Effects of resveratrol on changes in trimethylamine-N-oxide and circulating cardiovascular factors following exercise training among older adults

Purpose: Trimethylamine-N-oxide (TMAO) is a gut-derived metabolite associated with cardiovascular disease (CVD). In preclinical and observational studies, resveratrol and exercise training have been suggested as potential strategies to reduce the systemic levels of TMAO. However, evidence from experimental studies in humans remains unknown. This project examined the dose-dependent effects of a combined resveratrol intervention with exercise training on circulating TMAO and other related metabolite signatures in older adults with high CVD risk. Methods: Forty-one older adults [mean ( ± SD) age of 72.1 (6.8) years] participated in a 12-week supervised center-based, multi-component exercise training intervention [2 × /week; 80 min/session] and were randomized to one of two resveratrol dosages [Low: 500 vs. High:1000 mg/day] or a cellulose-based placebo. Serum/plasma were collected at baseline and post-intervention and evaluated for TMAO and associated analytes. Results: After the 12-week intervention, TMAO concentration increased over time, regardless of treatment [mean ( ± SD) Placebo: 11262 ( ± 3970); Low:13252 ( ± 1193); High: 12661( ± 3359) AUC; p = 0.04]. Each resveratrol dose produced different changes in metabolite signatures. Low dose resveratrol upregulated metabolites associated with bile acids biosynthesis ( i.e. , glycochenodeoxycholic acid, glycoursodeoxycholic acid, and glycocholic acid). High dose resveratrol modulated metabolites enriched for glycolysis, and pyruvate, propanoate, β -alanine, and tryptophan metabolism. Different communities tightly correlated to TMAO and resveratrol metabolites were associated with the lipid and vascular inflammatory clinical markers [|r| > 0.4, p < 0.05]. Conclusion: These findings suggest a distinct dose-dependent adaptation response to resveratrol supplementation on circulating metabolite signatures but not on TMAO among high-risk CVD older adults when combined with an exercise training intervention.


Introduction
Aging and cardiovascular disease (CVD) are global public health concerns.CVD is the leading cause of death globally (Roth et al., 2018;S. et al., 2020), and >17.9 million people die prematurely due to major adverse cardiovascular events (Lisy et al., 2018;World Health Organization, n.d.).More than 70 % of those with CVD in the USA are ≥60 years old, and these patients have a higher risk of morbidity (World Health Organization, n.d.;Yazdanyar and Newman, 2009).Moreover, CVD is one of the fastest-increasing conditions, projected to affect >130 million adults and to cost >$1 trillion in the USA in 2035 (S. et al., 2020;World Health Organization, n.d.).Thus, it is critically important to elucidate the pathophysiological mechanisms underlying CVD among older adults and identify potential therapeutic targets and treatment strategies to prevent cardiovascular morbidity and mortality.
CVD is the byproduct of multiple metabolic imbalances and lifestyle choice factors that trigger cellular inflammation signaling pathways and cardiometabolic biomarkers that disrupt metabolism homeostasis.Metaorganismal pathways and downstream metabolites have also been shown to influence host metabolism and CVD (Witkowski et al., 2020).One of the recent gut-microbiota-dependent pathways is mediated by trimethylamine-N-oxide (TMAO).TMAO is a gut microbiota metabolite formed from several nutritional substrates derived from the metabolism of phosphatidylcholine/choline, carnitine, betaine, dimethylglycine and ergothioneine by gut microbiota.This pathophysiological factor or its precursors (e.g., choline and L-carnitine) have been associated with CVD, the need for arterial surgery, incident mortality (Guasch-Ferré et al., 2017;Senthong et al., 2016;Tang et al., 2013), and major adverse cardiovascular events (Nie et al., 2018;Tan et al., 2019;Wang et al., 2011), in a concentrationdependent manner (Schiattarella et al., 2017).In the liver, high concentrations of TMAO have been shown to inhibit the hepatic reverse transport of cholesterol, decreasing bile acids synthesis and increasing cholesterol deposition in this organ and in systemic circulation (Koeth et al., 2013).This process reduces the expression of cholesterol 7α-hydroxylase (CYP7A1), the rate-limiting step in the formation of bile acids from cholesterol (Wilson and L., 2017), inhibiting bile acids' neosynthesis (Ding et al., 2018) which further exacerbates cholesterol accumulation and deposition in the liver and blood.Physiologically, these metabolic alterations increase the risk for CVD, inducing endothelial and vascular injury (Chen et al., 2017;Seldin et al., 2016), cholesterol deposition in macrophages and the generation of foam cells on arterial walls (Wang et al., 2011).Despite these physiological effects on CVD in animal models and observational studies, therapeutic strategies to prevent TMAO-associated effects in humans remain inconclusive.
On the one hand, considerable evidence highlights the well-known cardiometabolic benefits of exercise training on CVD morbidity and mortality (Lavin et al., 2022), but the effects of exercise on circulating TMAO remain incompletely described.Resveratrol, a natural polyphenolic compound with beneficial metabolic, antioxidant, antiinflammatory, and cardioprotective effects, has also been proposed as a nutraceutical intervention to potentially counteract the effects of TMAO (Chen et al., 2016;Larrosa et al., 2009).In a preclinical study, resveratrol supplementation modulated TMAO levels, cholesterol, and bile acid metabolism via increased hepatic bile acid synthesis through partial downregulation of the enterohepatic farnesoid x receptor (FXR)fibroblast growth factor (FGF)-15 axis (Chen et al., 2016).Resveratrol has also been proposed as a potential beneficial adjuvant to exercise training (Harper et al., 2021;Layne et al., 2017), but the combined effects on TMAO are unknown.
Thus, the objective of this study was to: (1) evaluate the effects of 12 weeks of resveratrol supplementation intervention combined with exercise training on systemic levels of TMAO in older adults with cardiovascular risk and (2) explore the cellular mechanisms underpinning this relationship.Our central hypothesis was that compared with the placebo group, 12 weeks of resveratrol supplementation will improve circulating indices of TMAO, bile acid synthesis, and clinical outcomes when combined with exercise training in older adults with CVD risk.

Study design overview
This study is an ancillary study to a three-arm, double-masked multicenter randomized controlled trial (RCT) to evaluate the effects of two resveratrol dosages combined with exercise training to improve physical function in older adults with CVD and functional limitations.The complete methodological study design and procedures including the statistical power calculations of this RCT were previously described (Harper et al., 2021;Layne et al., 2017).Briefly, participants were recruited from Gainesville, FL, and Birmingham, AL (USA) areas through printed advertisements and direct mailing.Interested volunteers completed a telephone pre-screening interview to determine eligibility for an in-person screening visit.Following the informed consent process, participants completed a set of questionnaires, including medical history, physical activity, medication use, and other eligibility criteria.
After determining that volunteers fulfilled all inclusion and exclusion criteria, participants returned to the center for baseline assessments.
Participants also provided a fasted blood sample for evaluation of lipid profiles and a comprehensive metabolic panel.Afterward, participants were randomized to 1:1:1 to one of three supplementation conditions using a permuted block randomization scheme stratified by age (65-75; >75 years) and gender (Harper et al., 2021;Layne et al., 2017).Eligible volunteers were randomly assigned to receive: (i) 500 mg/day resveratrol; (ii) 1000 mg/day resveratrol; or (iii) placebo.In addition, participants were asked to enroll in a supervised center-based exercise training intervention.Assessment visits occurred at baseline, week six, and after 12-weeks of intervention.These assessment visits collected data on clinical and health markers and different biospecimens for clinical and study assays (e.g., blood).
The study was overseen by a Data Safety Monitoring Board alongside a comprehensive study staff team, including physicians, nurse practitioners, exercise physiologists, the principal investigator, and study staff.Treatment allocation was concealed from all involved in the trial until completion of data analyses.All study procedures were approved by the University of Florida and the University of Alabama at Birmingham Institutional Review Boards (IRB-300000423).This study was registered at www.clinicaltrials.gov(NCT02523274).All participants signed a written informed consent form before enrollment for future use of biospecimens and data, consistent with the ethical procedures of the 1964 Helsinki Declaration and its later amendments for human studies by the World Medical Association (World Health Organisation, 2013).

Participants
The forty-one participants of this study are a sub-set of randomized volunteers that fulfilled the following criteria:

Intervention
The intervention included the oral intake of resveratrol, or a cellulose-based placebo combined with a supervised center-based multicomponent exercise intervention.All study procedures related to resveratrol supplementation and the multicomponent exercise training intervention are fully described below.

Resveratrol supplementation
Participants were randomized to one of two resveratrol dosages: Low (500 mg/day) or High (1000 mg/day) or a vegetable cellulose-based placebo provided by Reserveage Organics (Gainesville, FL, USA) through identical capsules.Participants were instructed to consume two daily pills, one capsule before breakfast and another before dinner.Volunteers randomized to the 500 mg/day consume one pill of the corresponding resveratrol dose and one pill of the Placebo.Participants were instructed to bring all unused pills at each assessment visit allowing the study staff to conduct pill counting to assess supplementation adherence.

Multi-component exercise training
All groups performed a supervised center-based multi-component exercise training intervention [2×/week; 80 min/session] for 12 weeks.The multi-component exercise training consisted of aerobic exercise, whole-body resistance exercise, and flexibility and balance exercises as follows: after a brief warm-up, participants performed a moderate to vigorous treadmill walking or, if not well-tolerated, participants pedaled in a stationary recumbent bicycle (LifeFitness, Schiller Park, IL, USA) [20 min/session at moderate intensity (5-6 rating of perceived exertion (RPE) scale) and the last 10 min/session at vigorous intensity (7-8 RPE)].After the aerobic training, participants executed whole-body resistance exercises [30 min/session] focusing on the upper-and lower body.In the first session of the week, participants performed leg press, leg extension, chest press, and seated row, while in the second session of the week, participants performed leg curl, calf flexion, arm curl, and triceps extension.The intensity of the resistance exercise training was set at 5-6 RPE and progressively increased to 7-8 RPE on Borg CR10 RPE scale and was performed in isotonic resistance training equipment (LifeFitness, Schiller Park, IL, USA).The sessions ended with flexibility and balance exercises [10 min/session] to cool down.

Biospecimen collection
Whole-blood was collected after overnight fasting (~8 h) to monitor hematologic and metabolic abnormal responses to the interventions at baseline, week 6 and 12 weeks according to standard clinical procedures.Serum was collected from whole blood (3.5 mL) in serum separator collection tubes (BD Medical, Franklin Lakes, NJ, USA).Samples were positioned upright for 30-60 min at ambient temperature and centrifuged for 20 min (3380 ×g).The serum was aliquoted in low-binding cryovials and stored at − 80 • C until further analyses.Plasma was collected from whole blood in ethylenediaminetetraacetic acid collection tubes (BD Medical, Franklin Lakes, NJ, USA).Plasma was isolated by centrifugation at ambient temperature (3380 ×g; 10 min).The supernatant was collected and centrifuged again at room temperature (10,000 ×g; 10 min).The plasma was then aliquoted and stored at − 80 • C until additional analyses.

Outcomes
2.5.1.TMAO, resveratrol, and untargeted metabolomics 2.5.1.1.Sample preparation.Aliquots of plasma (100 μL) were added to ice-cold PBS (100 μL).A Bligh-Dyer liquid-liquid extraction procedure was applied, first adding methanol: chloroform (2:1, by volume, 750 μL), followed by chloroform (250 μL) and ddH 2 O (250 μL) to each sample.After thorough vortexing samples were then centrifuged at 3000 ×g for 10 min at 4 • C to facilitate phase separation.The top layer containing the water-soluble metabolites was then transferred to a clean glass test tube and dried under nitrogen gas.Samples were re-suspended in ice-cold 80 % methanol (500 μL) to aid in additional protein precipitation.Samples were again centrifuged at 3000 ×g for 10 min at 4 • C, and the supernatants were transferred to a new test tube for drying under nitrogen gas.Once evaporated, each sample was re-suspended in ddH 2O containing 0.1 % formic acid (200 μL) for mass spectrometry analysis.Pooled samples containing an equal portion of each biological specimen were created for quality control and to annotate ion features.(LC-MS) analysis.An aliquot (10 μL) of each sample (serum extracts, pooled samples and extraction blanks) was loaded onto a 2.1 × 100 mm, 1.6 μm Luna Omega, 80 Å reverse-phase column (Phenomenex, Torrance, CA).The mobile phases were A) ddH 2 O with 0.1 % formic acid and B) acetonitrile with 0.1 % formic acid, respectively.Metabolites were resolved with a linear gradient of 2-50 % mobile phase B for 5 min, then 50-98 % B until 6.0 min with a 1-min hold, and then re-equilibration at initial conditions for 3 min.This was carried out using an Exion UHPLC (Sciex, Toronto, Ontario) and a flow rate of 500 μL/min.A SCIEX 5600 TripleTOF mass spectrometer (SCIEX, Toronto, Canada) was used to analyze the eluted metabolites.The IonSpray voltages for positive and negative electrospray ionization modes were ±5000/4500 V; the decluttering's potential was ±80 V. Ionspray GS1/GS2 and curtain gases were set at 40 psi and 25 psi, respectively.The interface heater temperature was 400 • C. Time-of-flight (TOF) survey scan from m/z 50-1000 were collected at Hz throughout each LC-MS run.The pooled sample was run every serum samples to assess and ensure instrument stability.

Liquid-chromatography mass-spectrometry
The pooled samples containing an equal portion of each biological specimen were used to annotate ion features.The LC-MS conditions used were identical to the serum specimens.Data were collected according to the following duty cycle: a 100 ms m/z 50-1000 TOF MS scan, followed by up to eight product ion TOF scans on selected molecular ions over the range from m/z 50-1000; these scans were collected over 50 ms intervals using a collision energy spread of 15 eV with a set collision point of 35 eV.Spectra were centroided and de-isotoped by Analyst software, version 1.81 TF (SCIEX, Toronto, Canada).

Data processing and metabolite annotation.
LC-MS data were processed using MS-DIAL version 4.70 (RIKEN Center, Yokohama City, Kanagawa) in conjunction with the Public positive and negative MSMS databases (version 16) to annotate ion features occurring across all samples, and to determine their peak areas and retention times.An Excel .csvfile was created containing each ion feature's m/z, peak area, and retention time.In addition, the annotation of TMAO, resveratrol, and other metabolites was assessed against the IROA 600 Metabolite Standards Compound Library (IROA Technologies, Sea Girt, NJ).A local database was developed under the same LC and MS conditions as used for the serum extracts.Although a much smaller database than the Public MSMS database has provided higher confidence in annotation since retention times and identical MSMS spectra could be matched with less error.In addition, MSMS spectra of annotated ion feature was manually inspected using PeakView 2.2 software (SCIEX, Toronto, Ontario).

Metabolomics data analysis.
For spectral analysis, we used the online statistical analysis tool MetaboAnalyst 5.0 (https://www.metaboanalyst.ca/)(Pang et al., 2020).Initially, samples were normalized by the total ion current and mean centered using Pareto scaling.Univariate analysis of variance was performed on data from both ion modes to identify variations in peak areas between groups at baseline and after the 12-weeks intervention.Partial least squares discriminant analysis (PLS-DA) was used to detect the most critical discriminant metabolites in each comparison (i.e., Placebo vs. Low dose and Placebo vs. High dose) consistent with published guidelines for metabolomics cardiometabolic trials (Rankin et al., 2016).Results were visually displayed using volcano plots, variable importance in projection (VIP) plots, and heatmaps using a Log 2 fold change ≥1.5 and a P value of <0.05.We also performed follow-up enrichment pathway analysis using the Functional Analysis module in MetaboAnalyst built-in metabolite set library (i.e., KEGG for Homo sapiens metabolites).Differences in mean outcomes measures between groups at baseline and after the 12-week intervention on TMAO and resveratrol metabolites were assessed using a general mixed model analysis of variance with Bonferroni posthoc correction for multiple comparisons.Differential TMAO and resveratrol conjugated metabolites analysis were performed and visually displayed on GraphPad Prism (version 9.2.0).Resveratrol and its conjugated metabolites were removed from the dataset and multivariate analysis was re-run to determine treatment × time interaction effects.In addition, the enrichment pathway analysis was applied to determine the biological relevancy of the metabolites altered by the intervention over time in both ion modes.

Farnesoid X receptor-fibroblast growth factor 19 axis
2.5.2.1.Enzyme-linked immunosorbent assay analysis.The fibroblast growth factor (FGF)19 was chosen as a marker of the farnesoid X receptor (FXR)-FGF-19 axis since it is the human ortholog of FGF15 (Somm and Jornayvaz, 2018).This growth factor is a hormone released by ileal enterocytes after stimulation of nuclear FXR, typically by absorbed bile acids, and provides negative feedback for bile acid synthesis in hepatocytes.This serum marker provided valuable information on bile acid synthesis and cholesterol metabolism and was validated as a biomarker in bile acid dysfunctional conditions (Vijayvargiya et al., 2017).The change in FGF19 concentration was assessed through the human enzyme-linked immunosorbent assay (ELISA) commercial kit (ThermoFisher Scientific, catalog number EHFGF19, USA) on serum samples at baseline and week 12 to determine the effect of the intervention, according to manufacturer instructions.This ELISA kit has an analytical sensitivity of 30 pg/mL (assay range between 32.77 and 8000 pg/mL) and an intra-assay coefficient of variation of <10 %.The BioTek Gen5 30.4 software was utilized with the BioTek 800 TS microplate reader (BioTek Instruments, Inc., Winooski, VT, USA) to detect and quantify the ELISA absorbance of FGF19 across samples.

ELISA data analysis.
The change in serum FGF19 after the 12week intervention was assessed through a mixed-model analysis of variance to determine treatment × time interaction effects with the posthoc Bonferroni correction method for multiple comparisons.The main effects of each factor (treatment and time) were investigated when no meaningful interaction effects were observed.These differential data analyses were performed and visually displayed on GraphPad Prism (version 9.2.0).The statistical threshold for meaningful differences was set at a p-value < 0.05.

Lipid panel and clinical markers data analysis.
Differential changes on circulating lipids and clinical markers were evaluated through a general mixed-model analysis of variance to determine treatment × time interaction effects followed by posthoc Bonferroni correction method for multiple comparisons.Similar to previous statistical analysis, the main effects of the treatment factor were investigated when no meaningful interaction effects occurred.These analyses were completed on GraphPad Prism (version 9.2.0), and the statistical threshold was set at a p-value < 0.05.
2.5.4.Integrative network analysis of TMAO, resveratrol, and altered metabolites with lipid panel and vascular inflammatory clinical markers 2.5.4.1.Integrative network statistical analysis.For biological interpretation, data integration, and differential network visualization (Perez De Souza et al., 2020), we explored potential meaningful relationships between TMAO, resveratrol, and altered metabolites with clinical markers.The integrative network analysis was performed using partial least square regression analysis with the eigenvector as the method for centrality analysis.This multivariate approach for integrative network analysis was set to include associations with |r| > 0.4 and a statistical significance level of p < 0.05.The multilevel community detection method was performed using xMWAS (Uppal et al., 2018), a data-driven integration and network analysis tool (version 0.552), accessed using the server-hosted Shiny app: https://kuppal.shinyapps.io/xmwas/rusing the Placebo, Low and the High dose resveratrol group samples separately in both ion modes.

TMAO, resveratrol metabolites and FGF-19
Post-intervention, TMAO, and resveratrol-conjugated metabolites presented different change patterns.There were no statistically significant treatment × time interaction effects, nor a main effect of treatment in TMAO relative abundance, but there was a main effect of time (Fig. 1A).TMAO increased over time, regardless of treatment [mean (±SD) Placebo: 11262 (±3970); Low:13252 (±1193); High: 12661 (±3359); F[1,74] = 4.0; p = 0.04].The relative abundance of resveratrol metabolites showed a main effect of treatment (Fig. 1B-E).Compared to the Placebo group, both resveratrol dosages (i.e., Low and High dose groups) exhibited a higher relative abundance of resveratrol metabolites (i.e., resveratrol sulfate, glucuronide, and their adducts) after the 12week intervention.However, only the High dose resveratrol was statistically significant across the different resveratrol-conjugated metabolites (p < 0.05).The abundance of FGF19 did not show any treatment × time interaction, nor a main effect of treatment, nor time [p > 0.05 for all comparisons (Fig. 1F)].
Fig. 1.Post-intervention metabolomics analysis of trimethylamine-N-oxide (TMAO) and resveratrol metabolites in older adults with cardiovascular risk.After the 12-week intervention, the relative abundance of TMAO increased over time, regardless of treatment (A).Similarly, compared to the Placebo group, resveratrol and resveratrol conjugated metabolites and adducts increased in both resveratrol dosages but were only statistically significant in the higher resveratrol supplementation dose (B, C, D, and E).The fibroblast growth factor 19 (FGF19) did not reveal any treatment × time interaction nor a main effect of treatment nor time (F).Metabolomics raw data were normalized by total ion current and subjected to mean centering and Pareto scaling and analyzed on MetaboAnalyst (version 5.0) using a univariate one-way analysis of variance or a general mixed-models analysis of variance followed by posthoc correction methods, accordingly.The statistical significance was set at log fold change ≥1.5 and p < 0.05.The relative abundance of TMAO, resveratrol (un)conjugated metabolites, and FGF19 were visually displayed using the same method on GraphPad Prism.Data are expressed as group means ±1 SEM (n = 41).* Statistical significance P ≤ 0.05.Abbreviations: FGF19: fibroblast growth factor 19; TMAO: trimethylamine-N-oxide.

Placebo vs. low dose resveratrol
After 12 weeks of intervention, LC-MS led to 111 and 145 annotated ion features in the negative positive ion modes, respectively.Univariate analysis of variance revealed significant annotated ion features altered by each comparison in both ion modes.In the negative ion mode, glycoursodeoxycholic acid [log 2 fold change (FC) = 1.7, p = 0.03] and glycocholic acid [log 2 FC = 2.1, p = 0.04] were the most distinct metabolites altered in this comparison.PLS-DA also revealed other important metabolites that were differentially affected between each group (Fig. 2A).Follow-up pathway analysis showed that these different metabolites were associated with vitamin B6, bile acid and fatty acid biosynthesis (Fig. 2B).
In the positive ion mode, glycochenodeoxycholic acid [log 2 FC = 1.7, p = 0.04] and glycocholic acid [log 2 FC = 2.0, p = 0.04] were the most discriminating metabolites in this comparison.PLS-DA revealed other important metabolites distinguishing each group, including resveratrolconjugated metabolites (Fig. 2C).The functional annotation of these distinct metabolites revealed an association with bile acid biosynthesis (Fig. 2D).

Serum lipid profile and vascular inflammatory clinical markers
Post-intervention, there were distinct change patterns in the lipid panel and the systemic vascular inflammatory markers across treatment groups.The lipid panel components did not exhibit any treatment × time interaction effect nor a main effect of each factor alone [p > 0.05 for all comparisons (Supplemental Digital Content 1A-D)].The vascular inflammatory clinical markers presented a main effect of time on eselectin (F[1,77] = 4.2; p = 0.04 (Supplemental Digital Content 1E)) and oxLDL (F[1,74] = 26.2;p < 0.001 (Supplemental Digital Content 1G)) but not on VCAM-1 (Supplemental Digital Content 1F).There were no treatment × time interaction effects nor a main effect of treatment on the vascular inflammatory clinical markers.

Placebo vs. low dose vs. high dose resveratrol networks in positive ion mode
In the positive ion mode, the Placebo community network was also distinct from the Low and High dose resveratrol group networks (Fig. 7D-F).In the Placebo network (Fig. 7D), there were vascular inflammatory markers as central nodes in distinct communities, such as eselectin (C3: green) and oxLDL (C1: light orange).
In the High dose resveratrol network, TMAO was positively correlated with the urinary excretion marker creatinine [|r| = 0.7, p < 0.05] which, in turn, was inversely associated with total cholesterol.TMAO was also associated with phenylacetylglutamine, a uremic toxin [|r| = 0.6, p < 0.05] and negatively correlated with VCAM-1.In contrast, resveratrol-conjugated metabolites (C6: light orange) were inversely associated with betaine [|r| = − 0.6, p < 0.05] and oxLDL.This network showed distinct vascular inflammatory clinical markers communities and correlated metabolites as central nodes, as shown in Fig. 7F.

Discussion
Resveratrol and exercise training have both shown promising impacts on TMAO and related analytes in pre-clinical and observational studies (Chen et al., 2016;Erickson et al., 2019), but their effects in humansparticularly in combinationare unknown.To our knowledge, we present here the first findings of a combined intervention of resveratrol with a multi-component exercise training intervention on TMAO and related analytes in older adults with high-CVD risk.Our metabolomics-based approachcombined with a multilevel integrative network analysis integrating TMAO and resveratrol metabolites, vascular inflammatory clinical markers, and other altered metabolite signaturesshowed distinct dose-dependent treatment responses after the 12-week intervention.
Specifically, resveratrolregardless of dosedid not alter directly TMAO responses in our study population.However, after the 12-week intervention, low dose resveratrol altered vascular inflammatory markers (i.e., e-selectin and oxLDL) and metabolites related to bile acid metabolism, including glycoursodeoxycholic acid, glycochenodeoxycholic acid and glycocholic acid.The multilevel integrative network analysis also showed distinct communities tightly correlated with TMAO and resveratrol metabolites and different correlation patterns in altered metabolites and clinical markers related explicitly to azelaic acid, tryptophan, and betaine metabolism.These findings support, at least partially, our central hypothesis that low-dose resveratrol supplementation (i.e., 500 mg/day) enhances bile acid biosynthesis when combined with a multi-component exercise training intervention.
We did not observe a direct dose-dependent response in TMAO circulating levels in the High dose resveratrol combined treatment comparison.However, our integrative network analysis found an inverse correlation of resveratrol-conjugated metabolites with betaine (i.e., one of TMAO's precursors) and oxLDL.Notably, high dose resveratrol (i.e., 1000 mg/day) altered distinct metabolites associated with glycolysis/gluconeogenesis, pyruvate, propanoate, tryptophan, and β-alanine metabolism.Resveratrol and exercise training have been shown to regulate glucose and amino acid metabolism by upregulating different metabolic signaling pathways, cellular and molecular enzymatic targets, or end-products involved in these pathways (Rabinowitz and Enerbäck, 2020;Sylow et al., 2017;Wu et al., 2018;Zhang et al., 2019).However, further research is needed to fully understand the physiological intricacies modulated by our high-dose combined treatment strategy, as recently, it was suggested that a 1000 mg/day resveratrol dose might elevate the CVD risk level in older adults (Mankowski et al., 2020).
Our findings are partially consistent with previous animal experimental studies (Chen et al., 2016;Ding et al., 2018).In a preclinical study using just resveratrol treatment, Chen et al. (Chen et al., 2016) showed that resveratrol attenuated the systemic levels of TMAO through a different mechanism by increasing hepatic bile acid neo-synthesis via the re-modulation of gut microbiota.This resveratrol gut-derived modulation process downregulated the enterohepatic FXR-FGF15 axis while increasing CYP7A1 expression counteracting TMAO-associated effects on cholesterol.We are still determining the reasons for our study's null treatment effects on TMAO, FGF19, and serum lipid levels and the distinct treatment dose-dependent effects on bile acid biosynthesis as there are currently no studies in humans using both treatments.However, based on previous preclinical studies, these findings could be the result of: (1) physiological differences between animals and human Fig. 2. Partial least square discriminant analysis (PLS-DA) of the untargeted metabolomics and enrichment pathway analysis after the 12-week intervention in both ion modes in older adults with cardiovascular risk.In the negative ion mode, both comparisons (i.e., Placebo vs. Low dose and Placebo vs. High dose) altered distinct metabolites.The PLS-DA of the comparison of Placebo against the Low dose resveratrol showed an alteration in several (un)conjugated bile acids metabolites (A-B).In the positive ion mode, most of the significant metabolites affected by this comparison were related to resveratrol-conjugated metabolites (C).The follow-up enrichment pathway analysis showed an upregulation of bile acid metabolism (D).The comparison of Placebo versus the High dose resveratrol in the negative ion mode modulated distinct metabolites of gluconeogenesis, pyruvate, and propanoate metabolism (E-F).In the positive ion mode, most metabolites altered against the Placebo group revealed an upregulation of resveratrol-conjugated metabolites (G).The enrichment pathway analysis showed an association with beta-alanine and tryptophan metabolism (H).Metabolomics raw data were normalized by total ion current and subjected to mean centering and Pareto scaling and analyzed on MetaboAnalyst (version 5.0) using a univariate one-way analysis of variance followed by posthoc correction methods.The statistical significance was set at log 2 fold change ≥1.5 and p < 0.05.The most important features were also compared using the partial least-square discriminant analysis in each ion mode and visually displayed into VIP score graphs for the top 25 most important features.The follow-up enrichment pathway analysis was also performed in the same platform using built-in libraries for metabolites functional annotation (i.e., KEGG for Homo sapiens metabolites).Data were calculated for n = 41.Fig. 3. Treatment × time untargeted metabolomics analysis on negative ion mode in older adults with cardiovascular risk.In the comparison of the Placebo against the Low dose resveratrol, there were 77 metabolites altered from baseline until the end of the intervention (A).The linear mixed model's analysis showed that seven metabolites were differently modulated in this comparison, including glycoursodeoxycholic acid (B), glycocholate (C), an unannotated plasma metabolite (i.e., C 8 H 8 O 5 , Plasma ID-144) (D), 2-methlylglutaric acid (E), 3-(4-hydroxyphenyl)lactic acid (F), 1,3-dimethylurate (G) and theophylline (H).Over time, there was a similar treatment × time change pattern in all these metabolites, except on 3-(4-hydroxyphenyl) lactic acid.Compared to the Placebo group, the Low dose of resveratrol decreased the relative abundance of this metabolite after the 12-week intervention.The follow-up functional annotation showed that these metabolites were associated with bile acids biosynthesis and citric acid metabolism.Metabolomics raw data were normalized by total ion current and subjected to mean centering and Pareto scaling and analyzed on MetaboAnalyst (version 5.0) using a linear mixed-models analysis of variance.The statistical significance was set at log fold 2- change ≥ 1.5 and p < 0.05.Data are expressed as group means ± 1 SEM (n = 41).models (Somm and Jornayvaz, 2018;Zeisel and Warrier, 2017); (2) insufficient re-modulation of gut microbiota in TMAO producing bacteria (Rath et al., 2020), and in turn, modulation of other distinct intestinal taxon microbiota communities; or (3) to the upregulation of bile acid pathways that do not directly involve the activation of FXR-FGF19 axis (Jia et al., 2021).
Overall, TMAO formation is influenced by distinct multi-step factors like the nutritional precursors (Borrel et al., 2017;Koeth et al., 2019Koeth et al., , 2013;;Rath et al., 2020), host gut microbiota composition (Gao et al., 2015;Koeth et al., 2019;Rath et al., 2020), and the inhibition/activation of the enzymatic activity of the hepatic flavin monooxygenases (Warrier et al., 2015).Our experimental study did not assess gut microbiota composition, which is one of the study limitations.Still, our untargeted metabolomics approach and the integrative network analysis revealed inverse resveratrol correlations with TMAO's dietary precursors, such as betaine in the High-dose resveratrol comparison.In addition, we also found a strong modulation of different gut-derived metabolites in both resveratrol dosages associated with tryptophan metabolism, which supports our hypothesis of a potential treatment shift change of gut microbiota composition towards a tryptophan-producing bacteria with our combined intervention.Due to our study design, it remains unclear if resveratrol supplementation or exercise or, in turn, the combination of both are implicated in this treatment effect since both have been associated with the modulation of tryptophan metabolism (Cervenka et al., 2017;Fei et al., 2022).Still, α,β-dihydroresveratrol is considered a potential aryl hydrocarbon receptor ligand in the tryptophan signaling pathway (Fei et al., 2022) which in our study was also positively associated with several metabolites of tryptophan metabolism in both resveratrol dosages.
Despite the dissimilarities in rodent and human bile acid metabolism since rodents only conjugate bile acids with taurine and not with glycine (Somm and Jornayvaz, 2018;Zeisel and Warrier, 2017), we also found an increase in bile acid biosynthesis in humans similar to those of Chen et al. (Chen et al., 2016) in a preclinical study.Specifically, we found an increase in glycochenodeoxycholic acid and glycocholic acid.These are primary bile acids that are glycine-conjugated and result from bile acid synthesis involving multi-step reactions via the classical 7α-hydroxylase, CYP7A1, and from alternate pathways involving CYP7B1, oxysterol 12αhydroxylase (CYP8B1) and sterol 27-hydroxylase (CYP27A1) (Jia et al., 2021).The increase in bile acid biosynthesis using our combined Low dose resveratrol treatment did not affect the FGF19 activity, nor the elevation in TMAO circulating levels postintervention suppressed bile acid biosynthesis.Therefore, our findings suggest that these conjugated glycine forms stimulated both classical and alternative bile acid biosynthesis pathways, maintaining a delicate homeostatic balance between bile acid synthesis and the FXR-FGF19 axis activation.Interestingly, chenodeoxycholic acid conjugated forms are the major cause of FXR activation/inhibition, and CYP8B1 is more sensitive to FXR activation in the liver, further stimulating/repressing bile acid neo-synthesis (Ding et al., 2018;Kim et al., 2007;Wang et al., 2003).We did not Fig. 4. Treatment × time untargeted metabolomics analysis on positive ion mode in older adults with cardiovascular risk.In the comparison of the Placebo against the Low dose resveratrol, there were 84 metabolites altered from baseline to postintervention in this ion mode (A).The linear mixed model's analysis of variance showed that four metabolites were differently modulated in this comparison, including piperine (B), 3-indolepropionic acid (C), and caffeine (D).Over time, there was a similar treatment × time change pattern in all these metabolites.The follow-up functional annotation analysis showed that these metabolites were related to caffeine and phenylacetate metabolism.Metabolomics raw data were normalized by total ion current and subjected to mean centering and Pareto scaling and analyzed on MetaboAnalyst (version 5.0) using a linear mixed-models analysis of variance.The statistical significance was set at log 2 fold change ≥1.5 and p < 0.05.Data are expressed as group means ± 1 SEM (n = 41).Fig. 5. Treatment × time untargeted metabolomics analysis on negative ion mode in older adults with cardiovascular risk.In the comparison of the Placebo against the High dose resveratrol, there were 78 metabolites altered from baseline until the end of the intervention (A).The linear mixed model's analysis of variance showed that ten metabolites were differently modulated in this comparison, including citraconic acid (B), 3-(2-hydroxyphenyl) propanoate (C), tryptophan (D), β-hydroxybutyric acid (E), kynurenine (F), inosine (G), 2-hydroxy-4-methylpentanoate (H), indoleacetic acid (I), and DL-3-(4-hydroxyphenyl) lactic acid (J).Overall, there was a similar treatment × time change pattern in all these metabolites except for tryptophan and kynurenine.Compared to the placebo group, the High dose of resveratrol decreased the relative abundance of tryptophan and kynurenine metabolites after the 12-week intervention.The follow-up functional annotation analysis showed that these metabolites were related to fatty acid biosynthesis and tryptophan metabolism.Metabolomics raw data were normalized by total ion current and subjected to mean centering and Pareto scaling and analyzed on MetaboAnalyst (version 5.0) using a linear mixed-models analysis of variance.The statistical significance was set at log 2 fold change ≥1.5 and p < 0.05.Data are expressed as group means ± 1 SEM (n = 41).evaluate the activity of CYP8B1, which regulates the shift to the alternative pathway once activated.Thus, further mechanistic studies are needed to understand these interconnected relationships of gut microbiota composition, and the cellular mechanisms of TMAO production, FXR-FGF19 axis, the CYP7A1 and CYP8B1 activation, and resveratrol on cholesterol and bile acids metabolism in this high-risk population.
While we found promising findings with the combined low-dose resveratrol and exercise training treatment strategy, several limitations should be acknowledged, including: (1) analysis with a relatively small sample size and the lack of a control group without the exercise training intervention does not allow to establish a causal effect of exercise training in combination with resveratrol supplementation and, thus, some of our findings need further confirmation; (2) the lack of assessment of participants' gut microbiota composition and their dietary patterns as well as other markers of the FXR-FGF19 axis like CYP7A1 and CYP8B1 expression may have hindered the complete understanding of these interconnected mechanisms; and (3) lastly, the use of statistical fold-change thresholds in our untargeted metabolomic data analysis and in the multilevel integrative network analysis may mislead biologically meaningful metabolic signatures even if they do not meet the relevant statistical thresholds.
In summary, our study provides the first evidence profiling the effects of two resveratrol dosages combined with exercise training on TMAO circulating levels using a metabolomics comparison approach and a multilevel integrative community network analysis in older adults with high CVD risk.Our data suggest a distinct resveratrol dose-dependent response on systemic levels of bile acid biosynthesis and different treatment metabolic signatures when combined with an exercise training intervention, but not an effect of treatment on TMAO circulating levels.The different dose-dependent metabolite signatures regulate distinct metabolic pathways and clinical systemic markers responses that should be further explored in the low-dose resveratrol treatment supplementation.Thus, a more extensive powered randomized clinical trial is needed to confirm these findings.
Supplementary data to this article can be found online at https://doi.org/10.1016/j.exger.2024.112479.Fig. 6.Treatment × time untargeted metabolomics analysis on positive ion mode in older adults with cardiovascular risk.In the comparison of the Placebo against the High dose resveratrol, there were 89 metabolites affected from baseline until the end of the intervention in the positive ion mode (A).The linear mixed model's analysis of variance showed that four metabolites were altered in this comparison, including creatine (B), carnitine (C), cotoin (D) and tryptophanol (E).Over time, the relative abundance of these metabolites revealed a similar change pattern (i.e., increased), regardless of the treatment group.The follow-up enrichment pathway analysis showed that most metabolites were associated with beta-oxidation of very long-chain fatty acids and carnitine synthesis (F).Metabolomics raw data were normalized by total ion current and subjected to mean centering and Pareto scaling and analyzed on MetaboAnalyst (version 5.0) using a linear mixed-models analysis of variance.The statistical significance was set at log 2 fold change ≥1.5 and p < 0.05.Data are expressed as group means ± 1 SEM (n = 41).
(a) age ≥ 65 years; (b) inactive: <150 min (min)/week of moderate physical activity on the Community Health Activities Model Program for Senior (CHAMPS) questionnaire; (c) randomized to either treatment group; (d) willingness to participate in all study procedures; (e) high 10-year atherosclerotic cardiovascular disease risk ≥ 5 % defined by the American Heart Association/American College of Cardiology criteria.Participants were excluded if: (a) failed to provide consent; (b) on resveratrol supplementation; (c) involved in a supervised rehabilitation program; (d) presented an absolute contraindication to exercise training according to the American College of Sports Medicine recommendations; (e) significant cognitive impairment or a Mini-mental State Examination (MMSE) score < 24; (f) severe medical conditions which did not allow safe participation; or (g) simultaneously participating in another intervention trial.

Table 1
Baseline characteristics of the study sample per group.
Notes: Data are presented as mean ± SD or n (percentage).Abbreviations: ASCDV: 10-year atherosclerosis cardiovascular disease risk score; BMI: Body mass index; DBP: Diastolic blood pressure; HDL: High-density lipoprotein; High: Resveratrol 1000 mg/day; LDL: Low density lipoprotein; Low: Resveratrol 500 mg/day; SBP: Systolic blood pressure.aThe10-year atherosclerosis cardiovascular disease risk percentage score was calculated only for subjects at University of Alabama at Birmingham through the American College of Cardiology and the American Heart Association on-line calculator (https://tools.acc.org/LDL/ascvd_risk_estimator/index.html#!/ calulate/estimator/). L.C.Baptista et al.