Association of Coronary Artery Disease and Metabolic Syndrome: Usefulness of Serum Metabolomics Approach

Introduction Individuals with metabolic syndrome (MetS) are at increasing risk of coronary artery disease (CAD). We investigated the common metabolic perturbations of CAD and MetS via serum metabolomics to provide insight into potential associations. Methods Non-targeted serum metabolomics analyses were performed using ultra high-performance liquid chromatography coupled with Q Exactive hybrid quadrupole-orbitrap high-resolution accurate mass spectrometry (UHPLC-Q-Orbitrap HRMS) in samples from 492 participants (272 CAD vs. 121 healthy controls (HCs) as cohort 1, 55 MetS vs. 44 HCs as cohort 2). Cross-sectional data were obtained when the participants were recruited from the First Affiliated Hospital of Zhengzhou University. Multivariate statistics and Student’s t test were applied to obtain the significant metabolites [with variable importance in the projection (VIP) values >1.0 and p values <0.05] for CAD and MetS. Logistic regression was performed to investigate the association of identified metabolites with clinical cardiac risk factors, and the association of significant metabolic perturbations between CAD and MetS was visualized by Cytoscape software 3.6.1. Finally, the receiver operating characteristic (ROC) analysis was evaluated for the risk prediction values of common changed metabolites. Results Thirty metabolites were identified for CAD, mainly including amino acids, lipid, fatty acids, pseudouridine, niacinamide; 26 metabolites were identified for MetS, mainly including amino acids, lipid, fatty acids, steroid hormone, and paraxanthine. The logistic regression results showed that all of the 30 metabolites for CAD, and 15 metabolites for MetS remained significant after adjustments of clinical risk factors. In the common metabolic signature association analysis between CAD and MetS, 11 serum metabolites were significant and common to CAD and MetS outcomes. Out of this, nine followed similar trends while two had differing directionalities. The nine common metabolites exhibiting same change trend improved risk prediction for CAD (86.4%) and MetS (90.9%) using the ROC analysis. Conclusion Serum metabolomics analysis might provide a new insight into the potential mechanisms underlying the common metabolic perturbations of CAD and MetS.


INTRODUCTION
Nowadays, the significant increase of coronary artery disease (CAD) populations has become a serious challenge all over the world (1). According to the Global Burden of Disease Study 2016 (2), the incidence and years lived with disability (YLDs) of cardiovascular diseases were 54.1 million and 33.5 million that year, respectively. Interestingly, the risk factor-adjusted proportional-hazards regression of CAD mortality was doubled for the subjects with metabolic syndrome (MetS) in a 13-year follow-up report (3). Using the Framingham database, the ageadjusted relative risks for CAD with the MetS were 2.54 and 1.54 in men and women, respectively (4). If the association of CAD and MetS can be explicated, its occurrence and progression might be predicted. However, the association of CAD and MetS remains ambiguous.
Multiple and complex molecular events characterize CAD, which mainly refers to the myocardial dysfunction and/or organic lesions caused by coronary artery stenosis and insufficient blood supply (5,6). Meanwhile, the underlying mechanism of MetS is also complicated depending on its own diverse features. According to the American Heart Association/ National Heart, Lung and Blood Institute (AHA/NHLBI) criteria, metabolic syndrome is a clustering of cardiovascular disease risk factors, which include abdominal adiposity, insulin resistance, inflammations, genetic factor, abnormal neuroendocrine, unhealthy lifestyle, intrauterine malnutrition, and so on. Among these factors, abdominal adiposity and insulin resistance play a key role in the incidence of MetS (7)(8)(9). To be clear, there is a general consensus that cardiac risk factors should be aggressively managed in individuals with MetS (10).
Metabolomics is powerful omics technology that has been widely used for the unbiased identification of metabolic alterations of diseases. Metabolites could clarify the common metabolic perturbations of CAD and MetS. Fan et al. (11) reported 12 panels of metabolomics-based biomarkers in various clinical subgroups of CAD. Eighty-nine different metabolites were identified, and the altered metabolic pathways included downregulated phospholipid catabolism, tricarboxylic acid cycle, biosynthesis of primary bile acid, and upregulated amino acid metabolism, short-chain acylcarnitines. Another study exhibited the similar alterations in the gut microbiota and serum metabolites in different CAD subgroups (12). Besides, a 90 plasma-based metabolomics studies of MetS showed tyrosine, alanine, and propionylcarnitine increased and asparagine, tryptophan/large neutral amino acid ratio decreased (13). Nevertheless, serum metabolomics analysis to explore the common metabolic perturbations of CAD and MetS is still lacking. Therefore, this work employs a serum metabolomic approach to describe metabolic alterations of CAD and MetS by ultra-high performance liquid chromatography coupled with Q Exactive hybrid quadrupole-orbitrap high-resolution mass spectrometry (UHPLC-Q-Orbitrap HRMS). The study's specific objectives were as follows: (i) to identify metabolic signatures associated with CAD and MetS; (ii) to reveal the association of metabolic perturbations between CAD and MetS; (iii) to explore the risk predictive performance of common changed metabolites of CAD and MetS. To our knowledge, this study is the first time to explore the common metabolic perturbations of CAD and MetS by metabolomics, which could provide an insight into potential associations of the two diseases.

Reagents and Chemicals
The HPLC-grade methanol and acetonitrile were obtained from Fisher Scientific (Fair Lawn, NJ, USA). HPLC-grade formic acid was acquired from Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). HPLC-grade H 2 O was prepared by the Millipore system (Shanghai, China). The internal standards and all the metabolite standards used in the method were purchased from J&K Scientific Ltd. (Beijing, China) and Sigma-Aldrich (St. Louis, MO, USA). Each standard substance was dissolved in methanol and prepared into a mixed standards solution (1.0 mg/ml for each compound).

Study Design and Participants
A total of 492 participants, including 272 CAD, 55 MetS patients, and 165 HCs, were recruited from the First Affiliated Hospital of Zhengzhou University. CAD groups were diagnosed according to "Nomenclature and diagnostic criteria of ischemic heart disease" by International Society and Federation of Cardiology (ISFC) combined with World Health Organization (WHO) (14), which included chest pain symptoms, cardiovascular risk factors, pathological Q-wave on ECG, or elevated myocardial enzymes. Coronary angiography was simultaneously applied to confirm the diagnosis. MetS was diagnosed by the Chinese Diabetes Society (CDS) criteria (15), which satisfied three or more of the following index: overweight or obese [Body mass index (BMI) ≥25.0]; hyperglycemia [fasting plasma glucose (FPG) ≥6.1 mmol/L and/ or plasma glucose (PG) ≥7.8 mmol/L after 2 h]; hypertension [systolic blood pressure (SBP)/diastolic blood pressure (DBP) ≥140/ 90 Hg]; dyslipidemia [triacylglycerol (TG) ≥1.7 mmol/L and/or high-density lipoprotein cholesterol (HDL-C) <0.9 mmol/L for men or <1.0 mmol/L for women]. The HCs were matched for age, sex, and BMI, and several simple overweight or obese people who were "healthy obese" were also included. Details of inclusion and exclusion criteria are in the Supplementary Table 1.

Biochemical and Other Measures at Baseline
Clinical characteristics of the 492 participants were recorded, such as sex, age, height, weight. BMI was calculated as kg/m 2 . An automated validated device was used to measure sitting brachial blood pressure after a 10-min rest. HbA 1c levels were measured by a Bio-Rad Variant II hemoglobin testing system, and FPG data were recorded. The alanine aminotransferase (ALT) and aspartate transaminase (AST) levels were measured by a sandwich enzyme-linked immunosorbent assay (ELISA) system. Plasma total cholesterol (TC), triglycerides (TG), lowdensity lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and serum creatinine were measured using standardized methods from venous samples. Glomerular filtration rate (eGFR) was calculated from serum creatinine using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation.

Sample Collection and Preparation
The blood samples were collected in Na2 EDTA tubes. Each sample was centrifuged at 3,000 rpm for 10 min at 4°C to obtain supernatant. A 100 ml serum of each sample was then precipitated with 300 ml methanol containing 500 ng/ml ketoprofen and 50 ng/ml 2-chloro-L-phenylalanine as internal standard. The mixture was vortexed for 1 min followed by centrifugation at 13,000 rpm for 10 min at 4°C. Then, a 200 ml aliquot of supernatant was transferred into vials for analysis. For the assessment in parallel of reproducibility and stability along the run, quality control (QC) samples were obtained by mixing all serum samples, respectively. One QC sample was injected after every 10 samples throughout the run.
The mass spectrometry was performed on Q-Exactive orbitrap system (Thermo Fisher Scientific, San Jose, USA) equipped with a heated electrospray ionization source operated in positive ion modes. The MS parameters were optimized and set as follows: collision energy at 20, 40, and 60 eV, ion source temperature at 350°C, capillary temperature at 320°Ct, spray voltage at 3.50 kV, sheath gas flow rate at 40 arb, auxiliary gas flow rate at 10 arb. Metabolomic profiles were acquired with a mass range of 80-1,200 m/z. The full scan spectra and MS/MS data were collected with the resolution of 70,000 and 17,500 FWHM, respectively. The samples were injected in random order, and all the mass data were acquired and processed using Thermo Xcalibur 3.0 software.

Data Processing and Statistical Analysis
The mass spectrometry raw data were conducted by Compound Discoverer 2.1 software (Thermo Fisher Scientific, San Jose, USA). The spectra were selected from input LC-MS data files, and retention time alignment was accomplished based on mass tolerance and time shift criteria. Preliminary identification of metabolites was realized by searching databases including ChemSpider, Mass Lists, mzCloud, and mzVault. Multiple nodes such as "Align Retention Times", "Detect Unknown Compounds", "Group Unknown Compounds", "Predict Compositions", "Fill Gaps", and "Normalize Areas" were combined to form an untargeted metabolomics workflow for raw data processing (Supplementary Figure 1).
The data matrix obtained from Compound Discoverer was imported into SIMICA 14.1 software (Umetrics AB, Umea, Sweden) for multivariate statistical analysis, including unsupervised principal component analysis (PCA), supervised partial least-squares discriminant analysis (PLS-DA), and orthogonal partial least-squares discriminant analysis (OPLS-DA). The PCA analysis was applied to assess the reproducibility and stability of QC samples, and the PLS-DA model was established to describe general separation of samples from different groups. The variables responsible for the discrimination between two groups were identified by OPLS-DA, and permutation test was performed 200 times to assess the risk of overfitting for the OPLS-DA model. In addition to the multivariate statistical method, the p value was calculated by Student's t-test and adjusted by the Benjamini-Hochberg method (16). The metabolites with variable importance in the projection (VIP) values >1.0 and p values <0.05 were screened as significant altered metabolites for CAD and MetS, respectively, the structures of which were further confirmed based on available reference standards.
Logistic regression was performed to investigate whether the significant metabolites for CAD and MetS were independent to clinical cardiac risk factors. Firstly, the spectra area of the metabolites and the clinical characteristics were subjected to MetaboAnalyst 4.0 for log transformation and pareto scaling to acquire the normalized data under the same dimension. Then the odds ratios (ORs) and 95% confidence interval were obtained adjusting for three models (Model 1: adjustment for BMI; model 2: adjustment for TC, TG, LDL-C, and HDL-C; model 3: further adjustment for BMI, TC, TG, HDL-C, LDL-C, SDP, DBP, FGP, HbA1c); the workflows were shown in Supplementary Figure 2. Sample-size calculation was done to assess the sample size needed for replicating the metabolic signatures associations of CAD with clinical risk factors (http://www.powerandsamplesize.com/).
For association analysis of the metabolic perturbations of CAD and MetS, the global comparison of 30 significant metabolites for CAD and 26 significant metabolites for MetS were conducted. The different metabolites were divided according to the disturbed pathways. The association of significant metabolic perturbations between CAD and MetS was visualized by Cytoscape software 3.6.1. As well, the fold change values described above were regarded as the trends of metabolites levels variation.
For further confirmation of the association of CAD and MetS, the binary logistic regression and ROC analysis were performed to explore the risk predictive performance of the common changed metabolites (17,18). The covariates of significant metabolites with same change trend were used as biomarkers panel. Based on the protocol, logistic regression was proceeded after log transformation and pareto scaling of the spectra area of 14 biomarkers to obtain the p value, then the AUC under the classical univariate ROC curve was computed using biomarker analysis of MetaboAnalyst 4.0.

RESULTS
A total of 492 participants were allocated to two cohorts. As shown in Figure 1, 272 CAD patients and 121 HCs formed cohort 1 and 55 MetS patients and 44 HCs formed cohort 2.
Supplementary Table 2 showed the clinical characteristics and biochemical index of all the participants in this study. For CAD vs. HCs in cohort 1, age, sex, BMI, DBP, ALT, AST, TC, eGFR, and CR were comparable among the two populations (p > 0.05). However, compared with HCs, individuals with CAD had higher SBP, HbA 1c , FPG, TG, and LDL-C, but lower HDL-C levels. Similarly, for MetS vs. HCs in cohort 2, MetS patients had higher SBP, HbA 1c , FPG, TC, TG, and LDL-C, but lower levels of HDL-C. No obvious significant difference of age, sex, BMI, DBP, ALT, AST, eGFR, and CR was found between MetS and HCs.

Metabolic Signatures Associated With CAD
The obtained typical total ion chromatograms (TIC) from the three representative populations are shown in Supplementary  Figure 3. For the comparison of CAD vs. HCs in the cohort 1, a total of 1,673 features from positive mode were detected, and 933 ions showed significantly changed with p < 0.05. PCA, PLS-DA, and OPLS-DA score plots were performed to identify the differences of the metabolic profiles between CAD and HCs, which showed remarkable separations with cumulative R 2 Y at 0.873 and Q 2 at 0.749 (Figures 2A, B and Supplementary Figure 4). The volcano plot represents the variation of metabolites amount for CAD vs. HCs according to the -log2 (fold change) ( Figure 2C). In total, 30 identified metabolites showed significant by filtered with VIP > 1 and adjusted p values < 0.05. The information of the significant metabolites is presented in Table 1. And the altered pathways are shown in Figure 2D. The heatmap for the different metabolites in CAD and HCs is exhibited in Figure 2E. Combining with abundance comparison (Supplementary Figure 5), we found that the CAD patients had significantly different metabolite profiles compared with HCs. Of the 30 metabolites, most amino acids, lipid, primary bile acid, short-chain acylcarnitines, and purine were similar with the previous study (13,14). The novel findings-such as pseudouridine and dihydrothymine belong to pyrimidine, niacinamide belongs to nicotinate and nicotinamide metabolism, 4a-Carbinolamine tetrahydrobiopterin belongs to folate-have not been reported previously.

Metabolic Signatures Associated With MetS
For the comparison of MetS vs. HCs in the cohort 2, a total of 920 features from positive mode were detected, and 475 ions showed significantly changed (p < 0.05). After further filtering by VIP > 1 from PCA, PLS-DA, and OPLS-DA score plots ( Figures 3A, B and Supplementary Figure 6), 26 identified metabolites showed significant. The OPLS-DA showed remarkable separations with cumulative R 2 Y at 0.978 and Q 2 at 0.701. Figure 3C exhibited volcano plot of the variation of metabolites amount for MetS vs.
HCs. And the pathways were showed in Figure 3D. The heatmap for the different metabolites in MetS and HCs was exhibited in Figure 3E. The information of the 26 significant metabolites is presented in Table 1. Combining with abundance comparison ( Supplementary Figure 7), we found that most of the metabolism descended for amine acids, lipids, purine, pyrimidine, primary bile acid, and steroid hormone, except for elevated fatty acids and niacinamide. Compared to a previous study (13,19), new metabolites for MetS, including uracil, hypoxanthine and paraxanthine, dihydrothymine, niacinamide, cortisol, and tetrahydrocorticosterone, have not been reported previously.

The Association of Metabolic Perturbations With Clinical Cardiac Risk Factors
Logistic regression was performed to evaluate the associations of metabolic perturbations with clinical cardiac risk factors, respectively. As shown in Figure 4 and Supplementary Table 3 For the comparison of MetS vs. HCs, all the 26 metabolites were significant after adjustment for model 1 (p < 0.05). On further adjustment for model 2, associations were overall attenuated, but 17 remained significant. And even for model 3, 15 metabolites remained significant (Supplementary Figure 8 and Supplementary Table 4).
In sample size calculation for estimating the metabolic signatures associations of CAD and MetS with clinical risk factors at power of 0.8, alpha level of 0.05 (20), required sample sizes for cohort 1 (metabolites for CAD vs. HCs) and cohort 2 (metabolites for MetS vs. HCs) are 426 and 21, respectively (Supplementary Table 5).

The Association of the Metabolic Perturbations of CAD and MetS
Herein, association analysis for the exploration of common metabolic perturbations of CAD and MetS were conducted. The heatmap of correlation coefficients calculated among the significant metabolites of CAD or MetS is shown in Supplementary Figures 9, 10. The metabolites were further subdivided according to the KEGG and HMDB databases ( Figure 5A). We discovered 11 metabolites both changed for CAD and MetS. Of the 11 metabolites, pyroglutamic acid, LysoPC(P-16:0), LysoPC(16:0/0:0), phosphorylcholine, 2-Acylsn-glycero-3-phosphocholine, uracil, hypoxanthine, and chenodeoxyglycocholic acid were downregulated, and acetylcarnitine was upregulated in both CAD and MetS patients. Conversely, dihydrothymine was upregulated in CAD but downregulated in MetS, and niacinamide was downregulated in CAD but upregulated in MetS. As shown in Figure 5B, the involved pathways included reduced amino acids metabolism, lipid metabolism, purine metabolism, primary bile acid biosynthesis, and increased fatty acid metabolism. Nevertheless, pyrimidine metabolism and nicotinate and nicotinamide metabolism changed oppositely for CAD and MetS.

The Risk Predictive Performance of Common Changed Metabolites
Accurate risk prediction is a prerequisite for effective management of diseases (21). The covariates of nine significant metabolites with same change trend for CAD and MetS were used as biomarkers panel to explore the risk predictive performance, as well as further confirmation of the association of CAD and MetS. For the comparison of CAD vs. HCs in the cohort 1, the areas under the AUC curve, sensitivity, and specificity were 0.909, 82.7%, and 88.5% ( Figure 6A). Predictive value was 86.4% ( Figure 6C). For the comparison of MetS vs. HCs in the cohort 2, the areas under the AUC curve, sensitivity, and specificity were 0.948, 86.4%, and 92.7% ( Figure 6B), and predictive value was 90.9% ( Figure 6D). The ROC curve of each metabolite in biomarker panel was shown in Supplementary Figure 11. Odds ratios of the biomarker panel and several clinical data were provided in Supplementary Table 6.

DISCUSSION
It is noteworthy that MetS is associated with the risk factors of CAD, which leads to approximately twice occurrence of the CAD event in individuals with MetS (22)(23)(24)(25). Nevertheless, the association of CAD and MetS is still only understood partially in humans. Here we report a serum metabolomics profile with UHPLC-Q-Orbitrap HRMS to explore the common metabolic perturbations of CAD and MetS. Impressively, all the 30 identified metabolites for CAD and 15 of 26 metabolites for MetS remained significant after adjustments of combinatorial variables of clinical risk factors. The association of CAD and MetS has been initially explored. As shown in Figure 7, the increased short-chain acylcarnitines levels suggest activated fatty acid metabolism in CAD and MetS (26). Free fatty acid (FFA) could increase the levels of reactive nitrogen species (RNS) and ROS to induce oxidative stress, which lead to b cell dysfunction and insulin resistance by regulating related signaling pathways (27,28). On the other hand, CAD and MetS patients have downregulated LysoPC(16:0/ 0:0), LysoPC(P-16:0), phosphorylcholine, and 2-Acyl-snglycero-3-phosphocholine. As known, a large number of lysophosphatidylcholines in serum are generated from phosphatidylcholines by the activity of lecithin cholesterol acyltransferase (LCAT). Obviously, low activity of the LCAT has been linked to CAD and MetS (29,30). The lower pyroglutamic acid related to glutathione metabolism is reduced because CAD and MetS patients have increasing oxidative stress and decreased antioxidant capacity, which lead to lower GSH and higher GSSH (31). The farnesoid X receptor, an endogenous sensor for bile acids, was activated by chenodeoxyglycocholic acid. The inverse correlation has been reported between it and CAD or MetS (32). The altered uracil, hypoxanthine, and dihydrothymine are likely due to the disturbed purine metabolism and pyrimidine metabolism. Uric acid (UA) is the end product of purine metabolism, and either overproduction or underexcretion of UA may cause hyperuricemia (HUA), which is positive correlated to CAD and MetS (33,34). It has been reported that, by increasing reactive oxygen species (ROS) production in ApoE-knockout mice, hypoxanthine increased the levels of serum cholesterol and the area of atherosclerotic plaque (35). On the other hand, hypoxanthine could aggravate myocardial and renal graft ischemia/reperfusion injury by ROS (36,37). It is considered that xanthine oxidoreductase (XOR) leads to reduced uracil and hypoxanthine and increased uric acid in our study (38,39).
As number of water-soluble vitamins, niacinamide is a precursor of NAD and NADP to maintain the cellular metabolisms, including energy metabolism, DNA repair and aging, and oxidative stress. Nevertheless, excessive nicotinamide could exert the toxic effect of MetS by inducing the depletion of free methyl pool, to influence the NADdependent enzymatic reaction and trigger oxidative stress (40). Furthermore, niacinamide could be converted to N1methylnicotinamide (MNAM) by N-methyltransferase (NNMT) (41). It has been reported that MNAM reduces homocysteine secretion by 50%, and elevated homocysteine stimulates atherosclerosis via oxidative stress, endothelial dysfunction, and thrombosis and could be regarded as a marker of atherosclerosis (42)(43)(44).
Our study has several strengths. First, some significant metabolites, including pseudouridine, dihydrothymine, niacinamide, 4a-Carbinolamine tetrahydrobiopterin for CAD, uracil, hypoxanthine, paraxanthine, dihydrothymine, niacinamide, cortisol, and tetrahydrocorticosterone for MetS, have not been reported elsewhere (11)(12)(13)(14)19). Second, for the Our study also has several limitations. In the current design, the identified metabolite signatures might reflect common perturbation in CAD and MetS attributed to clinical cardiac risk. Nevertheless, prospective cohort-based metabolomics study should be a more proper design to study the specific association between MetS and CAD among those patients after certain follow-up years. Second, the CAD and MetS patients recruited in this study were firstly diagnosed and have no relevant medication history. In addition, the sample size of MetS vs. HCs in the cohort 2 is a little small to be sufficiently powered, which needs other replication in an independent cohort with appropriate sample sizes for stratification. Finally, the quantitative study of the altered metabolites is still lacking because of insufficient metabolite standards, and the relative molecular mechanism research to confirm the functions of differential expressed metabolites in MetS and CAD is recommended in future studies.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by the ethics committee of the First Affiliated Hospital of Zhengzhou University (2019-KY-296). The patients/ participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

AUTHOR CONTRIBUTIONS
Conceptualization, ZJ and LL. Methodology, ZJ, YS, and QD. Investigation and data curation, DZ. Visualization, ZJ, LL, LZ, and SD. Writing-review and editing, ZJ and LL. Project administration, XZ and ZS. All authors contributed to the article and approved the submitted version.