Methylation of p15INK4b and Expression of ANRIL on Chromosome 9p21 Are Associated with Coronary Artery Disease

Background Genome-wide association studies have identified that multiple single nucleiotide polymorphisms on chromosome 9p21 are tightly associated with coronary artery disease (CAD). However, the mechanism linking this risk locus to CAD remains unclear. Methodology/Principal Findings The methylation status of six candidate genes (BAX, BCL-2, TIMP3, p14ARF, p15INK4b and p16INK4a) in 205 patients and controls who underwent coronary angiography were analyzed by quantitative MethyLight assay. Rs10757274 was genotyped and expression of INK4/ARF and antisense non-coding RNA in the INK4 locus (ANRIL) was determined by real-time RT-PCR. Compared with controls, DNA methylation levels at p15INK4b significantly increased in CAD patients (p = 0.006). To validate and dissect the methylation percentage of each target CpG site at p15INK4b, pyrosequencing was performed, finding CpG +314 and +332 remarkably hypermethylated in CAD patients. Further investigation determined that p15INK4b hypermethylation prevalently emerged in lymphocytes of CAD patients (p = 0.013). The rs10757274 genotype was significantly associated with CAD (p = 0.003) and GG genotype carriers had a higher level of ANRIL exon 1–5 expression compared among three genotypes (p = 0.009). There was a stepwise increase in p15INK4b and p16INK4a methylation as ANRIL exon 1–5 expression elevated (r = 0.23, p = 0.001 and r = 0.24, p = 0.001, respectively), although neither of two loci methylation was directly linked to rs10757274 genotype. Conclusions/Significance p15INK4b methylation is associated with CAD and ANRIL expression. The epigenetic changes in p15INK4b methylation and ANRIL expression may involve in the mechanisms of chromosome 9p21 on CAD development.


Introduction
Genome-wide association studies (GWAS) have found that single nucleotide polymorphisms (SNPs) on chromosome 9p21 (Chr9p21) affect susceptibility to coronary artery disease (CAD) in Caucasian population [1,2,3,4], and these associations have been reproduced in other populations [5,6,7,8]. However, the mechanisms of Chr9p21 for CAD remain elusive. Most of SNPs are highly correlated and located within a roughly 53-kb linkage disequilibrium (LD) region in which a long non-coding RNA, known as antisense non-coding RNA in the INK4 locus (ANRIL), is transcribed. It has been extensively reported that ANRIL transcripts are assembled with many exons and multiple isoforms of ANRIL transcripts coexist in a variety of cell types [9]. The genetic sequences upstream to Chr9p21 encode a well-characterized cluster of tumor suppressor genes, p14 ARF , p15 INK4b and p16 INK4a , alias INK4/ARF, all of which are transcribed from the opposite strand to ANRIL (Figure 1).
Epigenetics is defined as stable and heritable changes that are not due to disrupting the coding sequences of disease genes which has been shown to play an important role in various diseases including cancer, type 2 diabetes, systemic lupus erythematosus, etc. [21,22,23]. Cancer cells armed with both a loss of global methylation and a gain of methylation at tumor suppressor genes, such as INK4/ARF, often show unlimited proliferation [24,25]. Atherosclerostic plaque also has characteristics of excessive vascular smooth muscle cells (VSMCs) and macrophages proliferation. Inspired by these evidences and the strong association of Chr9p21 with CAD, we hypothesized that INK4/ARF hypermethylation also had its role in CAD development. Due to the little knowledge about the role of DNA methylation of specific loci in chromosome on cardiovascular disease so far, this study would help to explore a novel aspect in understanding cardiovascular disease.
In this study, we sought to determine whether DNA methylations of selected CpG islands, especially those in INK4/ARF locus, were involved in CAD. Since multiple SNPs associated with CAD do not appear to directly affect INK4/ARF expression, the altered expression of INK4/ARF is likely to be modulated by ANRIL or other epigenetic changes. Toward this end, we attempted to explore the association of DNA methylation with risk genotypes and altered ANRIL expression on Chr9p21.

Ethnic statement
This study was approved by the Ethics Committee of Shanghai Tenth People's Hospital. All patients gave written informed consent.

Study population
A total of 95 patients who were diagnosed as CAD by angiography and 110 sex-and age-matched participants without CAD after angiography were recruited in Department of Cardiology, Shanghai Tenth People's Hospital from March 2011 to October 2011. Of note, patients with cancer, acute myocardial infarction, severe heart failure (left ventricular ejection fraction #30%), cardiomyopathy, active infection and connective tissue disease were excluded. Hypertension was defined as systolic or diastolic blood pressure $140/90 mm Hg, under antihypertensive medications for one year before admission. Diabetes was defined as fasting blood glucose $7 mmol/L, non-fasting plasma glucose level $11.1 mmol/L or known treatment for diabetes. Peripheral venous blood (20 mL) was drawn into adequate tubes from each subject. A white differential cell count on whole blood using automated counter was performed.

Coronary angiography
Quantitative assessment of CAD was performed using coronary angiography as previously described [26]. In brief, significant CAD was defined as the presence of luminal diameter narrowing $50% in the left anterior descending artery, left circumflex artery, right coronary artery and their main branches. Left main trunk stenosis was considered as two-vessel disease. Severity of coronary atherosclerosis was further categorized as 1-, 2-or $3-vessel disease according to number of coronary vessels with significant stenosis.

Isolation of neutrophils and lymphocytes from peripheral blood
Human neutrophils and lymphocytes were isolated from 5 ml heparin-anticoagulated blood drawn from 26 healthy participants and 38 CAD patients as previously described [27], with minor modifications. Briefly, the 1:1 mixtures of peripheral blood and HBSS (without Ca 2+ and Mg 2+ ) were added into 2 ml Ficoll-Paque Plus (GE Healthcare, USA), followed by centrifugation at 1,500 rpm for 15 min without brake. Lymphocytes were collected at the interphase and neutrophils were collected by carefully removing the layer immediately above the red blood cells, followed by addition of 6% Dextran 500 prepared in 0.9% NaCl solution. After allowing RBC to settle for 30 to 60 min at room temperature, neutrophils in the supernatant were harvested. The cell type and purify were appraised by fluorescence-activated cell sorting as CD45-high SSC-low for lymphocytes and CD45-low SSC-high for neutrophils.

DNA extraction and genotyping
All peripheral blood samples were taken in the morning with patients fasting from midnight onward. Genomic DNA was extracted from whole blood cells, neutrophilis and lymphocytes using commercial available kit (Tiagen Biotech, Beijing, China). Based on the validation of GWAS and independent studies [2,12], we selected SNP rs10757274 for genotying, which is a representative marker of atherosclerostic diseases as Chr9p21 was a highly LD region. Genotyping was performed with TaqMan allelic discrimination by means of an ABI 7900HT (Applied Biosystems, CA, USA), in 384-well format. The TaqMan Assay kits as well as probes were purchased from Applied Biosystems. Data were analyzed using the ABI Prism SDS software version 2.3.

Bisulfite treatment and MethyLight analysis
Genomic DNA was then chemically modified by sodium bisulfite to convert all unmethylated cytosines to uracils while leaving methylcytosines unaltered (EZ Zymo Methylation Kit, Zymo Research, CA, USA). DNA methylation analysis was performed by MethyLight as previously described [28]. The gene names, locations, primer and probe sequences are summarized in Table S1. The b-Actin (ATCB) repeats were used as an internal reference to normalize the input DNA and to generate a standard curve. The amount of methylated DNA was determined by the threshold cycle number (Ct) for each sample, compared against a standard curve generated from CpGenome Universal Methylated DNA (Chemicon International Inc, CA, USA). The percentage methylated of reference (PMR) value was calculated by dividing the GENE/ACTB ratio of a sample by the GENE/ACTB ratio of a positive control, CpGenome Universal Methylated DNA, and multiplying by 100. The methylation status of each sample was determined as positive when PMR.4 [29].

Pyrosequencing quantitative methylation analysis
Pyrosequencing was applied to validate and dissect the methylation alterations in the observed target CpG regions of p15 INK4b according to the manufacturer's instructions. Briefly, a biotin-labeled primer and bisulfite converted DNA were mixed and performed with PCR, allowing for isolation of the amplicon.
Subsequently, the PCR products were denaturated and released to single strand products for pyrosequencing using the PyroMark Q24 system (Qiagen, Hilden, German). DNA methylation percentage at each CpG site was analyzed by PyroMark Q24 version 1.0.10 software in the CpG analysis mode. The p15 INK4b forward primer: GGG AGG GTA ATG AAG TTG AGT; reverse primer: Biotin-CTA CCC CCC CCA CTA AAC ATA CCC TTA T; sequencing primers: TTG AGT TTA GGT TTT TTA GGA and GGA GTA GAG TGG GAA AGA A.

RNA isolation and quantitative reverse-trasnscript polymerase chain reaction (RT-PCR)
Total RNA from whole blood cells was extracted using an RNeasy Mini Kit (Qiagen, Hilden, Germany) and 1 mg RNA was reverse transcribed with a PrimeScript RT reagent Kit (Takara Biotechnology, Tokyo, Japan). Relative quantification of gene expression was performed in duplicate. Specific primers for p14 ARF , p15 INK4b , p16 INK4a and GAPDH were designed for relative quantitative RT-PCR (Table S2). The mRNA expression was determined using SYBR Premix Ex Taq TM (Takara Biotechnology, Tokyo, Japan). Ct values for each target gene were normalized to GAPDH. Given the low expression levels and the established splice variants of ANRIL transcripts, three specific primers and probes as previously designated were used (Table S2) [15]. As marked in Figure 1, these splices were proposed to indicate the expression of proximal, central and distal exons of ANRIL, since there is devoid of a deep understanding of ANRIL splice variants and an economic high-throughput approach to analyzing the full length of long non-coding RNA.

Statistical analysis
The normal distribution of data was tested by Kolmogorov-Smirnov test. While the values of PMR and mRNA expression were highly skewed, Mann-Whitney U test was undertaken to examine differences between two groups. The significant differences between categorical variables were determined using x 2 test. Correlations between p14 ARF , p15 INK4b , p16 INK4a and ANRIL expression were tested using the Spearman's nonparametric correlation test. ANOVA test and Bonferroni correction were then used to compare gene expression and PMR values across the genotypes. Logistic regression analysis including environmental and genetic risk factors (ie., age, gender, smoking, hypertension, diabetes, lipid profiles, genotype, candidate gene expression and methylation) was performed to identify the independent determinants of CAD. All statistics were performed with SPSS 14.0 (SPSS Inc, Chicago, IL, USA). A value of p,0.05 was considered significant (two tailed).

Methylation of candidate genes and expression of INK4/ ARF and ANRIL in patients with CAD
Baseline characteristics of the patients and controls are listed in Table 1. Their mean age was 64 years, 56.1% were male, 31.7% had a history of diabetes and 66.8% had a diagnosis of hyptertension. Apart from statin treatment in a pre-hospital setting (p,0.001), no significant differences in baseline characteristics were seen between the CAD patients and controls. Similarly, there was no discrepancy in the peripheral total white blood cell count, differential count and red blood cell profile between two groups (Table S3). First of all, for the purpose of identification of  candidate hypermethylated genes in peripheral blood cells of CAD patients, we searched previous studies and explored whether these genes were hypermethylated in CAD patients. Seven predefined apoptosis-related genes (LOX-1, CASP3, BCL-2, BAX, TIMP3, ANXA5 and cIAP-3), which were found hypermethylated in vitro [30], and INK4/ARF were screened for CpG island methylation events in a cohort of 40 CAD patients using methylation-specific PCR. While four genes were absences of methylation (LOX-1, CASP3, ANXA5 and cIAP-3, data not shown), we then applied quantitative analyses to determine the methylation levels of the remaining three genes and INK4/ARF. As delineated in Figure 2, the methylation levels of BCL-2, BAX, TIMP3 and p14 ARF were at barely detectable levels in both CAD patients and controls. p15 INK4b was the only one showing significant methylation among candidate genes in CAD patients. Next, the gene expression and methylation on Chr9p21 were compared between CAD patients and controls (Table 2). A dramatically increase in methylation levels of p15 INK4b in the CAD group was observed compared with control (p = 0.006). CAD patients had a higher expression level of p14 ARF (p = 0.048), yet no differences in p16 INK4a methylation and expression of other genes were observed. Multivariate regression analysis showed that rs10757274 (OR = 1.80; 95% CI: 1.21-3.00), p15 INK4b methylation (OR = 2.55; 95% CI: 1.26-5.01) and statin were the independent determinants of CAD (Table 3). For the purpose of validation and dissection of p15 INK4b methylation percentages at interested CpG sites, we then applied pyrosequencing in 22 randomly selected controls and 22 CAD patients. DNA methylation percentages were obtained for seven CpGs covering 81 bp of p15 INK4b ( Figure 3A). As shown in Figure 3B, five analyzed CpG sites, located 269, 272, 280, 303 and 321 bp downstream (CpG +269, +272, +280, +303 and +321) of the transcription start site, were slightly methylated in both groups without between-group difference, whereas the degrees of methylation at CpG +314 and +332 were significantly increased in CAD patients compared with controls (p = 0.01 and 0.03, respectively). Additionally, the pyrosequencing results at CpG +314 and +332 were tightly correlated with p15 INK4b methylation levels measured by Methy-Light assay ( Figure 3C).
To determine which types of leukocytes contributed to p15 INK4b methylation in peripheral blood cells, genomic DNA was extracted from whole blood cells, purified neutrophils and lymphocytes. Neutrophils and lymphocytes from peripheral blood were .93% pure by fluorescence-activated cell sorting and morphology ( Figure 4C). Baseline characteristics of the second sample are listed in Table S4. As shown in Figure 4A, a significant increase in p15 INK4b methylation was observed in both whole blood cells and lymphocytes (p = 0.011 and 0.013, respectively). Likewise, p15 INK4b methylation in lymphocytes was highly correlated with that in whole blood cells (r = 0.54, p = 0.002) ( Figure 4B).
We further investigated the association between the number of culprit vessels and ANRIL expression and p15 INK4b /p16 INK4a methylation. Neither p15 INK4b nor p16 INK4a had statistically significant difference in methylation among the severity of CAD, whereas there was a gradual increase in ANRIL exon 4-5 expression as number of culprit vessels increased (p = 0.042) ( Figure S1).   differences in other gene expression and DNA methylation among three genotypes of rs10757274 (Table 4).

Correlation among p15 INK4b /p16 INK4a methylation and gene expression on Chr9p21
There was a strong correlation among INK4/ARF expression (r = 0.28 to 0.84, p,0.001) ( Figure 5). p15 INK4b and p16 INK4a methylation was inversely correlated with their corresponding genes expression (r = 20.17, p = 0.011 and r = 20.26, p = 0.002, respectively), which was in accordance with the theory that aberrant methylation of the CpG islands at promoters and exons is linked to loss of genes expression and their function. The significant associations of p15 INK4b /p16 INK4a methylation with serum levels of lipid profile or glucose were not found (data not shown).

Discussion
This work presents the first quantitative analysis of specific genes methylation in CAD. Our data indicated that p15 INK4b methylation was an important event in atherosclerosis, and such potential bridge between genotype and p15 INK4b methylation might be mediated by altered expression of ANRIL.
Although epigenetic changes are of crucial importance in the pathophysiology of atherosclerosis in response to multiple genetic and modifiable risk factors, there is still little data about the methylation status at specific loci [31]. Studies attempting to explore global methylation in patients with cardiovascular diseases did not reach a consensus, which mainly resulted from variance in subjects' selection criteria of each study [32,33,34,35]. In the present study, CAD patients diagnosed by angiography and age-, sex-and concomitant diseases-matched participants were enrolled, thereby avoiding the confounding effects of established risk factors on DNA methylation. Additionally, the results from previous studies failed to identify whether the methylation events emerged at specific loci. A previous model suggested that the exposure of HUVECs to ox-LDL induced other candidate gene methylation, yet these results could not be replicated in CAD patients so far [30].
INK4/ARF transcripts participate in the regulation of cell cycle arrest via p53 and Rb pathways and play important roles in cell proliferation and senescence [36]. In our study, p15 INK4b , whose hypermethylation has been proved to involve in the initiation and development of multiple types of cancers, was also strikingly hypermethylated in CAD patients compared with controls. Because there is no discrepancy observed in cellular composition of the blood samples between two groups, it is thus plausible that the changes in p15 INK4b methylation and gene expression depend on alterations in one or multiple given blood cell types. Therefore, another central conundrum is which types of leukocytes in peripheral blood are responsible for p15 INK4b methylation since epigenomic alterations vary from cell type to cell type in contrast to genetic variants. Pointing to this question, our results showed highly parallel levels of p15 INK4b methylation observed in whole blood cells and lymphocytes after careful collections of neutrophils and lymphocytes from peripheral blood. Among the differential leukocytes, lymphocytes have been considered to mediate immune and inflammatory response through altered DNA methyltransferases expression and alterations to inflammation-related DNA methylation [37,38]. Furthermore, the observations that significantly increased p15 INK4b methylation in CAD patients coincided with decreased p15 INK4b and p16 INK4a expression could be predominantly explained by the fact that p15 INK4b methylation could repress INK4/ARF expression, subsequently contributing to atherosclerosis. Prior studies conducted in INK4/ARF knockout mice model also found that either deficiency of p14 ARF or p16 INK4a was closely associated with atherosclerosis, thus providing an elegant rationale for our preliminary results [39,40]. Nevertheless, the p15 INK4b and p16 INK4a expression failed to be inversely correlated with CAD. Indeed, notwithstanding a significant increase of p15 INK4b methylation in CAD patients compared to that in controls, the average methylation level in CAD patients modestly exceeded the normal value. The significant but rather small changes in p15 INK4b methylation were subsequently dissected by pyrosequencing, finding that two of the seven observed CpGs (CpG +314 and +332) were markedly hypermethylated in CAD patients compared with controls. Indeed, the identification of CpG +314 and +332 hypermethylation at p15 INK4b is of important value, whereas two sites seem incapable of remarkably attenuating p15 INK4b expression in CAD patients compared with controls.
All available evidence to date have indicated that long noncoding RNAs attenuated expression of associated genes through diverse mechanisms such as heterochromatin formation, histone modifications, RNA interference, DNA methylation, etc. [19,20,41,42,43]. Our findings presented herein suggested that specific ANRIL species had close associations with p15 INK4b and p16 INK4a methylation, giving rise to the down-regulation of the corresponding gene expression. It is demonstrated that polycomb repressor complexes (PRCs) are of particular importance in repressing INK4/ARF expression. Current in vivo and in vitro studies unveiled that known members of PRC family, such as EZH2, CBX7 and SUZ12, contributed to initiation and maintenance of DNA hypermethylation via interaction with DNA methyltransferases [44,45,46]. On the other side of the coin, ANRIL may directly recruit some components of PRC family, ensuing repression of p15 INK4b and p16 INK4a through histone modifications [19,20], which are associated with aberrant methylation [47]. It is thus conceivable that ANRIL, coupled with PRCs, may cause a decrease in expression of p15 INK4b and p16 INK4a by programming epigenome including DNA methylation and histone modifications.
Together with the knowledge of CAD-associated SNPs adjacent to INK4/ARF locus, we came to the hypothesis that individuals may have an intrinsic propensity to p15 INK4b /p16 INK4a methylation and risk SNPs located in Chr9p21 may function through the contribution of ANRIL to neighboring INK4/ARF methylation. In this regard, our results found ANRIL exon 1-5 in carriers of risk genotype was overexpressed and had a strong association with p15 INK4b /p16 INK4a methylation. Although the mechanism underly- Our findings add to the body of knowledge on epigenetic changes in CAD, mainly p15 INK4b methylation on Chr9p21. However, our study poses several limitations that deserved further consideration. First and above all, this study was retrospective in nature and increased the likelihood of selection bias, although the patients in two groups were matched on up to 15 variables to adjust for the differences in baseline data. Therefore, the INK4/ ARF methylation and its influence on risk for CAD need to be investigated by more studies such as cohort studies.
Conclusions p15 INK4b methylation is associated with CAD and ANRIL expression, both of which are directly affected by gene polymorphisms on Chr9p21. These results point to a potential role of epigenetic changes as mediators from Chr9p21 polymorphisms to CAD.