Graphical abstract
Highlights
-
•
Plasma metabolomics revealed signatures in normal, NobCAD and ObCAD patients.
-
•
Metabolic signatures involved amino acids, carbohydrates, and lipid pathways.
-
•
These metabolic signatures increases our knowledge in atherosclerosis.
Glossary
- AMI
acute myocardial infarction
- BARD
Brazilian Artery Disease Study
- BCAA
Branched Chain Amino Acids
- BMI
body mass index
- CAD
coronary artery disease
- CTA
computed coronary tomography angiography
- HbA1C
glycated hemoglobin
- HDL
high-density lipoprotein
- HMDB
Human Metabolome Database
- HRP
high-risk plaques
- LDL
low-density lipoprotein
- NobCAD
non-obstructive coronary artery disease
- ObCAD
obstructive coronary artery disease
1. Introduction
More recently metabolomics, ie, the systematic study of small molecule byproducts of cellular metabolism, have gained substantial prominence for elucidating the molecular basis of diverse diseases, including cancer and cardiovascular disorders. Human plasma serves as a unique source of biologically relevant molecules essential for organismal function in both health and diseases, thereby enabling “liquid biopsy” approaches with high informational yield. Advances in mass spectrometry techniques have markedly expanded the analytical capabilities of these techniques [[1], [2], [3]]. Importantly, plasma metabolites are germane to basic mechanism of atherosclerosis including inflammation, endothelial functions, platelet function, matrix metalloproteinases, troponin, lipid and carbohydrates metabolism, and cellular proliferation [[4], [5], [6], [7]].
Several biological processes, such as inflammation and endothelial dysfunction, are common through all phases of atherosclerotic plaque formation, progression and complication, which lead to significant vasomotion alterations. These phases may occur silently over decades and often overlap temporally [4,5,7].
Metabolomics profiling may therefore uncover molecular signatures biomarkers capable of identifying individuals at risk who are not recognized by standard diagnostic tools, such as coronary computed tomography angiography (CTA) or traditional clinical risk scores. While CTA provides structural information, it does not capture the underlying molecular processes that drive biochemical processes underlying atherosclerosis [8].
Prior studies on metabolomics focused on specific aspects. Shah et al. [9], studied 478 patients following CABG aiming to identify biomarkers to improve risk stratification. Targeted mass/spectroscopy profiling of 69 metabolites allowed to identify three groups of metabolites associated with worse out-comes: a. Short-chain dicarboxylacyl carnitines; b. Ketone related metabolites; c. Short-chain acylcarnitines; they were independently predictive of adverse outcomes. Hence the aa. Concluded that metabolomics profiles are independently associated with adverse outcomes following coronary bypass grafting.
Bhatlacharya et al. [10], analyzed 63 metabolites from 1983 sequential patients subjected to cardiac catheterization; 995 had significant CAD, 23 had no CAD and 610 had no previous cardiac events. Two clusters of metabolites were associated with CAD: a. BCAA; b. Short chain acylcarnitines which included byproducts of BCAA metabolism.
Anlar et al. [11], compared metabolomics profiles in 221 patients, including normal and individuals with hypercholesterolemia, diabetes and hypertension. They observed changes in carbohydrate and lipid metabolism among those with CAD risk. Specifically, manitol/sorbitol/mannose, glucose and ribitol were increased while pregnenediol sulfate, aleoylearnitine and quinolinate were decreased among those with CAD risk.
Coronary atherosclerosis presents clinically as obstructive CAD (ObCAD) or non-obstructive CAD (NobCAD) –also known as subclinical atherosclerosis - although considerable superposition may occur. Importantly, this latter condition, once considered benign, has been associated with significant adverse outcomes, including myocardial infarction and mortality [12,13]. However, the molecular features associated with these phenotypes remain largely unexplored.
The aim of the present study was to characterize plasma metabolomics signatures of individuals with ObCAD and NobCAD and compare them with metabolomics profiles of individuals without coronary disease. Identifying metabolic patterns associated with different CAD phenotypes may deepen our understanding of atherosclerosis pathophysiology and ultimately support improved risk stratification and prevention strategies. This study represents an initial, hypothesis-generating step toward molecular characterization of atherosclerotic disease evolution.
2. Methods
2.1. Study population
This was a cohort study which included 60 patients from the Heart Institute (Incor) of the School of Medicine, University of São Paulo, Brasil, who were subjected to computed coronary angiography (CTA) for diagnostic purposes between 2011 and 2017. They were clinically stable. Reasons for the study included chest pain, positive functional ischemic tests or routine evaluation of cardiovascular risks. Three groups were analyzed: 1. Normal individuals (n = 17); group 2. Patients with non—obstructive CAD (NobCAD; n = 21; group 3- patients with obstructive CAD (ObCAD; n = 22. NobCAD is defined as stenosis < 50%, and ObCAD was considered if at least 1 lesion ≥ 50% was present in any coronary artery.
Exclusion criteria included: a-non– atherosclerotic cardiac diseases (valvulopathies, cardiomyopathies, pericardial diseases); b-chronic Kidney diseases (GFR < 30/mL/min/BSA); c-cancers; d-acute coronary syndrome within 90 days of study entry; e− DM1 (insulin dependent diabetes); f-current smoking.
Sample size- We used the Probabilistic Principal Component Analysis since we did not know previously the participant proteins considering a FDR (False Discovery Rate) 5% bins in 200, and significant proportional difference of 20%, the minimum n was calculated as 17 for each group (Software MetSize R version 2.0), [14].
The patients are part of the BARD Study (Brazilian Artery Disease) and provided written informed consent.
2.2. Metabolite extraction
Plasma samples were extracted using 200 μL of ice-cold extraction buffer (acetonitrile:methanol, 1v:1v), containing stable isotope reference compounds, at a concentration of 1 mg.mL-1 [succinic acid (D4, 98% - DLM 584-5), miristic acid (1,2,3–13C3, 99% - CLM 3665- 0.5) and palmitic acid (1,2,3,4 – 13C4)] at a 1:2 (v:v) sample: extraction buffer ratio (100μL). After vortexing for 5 min, the samples were centrifuged (15,000×g) for 10 min at 4°C for deproteinization. The supernatant fractions were then collected and evaporated to dryness using a vacuum concentrator. The resulting dried residues were suspended in 200 μL of methanol:water solution (4v:1v), vortexed and centrifuged again at 15,000×g for 10 min at 4°C. The supernatant was recovered and transferred to a new vial and kept at −80°C till analysis.
An aliquot (30 μL) of the extracted samples was lyophilized at −50°C under vacuum, derivatized with 30 μL of methoxamine-HCl (20 mg. mL-1) in pyridine for 16 h at room temperature, in the dark. Trimethylsilylation was performed by adding 30 μL of N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) containing 1% trimethylchlorosilane (TMCS) to the samples, followed by incubation, for 1 h at room temperature, in the dark. After the silylation, 30 μL of heptane was added.
The samples were analyzed in a randomized manner 30 min after the addition of heptane [15], using Gas chromatography/Mass spectrometry (GC-TOF/MS (Pegasus BT, LECO, St. Joseph, MI, USA) along with a control sample (blank) and a series of n-alkanes (C12 - C40), enabling the calculation of retention time indices [16]. The derivatized samples (1 μL) were automatically injected using a split/splitless injector in splitless mode and an automatic sampler (auto sampler-CTC Combi Pal Xt Duo (CTC Analytics AG, Switzerland), coupled to a gas chromatograph (8890 GC Agilent Technologie), equipped with 1 fused silica column [20 m length x 0.18 mm internal diameter (i.d) x 0.18 μm thickness (Agilent J&W Scientific)]. The injector temperature was set at 280°C, the septum purge rate was 20 mL.min-1, and it started after 300 s from the beginning of data acquisition. The gas flow through the column was 1 mL.min-1. The initial temperature was maintained at 80°C for 2 min, then ramped at 15°C. min-1 until reaching 305°C, and held at this temperature for 2 min. The column effluent was introduced into the ionization source of the mass spectrometer (GC-TOF/MS Pegasus BT (LECO Corporation, St. Joseph, MI, USA). The ionization source temperature was set to 250°C. Ions were generated by an electron source (70-eV) at an ionization current of 2.0 mA, and 20 spectros. s-1 were acquired in a mass range of 45–800 m/z, with the detector voltage adjusted to 1500 V.
All mass spectrometry (MS) data obtained were exported to ChromaTOF 2.12 Software (LECO Corporation), processed and identified. All adjustments such as baseline correction, noise exclusion, spectral deconvolution, peak detection and integration, and chromatogram alignment were performed using the R Software. The peaks identified by GC-MS analysis were individually verified, and the areas of compounds exhibiting isomerism were summed.
2.3. Coronary CTA acquisition and interpretation
CTA examinations were performed using a 320-detector row scanner (Canon Aquilion; 16 cm coverage) in 90% of cases and a 64-detector system in 10%, following standard protocols. Each CTA was independently reviewed by two trained technicians who compared their analyses with reports from two experienced angiographers. In cases of inconsistencies, the records were reviewed by two senior cardiologists to reach a consensus. Interobserver variability was 2.2%. No procedural complications occurred.
For the purpose of this study, the coronary circulation tree was divided into the major segments:.
-
(a)
left main, left anterior descending artery (and diagonal/septal branches),
-
(b)
circumflex artery (and marginal branches),
-
(c)
right coronary artery (and marginal branches).
ObCAD patients were additionally classified as having single-, double-, or triple-vessel disease. High-risk plaque (HRP) features were defined by the presence of ≥ 2 of the following: low attenuation, positive remodeling, calcification, or napkin-ring sign.
The study was approved by the institutional ethics committee and registered at the ClinicalTrials.gov (NCT06900270).
2.4. Protocol
A venous blood sample was collected at entry in the study for laboratorial measurements (Table 1). The following information was obtained: (Table 1). Age, gender, weight, blood pressure and height. BMI was derived from weight and height. Hypertension was defined as systolic blood pressure > 130/85 or use of antihypertensive medication. Obesity was defined as BMI > 30 and overweight as BMI > 26 but up to 29. Regarding exercise, patients were classified as sedentary or active (≥3 sessions per week). Positive family history was considered when at least one of the parents suffered cardiovascular events or interventions before 60 years of age. Active (≥10/cigarettes/day) smokers were excluded. Diabetes mellitus was considered when fasting plasma glucose ≥ 126 mg/dL, casual glucose > 140 mg/dL, glucose tolerance test > 200 mg/dL at 2 h, HbA1C higher than 6.5% or use of hypoglycemic agents. Dyslipidemia was considered when low density lipoprotein (LDL) was > 130 mg/dL, and triglycerides > 150 mg/dL, or when the patient referred to the use of anti-lipidemic medications.
Table 1.
Demographics.
| Demographics |
Group 1 (n = 17) |
Group 2 (n = 21) |
Group 3 (n = 22) |
P* |
|
|---|---|---|---|---|---|
|
Sex |
Female |
9 (53%) |
5 (23.8%) |
5 (22.7%) |
0.08 |
|
Male |
8 (47%) |
16 (76.2%) |
17 (77.3%) |
||
|
Race |
White |
14 (82.4%) |
20 (95.2%) |
18 (81.8%) |
|
|
Brown |
1 (5.9%) |
- |
1 (4.5%) |
||
|
Asian |
2 (11.8%) |
1 (4.7%) |
3 (13.6%) |
0.35 |
|
|
Age |
Mean (SD) |
66.0 ± 4.4 |
67.1 ± 3.9 |
66.1 ± 2.9 |
0.80 |
|
CAD family history |
9 (53%) |
15 (71.4%) |
13 (59.1%) |
0.5 |
|
|
Diabetes Mellitus |
3 (17.6%) |
4 (19%) |
5 (22.7%) |
0.92 |
|
|
Dyslipidemia |
7 (41.2%) |
17 (81%) |
22 (100%) |
0.001 |
|
|
Hypertension |
7 (41.2%) |
14 (66.6%) |
15 (68.2%) |
0.17 |
|
|
Sedentarism |
10 (58.8%) |
11 (52.4%) |
16 (72.7%) |
0.52 |
|
|
Obesity |
5 (29.4%) |
4 (19%) |
5 (22.7%) |
0,10 |
|
|
Ex- smoker |
5 (29.4%) |
11 (52.4%) |
9 (40.9%) |
0.35 |
|
|
Previous events |
AMI |
- |
- |
2 (9.1%) |
NA |
| Medications |
Antihypertensive |
8 (47%) |
17 (80.9%) |
19 (86.4%) |
0.001 |
|
Antilipidemic |
10 (58.8%) |
17 (80.9%) |
22 (100%) |
0.02 |
|
| Antidiabetic | 3 (17.6%) | 6 (28.6%) | 11 (50%) | 0.88 | |
| Laboratory tests | |||||
| Glucose (mg/dL) | 105.2 ± 26.1 | 103.5 ± 25.4 | 108.5 ± 20.3 | 0.78 | |
| Glycated Hemoglobin (HbA1c) (%) | 5.9 ± 0.96 | 5.9 ± 0.98 | 5.9 ± 0.72 | 0.94 | |
| Total Cholesterol (mg/dL) | 191.7 ± 48.4 | 173.6 ± 45.4 | 149.5 ± 34.8 | 0.012 | |
| HDL (mg/dL) | 53.5 ± 19.8 | 51.0 ± 10.5 | 43.3 ± 11.0 | 0.06 | |
| LDL (mg/dL) | 116.2 ± 41.9 | 101.0 ± 39.2 | 85.5 ± 30.9 | 0.045 | |
| Triglycerides (mg/dL) | 103.1 ± 36.4 | 108.0 ± 54.2 | 103.9 ± 45.5 | 0.93 | |
| Sodium (mEq/L) | 141.7 ± 1.5 | 142.4 ± 2.1 | 141.9 ± 1.8 | 0.51 | |
| Potassium (mEq/L) | 4.4 ± 0.26 | 4.3 ± 0.35 | 4.4 ± 0.33 | 0.51 | |
| CRP (mg/dL) | 1.9 ± 1.86 | 1.6 ± 2.51 | 2.85 ± 7.56 | 0.70 | |
| Urea (mg/dL) | 32.6 ± 8.8 | 32.0 ± 6.7 | 35.8 ± 10.6 | 0.34 | |
| Creatinine (mg/dL) | 0.9 ± 0.2 | 0.9 ± 0.1 | 1.0 ± 0.3 | 0.19 | |
| Alkaline Phosphatase (U/L) | 84.1 ± 23.9 | 78.0 ± 21.2 | 68.8 ± 6.7 | <0.001 | |
| AST (U/L) | 28.4 ± 16.4 | 27.5 ± 15.2 | 28.8 ± 10.6 | 0.95 | |
| ALT (U/L) | 27.2 ± 7.7 | 26.7 ± 11.0 | 28.9 ± 9.7 | 0.74 | |
| Gama GT (U/L) | 30.1 ± 13.9 | 32.4 ± 15.3 | 29.6 ± 12.6 | 0.78 | |
| Troponin I (mg/dL) | 3,2 ± 4.0 | 3.8 ± 5.6 | 5.0 ± 7.2 | 0.62 | |
Demographics and laboratory date at entry. AMI: acute myocardial infarction; AST: aspartate amino transferase; ALT: alanine aminotransferase; Gama GT: gama glutamil transferase; CRP – C reactive protein; HDL: high density cholesterol; LDL: low density cholesterol.
2.5. Statistical analyses
Categorical variables are expressed as absolute and relative frequencies. Continuous variables are reported as mean and standard deviation (SD). Baseline characteristics and events were compared. The association between categorical variables was evaluated by the chi-square test, continuous variables by ANOVA and post hoc comparisons by Bonferroni test. A significance level of 5% was adopted. SPSS for Windows v.25 was used for the analysis.
Metabolomic data were processed using MetaboAnalyst 6.0 [17]. Data were log-transformed and Pareto-scaled. Group differences were evaluated by ANOVA (FDR-adjusted p ≤ 0.05). Partial least squares discriminant analysis (PLS-DA), heatmaps, and variable importance in projection (VIP) scores were used to identify discriminating metabolites.
3. Results
3.1. Demographic characteristics
Sixty patients (mean age 66.4 ± 3.7 years) were included. Anti-hypertensive and anti-lipidemic medications were more frequently used in groups 2 and compared to group 1. Comparing group 1 with the sum of groups 2 and 3, anti-hypertensives and anti-lipidemic medications were more frequently used in the latter (p = 0.004). Also dyslipidemia was more frequent in group 3. There were no significant differences among the groups regarding laboratory data, except for total cholesterol LDL and alkaline phosphatase which are lower in group 3 (Table 1).
3.2. Metabolomics
Metabolic profiling was performed by GC/TOF-MS to identify changes in metabolite levels among three groups (Group 1: Normal individuals n = 17), Group 2: Individuals with NobCAD (n = 21), and Group 3: Individuals with ObCAD (n = 22). A total of 45 metabolites were identified. We found 42 metabolites differentially abundant among groups Table 2 (p ≤ 0.05, FDR adjusted; (Apendix A). In Table 3 the 5 most discriminating metabolites are shown. The differences among groups were explored by the supervised method of PLS-DA (Fig. 1). Based on it we found a clear separation among groups. Components 1 and 2 contribute with 31.2% and 14.4%, respectively, of the total (45.6%) variance. PLS-DA model fit was evaluated using the R2 and Q2 cross-validation performance measures [18], both of which vary between 0 and 1. R2 provides an indication of goodness of fit from the model, and Q2 provides an indication of the predictive ability. Our PLS-DA model comparing the 3 groups had R2 and Q2 values of 0.93 and 0.90, respectively.
Table 2.
The more abundant metabolites.
| Metabolite | p.value (≤0.05%) | FDR (≤0.05%) |
|---|---|---|
| Glyceric acid | 9.09E-43 | 3.82E-41 |
| Cholest-4-en-3-ol | 7.79E-22 | 1.64E-20 |
| Pseudo uridine | 1.91E-20 | 2.68E-19 |
| 3-Deoxytetronic acid/2-Deoxytetronic acid | 2.87E-20 | 3.02E-19 |
| Talofuranose | 4.33E-19 | 3.64E-18 |
| Tocopherol | 2.46E-17 | 1.72E-16 |
| Glycerol | 2.60E-16 | 1.56E-15 |
| Palmitelaidic acid | 6.91E-15 | 3.63E-14 |
| Glutamic acid | 1.35E-13 | 6.29E-13 |
| Asparagine | 2.38E-13 | 1.00E-12 |
| 4-Methyl 2-keto pentanoic acid enol | 7.04E-12 | 2.69E-11 |
| 2,3-Dihydroxybutanoic acid | 1.50E-10 | 5.25E-10 |
| Ribose | 7.07E-10 | 2.28E-09 |
| Isoleucine | 3.65E-09 | 1.10E-08 |
| Allofuranose | 1.88E-07 | 5.25E-07 |
| Phenylalanine | 1.39E-05 | 3.66E-05 |
| Stearic acid | 1.95E-05 | 4.82E-05 |
| Erythronic acid | 0.00010066 | 0.00023488 |
| Palmitic Acid | 0.0001367 | 0.00030219 |
| 9,12-Octadecadienoic acid | 0.00042433 | 0.00089108 |
| Glycine | 0.00073578 | 0.0014071 |
| Tyrosine | 0.00073704 | 0.0014071 |
| Citric acid | 0.003646 | 0.006658 |
| 2-Aminomalonic acid | 0.0050733 | 0.0088782 |
| 2-Butenedioic acid | 0.0085631 | 0.014386 |
| 9-Octadecenoic acid | 0.009379 | 0.015151 |
| Lysine | 0.014447 | 0.022473 |
| Cystine | 0.022118 | 0.033177 |
| Valine | 0.02656 | 0.038466 |
ANOVA showing the significantly statistical differences among the average of the abundance of 29 metabolites out of the total 42 identified for the 3 groups (G_01, G_02 and G_03) analyzed, with (p≤0.05) and FDR ajusted ( ≤ 0.05).
Table 3.
The five more abundant metabolites in the whole population.
| Metabolite | Class | Sub-Class | Vip Score | p-Value (≤0.05) |
|---|---|---|---|---|
| Isoleucine | Carboxylic acids and derivatives | Amino acids, peptides, and analogues | 1.96 | 3.65E-09 |
| Talofuranose | Organooxygen compounds | Carbohydrates and carbohydrate conjugates | 1.93 | 4.33E-19 |
| 2,3-Dihydroxybutanoic acid | Organooxygen compounds | Carbohydrates and carbohydrate conjugates | 1.91 | 1.50E-10 |
| 3-Deoxytetronic acid/2-Deoxytetronic acid | Organooxygen compounds | Carbohydrates and carbohydrate conjugates | 1.87 | 2.87E-20 |
| Palmitelaidic acid | Fatty Acyls | Fatty acids and conjugates | 1.65 | 6.91E-15 |
The 5 most discriminating metabolites identified.
Fig. 1.
PLS-DA scores plot showing a significant separation (R2 = 0.92 and Q2 = 0.89) among groups. Component 1 and 2 contribute with 31.8% and 14%, respectively, to the total (45.8%) variance. The total number of patients were 17, 21 and 22 for group 01, 02 and 03, respectively.
The measurement of variance importance of interested metabolites was done by Variable Importance in Projection (VIP) score to identify the potential biomarker that contributed to the model (Fig. 2). The red-, blue- and yellow-colored boxes on the right side of VIP scores plot indicate a high, low and intermediate relative abundance of the metabolites in each group. VIP is a weighted sum of squares of the PLS loadings which indicates the importance of the variable to the whole model. To identify the variables that had the most significant contribution in discriminating metabolite profiles among the groups, we considered the variable importance in the projection (VIP ≥1.0) combined with (p ≤ 0.05) from ANOVA. Based on it, we found 14 metabolites differentially abundant among groups, distributed according to their chemical classes (Fig. 2; appendix B). Specifically, in Group 1, 9 metabolites were highly abundant, ie, glycerol, palmitelaidic acid, cholesterol4—en-3-ol, talofuranose, 3 deoxytetronic acid/2 deoxytetronic acid, palmitic acid and glyceric acid; also 6 metabolites had low abundance. In Group 2, median or low abundance of all metabolites were observed. Finally, in Group 3, 6 metabolites were more abundant, ie.,isoleucine, 2,3 dihidroxybutanoic acid, glutamic acid, asparagine, 4 methyl 2-keto pentanoic acid ribose; and low abundance of 8 metabolites as shown in Fig. 3. Thus, there was clear differences between Groups I and 3. Overall, these metabolites are involved in carbohydrates and lipid metabolism, as well as energy production.
Fig. 2.
The variable importance in projection (VIP) scores plot from PLS-DA analysis of interested metabolites, among groups.
Fig. 3.
Heatmap based on the relative abundance of the 29 metabolites showing major statistically significant differences (p ≤ 0.05, FDR ≤ 0.05) among groups (G_01, G_02 and G_03). The colours of the value bar code: red, blue and white, show the high, low and intermediate abundance of the metabolites in each group, illustrating the magnitude of the variability among groups.
These metabolites mainly mapped to pathways involving carbohydrate metabolism, lipid metabolism, and energy production. The heatmap (Fig. 3) illustrates clear separation among groups based on relative abundance of the 25 most significantly altered metabolites (p ≤ 0.05; FDR ≤ 0.05).
3.3. Characteristics of coronary circulation
Considering the extent of CAD, in the NobCAD only lesions < 50% were present. On the other hand, among ObCAD, 2 patients had 1 vessel disease, 8 had two vessels and 12, tri vessels. ObCAD had far more extensive diseases than Group 2. On the other hand, only 4 pts had high risk plaques, of which 3 were in Group 3 and 1 in Group 2. Such small numbers preclude any statistical comparison.
4. Discussion
To date the classification and management of coronary artery disease (CAD) have relied primarily on classical risk factors and on structural assessments of the coronary arteries and myocardium. In the present study, using an untargeted metabolomic approach, we identified a novel dimension of disease characterization: metabolic signatures that differ significantly among healthy individuals, patients with non-obstructive CAD (NobCAD), and those with obstructive CAD (ObCAD). These findings suggest that coronary atherosclerosis is not merely a linear continuum of anatomical severity but is instead characterized by distinct metabolic states, reflecting qualitatively different pathobiological processes across CAD phenotypes.
The differentially abundant metabolites were primarily associated with pathways related to amino acids, carbohidrates and fatty acids as shown in Table 2. Isoleucine is a BCAA, which play pivotal roles in obesity and insulin resistance; its basic actions occur in the mitochondria through oxidation leading to generation of acetyl-CoA in the TCA cycle, therefore contributing to cellular byosynthetic pathways. In a clinical study, Bhattacharya et al. [10], observed that BCAA and short chain acylcarnitines were associated with CAD in patients referred for cardiac catheterization. Given their ample chemical actions, BCAA are involved in CVD such as hypertension and atherosclerosis [19].
Carbohydrate metabolism (Talofuronase, 2,3-Dihydroxybutanoic acid, 3-Deoxytetronic acid)) is a well-recognized participant of atherosclerosis [20]. Also, Muñoz –Cabrejas [21] observed an inverse relationship between a carbohydrate quality index and subclinical atherosclerosis in the femoral arteries of 2074 middle-aged males, indicating that high quality carbohydrates are protective against atherosclerosis. In addition Anlar et al. [11], documented distinct changes in carbohydrates and lipid metabolism linked to CAD risk. Particularly, manitol/sorbitol, mannose, glucose and ribitol were increased in higher CAD risk, while pregnenediol sulfate, oleoyl carnitine and quinolinate were decreased with higher CAD risk.
Finally, palmitelaidic acid is a trans-fatty acid linked to inflammation which is a fundamental mechanism of atherosclerosis as demonstrated in the Cantos trial [22]. Hence, these metabolites are all biologically relevant to the development and progression of atherosclerosis [5].
As shown in Fig. 3, normal individuals demonstrated predominance of lipid- and energy-related metabolites; NobCAD patients exhibited generally low abundance across all analyses; and ObCAD patients displayed metabolic enrichment primarily in amino acid–related and energy-associated compounds. These findings suggest distinct biochemical processes in advanced CAD compared with milder or absent disease.
A key finding of the present study is that patients with NobCAD do not simply represent an intermediate metabolic phenotype between normal individuals and those with obstructive disease. Instead, this group exhibited a globally attenuated metabolomics profile, with lower relative abundance of most discriminant metabolites. From a preventive perspective, this observation reinforces the concept that NobCAD is a clinically meaningful entity, associated with adverse outcomes despite the absence of flow-limiting lesions, and may represent a window of opportunity for preventive interventions.
In contrast, patients with ObCAD showed a marked predominance of metabolites related to amino acid metabolism (isoleucine, 2, 3 dihydroxy butanoic acid, glutamic acid, asparagine, 4-metyl- 2- ketopentanoic acid) and energy production (ribose). On the other hand, 8 metabolites had low abundance (Fig. 3).
Such metabolic alterations may reflect systemic consequences of chronic inflammation, endothelial dysfunction and plaque progression as well as underlying ischemia all of which are central targets in preventive cardiology.
Notably, the NobCAD and ObCAD groups exhibited a higher prevalence of hypertension and greater use of lipid-lowering therapies and anti-hypertensives compared with normal individuals. Aldosterone receptor antagonists [23] and statins [24] have been shown to promote plaque stabilization and reduce plaque volume, respectively, and may therefore have influenced the observed metabolic patterns, although the magnitude of this effect remains uncertain. Clinical events were infrequent, with only two cases of acute myocardial infarction, both occurring in the ObCAD group. Collectively, these findings underscore the limitations of risk-stratification strategies that rely solely on clinical characteristics.
The observed biochemical profiles aligned with disease extent. While NobCAD patients predominantly exhibited mild lesions, individuals with ObCAD more frequently had two- or three-vessel disease. Although high-risk plaque (HRP) features were uncommon, three of four HRPs occurred in ObCAD, limiting any conclusions regarding plaque vulnerability.
Our findings must be interpreted with caution, as residual confounding cannot be excluded. In addition to the antihypertensive agents and statins previously discussed, factors such as diet, physical activity, and fasting status were not assessed. Furthermore, the absence of a validation cohort limits the strength of the conclusions that can be drawn. Given the small sample size — and considering that only metabolites present in at least 70% of samples within each group were deemed valid — we focused on the first five metabolites listed in Table 3, which demonstrated the most significant differences among the three groups. This strategy aimed to reduce the risk of overfitting, even after applying FDR correction. Importantly, the lack of functional assays precludes causal inference; therefore, no cause–effect relationships can be established.
Overall, these metabolites may represent downstream consequences, compensatory responses, or unmeasured confounders at a specific stage of atherosclerosis evolution. Hence, they should be interpreted as systemic markers that differentiate coronary phenotypes, rather than mechanistic drivers of disease. In conclusion, the present study must be considered only as a hypothesis generating contribution which may inspire future investigations.
5. Strengths and limitations
This is one of the first studies to compare metabolomics signatures across normal, NobCAD, and ObCAD phenotypes using an untargeted approach. High-quality CTA interpretation with low interobserver variability ensured robust phenotypic classification. Use of stringent statistical methods (PLS-DA, FDR correction, VIP analysis) enhances confidence in the identified metabolic signatures.
Some limitations must be acknowledged. Because this is a cross-sectional analysis, it provides only a metabolic “snapshot” and does not capture temporal changes. Plasma metabolomics reflects a dynamic biological process, and a single time-point measurement may not fully represent the spectrum of metabolic alterations occurring throughout the course of atherosclerosis. Functional assays were not performed, preventing mechanistic or causal inference. The sample size is adequate for exploratory metabolomics analysis but remains limited for evaluating associations with clinical outcomes. The low prevalence of high-risk plaque features precludes conclusions regarding metabolomics correlations of plaque vulnerability.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
To Dr. Michelle Aparecida Pereira for her competent work as Research Assistant. To the tecnicians Lanai N. Grossi, Giovana Sisnandes dos Santos, Natalia S. Videira, Manuela M. Ventura Ferreira, for valuable assistance. We thank specifically the Laboratório Max Feffer de Genética de Plantas Departamento de Genética-Esalq-Univ de São Paulo, Br which performed the metabolomic analysis.
This work was supported by the São Paulo Research Foundation (FAPESP), Grant No. 2023/13581-3; Fundação E. J. Zerbini; Instituto de Gestão, Planejamento e Desenvolvimento da Vitivinicultura do Estado do Rio Grande do Sul; and Cooperativa de Crédito Serrana do Rio Grande do Sul.
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.athplu.2026.100611.
Appendix A. Supplementary data
The following are the Supplementary data to this article.
References
- 1.McGarrah R.W., Crown S.B., Zhang G.F., Shah S.H., Newgard C.B. Cardiovascular metabolomics. Circ Res. 2018 Apr 27;122(9):1238–1258. doi: 10.1161/CIRCRESAHA.117.311002. PMID: 29700070; PMCID: PMC6029726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Wishart D.S., Guo A., Oler E., Wang F., Anjum A., Peters H., et al. Hmdb 5.0: the human metabolome database for 2022. Nucleic Acids Res. 2022;50(D1):D622–D631. doi: 10.1093/nar/gkab1062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Pibiri M., Noto A., Dalu A., Muntoni S., Kopeć K.K., Spada M., et al. Metabolomics signatures of atherosclerosis in cardiovascular disease: a narrative systematic review. J Clin Med. 2025 Nov 12;14(22):8028. doi: 10.3390/jcm14228028. PMID: 41303064; PMCID: PMC12653451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ross R. Atherosclerosis--an inflammatory disease. N Engl J Med. 1999 Jan 14;340(2):115–126. doi: 10.1056/NEJM199901143400207. PMID: 9887164. [DOI] [PubMed] [Google Scholar]
- 5.Libby P. The changing landscape of atherosclerosis. Nature. 2021 Apr;592(7855):524–533. doi: 10.1038/s41586-021-03392-8. Epub 2021 Apr 21. PMID: 33883728. [DOI] [PubMed] [Google Scholar]
- 6.Ference B.A., Braunwald E., Catapano A.L. The LDL cumulative exposure hypothesis: evidence and practical applications. Nat Rev Cardiol. 2024 Oct;21(10):701–716. doi: 10.1038/s41569-024-01039-5. Epub 2024 Jul 5. PMID: 38969749. [DOI] [PubMed] [Google Scholar]
- 7.Gimbrone MA Jr, Topper J.N., Nagel T., Anderson K.R., Garcia-Cardeña G. Endothelial dysfunction, hemodynamic forces, and atherogenesis. Ann N Y Acad Sci. 2000 May;902:230–239. doi: 10.1111/j.1749-6632.2000.tb06318.x. discussion 239-40. PMID: 10865843. [DOI] [PubMed] [Google Scholar]
- 8.Serruys W., Hara H., Garg S., Kawashima Hideyuki, Nørgaard Bjarne L., Dweck Marc R., et al. Coronary computed tomographic angiography for complete assessment of coronary artery disease: JACC state-of-the-art review. J Am Coll Cardiol. 2021;78(7):713–736. doi: 10.1016/j.jacc.2021.06.019. [DOI] [PubMed] [Google Scholar]
- 9.Shah S.H., Bain J.R., Muehlbauer M.J., Stevens R.D., Crosslin D.R., Haynes C., et al. Association of a peripheral blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular events. Circ Cardiovasc Genet. 2010 Apr;3(2):207–214. doi: 10.1161/CIRCGENETICS.109.852814. Epub 2010 Feb 19. PMID: 20173117. [DOI] [PubMed] [Google Scholar]
- 10.Bhattacharya S., Granger C.B., Craig D., Haynes C., Bain J., Stevens R.D., et al. Validation of the association between a branched chain amino acid metabolite profile and extremes of coronary artery disease in patients referred for cardiac catheterization. Atherosclerosis. 2014;232(1):191–196. doi: 10.1016/j.atherosclerosis.2013.10.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Anlar G.G., Anwardeen N., Al Ashmar S., Pedersen S., Elrayess M.A., Zeidan A. Metabolomics profiling of stages of coronary artery disease progression. Metabolites. 2024;14(6):292. doi: 10.3390/metabo14060292. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Da Luz P.L., Favarato D., Abizaid A.A.C., Cezar L., Serrano C., Chagas A.C.P., et al. Comparison of coronary tomography angiography and invasive coronary angiography in non-obstructive CAD: Brazilian artery disease study (BARD) Arq bras cardiol. 2026;123(3) doi: 10.36660/abc.20250702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ahmadi A., Argulian E., Leipsic J., Newby D.E., Narula J. From subclinical atherosclerosis to plaque progression and acute coronary events: JACC state-of-the-art review. J Am Coll Cardiol. 2019 Sep 24;74(12):1608–1617. doi: 10.1016/j.jacc.2019.08.012. PMID: 31537271. [DOI] [PubMed] [Google Scholar]
- 14.Pang Z., Lu Y., Zhou G., Hui F., Xu L., Viau C., et al. MetaboAnalyst 6.0: towards a unified platform for metabolomics data processing, analysis and interpretation. Nucleic Acids Res. 2024 Jul 5;52(W1):W398–W406. doi: 10.1093/nar/gkae253. PMID: 38587201; PMCID: PMC11223798. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Gullberg J., Jonsson P., Nordström A., Sjöström M., Moritz T. Design of experiments: an efficient strategy to identify factors influencing extraction and derivatization of Arabidopsis thaliana samples in metabolomic studies with gas chromatography/mass spectrometry. Anal Biochem. 2004 Aug 15;331(2):283–295. doi: 10.1016/j.ab.2004.04.037. PMID: 15265734. [DOI] [PubMed] [Google Scholar]
- 16.Schauer N., Steinhauser D., Strelkov S., Schomburg D., Allison G., Moritz T., et al. GC-MS libraries for the rapid identification of metabolites in complex biological samples. FEBS Lett. 2005 Feb 28;579(6):1332–1337. doi: 10.1016/j.febslet.2005.01.029. Epub 2005 Jan 28. PMID: 15733837. [DOI] [PubMed] [Google Scholar]
- 17.Xia J., Wishart D.S. Web-based inference of biological patterns, functions and pathways from metabolomic data using MetaboAnalyst. Nat Protoc. 2011 Jun;6(6):743–760. doi: 10.1038/nprot.2011.319. Epub 2011 May 5. PMID: 21637195. [DOI] [PubMed] [Google Scholar]
- 18.Human metabolome database (HMDB). Available from: https://hmdb.ca.Acessed 22 January 2026.
- 19.McGarrah R.W., White P.J. Branched-chain amino acids in cardiovascular disease. Nat Rev Cardiol. 2023;20(2):77–89. doi: 10.1038/s41569-022-00760-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Luo S., Li L., Chen H., Wei J., Yang D. Glucose metabolism reprogramming of vascular endothelial cells and its implication in development of atherosclerosis. Rev Cardiovasc Med. 2024 Nov 22;25(11):423. doi: 10.31083/j.rcm2511423. PMID: 39618869; PMCID: PMC11607510. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Muñoz-Cabrejas A., Laclaustra M., Guallar-Castillón P., Casasnovas J.A., Jarauta E., Sandoval-Insausti H., et al. High-quality intake of carbohydrates is associated with lower prevalence of subclinical atherosclerosis in femoral arteries: the AWHS study. Clin Nutr (Phila) 2021 Jun;40(6):3883–3889. doi: 10.1016/j.clnu.2021.04.049. Epub 2021 May 14. PMID: 34134004. [DOI] [PubMed] [Google Scholar]
- 22.Ridker P.M., Everett B.M., Thuren T., MacFadyen J.G., Chang W.H., Ballantyne C., et al. CANTOS trial group. Antiinflammatory therapy with canakinumab for atherosclerotic disease. N Engl J Med. 2017 Sep 21;377(12):1119–1131. doi: 10.1056/NEJMoa1707914. Epub 2017 Aug 27. PMID: 28845751. [DOI] [PubMed] [Google Scholar]
- 23.Alkatiri A.H., Firman D., Alkatiri A.A., Suryajaya P.I., Sudharsono A. The role of angiotensin antagonism in coronary plaque regression: insights from the glagovian model. Int J Vasc Med. 2021 Apr 5;2021 doi: 10.1155/2021/8887248. PMID: 33880191; PMCID: PMC8046567. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dawson L.P., Lum M., Nerleker N., Nicholls S.J., Layland J. Coronary atherosclerotic plaque regression: JACC state-of-the-art review. J Am Coll Cardiol. 2022 Jan 4;79(1):66–82. doi: 10.1016/j.jacc.2021.10.035. PMID: 34991791. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.