- Email: [email protected]
Interactions among genetic variants from SREBP2 activating-related pathway on risk of coronary heart disease in Chinese Han population
Atherosclerosis 208 (2010) 421–426
Contents lists available at ScienceDirect
Atherosclerosis journal homepage: www.elsevier.com/locate/atherosclerosis
Interactions among genetic variants from SREBP2 activating-related pathway on risk of coronary heart disease in Chinese Han population Xiaoli Liu a,b , Yun Li a , Xiangfeng Lu a , Laiyuan Wang a , Qi Zhao a , Wei Yang a , Jianfeng Huang a , Jie Cao a , Hongfan Li a , Dongfeng Gu a,b,∗ a Department of Evidence Based Medicine and Division of Population Genetics, Cardiovascular Institute and Fu Wai Hospital, Peking Union Medical College and Chinese Academy of Medical Sciences, Beijing, China b National Human Genome Center at Beijing, Beijing, China
a r t i c l e
i n f o
Article history: Received 9 December 2008 Received in revised form 7 June 2009 Accepted 9 August 2009 Available online 14 August 2009 Keywords: Case-control study Coronary heart disease Gene–gene interaction Single nucleotide polymorphism Sterol regulatory element binding protein
a b s t r a c t Sterol regulatory element binding proteins (SREBPs), as a family of membrane-bound transcription factors, control the metabolism of cholesterol and fatty acids. We conducted a case-control study to investigate whether the common variants of genes from the SREBP2 activating-related pathway, involving SREBP2, SCAP, INSIG1 and INSIG2 genes, were associated with coronary heart disease (CHD) of Chinese Han population individually or interactively. Three, four and two single nucleotide polymorphisms (SNPs) from the INSIG1, INSIG2 and SCAP genes were chosen as haplotype-tagging SNPs (htSNPs), respectively, and one nonsynonymous coding SNP was selected from SREBP2. All of the SNPs were genotyped in 853 CHD cases and 948 unrelated control subjects. The interactions among SNPs of the four genes were evaluated with multifactor-dimensionality reduction (MDR) and logistic regression models (LRM). The results from MDR indicated that there existed the SNP-SNP interactive effect of INSIG1 gene on CHD (best prediction accuracy = 56.09%, p = 0.002 on 1000 permutations). The results from LRM also identified the 2-locus interaction model (adjusted p ≤ 0.001 for interaction) as well as the 3-locus gene–gene interaction (adjusted p = 0.026 for interaction). Single polymorphism analysis showed that the rs4822063 of SREBP2 was associated with LDL-C in the controls. The genotype CC carriers had higher LDL-C than the major allele G carriers (3.44 ± 0.90 mmol/L versus 3.17 ± 0.84 mmol/L, adjusted p = 0.038). Our results suggested that the INSIG1 gene was associated with CHD; there might be potential interactive effect on CHD among genes from SREBP2 activating-related pathway; and the SREBP2 gene might be associated with plasma lipid level. © 2009 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Recently, many efforts were made to elucidate the genetic contribution to coronary heart disease (CHD), a complex disease
Abbreviations: BMI, Body mass index; BP, Blood pressure; CHD, Coronary heart disease; CI, Confidence interval; DBP, Diastolic BP; ER, Endoplasmic reticulum; HDL-C, High-density lipoprotein-cholesterol; HMG-CoA, 3-Hydroxy3-methylglutaryl-CoA; htSNP, Haplotype-tagging SNP; HWE, Hardy–Weinberg equilibrium; INSIG, Insulin induced gene; TG, Triglyceride; LD, Linkage disequilibrium; LDL-C, Low-density lipoprotein-cholesterol; LRM, Logistic regression model; MAF, Minor allele frequency; MDR, Multifactor-dimensionality reduction; OR, Odds ratio; SBP, Systolic BP; SCAP, SREBP cleavage-activating protein; SNP, Single nucleotide polymorphism; SREBP, Sterol regulatory element binding protein; TC, Total cholesterol. ∗ Corresponding author at: Department of Evidence Based Medicine and Division of Population Genetics, Cardiovascular Institute, Fu Wai Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 167 Beilishi Road, Beijing, 100037, China. Tel.: +86 10 68331752; fax: +86 10 88363812. E-mail address: [email protected] (D. Gu). 0021-9150/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.atherosclerosis.2009.08.011
resulting from both genetic and environmental factors. To identify genetic determinants of CHD, one strategy is to analyze associations between candidate gene variants of risk factors such as plasma lipoprotein profiles and CHD. Sterol regulatory element binding proteins (SREBPs), as a family of membrane-bound transcription factors, including SREBP-1a, SREBP-1c, and SREBP-2, play important roles in feed back regulation of cellular cholesterol and fatty acid metabolism by activating related genes such as fatty acid synthase, 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) reductase, HMG CoA synthase, and the low density-lipoprotein (LDL) receptor [1,2]. The SREBPs are synthesized as inactive precursors residing in the endoplasmic reticulum (ER) membrane and activation of SREBPs require cleavage by two proteases in the Golgi apparatus [3]. The transportation of SREBPs from ER to Golgi needs another two kinds of proteins, namely SREBP cleavage-activating protein (SCAP) [4] and insulin induced gene (INSIG) proteins including INSIG1 and INSIG2. While the SCAP interacts with SREBPs through its cytoplasmic carboxyl terminus [5], its transmembrane sterolsensing domain interacts with sterols which results in changing
422
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
Fig. 1. Model for the sterol regulation of SREBP transporting and activating pathway. In the present of high concentrations of cholesterol in the ER, SCAP undergoes a conformational change, which increases the affinity of SCAP for Insig1 or 2. The Insigs/SCAP/SREBP complex is thus retained in the ER. In the absence of cholesterol, the affinity of SCAP for Insigs is decreased. This allows the SCAP/SREBP complex to interact with the COPII proteins and bud into COPII-coated vesicles for transporting to the Golgi. In the Golgi the SREBP undergoes two proteolytic cleavages by S1P and S2P proteases. These cleavages release the transcriptional active NH2 fragment which can enter into the nucleus and regulate the transcription of sterol response genes. bHLH basic-helix-loop-helix-leucine zipper; S1P site-1 proteases; S2P site-2 proteases.
conformation of SCAP [6]. INSIG1 and INSIG2 are closely related proteins of the ER, encoded by individual gene [7,8], and mediate feedback control of lipid synthesis by sterol-dependent binding to SCAP. In sterol depleted cells, SCAP transports SREBPs from the ER to the Golgi apparatus. The accumulation of cholesterol triggers changing the conformation [6] and binding of SCAP to either the INSIG1 or the INSIG2, which is a reaction that leads to the ER retention of SCAP/SREBPs complex and prevents the delivery of SREBPs to the Golgi apparatus [7–9]. In a word, the interaction between INSIGs, SCAP and SREBPs play crucial role in feedback regulation of lipid metabolism by sterol-dependent forming complex (the model shown in Fig. 1). There were few studies focused on the relationship between the SREBP2 or SCAP gene and CHD, including AMI, or atherosclerosis [10,11]. The results indicated that the SREBP-2-595A isoform was associated with an increased risk of early-onset MI among U.S. men, and the SCAP polymorphism A2386G appeared to modify the associations of SREBF-2 genotype with MI risk [11]. Other investigation by Robinet et al. [10] indicated the same variant of SREBP2 gene was associated with carotid intima-media thickness (IMT). Furthermore, some studies focused on association of SREBP2 or SCAP genes with plasma lipids, but the results were not all consistent and most of which were conducted in hypercholesterolaemia subjects [10,12,13]. However, there were not any evidences of association of INSIGs gene with CHD or plasma lipids. In present study, we hypothesized that the genetic variants from SREBPs activating pathway were associated with CHD individually or interactively. Thus we conducted a case-control study in Chinese Han population to investigate the relationship between the genetic variants of the SREBPs activating pathway and risk of CHD, involving SREBP2, SCAP, INSIG1 and INSIG2 genes.
2. Materials and methods 2.1. Subjects A total of 1801 unrelated subjects were included in this study. The enrolment criteria of the CHD cases and controls for the Beijing Atherosclerosis Study have been reported in detail previously [14]. We recruited 853 cases with CHD from patients hospitalized at the Fuwai Hospital and Cardiovascular Institute (Beijing, China) between October 1997 and December 2001. Eligible cases were those who survived an acute myocardial infarction or were documented by coronary angiography to have evidence of at least a 70% stenosis in any major coronary artery. Patients with congenital heart disease, cardiomyopathy, valvular disease and renal or hepatic disease were excluded. A total of 948 control subjects were randomly selected from individuals participating in a community-based survey of cardiovascular risk factors. The controls were judged to be free of ischemic changes by ECG, without symptoms of chest pain and to be free of CHD by medical history, the Rose questionnaire and clinical examination. This study was approved by the local Research Ethics Committee, and all subjects gave written informed consent. A set of questionnaires was completed by trained interviewers that included details of medical history, family history of CHD, drug intake, smoking and alcohol consumption. Blood pressure (BP), weight, height, waistline and hip circumference were recorded, and body mass index (BMI) (kg/m2 ) was calculated. Overnight fasting blood samples were drawn by venipuncture to measure serum biochemical measurements including total cholesterol, HDL cholesterol, and triglycerides. Blood specimens were processed in the central clinical laboratory at the Department of
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
Population Genetics at Fuwai Hospital of the Chinese Academy of Medical Sciences in Beijing. This laboratory participates in the Lipid Standardization Program of the US Centers for Disease Control and Prevention. Total cholesterol, HDL cholesterol, and triglycerides were analyzed enzymatically on a Hitachi 7060 Clinical Analyzer (Hitachi High-Technologies Corp). The low-density lipoprotein cholesterol (LDL-C) levels were calculated by use of the Friedewald equation. Three blood pressure measurements were obtained with the participant in the seated position after 5 min of rest. Participants were advised to avoid cigarette smoking, alcohol, caffeinated beverages, and exercise for at least 30 min before their blood pressure measurement. Hypertension was defined as systolic blood pressure (SBP) ≥140 mmHg and/or diastolic blood pressure (DBP) ≥90 mmHg, and/or the patients taking antihypertensive medication. Diabetes was defined as having a fasting plasma glucose level≥7.0 mmol/L and/or self-reported current treatment with antidiabetes medication (insulin or oral hypoglycemic agents). For participants who reported having taken antidiabetes, antihypertensive or lipid-lowering drugs during the previous two weeks, drug intake was documented. 2.2. SNPs selection In this study, a whole gene-based haplotype-tagging SNP (htSNP) approach was used to select candidate SNPs for SCAP, INSIG1 and INSIG2 gene. The htSNPs were selected according to following criteria and steps. Firstly, all SNPs genotype data of each gene were downloaded from HapMap (http://www.hapmap.org/) based on the genotyped SNPs in the Han Chinese of Beijing (CHB) of the HapMap project (the Phase II database, April 2007). The focused region for each gene spanned whole gene and extended 2 kb from 5 flanking upstream and 3 flanking downstream, respectively. Secondly, htSNPs were chosen with Haploview 4.0 software (http://www.broad.mit.edu/mpg/haploview/) based on downloaded genotype data. The candidates SNPs were restricted to minor allele frequency (MAF) ≥0.05 with examined haplotypes frequency ≥0.05 in CHB, and the block definition was taken from Gabriel et al. [15], where the upper confidence interval (CI) limit for strong pairwise linkage disequilibrium was 0.98 and the lower CI limit was 0.7. The focused region in SCAP, INSIG1 and INSIG2 gene had only one block according to this definition. Finally, the program was performed to select htSNPs. In results, three, four and two SNPs were selected as htSNPs for the INSIG1, INSIG2 and SCAP gene, respectively. The haplotypes combined by selected htSNPs could cover 99.9%, 92.9% and 92.2% of haplotype diversities in CHB for INSIG1, INSIG2 and SCAP genes, respectively. In addition, one nonsynonymous-coding SNP rs4822063 in SREBP2 gene (previously denoted as A595G [13] or 1784G>C [10]) was selected. All genotyped SNPs were listed in supplementary material. 2.3. Genotyping Genomic DNA was isolated from white blood cells by the standard phenol/chloroform method and was stored in 400 l of TE [10 mM Tris–HCl and 1 mM EDTA (pH 8.0)]. The 10 selected SNPs were genotyped using standard polymerase chain reaction and restriction-fragment-length polymorphism methods. Primers and conditions for amplification are available from the authors upon request. Re-genotyping all SNPs was performed in 96 randomly selected individuals by a team member with complete concordance as quality control.
423
such as sex, drinking, smoking and with t-test for the continuous variables such as age and blood pressure. The deviations from the Hardy–Weinberg equilibrium (HWE) of the selected SNPs were tested by the 2 -test among the controls. For single locus analyses, genotype and allele frequency distributions across controls and cases were tested by 2 -test, and association was analyzed under additive, dominant and recessive models, respectively. The logistic regression model (LRM) and general linear model (GLM) was performed to examine the independent effect of the each polymorphism on CHD and plasma lipids level under different models with adjustment for covariates or not. We conducted full model regression fitting potential covariates to identify covariates needed to be adjusted including age, gender, BMI, smoking, alcohol consumption, hypertension, diabetes, TG, TC and HDL-C, before analyzing on SNPs. The 2 -test, GLM and LRM were performed using the SPSS13.0 software. We used p < 0.05 to define statistical significance. The gene–gene interaction was examined using both the Multifactor Dimensionality Reduction (MDR) program and LRM. The MDR program (http://sourceforge.net/projects/mdr/) as a nonparametric and genetic model free approach was introduced to analyze high-order joint effects of loci in genetic association [16,17]. The MDR method includes a combined cross-validation/permutation testing procedure that minimizes the false-positive findings that may otherwise result from multiple comparisons. With 10-fold cross-validation, the data are divided into 10 equal parts, and the model is developed on 9/10 of the data (training set) and then tested on 1/10 of the remaining data (testing set). This is repeated for each possible 9/10 and 1/10 of the data, and the resulting 10 prediction accuracies are averaged to get an averaged prediction accuracy used to measure how many instances were correctly classified using the MDR model. It was thought the higher the accuracy, the better the model. In addition to the prediction accuracy, it also reports the cross-validation consistency, which is a measure of how many times out of 10 divisions of the data MDR find the same best model thus is thought that the higher the consistency, the better the model. The cross-validation analysis can be conducted several times using different random seeds and the results averaged to avoid spurious results due to chance divisions of the data. In our study, we repeated cross-validation process ten times totally by supplied different random seeds then averaged all 10 results as final prediction accuracy and cross-validation consistency. Permutation testing can be performed to assess the probability of obtaining a testing accuracy as large as or larger than the observed in the original data, given that the null hypothesis of no association is true. This is carried out by randomizing the casecontrol labels 1000 times and repeating the MDR analysis on each randomized dataset. This process yields an empirical distribution of testing accuracies under the null hypothesis, which is in turn used to calculate a P-value. The model with maximum average testing balanced accuracy and best cross-validation consistency will be selected as the optimal one. With LRM, we had examined all models with statistic significance (p ≤ 0.05) from MDR analyses. Because the genetic risk model was unknown for most of the SNPs we studied, we adopted an additive model, in which the major allele homozygote was defined as “0” as well as “1” for heterozygote and “2” for minor allele homozygote, respectively.
3. Results
2.4. Statistical analysis
3.1. Clinical characteristics
The statistical analyses on the characteristics of the subjects were performed with Pearson 2 -test for the categorical variables
The demographic details of the cases and controls were given in Table 1. Every characteristic was significantly different between
424
X. Liu et al. / Atherosclerosis 208 (2010) 421–426 Table 1 Characteristics of the cases with CHD compared with the controls.
Age (years) Male Hypertension Diabetes Smokers Drinkers BMI (kg/m2 ) TC (mmol/L) TG (mmol/L) HDL-C (mmol/L) LDL-C (mmol/L) Glu (mmol/L) SBP (mmHg) DBP (mmHg) Family history
Controls (n = 948)
Cases (n = 853)
52.0 ± 10.3 699 (73.7) 305 (32.2) 77 (8.1) 540 (57.0) 394 (41.6) 24.80 ± 3.28 5.15 ± 0.97 1.45 ± 0.96 1.29 ± 0.30 3.18 ± 0.85 5.52 ± 1.55 127.26 ± 18.17 80.25 ± 9.93 –
54.6 ± 8.9 667(78.2) 536(62.9) 220(25.8) 527(61.8) 398(46.7) 26.52 ± 3.21 5.24 ± 1.13 1.83 ± 1.22 1.09 ± 0.25 3.30 ± 1.03 6.01 ± 2.06 131.10 ± 20.55 76.31 ± 11.02 292(34.2)
p-value <0.001 0.027 <0.001 <0.001 0.038 0.052 <0.001 0.049 <0.001 <0.001 0.006 <0.001 <0.001 <0.001 –
Mean ± standard deviation values for quantitative variables and n (%) for qualitative variables. BMI, body mass index; TC, total cholesterol; TG, triglyceride; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; Glu, glucose; SBP, systolic blood pressure; DBP, diastolic blood pressure.
cases and controls except alcohol consumption. Compared with the control group, the CHD group had more male, older and smoking cases, and more individuals with hypertension, diabetes. Moreover, the cases had higher mean BMI and SBP, higher levels of serum TC, TG, LDL-C and fasting glucose, and lower HDL-C levels than the controls. 3.2. Single polymorphism analysis No significant deviation from HWE was found among the controls for all the polymorphisms. The distributions of genotypes and alleles for the SNPs for the studied gene among groups were shown in Table 2. Before analyzing on single locus, we identified a set of covariates including age, hypertension, diabetes, BMI, HDL-C and TC needed to be adjusted by conducting full model logistic regression. Then we performed full model logistic regression to test whether the single SNPs were associated with CHD with adjustment for the six identified covariates or without. The results did not show association for any genotyped SNPs with CHD whether adjusted for covariates or not. Similarly, we identified that age, gender, diabetes status and BMI were significantly associated with plasma lipids by con-
ducting full model general linear regression, irrespective of genetic variants. Thus we made adjustment for the four covariates when performing analysis on effects of SNPs on lipids. The results in the controls excluding seventeen individuals taking lipid-lowering drugs showed that the rs4822063 of SREBP2 was associated with plasma LDL-C and TC after adjustment for identified covariates. The genotype CC carriers had higher LDL-C and TC than the major allele G carriers, but the significant relationship disappeared after adjustment for multiple tests (shown in Table 3).
3.3. Multi-loci interaction analyses All the genotyped SNPs were involved in MDR analysis for testing association between the SNPs and CHD, and the results for 2-locus to 4-locus were listed in Table 4. We found that the 2-locus model involving the rs10271719 and rs9719268 of the INSIG1 gene showed the highest level of testing accuracy (56.09%, p = 0.002 on 1000 permutations) and the best 10-fold cross-validation consistency. Moreover, other two models involving INSIG1, INSIG2 and SCAP genes also showed significant prediction accuracy but lower values than the 2-locus model.
Table 2 Distribution of genotypes and alleles in controls and cases with CHD. . Gene
SNPs
Genotype
Control
Case
INSIG1
rs10271719
GG/GT/TT G/T GG/GT/TT G/T AA/AG/GG A/G
0.504/0.419/0.077 0.713/0.287 0.777/0.208/0.014 0.882/0.118 0.761/0.219/0.020 0.870/0.130
0.516/0.385/0.099 0.708/0.292 0.752/0.228/0.019 0.867/0.133 0.736/0.251/0.013 0.861/0.138
AA/AG/GG A/G TT/TC/CC T/C AA/AG/GG A/G TT/TC/CC T/C
0.779/0.207/0.014 0.882/0.118 0.419/0.458/0.123 0.648/0.352 0.563/0.379/0.058 0.753/0.247 0.252/0.507/0.241 0.505/0.495
0.782/0.206/0.012 0.885/0.115 0.430/0.451/0.118 0.657/0.343 0.572/0.373/0.055 0.758/0.242 0.266/0.489/0.245 0.510/0.490
TT/TC/CC T/C GG/GA/AA G/A
0.728/0.253/0.019 0.854/0.146 0.276/0.492/0.232 0.522/0.478
0.727/0.254/0.019 0.854/0.146 0.260/0.507/0.233 0.514/0.486
GG/GC/CC G/C
0.619/0.330/0.051 0.784/0.216
0.625/0.327/0.048 0.789/0.211
rs9719268 rs9769826 INSIG2
rs10197745 rs4848492 rs17047757 rs9308762
SCAP
rs17079634 rs4858868
SREBP2
rs4822063
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
425
Table 3 Association between SREBP2-rs4822063 and plasma lipids level in controls.# SNP rs4822063 # *
Genotype
LDL-C (mmol/L)
CC GG + GC
3.44 ± 0.90 3.17 ± 0.84
TC (mmol/L) 5.43 ± 1.06 5.13 ± 0.97
*
P = 0.038
P* = 0.051
Excluding individuals taking lipid-lowering drugs. Adjusted for age, gender, diabetes, BMI.
Table 4 Multi-loci interaction models by MDR and logistic regression analyses. Locus No. and combination
2-Locus: INSIG1-rs10271719-rs9719268 3-Locus: INSIG1-rs10271719-rs9719268 INSIG2-rs17047757 4-Locus: INSIG1rs10271719-rs9719268 INSIG2-rs9308762SCAP-rs4858868 * # a
MDR analysis
Logistic regression
Cross-validation consistency
Testing accuracy (%)
p-value*
Unadjusted p-valuea
Adjusted p-value# , a
10 8.3
56.09 55.56
0.002 0.004
≤0.001 0.045
≤0.001 0.026
9.9
54.60
0.03
0.276
0.178
p-value based on 1000 permutation. Adjusted for age, BMI, hypertension, diabetes, TC and HDL-C. p-values for interaction.
Subsequently, we performed LRM to evaluate the models from MDR. The 2-locus, 3-locus models resulted from MDR still remained significant interaction whether with adjustment for identified six risk factors of CHD or without (unadjusted and adjusted p-values for interaction were < 0.001 and <0.001 for 2-locus, 0.045 and 0.026 for 3-locus, respectively). But the 4-locus model did not show statistical significance (shown in Table 4). 4. Discussion The aim of the present study was to investigate the association between the common variants from the SREBP2 activating-related pathway and CHD. The results of the study indicated that there might be potential interaction on the risk of CHD between the genes from the focused pathway; and the SREBP2 gene was likely associated with plasma lipids level. Previously Miserez et al. [13] discovered a C to G transversion in exon 10 of the SREBP2 gene which caused a substitution of alanine to glycine (A595G) at amino acid position 595 on the protein level. The authors suggested that there was an impaired interaction and/or decreased stability of the complex between SCAP and the COOH-terminal domain of the SREBP2-595A isoform compared to the SREBP2-595G isoform. The SREBP2-595A isoform (genotype AA) was associated with significantly higher plasma TC concentrations in hypercholesterolemic subjects. Klos and Sing [18] reported that a single haplotype of SREBP2 was associated with apoB level in whites. In our study, we found that the subjects with genotype CC, same as previous reported 595A-homozygous, had higher TC and LDL-C than the major allele G carriers. Based on those findings and importance in lipid metabolism, we inferred that the SREBP2 was associated with plasma lipid levels. However, our results showed only the SREBP2 gene, but not the INSIG1, INSIG2 or SCAP genes, was associated with plasma lipid levels in controls. The results might be due to that the SREBP2 gene had more direct effect on cholesterol metabolism than other genes in the focused pathway by activating related genes transcription (shown in Fig. 1). MDR and LRM have individual characteristics when dealing with multi-locus interaction analysis. MDR, as a non-parametric method, does not require or assume any specific parametric form as well as genetic model for the relation between independent and dependent variables. LRM, as a parametric approach, requires specific parametric form as well as special genetic model and can detect
only low-order interactions as the model complexity increases with the order of interactions. However, MDR exhaustively searches across all multi-locus genotypes and identifies just one that best predicts the disease outcome which might miss other existing interactions due to interaction between dependent variables. Furthermore, the interaction model resulted from MDR does not seem to have clear biological interpretation but that is easy to get by LRM. Therefore, one can use both two methods as well as other methods to examine the gene–gene interaction. The strategy using both two methods had been performed by Briollais et al. to examine gene–gene interaction in breast cancer susceptibility [19] and by Qi et al. to identify the gene–gene interactions between HNF4A and KCNJ11 in predicting Type 2 diabetes in women [20]. The results from the MDR (non-parametric) and LRM (parametric) analyses consistently identified potential interaction between SNP rs10271719 and rs9719268 of the INSIG1 gene. In previous study, we found that the haplotype of the INSIG1 gene including those two SNPs was associated with CHD [21]. Combining the findings, we concluded that the INSIG1 gene was associated with CHD. In addition, besides the 2-locus model with the best prediction accuracy and cross-validation consistency and the least permutation p value, other two models, a 3-locus and a 4-locus model, were shown with permutation p-value less than 0.05 as well (shown in Table 4). With further LRM analysis on those models, we identified the significant 3-locus interaction model form MDR involving INSIG1, INSIG2 genes. The results from MDR and LRM confirmed our hypothesis that the genes from SREBP2 activating-related pathway might have individual or interactive effect on risk of CHD. Our results also suggested that the use of multiple statistical approaches rather than a single methodology could be the best strategy to elucidate complex gene interactions that had generally very different patterns. A major strength of the study was that the gene–gene interactions were consistently examined in analyses using different statistical models (parametric and non-parametric). However, there were a few limitations in our present study. Firstly, instead of selecting htSNPs for the SREBP2 gene, we had just chosen one nonsynonymous-coding SNP that might decrease the power to detect the association between the SREBP2 gene and CHD. Secondly, we had just selected the SREBP2 gene, but not other genes, from the SREBPs family as studied candidate gene in our study. Through biologic functional analyses and reviewing findings from previous studies, we supposed the SREBP2 had more possibility to
426
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
be associated with CHD. At the same time, it might lead to loss some important information. We hope further studies that will evaluate the effect of the SREBP genes on lipid metabolism and CHD, involving more SREBP genes and more SNPs covering whole genes. Moreover, we looked only at survivors of a hospitalized CHD patients so we had therefore omit the subjects that died in hospital thus might weaken the association between the SNP and CHD, so we should explain the results careful and need prospective cohort study to test our results. Our results from MDR and LRM provided evidence of interaction between the genes from SREBP2 activating-related pathway on risk of CHD for first time so our results were difficulty to generalize to other population due to lacking additional evidences from others. In addition, it also needed further replication in various populations due to complex genetic background between different populations even our result that the SREBP2 gene was associated with plasma lipid level was similar with other results from others. Our findings warranted further studies to replicate our results in other peoples and to elucidate the biological mechanism. Acknowledgments This work was supported by National Basic Research Program of China (Grant No. 2006CB503805). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.atherosclerosis.2009.08.011. References [1] Brown MS, Goldstein JL. The SREBP pathway: regulation of cholesterol metabolism by proteolysis of a membrane-bound transcription factor. Cell 1997;89(3):331–40. [2] Horton JD, Shah NA, Warrington JA, et al. Combined analysis of oligonucleotide microarray data from transgenic and knockout mice identifies direct SREBP target genes. Proc Natl Acad Sci USA 2003;100(21):12027–32. [3] Brown MS, Goldstein JL. A proteolytic pathway that controls the cholesterol content of membranes, cells, and blood. Proc Natl Acad Sci USA 1999;96(20):11041–8.
[4] Hua X, Nohturfft A, Goldstein JL, et al. Sterol resistance in CHO cells traced to point mutation in SREBP cleavage-activating protein. Cell 1996;87(3):415–26. [5] Sakai J, Nohturfft A, Cheng D, et al. Identification of complexes between the COOH-terminal domains of sterol regulatory element-binding proteins (SREBPs) and SREBP cleavage-activating protein. J Biol Chem 1997;272(32):20213–21. [6] Brown AJ, Sun L, Feramisco JD, et al. Cholesterol addition to ER membranes alters conformation of SCAP, the SREBP escort protein that regulates cholesterol metabolism. Mol Cell 2002;10(2):237–45. [7] Yang T, Espenshade PJ, Wright ME, et al. Crucial step in cholesterol homeostasis: sterols promote binding of SCAP to INSIG-1, a membrane protein that facilitates retention of SREBPs in ER. Cell 2002;110(4):489–500. [8] Yabe D, Brown MS, Goldstein JL. Insig-2, a second endoplasmic reticulum protein that binds SCAP and blocks export of sterol regulatory element-binding proteins. Proc Natl Acad Sci USA 2002;99(20):12753–8. [9] Goldstein JL, DeBose-Boyd RA, Brown MS. Protein sensors for membrane sterols. Cell 2006;124(1):35–46. [10] Robinet P, Vedie B, Chironi G, et al. Characterization of polymorphic structure of SREBP-2 gene: role in atherosclerosis. Atherosclerosis 2003;168(2):381–7. [11] Friedlander Y, Schwartz SM, Durst R, et al. SREBP-2 and SCAP isoforms and risk of early onset myocardial infarction. Atherosclerosis 2007. [12] Duan X, Zhu W, Li Y, et al. The effect of sterol regulatory element-binding protein 2 polymorphism on the serum lipid in northern Chinese subjects. J Lipid Res 2005;46(2):252–7. [13] Miserez AR, Muller PY, Barella L, et al. Sterol-regulatory element-binding protein (SREBP)-2 contributes to polygenic hypercholesterolaemia. Atherosclerosis 2002;164(1):15–26. [14] Zhou X, Huang J, Chen J, et al. Thrombospondin-4 A387P polymorphism is not associated with coronary artery disease and myocardial infarction in the Chinese Han population. Clin Sci (Lond) 2004;106(5):495–500. [15] Gabriel SB, Schaffner SF, Nguyen H, et al. The structure of haplotype blocks in the human genome. Science 2002;296(5576):2225–9. [16] Hahn LW, Ritchie MD, Moore JH. Multifactor dimensionality reduction software for detecting gene-gene and gene-environment interactions. Bioinformatics 2003;19(3):376–82. [17] Ritchie MD, Hahn LW, Roodi N, et al. Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am J Hum Genet 2001;69(1):138–47. [18] Klos KL, Sing CF, Boerwinkle E, et al. Consistent effects of genes involved in reverse cholesterol transport on plasma lipid and apolipoprotein levels in CARDIA participants. Arterioscler Thromb Vasc Biol 2006;26(8):1828–36. [19] Briollais L, Wang Y, Rajendram I, et al. Methodological issues in detecting genegene interactions in breast cancer susceptibility: a population-based study in Ontario. BMC Med 2007;5:22. [20] Qi L, van Dam RM, Asselbergs FW, et al. Gene-gene interactions between HNF4A and KCNJ11 in predicting Type 2 diabetes in women. Diabet Med 2007;24(11):1187–91. [21] Liu X, Li Y, Wang L, et al. The INSIG1 gene, not the INSIG2 gene, associated with coronary heart disease: tagSNPs and haplotype-based association study. The Beijing Atherosclerosis Study. Thromb Haemost 2008;100(5):886–92.
Contents lists available at ScienceDirect
Atherosclerosis journal homepage: www.elsevier.com/locate/atherosclerosis
Interactions among genetic variants from SREBP2 activating-related pathway on risk of coronary heart disease in Chinese Han population Xiaoli Liu a,b , Yun Li a , Xiangfeng Lu a , Laiyuan Wang a , Qi Zhao a , Wei Yang a , Jianfeng Huang a , Jie Cao a , Hongfan Li a , Dongfeng Gu a,b,∗ a Department of Evidence Based Medicine and Division of Population Genetics, Cardiovascular Institute and Fu Wai Hospital, Peking Union Medical College and Chinese Academy of Medical Sciences, Beijing, China b National Human Genome Center at Beijing, Beijing, China
a r t i c l e
i n f o
Article history: Received 9 December 2008 Received in revised form 7 June 2009 Accepted 9 August 2009 Available online 14 August 2009 Keywords: Case-control study Coronary heart disease Gene–gene interaction Single nucleotide polymorphism Sterol regulatory element binding protein
a b s t r a c t Sterol regulatory element binding proteins (SREBPs), as a family of membrane-bound transcription factors, control the metabolism of cholesterol and fatty acids. We conducted a case-control study to investigate whether the common variants of genes from the SREBP2 activating-related pathway, involving SREBP2, SCAP, INSIG1 and INSIG2 genes, were associated with coronary heart disease (CHD) of Chinese Han population individually or interactively. Three, four and two single nucleotide polymorphisms (SNPs) from the INSIG1, INSIG2 and SCAP genes were chosen as haplotype-tagging SNPs (htSNPs), respectively, and one nonsynonymous coding SNP was selected from SREBP2. All of the SNPs were genotyped in 853 CHD cases and 948 unrelated control subjects. The interactions among SNPs of the four genes were evaluated with multifactor-dimensionality reduction (MDR) and logistic regression models (LRM). The results from MDR indicated that there existed the SNP-SNP interactive effect of INSIG1 gene on CHD (best prediction accuracy = 56.09%, p = 0.002 on 1000 permutations). The results from LRM also identified the 2-locus interaction model (adjusted p ≤ 0.001 for interaction) as well as the 3-locus gene–gene interaction (adjusted p = 0.026 for interaction). Single polymorphism analysis showed that the rs4822063 of SREBP2 was associated with LDL-C in the controls. The genotype CC carriers had higher LDL-C than the major allele G carriers (3.44 ± 0.90 mmol/L versus 3.17 ± 0.84 mmol/L, adjusted p = 0.038). Our results suggested that the INSIG1 gene was associated with CHD; there might be potential interactive effect on CHD among genes from SREBP2 activating-related pathway; and the SREBP2 gene might be associated with plasma lipid level. © 2009 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Recently, many efforts were made to elucidate the genetic contribution to coronary heart disease (CHD), a complex disease
Abbreviations: BMI, Body mass index; BP, Blood pressure; CHD, Coronary heart disease; CI, Confidence interval; DBP, Diastolic BP; ER, Endoplasmic reticulum; HDL-C, High-density lipoprotein-cholesterol; HMG-CoA, 3-Hydroxy3-methylglutaryl-CoA; htSNP, Haplotype-tagging SNP; HWE, Hardy–Weinberg equilibrium; INSIG, Insulin induced gene; TG, Triglyceride; LD, Linkage disequilibrium; LDL-C, Low-density lipoprotein-cholesterol; LRM, Logistic regression model; MAF, Minor allele frequency; MDR, Multifactor-dimensionality reduction; OR, Odds ratio; SBP, Systolic BP; SCAP, SREBP cleavage-activating protein; SNP, Single nucleotide polymorphism; SREBP, Sterol regulatory element binding protein; TC, Total cholesterol. ∗ Corresponding author at: Department of Evidence Based Medicine and Division of Population Genetics, Cardiovascular Institute, Fu Wai Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, 167 Beilishi Road, Beijing, 100037, China. Tel.: +86 10 68331752; fax: +86 10 88363812. E-mail address: [email protected] (D. Gu). 0021-9150/$ – see front matter © 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.atherosclerosis.2009.08.011
resulting from both genetic and environmental factors. To identify genetic determinants of CHD, one strategy is to analyze associations between candidate gene variants of risk factors such as plasma lipoprotein profiles and CHD. Sterol regulatory element binding proteins (SREBPs), as a family of membrane-bound transcription factors, including SREBP-1a, SREBP-1c, and SREBP-2, play important roles in feed back regulation of cellular cholesterol and fatty acid metabolism by activating related genes such as fatty acid synthase, 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) reductase, HMG CoA synthase, and the low density-lipoprotein (LDL) receptor [1,2]. The SREBPs are synthesized as inactive precursors residing in the endoplasmic reticulum (ER) membrane and activation of SREBPs require cleavage by two proteases in the Golgi apparatus [3]. The transportation of SREBPs from ER to Golgi needs another two kinds of proteins, namely SREBP cleavage-activating protein (SCAP) [4] and insulin induced gene (INSIG) proteins including INSIG1 and INSIG2. While the SCAP interacts with SREBPs through its cytoplasmic carboxyl terminus [5], its transmembrane sterolsensing domain interacts with sterols which results in changing
422
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
Fig. 1. Model for the sterol regulation of SREBP transporting and activating pathway. In the present of high concentrations of cholesterol in the ER, SCAP undergoes a conformational change, which increases the affinity of SCAP for Insig1 or 2. The Insigs/SCAP/SREBP complex is thus retained in the ER. In the absence of cholesterol, the affinity of SCAP for Insigs is decreased. This allows the SCAP/SREBP complex to interact with the COPII proteins and bud into COPII-coated vesicles for transporting to the Golgi. In the Golgi the SREBP undergoes two proteolytic cleavages by S1P and S2P proteases. These cleavages release the transcriptional active NH2 fragment which can enter into the nucleus and regulate the transcription of sterol response genes. bHLH basic-helix-loop-helix-leucine zipper; S1P site-1 proteases; S2P site-2 proteases.
conformation of SCAP [6]. INSIG1 and INSIG2 are closely related proteins of the ER, encoded by individual gene [7,8], and mediate feedback control of lipid synthesis by sterol-dependent binding to SCAP. In sterol depleted cells, SCAP transports SREBPs from the ER to the Golgi apparatus. The accumulation of cholesterol triggers changing the conformation [6] and binding of SCAP to either the INSIG1 or the INSIG2, which is a reaction that leads to the ER retention of SCAP/SREBPs complex and prevents the delivery of SREBPs to the Golgi apparatus [7–9]. In a word, the interaction between INSIGs, SCAP and SREBPs play crucial role in feedback regulation of lipid metabolism by sterol-dependent forming complex (the model shown in Fig. 1). There were few studies focused on the relationship between the SREBP2 or SCAP gene and CHD, including AMI, or atherosclerosis [10,11]. The results indicated that the SREBP-2-595A isoform was associated with an increased risk of early-onset MI among U.S. men, and the SCAP polymorphism A2386G appeared to modify the associations of SREBF-2 genotype with MI risk [11]. Other investigation by Robinet et al. [10] indicated the same variant of SREBP2 gene was associated with carotid intima-media thickness (IMT). Furthermore, some studies focused on association of SREBP2 or SCAP genes with plasma lipids, but the results were not all consistent and most of which were conducted in hypercholesterolaemia subjects [10,12,13]. However, there were not any evidences of association of INSIGs gene with CHD or plasma lipids. In present study, we hypothesized that the genetic variants from SREBPs activating pathway were associated with CHD individually or interactively. Thus we conducted a case-control study in Chinese Han population to investigate the relationship between the genetic variants of the SREBPs activating pathway and risk of CHD, involving SREBP2, SCAP, INSIG1 and INSIG2 genes.
2. Materials and methods 2.1. Subjects A total of 1801 unrelated subjects were included in this study. The enrolment criteria of the CHD cases and controls for the Beijing Atherosclerosis Study have been reported in detail previously [14]. We recruited 853 cases with CHD from patients hospitalized at the Fuwai Hospital and Cardiovascular Institute (Beijing, China) between October 1997 and December 2001. Eligible cases were those who survived an acute myocardial infarction or were documented by coronary angiography to have evidence of at least a 70% stenosis in any major coronary artery. Patients with congenital heart disease, cardiomyopathy, valvular disease and renal or hepatic disease were excluded. A total of 948 control subjects were randomly selected from individuals participating in a community-based survey of cardiovascular risk factors. The controls were judged to be free of ischemic changes by ECG, without symptoms of chest pain and to be free of CHD by medical history, the Rose questionnaire and clinical examination. This study was approved by the local Research Ethics Committee, and all subjects gave written informed consent. A set of questionnaires was completed by trained interviewers that included details of medical history, family history of CHD, drug intake, smoking and alcohol consumption. Blood pressure (BP), weight, height, waistline and hip circumference were recorded, and body mass index (BMI) (kg/m2 ) was calculated. Overnight fasting blood samples were drawn by venipuncture to measure serum biochemical measurements including total cholesterol, HDL cholesterol, and triglycerides. Blood specimens were processed in the central clinical laboratory at the Department of
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
Population Genetics at Fuwai Hospital of the Chinese Academy of Medical Sciences in Beijing. This laboratory participates in the Lipid Standardization Program of the US Centers for Disease Control and Prevention. Total cholesterol, HDL cholesterol, and triglycerides were analyzed enzymatically on a Hitachi 7060 Clinical Analyzer (Hitachi High-Technologies Corp). The low-density lipoprotein cholesterol (LDL-C) levels were calculated by use of the Friedewald equation. Three blood pressure measurements were obtained with the participant in the seated position after 5 min of rest. Participants were advised to avoid cigarette smoking, alcohol, caffeinated beverages, and exercise for at least 30 min before their blood pressure measurement. Hypertension was defined as systolic blood pressure (SBP) ≥140 mmHg and/or diastolic blood pressure (DBP) ≥90 mmHg, and/or the patients taking antihypertensive medication. Diabetes was defined as having a fasting plasma glucose level≥7.0 mmol/L and/or self-reported current treatment with antidiabetes medication (insulin or oral hypoglycemic agents). For participants who reported having taken antidiabetes, antihypertensive or lipid-lowering drugs during the previous two weeks, drug intake was documented. 2.2. SNPs selection In this study, a whole gene-based haplotype-tagging SNP (htSNP) approach was used to select candidate SNPs for SCAP, INSIG1 and INSIG2 gene. The htSNPs were selected according to following criteria and steps. Firstly, all SNPs genotype data of each gene were downloaded from HapMap (http://www.hapmap.org/) based on the genotyped SNPs in the Han Chinese of Beijing (CHB) of the HapMap project (the Phase II database, April 2007). The focused region for each gene spanned whole gene and extended 2 kb from 5 flanking upstream and 3 flanking downstream, respectively. Secondly, htSNPs were chosen with Haploview 4.0 software (http://www.broad.mit.edu/mpg/haploview/) based on downloaded genotype data. The candidates SNPs were restricted to minor allele frequency (MAF) ≥0.05 with examined haplotypes frequency ≥0.05 in CHB, and the block definition was taken from Gabriel et al. [15], where the upper confidence interval (CI) limit for strong pairwise linkage disequilibrium was 0.98 and the lower CI limit was 0.7. The focused region in SCAP, INSIG1 and INSIG2 gene had only one block according to this definition. Finally, the program was performed to select htSNPs. In results, three, four and two SNPs were selected as htSNPs for the INSIG1, INSIG2 and SCAP gene, respectively. The haplotypes combined by selected htSNPs could cover 99.9%, 92.9% and 92.2% of haplotype diversities in CHB for INSIG1, INSIG2 and SCAP genes, respectively. In addition, one nonsynonymous-coding SNP rs4822063 in SREBP2 gene (previously denoted as A595G [13] or 1784G>C [10]) was selected. All genotyped SNPs were listed in supplementary material. 2.3. Genotyping Genomic DNA was isolated from white blood cells by the standard phenol/chloroform method and was stored in 400 l of TE [10 mM Tris–HCl and 1 mM EDTA (pH 8.0)]. The 10 selected SNPs were genotyped using standard polymerase chain reaction and restriction-fragment-length polymorphism methods. Primers and conditions for amplification are available from the authors upon request. Re-genotyping all SNPs was performed in 96 randomly selected individuals by a team member with complete concordance as quality control.
423
such as sex, drinking, smoking and with t-test for the continuous variables such as age and blood pressure. The deviations from the Hardy–Weinberg equilibrium (HWE) of the selected SNPs were tested by the 2 -test among the controls. For single locus analyses, genotype and allele frequency distributions across controls and cases were tested by 2 -test, and association was analyzed under additive, dominant and recessive models, respectively. The logistic regression model (LRM) and general linear model (GLM) was performed to examine the independent effect of the each polymorphism on CHD and plasma lipids level under different models with adjustment for covariates or not. We conducted full model regression fitting potential covariates to identify covariates needed to be adjusted including age, gender, BMI, smoking, alcohol consumption, hypertension, diabetes, TG, TC and HDL-C, before analyzing on SNPs. The 2 -test, GLM and LRM were performed using the SPSS13.0 software. We used p < 0.05 to define statistical significance. The gene–gene interaction was examined using both the Multifactor Dimensionality Reduction (MDR) program and LRM. The MDR program (http://sourceforge.net/projects/mdr/) as a nonparametric and genetic model free approach was introduced to analyze high-order joint effects of loci in genetic association [16,17]. The MDR method includes a combined cross-validation/permutation testing procedure that minimizes the false-positive findings that may otherwise result from multiple comparisons. With 10-fold cross-validation, the data are divided into 10 equal parts, and the model is developed on 9/10 of the data (training set) and then tested on 1/10 of the remaining data (testing set). This is repeated for each possible 9/10 and 1/10 of the data, and the resulting 10 prediction accuracies are averaged to get an averaged prediction accuracy used to measure how many instances were correctly classified using the MDR model. It was thought the higher the accuracy, the better the model. In addition to the prediction accuracy, it also reports the cross-validation consistency, which is a measure of how many times out of 10 divisions of the data MDR find the same best model thus is thought that the higher the consistency, the better the model. The cross-validation analysis can be conducted several times using different random seeds and the results averaged to avoid spurious results due to chance divisions of the data. In our study, we repeated cross-validation process ten times totally by supplied different random seeds then averaged all 10 results as final prediction accuracy and cross-validation consistency. Permutation testing can be performed to assess the probability of obtaining a testing accuracy as large as or larger than the observed in the original data, given that the null hypothesis of no association is true. This is carried out by randomizing the casecontrol labels 1000 times and repeating the MDR analysis on each randomized dataset. This process yields an empirical distribution of testing accuracies under the null hypothesis, which is in turn used to calculate a P-value. The model with maximum average testing balanced accuracy and best cross-validation consistency will be selected as the optimal one. With LRM, we had examined all models with statistic significance (p ≤ 0.05) from MDR analyses. Because the genetic risk model was unknown for most of the SNPs we studied, we adopted an additive model, in which the major allele homozygote was defined as “0” as well as “1” for heterozygote and “2” for minor allele homozygote, respectively.
3. Results
2.4. Statistical analysis
3.1. Clinical characteristics
The statistical analyses on the characteristics of the subjects were performed with Pearson 2 -test for the categorical variables
The demographic details of the cases and controls were given in Table 1. Every characteristic was significantly different between
424
X. Liu et al. / Atherosclerosis 208 (2010) 421–426 Table 1 Characteristics of the cases with CHD compared with the controls.
Age (years) Male Hypertension Diabetes Smokers Drinkers BMI (kg/m2 ) TC (mmol/L) TG (mmol/L) HDL-C (mmol/L) LDL-C (mmol/L) Glu (mmol/L) SBP (mmHg) DBP (mmHg) Family history
Controls (n = 948)
Cases (n = 853)
52.0 ± 10.3 699 (73.7) 305 (32.2) 77 (8.1) 540 (57.0) 394 (41.6) 24.80 ± 3.28 5.15 ± 0.97 1.45 ± 0.96 1.29 ± 0.30 3.18 ± 0.85 5.52 ± 1.55 127.26 ± 18.17 80.25 ± 9.93 –
54.6 ± 8.9 667(78.2) 536(62.9) 220(25.8) 527(61.8) 398(46.7) 26.52 ± 3.21 5.24 ± 1.13 1.83 ± 1.22 1.09 ± 0.25 3.30 ± 1.03 6.01 ± 2.06 131.10 ± 20.55 76.31 ± 11.02 292(34.2)
p-value <0.001 0.027 <0.001 <0.001 0.038 0.052 <0.001 0.049 <0.001 <0.001 0.006 <0.001 <0.001 <0.001 –
Mean ± standard deviation values for quantitative variables and n (%) for qualitative variables. BMI, body mass index; TC, total cholesterol; TG, triglyceride; HDL-C, high density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; Glu, glucose; SBP, systolic blood pressure; DBP, diastolic blood pressure.
cases and controls except alcohol consumption. Compared with the control group, the CHD group had more male, older and smoking cases, and more individuals with hypertension, diabetes. Moreover, the cases had higher mean BMI and SBP, higher levels of serum TC, TG, LDL-C and fasting glucose, and lower HDL-C levels than the controls. 3.2. Single polymorphism analysis No significant deviation from HWE was found among the controls for all the polymorphisms. The distributions of genotypes and alleles for the SNPs for the studied gene among groups were shown in Table 2. Before analyzing on single locus, we identified a set of covariates including age, hypertension, diabetes, BMI, HDL-C and TC needed to be adjusted by conducting full model logistic regression. Then we performed full model logistic regression to test whether the single SNPs were associated with CHD with adjustment for the six identified covariates or without. The results did not show association for any genotyped SNPs with CHD whether adjusted for covariates or not. Similarly, we identified that age, gender, diabetes status and BMI were significantly associated with plasma lipids by con-
ducting full model general linear regression, irrespective of genetic variants. Thus we made adjustment for the four covariates when performing analysis on effects of SNPs on lipids. The results in the controls excluding seventeen individuals taking lipid-lowering drugs showed that the rs4822063 of SREBP2 was associated with plasma LDL-C and TC after adjustment for identified covariates. The genotype CC carriers had higher LDL-C and TC than the major allele G carriers, but the significant relationship disappeared after adjustment for multiple tests (shown in Table 3).
3.3. Multi-loci interaction analyses All the genotyped SNPs were involved in MDR analysis for testing association between the SNPs and CHD, and the results for 2-locus to 4-locus were listed in Table 4. We found that the 2-locus model involving the rs10271719 and rs9719268 of the INSIG1 gene showed the highest level of testing accuracy (56.09%, p = 0.002 on 1000 permutations) and the best 10-fold cross-validation consistency. Moreover, other two models involving INSIG1, INSIG2 and SCAP genes also showed significant prediction accuracy but lower values than the 2-locus model.
Table 2 Distribution of genotypes and alleles in controls and cases with CHD. . Gene
SNPs
Genotype
Control
Case
INSIG1
rs10271719
GG/GT/TT G/T GG/GT/TT G/T AA/AG/GG A/G
0.504/0.419/0.077 0.713/0.287 0.777/0.208/0.014 0.882/0.118 0.761/0.219/0.020 0.870/0.130
0.516/0.385/0.099 0.708/0.292 0.752/0.228/0.019 0.867/0.133 0.736/0.251/0.013 0.861/0.138
AA/AG/GG A/G TT/TC/CC T/C AA/AG/GG A/G TT/TC/CC T/C
0.779/0.207/0.014 0.882/0.118 0.419/0.458/0.123 0.648/0.352 0.563/0.379/0.058 0.753/0.247 0.252/0.507/0.241 0.505/0.495
0.782/0.206/0.012 0.885/0.115 0.430/0.451/0.118 0.657/0.343 0.572/0.373/0.055 0.758/0.242 0.266/0.489/0.245 0.510/0.490
TT/TC/CC T/C GG/GA/AA G/A
0.728/0.253/0.019 0.854/0.146 0.276/0.492/0.232 0.522/0.478
0.727/0.254/0.019 0.854/0.146 0.260/0.507/0.233 0.514/0.486
GG/GC/CC G/C
0.619/0.330/0.051 0.784/0.216
0.625/0.327/0.048 0.789/0.211
rs9719268 rs9769826 INSIG2
rs10197745 rs4848492 rs17047757 rs9308762
SCAP
rs17079634 rs4858868
SREBP2
rs4822063
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
425
Table 3 Association between SREBP2-rs4822063 and plasma lipids level in controls.# SNP rs4822063 # *
Genotype
LDL-C (mmol/L)
CC GG + GC
3.44 ± 0.90 3.17 ± 0.84
TC (mmol/L) 5.43 ± 1.06 5.13 ± 0.97
*
P = 0.038
P* = 0.051
Excluding individuals taking lipid-lowering drugs. Adjusted for age, gender, diabetes, BMI.
Table 4 Multi-loci interaction models by MDR and logistic regression analyses. Locus No. and combination
2-Locus: INSIG1-rs10271719-rs9719268 3-Locus: INSIG1-rs10271719-rs9719268 INSIG2-rs17047757 4-Locus: INSIG1rs10271719-rs9719268 INSIG2-rs9308762SCAP-rs4858868 * # a
MDR analysis
Logistic regression
Cross-validation consistency
Testing accuracy (%)
p-value*
Unadjusted p-valuea
Adjusted p-value# , a
10 8.3
56.09 55.56
0.002 0.004
≤0.001 0.045
≤0.001 0.026
9.9
54.60
0.03
0.276
0.178
p-value based on 1000 permutation. Adjusted for age, BMI, hypertension, diabetes, TC and HDL-C. p-values for interaction.
Subsequently, we performed LRM to evaluate the models from MDR. The 2-locus, 3-locus models resulted from MDR still remained significant interaction whether with adjustment for identified six risk factors of CHD or without (unadjusted and adjusted p-values for interaction were < 0.001 and <0.001 for 2-locus, 0.045 and 0.026 for 3-locus, respectively). But the 4-locus model did not show statistical significance (shown in Table 4). 4. Discussion The aim of the present study was to investigate the association between the common variants from the SREBP2 activating-related pathway and CHD. The results of the study indicated that there might be potential interaction on the risk of CHD between the genes from the focused pathway; and the SREBP2 gene was likely associated with plasma lipids level. Previously Miserez et al. [13] discovered a C to G transversion in exon 10 of the SREBP2 gene which caused a substitution of alanine to glycine (A595G) at amino acid position 595 on the protein level. The authors suggested that there was an impaired interaction and/or decreased stability of the complex between SCAP and the COOH-terminal domain of the SREBP2-595A isoform compared to the SREBP2-595G isoform. The SREBP2-595A isoform (genotype AA) was associated with significantly higher plasma TC concentrations in hypercholesterolemic subjects. Klos and Sing [18] reported that a single haplotype of SREBP2 was associated with apoB level in whites. In our study, we found that the subjects with genotype CC, same as previous reported 595A-homozygous, had higher TC and LDL-C than the major allele G carriers. Based on those findings and importance in lipid metabolism, we inferred that the SREBP2 was associated with plasma lipid levels. However, our results showed only the SREBP2 gene, but not the INSIG1, INSIG2 or SCAP genes, was associated with plasma lipid levels in controls. The results might be due to that the SREBP2 gene had more direct effect on cholesterol metabolism than other genes in the focused pathway by activating related genes transcription (shown in Fig. 1). MDR and LRM have individual characteristics when dealing with multi-locus interaction analysis. MDR, as a non-parametric method, does not require or assume any specific parametric form as well as genetic model for the relation between independent and dependent variables. LRM, as a parametric approach, requires specific parametric form as well as special genetic model and can detect
only low-order interactions as the model complexity increases with the order of interactions. However, MDR exhaustively searches across all multi-locus genotypes and identifies just one that best predicts the disease outcome which might miss other existing interactions due to interaction between dependent variables. Furthermore, the interaction model resulted from MDR does not seem to have clear biological interpretation but that is easy to get by LRM. Therefore, one can use both two methods as well as other methods to examine the gene–gene interaction. The strategy using both two methods had been performed by Briollais et al. to examine gene–gene interaction in breast cancer susceptibility [19] and by Qi et al. to identify the gene–gene interactions between HNF4A and KCNJ11 in predicting Type 2 diabetes in women [20]. The results from the MDR (non-parametric) and LRM (parametric) analyses consistently identified potential interaction between SNP rs10271719 and rs9719268 of the INSIG1 gene. In previous study, we found that the haplotype of the INSIG1 gene including those two SNPs was associated with CHD [21]. Combining the findings, we concluded that the INSIG1 gene was associated with CHD. In addition, besides the 2-locus model with the best prediction accuracy and cross-validation consistency and the least permutation p value, other two models, a 3-locus and a 4-locus model, were shown with permutation p-value less than 0.05 as well (shown in Table 4). With further LRM analysis on those models, we identified the significant 3-locus interaction model form MDR involving INSIG1, INSIG2 genes. The results from MDR and LRM confirmed our hypothesis that the genes from SREBP2 activating-related pathway might have individual or interactive effect on risk of CHD. Our results also suggested that the use of multiple statistical approaches rather than a single methodology could be the best strategy to elucidate complex gene interactions that had generally very different patterns. A major strength of the study was that the gene–gene interactions were consistently examined in analyses using different statistical models (parametric and non-parametric). However, there were a few limitations in our present study. Firstly, instead of selecting htSNPs for the SREBP2 gene, we had just chosen one nonsynonymous-coding SNP that might decrease the power to detect the association between the SREBP2 gene and CHD. Secondly, we had just selected the SREBP2 gene, but not other genes, from the SREBPs family as studied candidate gene in our study. Through biologic functional analyses and reviewing findings from previous studies, we supposed the SREBP2 had more possibility to
426
X. Liu et al. / Atherosclerosis 208 (2010) 421–426
be associated with CHD. At the same time, it might lead to loss some important information. We hope further studies that will evaluate the effect of the SREBP genes on lipid metabolism and CHD, involving more SREBP genes and more SNPs covering whole genes. Moreover, we looked only at survivors of a hospitalized CHD patients so we had therefore omit the subjects that died in hospital thus might weaken the association between the SNP and CHD, so we should explain the results careful and need prospective cohort study to test our results. Our results from MDR and LRM provided evidence of interaction between the genes from SREBP2 activating-related pathway on risk of CHD for first time so our results were difficulty to generalize to other population due to lacking additional evidences from others. In addition, it also needed further replication in various populations due to complex genetic background between different populations even our result that the SREBP2 gene was associated with plasma lipid level was similar with other results from others. Our findings warranted further studies to replicate our results in other peoples and to elucidate the biological mechanism. Acknowledgments This work was supported by National Basic Research Program of China (Grant No. 2006CB503805). Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.atherosclerosis.2009.08.011. References [1] Brown MS, Goldstein JL. The SREBP pathway: regulation of cholesterol metabolism by proteolysis of a membrane-bound transcription factor. Cell 1997;89(3):331–40. [2] Horton JD, Shah NA, Warrington JA, et al. Combined analysis of oligonucleotide microarray data from transgenic and knockout mice identifies direct SREBP target genes. Proc Natl Acad Sci USA 2003;100(21):12027–32. [3] Brown MS, Goldstein JL. A proteolytic pathway that controls the cholesterol content of membranes, cells, and blood. Proc Natl Acad Sci USA 1999;96(20):11041–8.
[4] Hua X, Nohturfft A, Goldstein JL, et al. Sterol resistance in CHO cells traced to point mutation in SREBP cleavage-activating protein. Cell 1996;87(3):415–26. [5] Sakai J, Nohturfft A, Cheng D, et al. Identification of complexes between the COOH-terminal domains of sterol regulatory element-binding proteins (SREBPs) and SREBP cleavage-activating protein. J Biol Chem 1997;272(32):20213–21. [6] Brown AJ, Sun L, Feramisco JD, et al. Cholesterol addition to ER membranes alters conformation of SCAP, the SREBP escort protein that regulates cholesterol metabolism. Mol Cell 2002;10(2):237–45. [7] Yang T, Espenshade PJ, Wright ME, et al. Crucial step in cholesterol homeostasis: sterols promote binding of SCAP to INSIG-1, a membrane protein that facilitates retention of SREBPs in ER. Cell 2002;110(4):489–500. [8] Yabe D, Brown MS, Goldstein JL. Insig-2, a second endoplasmic reticulum protein that binds SCAP and blocks export of sterol regulatory element-binding proteins. Proc Natl Acad Sci USA 2002;99(20):12753–8. [9] Goldstein JL, DeBose-Boyd RA, Brown MS. Protein sensors for membrane sterols. Cell 2006;124(1):35–46. [10] Robinet P, Vedie B, Chironi G, et al. Characterization of polymorphic structure of SREBP-2 gene: role in atherosclerosis. Atherosclerosis 2003;168(2):381–7. [11] Friedlander Y, Schwartz SM, Durst R, et al. SREBP-2 and SCAP isoforms and risk of early onset myocardial infarction. Atherosclerosis 2007. [12] Duan X, Zhu W, Li Y, et al. The effect of sterol regulatory element-binding protein 2 polymorphism on the serum lipid in northern Chinese subjects. J Lipid Res 2005;46(2):252–7. [13] Miserez AR, Muller PY, Barella L, et al. Sterol-regulatory element-binding protein (SREBP)-2 contributes to polygenic hypercholesterolaemia. Atherosclerosis 2002;164(1):15–26. [14] Zhou X, Huang J, Chen J, et al. Thrombospondin-4 A387P polymorphism is not associated with coronary artery disease and myocardial infarction in the Chinese Han population. Clin Sci (Lond) 2004;106(5):495–500. [15] Gabriel SB, Schaffner SF, Nguyen H, et al. The structure of haplotype blocks in the human genome. Science 2002;296(5576):2225–9. [16] Hahn LW, Ritchie MD, Moore JH. Multifactor dimensionality reduction software for detecting gene-gene and gene-environment interactions. Bioinformatics 2003;19(3):376–82. [17] Ritchie MD, Hahn LW, Roodi N, et al. Multifactor-dimensionality reduction reveals high-order interactions among estrogen-metabolism genes in sporadic breast cancer. Am J Hum Genet 2001;69(1):138–47. [18] Klos KL, Sing CF, Boerwinkle E, et al. Consistent effects of genes involved in reverse cholesterol transport on plasma lipid and apolipoprotein levels in CARDIA participants. Arterioscler Thromb Vasc Biol 2006;26(8):1828–36. [19] Briollais L, Wang Y, Rajendram I, et al. Methodological issues in detecting genegene interactions in breast cancer susceptibility: a population-based study in Ontario. BMC Med 2007;5:22. [20] Qi L, van Dam RM, Asselbergs FW, et al. Gene-gene interactions between HNF4A and KCNJ11 in predicting Type 2 diabetes in women. Diabet Med 2007;24(11):1187–91. [21] Liu X, Li Y, Wang L, et al. The INSIG1 gene, not the INSIG2 gene, associated with coronary heart disease: tagSNPs and haplotype-based association study. The Beijing Atherosclerosis Study. Thromb Haemost 2008;100(5):886–92.