Mice subjected to TAC surgery were used to generate a pressure overload-induced heart failure model, and the details of the surgery were similar to those in Rockman's study (11). Male C57BL/6J specific pathogen-free (SPF) mice (6–8 weeks old) weighing 20–25 g were purchased from JSJ Company (Grand Haven, MI, USA). All mice were housed in an SPF environment with a 12-h light/a12-h dark cycles and were fed sterile food (SWS9102, Xie Tong) and water ad libitum. Two weeks later, the animals were randomized into two groups for either TAC or sham surgery. Each month after surgery, cardiac function was assessed by a transthoracic echo-cardiographic examination and measurement of brain natriuretic peptide (BNP). The BNP level was measured with an ELISA kit (E-EL-MO2O4c, Elabscience). Cardiac fibrosis was determined by Masson's Trichrome Staining (G1340, Solarbio).

The animal experiment was approved by the Animal Ethics Committee of Shanghai Medical College, Fudan University.

To evaluate the role of the microbiota in blood glucose homeostasis, each month after the surgery, five mice were randomly selected from the TAC and the sham surgical groups and treated with antibiotic cocktails for 10 days (antibiotics were stopped for the last 3 days) to produce TAC germ-free (TAC-GF) mice and sham-GF mice, as shown in Figure 1A. To establish the GF mouse model, mice were given an antibiotic cocktail (MedChemExpress; containing 100 mg/ml neomycin, 50 mg/ml streptomycin, 100 mg/ml ampicillin, 50 mg/ml vancomycin, 100 mg/ml metronidazole, 1 mg/ml bacitracin, 125 mg/ml ciprofloxacin, and 100 mg/ml ceftazidime) in their drinking water.

Blood glucose was measured monthly using a glucose meter (Accu-Chek222, Roche) in five mice each from the TAC-GF and the sham-GF groups and the matched mice in the TAC and the sham groups. An intraperitoneal glucose tolerance test (IPGTT) was performed after 8 h of fasting. Mice were injected with 1 g/kg glucose, and blood was taken by tail blood sampling at 0, 15, 30, 60, and 120 min after glucose injection to measure the glucose levels. Fasting insulin and glucagon levels were measured by using ELISA kits (CEA448Mu, CEB266Mu, Cloud-Clone). These assays were carried out according to the manufacturer's instructions. Insulin resistance was estimated by homeostasis model assessment for insulin resistance (HOMA-IR), defined as [fasting glucose (mmol/L) × fasting insulin (mIU/L)/22.5].

The cecum and cecal contents were collected at the time of sacrifice. Intestinal mucins were detected by Alcian blue and periodic acid-Schiff (AB/PAS) staining (G1285, Solarbio). Following the manufacturer's instructions, AB staining was performed first (at pH 2.5) for 10–20 min, followed by periodic acid-Schiff staining. Fecal lactate levels were measured by a colorimetric assay kit (A019-2-2, Jiancheng). The fecal wet-to-dry ratio was defined as (wet weight–dry weight)/dry weight. Fecal noradrenaline and adrenaline levels were measured by targeted high-performance liquid chromatography mass spectrometry (HPLC-MS; AB ExionLC, AB Sciex Qtrap 6500+, AB Sciex).

Fresh fecal samples from the donor mice were suspended in phosphate-buffered saline (PBS) buffer after 10 min of vortexing (1 ml of PBS was added per 100 mg of feces) and were then centrifuged for 1 min at 500 × g. The supernatant was pipetted for gavage, and all procedures were performed within 4 h. The feces used for transplantation were obtained from the mice in the surgery group in the third month. Twenty-four conventional GF mice were prepared for transplantation and were randomized to receive FMT from either the TAC (FMT-TAC) or the TAC-GF (FMT-TG) group or the sham (FMT-sham) or the sham-GF (FMT-SG) group in a one-to-one correspondence (i.e., one donor to one recipient). Each GF mouse was administered 200 μl of fecal sample suspension by oral gavage two times per week for the first 2 weeks and weekly thereafter for a total of 4 weeks, and the colonization was confirmed by 16S rRNA sequencing.

The fecal samples of the remaining mice in the sham and the TAC groups were collected and stored at −80°C for 16S rRNA sequencing. The gut microbiota DNA concentration and integrity were determined with a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis, respectively. PCR amplification of the V3–V4 hypervariable regions of the bacterial 16S rRNA gene was carried out in a 25-μl reaction using a universal primer pair (343F: 5′-TACGGRAGGCAGCAG-3′; 798R: 5′-AGGGTATCTAATCCT-3′). Sequencing was performed on an Illumina MiSeq (Illumina Inc., San Diego, CA, USA) with two paired-end read cycles of 300 bases each.

The FASTQ format raw data were preprocessed using Trimmomatic software. Sequences were further denoised with QIIME software (version 1.8.0). Clean reads were subjected to primer sequence removal and clustering to generate operational taxonomic units (OTUs) using VSEARCH software with a 97% similarity cutoff. The representative read of each OTU was selected using the QIIME package. All representative reads were annotated and blasted against the Silva database version 138 (16S/18S rDNA) using the Ribosomal Database Project (RDP) classifier. Principal component analysis (PCA) and principal coordinate analysis (PCoA) were conducted by using the R language. To characterize microbial functions, we annotated all of the genes in our catalog to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database.

Untargeted metabolomics profiling of fecal and serum samples was performed. Briefly, 40 mg of each fecal sample from the surgical mice or 80 μl of each serum sample from the FMT mice was extracted with methanol before lyophilization. Then, 15 mg/ml of methoxylamine hydrochloride in pyridine was subsequently added and incubated at 37°C for 90 min. Bis (trimethylsilyl)trifluoroacetamide (BSTFA; with 1% trimethylchlorosilane) and n-hexane were added to the mixture, which was derivatized at 70°C for 60 min. The samples were placed at ambient temperature for 30 min. The derivatized samples were analyzed on an Agilent 7890B Gas Chromatography System coupled to an Agilent 5977AMSD System (Agilent Technologies Inc, CA, USA).

The metabolomics data were converted by Abf Converter and analyzed by MS-DIAL. Metabolite characterization was based on the LUG database (Untargeted database from Lumingbio). In each sample, all peak signal intensities were segmented and normalized according to the internal standards with relative standard deviations (SDs) >0.3 after screening to obtain the data matrix. The matrix was imported into R to carry out a partial least-squares-discriminant analysis (PLS-DA) to distinguish the metabolites that differed between groups. To prevent overfitting, 7-fold cross-validation and 200 response permutation testing (RPT) were used to evaluate the quality of the model. A two-tailed Student's t-test was further used to verify whether the differences in metabolites between groups were significant. Metabolites with variable importance in the projection (VIP) values >1.0 and p-values <0.05 were selected as differential metabolites. Allocation of the metabolites within designated pathways was supported by KEGG.

Statistical analysis was performed using SPSS (version 25) software. Continuous variables are presented as the mean ± standard error of the mean (SEM). Two-group comparisons were performed using Student's non-paired t-test or the Wilcoxon (Mann-Whitney U) test. Variables with more than two groups were analyzed by a one-way ANOVA. The statistics regarding the microbiome subsection analyses are described in the gut microbiota profiling section of “Methods.” Correlations were calculated with the Spearman rank correlation coefficient. A level of p < 0.05 was considered statistically significant.

During the 3 months after surgery, the heart function of the TAC group decreased gradually, as shown by transthoracic echocardiography and BNP levels. At the end of the third month, all of the TAC group mice had an ejection fraction (EF) <50% and significantly increased perivascular fibrosis, left ventricular posterior wall dimensions at end-diastole (LVPWd), left ventricular internal diameter diastole (LVIDd), BNP, and heart weight (HW) (Figures 1B–H); however, the body weight (BW) of the mice was not altered significantly (Figure 1I), and antibiotics did not affect these parameters (Supplementary Figure 1). Fasting and random blood glucose, insulin, glucagon, and IPGTT levels did not show any differences between these groups in the first 2 months (Figures 2A,B), whether the microbiota was cleared or not, but in the third month after surgery, the IPGTTs of the TAC-GF mice revealed decreased glucose levels at the 15-, 30-, and 60-min time points, which did not occur in the other groups (Figure 2C). Significant decreases in the area under the curve (AUC), insulin levels, glucagon levels, and HOMA-IR were also observed in the TAC-GF group. Antibiotics did not affect blood glucose, insulin, or glucagon levels in the other groups (Figures 2D–G).

Thoracic aortic constriction surgery changed intestinal and fecal functions. A decrease in intestinal acidic mucin levels was observed in the TAC group (Figures 3A,B). The feces of the TAC group had higher water and lactate contents and a lower pH value (Figures 3C–E). Antibiotics did not influence the intestinal acidic mucin levels, fecal water content, or pH value; however, they decreased the lactate levels in the feces in both the TAC-GF and the sham-GF groups (Supplementary Figure 2).

Three months after surgery, fecal samples from 14 TAC and 8 sham group mice were successfully collected, and 16S rRNA sequencing showed that the gut microbiota was changed after TAC surgery. The numbers of OTUs were distributed between 2,193 and 2,774 in the 22 samples. Firmicutes and Bacteroidetes were the most abundant phyla in both groups, but a higher Firmicutes to Bacteroidetes ratio (F/B) was observed in the TAC group than in the sham group (Figures 4A,B). PCA revealed that the microbiota of the two groups was completely separated from each other (Figure 4C). Among the 43 most abundant differential bacteria at the genus level, the abundances of 22 were increased in the TAC group that included Helicobacter, Rikenellaceae, Colidextribacter, and Lactobacillus, while those of Prevotellaceae, Muribaculum, and Rikenella were decreased (Figure 4D). Coabundance analysis showed that Muribaculum was negatively correlated with Lactobacillus, Lachnospiraceae, Rikenellaceae, and Eubacterium, while Parasutterella was positively correlated with Bifidobacterium and Prevotellaceae (Figure 4E). The heart function data, such as HW/BW, LVPWd, LVIDd, and BNP levels, were positively correlated with Lactobacillus, Oscillibacter, Rikenellaceae, Lachnospiraceae, and Tuzzerella. The pH value was negatively correlated with Lactobacillus and Lachnospiraceae (Figure 5).

Partial least-squares-discriminant analysis indicated that the metabolites of the gut microbiota were significantly different between the TAC and the sham groups (Figure 6A). KEGG pathway enrichment showed that the metabolic alterations were enhanced in the regulation of lipolysis in adipocytes, the cyclic adenosine 3,5-monophosphate (cAMP) signaling pathway, and purine metabolism (Figure 6B). A total of 49 differential metabolites were identified. Noradrenaline and adrenaline were upregulated in the TAC group. Correlation analysis showed that noradrenaline and adrenaline were positively related to Lactobacillus and Oscillibacter (Figure 6C).

Since we found positive results in the third month after surgery and in the GF mice, we collected feces from all four groups of mice for fecal transplantation (Figure 7A). After transplantation of the TAC mouse fecal samples, the IPGTT-AUC, insulin level, glucagon level, and HOMA-IR of the FMT-TAC mice were all higher than those of the FMT-sham group mice, while the FMT mice treated with TAC-GF mouse feces (FMT-TG) and sham-GF mouse feces (FMT-SG) were not different from each other (Figures 7B–F). 16S rRNA sequencing suggested that there was a clear separation between the FMT-TAC group and the FMT-sham group, and the microbiota was similar between the donor and recipient mice (Figures 7G,H). FMT did not change the function of the heart (Supplementary Figure 3).

The serum metabolism of the FMT mice showed that there were 47 metabolites enriched in the FMT-TAC group, many of which were involved in gluconeogenesis, such as beta-alanine, succinic acid, alpha-ketoglutarate, malic acid, and dihydroxyacetone phosphate (Figure 8A). The KEGG pathway enrichment analysis showed that the glucagon pathway, citrate cycle, and tyrosine metabolism were significantly enriched (Figure 8B). When comparing the serum and fecal metabolites, we found that 46 products were upregulated between fecal and serum samples, such as noradrenaline, L-glutamic acid, and serotonin. The KEGG pathway analysis indicated that these products in serum were positively correlated with the cAMP signaling pathway, pentose and glucuronate interconversions, and purine metabolism (Supplementary Figure 4).

Noradrenaline and adrenaline are normal hormones whose levels are increased during heart failure, and they may penetrate through the intestine into the feces (12). To determine the origin of noradrenaline and adrenaline, we measured noradrenaline and adrenaline in the fecal samples of TAC and TAC-GF mice. The results showed that noradrenaline and adrenaline decreased significantly in the TAC-GF mouse fecal samples, and the same result was shown in FMT-TG mice (Figures 8C–F). The correlation analysis showed that the noradrenaline level was positively correlated with the host IPGTT-AUC, glucagon level, and HOMA-IR (Figures 8G–J).