Introduction

Cardiovascular diseases (CVDs) are the leading cause of death worldwide. Among them, ischemic heart disease (IHD), also called coronary heart disease (CHD), represents a major cause of mortality and disability1,2, accounting for over 9 million deaths annually3,4,5. IHD, also known as coronary heart disease and including asymptomatic myocardial ischemia, angina pectoris, ischemic cardiomyopathy, and myocardial infarction, arises from inadequate myocardial perfusion in the vast majority of cases due to obstruction of the coronary arteries by atherosclerosis6,7. Clinical evidence suggests that the incidence and prevalence of IHD generally increase with age8 alongside other contributing factors such as hypertension, diabetes mellitus, dyslipidemia, and smoking9. These factors are components of the metabolic syndrome, which promotes vascular inflammation, atherosclerotic plaque formation, and thrombosis, thereby increasing the risk of CVDs10,11.

The development of atherosclerosis involves multiple stages12. Plaque formation is triggered by the retention and modification of lipoproteins, mainly low-density lipoprotein (LDL) and remnants of triglyceride-rich lipoproteins, in the intima of the vessel wall. This is accompanied by endothelial dysfunction, which leads to pathogenic changes in microvascular tone, permeability, and leukocyte adhesion13. Upon entering the intima, monocytes differentiate to macrophages and take up modified lipoproteins to give rise to foam cells, which exacerbate vascular inflammation, remodeling and lesion progression. Plaque growth stimulates neoangiogenesis as an adaptive response to hypoxic conditions. Vascular smooth muscle cells migrate into the intima and proliferate, leading to fibrous cap formation, plaque destabilization and rupture, ultimately culminating in thrombosis and contributing to acute coronary syndrome14.

Currently, the diagnosis of IHD primarily relies on coronary angiography, which assesses the anatomy and functional status of the coronary arteries11. However, accurate diagnosis remains challenging due to technical limitations, the complexity of coronary anatomy, and the heterogeneous nature of atherosclerotic plaques15. Early disease detection and timely intervention are crucial for preventing progression and reducing morbidity and mortality16. In this context, identifying reliable biomarkers for the early diagnosis and risk stratification of CVDs holds significant scientific and clinical importance.

Metabolomics, an emerging field within omics sciences, is rapidly advancing in biomedical research. In recent years, Nuclear Magnetic Resonance (NMR) spectroscopy has been widely used for quantitative metabolite analysis, enabling a comprehensive and unbiased profiling of small molecules in biofluids and tissues17,18,19,20. These advancements have greatly improved our understanding of molecular biomarkers associated with various diseases, facilitating early diagnosis and the development of targeted therapeutic strategies18,21. Metabolome reflects the ongoing (patho)physiological state of the body or of a cell/tissue, and provides valuable insights into the biochemical processes underlying a disease that potentially could be used as targets for drug development22,23. In the coming years, metabolomics is expected to become a routine tool for disease diagnosis, health monitoring, aging research, and drug development24. Metabolomics is particularly valuable for metabolic disorders such as CVDs, where function and metabolism are inextricably linked at both the systemic and cardiac levels25. Risk factors for CVDs including obesity, type 2 diabetes, or hyperlipidemia have a metabolic origin and directly impact cardiac metabolism. The heart is a highly active metabolic organ able of converting chemical energy to mechanical energy. This occurs by metabolising various types of energy-providing substrates that are taken up from the blood circulation, and by changing the efficiency of utilization of different substrates as an adaptive response to altered physiological or metabolic environment26. Each of these substrates is converted into metabolic intermediates, entering the tricarboxylic acid (TCA) cycle either as acetyl-CoA or as a cycle intermediate for the generation of reducing equivalents, which are used for ATP synthesis through the mitochondrial oxidative phosphorylation. In the absence of oxygen, ATP production from mitochondrial oxidative phosphorylation declines in relation to the degree of ischemia, and the heart switches to alternative energy substrates for anaerobic ATP production26.

All these aspects can be captured by metabolomics and especially in the context of IHD, which involve disturbances in cardiac metabolism25,27,28 as a consequent of ischemia, leading to changes in substrate metabolism and contractile function. Correspondingly, metabolomics approaches on animal models and human cohort samples have been deployed for pathway identification and novel predictive/diagnostic/prognostic biomarker discovery in CVDs29,30,31,32, and for improving disease classification, risk stratification, and personalized medicine. As myocardial tissue samples are difficult to be obtained in humans, metabolomic profiling of serum or plasma has been extensively used, though reflective of changes in systemic metabolism28, not cardiac metabolism per se. Findings from human blood metabolomics suggested a role for some lipid species, amino acids, organic acids, and TCA cycle intermediates as potential biomarkers in heart diseases, including IHD33,34,35. However, discrepancies in IHD metabolites exist among studies due to differences in clinical cohorts, control conditions, and the metabolomic platforms used. Metabolomic platforms consist in mass spectrometry (MS)-based techniques, particularly liquid chromatography–mass spectrometry (LC-MS) and gas chromatography–mass spectrometry (GC-MS) and nuclear magnetic resonance spectroscopy (NMR). MS based platforms offer high sensitivity and are capable of detecting a wide range of metabolites, often at very low concentrations. They are typically used in targeted or semi-targeted approaches for the detection of specific metabolite classes, such as lipids, amino acids, or organic acids. In contrast, NMR spectroscopy provides a non-destructive, highly reproducible, and unbiased method for untargeted metabolic profiling of biofluids. Although NMR is generally less sensitive than MS-based techniques, it allows for absolute quantification without the need for external standards, requires minimal sample preparation, and offers rich structural information. Moreover, the robustness and reproducibility of NMR make it particularly well-suited for clinical research and longitudinal studies22,36.

Moreover, pericardial fluid has gained growing interest for the diagnosis and treatment of CVDs due to its anatomical proximity to the myocardium, its low clearance rates and origin as output of plasma ultrafiltration from the epicardial capillaries, secretions from pericardial mesothelial cells and contributions of the myocardial interstitial space37. Being a concentrate of heart-released factors, molecular signature of pericardial fluid could reliably inform on structural and functional parameters of the heart and its pathophysiological status38,39. Notwithstanding this, pericardial fluid is accessed during surgery, thus few reports have investigated the molecular features of pericardial fluid in CVDs38,40 including IHD, and even less through metabolomics39. Metabolite profiles with the ability to reliably discriminate ischemic from non-ischemic patients would be of great value and could have crucial implications in mechanistic elucidation of heart disease and cardiac clinical practice.

The present study aims at identifying key metabolites and metabolic pathways of IHD that may allow for an earlier intervention and/or a more effective approach to treatment. To this aim, we analyse both plasma and pericardial fluid, which has emerged as a sensitive tool for studying pathophysiology and biomarkers in heart diseases38,41, but has never been scrutinized for its metabolome signature in IHD42. Plasma and pericardial fluid samples are obtained from patients diagnosed with IHD and compared with samples from non-ischemic patients with valvular heart disease. Using a metabolomic approach based on ¹H-NMR spectroscopy combined with multivariate data analysis, we identify significant metabolic variations between the two groups, both in plasma and pericardial fluid. We observe increased concentrations of 3-hydroxybutyrate in both biological fluids, together with elevated succinate in pericardial fluid, indicating alterations in mitochondrial energy metabolism. Additional changes involve pathways linked to substrate utilization and redox balance. The metabolic signature of IHD may facilitate disease stage differentiation, offering insights relevant to early diagnosis and clinical characterization.

Materials and methods

Study design

The present research was part of the “cardio-miRNA” study, which had the objective to examine microRNA signature in IHD patients. The study was conducted following the Declaration of Helsinki and approved by the Ethics Committee of ASL Lecce-Puglia Salute in May 2023 (ethical clearance n. 47752/2023). All participants signed an informed consent to participate. Patients (IHD and controls) undergoing surgery for cardiovascular diseases. The IHD group (n = 10) consisted of patients diagnosed with IHD. The control group (n = 10) comprises patients with mitral valve insufficiency, where the surgical procedure was strictly related to a valvular malfunction and not to myocardial ischemia. Inclusion and exclusion criteria were as follows:

Inclusion criteria:

  1. 1.

    Male and female subjects aged between 18 and 70 years;

  2. 2.

    Body Mass Index (BMI) between 25 and 30;

  3. 3.

    Patients with: a) mitral valve insufficiency due to mitral valve prolapse requiring surgery to be performed via mini-thoracotomy or undergoing coronary angiography with intact coronary arteries; b) IHD undergoing bypass surgery with the left internal mammary artery graft and/or with saphenous vein;

  4. 4.

    Patients with a left ventricular ejection fraction >50%;

  5. 5.

    Elective patients: non-urgent surgery;

  6. 6.

    Signed informed consent by the patient or legal guardian.

Exclusion criteria:

  1. 1.

    Diagnosis of diabetes;

  2. 2.

    Presence of systemic diseases and/or inflammations (chronic renal failure requiring dialysis, liver failure, liver cirrhosis, etc.);

  3. 3.

    Presence of pathological pleural and pericardial effusions (pericardial >1 cm in thickness; pleural >4 cm in thickness);

  4. 4.

    History of pleuritis and endocarditis;

  5. 5.

    Presence of hematologic and blood coagulation disorders;

  6. 6.

    Presence of autoimmune diseases except for autoimmune thyroiditis;

  7. 7.

    Presence of severe diseases (systemic lupus erythematosus, rheumatoid arthritis, COPD, etc.) not related to CVD, but capable of disturbing the profile of circulating markers.

Plasma was collected in EDTA-tubes from each patient before the surgical procedure. Additionally, about 4 ml of pericardial fluid was collected from each patient during the surgery. Plasma and pericardial fluid samples were immediately centrifuged (3000 rpm at 4 °C for 10 min), aliquoted and stored at −80 °C until analysis.

Plasma and pericardial fluid preparation for NMR analysis

Plasma samples were prepared in accordance with standard protocols for NMR metabolomics investigations36. Briefly, prior to NMR analysis, the frozen samples were thawed at room temperature and shaked. 200 μL aliquots were mixed with 400 μL of a saline buffer solution (50 mM sodium phosphate buffer in D2O containing 0.05% of the internal standard TSP, trimethylsilyl) propionate-2,2,3,3,-d4, for chemical shift calibration, and 4.5 mM NaN3, at a pH 7.4) to minimize the variation in the pH. The entire prepared mixture was then transferred into the NMR tube to a 5 mm NMR tube for analysis.

Pericardial fluid samples were prepared according to a previous reported procedure42. Before NMR analysis, the frozen samples were thawed and shaked, then centrifuged for 10 min at 13000 g at 4 °C. For the NMR analysis, 300 μl of filtered PF were mixed with 400 μl of a 0.09 M phosphate buffer solution at pH = 7.4, in D2O containing 0.05% of the internal standard TSP. 650 μl of the final solution were transferred into a 5 mm NMR tube.

NMR experiments

NMR experiments were conducted on a Bruker Avance III spectrometer (Bruker, Ettlingen, Germany) operating at 600.13 MHz for 1H observation. The spectrometer was equipped with a TCI cryoprobe featuring a z-axis gradient coil and automatic tuning-matching functionality. Measurements were carried out at 300 K in automation mode following sample loading via the Bruker Automatic Sample Changer, which was integrated with IconNMR software version 5.0. (Bruker).

For each sample, a standard 1D 1H NMR spectrum was recorded using a Carr–Purcell–Meiboom–Gill (CPMG) spin-echo sequence to suppress water and broad protein signals. This spectrum acquisition involved 128 transients, 16 dummy scans, a 5 s relaxation delay, and an FID size of 64 K data points with a spectral width of 12,019.230 Hz (equivalent to 20.0276 ppm). The acquisition time was set to 1.36 s, with a fixed echo delay of 1.2 ms (d20). The total spin-spin relaxation delay duration was 302.4 ms, during which solvent signal suppression was applied. Before Fourier transformation, automated phasing, and baseline correction, the resulting FIDs were multiplied by an exponential weighting function corresponding to a line broadening of 0.3 Hz. Additionally, 2D NMR spectra—including 1H JRES1,H-1H COSY1,H-13C HSQC, and 1H-13C HMBC, Figs. S1–S8—were occasionally acquired to facilitate spectral assignment by comparing results with published literature and public databases43. All spectra were processed and analyzed using Topspin 3.6.1 and Amix 3.9.13 software (Bruker Biospin, Italy), allowing for visual inspection alongside the subsequent bucketing process required for multivariate statistical analysis.

Multivariate analysis of NMR data

The bucketing pre-processing procedure was applied on the CMPG spectra, covering the range 10.0–0.5 ppm, with the exclusion of the spectral region between 5.00-4.7 ppm (containing the residual peak from the suppressed water resonance) obtaining 475 buckets. Each spectrum was segmented using the simple rectangular bucket option of the AMIX software (Bruker Biospin) in fixed rectangular regions (buckets) of 0.02 ppm width which were successively integrated. The total sum normalization was applied to reduce small differences due to sample concentration and/or experimental conditions among samples. The CPMG spectra were processed and multivariate statistical analyses (unsupervised principal component analysis, PCA and the supervised orthogonal partial least squares discriminant analyses, OPLS-DA), were performed using Metaboanalyst 6.0 software.

The Partial Least Squares, PLS, regression models were performed using SIMCA 14 software.

Key metabolites identified through discriminant loadings in the OPLS-DA S-line plot were subsequently quantified by evaluating the integrals of select unbiased NMR signals using Amix 3.9.14 (Bruker Biospin, Italy). Results, presented as mean intensities with corresponding standard deviations (SD) of distinct NMR peaks, were statistically validated through a univariate t-test via the MetaboAnalyst platform. Statistical significance was determined at a minimum threshold of p < 0.05 with a 95% confidence interval.

Potential biomarkers identification

The metabolites identified as potential biomarkers in the S-line plots, derived from the OPLS-DA models, were further analysed with the characteristic (ROC) curve analysis using Metaboanalyst 6.0 software. The area under curve (AUC) value was used to evaluate the predictive performance of selected biomarker. ROC curve analysis was performed on all characteristic metabolites identified between the IHD and control groups, calculating AUC values for each metabolite. AUC values > 0.9 were considered as potential biomarkers for IHD early detection.

Metabolic pathway analysis

The MetaboAnalyst 6.0 software was used for the analysis of metabolic pathways based on the significant metabolites identified through prior supervised OPLS-DA models. Metabolites of interest were selected by targeting distinct, unbiased NMR spectral bins (VIP > 1), as indicated in the corresponding VIP plots, which then served as the input matrix for the metabolic pathway analysis. The pathway impact was quantified by summing the importance measures of metabolites aligned to a specific pathway and normalizing this value against the total importance measures of all metabolites associated with that pathway. This approach provided insight into the pathways most significantly altered within the study44.

Statistics and reproducibility

Statistical analyses were performed using MetaboAnalyst 6.0 software and SIMCA 14 software. To enhance the interpretability and statistical robustness of metabolomic comparisons between plasma and pericardial fluid samples, a mean centering and a pareto scaling of the bucketed obtained data was applied. For the plasma samples, 20 independent biological replicates were collected (10 from individuals with ischemic heart disease (IHD) and 10 from individuals without IHD). Each biological replicate was analyzed in duplicate, resulting in a total of 40 plasma spectra. For pericardial fluid, 10 biological replicates were included (5 per condition (with and without IHD)) and each was also analyzed in duplicate, yielding 20 spectra in total. Replicates were defined as biologically independent samples processed and analyzed under identical conditions. All statistical thresholds (p  <  0.05) are specified where applicable.

Results

Patient characteristics

Patients’ characteristics are summarized in Table 1. No significant differences were detected between IHD and control subjects. Notably, hematological parameters show a trend toward higher values in the IHD group, with red blood cell counts, hemoglobin, and hematocrit all slightly elevated compared to controls. This trend could suggest an adaptive response to chronic ischemic stress, possibly linked to compensatory mechanisms aimed at improving oxygen delivery. Baseline characteristics, including complete blood count and standard biochemical parameters shown in Table 1, were derived from peripheral blood samples obtained from all enrolled patients. Parameters related to metabolic status, including LDL cholesterol, fibrinogen, and glucose, are similar between the two cohorts, indicating that aside from the trends in hematological and injury-related markers, the overall metabolic profiles of patients with IHD and those with mitral insufficiency are largely comparable before surgical intervention. In our analysis of preoperative plasma biomarkers, we observed that total CPK levels were higher in patients with ischemic heart disease compared to those with mitral insufficiency. This elevation likely reflects an ongoing ischemic event, leading to increased release of intracellular enzymes because of myocardial stress and injury. Patients with ischemic heart disease included in this study had stable coronary artery disease and were scheduled for elective cardiac surgery; no acute coronary syndromes were documented prior to enrollment. Medication data reported in Table 1 refer to pharmacological treatments either ongoing at the time of hospital admission or initiated during hospitalization, prior to pericardial fluid collection. The recorded drug classes included ACE inhibitors, angiotensin receptor blockers, potassium-sparing diuretics, statins, SGLT2 inhibitors, chronic use of NSAIDs, corticosteroids, anticoagulants, calcium channel blockers, beta-blockers, diuretics, antiplatelet agents, and anxiolytics.

Table 1 Demographic, anthropometric, laboratory and clinical characteristics of IHD and control patients
Full size table

Multivariate data analysis of human plasma and pericardial fluid

NMR analysis of plasma was performed on 10 diseased IHD patients and 10 non-IHD controls (mitral valve insufficiency), in duplicate. Among these, pericardial fluid analysis was conducted on 5 IHD patients and 5 controls, in duplicate. Representatives 1H CPMG NMR spectra for each type of sample (plasma or pericardial fluid) from IHD patients are shown in Fig. 1. The visual inspection of the spectra evidenced some variability depending on the biofluid under study, plasma or pericardial fluid. Chemical shifts (δ) and assignments of metabolite resonances in the 1H NMR spectra are reported in Table S1. A total of 26 metabolites for plasma and 24 metabolites for pericardial fluid were identified, according to reported literature data and referring to further spectral acquisition(including 1H JRES1,H-1H COSY1,H-13C HSQC, and 1H-13C HMBC), Figs. S1–S8.

The main differences between the two biofluids content is related to the presence of acetate, creatine, phosphocreatine, 1-methylhistidine and glutathione in plasma and of acetoacetate, saccharopine and histidine in pericardial fluid (Fig. 1).

Fig. 1: Representative 1H CPMG NMR spectra of plasma and pericardial fluid samples derived from patients affected of IHD.
Fig. 1: Representative 1H CPMG NMR spectra of plasma and pericardial fluid samples derived from patients affected of IHD.
Full size image

A plasma; (B) pericardial fluid of IHD patients. From left to right: for formate, 1-me-his 1- methyl histidine, his histidine, phe phenylalanine, tyr tyrosine, thr threonine, lac lactate, m-ino myoinositol, tau taurine, cre/creP creatine/creatinephosphate, GSH glutathione, gln glutamine, suc succinate, glu glutamate, NAG = N-acetylglicoprotein, ace acetate, ala alanine, 3-OH-but 3-hydroxybutyrate, BCAA Branched Chain Amino acids (isoleucine, leucine, valine) 2-am-but 2-aminobutyrate.

Overall, the ¹H-NMR spectra of plasma and pericardial fluid of IHD patients are characterized by signals attributed to various metabolite classes, including amino acids (alanine, glutamine, glutamate, glycine, histidine, isoleucine, leucine, methionine, phenylalanine, tyrosine, valine, 1-methylhistidine), sugars (α- and β-glucose), organic acids (2-aminobutyrate, 3-hydroxybutyrate, acetate/acetoacetate, lactate, citrate, succinate), lipoproteins (very-low-density lipoproteins, VLDL, low-density lipoproteins, LDL), osmolytes (myo-inositol, taurine), other metabolites (creatine, phosphocreatine, glutathione).

Unsupervised principal component analysis was employed to provide a general overview of the trends and patterns in the data for both plasma and pericardial fluid. This approach allowed us to detect potential trends, the eventual presence of outliers (thus ensured that no strong outliers biased the model) and the natural groupings of samples without introducing any a priori classification (IHD or CTRL).

In relation to the choice of representation of specific components for the PCA models in Fig. 2 our goal was to better highlight the discrimination between the CTRL and IHD groups in each biofluid which already occurs in the unsupervised models without providing classes information (PCA). Each fluid (plasma and pericardial fluid) was analysed independently and the best components selection (PC2, PC3 for plasma and PC1, PC2 for pericardial fluid) was chosen.

Fig. 2: Principal component analysis, PCA, applied to the data buckets of 1H CPMG NMR spectra obtained from plasma and pericardial fluid samples.
Fig. 2: Principal component analysis, PCA, applied to the data buckets of 1H CPMG NMR spectra obtained from plasma and pericardial fluid samples.
Full size image

A PCA of Plasma, P, samples (n = 20, in technical duplicate) Principal component 2 and 3 (PC2 vs PC3) score plot explain the 29% of the total variance of the dataset. B PCA of Pericardial fluid, PF, samples (n = 10, in technical duplicate). (PC1 vs PC2) score plot explain the 77.1% of the total variance of the dataset. CTRL control group (red), IHD Ischemic heart disease affected patients (green).

The PCA models, shown in Fig. 2, were built using the components that better highlight the discrimination between the CTRL and IHD, thus the second and third principal components for plasma, and the first two components for pericardial fluid. The proportions of variance explained by each principal component, represented as Scree plots, are included in Fig. S9.

The PCA models, Fig. 2A, B, showed clear separation in plasma and pericardial profiles between controls (red) and IHD (green) patients, with minimal overlap observed in pericardial fluid. The PCA models highlighted appreciable differences between the metabolic profile of controls and IHD patients in both the biofluids analysed.

Following this exploratory step, we employed Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) as a supervised method to better discriminate between the metabolic profiles of IHD patients and control subjects. OPLS-DA maximizes class separation by removing variability unrelated to class discrimination, thus improving model interpretability identifying the variables that most contribute to group differentiation.

The supervised OPLS-DA analyses was applied to elucidate the most reliable class-discriminating variables for group separation, for both plasma and pericardial fluid samples (Fig. 3A and C). The differences in the metabolic patterns were studied by the analysis of the corresponding S-line plots (Fig. 3B and D).

Fig. 3: Orthogonal partial least squares discriminant analysis (OPLS-DA) score plots  (left zone) and their corresponding S-line plots(right zone) applied to data buckets of 1H CPMG NMR spectra obtained from plasma and pericardial fluid samples.
Fig. 3: Orthogonal partial least squares discriminant analysis (OPLS-DA) score plots  (left zone) and their corresponding S-line plots(right zone) applied to data buckets of 1H CPMG NMR spectra obtained from plasma and pericardial fluid samples.
Full size image

A, B Plasma (n = 20, in technical duplicate); (C, D) Pericardial fluid (n = 10, in techinical duplicate). Upper right zone of S-line plots (B-D): metabolites that are more abundant in patients with IHD than controls. Lower left zone of S-line plots (B-D): metabolites that are more abundant in controls than in patients with IHD. CTRL control group (red), IHD Ischemic heart disease affected patients (green).

OPLS-DA models (Fig. 3A and C) were obtained with one predictive and two orthogonal components. A notable separation between the two patient’s groups (controls and IHD) was observed, indicating a different metabolic profile among the two classes of patients and confirmed in both plasma and pericardial fluid. The S-line plots analysis (Figs. 3B and D) identified several metabolites, such as 3-hydroxybutyrate, acetoacetate increase, and alanine and tyrosine decrease in patients affected with IHD with respect to the controls. Other metabolites resulted to be altered specifically in the different fluids. In detail, for plasma an increased content of acetone, valine and leucine and a decrease of lactate, and for pericardial fluid an increase of isoleucine and succinate content was also observed in IHD patients.

Among the metabolites highlighted in the S-line plots, the following ROC curve analysis demonstrated excellent discriminatory power between the two groups (control vs. IHD) for 3-hydroxybutyrate and acetoacetate in plasma (Fig. 4A and Figs. S10), and for 3-hydroxybutyrate and succinate in pericardial fluid (Fig. 4B and Fig. S11), confirming these metabolites as possible useful biomarkers for IHD in plasma and pericardial fluid, respectively.

Fig. 4: ROC curve analysis for identified biomarkers, and Box and Whiskers plots of their content distribution in the different conditions (control and IHD).
Fig. 4: ROC curve analysis for identified biomarkers, and Box and Whiskers plots of their content distribution in the different conditions (control and IHD).
Full size image

A Plasma; (n = 20, in technical duplicate); (B) Pericardial fluid; (n = 10, in technical duplicate). CTRL control group (red), IHD Ischemic heart disease affected patients (green). In box and whiskers plots boxes represent interquartile ranges, whiskers indicate data spread, and dots denote outliers.

The further PLS regression models on plasma and pericardial fluid samples, the predictor variables (loadings) were related to response variables of Table 1. The axes (w*c1 and w*c2) represented how strongly each variable is correlated with the first two PLS components. For plasma, it was evidenced a correlation between higher levels of 3-hydroxybutyrate and acetoacetate and higher levels of PLT, Fibrinogen, GOT-Ast, Creatinine, glucose, GPT-Alt in patients affected by IHD, Fig. 5A–C. PLS regression model on pericardial fluid, Fig. 5D–F, evidenced the higher general content of 3-hydroxybutyrate and succinate correlating with the same parameters found for plasma: PLT, Fibrinogen, GOT-Ast, Creatinine, GPT-Alt (with except for glucose) in IHD patients. Anyway, for pericardial fluid, not high predictive parameters were obtained (for the lower number of samples with respect to plasma), thus this aspect need future further evaluation in enlarged cohort of patients.

Fig. 5: Partial Least Square (PLS) regression model for plasma (n = 20, in technical duplicate) and pericardial fluid (n = 10, in technical duplicate) of patients affected from IHD in comparison with non affected (CTRL).
Fig. 5: Partial Least Square (PLS) regression model for plasma (n = 20, in technical duplicate) and pericardial fluid (n = 10, in technical duplicate) of patients affected from IHD in comparison with non affected (CTRL).
Full size image

A–C Plasma (2 components, R2X = 0.417; R2Y = 0.298; Q2 = 0.042. From left to right: Score plot; relative loading plot; expansion of the loading plot with highlighted NMR variables. D–F Pericardial fluid (2 components, R2X = 0.816; R2Y = 0.566; Q2 = 0.203). From left to right: Score plot; relative loading plot; expansion of the loading plot with highlighted NMR variables. Y variables = LDL LowDensity Lipoprotein, WBC White Blood Cell count, RBC Red Blood Cell, Hb Hemoglobin, Hct Hematocrit, PLT Platelets, GotAst Glutamate Oxaloacetate Transaminase-Aspartate Aminotransferase, Gpt-Alt Glutamate Pyruvate Transaminase-Alanine Aminotransferase, CPK Creatine Phosphokinase, CPK-MB Creatine Phosphokinase MB.

In Fig. 6, the VIP scores showed the top 15 metabolites that were highly variated in plasma (Fig. 6A, S12) and in pericardial fluid (Fig. 6B, S13) of IHD patients with respect to controls. Higher VIP score ( > 1) indicates that the metabolite is more important in differentiating between groups. The further two-sample t-tests, also considering the FDR (Table S2 e S3), evidenced ten metabolites (3-hydroxybutyrate, isoleucine, alanine, valine, acetone, acetoacetate, leucine, lactate, phenylalanine, acetate) as significantly changed in the plasma, and three (acetoacetate, succinate and 3-hydroxybutyrate) in the pericardial fluid of IHD patients in comparison with controls, highlighted by the red dotted circle in VIP scores models in Fig. 6A, B.

Fig. 6: Variable importance in projection (VIP) plot displaying the top 15 most important metabolite features identified by OPLS-DA models.
Fig. 6: Variable importance in projection (VIP) plot displaying the top 15 most important metabolite features identified by OPLS-DA models.
Full size image

A Plasma; (B) Pericardial fluid. Colored boxes on the right indicate relative concentration of corresponding metabolite. Red and blue colors mean higher and lower amount of metabolites, respectively, in control and IHD comparisons.

The pathway enrichment and metabolic network analysis of the differentiated metabolites found in plasma and in pericardial fluid was performed and showed in Fig. 7 and Fig. S14. The analyses revealed pathways peculiarly enriched, such as catecholamine biosynthesis, thyroid hormone synthesis, Warburg effect, pyruvate metabolism, gluconeogenesis in plasma, and butyrate metabolism, oxidation and branched chain fatty acids, mitochondrial electron transport chain, phytanic acid peroxisomal oxidation and TCA cycle in pericardial fluid.

Fig. 7: Comprehensive metabolic changes associated with IHD.
Fig. 7: Comprehensive metabolic changes associated with IHD.
Full size image

Top panels: Pathway enrichment analysis of the differentiated metabolites in plasma (A) and pericardial fluid (B). The vertical dotted line represents the cut-off value. Statistical significance was assessed using two-sided test. Bottom panels: Scatter plot representing the overall enrichment of metabolic pathways from metabolomic pathway analysis in plasma (C) and pericardial fluid (D). Metabolite comparisons were conducted using two-sided t-test. To account for multiple testing, p-values were corrected using the False Discovery Rate (FDR) method. Colours, varying from yellow to red, mean that the metabolites are in the data with different levels of significance (yellow: p-value < 0.1; light orange: p-value < 0.05; orange: p-value < 0.01; and red: p-value < 0.001). Circle size, varying from small to big, means the impact of metabolites in data on the pathway alteration (small circle−low impact and large circle−high impact).

The interaction network among the enriched metabolic pathways shown in Fig. S14 revealed that the metabolic pathways enriched in pericardial fluid exhibit more robust interactions compared to those observed in plasma. This finding suggests highly coordinated metabolic alterations within the pericardial fluid microenvironment, as previously reported39. The significantly altered metabolites, as evidenced in VIP scores of Fig. 6, were also considered as entry for the pathway analysis. As shown in Fig. 7C, D, the analysis suggested phenylalanine, tyrosine and tryptophan biosynthesis as pathway significantly altered in IHD, as observed for both biofluids.

Discussion

IHD remains the leading cause of death worldwide3. Diagnosing IHD and its outcomes can be challenging due to several factors, including variability in symptoms, limitations of diagnostic tests, and the presence of comorbidities45. Moreover, many pathophysiological aspects of IHD remain poorly understood. For these reasons, a comprehensive understanding of the disease, as well as the search for new predictive/diagnostic/prognostic tools, is highly desirable.

In this study, we employed a ¹H NMR metabolomic approach to analyse the metabolic profile of patients with IHD in comparison to non-ischemic patients with valvular heart disease. To gain a deeper insight into disease mechanisms, analyses were simultaneously conducted on both plasma and pericardial fluid.

Many studies have assessed the metabolomic profile on plasma of patients with coronary heart disease and found significant associations of metabolite changes with disease incidence, pathogenesis and outcomes28. Pericardial fluid has been less studied in this context39, with some evidence identifying metabolite alterations as markers for risk of post-operative atrial fibrillation in patients undergoing coronary artery bypass grafting surgery. Far from being a static microenvironment, pericardial fluid not only provides homeostatic support to the heart, but also dynamically reflects cardiac injury and regulates its function by stocking bioactive molecules, some of which have been identified including cytokines, chemokines, growth factors, matrix metalloproteases (MMPs), or lipid mediators38,39,40,46.

In our study, multivariate data analysis of both biofluids revealed a distinct metabolic profile between IHD and control patients, indicating perturbed metabolic pathways during ischemia. Specific IHD-associated biomarkers, which may have biological significance in distinguishing between control and IHD conditions, were identified and further confirmed through ROC curve analysis. 3-Hydroxybutyrate emerged as a key biomarker of IHD, as it was significantly altered in both biofluids, i.e. plasma and pericardial fluid. This result suggests that some ischemia-associated metabolite changes in pericardial fluid may be reflected in circulating metabolite profile, which therefore represents a potential and more easily accessible perioperative tool for detecting the metabolic status of the myocardium.

3-Hydroxybutyrate and acetoacetate are ketone bodies predominantly synthetized in the liver from acetyl CoA derived from fatty acid oxidation. They are preferentially produced during periods of starvation or low glucose/high fatty acid availability and are important alternative energy-producing substrates for the heart. A key determinant of their rates of oxidation by the heart is the plasma concentration of acetoacetate and 3-hydroxybutyrate. Once entered the heart, ketone bodies are transferred to the mitochondria to be oxidized, finally producing acetyl-CoA for the TCA cycle and ATP production. Under ischemic conditions, the heart experiences restricted oxygen and glucose oxidation, which disrupts normal aerobic ATP production47. The heart readily switches substrates among fatty acids, glucose, ketone bodies, lactate and other substrates depending on their availability, neurohumoral, environmental, and hemodynamic load. The observed increased plasma levels of 3-hydroxybutyrate and acetoacetate in IHD may indicate a metabolic adaptation in response to cardiac ischemia and impaired glucose oxidation, which force the heart toward ketone body metabolism as alternative energy source48. Consistently, recent studies suggest that 3-hydroxybutyrate can act as a signaling metabolite in cardiovascular disease49,50. However, this had not been previously demonstrated through metabolomic analysis. Pericardial fluid is an ultrafiltrate of plasma and contains proteins, electrolytes, and phospholipids51. In IHD patients, we found a higher content of 3-hydroxybutyrate also in pericardial fluid. Excessive ketone production combined with impaired cardiac metabolism may cause elevated diffusion or leakage of 3-hydroxybutyrate into the pericardial fluid. Consequently, high levels of 3-hydroxybutyrate in the pericardial fluid signal a greater dependence on ketone body metabolism and possibly impaired clearance during ischemic stress52.

Succinate was uniquely altered in pericardial fluid of IHD patients compared with controls. This result is consistent with previous studies that reported cardiac succinate accumulation during ischemia53, but is the first demonstration of ischemic succinate in the pericardial fluid. Excess succinate could escape from cells into the pericardial fluid, reflecting mitochondrial stress and metabolic dysregulation associated with IHD. Succinate, a key intermediate of the TCA cycle, tends to accumulate in mitochondria under ischemic conditions when the electron transport chain slows due to reduced oxygen availability54. However, by using stable isotope labeling in isolated perfused mouse hearts, recent models on the mechanism and origin of ischemic succinate accumulation proposed that succinate primarily derives from the existing TCA cycle metabolite pool and canonical TCA cycle activity, with only a minor contribution to ischemic succinate of mitochondrial complex II reversal (reducing fumarate to succinate) and aminotransferase anaplerosis55 (originating fumarate from aspartate). Succinate accumulation in ischemia has been proposed to boost ischemic energetics55,56 via with substrate-level phosphorylation by succinyl-CoA synthetase, which improves cardiac ischemic energetics. Furthermore, succinate has shown to regulate vascular tone57. However, succinate has been also regarded as a detrimental metabolite in ischemia-reperfusion injury due to its rapid oxidation at the onset of tissue reperfusion, which drives reactive oxygen species generation and necrotic cell death58. Moreover, the accumulated succinate is released from the heart into the circulation at reperfusion59, potentially activating the G-protein-coupled succinate receptor (SUCNR1) signaling and engaging inflammatory signaling60. Elevated levels of both 3-hydroxybutyrate and succinate in pericardial fluid point to significant metabolic adjustments—such as increased ketone body utilization—and mitochondrial dysfunction, characterized by succinate build-up55,61. These biomarkers offer valuable biochemical insights into the local metabolic and inflammatory state of the ischemic myocardium. The observed elevation of ketone bodies, specifically 3-hydroxybutyrate and acetoacetate in plasma, and 3-hydroxybutyrate in pericardial fluid, in patients with ischemic heart disease (IHD) may reflect a metabolic adaptation to chronic myocardial energy stress. Under ischemic conditions, the myocardium downregulates glucose oxidation and shifts toward alternative energy substrates, including free fatty acids and ketone bodies. Ketone bodies have been shown to be more oxygen-efficient compared to fatty acids and may serve as a compensatory fuel in the setting of impaired mitochondrial function.

Moreover, elevated succinate in the pericardial fluid of IHD patients may signal mitochondrial metabolic dysregulation, as succinate accumulation is a known feature of ischemia-related Krebs cycle perturbation. Succinate has also been implicated in redox signaling and inflammation during reperfusion injury. Taken together, the metabolic signatures observed in our study may reflect a combination of enhanced ketone body metabolism and mitochondrial stress in chronically ischemic myocardium. These findings are consistent with recent reports describing a “fuel shift” in failing or ischemic hearts, and suggest that pericardial fluid may capture local metabolic remodeling not evident in plasma alone50,51.

Another finding was the increased levels of the branched chain amino acids (BCAAs), leucine, isoleucine, and valine, in both plasma and, for the first time, pericardial fluid of IHD patients compared with controls. BCAAs are additional sources of acetyl-CoA or succinyl CoA (anaplerosis) for the TCA cycle and ATP production. In accordance with our results, previous studies found higher plasma concentrations of BCAAs (particularly isoleucine) in patients with IHD62,63,64. Elevated plasma BCAAs were highly correlated with insulin resistance in CVD patients65, and predicted the risk of diabetes as well as CVDs66,67,68,69,70. BCAAs are reported to directly activate mammalian target of rapamycin (mTOR) signaling71 and in particular mTOR complex (mTORC), stimulating anabolic processes including protein synthesis and causing insulin resistance and cardiac hypertrophy and dysfunction72. While BCAAs oxidation is thought to be nonessential for cardiac energy production73, BCAA metabolism may play a pivotal role in the regulation of cardiac substrate metabolism. Indeed, a mouse model of impaired (gene knockout) BCAA catabolism showed that accumulation of BCAAs or their intermediate metabolites (branched-chain α-keto acids, BCKAs) in the heart suppressed glucose metabolism by decreasing glucose uptake and mitochondrial pyruvate utilization, a key step of glucose oxidation, and promoted fatty acid oxidation72,74,75. Defective BCAAs catabolism also resulted in mitochondrial dysfunction, oxidative stress, impaired contractile function, and increased the vulnerability of the heart to ischemic injury72. Evidence in both murine and human failing hearts demonstrated a coordinated downregulation of key BCAA catabolic genes as a possible mechanism underlying the marked accumulation of BCAAs and BCKAs in the heart and plasma76. Therefore, the present study and others of the literature suggest that BCAAs metabolism is another integral part of the metabolic remodeling in ischemia and may provide metabolic biomarkers and therapeutic targets for IHD.

We also found that plasma and pericardial fluid displayed decreased levels of aromatic amino acids, such as tyrosine and phenylalanine in IHD with respect to control. The epidemiology of aromatic amino acids in relation to CVDs is complex. Our finding agrees with previous plasma metabolomic studies77, while contrasts with other reports that suggested an association between elevated levels of specific aromatic amino acids and the increased risk of CVDs78,79,80. Through targeted metabolomic profiling of plasma, Markin et al.62 has recently found decreased tyrosine but increased phenylalanine in IHD patients compared with controls. Tyrosine and phenylalanine (through the conversion into tyrosine) are involved in the biosynthesis of neurotransmitters, including catecholamines, and hormones81. Aromatic amino acids are mainly related to hepatic metabolism82 and may represent an index of protein turnover (synthesis and proteolysis)81. During acute ischemia, proteolysis is decreased83 and may result in decreased rate of amino acid release. The observed changes in aromatic amino acids between ischemic and non-ischemic heart disease suggest that alterations in amino acid metabolism may be associated with IHD-related mechanisms. Further studies are warranted to identify the underlying metabolic mechanisms, and the potential causative role of dysregulated aromatic amino acids in IHD.

Notably, plasma levels of lactate, another energy-providing substrate, were higher in controls compared to IHD patients. It was expected to find an increase in lactate production as a result of acute myocardial ischemia, which blunts the mitochondrial oxidative phosphorylation and leads to an increase in anaerobic (glycolytic) metabolism generating pyruvate and then lactate. Moreover, the heart increases lactate uptake and utilization via the TCA cycle and, mostly, lactate glycolytic production84,85, resulting in an increase in circulating lactate. Our discordant result could be interpreted considering the time window of blood collection before surgery. It might be that we collected blood samples in time points before lactate rise occurred. Concordantly, a delay (1–2 h after the ischemia) in the plasma rise of lactate was found in a metabolomics study conducted in a model of balloon-induced ischemia34.

Comprehensive pathway analysis revealed several pathways enriched in both biofluids (plasma and pericardial fluid) including fatty acid biosynthesis, ketone body metabolism, valine, leucine, and isoleucine degradation, as well as tyrosine and phenylalanine metabolism. This finding complies with and expand previous studies assessing plasma metabolic profile in CVDs in relation to IHD and its cardiovascular outcomes62,86,87. The overlapping of enriched metabolic pathways between pericardial fluid and plasma further highlights the highly coordinated metabolic remodeling between plasma and pericardial fluid and suggests that these pathways and the correlated metabolites could hold a pathophysiological importance in IHD. Furthermore, some enriched pathways including alanine, aspartate and glutamate metabolism were more evident in pericardial fluid. These data outline that pericardial fluid could share with plasma the same profile for most metabolites thus reflecting broader pathophysiological changes in ischemia, but some differences may exist in pericardial fluid that could inform on the specific and direct influence of the ischemic heart microenvironment in shaping the metabolomic profile.

Our analysis indicates that pathways related to energy metabolism (e.g., fatty acid metabolism, mitochondrial electron transport chain, TCA cycle) and amino acid metabolism (e.g., branched-chain amino acids, phenylalanine, tyrosine, and tryptophan) were particularly affected. These alterations align with known ischemic heart adaptations, where oxygen deprivation leads to a compensatory increase in glycolysis and a potential reliance on alternative energy substrates88. The heart is widely recognized for its ability to utilize a broad range of energy sources, including fatty acids, glucose, ketone bodies, pyruvate, lactate, amino acids, and even its own structural proteins, in descending order of preference. The energy derived from these substrates not only powers mechanical contractions of the heart, but also supports transmembrane pumps and transporters essential for maintaining ionic balance, electrical signaling, metabolic processes, and catabolic activity88.

Several limitations of the present study need to be considered. We used a small sample size and therefore further analysis using a larger and prospective cohort would validate these associations between differential metabolites and IHD.

Although our findings are mostly supported by previous studies, the underlying molecular mechanisms and causal roles need to be investigated. Nonetheless, to our knowledge this is the first, albeit preliminary, study exploring the metabolomic signature of pericardial fluid in IHD. We identified a previously unknown metabolites profile in pericardial fluid of IHD patients compared to non-IHD subjects, which is to some extent different from that observed in the blood, highlighting that the pericardial fluid has its own unique environment. In line with our observation, a previous study40 found that the inflammatory profile of pericardial fluid was different from that of serum in IHD, and IHD potentiated the profibrotic activity of pericardial fluid compared with non-IHD (valvular disease), suggesting a greater impact of IHD on local microenvironment surrounding the heart. This finding and those of our study would pave the way for leveraging the intraoperative evaluation of pericardial fluid for diagnostic/prognostic and therapeutic purposes and for molecular investigation of IHD pathogenesis. While our findings reveal distinct metabolomic profiles in pericardial fluid associated with IHD, we acknowledge that these results are exploratory in nature. The study was not designed to assess diagnostic or prognostic utility, and as such, no definitive clinical conclusions can be drawn at this stage. Nevertheless, several metabolites identified in our analysis may represent promising candidates for further validation as potential biomarkers.

Future studies in larger, independent cohorts, ideally with longitudinal follow-up and integrated clinical endpoints, will be essential to determine whether the observed metabolic signatures can contribute to risk stratification, patient phenotyping, or therapeutic guidance in cardiovascular disease. At present, these findings should be regarded as hypothesis-generating, laying the groundwork for translational follow-up investigations.

Although our findings describe distinct metabolomic signatures in the pericardial fluid and plasma of patients with IHD, we acknowledge that the present study does not allow for disease staging or early diagnostic assessment, given the comparison between two phenotypic extremes: patients without overt IHD and those undergoing surgery for multivessel disease. Our intention was not to propose these metabolites as immediate diagnostic biomarkers for IHD but rather to explore the local metabolic changes that reflect myocardial stress and dysfunction.

Importantly, we believe that pericardial fluid metabolomic profiling may offer unique insights in clinical scenarios where conventional diagnostics fall short. One such context is microvascular dysfunction, in which patients present with angina or ischemia-related symptoms in the absence of obstructive coronary artery disease. In these patients, conventional biomarkers and imaging often lack sensitivity, and metabolic signatures, especially if accessible via minimally invasive or perioperative collection, may contribute to improved phenotyping or risk stratification.

Therefore, while the translational impact of our findings remains preliminary, they provide a basis for future studies aimed at investigating pericardial fluid metabolites in less well-defined or diagnostically challenging cardiovascular syndromes. These results should be viewed as hypothesis-generating and potentially complementary to existing diagnostic approaches, rather than as standalone tools.

In summary, this metabolomics study tried to dissect the metabolic remodeling of cardiac ischemia interrogating both the blood and pericardial space environments before cardiac surgery. Our results emphasize the intricate nature of metabolic reprogramming in IHD while also highlighting potential novel biomarkers for improved disease management.