9 47.3±8.7 0.02 Age at time of diagnosis (years) 42.7±11.3 — — Height (m) 1.63±0.08 1.60±0.08 o0.001 Weight (kg) 67.2±12.2 60.1±11.1 o0.001 BMI (kg mÀ2) 25.3±3.3 23.3±3.1 o0.001 Systolic blood pressure (mm Hg) 159.4±16.0a 114.2±15.4b o0.001 Diastolic blood pressure (mm Hg) 99.2±9.1a 74.8±9.9b o0.001 Mean arterial pressure (mm Hg) 119.2±9.7a 88.0±11.4 o0.001 Fasting serum glucose (mmol lÀ1) 5.56±1.21 5.05±0.76 o0.001 Fasting serum total cholesterol (mmol lÀ1) 5.55±0.93 5.46±0.89 0.169 Fasting serum LDL-C (mmol lÀ1) 3.38±0.90 3.41±0.79 0.498 Fasting serum HDL-C (mmol lÀ1) 1.32±0.32 1.52±0.41 o0.001 Fasting triglycerides (mmol lÀ1) 1.90±1.18 1.22±0.74 o0.001 Current smoker (yes:no)c 45:422 42:433 0.736 Alcohol intake (yes:no)d 53:414 54:421 0.267

Abbreviations: BMI, body mass index; HDL-C, high-density lipoprotein-cholesterol; LDL-C, low-density lipoprotein-cholesterol. P-values were calculated using the w2-test for gender, smoking and alcohol intake, and 1-way analysis of variance for the other variables; both tests within SPSS for Windows. Values are given as mean±s.d. aThe systolic and diastolic readings denote the untreated blood pressure (BP) at time of diagnosis. Mean arterial pressure was derived from the untreated BP. bThe systolic and diastolic readings denote the mean sitting BP on entry into study. cCurrent smoking included persons who stopped smoking within previous 3 months. dYes, if regularly taking alcoholic drink within previous 3 months.

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 556 slightly younger than those without FHH (46.3±8.1 Genotyping of intronic SNPs vs 49.1±9.5 years, P ¼ 0.01) (Supplementary Table As the generic dsDNA melt-profiling method de- 2). None of the 942 unrelated individuals had fasting scribed above could not reliably distinguish be- serum glucose X7.0 mmol lÀ1. The cases had slightly tween the three genotypes of a SNP, another type of higher serum glucose than the controls, and slightly melt-profiling assay was used to perform rapid and lower serum HDL-cholesterol and higher triglycer- definitive genotyping for the three intronic SNPs. ides (Table 1). The rates of smoking and alcohol High-resolution melt-curve analysis of asymmetri- intake were similar in the cases and the controls cally amplified ssDNA with an unlabelled allele- (Table 1). specific oligoprobe reliably distinguished between the three genotypes, and automated genotype calling occurred consistently. Pearson’s goodness-of-fit w2-test showed that the Screening for genetic variants three intronic sites for cases and controls did Using high-resolution dsDNA melt-curve analysis of not deviate significantly from Hardy-Weinberg ex- introns 1 and 5, and exons 3 and 4, we screened pectations. eight putative SNPs of the EMILIN1 gene then listed Genotypic and allelic frequencies of the three in dbSNP (Build 125, April 2006). Out of these SNPs in the cases and controls are shown in Table 2. eight SNPs, only three intronic SNPs, rs2289360, The three intronic SNPs were in strong pair-wise rs2011616 and rs742455, were observed and vali- linkage disequilibrium with one another (r240.89, dated to be polymorphic in our Singaporean Figure 1). We observed significant differences in Chinese population. The three SNPs yielded differ- genotypic frequency at rs2289360 and rs7424556, ent melt-curve profiles compared with reference but not at rs2011616, between hypertensive patients sequenced samples, and were confirmed by PCR- and the 166 control subjects who had no FHH restriction fragment length polymorphism and di- (Table 2). However, when 309 control subjects who rect sequencing. had FHH were included in the comparison with

Table 2 Genotype and allele frequencies of 3 intronic SNPs in the EMILIN1 gene

SNP Frequency P

Cases Controls without All controls Cases vs controls Cases vs (n ¼ 467) FHH (n ¼ 166) (n ¼ 475) without FHH all controls

rs2289360 Genotype AA 0.0471 0.0361 0.0400 0.015 0.071 AG 0.3041 0.4277 0.3747 GG 0.6488 0.5361 0.5853 Allele A 0.1991 0.2500 0.2274 0.052 0.135 G 0.8009 0.7500 0.7726

rs2011616 Genotype AA 0.0538 0.0485 0.0464 0.091 0.19 AG 0.3011 0.3939 0.3565 GG 0.6452 0.5576 0.5970 Allele A 0.2043 0.2455 0.2247 0.118 0.282 G 0.7957 0.7545 0.7753

rs7424556 Genotype CC 0.0558 0.0482 0.0421 0.042 0.122 CG 0.3026 0.4096 0.3621 GG 0.6416 0.5422 0.5958 Allele C 0.2071 0.2530 0.2232 0.082 0.396 G 0.7929 0.7470 0.7768

Abbreviations: EMILIN1, elastin microfibril interfacer-1 gene; FHH, family history of hypertension in first-degree relatives; SNPs, single- nucleotide polymorphisms. In both the cases and the controls, the genotype distribution of the three polymorphisms was in Hardy-Weinberg equilibrium, which we assessed in each study group using Pearson’s goodness-of-fit w2-test. P-values were computed using the w2-test within SPSS for Windows, Rel.16.0.1.2007: SPSS Inc., Chicago, IL, USA.

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 557 2 2 r cases = 0.9478; r controls= 0.8892

rs2289360 rs7424556

Exon 1 Exon 2 Exon 3 Exon 4 Exon 5

Intron 1 Intron 2 Intron 3 Intron 4 Intron 5

rs2011616

2 2 r cases = 0.9437; r controls= 0.8947

2 r cases = 0.9363; 2 r controls= 0.914

Figure 1 Location and linkage disequilibrium (r 2) of three SNPs, rs22289360, rs2011616 and rs7424556, of the EMILIN1 gene on chromosome 2p23.3–p23.2. patients, the differences in frequency of both quency for the risk haplotype GGG (P ¼ 0.043). SNPs were no longer significant (Table 2). In regard Another haplotype, AGG, was present in the healthy to genotype frequency, the controls with FHH control population but not in the affected cases were not significantly different from the controls (Table 5). without FHH, or from the cases (Supplementary Using the Zclump statistic, we found a significant Table 3). overall association between the haplotypes and In the recessive genetic model, significant asso- hypertension (empirical P ¼ 0.0072). ciations between the cases and controls without FHH were detected at rs2289360, rs2011616 and rs7424556 (P ¼ 0.010, P ¼ 0.038 and P ¼ 0.023, re- Discussion spectively) (Table 3). A significant association between the cases and all the controls was found We detected a significant association between EH at rs2289360 only (P ¼ 0.049). The G alleles of and three intronic SNPs of EMILIN1 (rs2289360, rs2289360, rs2011616 and rs7424556 were the rs2011616 and rs7424556) in an association study high-risk alleles for EH (Table 3). in 942 non-diabetic ethnic Chinese patients and We made an analysis of covariance, adjusting for controls. Our study thus implicates EMILIN1 as a eight possible confounders, on the strongly asso- candidate gene for human EH. Though intronic ciated marker rs2289360 to investigate the effect of SNPs do not directly alter cellular function, they the genotypes on BP (Table 4). There were signifi- might predispose patients to EH or influence their cant differences in systolic BP and mean arterial response to drug treatment. pressure (adjusted P ¼ 0.048 and 0.042, respectively) The observation that FHH occurred more often in between the three genotypes for rs2289360 in the the 467 EH cases (88.1%) than in 475 normotensive pooled sample (n ¼ 633) of cases and controls with- control subjects (65.1%) is expected. We found a out FHH. However, the absolute differences in BP significant difference in the GG genotype frequency between genotypes were small, and therefore un- of each of the three intronic SNPs between the 467 likely to be physiologically important. We found no cases and the 166 normotensive controls without association of systolic and diastolic pressures, and FHH (Table 3). Although the controls without FHH mean arterial pressure, with both rs7424556 and were relatively few, this result probably arose rs2011616 (data not shown). because phenotypic extremes were compared, namely selected controls (FHH in 0%) versus EH cases (FHH in 85.1%). The difference vanished Haplotype analysis when the 309 controls with FHH were included in Eight haplotypes were estimated from the three the comparison (Table 3, Supplementary Table 3), SNPs in the pooled sample of cases and controls just as Shimodaira et al. had found.15 Our observa- without FHH. More than 90% of individuals had tion thus confirms that excluding the controls with two common haplotypes (GGG or AAC). We ob- FHH increases the likelihood of detecting genetic served a significant difference in haplotype fre- differences.

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 558 Table 3 Recessive genetic model of three intronic SNPs in the EMILIN1 gene in 942 subjects, comprising 467 hypertensive cases and 475 normotensive control subjects with and without familial hypertension

SNP Genotype Frequency Cases (n ¼ 467) Cases (n ¼ 467) Cases (n ¼ 467) vs controls without vs controls with vs all controls FHH (n ¼ 166) FHH (n ¼ 309) (n ¼ 475)

rs2289360 GG 0.6488 Odds ratio 1.69 Odds ratio 1.17 Odds ratio 1.35 AG-AA 0.3512 95% CI 1.13-2.50 95% CI 0.83-1.64 95% CI 1.00-1.81 P ¼ 0.010 P ¼ 0.366 P ¼ 0.049 rs2011616 GG 0.6452 Odds ratio 1.52 Odds ratio 1.09 Odds ratio 1.23 AG-AA 0.3548 95% CI 1.02-2.27 95% CI 0.78-1.54 95% CI 0.92-1.66 P ¼ 0.038 P ¼ 0.606 P ¼ 0.167 rs7424556 GG 0.6416 Odds ratio 1.59 Odds ratio 1.08 Odds ratio 1.24 CG-CC 0.3584 95% CI 1.07-2.36 95% CI 0.77-1.51 95% CI 0.92-1.67 P ¼ 0.023 P ¼ 0.675 P ¼ 0.152

Abbreviations: EMILIN1, elastin microfibril interfacer1 gene; FHH, family history of hypertension in first-degree relatives; 95% CI, 95% confidence interval. In both the cases and the controls, the genotype distribution of the three polymorphisms was in Hardy-Weinberg equilibrium, which we assessed in each study group by means of Pearson’s goodness-of-fit w2-test. In every instance, the recessive genetic model was used to compare common homozygotes with the pooled heterozygotes and rare homozygotes as the reference group. P-values were calculated by logistic regression, using a web-based SNP analysis tool,23 with gender, age, BMI, fasting serum glucose, serum HDL-cholesterol, and serum triglycerides as covariates.

Table 4 Association between blood pressure and rs2289360 genotypes in 633 Singaporean Chinese subjects, comprising 467 hypertensive cases and 166 normotensive controls without familial hypertension

Blood pressure (BP) Genotype 633 Subjects (cases, n ¼ 467; controls without FHH, n ¼ 166) P

Frequency Mean Standard deviation (mm Hg) (mm Hg)

Systolic BP GG 0.6193 149.6 24.9 0.016 (0.048)a AG 0.3365 143.4 26.1 AA 0.0442 148.9 27.5 Diastolic BP GG 0.6193 93.6 14.3 0.048 (0.058)a AG 0.3365 90.6 14.6 AA 0.0442 93.5 14.6 Mean arterial pressure GG 0.6193 112.3 16.9 1 2 a (3systolic BP+3diastolic BP) AG 0.3365 108.2 17.7 0.021 (0.042) AA 0.0442 111.9 18.2

Abbreviations: BP, blood pressure; FHH, family history of hypertension in first-degree relatives. P-values were computed by analysis of covariance using SPSS for Windows, Rel.16.0.1.2007: SPSS Inc., Chicago, IL, USA. aAdjusted for gender, age, body mass index, fasting serum glucose, serum HDL-cholesterol, serum triglycerides, smoking habit, and drinking habit. Smoking habit included persons who stopped smoking within previous 3 months; drinking indicated regularly taking alcoholic drink within previous 3 months (Table 1).

Secondly, normotensive persons with FHH might diabetes and dyslipidaemia, and to increase the be non-penetrant carriers of high-risk alleles. In likelihood of detecting SNPs predisposing to hyper- physiological terms, healthy persons with hyperten- tension. Significant but small differences in fasting sive parents might have stiffer arteries26 and higher serum glucose, high density lipoprotein cholesterol 24-h ambulatory blood pressure27 than those with- and triglycerides occurred between the cases and out FHH. Lastly, a few controls with FHH might later controls (Table 1), because hypertensive persons develop hypertension. The evidence therefore ad- tend to be heavier and to have lower serum high vocates enroling elderly normotensive controls density lipoprotein cholesterol and higher triglycer- without FHH in any case–control association study ides.1,16 The low rates of smoking and alcohol intake of hypertension. (Table 1) are consistent with those in the general The present study carefully phenotyped non- population1,17 and do not represent subject selection obese Chinese persons, whose fasting serum glucose bias. was o7.0 mmol lÀ1. The subjects therefore represent All of the three intronic SNPs identified in the a sample of race-matched, non-obese and non- present study (rs2289360, rs2011616 and rs7424556) diabetic cases, and controls. The intention of the are significantly associated with EH, only when the subject selection criteria, which preferred lean and cases are compared with the controls without FHH euglycaemic subjects,16 was both to decrease the (Table 3). Assessing a larger number of controls effect of genes predisposing persons to obesity, without FHH might have yielded larger odds ratios.

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 559 Table 5 Haplotype frequencies of EMILIN1 in 633 Singaporean We did not identify one SNP that alone contri- Chinese subjects, comprising 467 hypertensive patients and 166 butes significantly to the global association of the normotensive controls without familial hypertension haplotypes of EMILIN1 (Table 5). Shimodaira et al.15 Haplotypesa Frequency P found EH linked to the GGT haplotype, just as the present study observed EH association with Cases Controls without the common haplotype GGG (Table 5). However, (n ¼ 467) FHH (n ¼ 166) the two results are not comparable, because only rs2011616 was common to both studies. Shen et al.28 GGG 0.7835 0.7078 0.043 did not report a haplotype analysis. AAC 0.1873 0.2158 0.414 After evaluating all the above information, we GAC 0.0098 0.0093 0.883 believe that the positive associations of the AGC 0.0075 0.0075 0.754 AAG 0.0045 0.0064 0.776 three intronic SNPs of EMILIN1 with EH observed GGC 0.0043 0.0265 0.023 in the present study are valid and deserve further GAG 0.0030 0.0094 0.109 investigation. AGG 0.0000 0.0173 0.004 In recent genome-wide association studies of hypertension using DNA microarray technology, Zclump empirical 0.0072 P-value the chips (for example, Affymetrix array 5.0, Affymetrix 6.0 and Illumina 1M29–31) contained up Abbreviations: EMILIN1, elastin microfibril interfacer-1 gene; FHH, to 1.2 million SNPs. The Illumina 1 M chip included family history of hypertension in first-degree relatives. the EMILIN1 SNPs rs2289360 and rs2011616, but aHaplotypes consisting of rs2289360, rs2011616,andrs7424556. Haplotype distributions were compared between cases and controls not rs7424556. No ‘hits’ were reported in the three by the w2-test. Overall association of all the haplotypes with genome-wide association studies for the association hypertension was calculated using the Zclump statistics test at one 25 of hypertension with rs2289360 and rs2011616,or degree of freedom. with 30 other SNPs within 1000 bp of EMILIN1.32–34 The modest associations that we found in the present study might not be sufficiently strong to be When cases are compared with all the controls, only detected on a genome-wide scale. rs2289360 is associated with EH (Table 3), suggest- In conclusion, our association study in EH cases ing that EH risk alleles in some controls with FHH and normotensive controls in a Singaporean Chi- decreased the odds ratios for the other SNPs. nese population detected a significant association Similarly, for the quantitative BP traits, only the between hypertension and three intronic SNPs, SNP rs2289360 shows a significant positive associa- rs2289360, rs2011616 and rs7424556,ofEMILIN1. tion with systolic BP and mean arterial pressure in In case–control comparisons, selecting control sub- the pooled sample of cases and controls without jects without FHH increases the likelihood of FHH (Table 4). detecting association with hypertension. The work The SNP rs2289360 was omitted in a Japanese affirms EMILIN1 as a candidate gene for EH. To study by Shimodaira et al.15 because of deviation elucidate the role of EMILIN1 in the pathophysiol- from the Hardy-Weinberg equilibrium. Neither a ogy of hypertension, the gene should first be study by Shen et al. in northern Han Chinese28 nor a evaluated across the major ethnic groups both study by Shimodaira et al.15 examined rs7424556. within and outside Asia. We identified rs7424556 after screening for genetic variants. Shen et al.28 and Shimodaira et al.15 might have missed rs7424556, because they selected their What is known about this topic SNPs from public databases, such as those of the K Studies in animals suggest that EMILIN1, the gene International HapMap Project and the NCBI. encoding elastin microfibril interfacer-1 protein, contributes to the pathogenesis of hypertension, but The study by Shen et al. showed that northern evidence that EMILIN1 has a causative role in human Han Chinese aged 460 years with the A allele (AG/ hypertension is scanty. AA genotype) of rs2011616 had increased risk for K Only two other studies in humans have reported hypertension, but persons aged o50 years had a evidence for association of SNPs in EMILIN1 with decreased risk.28 This discrepant allelic effect on the hypertension. BP is difficult to interpret. The present study in What this study adds Singaporean Chinese subjects (mean age below 50 K In this case–control study in Chinese persons, we found years) observed a similar association of EH with three polymorphisms in introns 1 and 5 of EMILIN1 that rs2011616 (Table 3) as found in northern Han were in strong pair-wise linkage disequilibrium; three 28 SNPs, rs2289360, rs201161 and rs7424556, which were Chinese subjects aged o50 years. Similarly, the associated with EH; and a significant association between Japanese study by Shimodaira et al. in hypertensive the common risk haplotype GGG and EH. patients and controls without FHH showed a K The evidence supports EMILIN1 as a candidate gene for genotypic association of EH with the G allele (AG/ essential hypertension that requires further examination. 15 K To increase test sensitivity in case–control studies, GG genotype) of rs2011616. The present study also investigators should recruit elderly controls without detected an association of the G allele of rs2011616 hypertensive first-degree relatives. with hypertension.

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 560 Conflict of interest 10 Li B, Khanna A, Sharma V, Singh T, Suthanthiran M, August P. TGF-beta-1 DNA polymorphisms, The authors declare no conflict of interest. protein levels, and blood pressure. Hypertens 1999; 33: 271–275. 11 Cambien F, Ricard S, Troesch A, Mallet C, Generenaz L, Evans A et al. Polymorphism of the transforming Acknowledgements growth factor-beta-1 gene in relation to myocardial infarction and blood pressure: The Etude Cas-Temoin We thank Linda Gan and Rita Yong for their help in de l’Infarctur du Myocarde (ECTIM) Study. Hypertens the Molecular Genetics Laboratory of the Defence 1996; 8: 881–887. Medical & Environmental Research Institute, DSO 12 Suthanthiran M, Li B, Song JO, Ding R, Sharma VK, Labs, Singapore and Bernadette GC Er for expert Schwartz JE et al. Transforming growth factor-beta-1 help in subject recruitment. We also thank Nicole PS hyperexpression in African-American hypertensives: a Choa for help in manuscript preparation. The study novel mediator of hypertension and/or target organ damage. Proc Natl Acad Sci 2000; 97: 3479–3484. was mainly supported by the National Medical 13 August P, Suthanthiran M. Transforming growth factor Research Council, Singapore (grant no. NMRC/ beta signaling, vascular remodeling, and hypertension. 1067/2006) (BMC, BGCE). Other funding sources N Engl J Med 2006; 54: 2721–2723. were Yong Loo Lin School of Medicine, National 14 Raman M, Cobb MH. TGF-beta regulation by Emilin1: University of Singapore (VMSO), Shaw Foundation new links in the etiology of hypertension. Cell 2006; and Lee Foundation (BMC, BGCE), Singapore Heart 124: 893–895. Foundation (grant no. R-172-000-143-593) (BMC, 15 Shimodaira M, Nakayama T, Sato N, Naganuma T, BGCE), National Healthcare Group (R-172-000-146- Yamaguchi M, Aoi N et al. Association study of the 490) (BMC, BGCE), the National University Hospital elastin microfibril interface 1 (Emilin1) gene in Patient Care Research Fund (grant no. R-172-000- essential hypertension. Am J Hypertens 2010; 22: 547–555. 139-592) (BMC, BGCE) and the Defence Science & 16 Lee YW, Oh VMS, Garcia E, Taylor EA, Wu H, Yap EP Technology Agency, Singapore (EPHY). et al. Haplotypes of the beta2-adrenergic receptor gene are associated with essential hypertension in a Singaporean Chinese population. JHypertens2004; 22: 2111–2116. References 17 Epidemiology and Disease Control Department, Min- istry of Health, Singapore. National Health Survey 1 Epidemiology and Disease Control Department, Min- 1998. http://www.moh.gov.sg/mohcorp/uploadedfiles/ istry of Health, Singapore. National Health Survey Publications/Reports/1999/nhs.pdf. (accessed 4 March 2004. http://www.moh.gov.sg/mohcorp/uploadedfiles/ 2009). Publications/Reports/2005/NHS_2004(Part3).pdf (ac- 18 Caulfield M, Munroe P, Pembroke J, Samani N, cessed 4 March 2009). Dominiczak A, Brown M et al. MRC British Genetics 2 Chobanian AV, Prescott MF, Haudenschild CC. Recent of Hypertension Study. Genome-wide mapping of advances in molecular pathology. The effects of human loci for primary hypertension. Lancet 2003; hypertension on the arterial wall. Exp Mol Pathol 9375: 2118–2123. 1984; 41: 153–169. 19 Chomczynski P, Sacchi N. The single-step method of 3 Shepherd JT. Franz Volhard lecture. Increased sys- RNA isolation by acid guanidinium thiocyanate– temic vascular resistance and primary hypertension: phenol–chloroform extraction: twenty-something the expanding complexity. J Hypertens Suppl 1990; 8: years on. Nat Protoc 2006; 1: 581–585. S15–S27. 20 Zhou L, Mayers AN, Vandersteen JG, Wang L, Wittwer 4 Arribas SM, Hinek A, Gonza´lez MC. Elastic fibers and CT. Closed-tube genotyping with unlabeled oligonu- vascular structure in hypertension. Pharmacol Ther cleotide probes and a saturating DNA dye. Clin Chem 2006; 111: 771–791. 2004; 50: 1296–1298. 5 Faury G, Pezet M, Knutsen RH, Boyle WA, Heximer SP, 21 Bennett CD, Campbell MN, Cook CJ, Eyre DJ, Nay LM, McLean SE et al. Developmental adaptation of the Nielsen DR et al. The LightTyper: high-throughput mouse cardiovascular system to elastin haploinsuffi- genotyping using fluorescent melting-curve analysis. ciency. J Clin Invest 2003; 112: 1419–1428. Biotechniques 2003; 34: 1288–1292. 6 Bressan GM, Daga-Gordini D, Colombatti A, Castellani 22 Lemeshow S, Hosmer Jr DW, Klar J. Sample size I, Marigo V, Volpin D. Emilin, a component of elastic requirements for studies estimating odds ratios or fibers preferentially located at the elastin-microfibrils relative risks. Stat Med 1988; 7: 759–764. interface. J Cell Biol 1993; 121: 201–212. 23 Sole X, Guino E, Valls J, Iniesta R, Moreno V. SNPStats: 7 Miriam Z, Braghetta P, Sabatelli P, Mura I, Doliana R, a web tool for the analysis of association studies. Colombatti A et al. Emilin-1 deficiency induces Bioinformatics 2006; 22: 1928–1929. elastogenesis and vascular cell defects. Mol Cell Biol 24 Stephens M, Smith NJ, Donnelly P. A new statistical 2004; 24: 638–650. method for haplotype reconstruction from population 8 Zacchigna L, Vecchione C, Notte A, Cordenonsi M, data. Am J Hum Genet 2001; 68: 978–989. Dupont S, Maretto S et al. Emilin1 links TGF-beta 25 Sham PC, Curtis D. Monte Carlo tests for associations maturation to blood pressure homeostasis. Cell 2006; between disease and alleles at highly polymorphic 24: 929–942. loci. Ann Hum Genet 1995; 59: 97–105. 9 Lijnen PJ, Petrov VV, Fagard RH. Association between 26 Zhou L, Chen Y, Sun N, Liu X. Family history of transforming growth factor-beta and hypertension. Am hypertension and arterial elasticity characteristics in J Hypertens 2003; 6: 604–611. healthy young people. Hypertens Res 2008; 31: 833–839.

Journal of Human Hypertension Association of EMILIN1 gene with hypertension VMS Oh et al 561 27 Goldstein IB, Shapiro D, Weiss RE. How family history 31 http://www.affymetrix.com/browse/products.jsp? pro and risk factors for hypertension relate to ambulatory ductId ¼ 131459&navMode ¼ 34000&navAction ¼ jump& blood pressure in healthy adults. J Hypertens 2008; 26: aId ¼ productsNav#1_3 (accessed 18 December 2010). 276–283. 32 Cho YS, Go MJ, Kim YJ, Heo JY, Oh HJ, Ban H-J et al. A 28 Shen C, Lu X, Li Y, Zhao Q, Liu X, Hou L et al. Emilin 1 large-scale genome-wide association study of Asian gene and essential hypertension: a two-stage associa- populations uncovers genetic factors influencing eight tion study in a northern Han Chinese population. BMC quantitative traits. Nat Genet 2009; 41: 527–534. Med Genet 2009; 10: 118. 33 Newton-Cheh C, Johnson T, Gateva V, Tobin MD, 29 http://www.illumina.com/products/human1m_duo_d- Bochud M, Coin L et al. Genome-wide association na_analysis_beadchip_kits.ilm (accessed 18 December study identifies eight loci associated with blood 2010). pressure. Nat Genet 2009; 41: 666–676. 30 http://www.affymetrix.com/estore/browse/products.jsp? 34 Levy D, Ehret GB, Rice K, Verwoert GC, Launer LJ, navMode ¼ 34000&productId ¼ 131533&navAction ¼ Dehghan A et al. Genome-wide association study of jump&aId ¼ productsNav#1_1 (accessed 18 December blood pressure and hypertension. Nat Genet 2009; 41: 2010). 677–687.

Supplementary Information accompanies the paper on the Journal of Human Hypertension website (http:// www.nature.com/jhh)

Journal of Human Hypertension