The murine PCa cell line RM-1 was obtained from Shanghai Zhong Qiao Xin Zhou Biotechnology Co., Ltd. and authenticated by STR profiling. RM-1 cells were cultured in RPMI-1640 medium (Gibco, United States) supplemented with 10% fetal bovine serum (FBS) (Gibco, United States), 1% penicillin, and 1% streptomycin (Gibco, United States). The cell line was maintained in a humidified incubator at 37°C with 5% CO₂.

All C57 mice were obtained from Beijing Vital River Laboratory Animal Technology Co., Ltd. All animal procedures and experimental protocols were conducted in accordance with the Guide for the Care and Use of Laboratory Animals and approved by the Laboratory Animal Welfare and Ethics Committee of Army Medical University, China (Approval No.: AMUWEC2020186). Every effort was made to minimize animal suffering and reduce the number of animals used. Male C57 mice (6 weeks old) were used in the experiments. The mice were randomly assigned to the Ctrl, CH-ABX, and CH-Ctrl groups (n = 10 per group) and housed individually under standard 12-h light/dark conditions. Mice in the CH-ABX group received antibiotic treatment for 2 weeks prior to the experiment via drinking water containing 0.2 g/L ampicillin, neomycin, metronidazole, and 0.1 g/L vancomycin daily. Approximately 2 × 10⁶ RM-1 cells were subcutaneously injected into the inner thigh of each mouse. On day 7 after establishing the tumor-bearing mouse model, CH (Chengdu MUST Bio-technology Co., Ltd, China) was dissolved in water and administered daily by gavage at 0.2 mg CH per mouse in the CH-ABX and CH-Ctrl groups, while the Ctrl group received water alone. After 4 weeks, the mice were weighed and euthanized via sodium pentobarbital injection. Subcutaneous tumors and intestinal content samples were collected, and tumor weight was measured.

DNA was extracted from samples in the Ctrl, CH-ABX, and CH-Ctrl groups using a fecal DNA extraction kit (DP712, Tiangen Company, Beijing, China), with 10 replicates for each. Quality assessment of the DNA was done on 1% agarose gels. The V3-V4 region of the 16S-rDNA gene was targeted for amplification via PCR with specific primers: 341F: CCTACGGGNGGCWGCAG; 806R: GGACTACHVGGGTATCTAAT, each with a unique eight-base barcode sequence. PCR was conducted in 30 μL volumes containing 15 μL of Phusion^® High-Fidelity PCR Master Mix, 0.2 μM of primers, and 10 ng of DNA template. The thermal profile included an initial denaturation at 98°C for 1 min, followed by 30 cycles of 98°C for 10 s, 50°C for 30 s, and 72°C for 30 s, concluding with a final extension at 72°C for 5 min. PCR products were visualized on 2% agarose gels and purified using the GeneJET Gel Extraction Kit. Sequencing libraries were prepared with the Illumina TruSeq DNA PCR-Free Library Preparation Kit and indexed. Sequencing was performed on the Illumina Hiseq platform by Genecloud Co. Ltd, Chongqing, China. QIIME software package 2 (version 2020.2) was used for sequence analysis. Operational taxonomic units (OTUs) were determined at a similarity threshold of ≥97% using the MOTHUR pipeline, with the most abundant tag sequence chosen as representative. Alpha-diversity metrics such as Observed OTU, Shannon, and Simpson indices were calculated with the “vegan” R package. Beta-diversity was analyzed using Principal coordinates analysis (PCoA) and Partial Least Squares Discriminant Analysis (PLS-DA) through R. Taxonomic comparisons at phylum and genus levels were made using the Wilcoxon rank-sum test. LEfSe method was employed for differential abundance analysis of bacterial communities, with Kruskal–Wallis rank-sum test for significance (padj <0.05) and LDA score ≥2.5 indicating substantial effect size. Metabolic pathways were analyzed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database, considering pathways enriched when padj <0.05. To enhance statistical robustness and reduce the probability of Type I errors, a false discovery rate (FDR) correction was applied. Correlation between microbial flora and metabolites was assessed with Pearson’s correlation coefficient via R’s mixOmics package.

Prior to metabolite extraction, samples were weighed and lyophilized. They were then ground in a 2 mL Eppendorf tube with a 5 mm tungsten bead for 60 s at 65 Hz using a Grinding Mill. The extraction process involved a 1 mL mixture of methanol, acetonitrile, and water (in a ratio of 2:2:1), followed by an hour of ultrasonication in an ice bath. Afterward, the samples were chilled at −20°C for an hour and centrifuged at 14,000 g for 20 min at 4°C. The supernatants were collected and evaporated to dryness under vacuum conditions. Metabolomics profiling was conducted using a UPLC-ESI-Q-Orbitrap-MS system, which included a UHPLC Shimadzu Nexera X2 LC-30AD from Shimadzu, Japan, and a Q-Exactive Plus from Thermo Scientific, San Jose, United States. For LC separation, an ACQUITY UPLC^® HSS T3 column (2.1 × 100 mm, 1.8 μm) from Waters, Milford, MA, United States was utilized. The flow rate was set at 0.3 mL/min with a mobile phase consisting of A: 0.1% FA in water and B: 100% acetonitrile (ACN). The gradient elution started at 0% buffer B for 2 min, increased linearly to 48% over 4 min, then to 100% in another 4 min, maintained this level for 2 min, and finally decreased back to 0% buffer B in 0.1 min with a 3-min reequilibration period. ESI in both positive and negative modes was used for MS data collection. The HESI source parameters were as follows: Spray Voltage of 3.8 kv (positive) and 3.2 kv (negative); Capillary Temperature of 320°C; Sheath Gas (nitrogen) flow at 30 arb; Aux Gas flow at five arb; Probe Heater Temp at 350°C; S-Lens RF Level at 50. The mass range for full MS scans was set from 70 to 1050 Da with a resolution of 70,000 at m/z 200 and 17,500 for MS/MS scans. The injection time was set to 100 ms for MS and 50 ms for MS/MS. The isolation window for MS2 was set to 2 m/z with collision energies of 20, 30, and 40 for fragmentation. Raw MS data were processed using MS-DIAL for peak alignment and area extraction. R (version: 4.0.3) and its packages were employed for multivariate data analysis and modeling with Pareto scaling for mean-centering. Principal component analysis (PCA) was used to construct models, and VIP scores were calculated to assess variable significance with values over 1 considered significant. Metabolites with VIP values above 1.0 and p values below 0.05 were deemed statistically significant. Fold change was determined by calculating the logarithm of the average mass response ratio between two classes. Differential metabolites were subjected to cluster analysis using R packages. Perturbed biological pathways were identified through KEGG pathway analysis with differential metabolite data using the KEGG database and Fisher’s exact test with FDR correction for multiple testing. To enhance statistical robustness and reduce the probability of Type I errors, a FDR correction was applied. Pathways with padj <0.05 were considered statistically significant after enrichment analysis.

Total RNA was extracted utilizing the TRIzol reagent (Invitrogen, CA, United States) following the manufacturer’s guidelines. Subsequently, libraries were prepared with the VAHTS Universal V6 RNA-seq Library Prep Kit as directed by the manufacturer. Transcriptome sequencing and analysis were executed by OBiO Technology Corp., Ltd. (Shanghai, China). Sequencing was performed on the Illumina Novaseq 6000 platform, yielding 150 bp paired-end reads. Initial processing of raw reads was conducted using fastp (v0.22.0) to eliminate low-quality sequences and retain clean reads. These clean reads were then aligned to the reference genome with HISAT2 (v2.1.0). Gene expression levels were quantified by calculating FPKM values, and read counts for each gene were obtained through feature Counts. PCA was performed on samples using standardized expression values to reduce data complexity, thoroughly explore relationships among samples and variation magnitude, and assess sample reproducibility. To enhance statistical robustness and reduce the probability of Type I errors, a FDR correction was applied. Differential expression analysis was conducted using DESeq2, with significance defined as padj <0.05. To illustrate gene expression patterns across different groups and samples, hierarchical cluster analysis of differentially expressed genes (DEGs) was performed using R (v3.2.0). Additionally, GO and KEGG pathway enrichment analyses of DEGs were conducted based on the hypergeometric distribution to identify significantly enriched terms, utilizing R (v3.2.0) for each analysis.

Spearman correlation coefficients were used to perform pairwise correlation analyses among differential bacterial genera, genes, and metabolites, and a correlation network was constructed to elucidate the relationships between differential bacterial genera, metabolites, and genes. We applied FDR correction to minimize the likelihood of Type I errors to the greatest extent possible.

The data are presented as the mean ± standard deviation (SD). We employed SPSS software (SPSS 26.0, IBM Corporation, United States) for statistical analyses and GraphPad Prism (version 8.0; GraphPad Software, LLC, United States) to create graphs. To compare the data before and after treatment, paired t - tests were carried out. For comparisons among multiple groups, one - way ANOVA was utilized. When it came to further comparing two groups, the least significant difference method was applied. For non - normally distributed data, the Mann - Whitney U test was used to evaluate group differences. Statistical significance was defined as a P - value of less than 0.05.

The body weight and tumor weight of mice in the Ctrl, CH-Ctrl, and CH-ABX groups were measured and grossly examined. We found no significant differences in body weight among the groups (Figure 1A). However, the tumor weight in the CH-Ctrl group was significantly lower compared to both the Ctrl and CH-ABX groups. Although the CH-ABX group showed a decreasing trend in tumor weight relative to the Ctrl group, the difference was not statistically significant (Figures 1B,C). These results suggest that the gut microbiota may mediate the inhibitory effect of CH on PCa growth.

We next performed 16S rDNA amplicon sequencing analysis on the gut microbiota of each group. The results showed no significant differences in Shannon and Simpson α-diversity indices between the Ctrl and CH-Ctrl groups (Figures 2A,B). PLS-DA analysis revealed clear inter-group separation and intra-group clustering between the Ctrl and CH-Ctrl groups (Figure 2C). At the phylum level (relative abundance >0.1%) and genus level (relative abundance >1%), the overall structure of gut microbiota showed minimal changes between the two groups, though significant alterations were observed in the relative abundance of specific phyla and genera (Figures 2D,E). Specifically, compared with the Ctrl group, the CH-Ctrl group exhibited decreased relative abundance of p_Bacteroidetes but increased relative abundance of p_Deferribacteres, p_Epsilonbacteraeota, and p_Proteobacteria at the phylum level (Figure 2D). At the genus level, the CH-Ctrl group showed reduced relative abundance of g_Alistipes, g_Bacteroides, and g_Lachnospiraceae_NK4A136_group, while demonstrating increased relative abundance of g_Blautia, g_Helicobacter, g_Mucispirillum, g_Ruminiclostridium, and g_Ruminiclostridium_9 compared to the Ctrl group (Figure 2E). Linear discriminant analysis (LDA) with an effect size threshold (LDA score >2.5) identified significantly different microbial biomarkers between the two groups (Figure 2F). Subsequent KEGG pathway analysis of gut microbiota revealed 23 significantly different metabolic pathways between the groups (Figure 2G). Finally, correlation analysis between the significantly different gut microbiota and KEGG metabolic pathways in Ctrl and CH-Ctrl groups generated a positive/negative correlation map (Figure 2H).

Comparative analysis between CH-Ctrl and CH-ABX groups revealed increased Shannon and Simpson α-diversity indices in the CH-Ctrl group, though without statistical significance (Figures 3A,B). Principal Coordinates Analysis (PCoA) demonstrated distinct inter-group separation and intra-group clustering between CH-ABX and CH-Ctrl groups (Figure 3C). Significant structural alterations were observed at both phylum (relative abundance >0.1%) and genus (relative abundance >1%) levels, accompanied by marked changes in relative abundance of specific taxa (Figures 3D,E). At the phylum level, CH-Ctrl group exhibited reduced or undetectable levels of p_Bacteroidetes, p_Deferribacteres, and p_Verrucomicrobia, while showing increased relative abundance of p_Epsilonbacteraeota and p_Firmicutes compared to CH-ABX group (Figure 3D). Genus-level analysis demonstrated decreased or absent g_Akkermansia, g_Bacteroides, g_Lachnoclostridium, g_Mucispirillum, and g_Prevotellaceae_UGG_001, contrasted with elevated levels of g_Blautia, g_Helicobacter, g_Oscillibacter, g_Ruminiclostridium, and g_Ruminiclostridium_9 in CH-Ctrl group (Figure 3E). Linear discriminant analysis (LDA) with an effect size threshold (LDA score >2.5) identified significantly differential microbial biomarkers between the groups (Figure 3F). KEGG pathway analysis revealed 19 significantly different metabolic pathways (Figure 3G). Correlation analysis between these differential gut microbiota and metabolic pathways generated a comprehensive positive/negative correlation network (Figure 3H).

Comprehensive analysis integrating the relative abundance trends of differential microbiota between CH-Ctrl vs Ctrl and CH-Ctrl vs CH-ABX groups suggests that CH may influence PCa progression by increasing the relative abundance of g_Blautia, g_Lactobacillus, and g_Butyricicoccus while decreasing g_Akkermansia abundance.

Our analysis of untargeted metabolomics in intestinal contents from PCa mice across groups revealed that the CH-Ctrl group showed distinct separation from the Ctrl group along the PC1 axis (21.49%) (Figure 4A) and from the CH-ABX group along the PC2 axis (19.51%) (Figure 4B). Volcano plot analysis identified 240 and 123 significantly differential metabolites when comparing CH-Ctrl with Ctrl and CH-ABX groups, respectively (Figures 4E,F). Heatmap clustering of the top 50 differential metabolites demonstrated that, compared to the Ctrl group, the CH-Ctrl group exhibited 24 significantly upregulated and 26 downregulated metabolites (Figure 4C). Similarly, relative to the CH-ABX group, there were 22 upregulated and 28 downregulated metabolites in the CH-Ctrl group (Figure 4D). The metabolites consistently upregulated in CH-Ctrl compared to both Ctrl and CH-ABX groups were ZM323881, acetylglycine, PC(17:0-20:3)-d5_ISTD, 6-aminododecanedioic acid, and chloride. Conversely, PC (18:1/18:2) and 2-naphthalenesulfonic acid were consistently downregulated. KEGG pathway enrichment analysis of these differential metabolites identified choline metabolism in cancer as the common significantly enriched pathway between CH-Ctrl versus both Ctrl and CH-ABX comparisons (Figures 4G, H).

Simultaneously, we performed transcriptome sequencing analysis on PCa tissues from the mice. Overall, the CH-Ctrl group exhibited 166 significantly upregulated and 413 downregulated genes compared to the Ctrl group (Figure 5A), while showing 312 upregulated and 218 downregulated genes relative to the CH-ABX group (Figure 6A). KEGG pathway enrichment analysis of the differentially expressed genes revealed that, compared to both Ctrl and CH-ABX groups, the CH-Ctrl group showed concurrent significant enrichment in the top 20 KEGG pathways for both increased pathways (Cell adhesion molecules and Neuroactive ligand-receptor interaction; Figures 5B, 6B) and decreased pathways (Staphylococcus aureus infection and Arrhythmogenic right ventricular cardiomyopathy; Figures 5C, 6C). Subsequent GO functional enrichment analysis demonstrated that, relative to both Ctrl and CH-ABX groups, the CH-Ctrl group displayed simultaneous significant enrichment in increased GO terms (nervous system process; Figures 5D, 6D) and decreased GO terms (epidermis development, cell periphery, extracellular region, and extracellular space; Figures 5E, 6E).

Through comprehensive correlation analysis of significantly altered gut microbiota, intestinal metabolites, and transcriptome data - combined with our previous findings - we identified key differential gut microbes, metabolites, and genes potentially involved in CH-mediated suppression of PCa growth. CH treatment significantly increased the relative abundance of g_Blautia, g_Lactobacillus, and g_Butyricicoccus while inhibiting g_Akkermansia proliferation. This microbial shift enhanced Acetylglycine production, consequently upregulating Ttpa, Gm14964, Shc3, and Elovl4 gene expression while downregulating Gm10531 and Bc021767 (Figure 7).