Abstract Introduction Prediabetes and preclinical atherosclerosis are interrelated conditions contributing to cardiovascular risk, even in apparently healthy individuals. Metabolomics provides insights into the early metabolic alterations underpinning these diseases. Objectives This study aimed to investigate the shared and distinct metabolic signatures associated with prediabetes and preclinical atherosclerosis in a population with low to moderate cardiovascular risk, using a targeted metabolomic approach. Methods A cross-sectional analysis was performed on 447 participants (mean age 39.7 ± 9.6 years) from the Białystok PLUS cohort. Prediabetes was diagnosed based on HbA1c and OGTT criteria. Preclinical atherosclerosis was assessed by carotid ultrasound. Targeted metabolomics profiling encompassed 434 metabolites and 218 metabolite sums or ratios using HPLC–MS/MS. Statistical analyses included ANOVA, linear regression, correlation analysis, and metabolite set enrichment analysis (MSEA). Results Prediabetes was significantly associated with preclinical atherosclerosis (30.8% vs. 19.5%, p = 0.006). Prediabetes had a broader metabolic impact than atherosclerosis, particularly affecting amino acid and lipid metabolism. Glutamic acid, lactic acid, and L-alanine were strongly associated with prediabetes. Trimethylamine N-oxide (TMAO) was uniquely linked to both prediabetes and its interaction with atherosclerosis, suggesting a context-dependent metabolic response. Glutaminase activity emerged as a robust shared metabolic feature of both conditions. Pathway analyses revealed converging disturbances in glutathione and folate metabolism, mitochondrial function, and redox regulation. Conclusions Prediabetes is associated with more pronounced metabolic alterations than preclinical atherosclerosis. TMAO and glutaminase activity may represent key metabolic links between these conditions. These findings highlight the potential of metabolomics in identifying early biomarkers and mechanisms relevant to the prevention of cardiometabolic diseases. Supplementary Information The online version contains supplementary material available at 10.1186/s12933-025-02841-2. Introduction Atherosclerosis represents a leading pathological basis of cardiovascular disease (CVD), encompassing major clinical events such as myocardial infarction and stroke [[42]1]. Prediabetes has increasingly been acknowledged as a critical determinant in the pathogenesis of CVD, notably atherosclerosis [[43]2]. Contemporary research underscores that prediabetes is not merely a prodromal phase of type 2 diabetes (DM), but rather a distinct and independent risk factor that actively contributes to the initiation and advancement of atherosclerotic processes [[44]3]. Recent investigations have revealed that individuals with prediabetes are at a markedly elevated risk of developing preclinical atherosclerosis. In our cross-sectional study comprising 1,431 participants, prediabetes, identified through HbA1c levels and oral glucose tolerance test (OGTT) criteria, was significantly associated with an increased prevalence of carotid artery plaques and more advanced preclinical atherosclerotic changes, particularly among individuals classified as having low to moderate cardiovascular risk [[45]4]. Likewise, another study employing propensity score matching found that, even after adjusting for age, sex, hypertension, and HDL-C levels, prediabetes independently increased the odds of exhibiting preclinical atherosclerotic plaques by 29%. [[46]5]. These findings highlight the critical importance of early identification of both clinical conditions in order to implement timely preventive strategies. Metabolomics has emerged as a powerful approach for elucidating the intricate pathophysiological mechanisms underlying atherosclerosis and CVD [[47]6]. Numerous epidemiological studies have leveraged this technology to achieve precise and comprehensive assessments of how environmental exposures influence health outcomes [[48]7, [49]8]. Metabolites, defined as small molecules that serve as intermediate or end products of enzymatically catalysed metabolic reactions within cells, reflect the complex interplay between genetic predisposition and environmental influences [[50]9]. Consequently, metabolites offer a valuable framework for dissecting gene-environment interactions and advancing our insights into the pathogenesis of multifactorial conditions. Among the metabolites of emerging interest, trimethylamine N-oxide (TMAO), a compound generated by hepatic oxidation of gut microbiota-derived trimethylamine, has been identified as a key contributor to atherogenesis [[51]10]. Elevated circulating concentrations of TMAO have been consistently associated with increased vascular inflammation, endothelial dysfunction, and foam cell formation, all of which represent hallmarks of early atherosclerotic changes [[52]10]. Mechanistically, TMAO has been shown to upregulate macrophage scavenger receptors, inhibit reverse cholesterol transport and bile acid synthesis, and induce pro-inflammatory cytokine production, thereby promoting lipid accumulation and vascular injury [[53]11]. Notably, recent studies suggest that individuals with prediabetes display distinct alterations in gut microbial composition and host metabolic profiles [[54]12], including elevated TMAO levels, which may in part mediate their increased susceptibility to atherosclerotic cardiovascular disease. The objective of the present study was to evaluate the association between preclinical atherosclerosis and prediabetes in apparently healthy population using a comprehensive metabolomic approach allowing to identify of potential metabolic mediators, such as TMAO, that may explain the link between prediabetes and the development of early atherosclerotic changes. Material and methods Study population The analysis encompassed 447 participants (mean age 39.7 ± 9.6 years; 44.7% males) enrolled in the Bialystok PLUS study. The Białystok PLUS Study is a population-based cohort designed to investigate the determinants of chronic non-communicable diseases in adults residing in Białystok, Poland. Additional details regarding the study design and methodology have been described in previous publications [[55]13]. Participants included in the cross-sectional analysis met the following inclusion criteria: (1) classification within low to moderate cardiovascular risk (CVR) categories; (2) no current or prior treatment with cholesterol-lowering agents, including statins; (3) no history of diabetes mellitus or diagnosis during investigation; (4) no use of antidiabetic medications or pharmacotherapies approved for obesity management; and (5) no chronic use of non-steroidal anti-inflammatory drugs (NSAIDs) or systemic glucocorticoids. Data collection and assays All the clinical and biochemical measurements were conducted by qualified medical personnel. High reproducibility of the examinations was achieved by performing them based on verified Standard Operating Procedures. The details of the subjects’ medical history were collected from questionnaires at the time of study entry. In the morning of the visit, peripheral intravenous fasting blood samples were collected after at least eight hours of fasting. Comprehensive biochemical assessments were performed: total cholesterol (TC), LDL-cholesterol (LDL-C) and HDL-cholesterol (HDL-C) fractions, triglycerides (TG), glucose, glycated haemoglobin (HbA1c). The samples were then prepared for further analysis by centrifugation and storage at − 70 °C. As recommended by the World Health Organization (WHO), we performed an OGTT on fasting patients who did not report having a history of DM. 75 g of glucose dissolved in water was administered orally and then blood was taken sequentially after 1 and 2 h. Fasting glucose concentration and at 60 min and 120 min glucose levels were assessed in plasma drawn on EDTA with sodium fluoride using a reference enzymatic method with hexokinase (Cobas c111, Roche). Using the homogeneous enzymatic colorimetric method on the Cobas c111 from ROCHE (ROCHE, Meylan, Isère, France), the concentration of LDL-C and HDL-C was determined. TC and TG were determined using the enzymatic colorimetric method on the Cobas c111 from ROCHE. HbA1c was determined by ion-exchange high-performance liquid chromatography (HPLC) on D-10 from Bio-Rad (Bio-Rad, Hercules, CA, USA). Serum insulin and peptide C were determined by electrochemiluminescence (ECLIA) on the Cobas e411 device (ROCHE). Anthropometric measurements, including height and circumferences of the waist were taken (SECA 201 tape, Hamburg, Germany). Body mass index (BMI) was calculated as weight in kilograms divided by height in metres squared and is expressed in units of kg/m^2. Diagnosis of prediabetes Individuals with a history of DM or current use of antidiabetic medications were excluded from the analysis. Participants meeting the diagnostic criteria for DM based on the study measurements were also not included. Prediabetes was defined as an HbA1c level ≥ 5.7% and < 6.5%. Impaired fasting glucose (IFG) was diagnosed in participants with fasting plasma glucose levels ≥ 100 mg/dL but < 125 mg/dL. Impaired glucose tolerance (IGT) was identified in individuals with 2-h post-load glucose values ≥ 140 mg/dL but < 200 mg/dL during an OGTT. Due to the availability of only a single fasting glucose measurement and the lack of a confirmatory time point, participants with fasting glucose levels ≥ 125 mg/dL were classified as having IGT. Altogether, 201 (mean age 42.4 ± 9.3 years; 48.8% males) individuals fulfilled the diagnostic criteria for prediabetes and were categorized accordingly. Preclinical atherosclerosis assessment Ultrasound examination of the carotid arteries was used to evaluate early atherosclerotic lesions. The ultrasonography measurements were made using the ultrasound Vivid 9 (GE Healthcare, Chicago, IL, USA). The presence of any atherosclerotic plaques in (1) right common carotid artery (CCA), (2) left CCA, (3) right external carotid artery (ECA), (4) left ECA, (5) right internal carotid artery (ICA), (6) left ICA, (7) right bifurcation (BIF), (8) left BIF were evaluated. We assessed atherosclerotic plaques as binomial quality variables and marked them as present when 1 of the following criteria was fulfilled: (1) the local thickening IMT towards the lumen of the vessel, exceeding the surrounding IMT by > 0.5 mm, (2) the local thickening IMT towards the lumen of the vessel, surpassing the surrounding IMT by 50%, (3) IMT thickening > 1.5 mm [[56]14]. A constant element of the study was the estimation of intima media thickness (IMT) in right and left CCA and the result is presented as an average value from 5 measurements. Significant value to assess the presence of preclinical atherosclerotic lesion was adopted IMT > 0.9 mm. Metabolic assessment Metabolomic analyses were performed on plasma derived from blood collected in EDTA tubes. For this purpose, blood was first centrifuged for 10 min at 400 RCF. The resulting supernatant was transferred to a new tube and subjected to a second centrifugation for 10 min at 1800 RCF to remove residual cellular debris. The final plasma samples were immediately frozen and stored at − 80 °C until analysis. The samples collected from all study participants were subjected to targeted metabolomics using the high-performance liquid chromatography-tandem mass spectrometry ((HPLC–MS/MS) technique. In total, 630 metabolites were measured using an MxP Quant 500 kit (Biocrates Life Sciences AG, Innsbruck, Austria). The manufacturer's protocol and methodology were used for sample preparation and measurements. Analyses were performed on an HPLC–MS/MS system composed of a Shimadzu Nexera LC-40 HPLC coupled with a Sciex QTRAP 6500+ mass spectrometer. The raw spectral data processing, quantification, and normalization were performed using WebIDQ software (Biocrates Life Science AG, Innsbruck, Austria). Data normalization was followed as described in the Biocrates kit user's manual [[57]15]. In brief, intra-plate data normalization was applied based on the target values of the quality control (QC) sample level 2. For this purpose, correction factors were calculated by dividing the median QC2 metabolite concentration by the target concentration (specified by the manufacturer) for each kit. Metabolite concentrations were normalized by dividing each value by the calculated correction factor for each metabolite from all kits. Inter-plate normalization was performed by calculating the correction factor for each kit. For this purpose, the median metabolite concentrations for a single set were divided by the median metabolite concentrations from all sets, separately for each metabolite. Finally, metabolite concentrations were normalized by dividing each concentration value by the calculated correction factor. The resulting data were then filtered to reject metabolites that were below the limit of detection (LOD) in more than 20% of the samples. Data cleaning also included the exclusion of metabolites with low reproducibility, assessed based on the coefficient of variation (CV) calculated from triplicates of the quality control (QC) samples for each metabolite in each batch, with a rejection criterion of a CV value greater than 30%. The missing values in the remaining metabolites resulting from the value estimated below the LOD were replaced by ½ of the LOD for each metabolite in each batch [[58]15]. In addition, sums and ratios of metabolite concentrations were calculated using MetaboINDICATOR™ (Biocrates Life Sciences AG, Innsbruck, Austria) [[59]16]. Formulas were calculated based on the previously obtained metabolites. In total, 434 metabolites and 218 derived ratios and sums were available for analyses. Although 434 metabolites passed initial filtering and quality control, only 120 metabolites could be confidently mapped to identifiers compatible with MetaboAnalyst, allowing their use in metabolite set enrichment analysis (MSEA). To ensure consistency and comparability, all statistical analyses, including ANOVA, linear models, and correlation analyses, were likewise restricted to these 120 metabolites. Statistical analysis Prior to statistical analysis, raw metabolite abundance data were normalized to account for skewness and scale differences. Specifically, each metabolite was log-transformed (natural logarithm) and subsequently standardized using z-score scaling, i.e., mean-centered and divided by the standard deviation across samples. All samples were analysed within a single analytical batch, and therefore no batch correction was applied (see Supplementary Fig. 1 for normalization details). All statistical analyses and visualizations were performed in R (version 4.4.2) using packages such as tidyverse, broom, heatmap, limma, and ComplexHeatmap. To assess group differences and interactions, we used linear models including prediabetes status, atherosclerosis status, and their interaction, with age and sex and other clinical variables as covariates. To account for unequal group sizes and the presence of interaction terms, we applied Type III analysis of variance (ANOVA) using the Anova() function from the R package car, with contrast coding set to sum-to-zero. This approach allows for accurate testing of main effects and interactions in unbalanced designs. Clinical variables were compared across groups using one-way ANOVA, with Tukey's post hoc test applied where appropriate. Summary statistics (mean, SD, quartiles) were reported for each group. For metabolomic data (120 metabolites and 218 derived ratios/sums), both two-way ANOVA models (testing main effects and interaction between preclinical atherosclerosis and prediabetes) and multiple linear models (including age and sex as covariates) were used. All models were run independently for each metabolite. All p-values for effects of interest (atherosclerosis, prediabetes, interaction) were adjusted using the Benjamini–Hochberg false discovery rate (FDR) procedure. Metabolites with at least one significant FDR-adjusted p-value (< 0.05) in ANOVA models were retained for downstream analysis. Spearman rank correlation coefficients were calculated between serum metabolite concentrations and clinical variables, including lipid profile, glycemic indices, and vascular parameters. Categorical variables (e.g., sex, disease status) were not included in this analysis. P-values were adjusted using the Benjamini–Hochberg method to control the false discovery rate. Only correlations with FDR-adjusted p < 0.05 and an absolute rho (|ρ|) of at least 0.1 were visualized in annotated heatmaps. Non-significant associations were excluded from the figure for clarity. Metabolite set enrichment analysis was performed separately for metabolites associated with atherosclerosis and prediabetes using the MetaboAnalyst web platform ([60]https://www.metaboanalyst.ca). Analyses were based on SMPDB and KEGG pathway libraries. Enrichment results were visualized in R using log-scaled bar plots with coloring based on adjusted p-values. The STROBE guidelines for observational studies were applied in this study. Results The link between preclinical atherosclerosis and prediabetes in study sample Among individuals’ prediabetes-free, 19.5% were found to have preclinical atherosclerosis, compared to 30.8% among those with prediabetes. The difference between the groups was statistically significant (p = 0.006). The fundamental clinical characteristics of the study cohort (n = 447, 44.7% males) are presented in Table [61]1. The selected population sample had a mean age of 39.7 ± 9.6 years. The median BMI was 24.77 kg/m^2, indicating that most individuals were within the normal weight range. On average, both the lipid profile and glucose metabolism parameters were within normal ranges. Additionally, Supplementary Table 1 presents the characteristics of the selected cohort stratified by the presence or absence of prediabetes and preclinical atherosclerosis. Group 1 includes individuals with neither condition (n = 48), group 2 comprises participants with prediabetes only (n = 62), group 3 includes those with preclinical atherosclerosis only (n = 198) and group 4 consists of individuals presenting with both prediabetes and subclinical atherosclerosis (n = 139). This classification enabled the assessment of metabolic differences associated with each condition independently, as well as their potential synergistic effects. Table 1. Baseline clinical characteristic of study group (n = 447) Variable Descriptive statistics Age, years 39.7 ± 9.6 Sex, n (%) males 200 (44.7%) Body Mass Index, kg/m2 24.77 22.2 27.4 Total cholesterol, mg/dl 185 166 211 LDL-C, mg/dl 118.5 96.9 140.4 HDL-C, mg/dl 62.95 51.12 73.62 TG, mg/dl 83 61 115 HbA1c, % 5.2 5 5.5 Fasting glucose level, mg/dl 96 91 102 Serum glucose level 2h post-loading, mg/dl 112 99 130 Fasting C-peptide level, nmol/l 2.06 1.7 2.64 C-peptide level after 2 h, nmol/l 7.36 5.8 9.34 Fasting insulin level, mU/l 9.62 6.94 13.04 Serum insulin level after 2 h post loading, mU/l 43.54 26.68 63.77 [62]Open in a new tab The table presents medians and interquartile ranges (25th–75th percentile), except for age, which is shown as mean and standard deviation, and for sex, which is presented as percentages The assessment of effects of preclinical atherosclerosis and prediabetes on metabolites Figure [63]1 and Supplementary Table 2 present the results of a two-way ANOVA evaluating the associations between individual metabolite concentrations and two clinical conditions: atherosclerosis and prediabetes, including their interaction effect. After Benjamini–Hochberg false discovery rate (FDR) adjustment, a total of 20 metabolites were found to be significantly associated exclusively with prediabetes, 4 exclusively with preclinical atherosclerosis, and another 4 were significantly associated with both conditions, suggesting overlapping metabolic alterations. Only 1 metabolite, TMAO, showed a significant association with the interaction term. Among the metabolites with the strongest associations, Glu, CE(20:5) (cholesteryl ester of eicosapentaenoic acid), CE(20:3), and CE(18:3) were significantly associated with both atherosclerosis and prediabetes. Dehydroepiandrosterone sulfate (DHEAS), indole-3-propionic acid (IPA), and CE(16:1) were primarily associated with preclinical atherosclerosis. Lac, L-Ala, docosahexaenoic acid (DHA), eicosapentaenoic acid (EPA), and dihomo-gamma-linolenic acid (DGLA) were among the most strongly associated with prediabetes. Fig. 1. [64]Fig. 1 [65]Open in a new tab Venn diagram illustrating the number of metabolites significantly associated with atherosclerosis status, prediabetes status as identified by two-way ANOVA. Trimethylamine N-oxide (TMAO) was not directly correlated with prediabetes and preclinical atherosclerosis, but was associated with their interaction. A summary of the results of linear regression models assessing the independent and interaction effects of preclinical atherosclerosis, prediabetes on individual metabolite concentrations, adjusted for age and sex (including only those with statistically significant associations), is presented in Table [66]2. After applying false discovery rate (FDR) correction for multiple comparisons, no metabolites demonstrated statistically significant associations with preclinical atherosclerosis alone. In contrast, seven metabolites showed significant associations with prediabetes, including L-Ala, Glu, and Lac, which exhibited the strongest associations with dysglycemia. Additional metabolites, such as DHEAS, DGLA, and EPA, also remained significant following adjustment for sex and age. Supplementary Table 3 presents additional linear regression models adjusted not only for age and sex but also for established cardiovascular risk factors, including BMI, current smoking status, fasting glucose concentration, HOMA-IR, as well as LDL-C and HDL-C levels. Notably, TMAO remained a statistically significant predictor even after adjustment for these covariates, highlighting its potential role as a key metabolic mediator in the interplay between preclinical atherosclerosis and prediabetes. Table 2. The effects of atherosclerosis and prediabetes and their interaction on metabolite levels, adjusted for age and sex—summary of linear regression models Metabolite Preclinical atherosclerosis Prediabetes Their interaction raw p FDR adjusted p Partial η^2 raw p FDR adjusted p Partial η^2 raw p FDR adjusted p Partial η^2 Lactic acid 0.199 0.697 0.001 0.000 0.001 0.037 0.002 0.118 0.022 Glutamic acid 0.004 0.231 0.041 0.000 0.001 0.064 0.076 0.601 0.007 L-Alanine 0.357 0.823 0.001 0.000 0.003 0.041 0.123 0.738 0.005 Dihomo-gamma-linolenic acid 0.350 0.823 0.011 0.001 0.036 0.035 0.864 0.948 0.000 Eicosadienoic acid 0.752 0.927 0.005 0.002 0.036 0.034 0.821 0.938 0.000 Dehydroepiandrosterone sulfate 0.116 0.697 0.056 0.002 0.042 0.001 0.180 0.806 0.004 Trimethylamine N-oxide 0.012 0.231 0.003 0.002 0.042 0.010 0.000 0.014 0.033 [67]Open in a new tab Bold indicates the significant p value Linear regression models were fitted separately for each metabolite (n = 400) with preclinical atherosclerosis status, prediabetes status and their interaction as independent variables, and age and sex as covariates. Raw and false discovery rate (FDR)-adjusted p-values are reported for the three coefficients of interest. The table includes only metabolites for which at least one of the coefficients showed a statistically significant FDR-adjusted p-values (< 0.05) Notably, TMAO was the only metabolite demonstrating statistically significant association with both prediabetes and the interaction between atherosclerosis and prediabetes, suggesting a combined, non-additive effect. This interaction was visualized on a violin plot (Fig. [68]2), illustrating the differing effects of prediabetes across atherosclerosis strata. Post hoc stratified analysis (adjusted for age and sex) revealed that in individuals without atherosclerosis, TMAO levels were significantly higher in those with prediabetes compared to prediabetes-free participants (β = 1.11, p = 0.0235). In contrast, among individuals with atherosclerosis, TMAO levels did not differ significantly by prediabetes status (β =–2.69, p = 0.18). These findings imply a potential context-dependent effect of prediabetes on TMAO, modulated by the presence of preclinical atherosclerosis. Fig. 2. [69]Fig. 2 [70]Open in a new tab Trimethylamine N-oxide levels in relation to prediabetes and preclinical atherosclerosis The evaluation of the prediabetes and preclinical atherosclerosis effect on the modulation of metabolite profiles The results of the two-way ANOVA assessing the independent and interaction effects of prediabetes and preclinical atherosclerosis on metabolite sums and ratios are presented in Fig. [71]3 and Supplementary Table 4. Significant main effects of atherosclerosis were predominantly observed in lipid-related parameters, including the ratio of lysophosphatidylcholines (LPCs) containing polyunsaturated fatty acids (PUFAs) to those containing saturated fatty acids (SFAs), the sum of steroid hormones, and selected classes of cholesteryl esters—specifically those containing monounsaturated (MUFA-CEs) and long-chain fatty acids (LCFA-CEs). Fig. 3. [72]Fig. 3 [73]Open in a new tab Venn diagram of associations between metabolite sums and ratios with preclinical atherosclerosis, prediabetes, based on ANOVA results In contrast, the main effects of prediabetes showed a broader metabolic footprint, with significant associations related to AA metabolism and lipid signalling. These included elevated sums of aromatic and sulfur-containing amino acids (AAs), along with increased activities of phospholipase A2 (PLA2) and indoleamine 2,3-dioxygenase (IDO). Additionally, higher levels of very-long-chain fatty acid dihydroceramides (VLCFA-DH-Cer) and the overall pool of dihydroceramides (DH-Cer) were observed. Perturbations in the profiles of unsaturated phosphatidylcholines (UFA-PCs) and ether-linked phosphatidylcholine species (PC (O)s), including both unsaturated (UFA-PC (O)) and saturated (SFA-PC (O)) variants, were also detected. Furthermore, the ratio of saturated galactosylceramides (SG) to hexoses and concentrations of measured omega-3 fatty acids (ω-3 FAs) were significantly elevated in individuals with prediabetes. The sum of carboxylic acids and the calculated glutaminolysis rate were also significantly associated with dysglycemia-related metabolic disruptions. Several metabolic features were found to be dysregulated in both prediabetes and preclinical atherosclerosis. Glutaminase activity, assessed as Glu/Gln ratio [[74]17] and glutathione (GSH) pathway were significantly associated with both conditions. Furthermore, multiple PLA2 activities, specifically PLA2 Activity (3) and PLA2 Activity (5), indicate shared activation of phospholipase A2-mediated pathways. Other shared features included the sum of PUFA-containing phosphatidylcholines (PUFA-PCs), ether-linked PUFA-PC (O)s, and very-long-chain fatty acid cholesteryl esters (VLCFA-CEs). Table [75]3 provides a summary of the linear regression analyses, which evaluated associations between selected metabolic features (sums and ratios) and three factors: prediabetes, preclinical atherosclerosis, and their interaction, with adjustments for age and sex. Glutaminase activity emerged as the only shared predictor of both prediabetes and atherosclerosis, highlighting a potentially convergent glutamine-related metabolic mechanism. In contrast, the glutaminolysis rate, the sum of carboxylic acids, and the ratio of SG to hexose were uniquely associated with prediabetes. No distinct metabolic predictors were identified exclusively for atherosclerosis or for the interaction term. Supplementary Table 5 provides a summary of linear regression models in which CVR factors were reintroduced as covariates. Following adjustment for BMI and current smoking status, glutaminase activity continued to demonstrate a statistically significant association, indicating its potential independent contribution to the observed metabolic alterations. Table 3. The effects of atherosclerosis and prediabetes and their interaction on metabolite sums and ratios, adjusted for age and sex—summary of linear regression models Metabolite Ratios and Sums Preclinical atherosclerosis Prediabetes Their interaction raw p FDR adjusted p Partial η^2 raw p FDR adjusted p Partial η^2 raw p FDR adjusted p Partial η^2 Glutaminolysis Rate 0.064 0.424 0.000 0.000 0.000 0.036 0.001 0.157 0.023 Glutaminase Activity 0.000 0.025 0.040 0.000 0.000 0.054 0.007 0.229 0.016 Ratio of SG to Hexose 0.563 0.776 0.003 0.000 0.000 0.079 0.794 0.990 0.000 Sum of Carboxylic Acids 0.248 0.628 0.001 0.000 0.002 0.031 0.005 0.190 0.018 [76]Open in a new tab Bold indicates the significant p value The correlations between metabolites and clinical parameters Figure [77]4 presents the Spearman correlation matrix between selected metabolite levels and clinical parameters, including lipid profile components, glucose metabolism indicators, and preclinical atherosclerosis parameters. Only statistically significant correlations (FDR-adjusted p < 0.05 and|ρ|≥ 0.1) are shown. Among the most prominent observations, cholesteryl ester (CE) (20:4) exhibited moderate positive correlations with multiple glycaemic and insulinemic parameters, including fasting glucose, fasting insulin, and baseline C-peptide levels. Fatty acids also showed a positive correlation with parameters of glucose metabolism, especially postprandial. Glutamic acid (Glu) was consistently positively associated with both glucose and insulin concentrations across all assessed time points, suggesting a potential role in glucose homeostasis. In contrast, serine demonstrated inverse correlations with several metabolic markers, most notably with 120-min post-load glucose and C-peptide levels, as well as TG. Among sphingolipids, SM C16:1 showed positive associations with TC and its fractions and IMT, suggesting a possible role in early atherosclerotic processes. HDL-C fraction, in contrast to other lipemic parameters, negatively correlated with a broad spectrum of metabolites, including glutamic acid, aminoadipic acid, lactic acid (Lac), L-alanine (L-Ala), and L-cysteine (L-Cys). Fig. 4. [78]Fig. 4 [79]Open in a new tab Spearman correlation heatmap between serum metabolites and cardiometabolic biomarkers. Each cell represents the Spearman correlation coefficient between a metabolite and a clinical variable. The color scale ranges from red (positive correlation) to blue (negative correlation), with white indicating values near zero or associations that did not reach statistical significance. Only statistically significant correlations (FDR-adjusted p < 0.05) with absolute correlation coefficients ≥ 0.1 are shown. Exact Spearman rho values are indicated in each cell Metabolite set enrichment analysis Figure [80]5 and Supplementary Table 6 present the results of a Kyoto Encyclopedia of Genes and Genomes (KEGG) [[81]18] metabolic pathway enrichment analysis (MSEA) conducted to identify metabolic pathways significantly associated with preclinical atherosclerosis. The analysis was performed based on differentially abundant metabolites and includes multiple testing correction using the false discovery rate (FDR) approach. Pathways are ranked according to their enrichment ratio (logarithmic scale) and color-coded by adjusted p-values. The most significantly enriched pathways (FDR < 0.001) include glutathione metabolism, steroid biosynthesis, thiamine metabolism, butanoate metabolism, and nitrogen metabolism. This KEGG-based pathway enrichment analysis highlights a distinct metabolic signature associated with preclinical atherosclerosis, implicating redox balance (glutathione), lipid synthesis (steroid biosynthesis), and energy-related pathways (thiamine and butanoate metabolism). These findings suggest that early stages of atherosclerosis are linked to oxidative stress responses, disturbances in lipid biosynthesis, and alterations in energy metabolism. Additional significantly enriched pathways include taurine and hypotaurine metabolism, histidine metabolism, arginine and proline metabolism, alanine, aspartate, and glutamate metabolism, and cysteine and methionine metabolism, indicating broader disruptions in amino acid and nitrogen metabolism. Collectively, these results highlight the involvement of redox regulation, amino acid utilization, and bioenergetic pathways in the metabolic signature of preclinical atherosclerosis. Fig. 5. [82]Fig. 5 [83]Open in a new tab KEGG Pathway Enrichment Analysis for preclinical atherosclerosis Simultaneously, Fig. [84]6 and Supplementary Table 7 display results from a KEGG MSEA investigating metabolic alterations associated with prediabetes, based on significantly altered metabolites. A number of pathways show highly significant enrichment (adjusted p < 0.001), including biosynthesis of unsaturated fatty acids, alanine, aspartate, and glutamate metabolism, histidine metabolism, butanoate metabolism, as well as nitrogen metabolism. Additional enriched pathways include glutathione and porphyrin metabolism, both of which are involved in oxidative stress regulation. These results highlight a broad metabolic disruption in prediabetes, involving not only energy metabolism but also amino acid turnover and nitrogen balance. Fig. 6. [85]Fig. 6 [86]Open in a new tab KEGG Pathway Enrichment Analysis for prediabetes MSEA using the Small Molecule Pathway Database (SMPDB) [[87]19] also revealed distinct yet overlapping profiles in individuals with atherosclerosis and prediabetes. Figure [88]7 and Supplementary Table 8 showed that in preclinical atherosclerosis, the most significantly enriched pathways included cysteine metabolism, glutathione metabolism, androgen and estrogen metabolism, folate metabolism, and pathways related to AA metabolism such as pantothenate and CoA biosynthesis, glutamate metabolism, and amino sugar metabolism. Lipid-related pathways, such as arachidonic acid and alpha-linolenic acid metabolism, were also significantly represented. In contrast, prediabetes was marked by robust enrichment of the glucose–alanine cycle, glutathione metabolism, alanine metabolism, arachidonic acid metabolism, and glutamate metabolism, along with key alterations in the Warburg effect, tyrosine metabolism, and folate metabolism (Fig. [89]8 and Supplementary Table 9). Both conditions demonstrated significant enrichment of several shared pathways, including glutathione metabolism, glutamate metabolism, folate metabolism, the Warburg effect, glucose-alanine metabolism, malate-aspartate shuttle, and arachidonic acid metabolism, indicating common metabolic perturbations associated with oxidative stress, inflammation, and mitochondrial dysfunction. In particular, the involvement of tryptophan and glycerophospholipid metabolism highlights the importance of immune-metabolic crosstalk and membrane lipid remodeling in the pathogenesis of these interconnected conditions. Fig. 7. [90]Fig. 7 [91]Open in a new tab SMPDB Pathway Enrichment Analysis for preclinical atherosclerosis Fig. 8. [92]Fig. 8 [93]Open in a new tab SMPDB Pathway Enrichment Analysis for prediabetes Discussion This cross-sectional analysis of metabolomic profiles within the Białystok PLUS cohort highlights distinct and overlapping metabolic signatures associated with prediabetes and preclinical atherosclerosis in a population at low to moderate CVR. By limiting our study population to individuals free from comorbidities and chronic medication use, we aim to capture the initial pathophysiological developments of atherosclerosis in a less confounded clinical context. We demonstrated that several investigated metabolites were independently associated with both preclinical atherosclerosis and prediabetes, even after adjustment for age and sex in multivariable models. Prediabetes was found to be associated with more extensive and pronounced alterations across multiple metabolic domains, including amino acid metabolism, lipid signalling, and enzymatic activities, compared to preclinical atherosclerosis. Notably, TMAO was the only metabolite demonstrating statistically significant association with both prediabetes and the interaction between atherosclerosis and prediabetes, suggesting a context-dependent metabolic modulation. Among the many dysregulated metabolic features observed in both prediabetes and preclinical atherosclerosis, glutaminase activity uniquely emerged as the only robust, shared predictor after adjustment for age and sex. Although prediabetes and preclinical atherosclerosis exhibited distinct patterns in lipid and amino acid metabolism, enrichment analyses revealed converging disruptions in pathways related to glutathione and folate metabolism, as well as mitochondrial function. These shared alterations support the notion that oxidative stress and inflammation constitute common early mechanisms underlying cardiometabolic disease progression. Collectively, these findings underscore the utility of metabolomic profiling in elucidating shared mechanisms in early cardiometabolic risk conditions. Our results suggest that atherosclerosis may exert a modulatory influence on the association between prediabetes and TMAO levels, potentially indicating a context-dependent metabolic interaction. TMAO is produced by gut microbiota, and its levels are influenced by dietary choices [[94]11]. It triggers excessive mitophagy in endothelial cells, facilitating pyroptosis, which contributes to the initiation and progression of atherosclerosis [[95]20]. Other authors showed that dietary TMAO supplementation accelerates atherosclerosis in ApoE−/−mice [[96]21]. Moreover, TMAO levels were significantly higher in middle-aged patients with atherosclerotic cardiovascular disease compared to those with and without atherosclerosis risk factors [[97]22]. In another paper, TMAO concentrations were shown to be positively associated with atherosclerosis severity in patients with acute coronary syndrome [[98]23]. Moreover, earlier studies have demonstrated that plasma TMAO was associated with an increased prevalence of prediabetes, but not with insulin resistance or IFG longitudinally [[99]12]. Another study investigated the relationship between TMAO levels and incident DM in middle-aged and older adults, finding that higher serum TMAO concentrations were linked to an elevated risk of DM and increased fasting glucose levels [[100]24]. Furthermore, elevated TMAO levels have been significantly associated with long-term major adverse events in patients with CAD and concurrent DM [[101]25]. This association was also confirmed in a Chinese population, where TMAO levels were strongly correlated with DM among patients with CAD [[102]26]. Our analysis is distinguished by showing the interaction between TMAO metabolite levels and preclinical conditions of atherosclerosis and prediabetes. The observed interaction effect for TMAO indicates a biologically relevant context dependency, whereby prediabetes is associated with elevated TMAO levels only in individuals without subclinical atherosclerosis. This pattern was consistent in both the interaction model and the post hoc stratified analysis. One possible interpretation is that the metabolic impact of prediabetes on TMAO is more evident in the earlier stages of vascular change and may be masked or reversed in the presence of atherosclerosis-related pathophysiology. These findings highlight the importance of considering combined risk profiles in metabolite-based biomarker studies, as overlooking such interactions may obscure clinically relevant metabolic signatures. In our results, glutaminase activity emerged as the main shared predictor of both prediabetes and atherosclerosis, highlighting a potentially convergent glutamine-related metabolic mechanism. Glutaminase, an enzyme responsible for the conversion of glutamine to glutamate, exists in two primary isoforms: the liver-type and the kidney-type [[103]27]. The expression of hepatic glutaminase is upregulated during conditions such as starvation, DM, and consumption of a high-protein diet, whereas the kidney-type glutaminase shows increased activity exclusively in the kidney in response to metabolic acidosis [[104]28]. Recently, it was demonstrated that hepatic glutaminolysis contributes to the advancement of atherosclerosis by modulating the plasma glutamine-to-glutamate ratio, as well as other parameters, including triglyceride concentrations and the presence of monocytopenia [[105]29]. A prior study showed that the plasma glutamine-glutamate ratio is an independent risk factor for carotid plaque progression [[106]30]. Moreover, other researchers indicate that decreased glutaminase activity in adipocytes is linked to obesity-associated IR, which is a precursor to prediabetes [[107]31]. Furthermore, it was shown that reduced hepatic glutaminase levels lead to lower glucagon-stimulated glucose production, suggesting that glutaminase activity may play a significant role in glucose homeostasis and could be relevant in prediabetes [[108]32]. Collectively, our results identify glutaminase activity as a critical metabolic intersection linking prediabetes and atherosclerosis, and point to glutamine metabolism as a promising area for future research aimed at developing novel strategies for the prevention and management of diverse cardiometabolic disorders. The cross-sectional design limits causal inference between metabolic alterations, prediabetes, and preclinical atherosclerosis. Although the sample was well-characterized, its modest size may reduce power to detect subtle effects. Moreover, the exclusion of individuals with high CVR improves internal validity but may limit generalizability to broader populations. While the group sizes were not perfectly equal, the degree of imbalance between them was moderate. To address this, we used linear models with covariates and applied Type III ANOVA, which is specifically suited to unbalanced factorial designs with interactions. A statistically significant interaction was identified for one metabolite, suggesting a robust effect. However, we acknowledge that the relatively small size of one subgroup may have reduced the statistical power to detect weaker or more subtle interaction effects. Non-significant interactions should therefore be interpreted with caution. Despite these limitations, the study is based on a population-based cohort with strict inclusion criteria, minimizing confounding by comorbidities or pharmacotherapy. The integration of comprehensive metabolomic profiling with preclinical phenotypes allows for detailed exploration of metabolic perturbations and their potential pathophysiological relevance. Furthermore, the combined use of correlation analysis, multivariate modelling, and pathway enrichment strengthens the robustness and interpretability of the results. Together, these features provide valuable insights into the early metabolic signatures associated with cardiometabolic risk conditions. Conclusions Our cross-sectional analysis showed that prediabetes was associated with broader metabolic disturbances than preclinical atherosclerosis, particularly affecting amino acid and lipid metabolism. TMAO was the only metabolite significantly linked to both conditions and their interaction, indicating a potential context-dependent role. Glutaminase activity emerged as a shared feature, suggesting convergent metabolic mechanisms. Despite distinct profiles, both conditions exhibited common alterations in oxidative stress, redox regulation, and mitochondrial pathways. These findings highlight the value of metabolomics in identifying early metabolic signatures of cardiometabolic risk. Electronic supplementary material Below is the link to the electronic supplementary material. [109]12933_2025_2841_MOESM1_ESM.jpeg^ (583.7KB, jpeg) Supplementary Material 1. Figure 1. Distribution of metabolite concentrations before and after normalization. [110]Supplementary Material 2^ (67KB, xlsx) Acknowledgements