The present study was approved by the Local Ethical Commission of the National Laboratory Astana- Nazarbayev University on 24/09/2020, Approval No. 05-2020. The study aims, objectives, risks, and protocols were meticulously explained to all participants, and they were provided with leaflets containing relevant information before signing the informed consent form.

Patients were recruited through Karaganda Cardiac Hospital, “Multidisciplinary Hospital No. 2” in Kazakhstan, based on their medical diagnosis and patient profile, during the period from May 2022 to September 2023. The study was registered on ClinicalTrial.gov, ID: NCT06573892.

The inclusion criteria were patients (males and females) over 18 years of age, diagnosed with STEMI, who underwent primary percutaneous coronary intervention within 12 h after the onset of the disease. The diagnosis of STEMI was established based on clinical symptoms—chest pain lasting more than 20 min in combination with electrocardiographic changes (ST-segment elevation ≥ 1 mV in at least two adjacent leads (or a blockage of the left leg of the Gis bundle for the first time) and an increase in Creatine kinase and troponin levels.

The exclusion criteria were patients with cardiogenic shock, ineffective results of stenting of infarct-dependent arteries (weak filling or lack of filling of arteries), chronic intestinal diseases, or acute intestinal diseases at the screening time. Also excluded patients were those with dietary restrictions such as allergies or intolerance to grapes and those with clinically significant infections in the 6 months prior to the start of the study. For example, patients with infections requiring hospitalization or parenteral antimicrobial therapy, opportunistic infections or the presence of anamnesis of any disease for which (according to the researcher's assessment) an exacerbation is possible due to participation in the study. Other exclusion criteria include a history of oncological or lymphoproliferative disease 5 years before the baseline visit, psychiatric disorders, including recent (within the last year) or active suicidal thoughts, or patients who require additional coronary artery revascularization [Coronary Artery Bypass Grafting(CABG) or stenting].

Patients were randomized according to the type of dietary intervention by an independent research assistant who was not involved in any other aspect of the study. Participants were randomly assigned to one of the experimental groups (1:1) in sets of 6, utilizing the resource at https://www.sealedenvelope.com/simple-randomiser/v1/lists. The packets, which included either a polyphenol or a placebo, were pre-packaged based on the randomization scheme. As a result, participants, medical staff, and evaluators did not know which group each participant was in until the data was examined and the database was accessed.

Dietary intervention was carried out on the background of standard therapy after the acute phase of STEMI. The first group was prescribed daily for 3 months, no later than 30 minutes before meals, 15 ml of a concentrate of grape-extracted polyphenols, and the second group was prescribed a placebo. The polyphenol supplement used in this study has an established favorable safety profile. It conforms to all applicable state standards and regulations (Eurasian Economic Union Committee for Sanitary and Epidemiological Control Certificate No KZ.16.01.98.003.R.001043.11.21). Safety testing was conducted in an accredited laboratory for food product analysis (Nutritest No KZ.T.02.E0177). The only potential side effect of concern is a reaction to grape components, which was considered in the exclusion criteria. The placebo consisted of water with grape aromatizer and caramel coloring (E150) added at a low concentration to match the appearance of the polyphenol concentrate.

Participants were assessed at baseline (M0) and after 3 months of polyphenol intake (M3). Blood plasma and stool samples were collected from all study participants during hospitalization after primary percutaneous coronary intervention upon discharge from the hospital and after 3 months of intervention. In addition, clinical parameters were analyzed, including complete blood count (CBC); plasma lipid profile (including low-density lipoproteins (LDL); cholesterol (TC) and triglyceride (TG) levels; cardiac troponin I (cTnI); kidney function, including plasma creatinine; blood urea nitrogen (BUN); glucose (GLU); total protein (TPro); alanine transaminase (ALT); aspartate transferase AST; total bilirubin (TBil); creatine kinase-myocardial band (CK); urine test; electrocardiography and echocardiography.

Synergy between percutaneous coronary intervention with TAXus and cardiac surgery (SYNTAX) evaluation was conducted on all participants at the beginning of the study. Two interventional cardiologists (unaware of the TMAO results and clinical outcomes) evaluated the SYNTAX score using an online calculator (https://www.syntaxscore.com, version 2.28). The SYNTAX index was defined as the SYNTAX of the acute phase—the residual SYNTAX of the chronic phase, and the plaque progression group was defined as the highest tertile on the SYNTAX scale (≥3). For long-term risk prediction after percutaneous coronary intervention (PCI), the Logistic Clinical SYNTAX Score (LCSS) was assessed at the beginning of the study (33). All participants underwent an evaluation of the risk of developing CVD using Systematic Coronary Risk Evaluation 2 (SCORE2) at the M0 and M3. The risk was assessed as low, moderate and high (34). All participants received dietary guidance in accordance with Mediterranean diet and Dietary Approaches to Stop Hypertension (DASH) nutritional recommendations in accordance with the Republic of Kazakhstan's clinical protocol for STEMI patient management (35).

Quantitative determination of TMAO concentration in plasma samples was conducted utilizing a combination of high-performance liquid chromatography and tandem mass spectrometry (HPLC-MS/MS). The analytical instrumentation consisted of an Agilent 1260 Infinity chromatographic system coupled with a G6130A quadrupole mass spectrometer (both manufactured by Agilent Technologies, Santa Clara, CA, USA).

Blood specimens were collected in ethylenediaminetetraacetic acid (EDTA)—containing tubes and underwent immediate centrifugation (3,000 rpm, 15 min) to isolate plasma, which was subsequently cryopreserved at −80°C until analysis. The analytical protocol employed several high-purity reagents: 95% trimethylamine N-oxide standard, 95% formic acid, and 99.9% acetonitrile (all sourced from Sigma-Aldrich, St. Louis, MO, USA), along with ultrapure water (18.2 mg/L resistivity) generated using the Milli-Q purification system (Millipore, Burlington, MA, USA).

The sample preparation protocol involved protein precipitation by combining 100 μL of thawed plasma with 600 μL acetonitrile, followed by high-speed centrifugation (20,000 × g, 10 min) at 4°C. The resulting supernatant (100 μL) was diluted with an equal volume of water, and 10 μL of this preparation was injected into the HPLC-MS/MS system.

Chromatographic separation was achieved using a ZORBAX Eclipse XDB C-18 analytical column (2.1 × 75 mm, 3.5 μm particle diameter) with a corresponding guard column (12.5 × 4.6 mm, 3.5 μm) maintained at 80°C. The separation employed an isocratic elution method using a binary mobile phase system: solution A (0.125% formic acid in a 1:1 acetonitrile-water mixture) and solution B (0.125% formic acid in a 1:1 water mixture). The chromatographic conditions included a constant flow rate of 0.250 mL/min at a column temperature of 30°C. TMAO concentrations were determined by integrating the chromatographic peak areas from the resulting analytical curves.

The oxidant status (the potential of the body's antioxidant systems) was determined in blood plasma samples by a rapid test using the e-BQC apparatus (BioQuoChem-BQC redox technologies, Oviedo, Spain) with an assessment of the total antioxidant potential and the antioxidant capacity of fast and slow antioxidants (24). We immediately received results of fast (Q1) and slow (Q2) active antioxidants, as well as total charge (QT). According to standard curves, this data is further translated into CEAC (equivalent to vitamin C antioxidant capacity).

The composition of the gut microbiome was determined from stool samples. The collection of biomaterial (stool) for the gut microbiome analysis was carried out in DNA/RNA Shield™ test tubes, catalog number R1101 (Zymo Research). DNA extraction was performed using the ZymoBIOMICS™ DNA extraction kit according to the recommended protocol.

The 16S locus libraries were prepared using NEXTflex^® 16S V1-V3 Amplicon-SeqKit (Perkin Elmer, catalog number NOVA-4202-04). Amplicons were sequenced on the MiSeq device (Illumina). Analysis: Demultiplexing, filtering, and determination of amplicon sequence variant (ASV) and taxonomic identification were performed using the LotuS program. After the abundance filtering at M0, the polyphenol group remained 36, in the placebo group 41 samples, respectively; at M3, the polyphenol group remained 25, and in the placebo group 25 samples. The 16S amplicon sequencing data, starting from raw reads, underwent taxon density tables processing using the Less OTU Scripts 2 (LotuS2) pipeline. Critical steps in this process included demultiplexing, quality filtering, and dereplication of reads, which were carried out with the assistance of a straightforward demultiplexer (sdm). Additionally, chimeric sequences were identified and eliminated using UCHI Marker Examiner (UCHIME) algorithms. Taxonomic postprocessing and sequence clustering, employing the combined databases such as SILVA, Greengenes 2 (GG2), and Human Intestinal Tract database (HITdb), were performed using Lowest Common Ancestor (LCA) and Divisive Amplicon Denoising Algorithm 2 (DADA2) sequence clustering algorithms, respectively (36). The total number of reads accounted for 28,147,916, employing a similarity threshold of 97% for distance comparison. After filtering, the sequences were categorized into 7754 ASVs and were attributed to the bacterial domain, with 25,637,043 reads remaining in the matrix.

To further investigate the functional potential of the microbial community, metabolic pathway analysis was performed using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) version 2.5.0. Following default settings, PICRUSt2 predicted functional metagenomic profiles based on the 16S rRNA sequencing data. To determine the gene family copy numbers for each amplicon sequence variant (ASV), a reference tree with a Nearest Sequenced Taxon Index (NSTI) threshold of 2 was used.

Statistical analysis of results was performed on R v4.4.1 and Python v3.9.14. Independent T-test and Mann–Whitney U-test were used to compare groups. Tests were automatically selected based on normality and homoscedasticity criteria. The Shapiro and Levene tests were used to test these criteria. α and β diversity were evaluated using vegan v2.6.1 and OTUtable v1.1.2. Within-sample diversity was assessed using Shannon and Chao1 indices. Between-sample diversity was evaluated using Bray-Curtis metrics on Hellinger-transformed data. Differential analysis on taxonomic data was performed using Linear discriminant analysis Effect Size (LEfSe) v1.10.0 (LDA > 2 and p ≤ 0.05) from the microbiome Marker package and STAMP v2.1.3 on functional data. Only features with at least 30% prevalence were retained for the analysis. The correlation was computed using Pearson's r coefficient from SciPy v1.10.1. Right-hand outliers in the clinical data were replaced by 95%. Compositional data was transformed using the centered log-ratio (clr) method, and all other variables were standard scaled. All multiple comparisons were adjusted using the False Discovery Rate with Benjamini–Hochberg (FDR BH) method, and the significance level was set at 0.05. No numerous comparison adjustments were applied when comparing significantly differentially abundant parameters. Visualization was performed with ggplot2 v3.5.1, Matplotlib v3.7.1 and seaborn v0.11.2. The two-sided value of p < 0.05 was considered statistically significant.

In total, 132 patients were identified and invited to participate in the study; 12 declined to participate, and 10 did not respond. The remaining 110 patients underwent initial screening, after which three refused to continue in the study, and seven failed the inclusion criteria. The included patients were randomized into two groups of 50 each; the first group received polyphenol concentrate and the second placebo, respectively. However, five more patients in each group were lost as they either declined to continue or did not comply (Figure 1). A total of 45 participants from both study groups completed all study visits through month 3 (M3).

There was no significant difference in the average age of each group, with 64.0 [55.0–66.0] and 61.0 [55.0–68.0], p = 0.83, respectively. Similarly, there were no significant differences in gender, ethnicity, social status, and other disease-related characteristics such as smoking, BMI, SYNTAX Score I, LCSS, Score 2, cTnL, and Main diagnosis and accompanied comorbidities, p > 0.05 (Table 1).

Intergroup comparative analysis of clinical and laboratory characteristics at baseline (M0) and 3-month (M3) follow-up demonstrated statistically significant differences in TMAO levels that were reduced by half after 3 months in the polyphenol group, p < 0.01 compared to placebo (Table 2). The LDL level was high in the polyphenol group compared to placebo at the start of the experiment, p = 0.005 and remained unchanged (Table 2 and Supplementary Table S1).

Analysis of clinical and laboratory parameters from baseline (M0) to 3 months (M3) revealed a significant two-fold increase in TMAO levels in the placebo group (from 0.5 [0.3–0.6] to 0.7 [0.5–1.4], p < 0.001; Table 3), while TMAO levels remained stable in the polyphenol group (from 0.5 [0.2–0.9] to 0.4 [0.3–0.9], p > 0.05). Interestingly, the AST and ALT levels decreased in the polyphenol group from M0 to M3, p = 0.007 and p = 0.03, respectively. The LDL level was elevated in the placebo group from M0 to M3, p = 0.004. The ejection fraction slightly increased in both study groups, p ≤ 0.05 (Table 3).

There were significant differences in blood and urine tests; monocyte levels in blood and leucocytes in urine decreased in both study groups over the period, p ≤ 0.05. The urine pH decreased in the polyphenol group at M3 compared to M0, p = 0.007. Other parameters did not demonstrate any differences at the baseline and after 3 months (Supplementary Table S3).

Analysis of α diversity showed no significant differences between the study groups at baseline (M0) and 3-month (M3) follow-up (p > 0.05) (Figure 2A). Yet, significant differences were found in both Shannon and Chao1 indexes during intragroup analysis (baseline M0 vs. 3-month M3 comparison) in both Polyphenol and Placebo groups, with α diversity decreased at 3 months compared to baseline (Shannon index: p = 0.0002 and p = 0.021; Chao1 index: p ≤ 0.0001 and p = 0.0007, respectively), (Figure 2B).

Similarly, analysis of β diversity by the Bray-Curtis index did not reveal any compositional differences between the study groups at baseline (M0) and 3-month (M3) follow-up (p > 0.05, (Figure 2C). Meanwhile, significant intragroup differences (baseline M0 vs. 3-month M3 comparison) were observed, with distinct clusters formed by PCoA using Bray-Curtis dissimilarity (ANOSIM: R = 0.27, p ≤ 0.0001 for the Polyphenol group; R = 0.23, p ≤ 0.001 for Placebo group). The first two principal coordinates explained 30.7% of the total variance (PCoA1: 19.56%, PCoA2: 11.14%) in the Polyphenol group and 30.4% (PCoA1: 19.35%, PCoA2: 11.05%) in the Placebo group (Figure 2D). The Firmicutes/Bacteroidetes (F/B) ratio did not experience statistically significant shifts in the placebo group between M0 and M3, p = 0.826, and between two study groups at M0, p = 0.6540, and M3, p = 0.141. However, there was a significant decrease in the F/B ratio in the polyphenol group between M0 and M3, p = 0.04 (Figure 2E).

Most statistically significant changes in relative abundance were observed at the species level: Bacteroides dorei (polyphenol group p = 0.001, placebo group p = 0.007), Bacteroides eggerthii (polyphenol group p = 0.03, placebo group p = 0.002), and Parabacteroides merdae (polyphenol group p = 0.04, placebo group p = 0.02), decreased significantly in both study groups, between M0 and M3, while the abundance Bacteroides plebeius (p = 0.03) and Bacteroides vulgatus (p = 0.001) noticeably decreased in placebo group between M0 and M3. Roseburia significantly increased in the polyphenol group between M0 and M3 at the genus level, p = 0.01 (Figure 3A). At the family level, Akkermansiaceae significantly decreased in the Placebo group, between M0 and M3, p = 0.03, and at the order level, Lachnospirales were elevated in the Polyphenol group between M0 and M3, p = 0.01, and Oscillospirales was significantly higher in Placebo group compared to Polyphenol group at M3, p = 0.002 (Figures 3B–D). Refer to Supplementary Tables S3–S5 for p-values corresponding to these differences.

Linear discriminant analysis effect size analysis (LefSe) indicating significant taxonomic changes (LDA > 2, p ≤ 0.05) identified several discrepant microbial taxa in the polyphenol and placebo groups at M3 point (Figures 4A, B).

Specific taxa dominated in the polyphenol group compared to the placebo group: species Agathobaculum sp., p = 0.004, effect size = 2.97; Alistipes finegoldii, p = 0.04, effect size = 2.44; and genus Sellimonas, p = 0.002., effect size = 2.35. On the contrary, species Alistipes timonensis, p = 0.003, effect size = 2.28; Coprobacter secundus, p = 0.006, effect size = 2.37; Lactobacillus salivarius, p = 0.01, effect size 2.6; Parabacteroides goldsteinii, p = 0.04, effect size = 2.9; Prevotella timonensis, p = 0.02, effect size =2.11; Bifidobacterium dentium, p = 0.02, effect size = 2.09; Sanguibacteroides justesenii, p = 0.03, effect size 2.2; Bacteroides pectinophilus, p = 0.04, effect size 2.1; Ruminococcaceae bacterium, p = 0.04, effect size = 2.1; genera Prevotella, p = 0.01, effect size 3.4, Ligilactobacillus, p = 0.04, effect size 2.7; Sanguibacteroides, p = 0.03, effect size = 2.2; Coprobacter, p = 0.01, effect size = 2.7 etc. were enriched in the placebo group, and depleted in polyphenol group, respectively (Figures 4A, B). The specific bacterial taxa identified by LefSe analysis showed significant differences not only between study groups but also within each group (polyphenol or placebo) from baseline to 3 months, highlighting distinct patterns of microbial changes (p ≤ 0.05) (Figure 4C). Overall, our LEfSe analysis demonstrated domination of various taxa from the Firmicutes, Bacteroidetes, Cyanobacteriota and Actinobacteria phyla in the placebo group, while only a few taxa in the Polyphenol group. The remaining Lefse results are presented in Supplementary Tables S3–S5.

Correlation analysis between the taxa and clinical-laboratory parameters demonstrated significant relationships, thus negative correlations of s. Alistipes finegoldy (p = 0.04), s. B. pectinophilus (p = 0.01) with TMAO were observed, whilst s. Alistipes timonensis demonstrated a positive correlation with TMAO (p = 0.01). Regarding the rest of the positive correlations, s. Lactobacillus salivarius positively correlated with AST (p = 0.009) and TC (p = 0.03), s. B. pectinophilus with CK (p = 0.05), s. A. finegoldy with SCORE 2 (p = 0.003), respectively. s. S. justesenii negatively correlated with TPro (p = 0.02) (Figure 5A and Supplementary Table S6).

In addition, analysis of the predicted correlation between metabolic pathways and clinical-laboratory parameters demonstrated a few significant relationships. In contrast, P125-PWY-super pathway of (R, R)-butanediol biosynthesis, p = 0.0002, P161-PWY-acetylene degradation, p = 0.02, UDPNAGSYN-PWY-UDP-N-acetyl-D-glucosamine biosynthesis I, p = 0.02, demonstrated a significant negative correlation with EchoEF, PWY-6588-pyruvate fermentation to acetone demonstrated positive correlation with TMAO, p = 0.04 (Figure 5B and Supplementary Table S7).

Further, the Sankey plot illustrating the networks between LEfSe markers, metabolic pathways, and TMAO levels showed that specific taxa nodes demonstrated various flows to nodes of metabolic pathways and connected to TMAO in two study groups. s_Parabacteroides_goldsteinii increased in the placebo group and decreased in the polyphenol group, respectively, appears to have strong positive correlations with several metabolic pathways (K01035, PWY-6588, K00563) and with TMAO in the placebo group, as well as negative correlations with K03628 and TMAO, particularly under the polyphenol condition. s_Alistipes_timonensis showed a mix of positive with K00563 and negative correlations with B09108, K03628, and K02190 metabolic pathways; a notable one is complex negative connections with B09108 and with TMAO under the polyphenol condition. s. S. justesenii also demonstrated complex relationships, a remarkable negative correlation with K02190 and TMAO level under the polyphenol condition (Figure 5C). Alistipes finegoldii and Prevotella timonensis demonstrated opposite correlations with K02190, positive and negative, respectively, and Coprobacter secundus positively correlated with K00563 (Figure 5C). Although the observed correlations were relatively weak and moderate (Pearson's r ranging from −0.18 to 0.36), they were statistically significant and can be associated with the biological complexity of microbial-metabolite interactions.

Thus, the analysis of metabolic pathways between two study groups at M3 was conducted. Three pathways, PWY-6588 pyruvate fermentation to acetone (p = 0.027), P161-PWY-acetylene degradation (anaerobic), (p = 0.028), UDPNAGSYN-PWY-UDP-N-acetyl-D-glucosamine biosynthesis (p = 0.029) were significantly lower in polyphenol group comparing to the placebo group. In contrast, PWY-5005 biotin biosynthesis biotin was higher in the polyphenol group compared to the placebo group (p = 0.02) (Figure 5D).