To investigate the relationship between low-dose chronic cadmium exposure and antibiotic efficacy in vivo, we established a cadmium exposure mouse model by adding 1 mg/L CdCl₂ to the drinking water. Mice in both the control and cadmium (Cd)-exposure groups were administered ciprofloxacin 2 h following infection with Staphylococcus aureus (S. aureus) Newman, while the untreated group received phosphate-buffered saline (PBS). Twenty-four-hour post-infection, lungs, livers and kidneys were harvested for bacterial load assessment (Figure 1A). After 12 weeks of exposure, cadmium accumulation in various organs was quantified using inductively coupled plasma mass spectrometry (ICP-MS). As depicted in Figure 1B, cadmium concentrations significantly increased in all tested organs, with the highest accumulation observed in the lungs, livers and kidneys, showing increases of 16.1, 12.3, and 8.7-fold, respectively.

Following the successful establishment of the cadmium exposure mouse model, we assessed whether cadmium exposure leads to a reduction in antibiotic efficacy. Prior to evaluating the impact of cadmium exposure on antibiotic efficacy, we observed that S. aureus Newman colonization was significantly enhanced under specific heavy metal stress conditions, indicating increased host susceptibility to bacterial infections. Subsequently, we assessed the reduction of bacterial loads in the three organs with and without ciprofloxacin treatment to determine therapeutic efficacy. In the control group, a single dose of ciprofloxacin resulted in an approximately 100-fold reduction in bacterial loads within the organs. In contrast, the Cd-exposure group exhibited minimal reduction following antibiotic administration, suggesting diminished antibiotic efficacy (Figures 1C–E).

We then aimed to investigate the underlying mechanisms contributing to the observed antibiotic failure in the Cd-exposure physiological environment. Given the established role of contaminated food and water as significant sources of cadmium exposure (Zhao et al., 2025), and recognizing that the gut microbiota serves as the primary interface for orally ingested substances (Zhang et al., 2020), it is highly probable that cadmium exposure significantly alters the composition of the gut microbiota. Previous studies have demonstrated that acute high-dose cadmium exposure in mouse models disrupts beneficial gut microbiota while promoting harmful bacterial growth (Garg et al., 2025; Nehzomi and Shirani, 2025). To assess whether the disruption of gut microbiota by cadmium exposure critically impairs antibiotic efficacy, a microbiota transplantation experiment was conducted as previously described (Tan et al., 2022).

Initially, the native gut microbiota of the transplantation group mice was depleted by administering a cocktail of antibiotics in their drinking water for 1 week. Fecal samples from Cd-exposure mice were collected, and the extracts were continuously added to the drinking water of the transplantation group for 2 weeks, followed by a return to normal drinking water for the subsequent 9 weeks (Figure 1B). Thereafter, the efficacy of antibiotics was evaluated post-gut microbiota transplantation. In our study, we observed that, in contrast to direct cadmium exposure, microbiota transplantation from Cd-exposure mice did not result in an increased bacterial burden compared to the control group. This suggests that the heightened susceptibility to bacterial infection following cadmium exposure is not attributable to gut microbiota dysbiosis. However, similar to the Cd-exposure mice, a significant reduction in antibiotic efficacy was observed post-microbiota transplantation (Figures 1C–E). These findings led us to hypothesize that cadmium exposure alters the gut microbiota composition, resulting in significant changes in specific metabolites, which ultimately contribute to antibiotic treatment failure.

To investigate the mechanisms by which chronic cadmium exposure diminishes antibiotic efficacy, we conducted 16S rDNA sequencing to assess alterations in gut microbiota composition following cadmium exposure. Our analysis of the microbial community composition revealed a reduction in species diversity within the intestinal flora of Cd-exposure mice (Figure 2A). Additionally, we utilized the Chao and Shannon index to quantitatively compare microbiota diversity between the two groups. In a similar vein, both index exhibited a significant reduction following cadmium exposure (Figures 2B,C). Subsequently, we conducted an in-depth analysis to investigate the differences in the abundance of core and beneficial commensals between the two groups at the genus level. Among the top 20 bacterial genera, the abundance of eight genera, comprising both core and beneficial commensals, was markedly decreased. These included five beneficial gut microbiota genera—Clostridia UCG-014, Alistipes, Prevotellaceae UCG-001, Ruminococcus, and Butyricicoccus—and three core commensal genera—Bacteroides, Parabacteroides, and Alloprevotella (Figure 2D). This suggests that chronic cadmium exposure adversely affects the survival of the aforementioned beneficial bacterial genera. The observed decrease in genera such as Clostridia UCG-014 and Alistipes aligns with previous studies that associated cadmium exposure with gut dysbiosis, highlighting the essential role of these taxa in preserving gut barrier integrity and facilitating short-chain fatty acid production (Garg et al., 2025). Furthermore, the reduction in Butyricicoccus, a recognized butyrate producer, parallels findings from models of dysbiosis induced by heavy metals, where the depletion of butyrate was found to exacerbate metabolic and immune dysfunction (Nehzomi and Shirani, 2025). Collectively, long-term low-dose cadmium exposure in mice resulted in a diminished variability of gut microbiota species diversity, particularly impacting key beneficial bacteria.

Given that we have demonstrated the disruption of multiple core and beneficial commensals and the consequent alteration in gut microbiota balance post-cadmium exposure, we further investigated the associated changes in endogenous metabolites. An untargeted metabolomics analysis was performed to investigate changes in metabolite profiles in serum samples from control and cadmium-exposed mice following a 12-week feeding period. The results indicated significant variations in the levels of 1,500 metabolites, with 742 metabolites being upregulated and 758 downregulated (Figure 3A). Subsequent KEGG pathway enrichment analysis of the differentially expressed metabolites revealed a predominant enrichment in pathways related to organic acid and lipid metabolism, including sphingolipid metabolism, linoleic acid metabolism, glycine, serine, and threonine metabolism, glyoxylate and dicarboxylate metabolism, and carbon metabolism (Figure 3B). Additionally, the top 60 metabolites were identified and subjected to clustering analysis, comprising 22 upregulated and 38 downregulated metabolites (Figure 3C). Notably, several key organic-acid-related compounds, such as 2-hydroxybutyric acid (Xie et al., 2025) and DL-mevalonolactone (Yogev et al., 2023), were identified, which have been previously reported as regulatory factors in various diseases and disorders. Moreover, the classification of differential metabolites revealed a significant emphasis on lipids and organic acids, with 25% identified as lipids and lipid-like compounds and 21.67% as organic acids and derivatives (Figure 3D).

To investigate the relationship between alterations in gut microbiota composition and circulating metabolite levels, we performed a correlation analysis using the Spearman correlation algorithm. A heatmap illustrating the correlation coefficient matrix between the top 15 gut microbiota and metabolites is presented (Figure 4). As previously noted, the top 15 gut microbiota primarily consisted of various core and beneficial commensals, while the top 15 metabolites were predominantly organic-acid-related compounds and lipid-like molecules. Notably, we observed a significant reduction in the circulating level of DL-mevalonolactone in mice exposed to cadmium compared to control mice. The alteration in serum DL-mevalonolactone content exhibited a strong positive correlation with changes in the abundance of six core and beneficial commensal bacteria, including Butyricicoccus, Clostridia UCG-014, Alistipes, Ruminococcus, Parabacteroides, and Alloprevotella. This suggests that the significant reduction in serum DL-mevalonolactone levels following chronic cadmium exposure is closely associated with alterations in gut microbiota composition, characterized by a notable decrease in the species diversity of core and beneficial commensals. The observed positive correlation between Butyricicoccus and DL-mevalonolactone levels is corroborated by previous research indicating that butyrate-producing bacteria influenced host mevalonate pathway metabolites, which play a critical role in lipid metabolism and immune regulation (LeBlanc et al., 2013). Furthermore, the involvement of Parabacteroides in detoxification pathways (Dubey et al., 2019) may elucidate its association with DL-mevalonolactone, as both are involved in alleviating oxidative stress caused by cadmium exposure. Additionally, myo-inositol and N-acetylglutamine also demonstrated significant correlations with alterations in the gut microbiota, hence the preliminary screening of these candidates was conducted. Despite several metabolites exhibiting noteworthy associations, subsequent functional validation indicated that myo-inositol and N-acetylglutamine did not display measurable biological activity, while the effect of DL-mevalonolactone was remarkable (Supplementary Figure S1). Consequently, our subsequent mechanistic investigations were specifically focused on DL-mevalonolactone. Previous studies have identified DL-mevalonolactone as a δ-lactone form of mevalonate, serving as a precursor in the mevalonate pathway (Soto et al., 2011). Oral administration of DL-mevalonolactone has demonstrated efficacy against HMGCR mutation and statin-induced myopathy (Yogev et al., 2023). However, its regulatory effect on bacterial infections has not been documented, and further investigation is required to determine whether the reduced DL-mevalonolactone content is a primary factor contributing to diminished antibiotic efficacy. Furthermore, it remains to be investigated whether this endogenous target metabolite possesses direct antimicrobial activity or contributes to the inhibition of antibiotic tolerance, thereby reducing the formation of bacterial persisters when used in combination with antibiotics. Overall, our findings suggest that administering chronic low-dose cadmium to mice via drinking water may impact the mevalonate pathway and disrupt the biosynthesis of DL-mevalonolactone, potentially indicating an association with diminished antibiotic activity.

The aforementioned multi-omics analysis suggests that DL-mevalonolactone may play a previously unrecognized role in maintaining antibiotic efficacy. To explore the underlying mechanisms, we conducted further investigations. Initially, in vitro minimum inhibitory concentration (MIC) analysis revealed that DL-mevalonolactone alone exhibited weak antibacterial activity, with an MIC value of 8 mM, indicating a minimal direct antibacterial effect. Subsequently, checkerboard assays were conducted to assess the potential synergistic interaction between DL-mevalonolactone and antibiotics. The FIC index indicated that DL-mevalonolactone exhibited minimal in vitro synergistic activity with ciprofloxacin against S. aureus Newman. Consequently, the potentiation effect observed with bactericidal antibiotics in vivo cannot be attributed solely to drug–drug interactions (Figure 5A). We subsequently proposed that the human endogenous metabolite DL-mevalonolactone may induce bacterial activation and inhibit the formation of persisters. Thus, its reduced levels following chronic low-dose cadmium exposure could lead to the emergence of antibiotic tolerance, characterized by phenotypically tolerant bacteria without genetic alterations. To evaluate this hypothesis, we assessed the inhibitory effect of DL-mevalonolactone on the formation of S. aureus Newman persisters. Cultures of S. aureus Newman were exposed to varying concentrations of DL-mevalonolactone for 16 h, with the highest concentration (500 μM) showing no impact on bacterial growth. Subsequently, ciprofloxacin at 10 times the MIC was administered to eliminate non-persistent bacteria. As anticipated, the rate of persister formation was significantly reduced with the addition of DL-mevalonolactone (Figure 5B). Furthermore, we assessed the eradication effect of DL-mevalonolactone on S. aureus Newman persisters. Persister cells were exposed to a combination of ciprofloxacin at 10 times the MIC and varying concentrations of DL-mevalonolactone (ranging from 0 to 500 μM). The eradication rate of persisters was determined using untreated bacteria as a control. Our findings indicate that the addition of DL-mevalonolactone enhances the antibacterial efficacy of ciprofloxacin against S. aureus Newman persisters (Figure 5C).

To elucidate the mechanism by which DL-mevalonolactone reduces the formation rate of antibiotic-tolerant bacteria, we conducted a transcriptomic analysis. S. aureus Newman, treated with 500 μM DL-mevalonolactone for 12 h, underwent total RNA extraction and sequencing, leading to the identification of differentially expressed genes. The treatment with DL-mevalonolactone resulted in the upregulation of 206 genes and the downregulation of 182 genes (Figure 5D). The expression levels of several genes associated with energy metabolism were notably upregulated, including ATP-dependent ABC transporter genes varF and varG (log₂[FC] = 1.54 and 1.45, respectively), as well as genes involved in arginine biosynthesis, namely argB/C/F/H/J (log₂[FC] = 1.40, 1.74, 3.09, 1.73, and 1.55, respectively). This suggests that supplementation with DL-mevalonolactone promotes enhanced energy production and arginine metabolism. Conversely, the expression levels of several genes encoding key virulence factors of S. aureus were markedly downregulated, including the staphylocoagulase encoding gene coa, a multidrug efflux-related gene norB, a microbial surface component recognizing adhesive matrix molecule encoding gene spa, and a staphylococcal enterotoxin-like toxin X encoding gene selX (log₂[FC] = −1.49, −1.71, −3.63, and −1.98, respectively). These findings indicate that DL-mevalonolactone may also suppress the expression of virulence factors in S. aureus. The KEGG enrichment bar chart revealed that differentially expressed genes were predominantly associated with various metabolic pathways, notably including arginine biosynthesis, valine, leucine, and isoleucine biosynthesis, lysine biosynthesis, as well as alanine, aspartate, and glutamate metabolism (Figure 5E). Collectively, these findings suggest that DL-mevalonolactone significantly influenced the metabolism of S. aureus, particularly affecting amino acid and energy metabolism (Figure 5F).

Previous studies have indicated that a key factor contributing to antibiotic tolerance is the reduction in bacterial metabolism (Lobritz et al., 2015; Liu et al., 2022). This metabolic deceleration enables bacteria to enter a dormant state, rendering them less vulnerable to antibiotics, which typically target active cellular processes. Given that DL-mevalonolactone supplementation enhances S. aureus metabolism, we hypothesize that the reduction in DL-mevalonolactone levels due to chronic low-dose cadmium exposure diminishes bacterial metabolism and energy production, thereby resulting in decreased antibiotic susceptibility. To validate our hypothesis, we assessed alterations in bacterial intracellular ATP levels following treatment with a combination of ciprofloxacin and increasing concentrations of DL-mevalonolactone. Consistent with our expectations, the addition of DL-mevalonolactone significantly enhanced ATP production (Figure 5G), indicating that DL-mevalonolactone may facilitate the transition of dormant bacteria to a metabolically active state by promoting ATP synthesis. This mechanism of metabolic reactivation parallels the effects observed with other small-molecule adjuvants, such as 2-hydroxybutyrate, which also restored bacterial susceptibility by augmenting TCA cycle activity and ATP production (Xie et al., 2025). The association between metabolic activation and the reversal of antibiotic tolerance is further corroborated by studies indicating that the replenishment of essential metabolic intermediates could counteract bacterial dormancy in persistent infections (Lobritz et al., 2015).

The S. aureus Newman strain used in this study was kindly gifted by Prof. Wei Xia (Sun-Yat Sen University). Bacteria were cultured in the tryptic soy broth (TSB) media or on TSB agar plates at 37 °C except as specifically stated. Chemical reagents were purchased from Shanghai Sangon Biotech, MCE or J&K Scientific.

Female C57BL/6 mice (aged 6 weeks, 25–28 g) were supplied from Changzhou Kavens Laboratory Animal Co., Ltd. Female mice were selected based on their increased sensitivity to cadmium toxicity due to sex-specific differences in metal metabolism (Vahter et al., 2007), and more stable group-housing conditions for long-term microbiota studies, avoiding male aggression-related confounders (Van Loo and Baumans, 2003). All animal experiments performed in this study were complied with the Declaration of Helsinki and relevant regulations released by the Ministry of Health and the State Science and Technology Commission, People’s Republic of China. All animal experimental protocols were approved by the Ethics Committee of Wuxi people’s hospital, Jiangsu, China (permission number, DL2023019). The laboratory animal use license was SYXK-2020-0010, certified by the Science and Technology Agency of Jiangsu Province.

Drinking water for control group mice was Nongfu natural mineral water, whereas for Cd-exposure group mice, the drinking water contained 1 mg/L CdCl₂. Both groups were raised in conventional housing for 12 weeks before infection.

After 12 weeks of modeling, hearts, livers, spleens, lungs and kidneys of both group mice were harvested and weighed. The Cd concentrations of these internal organs were analyzed by ICP-MS after nitrolysis to verify the success of the Cd-exposure model.

Staphylococcus aureus Newman was cultured overnight at 37 °C and diluted 1/100 into fresh TSB media. The liquid culture was grown at 37 °C with shaking (220 r.p.m.) for about 4 h until the OD_600nm reached 0.6–0.8, and then bacteria were harvested and washed twice with ice-cold PBS buffer. Bacteria c.f.u. per milliliter was determined before infection.

To evaluate ciprofloxacin efficacy, control and Cd-exposure group mice were infected intraperitoneally with S. aureus Newman at a non-lethal dose (1.0 × 10⁷ c.f.u. per mouse). The infected mice received intraperitoneal rejections with a total dose of 20 mg/kg ciprofloxacin or sterile PBS as control at 2 h post-infection. Animals were euthanized 24 h following infection. The lungs, livers and kidneys of each mouse were aseptically removed, homogenized, serially diluted and plated on TSB plates to determine the bacterial c.f.u. in the absence or presence of ciprofloxacin treatments. Six mice were used for each experimental group.

Before gut microbiota transplantation, the transplantation group mice (n = 6 biologically independent mice per group) were initially fed with a cocktail of antibiotics in drinking water for 1 week to clear the original gut microbiota. Approximately 100 mg feces from Cd-exposure group mice were collected and resuspended in 1.5 mL sterile PBS, and then the gut microbiota supernatants were obtained after centrifugation. Supernatants were transplanted into the microbiota-depleted mice by gavage for 2 weeks. After transplantation, mice were fed with a standard diet and regular drinking water as the control group for another 10 weeks. Subsequently, the ciprofloxacin efficacy was evaluated as described above.

Fresh fecal samples of control and Cd-exposure group mice were collected after feeding for 12 weeks. DNA was extracted using the CTAB according to manufacturer’s instructions and 16S rDNA sequencing was performed by Shanghai Biotree Biotech. For amplicons of the V3-V4 region, primers were 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′). For amplicons of the Archae region, primers were F (5′-GYGCASCAGKCGMGAAW-3′) and R (5′-GGACTACHVGGGTWTCTAAT-3′). For amplicons of the V4 region, primers were 515F (5′-GTGYCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). For amplicons of the V4-V5 region, primers were F (5′-GTGCCAGCMGCCGCGG-3′) and R (5′-CCGTCAATTCMTTTRAGTTT-3′). The amplicon pools were prepared for sequencing and the size and quantity of the amplicon library were assessed on Agilent 2100 Bioanalyzer (Agilent, United States) and with the Library Quantification Kit for Illumina (Kapa Biosciences, Woburn, MA, United States), respectively. The libraries were sequenced using the NovaSeq PE250 platform. High-quality clean tags were obtained by fqtrim (v 0.94) and chimeric sequences were filtered using Vsearch software (v 2.3.4). Alpha diversity was applied in analyzing complexity of species diversity for a sample through 5 indices, including Chao1, Observed species, Goods coverage, Shannon, Simpson, and all these indices were calculated with QIIME2. Beta diversity was calculated by QIIME2, the graphs were generated by R package. Blast was used for sequence alignment, and the feature sequences were annotated with SILVA database for each representative sequence.

For metabolites extraction, 100 μL serum sample from control and Cd-exposure group mice after 12 weeks of feeding were mixed with 400 μL extraction solution (MeOH: ACN, 1:1) containing deuterated internal standards. Samples were vortexed for 30 s, sonicated for 10 min at 4 °C and incubated at −40 °C for 1 h to precipitate proteins. After centrifugation (13,800 g for 15 min at 4 °C), the supernatants were obtained for LC–MS/MS analysis.

The metabolite extracts were analyzed by a Vanquish UHPLC system (Thermo Fisher Scientific) with a Waters ACQUITY UPLC BEH Amide (2.1 mm × 50 mm, 1.7 μm) coupled to Orbitrap Exploris 120 mass spectrometer (Thermo Fisher Scientific). Water with 25 mmol/L ammonium acetate and 25 mmol/L ammonia hydroxide (pH = 9.75) (A) and acetonitrile (B) were used as the mobile phase. The auto-sampler temperature was 4 °C, and the injection volume was 2 μL.

The ESI source conditions were set as follows: sheath gas flow rate as 50 Arb, Aux gas flow rate as 15 Arb, capillary temperature 320 °C, full MS resolution as 60,000, MS/MS resolution as 15,000, collision energy: SNCE 20/30/40, spray voltage as 3.8 kV (positive) or −3.4 kV (negative), respectively. The raw data were converted to the mzXML format using ProteoWizard and processed with an in-house program, which was developed using R and based on XCMS, for peak detection, extraction, alignment, and integration. Then an in-house MS2 database (BiotreeDB) was applied in metabolite annotation. The cutoff for annotation was set at 0.3.

Relative abundance of top 15 differential genus and differentially expressed metabolites were used to calculate correlations via Spearman algorithm. The cor.test function was used to determine whether the correlation coefficient between the two numerical vectors was statistically significant. Matrix of correlation coefficients and matrix of p values were obtained, and the correlation heatmap was plotted using R package ggplot.

Monoclonal S. aureus Newman activated on TSB agar plate was picked and inoculated into fresh TSB medium for overnight culture at 37 °C. Then the bacterial solution was diluted 100 times into fresh TSB medium and the culture was continued for an addition 2 h until the OD_600nm reached 1.0. The 100-fold diluted bacterial solution was mixed with an equal volume of twofold gradient diluted DL-mevalonolactone in TSB medium, and the mixtures were added into a 96-well transparent microplate. After 16 h incubation at 37 °C, the lowest ciprofloxacin concentration that the growth of bacteria was invisible was defined as the MIC value. Each experiment was performed with biological replicates.

Bacterial solution preparation was performed as described in MIC assays. Each drug was serially diluted to 8 concentrations (final concentrations ranged from 0 to 4 × MIC) to create an 8 × 8 matrix. Equal volume of drug dilution was mixed with bacterial suspensions in a 96-well transparent microplate. After incubation for 16 h at 37 °C, the OD_600nm value of each well was detected by a multifunctional microplate reader. Three biological replicates were performed and the means were used for calculation. Fractional inhibitory concentration index (FICI) was calculated as the sum of the two MIC_combination/MIC_alone ratios of two drugs; FICI = FIC_A + FIC_B. Synergism was defined with FICI ≤ 0.5.

Staphylococcus aureus Newman was cultured overnight and diluted 1000 times into fresh TSB medium containing different concentrations of DL-mevalonalactone (final concentrations ranged from 0 to 500 μM). The culture was continued at 37 °C for 16 h, and then 10 × MIC ciprofloxacin was added to kill non-persistent bacteria for another 4 h. Samples were removed, serially diluted and plated onto TSB agar plates to calculate bacteria c.f.u. per ml at each time point as indicated.

To determine the effect of DL-mevalonolactone on the eradication of S. aureus Newman persisters, surviving bacteria after treatment with ciprofloxacin for 4 h were collected and resuspended in fresh TSB medium for further incubation. 10 × MIC ciprofloxacin in combination with different concentrations of DL-mevalonolactone (final concentrations ranged from 0 to 500 μM) were added for another 4 h treatment. The initial and remaining bacteria c.f.u. per ml were counted by coating plate method as described above, and the eradication rates were calculated.

Staphylococcus aureus Newman was grown to logarithmic phase in fresh TSB medium, then 500 μM DL-mevalonolactone were added into the dosing group. After 4 h at 37 °C, bacteria were collected and total RNAs were extracted using a Bacterial RNA Kit (Omega Bio-Tek). The transcriptomic sequencing and data analysis were performed by Shanghai Sangon Biotech. Total RNA of samples were sequenced using Illumina HiseqTM. Raw sequenced reads were evaluated by FastQC, filtered using Trimmomatic to obtain clean data and mapped against the genome of S. aureus Newman through Bowtie2. After normalization, differentially expressed genes were identified using DESeq2 with q value ≤ 0.05 and |log₂Fold Change| ≥ 1.

Intracellular ATP levels of S. aureus Newman persisters were determined using a ATP Content Assay Kit (Shanghai Sangon Biotech) as manufacturer’s instructions. S. aureus Newman persisters were prepared as described above and adjusted to OD_600nm = 0.6. After treatment with different concentrations of DL-mevalonolactone (final concentrations ranged from 0 to 500 μM) for 2 h, bacterial precipitates were collected by centrifugation. Lysostaphin was used to lyse S. aureus Newman, and supernatants were obtained for intracellular ATP levels determination. Detecting solution was prepared according to instructions and mixed with samples or standard solutions immediately before optical density A1 at 340 nm was determined. After incubation at 25 °C for 3 min, optical density A2 at 340 nm was determined. Total ATP levels (μmol/106 cells) were calculated as 0.125 × ∆A_sample÷∆A_standard.

For 16S rRNA data analysis, Wilcoxon tests were used for alpha diversity comparisons while PERMANOVA assessed beta diversity differences. Untargeted metabolomics data utilized Wilcoxon rank-sum tests with FDR correction to identify differential metabolites. Transcriptomic analysis of DEGs was performed using DESeq2’s Wald test. And Spearman’s correlation analysis with FDR adjustment was applied for multi-omics integration, with all analyses conducted in R. Other statistical analysis was performed by GraphPad Prism v.9.5.0 for Windows. All data were presented as mean ± s. d. unless indicated. Differences between two groups were evaluated using unpaired two-tailed Student’s t-tests and significance was indicated for p < 0.05.