The Chinese medicines in XRZYBXD prescription were purchased from the Second Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine. XRZYBXD was prepared by combining two times of decocted Chinese medicine solution. Q-Orbitrap high resolution liquid-mass combination was applied to identify the metabolites of XRZYBXD.

Mouse hepatoma cell H22 were purchased from China Typical Culture Storage Center (Wuhan, China); H22 cells were cultured in RPMI-1640 medium (Gibco) containing 10% FBS (fatal bovine serum) in cell incubator containing 5% CO₂ at 37°C.

Male C57BL/6 mice aged at 8–9 weeks (18–20 g) were provided by Experimental Animal Center of Guangzhou University of Chinese Medicine. All animals were raised in the specific pathogen-free (SPF) animal house with light and darkness for 12 h each, and appropriate temperature and humidity. Mice have free access to food and purified water. Animal experiments were approved by the Animal Ethics Committee of Guangzhou Ling Fu Tuopu biological Co., LTD (Approval No. LFTOP-IACUC-2024-0130).

After 7 days of adaptive feeding, 5 mice were randomly chosen as the normal group (N), and 25 mice were used for building H22 tumor-bearing mice model. As we previously described (Zeng et al., 2023), the H22 cells (2 × 10⁷/mL) were inoculated subcutaneously into the right axilla of mice. After successfully modeling, the mice were randomly divided into 5 groups: tumor model group (T), XRZYBXD high dose (H), medium dose (M) and low dose (L) groups, and positive drug group (S). Mice in the XRZYBXD group were administered with a low dosage (13.77g/kg), medium dosage (27.54 g/kg) and high dosage (55.08 g/kg/day) via gavage, once a day. Tegafur, Gimeracil and Oteracil Potassium Capsules (S-1) (Approved by H20100135, Jiangsu, China) was used as the positive drug. The efficacy of S-1 is clear, and it is a commonly used clinical chemotherapy drug for the treatment of digestive system tumors. According to the body surface area of 1.6 m² and the body weight of 60 kg for adults, the dosage of S-1 is 120 mg per day. According to the conversion formula of intragastric dose between mice and human, the intragastric dose of mice was 25 mg/kg, the intragastric volume was 0.1 mL/10 g, once a day. And the drug was configured to about 2.5 mg/mL. Mice in the normal group and tumor model group were given the same volume of saline. The body weight and tumor volume were measured every 3 days. The tumor volume was calculated by the formula V = (length × width²)/2. After 21 days of administration, mouse feces in model group and XRZYBXD high-dose group were collected to prepare for subsequent microbial manipulation experiments. Then the mice were anesthetized cervical vertebra dislocated and killed, and the blood of the eyeball, and the tumor, liver, kidney and ileum were taken out. The tumor, liver, kidney and ileum tissues were stained with HE, observed and photographed by light microscope (200×). Q-PCR (Quantitative real time polymerase chain reaction), and immunohistochemical (IHC) were performed using tumor and ileum tissues. Inflammatory factors including IL-1β, IL-18 in mice were assessed by ELISA (Enzyme-linked Immunosorbent Assay) kits.

Fresh feces were collected in sterilized EP (Eppendorf) tubes and stored in the −80°C refrigerator for later use. Quality tests are performed on the feces first. Qualified DNA samples were selected and configured with the PCR reaction system for PCR amplification. Agencourt AMPure XP magnetic beads were used to purify the PCR amplification products and dissolved in the Elution Buffer. Then they were labeled and applied for library construction. The fragment range and concentration of the library were detected using Agilent 2100 Bioanalyzer. The qualified library was sequenced by HiSeq platform according to the size of the inserted fragment. Filter the data of offline, remaining high-quality clean data for later analysis. Reads are spliced into Tags by the relationship of overlap between reads. The Tags are clustered into OTU and compared with the database to annotate the species. Based on OTU and annotation results, sample species complexity analysis, inter-group species difference analysis, association analysis and function prediction were conducted.

After 21 days of intervention, the feces of the tumor model group and the XRZYBXD high dose group were respectively tested for quality, and the qualified feces were detected for metabolites by LC-MS/MS (liquid chromatography tandem mass spectrometry). Bioinformatics analysis was used to screen differential metabolites.

A mixture of antibiotics (ABX) containing 1 mg/mL ampicillin, 1 mg/mL metronidazole, 1 mg/mL neomycin and 0.5 mg/mL vancomycin (Xu et al., 2020) was applied to build the pseudo-germ-free mouse model. The drinking water with ABX was changed three times a week. Further, in order to validate the anti-tumor and pro-pyroptosis effects of XRZYBXD-shaped intestinal flora, Mouse feces from the model group and the high-dose group of XRZYBXD of the first animal experiment were collected. Refer to previous literature (Zhong et al., 2022), 0.3g feces were added into 3 mL distilled water, thoroughly shaken and mixed, and impurities were filtered to make feces suspension, which was stored in the refrigerator at −80°C for future use. Pseudo-germ-free mice were administered with 200 μl fecal suspension by gavage for 21 days.

Q-PCR assay was applied as we previously reported (Zeng et al., 2023). Mouse PCR primers were designed and purchased by Sangon Biotech (Shanghai, China), including NLRP3, Caspase 1, GSDMD, IL-1β, IL-18, occluding, claudin-1, ZO-1, Muc2 and the housekeeping gene primer ACTB. The primer sequences are shown in Table 1. The relative expression levels of each gene were obtained by analyzing and calculating 2^–ΔΔCt.

Western blot assay was conducted as we previously reported (Zeng et al., 2023). The following antibodies were used in the Western blot assay including NLRP3 antibody [15101, Cell Signaling Technology (CST)], Cleaved caspase 1 antibody (89332, CST), GSDMD-N antibody (34667, CST), IL-1β antibody (12242, CST), IL-18 antibody (57058, CST), and β-actin antibody (4970, CST).

After paraffin section was dewaxed, citric acid antigen repair buffer was added for repair, sealed with 3%BSA at room temperature, then primary antibodies against occluding (27260-1-AP, Proteintech), claudin-1 (AF0127, Affinity) and ZO-1 (AF5145, Affinity) were added, and incubated at 4°C overnight. Next, secondary antibodies were used to incubate the slides for 50 min. DAB color solution was added for color development, and the time was controlled by microscope observation. When the color development turned to brownish yellow, it was terminated by washing with water and sealed with neutral gum, then observed and photographed under microscope and analyzed statistically. The total optical density (IOD) and area of each image were measured, and the average optical density (AOD) was calculated.

The statistical software SPSS 25.0 was applied for statistical analyses. When the data satisfy both normal distribution and homogeneity of variance, one-way analysis of variance (ANOVA) was used. When the data do not meet the normal distribution or the variance is not uniform, the non-parametric Kruskal-Wallis H test was used. Statistical graphs were plotted by Graphpad Prism v.8.0 (Graphpad Software, USA). The statistical results were statistically different with P < 0.05.

A total of 417 metabolites were obtained in XRZYBXD based on Q-Orbitrap high resolution liquid-mass combination. Among them, 30 metabolites with high matching degree with mzCloud database were shown in Table 2 and Supplementary Figure 1.

With different intervention, the growth trend of tumor volume in XRZYBXD high, medium dose group and S-1 group were significantly slower than that in tumor model group (Figure 1A). At the end of intervention, the tumor weight of XRZYBXD high, medium dose group and S-1 group were also significantly lower than that of the tumor model group (Figure 1B). Pathological examination showed that tumor cells in the three groups were loosely arranged, and their necrotic area of tumor tissue were significantly larger than that in the tumor model group (Figure 1D). Liver and kidney sections of mice suggested no significant abnormality in each group (Figure 1D). However, during the administration, compared with the other groups, the body weight of mice in the S-1 group showed a downward trend. At the end of intervention, the mice in S-1 group had the lowest body weight and were significantly lower than the other groups (Figure 1C). Therefore, it could be found that although the S-1 group had better anti-tumor effect, the mice lost weight more quickly, while XRZYBXD inhibited hepatoma cells growth without body weight loss side effects.

As shown in Figures 1E–I, the mRNA relative expression level of NLRP3, Caspase 1, GSDMD, IL-1β and IL-18 in XRZYBXD high dose group were significantly higher than those in the tumor model group by Q-PCR. The protein expression level of NLRP3, caspase 1, GSDMD, IL-1β and IL-18 in XRZYBXD high dose group were also significantly higher than those in the tumor model group by WB (Figures 1J–M and Supplementary Figures 2F–H). Besides, Elisa results of tumor tissue displayed that compared with tumor model group, the levels of IL-1β and IL-18 in XRZYBXD high dose group were increased (Figures 1N, O). These results indicated that XRZYBXD may promote pyroptosis of hepatoma cells.

H&E staining, IHC and Q-PCR were used to investigate the effect of XRZYBXD on intestinal barrier. As we can see from the normal group in H&E staining (Supplementary Figure 2A), the structure of each layer of ileum was clear, the villi were intact, the mucosal epithelial cells were densely arranged, and no obvious pathological variations can be observed. However, in the tumor model group, the ileum structure was disturbed, the villi were broken, and the goblet cells were reduced, the mucosal epithelium of the ileum was necrotic and exfoliated. After treatment of XRZYBXD high dose, the pathological changes of the ileum structure, mucosa and goblet cells were restored (Figure 4A). Further, occluding, claudin-1, ZO-1 and mucin 2(Muc2) were applied to assess intestinal barrier function under different invention. Compared with normal group, the expression of occluding, claudin-1, ZO-1 and Muc2 in tumor model group were prominently reduced, but they were all restored after the intervention of XRZYBXD (Figures 4B–G and Supplementary Figure 2B).

We further investigated the metabolites produced by gut flora using feces through untargeted metabolomics. Based on the threshold of fold change (FC) > 2 and P-value < 0.05, compared with tumor model group, we identified 592 down-regulated and 685 up-regulated metabolites as shown in volcano plot (Figure 4H). Further, we identified 79 differential metabolites whose retention time matched with BGI metabolites database (BMDM) or mzCloud database and presented the results using a heatmap and table (Figure 4I and Table 3). Among these differential metabolites, compared with the tumor model group, differential metabolites whose relative content increased following XRZYBXD intervention such as 8-hydroxyquinoline, Indole, N-(2,6-dimethylphenyl)-n-(methoxyacetyl) alanine, N-acetylmethionine, 2-methoxyresorcinol, D-ornithine, Quinoline, Ursolic acid, Dl-malic acid and Spermidine, while Betaine, Cholecalciferol, Dl-carnitine, Trans-cinnamaldehyde, 3-coumaric acid, 5-hydroxyindole-3-acetic acid, Carbaprostacyclin, Cuminaldehyde, Ethyl oleate, N-acetyldopamine, B -asarone and so on decreased prominently following XRZYBXD intervention.

Spearman correlation analysis was conducted to investigate the association between gut microbiota and tumor weight, as well as indicators of hepatoma cell pyroptosis and intestinal barrier function. We found that probiotic such as Akkermansia muciniphila, Parabacteroides distasonis, and Parasutterella excrementihominis were positively correlated with indicators of pyroptosis, while harmful bacteria such as Barnesiella intestinihominis, Clostridium scindens showed negative correlation with these indicators. However, only Akkermansia muciniphila had strong negative correlation with tumor weight (Figure 4J). Additionally, many microfloras such as Akkermansia muciniphila, Parasutterella excrementihominis, Barnesiella intestinihominis and Bacteroides_stercorirosoris also had close association with indicators of intestinal barrier function.

Then, we further explored the association between microflora and metabolites through spearman correlation analysis. The most important microflora Akkermansia muciniphila exhibited significant positive correlations with several differential metabolites including 3-(2-hydroxyphenyl) propanoate, Cryptotanshinone, Dantron, Gibberellin a4, Hematoxylin, N-acetyl-l-tyrosine, Tretinoin, Spermidine, (-)-caryophyllene oxide, 3,4-dihydroxybenzaldehyde, 3-acetyl-11-keto-β-boswellic acid, Perillartine and Hesperetin, while Akkermansia muciniphila showed prominent negative correlations with several differential metabolites including (±)-abscisic acid, Phenylacetylglycine, (±)11(12)-eet, 15(r),19(r)-hydroxy prostaglandin f2α, Butylparaben, Sedanolide, Medroxyprogesterone, Ethyl oleate, 3-acetyl-2,5-dimethylfuran, Diacetyl and Vanillin (Figure 5A).

Interestingly, as we can see from Figure 5B, when spearman correlation analysis was applied to investigate the relationship between differential metabolites and tumor weight, as well as indicators of pyroptosis, we found that the above-mentioned differential metabolites who had significant correlation with Akkermansia muciniphila, also had significant correlation with tumor weight and indicators of pyroptosis. These indicated that microfora especially Akkermansia muciniphila may have anti-tumor effect through their metabolites (Figure 5B).

As shown in Figure 6A, to verify the contribution of intestinal flora in the efficacy of XRZYBXD, ABX was used for eliminating intestinal flora in H22 tumor-bearing mice during XRZYBXD treatment. Intriguingly, the tumor volume and tumor weight of ABX + XRZYBXD high dose group were higher than that of XRZYBXD high dose group, but lower than that of tumor model group (Figures 6B, C), which indicated that the elimination of gut microbiota weakened the anti-tumor effect of XRZYBXD. In addition, compared with XRZYBXD high dose group, the mRNA and protein expression of NLRP3, caspase 1, GSDMD, IL-18 and IL-1β were markedly decreased in ABX + XRZYBXD high dose group, which suggested that the pro-pyroptosis effect of XRZYBXD in hepatoma cells was attenuated by the depletion of intestinal flora (Figures 6D–N and Supplementary Figures 2I–K). Additionally, through H&E staining, we found that although the damage degree of ileum in ABX + XRZYBXD high dose group mice was lower than that in model group, there were still some villi morphology disturbed, goblet cells reduced, and some epithelial cells shed (Supplementary Figure 2A). Following the depletion of microbiota, the expression of occluding, claudin-1, ZO-1 and Muc2 also had a noticeable decrease when compared with XRZYBXD high dose group (Figures 6O–U and Supplementary Figure 2C).

Furthermore, fecal microbiota from XRZYBXD high dose group and tumor model group were respectively transplanted into pseudo-sterile H22 tumor-bearing mice to confirm the role of microbiota in the pro-pyroptosis effect of XRZYBXD (Figure 7A). Compared with mice that were transplanted with fecal microbiota from the tumor model group (T(FMT) group), mice that were transplanted with fecal microbiota from the XRZYBXD high dose group (H(FMT) group) had significant decrease in tumor volume and weight, but there was no significant difference between T (FMT) group and tumor model group (Figures 7B, C). Meanwhile, the mRNA and protein expression levels of NLRP3, caspase 1, GSDMD, IL-1β and IL-18 in H(FMT) group were higher than that in T(FMT) group, while there were no noticeable difference between T(FMT) group and tumor model group (Figures 7D–N). Besides, H&E staining showed that the destroyed of FMT(T) group was similar to model group, but significantly restored after the intervention of FMT(H) (Supplementary Figure 2A). Meanwhile, the expression of occluding, claudin-1, ZO-1 and Muc2 were also increased following H(FMT) treatment (Figures 7O–U and Supplementary Figure 2D). These data suggested that the FMT experiment was successfully executed and XRZYBXD-shaped intestinal flora had a prominent promoting effect on hepatoma cell pyroptosis.