Author + information
- Received August 22, 2017
- Revision received September 29, 2017
- Accepted December 18, 2017
- Published online April 30, 2018.
- Leonardo P. de Carvalho, MD, PhDa,b,c,d,
- Sock Hwee Tan, PhDb,c,
- Ghim-Siong Ow, BSce,
- Zhiqun Tang, PhDe,f,
- Jianhong Ching, PhDg,
- Jean-Paul Kovalik, MD, PhDg,
- Sock Cheng Poh, BScb,
- Chee-Tang Chin, MBBSg,h,
- A. Mark Richards, MD, PhDb,i,
- Eliana C. Martinez, MD, PhDj,
- Richard W. Troughton, MD, PhDi,
- Alan Yean-Yip Fong, MBBSk,l,
- Bryan P. Yan, MBBSm,
- Aruni Seneviratna, MBBS, MPHb,
- Vitaly Sorokin, MD, PhDb,c,
- Scott A. Summers, PhDn,
- Vladimir A. Kuznetsov, PhD, SciDe,o,∗∗ ( and )
- Mark Y. Chan, MBBS, PhDb,c,∗ ()
- aFederal University of Sao Paulo State, Sao Paulo, Brazil
- bNational University Heart Center, Singapore, Singapore
- cCardiovascular Research Institute, Yong Loo Lin School of Medicine, National University of Singapore, Singapore
- dAlbert Einstein Hospital, São Paulo, Brazil
- eBioinformatics Institute, A*STAR, Singapore
- fInstitute of Molecular and Cell Biology, A*STAR, Singapore
- gProgram in Cardiovascular and Metabolic Disorders, Duke-National University of Singapore Graduate Medical School, Singapore
- hNational Heart Centre Singapore, Singapore
- iChristchurch Heart Institute, University of Otago Christchurch, Christchurch Hospital, Christchurch, New Zealand
- jMiller School of Medicine, University of Miami, Miami, Florida
- kClinical Research Centre, Sarawak General Hospital, Kuching, Malaysia
- lDepartment of Cardiology, Sarawak General Hospital, Kuching, Malaysia
- mDepartment of Medicine & Therapeutics, The Chinese University of Hong Kong, Hong Kong, China
- nUniversity of Utah, Salt Lake City, Utah
- oNanyang Institute of Technology in Health & Medicine, Nanyang Technological University, Singapore
- ↵∗Address for correspondence:
Dr. Mark Y. Chan, Cardiac Department, Yong Loo-Lin School of Medicine, National University of Singapore, 1E, Kent Ridge Road, NUHS Tower Block, Level 9, Singapore 119228.
- ↵∗∗Dr. Vladimir A. Kuznetsov, Department of Urology, Department of Biochemistry and Molecular Biology, 750 East Adams Street, Weiskotten Hall, Room 1271, SUNY Upstate Medical University, Syracuse, New York 13210.
• Targeted profiling of ceramides identified a 12-ceramide plasma signature that predicted 12-month cardiovascular death, MI, and stroke in 2 prospective cohorts of AMI patients.
• Among coronary artery bypass grafting patients, plasma ceramides were higher in those with recent AMI compared with those without recent acute MI.
• Analysis of rat ischemic myocardium revealed a consistent increase in ceramide levels and overexpression of 3 enzymes in ceramide biosynthesis.
We identified a plasma signature of 11 C14 to C26 ceramides and 1 C16 dihydroceramide predictive of major adverse cardiovascular events in patients with acute myocardial infarction (AMI). Among patients undergoing coronary artery bypass surgery, those with recent AMI, compared with those without recent AMI, showed a significant increase in 5 of the signature’s 12 ceramides in plasma but not simultaneously-biopsied aortic tissue. In contrast, a rat AMI model, compared with sham control, showed a significant increase in myocardial concentrations of all 12 ceramides and up-regulation of 3 ceramide-producing enzymes, suggesting ischemic myocardium as a possible source of this ceramide signature.
- acute coronary syndrome
- major adverse cardiovascular and cerebrovascular events
- risk prediction
As many as 10% of the patients hospitalized for acute myocardial infarction (AMI) will have another episode of AMI within 12 months of the index hospitalization (1). Apart from cholesterol, recent studies have highlighted the physiological role of other lipid classes in atherosclerosis and AMI, including ceramides, sphingomyelin, phosphatidylcholines, and cholesterol esters (2–4).
Of these lipids, ceramides are among the most bioactive membrane lipids regulating signal transduction pathways that lead to cell survival or death (5). Ceramides accumulate in coronary atheromatous plaque (3), and their glycosylated forms, glucosylceramides and lactosylceramides, are more abundant in regions of arterial tissue with visible plaque development (2). Myocardium can produce ceramides in response to ischemia and reperfusion, leading to an increase of ceramides that activate mitochondrial autophagy and apoptosis (6,7). As plasma ceramides are now readily quantifiable, it has become possible to examine the relationship between ceramides and cardiovascular death in stable and unstable coronary artery disease (CAD) cohorts (4,8,9). Although associations between ceramides and cardiovascular death have been observed, the association between plasma ceramides and incident MI has only recently emerged (4,10).
To evaluate the prognostic role and tissue origin of plasma ceramides in AMI, we studied the association of ceramides with cardiovascular death, recurrent MI, and stroke to construct a prognostic ceramide signature. The prognostic utility of the ceramide signature was validated in an independent AMI cohort. We then further characterized the arterial and myocardial profiles of the signature’s individual ceramides in human aortic biopsy samples and a rodent AMI model, respectively.
Plasma samples were obtained from patients undergoing invasive management for AMI at 2 tertiary hospitals in Singapore between 2011 and 2013 (Supplemental Figure 1). Blood samples were collected pre-angiography (day 1) and within 24 h post-angiography (day 2). We included 337 patients with AMI who were diagnosed following the criteria from the Third Universal Definition of Myocardial Infarction (11), as ascertained by the managing physician. All patients underwent coronary angiography within 7 days of symptom onset and within 3 days of hospitalization. Exclusion criteria included low hemoglobin concentration (<8 g/l for men and <7 g/l for women), unwillingness to give consent, or absence of obstructive CAD, defined as any stenosis ≥70% or left main stenosis ≥50% (12). All patients were contacted by phone up to 24 months after the index hospitalization to ascertain the incidence of major adverse cardiac and cerebrovascular events (MACCE).
Plasma samples were obtained from an independent cohort of 119 patients undergoing invasive management for clinically diagnosed AMI and who had obstructive CAD on coronary angiography at the Christchurch Hospital (Christchurch, New Zealand) between 2013 and 2014. Blood samples were collected at 24 h post-angiography (day 2), and all patients received follow-up phone calls up to 24 months after the index hospitalization (Supplemental Figure 1).
All animal experiments were approved by the Institutional Animal Care and Use Committee of the National University of Singapore and were performed in accordance with the established guiding principles for animal research (Supplemental Appendix). Briefly, male Wistar rats (250 to 300 g) underwent permanent ligation of the left anterior descending artery, sham thoracotomy, or anesthesia without surgery (AMI/sham/healthy groups). The combination of isoflurane for anesthesia, pentobarbital for euthanasia, and organ harvesting during anesthesia rather than post-euthanasia has been shown to minimally perturb the rodent metabolome (13).
Human arterial tissue and plasma
A total of 70 patients undergoing elective coronary artery bypass grafting (CABG) with (n = 35) or without (n = 35) who incurred AMI within 2 weeks prior to surgery were recruited. Standard full normothermic cardiopulmonary bypass (37°C) and cold (4°C) anterograde cardioplegia were performed. Proximal anastomoses were created by punch biopsies of the aorta during CABG; the punch biopsy tissues were snap-frozen immediately and stored at −196°C until used. Simultaneous peripheral blood samples were also collected, and plasma was stored at −80°C.
Please refer to the Supplemental Appendix for details regarding lipid extraction.
Measurement of lipid species
HPLC-grade acetonitrile and ammonium acetate were purchased from Sigma-Aldrich (St. Louis, Missouri). A total of 68 ceramide species (Supplemental Table 1) were quantified by hydrophilic interaction LC (HILIC) MS/MS at Duke-National University of Singapore Graduate Medical School (Singapore). Extracted lipids were separated on an HILIC column (Thermo Fisher Scientific, Waltham, Massachusetts) (100 × 2.1 mm; particle size 2.6 μm) using an Agilent 1260 LC system coupled to an Agilent 6430 triple quadrupole mass spectrometer (Agilent Technologies, Santa Clara, California). Briefly, the HILIC column was equilibrated with 100% mobile phase A (acetonitrile/water [95:5] with 10 mmol/l ammonium acetate, pH 8.0), increasing to 20% mobile phase B (acetonitrile/water [50:50] with 10 mmol/l ammonium acetate, pH 8.0) in 5 min, then held for 5 min. The column was finally re-equilibrated with 100% mobile phase A for 5 min. Peak detection was performed using the Quantitative Analysis for QqQ (version B.06.00, Agilent). Chromatography peaks were autointegrated and manually curated. Lipid concentrations were estimated by comparing the peak area of each species to the peak area of the corresponding internal standard. Blank runs (mobile phase B) were included every 5 samples to assess the extent of carryover.
RNA isolation and real-time PCR
Please refer to the Supplemental Appendix for details regarding RNA isolation and real-time PCR.
Baseline characteristics of continuous and categorical variables were presented as mean ± SD and percentage, respectively. Continuous variable sets were compared using the Student t test and Mann-Whitney U test for parametric and nonparametric tests, respectively, whereas categorical variable sets were compared using the chi-square and Fisher exact tests. Dependent variable comparison was carried out using the paired Student t test and Wilcoxon matched pairs test. Pearson and Kendall’s tau correlation coefficient values were calculated to understand correlation patterns between different variables.
Rat myocardial and human aortic ceramide intensities were normalized to tissues’ wet weights (Supplemental Appendix, Supplemental Figure 2). The fold-changes of paired human aortic tissue and plasma were calculated via the ratio of their median intensities. The statistical significance of the differential intensity values for the plasma and tissue samples was assessed via the nonparametric Mann-Whitney U test. For the comparison of rat myocardial and plasma samples among healthy, sham, and MI groups, we applied 2-way analysis of variance with ceramide concentration as one factor and healthy/sham/MI group as the other factor. The p value for interaction between these 2 factors was 0.01.
The prognostic association of individual ceramide species with MACCE and time of the event was modelled using Cox Proportional Hazards regression. Clinical risk factors, including components of the GRACE (Global Registry of Acute Coronary Events) risk score (14), were adjusted in the Cox regression model. Time-to-event survival functions for each ceramide were estimated and visualized via the Kaplan-Meier survival curve analyses, with low- and high- risk groups compared by the log-rank test.
We then used deep-learning predictor selection and risk stratification methods (data-driven grouping [DDg] and statistically-weighted voting grouping [SWVg], respectively) (Supplemental Figure 3). This model, which we had previously developed and validated (15–20), is described in the Supplemental Appendix. Briefly, at the first step of the method, we apply the DDg method to identify cut-off values (intensities) of individual prognostic variables (i.e., ceramides) that optimally separate patients into low- and high-risk MACCE groups, In the second step, the SWVg method utilizes the 1D-DDg–defined ceramide risk categories for each patient and combines ceramides in a stepwise multivariable fashion to produce a synergistic multiceramide signature. To define a robust prognostic signature, we only included component ceramides that were statistically significant in all 3 models: Cox regression, DDg, and SWVg. A 2-sided alpha of 0.05 was considered statistically significant, and the false discovery rate was set at 0.1. Because patients either do or do not have events and because of sample size limitations, multiple risk categories were avoided. Thus, 2 prognostically distinct risk categories, low-risk and high-risk, were considered. At this consideration, SVWg optimization procedure (17) defined the 12 ceramides that collectively provided most confidence discrimination of the low- and high-risk patients. Finally, a linear regression model was used for the quantitative analysis of the association between datasets and testing their scalability (Supplemental Appendix, Supplemental Figure 4). Regular statistical analyses were performed on GraphPad Prism version 5 (GraphPad, San Diego, California), Statistica 7.1 (Tulsa, Oklahoma), ComBat (21), and SurvExpress (22).
The study was conducted according to the Helsinki Declaration, and all institutions’ human ethics review boards approved the study protocol. Animal experiments were approved by the institutional animal care and use committee of the National University of Singapore.
The clinical characteristics and biochemical profiles of the AMI patients
The clinical characteristics and biochemical profiles of the AMI patients are shown in Table 1. Compared with the validation cohort, patients in the discovery cohort were significantly younger; more likely to be men and smokers; and more likely to have diabetes mellitus, dyslipidemia, and a lower body mass index (p < 0.05). Mean levels of triglycerides and low-density lipoprotein cholesterol were not significantly different between the 2 cohorts. Mean high-density lipoprotein cholesterol showed a nonsignificant trend toward lower levels in the Singapore discovery cohort.
Correlations and differences of the plasma ceramides measured pre-angiography and within 24 h post-angiography
Of the 68 ceramides measured, 14 ceramides and 6 dihydroceramides (DHCer) were reproducibly detected in plasma. The concentrations of 8 of 14 ceramides (Cer d18:1/14:0, d18:1/18:1[9Z], d18:1/20:0, d18:1/22:0, d18:1/23:0, d18:1/24:0, d18:1/25:0, and d18:1/26:0) and all 6 DHCer (DHCer d18:0/16:0, d18:0/18:0, d18:0/20:0, d18:0/22:0, d18:0/24:0, and d18:0/24:1[15Z]) were significantly lower at day 2 compared with day 1 (Supplemental Table 2).
During a median follow-up of 12 months, 26 of the 327 discovery cohort patients and 25 of the 119 validation cohort patients had their first MACCE event, respectively. Cox proportional hazards regression analysis performed for each ceramide, with adjustment for the GRACE score, indicated 11 and 14 of the quantifiable ceramides (at days 1 and 2, respectively) yielded significant hazard ratios for MACCE in the discovery cohort (Table 2).
Table 2 shows that 10 ceramides are survival significant at both days 1 and 2 (p ≤ 0.05); additionally, 1 ceramide was significant at day 1, and 4 ceramides at day 2. Kendall’s tau correlation coefficient between −log2 (p value) at days 1 and 2 was 0.507 (p = 0.001). The individual ceramide measurements at day 2 had smaller p values than at day 1.
Thus, we conclude that high-risk patients have an overexpression of the selected plasma ceramides on days 1 and 2; however, day 2 ceramides more accurately identified patients who were at high-risk of MACCE.
Identification of the multiceramide prognostic signature
Supplemental Figure 3 shows the workflow of the discovery and validation analyses. Supplemental Table 3 shows that 20 ceramides measured at day 2 were significant for MACCE-free survival (false discovery rate <0.05). Notably, for 18 of the 20 ceramides selected in the discovery cohort, the 1D-DDg–defined cut-off values better stratified patients into low- and high-risk patient groups compared with the GRACE score. Supplemental Table 3 summarizes the univariate 1D-DDg and multivariate SWVg models used in the training cohort. As shown in Figure 1A (left), there was no significant difference between relatively low- and high-risk groups defined by GRACE score (at cut-off GRACE score = 141; p = 0.41) in the discovery cohort. The combination of 12 ceramides was highly predictive of 12-month MACCE (p = 3.17 × 10−10) (Figure 1B, left). The survival curves after stratification into low- and high-risk groups using each of the top 12 ceramides are shown in Supplemental Figure 5.
Validation of the 12-ceramide prognostic plasma signature in an independent cohort
The intensities of the plasma ceramides from both the discovery and validation cohort showed similar distributions that were scalable and well-calibrated (Supplemental Appendix, Supplemental Figure 4). When we dichotomized the patients in the validation cohort using the same cut-off intensities of the ceramides defined in the training cohort, we found that 11 of the 12 dichotomized ceramides were directionally similar in the validation as in the training cohort. By the binomial test of the HO-hypothesis (random coincident events) this result was significant (p = 0.00468). Of the 12 ceramides, only Cer(d18:1/22:1) was statistically significant in the both cohorts (p < 0.05) (Figure 1C, Supplemental Figure 6). Several others trended toward statistical significance (for instance, Cer(d18:1/23:0), Cer(18:1/24:0), Cer(18:1/25:0), and Cer(d18:1/26:0) at p < 0.10) (see also Supplemental Figure 7). However, in validation cohort collectively, the 12-ceramide prognostic signature was statistically significant (p = 0.03875) (Figure 1B, right).
Due to the difference in sample size, with the available validation dataset being ∼3× smaller than the discovery dataset, we expected that the results of implementation of the prognostic model parameters, defined by the 1D-DDg method at validation step, might be less stable and significant than in the training dataset (Supplemental Figures 6 and 7). The relatively small size of the validation cohort prohibited us from calculating calibration measures such as the Hosmer-Lemeshow chi-square test. However, we can infer from Figures 1A to 1C as to whether the GRACE score, 12-ceramide-defined score, and single ceramide predictors predict MACCE event rates uniformly in the discovery and validation cohorts (Supplemental Table 3). By comparing the MACCE-free survival rates of the GRACE score–defined low- and high-risk groups in the discovery and validation cohorts, we see that the GRACE score had the poorest reproducibility in predicting actual survival rates of both the low- and high-risk groups (Figure 1A). In contrast, low- and high-risk groups defined by an individual ceramide predictor yielded more similar MACCE-free survival rates in the discovery and validation cohorts, especially among the high-risk group (Figure 1C). The most reproducible MACCE-free survival rates were observed with the 12-ceramide risk signature (Figure 1B).
Arterial and plasma ceramides in AMI
We then sought to investigate changes in the 12-ceramide prognostic signature in human arterial tissue and plasma obtained concurrently after a recent AMI. Human ascending aortic punch biopsies and concomitant plasma samples were obtained during CABG in patients with (n = 35) and without (n = 35) AMI within 2 weeks prior to CABG (Supplemental Table 4). Compared with patients without recent AMI, those with recent AMI had higher levels of plasma Cer(d18:1/16:0), Cer(d18:1/18:0), Cer(d18:1/20:0), Cer(d18:1/24:1(15Z)), and DHCer(d18:0/16:0) (Figure 2), all of which were components of the 12-ceramide prognostic signature (Table 3). Aortic tissue samples from the same patients demonstrated only a nonsignificant trend toward slightly higher ceramide levels in patient samples with versus those without recent AMI (Figure 2).
Myocardial ceramides and their regulatory enzymes in AMI
The use of an animal model of AMI allowed us to delineate the effect of acute ischemia and necrosis on plasma and myocardial ceramides and to study the related gene expression changes in enzymes involved in ceramide biosynthesis and degradation. The mortality at 1 h post-MI was 14.3% in our rat experiments. All deaths occurred during the first hour post–left anterior descending (LAD) ligation, whereas no mortality was observed in the sham groups. Following ligation of the LAD artery in rats (n = 6), we observed an increase in myocardial levels of individual ceramide species from the prognostic signature when compared with rats with or without sham surgery (n = 3) (Figure 3A, Supplemental Figures 8 and 9). Notably, the levels of 5 ceramides in myocardial tissue, Cer(d18:1/16:0), Cer(d18:1/18:0), Cer(d18:1/20:0), Cer(d18:1/22:0), and Cer(d18:1/24:0), were significantly higher in the AMI group than in the sham control groups. In plasma samples obtained at the same timepoints via tail vein sampling, only Cer(d18.1/24.0) was significantly increased in the AMI group compared with sham and healthy animals, whereas the other plasma ceramides showed a numerical increase that was not statistically significant (Figure 3B, Supplemental Figures 10 and 11).
We then profiled gene expression changes of the enzymes involved in ceramide synthesis and degradation pathways to identify changes in myocardial enzyme gene expression 24 h after ligation of the LAD. Supplemental Figure 12 shows that there was an overexpression of serine palmitoyl transferase-2, ceramide synthase 6 (CerS6), and neutral sphingomyelinase (nSMase), and an underexpression of acid ceramidase in AMI versus sham control subjects. The serine palmitoyl transferase-2 gene encodes a long-chain base subunit of serine palmitoyl transferase (SPT), an enzyme that catalyzes the first step of the biosynthesis of sphingolipids. CerS6 encodes ceramide synthase 6, which is primarily involved in the de novo synthesis of C16:0 ceramides. nSMase encodes neutral sphingomyelinase, which metabolizes sphingomyelin into ceramides. Acid ceramidase is a lipid hydrolase that degrades ceramides into sphingosine and free fatty acids.
Ceramides, highly bioactive membrane lipids, have been implicated in atherosclerotic plaque progression (23–25) and ischemia-induced cardiomyocyte apoptosis (6,26). We identified a 12-ceramide plasma prognostic signature that predicted long-term MACCE in patients with AMI, of which Cer(d18:1/22:1), Cer(d18:1/24:1[15Z]), and DHCer(d18:0/16:0) were the strongest predictors of MACCE. Further profiling of the ceramide signature in patients undergoing CABG showed that plasma levels of 4 ceramides and 1 dihydroceramide were significantly higher in patients with versus without recent AMI. Aortic tissue from the same patients showed only a modest and nonsignificant numerical trend toward increased concentrations of these ceramide species when comparing patients with versus without recent AMI. Additionally, temporal profiling of myocardial ceramides and their regulatory enzymes in a rodent model of AMI demonstrated myocardial up-regulation of 3 ceramide production enzymes, SPT, CerS6, and nSMase, and down-regulation of 1 ceramide degradation enzyme, acid ceramidase, leading to a net increased production of all 12 ceramides in ischemic myocardium within 24 h of coronary artery occlusion. Although paired plasma samples from the same rat AMI model showed a numerical increase in the concentrations of all 12 ceramides, only the increase in plasma Cer(d18:1/24:0) was statistically significant when compared with sham and healthy control subjects.
Ceramides and atherosclerosis
Atherosclerosis involves the ceramide pathway not only through ceramide-induced macrophage/foam cell apoptosis, but also via interactions with reactive oxygen species, nitric oxide, and inflammatory cytokines (23). In our study, the aortic tissue levels of ceramide species tended, albeit not significantly, to be higher in patients with versus without recent AMI undergoing bypass surgery. As confirmed in Supplemental Table 4, all patients in the bypass cohort had severe multivessel CAD. Aortic biopsies were obtained as remotely as 2 weeks after AMI, and more time-sensitive kinetic profiles might reveal greater enzyme turnover. The presence of a higher concentration of ceramides in plasma from the same patients with versus those without recent AMI suggests that sources other than arterial tissue, such as post-infarct myocardium, may be a continuing source of plasma ceramides 2 weeks after AMI.
Ceramides and ischemic myocardial injury
Mitochondrial dysfunction is a key event mediating cell death in ischemia-reperfusion injury, the insult leading to the many long-term sequelae in AMI, and ceramides are strongly implicated in mitochondrial-induced apoptotic cell death (6,27). Endogenous ceramides, generated from either de novo synthesis or sphingomyelinase activation in response to diverse cellular stress stimuli such as hypoxia, ischemia-reperfusion, chemotherapy, or γ-irradiation, mediate mitochondrial cytochrome C release and apoptosis of cardiomyocytes (27). Consistent with the pattern of increased ceramide production in our rat model of AMI, other studies have shown an increase in myocardial ceramide levels during ischemic-reperfusion injury with corresponding down-regulation of ceramidase activity (28,29) and dramatic reciprocal up-regulation of SPT (30) and sphingomyelinase (31). Serine palmitoyl transferase is a key rate-limiting enzyme catalyzing the first step in the de novo synthesis of ceramides (32). Interestingly, the inhibition of SPT with myriocin protects cardiomyocytes from lipotoxic injury, suggesting a potential role for therapeutic ablation of the de novo ceramide pathway to minimize myocardial injury (33). SMases hydrolyses sphingomyelin, releasing phosphocholines and ceramides. Ischemia-reperfusion leads to an elevation of SMase activities and subsequent activation of TNF-α mediated apoptosis (34). Meanwhile, the blockade of SMase activity with N-acetylcysteine abrogates TNF-α–induced apoptosis and improves post-infarct cardiac recovery (35). Taken together, the inhibition of SPT and SMases could potentially ameliorate myocardial ischemia-reperfusion injury, attenuate apoptosis, and facilitate cardiac recovery.
Ceramide synthases regulate the synthesis of ceramides in both the de novo and salvage ceramide pathways. There are 6 known ceramide synthases, and each has a unique specificity for acyl CoA chain length (36). We observed an increase in rat myocardial CerS6 during AMI, which has been shown to generate mainly shorter-chain ceramides such as C14:0 and C16:0 (37). In agreement with our human data, rat myocardial Cer(d18:1/14:0) and Cer(d18:1/16:0) increased significantly 24 h after AMI.
Odd-chain ceramides and dihydroceramides
We observed the presence of 2 odd-chain ceramides, Cer(d18:1/23:0) and Cer(d18:1/25:0), in the prognostic multiceramide signature. Odd-chain ceramides are derived from odd-chain saturated fatty acids, which are not endogenously produced in the human body. Dairy products and meat from ruminant animals are important sources of odd-chain saturated fatty acids (38). Population-based studies have found an inverse association between plasma odd-chain saturated fatty acids and the risk of coronary heart disease (39) and type 2 diabetes (40). Given the low abundance of the 2 odd-chain ceramides in both cohorts, we cannot exclude the possibility that they may represent an epiphenomenon rather than having a more direct role in AMI outcomes.
DHCer, which are transiently produced during de novo ceramide synthesis, have been deemed to be biologically inert (41). Recent evidence suggests that DHCer have biological functions that are distinct and nonoverlapping with those of the more prevalent ceramides, including autophagy, hypoxia, and cellular proliferation (41). Our prognostic ceramide signature only consists of 1 dihydroceramide, DHCer(18:0/16:0), but further research may better define the prognostic role of elevated plasma DHCer for secondary ischemic events.
First, our discovery and validation cohorts were relatively small compared with other studies. Second, the highly dimensional nature of our datasets (larger number of predictors than events) necessitated the use of deep learning models (DDg and SWVg) to derive predictive signatures. Although this approach is able to delineate a ceramide signature that classifies patients into low versus high risk, our sample size does not permit the development of a quantitative risk score to predict the incidence of MACCE. Third, our study lacked other reference prognostic biomarkers such as B-type natriuretic peptide and C-reactive protein. Fourth, the activation of lipoprotein lipase by heparin will continue to be a concern for all lipidomic studies. Lipoprotein lipase catalyzes the breakdown of triglycerides in lipoproteins, which may affect the plasma lipidome (42). The intensities of ceramides measured on day 2 were also less than those on day 1, likely because heparin was stopped immediately after coronary angiography with or without percutaneous coronary intervention in patients with AMI. Indeed, this should be strongly considered when measuring plasma ceramides in clinical samples; testing for plasma ceramides as prognostic biomarkers is now being offered as a clinical test for patients with CAD (43), and patients with CAD are often exposed to heparin in the inpatient setting. Fifth, our paired analysis of arterial tissue and plasma in patients with and without AMI utilized punch biopsies of the thoracic aorta instead of coronary artery segments. Although access to coronary artery tissue remains a limitation of many human studies similar to ours, numerous studies have established a strong correlation between atherosclerotic disease severity of the thoracic aorta and coronary arteries (44–46).
A key strength of our study is the verification of the presence of the 12-ceramide prognostic signature in paired aortic tissue-plasma and paired myocardial tissue-plasma samples. To our knowledge, no prior published studies have simultaneously profiled a singular ceramide signature in AMI cohorts with longitudinal outcomes, paired aortic tissue-plasma samples, and paired myocardial tissue-plasma samples.
A 12-ceramide plasma prognostic signature comprising 9 even-chain ceramides, 2 odd-chain ceramides, and 1 dihydroceramide of C14-26 chain length was associated with cardiovascular death, recurrent MI, and stroke after an AMI in ethnically diverse cohorts. Taken together with other studies, our data support a prognostic role of plasma ceramide species in AMI. The profiling of proposed multiceramide signature in arterial and myocardial tissue in recent and acute infarction, respectively, indicates an up-regulation of ceramide pathway enzymes in ischemic myocardium, which may be a source of plasma ceramides in AMI. Future studies are needed to clarify the role of odd-chain ceramides and dihydroceramide in mediating ischemic outcomes after AMI. The prognostic role of other complex ceramides such as glucosylceramides and lactosylceramides also deserves further study; such efforts will require a combination of quantitative lipidomic and glycomic platforms.
COMPETENCY IN MEDICAL KNOWLEDGE: Plasma ceramides predict cardiovascular death, recurrent MI, and stroke after an AMI. Taken together with other studies, our data support a prognostic role of plasma ceramide species in AMI. The optimal time at which to perform testing appears to be 24 to 72 h after symptom onset.
TRANSLATIONAL OUTLOOK: Future studies are needed to clarify the role of odd-chain ceramides and dihydroceramide in mediating ischemic outcomes after AMI. The prognostic role of other complex ceramides such as glucosylceramides and lactosylceramides also deserves further study; such efforts will require a combination of quantitative lipidomic and glycomic platforms.
The authors thank Ms. Xiaoxun Yang (Yong Loo Lin School of Medicine, National University of Singapore, Singapore) for blood sample accessioning and aliquoting, along with Professor Dominique de Kleijn (Utrecht Medical Centre, Utrecht, the Netherlands), and Dr. Jiongwei Wang (Department of Surgery, National University of Singapore, Singapore) for their valuable experimental guidance.
This work was supported by grants NMRC/CSA/028/2011 and NMRC/CG/014/2013 from the National Medical Research Council, Singapore. Dr. Richards has received salary support from NMRC/STaR/0022/2014. Drs. Ow, Tang, and Kuznetsov have received salary support from BII/A*STAR, Singapore. Dr. Summers has served as a consultant for and is a shareholder of Centaurus Therapeutics. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Kuznetsov and Chan contributed equally to this work and are joint senior authors.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Basic to Translational Science author instructions page.
- Abbreviations and Acronyms
- acute myocardial infarction
- coronary artery bypass graft
- coronary artery disease
- ceramide synthase 6
- data-driven grouping
- hydrophilic interaction LC
- left anterior descending
- major adverse cardiac and cerebrovascular events
- myocardial infarction
- neutral sphingomelinase
- serine palmitoyl transferase
- serine palmitoyl transferase-2
- statistically-weighted voting grouping
- Received August 22, 2017.
- Revision received September 29, 2017.
- Accepted December 18, 2017.
- 2018 The Authors
- Chaudhry S.I.,
- Khan R.F.,
- Chen J.,
- et al.
- Chatterjee S.
- Mukhin D.N.,
- Chao F.F.,
- Kruth H.S.
- Wang D.D.,
- Toledo E.,
- Hruby A.,
- et al.
- Sigruener A.,
- Kleber M.E.,
- Heimerl S.,
- Liebisch G.,
- Schmitz G.,
- Maerz W.
- Havulinna A.S.,
- Sysi-Aho M.,
- Hilvo M.,
- et al.
- Thygesen K.,
- Alpert J.S.,
- Jaffe A.S.,
- et al.
- Overmyer K.A.,
- Thonusin C.,
- Qi N.R.,
- Burant C.F.,
- Evans C.R.
- Chen L.,
- Jenjaroenpun P.,
- Pillai A.M.,
- et al.
- Kuznetsov V.A.,
- Tang Z.,
- Ivshina A.V.
- Ow G.S.,
- Kuznetsov V.A.
- Ow G.S.,
- Tang Z.,
- Kuznetsov V.A.
- ↵ComBat. Available at: https://www.bu.edu/jlab/wp-assets/ComBat/Abstract.html. Accessed January 12, 2018.
- ↵SurvExpress. Available at: http://bioinformatica.mty.itesm.mx:8080/Biomatec/SurvivaX.jsp. Accessed January 12, 2018.
- Edsfeldt A.,
- Duner P.,
- Stahlman M.,
- et al.
- Jin J.,
- Hou Q.,
- Mullen T.D.,
- et al.
- Parra V.,
- Moraga F.,
- Kuzmicic J.,
- et al.
- Reforgiato M.R.,
- Milano G.,
- Fabrias G.,
- et al.
- Park T.-S.,
- Hu Y.,
- Noh H.-L.,
- et al.
- Mizutani Y.,
- Kihara A.,
- Igarashi Y.
- Khaw K.-T.,
- Friesen M.D.,
- Riboli E.,
- Luben R.,
- Wareham N.
- Forouhi N.G.,
- Koulman A.,
- Sharp S.J.,
- et al.
- Siddique M.M.,
- Li Y.,
- Chaurasia B.,
- Kaddai V.A.,
- Summers S.A.
- Brunner M.P.,
- Shah S.H.,
- Craig D.M.,
- et al.
- ↵Mayo Clinic Laboratories. Test ID: CERAM. Ceramides, Plasma. Available at: http://www.mayomedicallaboratories.com/test-catalog/Clinical+and+Interpretive/65054. Accessed August 15, 2017.
- Fazio G.P.,
- Redberg R.F.,
- Winslow T.,
- Schiller N.B.