Circulating "LncPPARδ" From Monocytes as a Novel Biomarker for Coronary Artery Diseases.

To investigate long noncoding RNA NONHSAT112178 (LncPPARδ) as a biomarker for coronary artery disease (CAD) in peripheral blood monocyte cells, RT-qPCR was performed to validate the microarray results, receiver operating characteristic curve was applied to study the potential of LncPPARδ as a biomarker. Diagnostic models from LncPPARδ alone or combination of risk factors were constructed by Fisher criteria. The expression of genes neighboring the LncPPARδ gene was examined with RT-qPCR in THP-1 cell line treated with LncPPARδ siRNA. Using a diagnostic model by Fisher criteria, the consideration of risk factors increased the optimal sensitivity from 70.00% to 82.00% and decreased the specificity from 94.00% to 78.00%. The consideration of risk factors also increased area under the receiver operating characteristic curve from 0.727 to 0.785 (P = 0.001), from 0.712 to 0.768 (P = 0.01), and from 0.769 to 0.835 (P = 0.07), in the original, training, and test sets, respectively. Finally, we found that the expression of peroxisome proliferator-activated receptor δ (PPARδ), Adipose Differentiation-Related Protein (ADRP), and Angiopoietin-like 4 (ANGPTL4) were affected by LncPPARδ silencing.Our present study indicated that LncPPARδ, especially combined with risk factors, can be a good biomarker for CAD. LncPPARδ regulates the expression of neighboring protein-coding genes, PPARδ and its direct target genes ADRP and ANGPTL4.

INTRODUCTION

Cardiovascular diseases (CVD) are still a prevailing cause of death worldwide. The leading causes of the majority of CVD are atherosclerosis (AS) and its thrombotic complications. In the heart, AS causes coronary artery disease (CAD). The current therapeutic strategies for CAD, including lowering plasma low-density lipoprotein (LDL), antiplatelet agents, and/or anticoagulants following revascularization therapy, such as percutaneous coronary intervention and coronary artery bypass surgery, have proven effective, but the mortality has remained dismal. Early diagnosis of CAD may assist early interventions and improve the outcome. So an early identification of patients with CAD remains an emerging need.[1]

Long noncoding RNAs (lncRNAs), a novel class of no coding RNAs (ncRNAs), are longer than 200 nucleotides and without encoding-protein capacity.[2-4] Recently, thousands of lncRNAs have been identified in different species. Increasing evidence has suggested that lncRNAs play crucial roles in controlling gene expression and other cellular processes during developmental and differentiation processes.[5] LncRNAs can regulate gene expression at the levels of epigenetic control, transcription, RNA processing, and translation.[6,7] Several recent studies have identified some lncRNAs associated with diseased hearts, both in human patients and mouse models.[8-10] In those studies, the expression of cardiac-expressed or circulating lncRNAs was altered in patients with cardiomyopathy or heart failure.[11-14] What is more, several recent studies have shown that some lncRNAs are involved in the development of various types of CVD, such as cardiac hypertrophy,[15,16] heart failure,[12,14,17] and myocardial infarction.[18] The levels of some lncRNAs, such as myocardial infarct-associated transcript-1 (MIAT1), lincP21, and antisense noncoding RNA in the INK4 locus (ANRIL), were markedly increased in AS.[19-22] LncRNAs could serve as biomarkers for CVDs. For instance, lncRNA LIPCAR for heart failure postmyocardial infarction. Together, these findings provided evidence and support for the potential roles of lncRNAs in CAD development and progression. The expression of lncRNA might serve as biomarkers.

CAD is considered as a chronic inflammatory disease that contributes to the formation of plaques at all stages in large- and mid-sized arteries. Plaque formation results from the infiltration of peripheral blood monocytes (PBMCs) into the subendothelial space, where they differentiate into macrophages and subsequently internalize modified lipoproteins (ox-LDL), and finally differentiate into foam cells. Therefore, recruited PBMCs play a pivotal role in this event. Recently, many studies indicate that peroxisome proliferator-activated receptor δ (PPARδ) plays an important role in CADs.[23] Activation of PPARδ increases cholesterol efflux from the macrophages in the lesions and decreases transendothelial migration of leukocyte/monocytes into the arterial wall; thus reduces the atherosclerotic lesion size.[23] Moreover, stimulation of PPARδ in vivo is found to improve multiple cardiovascular risk factors, including AS, dyslipidemia, obesity, and insulin resistance, by reducing inflammation in the artery wall, improving the plasma lipid profile, decreasing adiposity, and increasing insulin sensitivity and fatty acids oxidation in the liver, skeletal muscle, and adipose tissue.[24]

Previous studies indicate that some no coding RNAs (ncRNAs) mediate the RNA-dependent transcriptional repression of their neighboring protein-coding genes in a cis-specific manner.[25] In our study, lncRNAs in the PBMCs, a populations of neutrophils that may play important roles in the inflammatory process during AS,[26] were screened by microarray analysis in a small-sized cohort with or without CAD for potential CAD biomarkers and then these biomarkers were validated in different populations. We found that a circulating lncRNA, NONHSAT112178, which is named LncPPARδ in this study, was a specific and sensitive biomarker for CAD, especially when combined with risk factors. LncPPARδ may repress its nearby protein-coding gene PPARδ, indicating its important role in the process of CAD development.

RESULTS

Independent Validation of LncPPARδ Expression in a Small CAD Cohort

The “microarray cohort” was composed of 15 male patients with or 15 males without CAD, respectively. The characteristics of these patients are shown in Table S2. The LncRNA profiles from PBMCs in these 2 groups patients differed significantly. We sorted all lncRNA transcripts that were included in the microarrays according to their average normalized intensity, which showed a normalized intensity of >6 in PBMCs, as illustrated in the hierarchical clustering shown in Figure 1. Microarray analysis indicated the expression of LncPPARδ in PBMCs from CAD patients increased more than twice compared with that of the control group. To independently validate the expression of LncPPARδ, we studied a set of PBMCs’ samples by RT-qPCR from 20 CAD patients and 20 control subjects, respectively. The characteristics of these patients are shown in the Table S3. Consistent with the result from the microarray analysis, LncPPARδ expression was increased 2.2-fold in CAD patients compared with control samples (Figure 2A). Next, receiver operating characteristic (ROC) curve analysis was performed to examine whether PBMC LncPPARδ could be used as diagnostic biomarkers for CAD using these 20 × 20 samples. It showed an area under the ROC curve (AUC) of 0.857 (Figure 2B). The result indicated that LncPPARδ might be a good candidate biomarker to predict CAD.

FIGURE 1

Heat map of lncRNA expression from microarray analysis of combined circulating monocyte samples of patients with CAD and control subjects (n = 3, PBMCs from 5 patients were pooled as 1 sample), respectively. The expression of lncRNA is hierarchically clustered on the y-axis, and CAD or control monocyte samples are hierarchically clustered on the x-axis. The relative lncRNA expression is depicted according to the color scale shown on the left. Red indicates up-regulation; green, down-regulation. “P1M-P3M” indicates CAD samples; “N1M-N3M,” control samples (A). Scatter plot of lncRNA expression in test samples versus normal samples. X-axis depicts data values of normal samples, Y-axis depicts data values of test samples. Dots located above the upper green line and below the under green line represent fold change ≥2.0. “Test” indicates CAD samples; “Normal,” control samples (B). CAD = coronary artery heart disease, PBMC = peripheral blood monocyte.

FIGURE 2

Expression levels and ROC curve analyses of LncPPARδ. (A) Expression of LncPPARδ in the PBMCs from CAD patients and control subjects determined by qRT-PCR (∗P < 0.001, vs Control, n = 20). (B) ROC curve analyses of LncPPARδ for diagnosis of CAD in pilot samples. AUC = area under the ROC curve, CAD = coronary artery heart disease, PBMC = peripheral blood monocyte, ROC = receiver operating characteristic.

Validation of Increased PBMC LncPPARδ in a Large CAD Cohort

Because above observation was made with a small population, we further studied the diagnostic value of LncPPARδ in a big population of patients (211 CAD patients and 171 control subjects) who enrolled during February 2013 to May 2014. We performed ROC curve analysis using the whole 382 subjects and found an AUC of 0.727 with a 95% confidence interval (CI) of 0.677 to 0.777 (Figure 3A). The clinical and demographic features of these patients were shown in the Table S4. These data suggested that LncPPARδ had a good potential to distinguish patients with CAD from the controls.

FIGURE 3

Diagnostic value of LncPPARδ with or without combination of other risk factors on CAD. Original set included 211 CAD patients and 171 control subjects (A); training set included 161 CAD patients and 121 control subjects (B); and test set included 50 CAD patients and 50 control subjects (C), respectively. CAD = coronary artery heart disease.

To further determine the diagnostic accuracy of LncPPARδ, we constructed a diagnostic model by Fisher criteria. The whole 382 samples were randomly divided into training and test sets. Fisher criteria was applied to establish discriminant function with the training set, and then validated by the test set.

We adopted random sampling method to construct the diagnostic model, with uniform distribution every time. The training set (n = 282) was consisted of 161 from 211 cases and 121 from 171 control samples. ROC curve analysis and discriminative function were repeatedly made by Fisher method. We repeated this procedure up to 100 times, and chose the optimal discriminant function, which corresponding to the 79th training set, as the final discriminant function. The AUC was 0.712 with a 95% CI of 0.653 to 0.771 (Figure 3B). The signature was defined as follow: f = −1.10492 × +1.239687, where “x” denoted the expression of LncPPARδ in PBMCs, a subject was classified as “CAD” if f < 0 according to the patient's LncPPARδ expression value and as “control” if not. The characteristics of these samples were shown in Table 1 and the 2 groups of patients differed significantly in sex (P = 0.04), age, tobacco and alcohol use (P < .001), arterial hypertension, and medicine treatment, including angiotensin receptor blocker (ARB), calcium channel blocker (CCB), and β-blocker (P = 0.001), respectively, but not in other clinical and pathological factors.

TABLE 1

Clinical Characteristics of Patients With Training and Testing Group

The lncRNA signature was then validated for its diagnostic accuracy in the test set of 100 patients. The same model and criteria as those done for the training set were applied. After running for 100 times, on average it correctly classified 28 and 40 patients of the test group into the case and control groups, respectively, corresponding to average sensitivity and specificity of 55.14% and 80.16%, respectively (Table S5-1). The optimally corrected classified number (79th time) for “CAD” cases and controls was 35 and 47, respectively, corresponding to the optimal sensitivity and specificity of 70.00% and 94.00%, respectively (Table S5-2). Then, we constructed ROC curves for continuous predictor with the test set to estimate the diagnostic sensitivity and specificity of LncPPARδ. The AUC was 0.769 with a 95% CI of 0.677 to 0.861 (Figure 3C). The 2 groups of patients differed significantly in sex (P = 0.01), height (P = 0.007), arterial hypertension (P = 0.01), CCB (P < 0.001), and ARB (P = 0.01), respectively, but not in other clinical and pathological factors (Table 1).

Correlation of LncPPARδ Level With Demographic and Clinical Factors

To examine whether or not the diagnostic value of LncPPARδ signature was independent of other risk factors for CAD, including age, sex, body mass index (BMI), alcohol use, tobacco use, hypertension, hyperlipidemia, diabetes mellitus, obstructive degree, and medication history (statin, angiotensin-converting enzyme inhibitor [ACEI], ARB, CCB, β-blocker, and antiplatlet), we analyzed the relationship between the LncPPARδ expression and these risk factors by Kruskal–Wallis H test and Mann–Whitney U test. Obvious differences were observed when CAD cases were stratified by sex, hypertension, tobacco use, and alcohol use history. No difference was found for expression of LncPPARδ in both control and CAD group patients with other risk factors, as shown in Table 2.

TABLE 2

Effects of Clinical Risk Factors of Patients With Coronary Artery Diseases on lncRNA Signature in 3 Sets

Increased Diagnostic Prediction of LncPPARδ After Combination With Risk Factors

As above indicated, we found correlations between PBMC LncPPARδ levels and sex, hypertension, tobacco use, and alcohol use. To show whether or not those factors had an additive effect on the prediction specificity and sensitivity, we made another ROC curve analysis of LncPPARδ in combination of these risk factors in the whole 382 samples. It resulted an increased diagnostic prediction compared to LncPPARδ alone with an AUC of 0.785 (95% CI: 0.740–0.830; P = 0.001; Figure 3A). Then, we constructed a diagnostic model with LncPPARδ in combination of these risk factors according to the same method and criteria as the former. The optimal discriminative function was obtained with the 32nd training set. The optimal AUC was 0.768 (95% CI: 0.714–0.822; Figure 3B). The corresponding signature was defined as follow: f = −0.69124x1 − 0.79499x2 − 0.15350x3 + 0.30563x4 − 0.33830x5 + 1.11826, where “x1” denoted the expression of LncPPARδ, “x2,” “x3,” “x4,” and “x5” indicated hypertension, sex, tobacco use, and alcohol use, respectively. The value of “x2,” “x3,” “x4,” and “x5” was “1” or “0,” respectively. Hypertension, male, smoking, and drinking, each was defined “1,” otherwise, it was “0.” A patient was classified as “CAD” if f < 0 according to the patient's LncPPARδ expression value and the risk factors, and as “control” if not. The clinical and pathological characteristics of the 2 groups of patients were shown in Table 3 .

TABLE 3

Clinical Characteristics of Patients in Training and Test Group (32nd)

Then, the lncRNA signature was also tested for its diagnostic accuracy with the test set using the same model and criteria as those derived from the training set. The AUC was 0.835 (95% CI: 0.757–0.913; Figure 3C). On average, after 100 runs, it correctly classified 33 and 37 patients of the test group into the case and control groups, respectively. The corresponding average sensitivity and specificity was 66.42% and 74.86%, respectively (Table S6-1). And the optimal discrimination number (32nd time) was 41 and 39 patients of the test set into the case and control groups, respectively. The corresponding sensitivity and specificity was 82.00% and 78.00%, respectively (Table S6-2). The clinical and pathological characteristics between the 2 groups were as mentioned above (Table 3 ).

Finally, to compare the sensitivity and specificity in diagnostic accuracy between the models constructed by LncPPARδ alone or LncPPARδ in combination of risk factors, we performed ROC curve analysis. According to the diagnostic models, in the original set and training set, there were significantly enhanced predictive values by the combination with risk factors (P = 0.001 and P = 0.01, respectively; Figure 3A and B). In the test set, the combined model had a higher AUC than the single signature (0.835 vs 0.769, Figure 3C) but no significantly different predictive ability was found (P = 0.07), probably owing to limited sample size. Taken together, the optimal sensitivity in the combined model increased significantly from 70.00% to 82.00%, and the specificity decreased from 94.00% to 78.00%, respectively (Tables S5-2 and S6-2), indicating that the diagnostic accuracy of the combined model was enhanced significantly for CAD discrimination.

Application of Biomarker LncPPARδ Alone or in Combination With Risk Factors in Discriminative Diagnosis of Cases of Clinical Suspect of CAD

Using the same 90 clinically suspected cases of CAD, we further validated the accuracy of biomarker LncPPARδ alone or in combination with risk factors for CAD prediction in a prospective fashion. First, we analyzed the values of single LncPPARδ or combined model (expression of LncPPARδ in these PBMC samples with risk factors) for these patients, and classified them on the basis of above discriminative functions. All 90 cases were then followed and determined by coronary angiography diagnosis. Results from the final angiography diagnosis served as a golden standard (100% accuracy) to test the accuracy of the LncPPARδ profile alone or combination model for CAD diagnosis. As shown in Table S7-1, it could correctly classify 34 and 36 patients into the case and control groups with the single LncPPARδ model, respectively, the corresponding sensitivity and specificity was 68.00% and 90.00%, respectively. While by the combination model, it correctly classified 38 and 33 patients into the case and control groups, respectively, the corresponding optimal sensitivity and specificity of the combination model among these suspicious cases was 76.00% and 82.50%, respectively (Table S7-2). The combined model had a significant higher sensitivity than the single LncPPARδ signature (76.00% vs 68.00%), although the specificity was decreased from 90.00% to 82.50%. The 2 groups of patients differed significantly in age (P = 0.02), but not in other clinical and pathological factors (Table S8).

Diagnostic Specificity of LncPPARδ in PBMCs for CAD

Because CVD is always associated with some CVD risk factors, including hyperlipidemia, hypertension, and diabetes. In order to understand the specificity of LncPPARδ, we also studied those CVD risk factors in the present study. To investigate whether PBMC expression of LncPPARδ were elevated in other CVDs, we categorized the patients according to specific cardiovascular-related diseases (Figure 4). Although the sample size was small, the PBMC level of LncPPARδ increased significantly in patients with CAD, compared with patients without CAD but with other CVDs or metabolic diseases, including arrhythmia (n = 8), valvular disease (n = 8), dilated cardiomyopathy (n = 6), hyperlipidemia (n = 8), hypertension (n = 8), Type 2 diabetes mellitus (n = 12), abdominal aortic aneurysm (n = 2), and viral myocarditis (n = 7). Horizontal lines indicated the median and the different populations had similar sex, age (matched sex and age information is in Table S9), alcohol use, tobacco use, and other clinical features.

FIGURE 4

Diagnostic specificity of LncPPARδ in PBMCs for CAD. LncPPARδ levels were checked in CAD and other cardiovascular diseases, including arrhythmia (n = 8), valvular disease (n = 8), dilated cardiomyopathy (n = 6), hyperlipidemia (n = 8), hypertension (n = 8), type 2 diabetes mellitus (n = 12), abdominal aortic aneurysm (n = 2), and viral myocarditis (n = 7). Horizontal lines indicate the median. CAD = coronary artery heart disease, PBMC = peripheral blood monocyte.

LncPPARδ Is Stable as a CAD Biomarker

Stability is a key for a reliable biomarker. So we studied the stability of LncPPARδ in PBMC, which extracted from the cells right after getting the samples, exposed to room temperature for varied time and different freeze–thaw cycles. As indicated in Figure 5A, exposure to room temperature for 6 h had no effect on LncPPARδ expression. Even after 48 h, LncPPARδ levels were still at 80% of the unexposed levels. Thus, LncPPARδ levels were stable at room temperature. Freeze–thaw cycle is another major factor affecting RNA level. We found that 3 freeze–thaw cycles had no effect on LncPPARδ levels, although 7 freeze–thaw cycles did reduce the LncPPARδ to half of its original levels (Figure 5B).

FIGURE 5

Stability of LncPPARδ in PBMC exposed at varying time periods at room temperature (A) and different freeze-thaw cycles (B). Expression of LncPPARδ in the PBMCs was determined by qRT-PCR (∗P < 0.001, vs control, n = 12). PBMC = peripheral blood monocyte.

Regulation of LncPPARδ on Expressions of PPARδ and Its Target Genes

LncPPARδ was near protein-coding gene PPARδ. In light of the involvement of PPARδ in the pathogenesis of AS and according to the molecular regulatory principles of lncRNAs, we asked whether the expression of PPARδ was regulated by LncPPARδ. We found that the transcript expressions of PPARδ were affected by the knockdown of LncPPARδ by siRNA (Figure 6A and B). The mRNA expression of its downstream target genes including ADRP and ANGPTL4 was increased in THP-1 cells transfected with LncPPARδ targeting siRNA (Figure 6C and D).

FIGURE 6

Effects of LncPPARδ on mRNA expressions of PPARδ, ADRP, and ANGPTL4 in THP-1 cells after transfection of LncPPARδ siRNA. Knock-down efficiency of LncPPARδ siRNA determined by qRT-PCR (A). The effects of LncPPARδ silencing on PPARδ, ADRP, and ANGPTL4 expression determined by qRT-PCR at 24 h after siRNA treatment (B–D). PPARδ = peroxisome proliferator-activated receptor δ. ∗P < 0.001, compared with control, n = 6.

DISCUSSION AND CONCLUSIONS

CAD is one of the major diseases affecting human health worldwide.[27,28] Inflammation is essential to the initiation, development, and progression of CAD.[29,30] The correlation between inflammatory markers in the circulating blood and CAD has attracted a great deal of interest. Previous epidemiological studies have reported the links between CAD and “downstream” inflammatory markers, such as C-reactive protein and fibrinogen.[31,32] However, human genetic evidence does not support a causal relationship between these liver-derived factor and CAD.[33-35]

In contrast, the “upstream” markers of inflammation, such as proinflammatory cytokines, are more likely to be directly related to CAD, because they control the inflammatory cascades.[36] Latest studies have shown that interleukin-6 (IL-6),[37-39] interleukin-18 (IL-18),[40] matrix metalloproteinase 9 (MMP-9),[41] soluble CD40 ligand (sCD40L),[42] and TNF-α[43] independently associated with the risk of CAD, and this risk is independent of conventional risk factors. In this study, we set to identify an lncRNA biomarker belonging to potential regulators of proinflammatory cytokines. We examined the lncRNA profiles in PBMCs, an important population of inflammatory cells contributing to AS, of CAD patients and paired control subjects, and discovered a lncRNA, LncPPARδ alone or in combination with risk factors as a biomarker for CAD with high sensitivity and specificity. And the stability of PBMC lncPPARδ at room temperature for 24 h and up to 4 freeze–thaw cycles is sufficient for the analysis of lncPPARδ for CAD diagnosis for most cases. However, LncPPARδ in the plasma did not serve as a sensitive predictor of CAD (data were not shown), indicating that LncPPARδ plays a role in PBMCs to regulate its inflammatory response.

In recent years, an increasing number of lncRNAs have been found to be associated with various diseases. The roles of lncRNAs in development of cardiac diseases are increasingly being studied, including CAD.[44] The correlation of lncRNA levels with the prognosis of patients with cancer has recently been reported for several malignancies, such as hepatocellular carcinoma,[45] breast cancer, and colorectal cancer.[46] However, circulating lncRNAs in CAD as biomarker have not been reported so far. Here, we investigated the differential LncPPARδ expression in PBMCs of a cohort of 502 patients with or without CAD, indicating a potential role of LncPPARδ in CAD. In present study, the lncRNA signature, identified in the training set, showed similar diagnostic value in both the test set and the independent cohort. Furthermore, in the stratified analysis, the lncRNA signature enhanced the specificity and sensitivity to diagnose CAD when combined with other risk factors. Our results suggest that lncRNAs can be powerful predictors for diagnosis of patients with CAD.

Most lncRNAs are not yet functionally annotated. However, we can infer the possible function of the lncRNAs in CAD by analysis of CAD-related mRNA expressions. We found that PPARδ was down-regulated in PBMCs of CAD. It is interesting to notice that PPARδ was a nearby gene of LncPPARδ. According to the regulatory principles of lncRNAs, we suppose LncPPARδ could affect PPARδ and its downstream target genes. Down-regulation of LncPPARδ by siRNA in THP-1 cells line reduced mRNA expressions of PPARδ and its target genes ADRP and ANGPTL4, indicating that LncPPARδ is involved in PPARδ inflammatory signaling pathways. Thus, it is a plausible inference that the lncRNA associated with CAD patients may be involved in AS and CAD development.

The patients included in this study are Han Chinese from Chongqing City. Although a large cohort of patients was studied in this experiment, we are not sure whether the conclusion is applicable to other races and patients from other cities. Therefore, more prospective cohorts were needed to be studied in the future in other centers including patients from other races.

In conclusion, our study has shown that LncPPARδ, especially combined with risk factors, can be a biomarker for CAD. More valuable is that LncPPARδ separates CAD patients from the controls with symptoms similar to CAD patients, including angina pectoris or chest pain. The regulation of LncPPARδ on expressions of its neighboring protein-coding gene PPARδ and direct target genes of PPARδ, ADRP, and ANGPTL4 might be involved in the CAD pathogenesis. Due to the limitation of patients from one city, further validation in prospective cohorts from areas and races is needed.

MATERIALS AND METHODS

Study Cohorts

This study was a single-center clinical trial at Third MMU-based National Institute of Health sponsored (NIH.gov clinical trial NCT01629225). Briefly, the study recruited 502 patients admitted to the Department of Cardiology, Daping Hospital (Chongqing, China) from February 2013 to May 2014, due to clinically diagnosed or suspected CAD. All patients complained of chest pain. The study was approved by the Ethics Committee of the Medical Faculty of the Daping Hospital. Written informed consent was obtained from all patients or their families in accordance with the Declaration of Helsinki. Diagnosis was based on the final diagnosis with coronary artery angiography at discharge according to ACC/AHA classification.[47] Angiographic findings were interpreted independently by 2 interventional cardiologists in a blinded manner. CAD was diagnosed by the percentage of narrowing of each coronary artery segment and defined by ≥50% narrowing of the lumen of at least 1 of the major coronary arteries (the left main coronary trunk, anterior descending branch, circumflex artery, and right coronary artery), while the controls were people with no arterial stenosis at any coronary arteries. To minimize potential profound influence of acute myocardial ischemia, patients with myocardial infarction were excluded. Further exclusion criteria in this study were as follows: patients with severe valvular heart disease, malignant tumors, and other severe systemic diseases (such as renal failure and hepatic disease); patients with serious acute infection in the last 4 weeks or active chronic inflammatory disease, suspected drug or alcohol abuse; and patients who rejected participation in this study.

Microarray and Computational Analysis

RNAs from PBMCs were subjected to microarray-based global transcriptome analysis. RNA was preamplified and then underwent microarrays (Arraystar, Human LncRNA array, version 2.0), which allowed for simultaneous detection of both 33,045 lncRNAs and 30,215 coding transcripts. To find a potential biomarker candidate lncRNA, we first screened all lncRNA transcripts in the lncRNA microarrays. To make lncRNA levels easily detectable in the clinic, we selected these monocyte lncRNAs expression in CAD patients, according to the following criteria: fold change > 2, normalized intensity > 6, and P < 0.001. After screening, 269 lncRNAs were found to meet those criteria, 157 lncRNA transcripts were up-regulated and 112 lncRNA transcripts were down-regulated. The microarray data analyzed in this study have been deposited in the NCBI Gene Expression Omnibus database under accession number GSE69587. (http://www.ncbi.nlm.nih.gov/geo/info/linking.html).

Collection and Purification of PBMCs

One milliliter (mL) of peripheral blood (anticoagulant: ethylenediaminetetraacetic acid) was incubated with 15 μL fluorochrome-labeled monoclonal mouse antibody against human PerCP-CD14 (340585, BD Biosciences, New Jersey) in the dark for 15 min. Thereafter, 15 mL of ammonium chloride lysis buffer was added to lyse red blood cells. Fifteen milliliters of stopping medium (Phosphate buffered saline (PBS) with 3% heat-inactivated serum and 0.1% sodium azide) was then added to stop the lysis reaction. After mixing gently, samples were centrifuged at 120 g for 5 min and then washed with PBS for 5 times. Thereafter, cells were suspended in 200 μL of staining medium, mixed and run on flow cytometer immediately by FACS (FC500, Beckman Coulter, Fort Lauderdale, FL). At least 5 million events were acquired from the Cytometer. Flow data were analyzed with Flowjo software (Treestar, Inc., California). Absolute mononuclear cell count was estimated as the sum of monocytes using a Coulter ACT/Diff cell counter (Beckman Coulter).

Cell Culture

THP-1 monocytes, a human acute monocytic leukemia cell line, were obtained from IBCB (Institute of Biochemistry and Cell Biology, Shanghai, China) and cultured according to the supplier's recommendations at 37°C in a humidified 5% (v/v) CO2 air atmosphere in RPMI 1640 medium (Gibco, Invitrogen, New York) supplemented with 0.1 mg/mL l-glutamine, 0.1 mg/mL streptomycin, 100 U/mL penicillin, and 10% (v/v) fetal calf serum.

Transfection

For transfection, THP-1 monocytes were transferred to fresh medium for 24 h and then transfected using Lonza (Cologne, Germany) Nucleofection technique with LncPPARδ-specific siRNA and common negative control siRNA according to the manufacturer's protocol. Subsequently, transfected cells were seeded into 24-well plates (1 × 106 cells/well) for further 24 h in Monocyte Nucleofector medium (Lonza) containing 50 μM β-mercaptoethanol, 1 mM sodium pyruvate, 1% (v/v) nonessential amino acids, 0.1 mg/mL l-glutamine, 0.1 mg/mL streptomycin, 100 U/mL penicillin, and 5% (v/v) human serum off the clot (PAA Laboratories GmbH). Finally, cells were incubated for 24 h with cytokines in serum-free Monocyte Nucleofector medium containing 1 mM sodium pyruvate, 1% (v/v) nonessential amino acids, 0.1 mg/mL l-glutamine, 0.1 mg/mL streptomycin, 100 U/mL penicillin, and 50 μM β-mercaptoethanol.

RNA Isolation and Quantitative RT-PCR

Total RNA was extracted from PBMC and THP-1 by TRIzol (Invitrogen) and purified with RNeasy kits (Qiagen, Hilden, Germany). After reverse transcription (Superscript II, Invitrogen, California), qPCR was performed using the Brilliant Green Mastermix-Kit and the MX4000 multiplex qPCR System from Stratagene. The primers of the lncRNAs used in RT-qPCR are listed in Table S1.

Statistical Analysis

Data were presented as mean ± standard deviation, mean rank or numbers of patients unless otherwise described. Horizontal lines indicated the median in the scatter plots of the lncRNAs expression. Shapiro–Wilk and Kolomogorov–Smirnov tests were used to test for non-Gaussian distribution. For continuous variables, the 2-tailed Student test was used for normal distribution and homogeneity of variance, and the Mann–Whitney U test for abnormal distribution. Discrete variables were compared by 2 × 2 contingency table analysis of χ2 test. The association of LncPPARδ with the risk of CAD was assessed by Kruskal–Wallis H test for age and BMI. ROC curves and AUC were used to assess the sensitivity and specificity of LncPPARδ as a novel diagnostic tool for the detection of CAD. And we examined whether LncPPARδ improved the diagnosis prediction accuracy of CAD when added to a diagnostic model with or without gender, hypertension, tobacco use, and alcohol use, constructed by Fisher criteria. The AUC between the 2 models was compared by u test. P value < 0.05 was considered statistically significantly. Statistical analyses were performed using Windows SPSS Statistics 20 (SPSS, Inc., Chicago, IL), and GraphPad Prism 5 (GraphPad Software, San Diego, CA), and the MATLAB [V2013a] program.

TABLE 3 (Continued)

Clinical Characteristics of Patients in Training and Test Group (32nd)