Gut microbiome mediates the protective effects of exercise after myocardial infarction

Gut microbiota plays important roles in health maintenance and diseases. Physical exercise has been demonstrated to be able to modulate gut microbiota. However, the potential role of gut microbiome in exercise protection to myocardial infarction (MI) remains unclear. Here, we discovered exercise training ameliorated cardiac dysfunction and changed gut microbial richness and community structure post-MI. Moreover, gut microbiota pre-depletion abolished the protective effects of exercise training in MI mice. Furthermore, mice receiving microbiota transplants from exercised MI mice had better cardiac function compared to mice receiving microbiota transplants from non-exercised MI mice. Mechanistically, we analyzed metabolomics in fecal samples from exercised mice post-MI and identified 3-Hydroxyphenylacetic acid (3-HPA) and 4-Hydroxybenzoic acid (4-HBA), which could be applied individually to protect cardiac dysfunction post-MI and apoptosis through NRF2. Together, our study provides new insights into the role of gut microbiome in exercise protection to MI, offers opportunities to modulate cardiovascular diseases by exercise, microbiome and gut microbiota-derived 3-HPA and 4-HBA. 2VpFEiBoi5zKv8qdp_bZN4 Video Abstract Video Abstract


Background
The number of genes associated with gut microbes vastly exceed the total complement of genes in the host [1]. The gut microbiome is critical for maintaining host physiology and homeostasis through metabolic exchange and co-metabolism of substrates [2]. Moreover, the gut microbiome has emerged as an important regulator in numerous facets of human health and disease. Mounting evidence has linked alterations of gut microbiome with a variety of disease including obesity, type 2 diabetes mellitus, fatty liver, hypertension, heart failure, and myocardial infarction (MI) [3][4][5][6][7][8][9]. As the gut microbiome is therapeutically modifiable, manipulation represents a novel opportunity to combat chronic diseases.
MI is a major cause of death worldwide [10]. With the advances in therapeutic approaches of mechanical reperfusion such as percutaneous coronary intervention (PCI), the acute mortality rates of MI have been significantly reduced while the post-infarction cardiac remodeling and heart failure has been a huge healthcare and economic burden [11]. Cardiac rehabilitation (CR) is a recommended effective adjunct therapy for patients of MI and exercise training is a powerful tool in CR programs [12,13]. Exercise-based CR can lead to improved exercise capacity and prognosis in MI patients [14]. Understanding the molecular mechanism underlying the protective effects of exercise in MI can help identify novel therapy for post-infarction cardiac remodeling and heart failure.

Open Access
*Correspondence: junjiexiao@shu.edu.cn Alterations of the gut microbiome have been reported upon MI in both animal models and human patients [15,16]. Mice receiving exercise training demonstrated favorable changes in the composition of gut microbiota [17]. In humans, athletes have a higher richness and diversity of the gut microbiome [18]. However, the role of gut microbiome in the protective effects of exercise in cardiac dysfunction post-MI is unclear. In our study, we aimed to explore the potential roles of gut microbiome in exercise protection in MI.

Animals
Male C57BL/6 mice aged 8-10 weeks were purchased from Cavens Lab Animal (Changzhou, China), and raised at the specific pathogen-free (SPF) laboratory animal facility of Shanghai University (Shanghai, China). Mice were maintained on a 12-h light/dark cycle at 25 °C and provided free access to commercial rodent chow (sterilized by Cobalt-60) and tap water (high-temperature sterilization) before initiation of the experiments. Randomized grouping was used and the same group of mice were co-housed with less than 5 animals per cage. All animal experiments were in accordance with the guidelines approved by the committee on the Ethics of Animal Experiments of Shanghai University.
Left anterior descending coronary artery was ligated to create the MI mice model, as we applied before [19]. Under sterile conditions, left anterior descending coronary artery was tied by a 7-0 silk suture. Sham-operated (Sham) mice were treated with the same surgery without tying the left anterior descending coronary artery. Exercise therapy via running was carried out 1-week after MI surgery.
In the running model, MI mice were placed in a treadmill, starting from 10 min with 5 m per minute and we increased distance 2 m per minute each day until 60 min at 15 m per minute [20]. Mice were sacrificed after training a total of 8 weeks of running.
The treadmill running test was used to measure endurance capacities of mice. Mice were placed in the treadmill to adapt to the environment for 5 min and then run at the speed of 15 m/min. We increased the speed by 1 m/min every 4 min and recorded the running speed and running time when the mice were exhausted. According to the speed and time, the general movement path of mice was calculated as endurance capacities.
Transthoracic echocardiography examination was used to determine cardiac function as demonstrated by left ventricular ejection fraction (EF) and fractional shortening (FS) performed with the VisualSonics Vevo 2100system (VisualSonics Inc, Toronto, Ontario, Canada) with a 30 MHz central frequency scan head and measured from M-mode images taken from the parasternal short-axis view at papillary muscle level. Each mouse was anesthetized with 1.5% isoflurane and measured at least three times.

16S rDNA profiling of gut microbiota in mice
At the endpoint of the experiments, one mouse was put in a cage and then the mice feces (100 mg per mice) collected under sterile conditions were frozen using liquid nitrogen and stored at -80 ℃. Fecal samples, packed with dry ice, were sent for analyses to the laboratory at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).
For 16S rRNA statistical analyses, Pan was calculated by Usearch. Alpha diversity was calculated by Mothur. Principal co-ordinates analysis (PCoA) and non-metric multidimensional scaling analysis (NMDS) were performed by R version 3.3.1, vegan, and mixOmics package, and significant differences were assessed by analysis of similarities (ANOSIM). Enterotype analysis was calculated by the statistical clustering method, which was based on Jensen-Shannon distance (JSD). Correlation analysis was assessed by Spearman. Heatmap, barplot, and pielot were performed by R version 3.3.1, vegan, and mixOmics package. Significant differences were assessed by Wilcoxon rank-sum with FDR.

Antibiotic treatment and fecal microbiota transplantation
The maximal dose of antibiotic cocktail was prepared by mixing ampicillin (0.25 mg/mL), metronidazole (0.25 mg/mL), neomycin (0.25 mg/mL), and vancomycin (0.125 mg/mL) in autoclaved water [15]. Animals that received the antibiotic cocktail were referred to as ABX mice, whereas drink autoclaved water were referred to as untreated mice. ABX mice drank maximal dose of antibiotic cocktail for 7 days post-MI surgery, followed by normal autoclaved water. 1/4 ABX mice received a 25% dose of the antibiotic cocktail until the end of the experiment.
For microbiota transplantation, the endpoint-mice fecal samples (100 mg per mice), collected under sterile conditions, were resuspended with pre-cooled PBS, centrifuged for supernatant with 1000 rpm for 5 min at 4 °C, and this step was repeated twice. Then glycerin (20%) was added before storage at -80 ℃. Mice drank maximal dose of antibiotic cocktail for 7 days post-MI surgery and then were orally inoculated (200 μL for each mouse) at 1-day interval for 1 week. Cardiac function and histopathological detection were detected after 8 weeks.

MicrobioMET
At the endpoint of the experiments, one mouse was put in a cage and then feces samples (100 mg per mice) collected under sterile conditions were frozen using liquid nitrogen and stored at -80 ℃. Fecal samples packed with dry ice were sent to the laboratory at Metabo-Profile Biotechnology Co., Ltd. (Shanghai, China). Metabo-Profile lnc. (Shanghai, China) performed the quantitation of bacterial metabolites as previously reported [23,24], using a database composed of 132 standards obtained from Sigma-Aldrich (St. Louis, MO, USA), Steraloids Inc. (Newport, RI, USA), and TRC Chemicals (Toronto, ON, Canada). In this project, the microbial metabolites were quantitated by ultra-performance liquid chromatography coupled to tandem mass spectrometry (UPLC-MS/MS) system (ACQUITY UPLC-Xevo TQ-S, Waters Corp., Milford, MA, USA).
Univariate statistical analyses were assessed either by student T test or Wilcoxon Test, depending on the normality of data and homogeneity of variance. Multivariate statistical analyses (OPLS-DA) were used to observe the similarities and dissimilarities among the groups. Pathway bubbleplot and network diagram were assessed by R language, ggplot2 package, and igragh and Cairo packets. Correlation analysis was assessed by Spearman.

Metabolites treatment
To prepare the metabolites supplement, 3-Hydroxyphenylacetic acid (3-HPA, SANTA CRUZ), 4-Hydroxybenzoic acid (4-HBA, SIGMA), and p-Hydroxyphenylacetic acid (4-HPA, SIGMA) were individually dissolved in autoclaved in saline. Mice drank maximal dose of the antibiotic cocktail for 7 days post-MI surgery and then (See figure on next page.) Fig. 2 Running training changes gut microbial richness and community structure in mice after MI. A Pan analysis for four groups based on genus level (n = 7:18:12:12). B Fecal bacterial community at the phylum level among Sham + control, MI + control, Sham + Run, and MI + Run. C PCoA analysis based on the relative abundance of genus between Sham + control and MI + control groups. D PCoA analysis based on the relative abundance of genus between MI + control and MI + Run groups. E A total of 49 samples were clustered into enterotype 1 (green), enterotype 2 (red), and enterotype 3 (blue) at the genus level. F The percentage of Sham + control, MI + control, Sham + Run, and MI + Run samples distributed in three enterotypes. G Five changed genera among Sham + control, MI + control, Sham + Run, and MI + Run. Significant differences were assessed by ANOSIM analysis in C, D; abund jaccard analysis in E; Wilcoxon rank-sum with FDR in F. *p < 0.05, **p < 0.01, ***p < 0. were orally inoculated (3-HPA, 4-HBA, 4-HPA, 6, 25 mg/ kg per day) at 1-day interval for 4 weeks.

Histological analysis
Histopathological detection was used to analyze ventricular remodeling. Hematoxylin-Eosin (HE) staining and wheat germ agglutinin (WGA) staining was used to measure cardiomyocyte size. For HE staining, mouse heart tissue fixed in 4% paraformaldehyde followed by buried with paraffin. Paraffin Sect. (

Cell isolation, culture, treatments, and immunofluorescent staining
Neonatal rat cardiomyocytes (NRCMs) were isolated and cultured as previously reported [25,26]. To induce OGD/R, NRCMs were firstly cultured for 8 h with serumfree no glucose DMEM (Gibco, U.S.A) in an air-tight chamber with a humidified hypoxic atmosphere containing 5% CO 2 and 95% N 2 at 37 °C. After exposure to oxygen glucose deprivation for 8 h, the culture medium was replaced with serum and glucose-containing DMEM and transferred to a normal incubator for recovery for 12 h. For metabolites treatment, 3-HPA and 4-HBA were incubated with indicated duration and concentration. Terminal deoxynucleotidyl transferase-mediated dUTP in situ nick-end-labeling (TUNEL) staining was conducted to detect apoptotic nuclei by confocal microscopy in α-actinin-labeled cardiomyocytes, as described before [27]. Fifteen fields/sample (200 × magnification) were viewed under a confocal microscope (Leica, Germany).

RNA isolation and relative quantitative RT-PCR
RNA isolation and relative quantification RT-PCR were performed as described previously [26]. The sequences of primers used for RT-PCR were as follows. ANP forward: AGC CGT TCG AGA ACT TGT CTT, ANP reward: CAG GTT ATT GCC ACTT AGG TTC A, BNP forward: GAG TCC TTC GGT CTC AAG GC, BNP reward: TACAG CCC AAA CGA CTG ACG. 18 s RNA were used as internal controls.

Statistical analysis
Data from the mouse and cell model were expressed as mean ± SD. Significant differences were assessed either by two-tailed student t test, one-way ANOVA followed by Bonferroni's post hoc test, or two-way ANOVA followed by Bonferroni's post hoc test when appropriate. P values less than 0.05 were statistically different. Analyses were performed using GraphPad Prism 7.0.4.

Data availability
The raw reads were deposited into the NCBI Sequence Read Archive (SRA) database (Accession Number: SRP287461). The data for this study were available by contacting the corresponding author upon reasonable request.

Exercise training mitigates cardiac dysfunction post-MI
To determine if exercise training could be protective for cardiac dysfunction post-MI, mice were subjected to treadmill running 1-week post-MI (Fig. 1A). Eightweek exercise training significantly improved cardiac function as demonstrated by elevated ejection fractions (EF) and fractional shortening (FS) (Fig. 1B) in post-MI mice. Moreover, endurance capacity of these mice was also enhanced (Fig. 1C), whereas the mRNA expression levels of atrial natriuretic peptide (ANP) and brain natriuretic peptide (BNP) were increased in MI and were decreased by exercise training (Fig. 1D and E). Moreover, Masson staining showed that exercise training could significantly reduce post-MI cardiac fibrosis (Fig. 1F), while hematoxylin-eosin (HE) and wheat germ agglutinin (WGA) staining displayed that the ultrastructure of myocardium was well-preserved and cardiomyocyte cross-sectional area was decreased in the exercised group post-MI ( Fig. 1G and H). Additionally, the protein expression levels of Collagen1, and the ratios of Bax/Bcl 2 and cleaved Caspase3/Caspase3 were increased in MI and were decreased by exercise training (Fig. 1I). Collectively, these data consistently demonstrate that exercise training can mitigate cardiac dysfunction post-MI.

Exercise training changes gut microbial richness and community structure post-MI
To identify whether the protective effects of exercise training for MI are associated with gut microbial content and composition, we performed 16S rDNA profiling of fecal samples from four groups, including Sham + control, MI + control, Sham + Run, and MI + Run. Bacterial DNA was extracted from fecal samples, sequenced on the Illumina platform, and an average of 56,638.65 ± 12,105.23 (SD) sequences per sample were generated (Additional file 1: Table S1). Pan analysis based on the genus level indicated that MI decreased the level of gene enrichment while running exercise rescued the decreased gene enrichment in MI ( Fig. 2A). Comparison of alpha diversity (the observed richness) based on the genus level indicated that MI and running did not change the community diversity of gut microbial (Additional file 1: Table S2 and S3).
In the feces of these 49 mice, Bacteroidota was predominant, represented by 42.30-57.92% of the 16S rRNA gene sequences. The second most abundant phylum was Firmicutes, with 37.68-50.83% in each group. Interestingly, in MI + Run group, Bacteroidota was higher abundant and Firmicutes was lower abundant than in the non-exercisegroup ( Fig. 2B and S1, Additional file 1: Table S4). At the genus level, detected ASVs were distributed among 136 different bacterial genera in total ( Fig. S2 and S3, Additional file 1: Table S5). We used principal coordinate analysis (PCoA) to examine the community structures of the gut microbiotas, and found that the gut microbiota of mice in MI were separated from Sham (Fig. 2C). In addition, the gut microbiota of MI mice in the running group were separated from sedentary group (Fig. 2D).

Antibiotic treatment inhibits the protective effects of exercise training in cardiac dysfunction post-MI
To investigate if gut microbiota in general is responsible for the beneficial effects of exercise training in cardiac dysfunction post-MI, we administered oral antibiotics to male mice 1-week post-MI for 7 days to pre-deplete the gut microbiota and started exercise training for another 8 weeks (Fig. S5A). As the maximal dose give to the mice, which was named ABX mice (ABX treated for 1 week as indicated in Fig. S5A), led to 42% mortality in MI mice, we also used a quarter of this dose (1/4 ABX group, 1/4 ABX treated for 9 weeks as indicated in Fig. S5A) that did not result in post-MI deaths ( Fig. S5B and C). As indicated in Fig. 3A, both ABX administration and 1/4 ABX administration impaired cardiac function of exercise MI mice, which is evidenced by the loss of beneficial effects of exercise training, including EF and FS (Fig. 3A), endurance capacity (Fig. 3B), the mRNA expression levels of ANP and BNP (Fig. 3C and D), cardiac fibrosis, cross-sectional cardiomyocyte area, and apoptosis ( Fig. 3E-H). Of note, ABX or 1/4 ABX had no significant effect on cardiac function in MI mice (Fig. S6). Thus, these data confirm that gut microbiota are involved in the protective effects of exercise training in cardiac dysfunction post-MI.
We also further confirmed the effects of ABX or 1/4 ABX in gut microbiota in MI by Pan analysis and alpha diversity, which showed that as compared to MI groups, ABX or 1/4 ABX decreased levels of gene enrichment ( Fig. S7A and B, Additional file 1: Table S8-S11). Based on the Community barplot analysis, the relative abundance of gut microbiota at the phylum level was changed by ABX or 1/4 ABX ( Fig. S7C and D, Additional file 1: Table S12-13). Moreover, the changes of genus ABX or 1/4 ABX in MI followed by running training were also showed in Fig. S8 and Additional file 1: Table S14.

Fecal microbial transplantation from exercise training post-MI is a transmissible trait
Many diseases' severities have been previously reported to be influenced by microbiota transplantation such as hypertension, metabolic syndrome, and MI [6,15,28]. Here, we sought to determine whether the beneficial effect of exercise training in MI is transmissible with fecal microbial transplantation by using fecal samples from exercised MI mice (Fig. 4A). Using echocardiographic analysis, we observed that mice receiving microbiota transplants from exercised-MI mice had better cardiac function as compared to those mice receiving microbiota transplants from MI mice (Fig. 4B). Although no significant difference of endurance capacity was observed, mice receiving microbiota transplants from exercised-MI mice had attenuated cardiac hypertrophy and fibrosis, decreased cross-sectional cardiomyocyte area, and apoptosis ( Fig. 4C-I).
We further investigated the effects of FMT in community richness and structure in MI mice treated with ABX. Pan analysis showed that mice received microbiota transplants from exercised-MI mice or non-exercised-MI mice could both increase community richness (Fig. S9A). Based on the community barplot analysis, the relative abundance of gut microbiota at the phylum level and genus level was changed by FMT ( Fig. S9B and C, Additional file 1: Table S15-16). We also used PCoA and NMDS analysis based on the relative abundance of genus to examine the community structures of the gut microbiotas, and found that the gut microbiota of mice received microbiota transplants from exercised-MI mice were separated from those mice received microbiota transplants from non-exercised-MI mice ( Fig. S9D and E). Moreover, the changes of genus by FMT were also showed in Fig. S10.

Metabolomics in exercise after MI
We performed MicrobioMET (Metabo-Profile, Shanghai, P.R. China) to analyze metabolomics in mice fecal samples from exercise post-MI. Overall, 116 metabolites including amino acids, fatty acids, indoles, organic acids, and phenols were quantitated using ultraperformance liquid chromatography coupled to tandem  Table S17). Metabolites may have similar functions because they are of the same nature. We divided 116 metabolites into 13 categories according to their nature. Relative abundance of metabolites category showed that phenols were increased in the MI + Run group compared to MI group ( Fig. 5A and B, Additional file 1: Table S18). 3-Hydroxyphenylacetic acid (3-HPA) and p-Hydroxyphenylacetic acid (4-HPA) are the main component of phenols. 3-HPA contents were decreased in MI group as compared to control group, and 4-HPA contents were increased in Sham + Run group compared to control group ( Fig. 5C and D, Additional file 1: Table S19 and S20). Consistently, 3-HPA was decreased in MI + control group as compared to Sham + control group, and 4-HPA was increased in MI + Run group as compared to MI + control group by volcano plot of OPLS-DA model (Fig. S11). These findings suggested 3-HPA and 4-HPA mediate beneficial effects of microbiota transplants from exercised MI mice.
To discriminate the metabolic profiles across groups, we performed clustering analyses based on orthogonal partial least square discriminant analysis (OPLS-DA). OPLS-DA was a partitioning of the X-data facilitates model interpretation and model prediction. The fecal samples from distinct groups were largely separated according to the OPLS-DA plots. Forty-two significant variable metabolites were identified between Sham + control group and MI + control group, while 37 significant variable metabolites were found between MI + control group and MI + Run group (Fig. S12, Additional file 1: Table S21, S22). Furthermore, we identified 4_Hydroxybenzoic acid (4-HBA) as a significant variable metabolite in both Sham + control and MI + Run group as compared to MI + control group (Additional file 1: Table S21, S22). These findings suggested 4-HBA might play a role in mediating beneficial effects of microbiota transplants from exercised MI mice.
To identify metabolic pathways that potentially play a role in the protective effects of exercise post-MI, we analyzed the pathway enrichment in both Sham + control and MI + Run group as compared to MI + control group, as well as Sham + Run group compared to Sham + control group. Interestingly, Alanine, aspartate, and glutamate metabolism pathway had a significant impact on MI + control vs. MI + Run ( Fig. 5E and F, Fig. S13, Additional file 1: Table S23-S28).
Collectively, these findings suggest that 3-HPA, 4-HPA, and 4-HBA may potentially contribute to the protective effects of exercise post-MI.

Two metabolites, 3-HPA and 4-HBA, protect cardiac dysfunction through activating NRF2
As 3-HPA, 4-HBA, and 4-HPA were indicated to potentially play a role in mediating beneficial effects of microbiota transplants from exercised MI mice, we checked whether their positive effects in cardiac function post-MI could be directly transferred. Every metabolite was tested at two doses, 6 mg/kg and 25 mg/kg. 3-HPA at both doses and 4-HBA at a dose of 25 mg/kg could significantly lead to improved EF and FS post-MI while 4-HPA failed to have beneficial effects (Fig. 6A). Besides, 3-HPA and 4-HBA at both doses could enhance endurance capacity in post-MI mice (Fig. 6B). Moreover, 3-HPA at both doses and 4-HBA at a dose of 25 mg/kg significantly attenuated cardiac hypertrophy and fibrosis, decreased crosssectional myocardium area, and apoptosis ( Fig. 6C-H). Therefore, supplementing these two metabolites, 3-HPA and 4-HBA, protects cardiac dysfunction post-MI.
We further checked the effect of 3-HPA and 4-HBA in vitro. As expected, 3-HPA and 4-HBA decreased NRCMs apoptosis induced by OGD/R as indicated by a reduction in TUNEL staining positive cardiomyocytes and the ratios of Bax/Bcl2 and cleaved Caspase3/ Caspase3 ( Fig. S14 and Fig. 7A-D). As NRF2 has been reported to be a target of phenols [29], and NRF2 could protect against MI and apoptosis induced by OGD/R [30,31], we further analyzed whether NRF2 mediated the protective effects of 3-HPA and 4-HBA on apoptosis. 3-HPA and 4-HBA increased the expression of NRF2, and inhibition of NRF2 prevented the protective effects of 3-HPA and 4-HBA on apoptosis (Fig. 7B, D-F). Together, our findings illustrate that 3-HPA and 4-HBA protect cardiomyocytes apoptosis by activating NRF2.

Discussion
There is a growing recognition of the contribution of gut microbiota to health and disease susceptibility. Herein, we described exercise training improved cardiac dysfunction and rescued gut microbial richness and community structure post-MI. Besides, we observed that two metabolites, 3-HPA and 4-HBA, played potential roles in mediating beneficial effects of exercise in MI mice through NRF2. Potential microbiota-altering therapies can lead to new approaches for preventing and managing cardiovascular diseases [32,33]. The gut microbiome is primarily composed of species within the Bacteroidetes, Firmicutes, Actinobacteria, Proteobacteria, and Cerrucomicrobia phylae [34]. Furthermore, Archaea, Verrucomicrobia, and Spirochaetes were reported in global gut metagenomes [35]. In the total bacterial species, the most abundant species are Bacteroidetes and Firmicutes [36]. Microbiota richness and diversity are important to maintain stability and performance and new biomarkers of health [37]. Loss of gut flora richness and biodiversity are associated with various diseases [3,38]. For example, hypertension patients dramatically decreased microbial richness and diversity [39]. In heart failure patients, a significantly decreased diversity of the intestinal microbiome has been observed [40]. Similarly, we found that MI decreased community richness of gut microbial based on Pan analysis. On genus level, Lactobacillus, Alistipes, and Turicibacter are more abundant in mouse gut microbiota [41]. In our study, we found that Lachnospiraceae_UCG-001 was increased in MI while Alistipes, Ruminococcus, Allobaculum, and Oscillospiraceae UCG-005 were decreased in MI. Lachnospiraceae has been reported to be higher in 7-day post-MI in rat, which is consistent with our data [42]. Allobaculum, an important healthy component of the mouse microbiome, was found to be decreased in obese mice or mice that lost circadian rhythm [43,44]. Here, we found that Allobaculum was decreased in MI and could be increased by running training in MI mice, which is consistent with the report that Allobaculum could be increased with exercise [45]. On the whole, our findings on gut microbiota in MI are supported by many hints in previous researches.
Physical exercise has been demonstrated to be able to modulate gut microbiota and increase the abundance of beneficial microbial species [46][47][48]. The microbiome of professional athletes exhibits higher richness compared to sedentary controls [49]. Consistently, we found that running treatments could increase community richness of gut microbial. In coronary heart disease, Phyla Bacteroidetes and Proteobacteria were decreased, whereas the Phyla Firmicutes were increased [50]. Interestingly, we found that running treatments could increase Bacteroidetes and reduce Firmicutes in MI + RNU group compared with MI group. Furthermore, we discovered that gut microbiota was responsible for the beneficial effects of exercise training in cardiac dysfunction post-MI, since orally antibiotics could prevent the beneficial effects of exercise training in cardiac fibrosis, cross-sectional area, and apoptosis. In human AMI patients, a higher microbial richness and diversity was found in the systemic microbiome of ST-segment elevation MI patients' blood [51]. However, our work illustrated the richness of gut microbiota was decreased 8-week post-MI but was increased by running training in MI. In addition, MI and running did not change the community diversity of gut microbial. We speculate that the content of intestinal flora is distinct in different stages of MI. Among Bacteroidetes, Alloprevotella was associated with decreased lifetime CVD risk. Among Firmicutes, Tyzzerella was enriched among those with high CVD risk profile [52,53]. But we expound no association between Alloprevotella, Tyzzerella_3, and cardiac function. MI was associated with a reorganization of the gut microbial community, such as a reduction in Lactobacillus. Supplementation of different species of Lactobacillus in rats showed cardioprotective effects when administered after MI [15]. In contrast, we found no association between Lactobacillus and cardiac function. Candida was correlated with heart failure severity [54]. But we did not observe this difference of Candida in MI mice. In general, our findings on the relationships between exercise, gut microbiota, and heart are consistent or inconsistent with previous studies. Further researches are needed to confirm the deeply link. Gut microbiome functions like an endocrine organ, generating bioactive metabolites such as short-chain fatty acids and bile acids that can impact the host physiology [55]. Diverse roles of the gut microbiota in human health and disease have been recognized [56]. Mounting evidence in mice and human is accumulating showing that gut microbiota is linked with cardiovascular health [57,58]. Metabolic therapy could confer benefit for treating heart failure [59]. Microbial transplantation studies could provide strong evidence to support the contribution of gut microbiota in host physiological processes and disease risks [60,61]. Metabolites play important roles in individual physiological stress, disease process, and drug development. The application of metabolites in cardiovascular disease is a rapidly expanding field [62]. Metabolite profiling in Finland, Southall, and Brent, British populations identified phenylalanine, monounsaturated fatty acids, and polyunsaturated fatty acids as biomarkers for cardiovascular risk [63]. Glycerophosphocholine metabolites modulate atherosclerosis and thus risk for cardiovascular disease [64]. Our work analyzed metabolomics in mice fecal samples from exercise after MI and screened 12 variable significant metabolites including 3-HPA and 4-HBA. 3-HPA is reported as a strong predictor of autism spectrum disorders [65]. Oral administration of 4-HBA decreased blood glucose levels in the diabetic rat by increasing glucose consumption [66]. Exercise induced alterations in metabolites such as alpha-tocopherol, myocardial high-energy phosphate metabolites in patients with Chagas heart disease [67,68]. However, the role of 3-HPA and 4-HBA in exercise and MI remains unknown. Herein, we discovered two metabolites, 3-HPA and 4-HBA, that can protect for cardiac dysfunction post-MI. In NRCMs, 3-HPA and 4-HBA could decrease apoptosis induced by OGD/R through activating NRF2. Further investigations are needed to find out the role of NRF2 in exercised MI mice.
In conclusion, our study revealed that gut microbiota mediated the protective effects of exercise after myocardial infarction based on the 16S rRNA sequencing analysis, oral antibiotics administration, and fecal microbial transplantation. We identified two metabolites from the fecal of exercised mice post-MI, 3-HPA and 4-HBA, protecting against cardiac dysfunction post-MI and apoptosis through NRF2. Our study provides insights into the composition of the gut microbiota in exercise and MI mice, enhances our understanding of the effect of 3-HPA and 4-HBA on MI, and launch approaches to modulate cardiovascular diseases by exercise, microbiome, and metabolites.

Conclusion
Together, our study provides new insights into the role of gut microbiome in exercise protection to MI, offers opportunities to modulate cardiovascular diseases by exercise, microbiome, and gut microbiota-derived 3-HPA and 4-HBA.
Additional file 1: Table S1. Data production of 49 samples in Sham+Control, MI+Control, Sham+Run, and MI+Run.  Table S26. Detailed information of metabolites metabolites pathways on network across Sham+Control and MI+Control. Table S27. Detailed information of pathways analysis across Sham+Control and Sham+Run. Table S28. Detailed information of metabolites metabolites pathways on network across Sham+Control and Sham+Run. Table S29. Detailed information of the relationship between metabolites and the top 50 genera Additional file 2: Figure S1. Relative abundance of gut microbiota in different groups, based on the phylum level. Figure S2. Relative abundance of gut microbiota based on the genus level among Sham+Control, MI+Control, Sham+Run, and MI+Run. Figure S3. Relative abundance of gut microbiota in different groups, based on the genus level. Figure S4. The relationship between the top 50 different genera and cardiac function index. Figure S5. MI mice displayed dose-dependent mortality after ABX treated. Figure S6. Gut microbiota pre-depletion by antibiotics does not affect cardiac function of MI mice. Figure S7. Gut microbiota pre-depletion by ABX decreased community richness and changed structure in mice after MI. Figure S8. Top 20 different genus across groups in the feces of mice after MI+Untreated+Run, MI+ABX+Run and MI+1/4ABX+Run. Figure S9. Fecal microbiota transplantation (FMT) increased community richness and changed structure in mice after MI+ABX. Figure S10. Different genus across groups in the feces of mice after FMT from MI+Control, FMI from MI+Run, or without FMT (PBS). Figure S11. 3-HPA and 4-HPA are identified by volcano plot of OPLS-DA model. Figure S12. 4-HBA is identified by OPLS-DA Model Discrimination. Figure S13. Pathway analysis bubble plot and relevant network. Figure S14. 3-HPA and 4-HBA decrease apoptosis with indicated duration and concentration.