Ultrapure water (18.25 MΩ) was prepared using an ultrapure water system (Chengdu Yue Chun Technology Co., Ltd., Chengdu, China). Acetonitrile, formic acid, and methanol (all high-performance liquid chromatography [HPLC] grade) were obtained from Tedia (Fairfield, OH, United States). Ammonium formate was purchased from Thermo Fisher Scientific Inc. (Shanghai, China). The broad-spectrum β-lactam antibiotic imipenem/cilastatin sodium (in equal quantities; trade name: Tienam) was purchased from Hangzhou MSD Pharmaceutical Co., Ltd. (Hangzhou, China). Analytical-grade chemicals were used in our experiments.

The CS herb was purchased from Nanning Shengyuantang Chinese Herbal Medicine Co., Ltd. (Nanning, China), and it was authenticated by Associate Professor Changming Mo at the Guangxi Academy of Agricultural Sciences. The CS herb was cleaned, dried, and cut into small pieces. Supplementary Figure 1 shows the TACS extraction protocol in our lab. After the extraction was finished, the TACS sample was stored in a desiccator.

UPLC-Q-TOF/MS was performed to characterize the chemical profile of TACS. Corresponding analytical methods and results are described in the Supplementary Materials.

Thirty healthy male Sprague–Dawley rats (weighing 100–140 g) were obtained from the Experimental Animal Center of Guangxi Medical University (Guangxi, China; Approval No.: SCXK-20140002). The rats were raised individually in plastic cages in a room maintained at 22 ± 3°C and 55 ± 15% humidity with a 12-h light–12-h dark cycle and free access to water. The rats were randomly divided into three groups (with 10 rats in each group): a control group, an antibiotic-only group, and a TACS group. The control group was given 1 ml/100 g i.g., normal saline. The antibiotic-only group received 50 mg/kg i.g. of the antibiotic preparation imipenem/cilastatin sodium. The TACS group received the same dose of imipenem/cilastatin sodium with 50 mg/kg oral TACS. The rats were treated once a day for 21 days (Zheng et al., 2011). The body weight was recorded daily as shown in Supplementary Figure 4.

This study was approved by the Ethics Committee for Animal Experiments of Guangxi Medical University (No. 201709003).

On day 20, urine samples were collected for 24 h from rats in metabolic cages (with one rat per cage). The urine samples were stored at -80°C. They were thawed at room temperature before analysis and centrifuged (13,000 rpm, 4°C, 10 min). Next, 500-μl supernatant was mixed with 500-μl water, and an aliquot of 5 μl was used for UPLC-Q-TOF/MS analysis.

Feces samples were also collected on day 20 and then transferred into sterile conical tubes. The feces samples were immediately frozen in liquid nitrogen and stored at -80°C. They were ground on ice and weighed. The samples were processed by adding ice-cold water at a ratio of 100 mg/500 μl, vortexing, blending, and centrifuging (13,000 rpm, 4°C, 15 min). The supernatant was collected, and the sample was further processed using the same method with ice-cold methanol. The double-extraction mixture was twirled and centrifuged (13,000 rpm, 20 min, 4°C), and an aliquot of 5 μl was used for UPLC-Q-TOF/MS analysis.

At 24 h after the last treatment administration, all the rats were sacrificed, and cecum samples were collected in clean centrifuge tubes. The cecum samples were immediately placed in dry ice and then stored at -80°C until microbiological analysis.

The microbial DNA was extracted from the cecum samples and prepared for 16S rRNA gene sequencing analysis. The analysis was performed as follows: (i) the microbial DNA was extracted; (ii) the DNA quality was assessed by electrophoresis; (iii) the V3–V4 hyper-variable regions of the bacterial 16S rRNA gene were amplified using a Thermocycler PCR system (Gene Amp 9700, ABI, United States); (iv) PCR products were extracted, purified, and then sequenced on an Illumina MiSeq platform (Illumina, San Diego, CA, United States); and (v) raw fastq data were processed and analyzed. Detailed steps are described in the Supplementary Materials. The raw 16S rRNA sequence data have been deposited in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) database under accession number PRJNA529316.

According to the references in the Greengenes (v13.5) database, PICRUSt was performed using the abundance predictions of the Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologs and the KEGG pathways of the bacterial communities (Langille et al., 2013). STAMP was used to calculate and visualize the functional differences between pairs of groups (Parks et al., 2014). We used two-sided Welch’s t-tests and the Benjamini–Hochberg false discovery rate (FDR) correction for statistical analysis in the two-group functional prediction analysis. Moreover, the prediction accuracy of PICRUSt was evaluated using the Nearest Sequenced Taxon Index (NSTI), with lower values indicating higher prediction accuracy (Hu et al., 2016).

Using high-precision docking simulations and molecular pathway maps, we performed a molecular docking analysis to comprehensively characterize the selectivity of five ligands and to illustrate how they act on a complex molecular network. The target protein in the analysis was CYP27A1 (Protein Data Bank [PDB]-ID: 1mfx), which is a key enzyme in the bile acid synthesis pathway. The chemical structures of the five ligands, which are the main alkaloid components of TACS (berberine, palmatine, chelerythrine, jatrorrhizine, and dehydrocavidine), are shown in Supplementary Figure 5. The molecular docking analysis was carried out using the Surflex dock module of the licensed software SYBYL-X version 2.0 (Jain, 2007) as follows: (1) the .mol2 format documents of chemical structures were prepared by SYBYL-X 2.0; (2) Tripos force field was used for the energy minimization calculation to optimize the structure regarding the combination of each ligand and the target protein; (3) the protein data were downloaded from PDB⁶; (4) the protein was prepared using SYBYL-X 2.0 by removing the substructure (chain B) and water, fixing the bumps, and incorporating hydrogen; and (5) the molecular docking analysis was implemented (Liang et al., 2018; Thillainayagam et al., 2018). The molecular conformations identified from the molecular docking analysis were modified using the molecular conformation based on the crystal structure in PyMOL software (Seeliger and de Groot, 2010).

One-way analysis of variance (ANOVA) was used for evaluating the between-group significant differences in potential biomarkers using Statistical Package for Social Sciences (SPSS) program software version 16.0 (Chicago, IL, United States). P < 0.05 was set as the significance threshold.

16S rRNA gene sequencing was performed to analyze the cecum microbiota diversity and composition. We assessed the gut microbial diversity and richness using the Chao 1 index, Shannon diversity index, and observed operational taxonomic units (OTUs). Figure 1A shows that significant changes in both diversity and richness were found after antibiotic administration. A principal coordinate analysis (PCoA) was carried out to compare the microbiota patterns in each group. The distinct separations in the scatterplot indicated the significant differences in community composition between the three groups (Figure 1B).

Around 99% of the total bacterial relative abundance was classified into nine phyla and 161 genera; p__Firmicutes and p__Bacteroidetes were the dominant phyla, followed by p__Actinobacteria and p__Proteobacteria (Figure 1C). The 19 dominant genera that had relative abundance > 1% in each group were subjected to further analysis (Figure 1D). Ten of these genera significantly changed in the antibiotic-only group (Table 1). Among these genera, g__Blautia, g__Intestinibacter, g__Parabacteroides, g__Hungatella, and g__Ruminococcus_torques_group were significantly upregulated in the antibiotic-only group compared to the control group, and the former two were obviously restored by TACS treatment. In contrast, g__Clostridium_sensu_stricto_1, g__Ruminococcus_1, g__norank_f_Lachnospiraceae, g__unclassified_f_Ruminococcaceae, and g__norank_f_Ruminococcaceae were significantly downregulated in the antibiotic-only group. Additionally, g__Blautia was completely restored to the control level by TACS, while there was incomplete restoration of g__Clostridium_sensu_stricto_1, g__Hungatella, and g__Intestinibacter. These results indicated that gut microbiota dysbiosis at the OTU level and genus level could be caused by antibiotic administration, and this kind of dysbiosis can be modulated by TACS treatment.

UPLC-Q-TOF/MS in the positive ion mode was performed to investigate urine samples from the control, antibiotic-only, and TACS groups, which represented physiological status, pathological condition, and intervening effects, respectively. Typical base peak intensity (BPI) chromatograms of the three groups are shown in Supplementary Figure 6. PCA demonstrated the separation of the three groups (Figure 2A). PLS-DA, a supervised multivariable statistical method that was carried out to better visualize the discriminated groups from PCA and to evaluate the metabolic patterns of all three groups, showed that the metabolic profile of the antibiotic-only group was clearly separated from that of the control group, while the TACS group also differed from the antibiotic-only group and was similar to the control group (Figure 2B). The results indicated that antibiotic administration induced significant metabolic changes in the urine and that these changes could be modulated by TACS treatment.

To enhance the separation of the antibiotic-only and control groups established in the PLS-DA, OPLS-DA was employed (Figure 2C). Seven-fold cross-validation was used to evaluate the predictive abilities of the constructed OPLS-DA model. Furthermore, a 200-times permutation test was carried out between the control and antibiotic-only groups to prevent model over-fitting (Supplementary Figure 7). Significantly altered metabolites in the urine were selected using the S-plot and the VIP values. We considered the variables far from the origin in the S-plot with VIP > 1 to be the potential biomarkers associated with gut microbiota dysbiosis. As revealed in Figure 2D, 11 variables contributing to the clustering from the S-plot, based on the differences in the urine metabolic profiles between the antibiotic-only and control groups, were treated as the potential biomarkers. Their retention times, m/z, and VIP values are listed in Supplementary Table 3. Based on the accurate mass and the MS^E spectra measurements obtained via Q-TOF/MS and comparison of the data with literature and/or online databases, such as (HMDB), METLIN, Massbank, and KEGG, 10 identified endogenous metabolites were characterized. These metabolites were creatine, N(6)-methyllysine, valyl-hydroxyproline, proline, 2-hydroxy-2-methylbutanenitrile, N4-acetylaminobutanoate, betonicine, isovalerylglycine, 1-methoxypyrene, and pentahomomethionine.

Among these metabolites, valyl-hydroxyproline, proline, 2-hydroxy-2-methylbutanenitrile, and 1-methoxypyrene were increased significantly in the antibiotic-only rats compared to the controls, while creatine, N(6)-methyllysine, N4-acetylaminobutanoate, betonicine, isovalerylglycine, and pentahomomethionine were clearly decreased in the antibiotic-only rats. After TACS treatment, the level of N(6)-methyllysine, N4-acetylaminobutanoate, betonicine, and isovalerylglycine were upregulated compared to the antibiotic-only group, while valyl-hydroxyproline and 1-methoxypyrene were downregulated. The significant changes in the identified urinary biomarkers related to gut microbiota dysbiosis are shown in the histogram (Figure 2E).

UPLC-Q-TOF/MS in the positive ion mode was performed to investigate the fecal samples from all three groups. The representative BPI chromatograms of each group are shown in Supplementary Figure 8. PCA and PLS-DA were carried out to discriminate between the three groups and to evaluate the metabolic patterns of the three groups. The results showed that the metabolic profile of the antibiotic-only group was different from that of the control group, while the TACS group obviously differed from the antibiotic-only group but was similar to the control group (Figure 3A,B). The results suggested that antibiotic administration induced significant metabolic changes in feces, and these changes could be modulated by TACS treatment.

OPLS-DA, which enhanced the established separation of the antibiotic-only and control groups, and the S-plot were employed to identify the metabolites that contributed to the clustering (Figure 3C,D). Seven-fold cross-validation was used to evaluate the predictive abilities of the constructed OPLS-DA model. Furthermore, to prevent against model over-fitting, a 200-times permutation test was performed between the control and antibiotic-only groups (Supplementary Figure 9). As revealed in Figure 3D, twenty-three significantly altered metabolites in the feces were identified as potential biomarkers. Their retention times, m/z, and VIP values are listed in Supplementary Table 4. The selected endogenous metabolites were characterized based on the accurate mass and the MS^E spectra measurements obtained via Q-TOF/MS and comparison of the data with literature and/or online databases. This method identified 23 metabolites found in feces as potential biomarkers of gut microbiota dysbiosis, comprising N-acetylputrescine, N-acetyl-D-glucosamine, hypoxanthine, N-acetyl-D-mannosamine, tyramine, isoleucine, valine, tryptophan, 2-formaminobenzoylacetate, 5-hydroxyindoleacetate, urobilin, hyocholate, vulpecholate, 3-oxocholate, cholate, murideoxycholate, chenodeoxycholate, nutriacholate, 7-hydroxy-3-oxocholanoate, 7a-hydroxy-3-oxo-5b-cholanoate, ursodeoxycholate, deoxycholate, and 12-ketodeoxycholate.

Among these metabolites, N-acetylputrescine, tyramine, isoleucine, valine, tryptophan, 2-formaminobenzoylacetate, 5-hydroxyindoleacetate, hyocholate, 3-oxocholate, and cholate were clearly increased in antibiotic-only rats compared to the controls, and N-acetyl-D-glucosamine, hypoxanthine, N-acetyl- D-mannosamine, urobilin, vulpecholate, murideoxycholate, chenodeoxycholate, nutriacholate, 7-hydroxy-3-oxocholanoate, 7a-hydroxy-3-oxo-5b-cholanoate, ursodeoxycholate, deoxycholate, and 12-ketodeoxycholate were significantly decreased in antibiotic-only rats compared to the controls. After TACS treatment, the levels of hypoxanthine, 2-formaminobenzoylacetate, 7-hydroxy-3-oxocholanoate, and ursodeoxycholate were upregulated compared to the levels in the controls, while isoleucine and 3-oxocholate were obviously downregulated. The significant changes in the identified fecal biomarkers related to gut microbiota dysbiosis are depicted in the histogram (Figure 3E).

To evaluate the differences in the functional capacity of the intestinal bacterial community between the control and antibiotic-only group, PICRUSt analysis was performed to analyze the KEGG pathways related to the intestinal microbiota. As shown in Figure 4A, the second-level KEGG pathway analysis showed that energy metabolism and amino acid metabolism were enriched in the antibiotic-only group compared to the control group, while the enzyme family was decreased. Consistent with the second-level KEGG pathways results, Figure 4B shows that the antibiotic-only group had significantly elevated relative abundances of genes involved in phenylalanine, tyrosine, and tryptophan biosynthesis; glycine, serine, and threonine metabolism; amino acid–related enzyme, alanine, aspartate, and glutamate metabolism; and valine, leucine, and isoleucine biosynthesis (p < 0.05). Meanwhile, the abundance of genes involved in carbohydrate metabolism, nucleotide metabolism, and drug metabolism of other enzymes was decreased (p < 0.05).

To identify the potential link between altered gut microbiota and potential biomarkers in urine and feces involved in gut microbiota dysbiosis (r < -0.5 or r > 0.5, p < 0.05), we performed a Spearman’s rank-order correlation analysis. This analysis identified multiple significant associations between the perturbed gut microbiota and altered metabolites in rats with antibiotic-induced gut dysbiosis (Figure 5).

According to the correlation analysis between altered gut microbiota and potential urinary biomarkers, betonicine was positively correlated with g__Ruminococcus_1, g__norank_f_Ruminococcaceae, and g__unclassified_f_Ruminococcaceae. Creatine, N4-acetylaminobutanoate, N(6)-methyllysine, isovalerylglycine, and pentahomomethionine were positively correlated with g__norank_f_Lachnospiraceae,g__norank_f_Ruminococcaceae, and g__unclassified_f_Ruminococcaceae. Meanwhile, creatine was also positively correlated with g__Ruminococcus_1. 2-Hydroxy-2-methylbutanenitrile and proline were positively correlated with g__Ruminococcus_torques_group, g__Hungatella, g__Parabacteroides, and g__Intestinibacter. Valyl-hydroxyproline was positively correlated with g__Blautia, g__Hungatella, g__Parabacteroides, and g__Intestinibacter, and 1-methoxypyrene was positively correlated with g__Blautia and g__Intestinibacter.

Regarding the fecal potential biomarkers, all the elevated potential biomarkers were positively correlated with g__Hungatella, and all the elevated biomarkers but for valine and 2-formaminobenzoylacetate were positively correlated withg__Parabacteroides and g__Intestinibacter. N-acetylputrescine, isoleucine, tyramine, valine, 5-hydroxyindoleacetate, 2-formaminobenzoylacetate, hyocholate, and cholate were positively correlated with g__ Ruminococcus_torques_group, while isoleucine was also positively correlated withg__Blautia. N-acetyl-D-mannosamine, chenodeoxycholate, and 12-ketodeoxycholate were positively correlated to g__Ruminococcus_1, g__norank_f_Lachnospiraceae, g__norank_f_Ruminococcaceae, and g__unclassified_f_Ruminococcaceae, and ursodeoxycholate and deoxycholate were positively correlated with g__Ruminococcus_1. Also, N-acetyl-D-glucosamine, nutriadeoxycholate vulpecholate, murideoxycholate, urobilin, and 7a-hydroxy-3-oxo-5b-cholanoate were positively correlated with g__norank_f_Lachnospiraceae, g__norank_f_Ruminococcaceae, and g__unclassified_f_Ruminococcaceae. 7-Hydroxy-3-oxocholanoic acid was positively correlated with g__norank_f_Lachnospiraceae, while hypoxanthine was positively correlated with g__Clostridium_sensu_stricto_1.

The key metabolic pathways were mapped using MetaboAnalyst 3.5⁷ by integration of all identified potential biomarkers. As shown in Supplementary Figure 10, six metabolic pathways were considered to be the key pathways that may be important to antibiotic-induced gut microbiota dysbiosis (impact values > 0.01) (Wang et al., 2012). They comprised the pathways associated with valine, leucine and isoleucine biosynthesis, tryptophan metabolism, amino sugar and nucleotide sugar metabolism, arginine and proline metabolism, tyrosine metabolism, and purine metabolism. According to Supplementary Tables 3, 4, in the TACS group, the levels of N4-acetylaminobutanoate, betonicine, isovalerylglycine, hypoxanthine, 7-hydroxy-3-oxocholate, and ursodeoxycholate were significantly elevated compared to antibiotic-only group, while the levels of valyl-hydroxyproline, 1-methoxypyrene, isoleucine, and 3-oxocholate were decreased significantly compared to the antibiotic-only group, suggesting that TACS can regulate disorders of the most relevant metabolic pathways in the pathological process of antibiotic-induced gut microbiota dysbiosis. The microbial diversity results indicated that TACS also modulated the gut microbiota dysbiosis. Spearman’s correlation analysis showed that the 10 potential biomarkers recovered by TACS were correlated with g__Blautia and g__Intestinibacter, suggesting that the intervening effects of TACS may be achieved via regulation of the most relevant metabolic pathways, including branched-chain amino acid (BCAA) metabolism, bile acid metabolism, arginine and proline metabolism, and purine metabolism.

Thus, the comprehensive metabolic network construction (using KEGG and based on the identified potential biomarkers) revealed that BCAA metabolism, bile acid metabolism, arginine and proline metabolism, purine metabolism, aromatic amino acid metabolism, and amino sugar and nucleotide sugar metabolism were the key metabolic pathways involved in the metabolic changes in rats with gut microbiota dysbiosis treated with TACS (Figure 6). However, after combining the metabolic pathway analysis with Spearman’s correlation analysis, only four metabolic pathways (BCAA metabolism, bile acid metabolism, arginine and proline metabolism, and purine metabolism) were identified as the most relevant metabolic pathways involved in the intervening effects of TACS against antibiotic-induced gut microbiota dysbiosis.

Five alkaloid components of TACS (berberine, palmatine, chelerythrine, jatrorrhizine, and dehydrocavidine) were used in a molecular docking analysis with CYP27A1, the key enzyme in bile acid synthetic pathways. The total docking simulation scores of berberine, palmatine, chelerythrine, jatrorrhizine, and dehydrocavidine were 9.5164, 8.5358, 6.8842, 4.4558, and 4.4558, respectively. As shown in Figure 7, the binding sites of berberine, palmatine, chelerythrine, jatrorrhizine, and dehydrocavidine were ASN-370, ALA-314 and ARG-446, GLY-437, VAL-367 and GLY-437, and ARG-94, respectively.

Branched-chain amino acids (including leucine, valine, and isoleucine) and their metabolites can stimulate protein synthesis, mediate levels of neurotransmitter, promote insulin secretion, and decrease appetite and body adiposity by activating the mechanistic target of rapamycin (mTOR) signaling pathway. They can also be transported to the brain and influence neurotransmitter synthesis, resulting in different effects on body behavior and brain function (Nishitani et al., 2005; Pedroso et al., 2015; Zhang et al., 2017; Holecek, 2018). BCAA dysmetabolism influences many diseases. For example, patients with cirrhosis and hepatic steatosis have a decreased level of blood BCAAs (Dasarathy and Hatzoglou, 2017; Hoyles et al., 2018), while in patients with diabetes and obesity, the BCAA level is increased (Borghi et al., 1985; She et al., 2007). Studies have shown that BCAAs can be consumed and oxidized in the intestine, and BCAAs participate in bacterial metabolism and in the regulation of intestinal microbial species and diversity (Zhang et al., 2017). Leucine can increase villus height in the duodenum in order to enhance intestinal development, while isoleucine supplements can enrich Lactobacillus and inhibit the growth of Escherichia coli to improve the gut microbial population, forming a healthy gut microbial environment (Zhao et al., 2014; Sun et al., 2015).

In our previous study, the elevated level of isoleucine in the plasma of rats with CCl₄-induced acute hepatic injury was downregulated by Corydalis saxicola Bunting (CS), indicating that CS might modulate BCAA metabolism by promoting protein synthesis (Liang et al., 2016). The elevated isoleucine in rats with antibiotic-induced gut dysbiosis in the present study probably occurred due to increased proteolysis and decreased diversity of BCAA-utilizing bacteria. Additionally, isoleucine was positively correlated with g__Blautia, which has been reported to be a contributor to non-alcoholic fatty liver disease (Shen et al., 2017). After TACS treatment, the disorders related to isoleucine and g__Blautia were corrected, suggesting that TACS regulates intestinal bacteria dysbiosis-induced BCAA dysmetabolism by inhibiting proteolysis and modulating gut bacteria diversity.

The primary bile acids (cholic acid and chendeoxycholic acid) produced by the liver and metabolized into secondary bile acids by the 7alpha-dehydroxylating gut microbiota promote intestinal nutrient absorption and biliary cholesterol secretion to maintain bile acid homeostasis. The bile acids are essential for protecting the liver and other tissues and cells from cholesterol and bile acid toxicity (Ramirez-Perez et al., 2017; Chiang and Ferrell, 2018). Hepatic lipid and glucose metabolic homeostasis and energy metabolism are regulated by bile acids, which are endogenous ligands that can activate nuclear receptor farnesoid X receptor (FXRα) and membrane G protein–coupled bile acid receptor-1 (GPBAR1 or TGR5) (Lefebvre et al., 2009; Chiang, 2013). Evidence has shown (Ridlon et al., 2013) that bile acids are synthesized by classic pathways in healthy livers. When sufficient bile acids are secreted into the small intestine, they are metabolized into secondary bile acids, which maintain the stabilized gut environment and protect the gut against damage from inflammatory factor. During hepatic fibrosis, CYP7A1 expression in the classic pathway is inhibited by pro-inflammatory cytokines, which results in dysmetabolism of bile acids. The bile acid–metabolizing gut bacteria then have insufficient energy, leading to decreased bacterial abundance, which reduces the small intestine’s defense against inflammatory factors such as lipopolysaccharide (LPS), leading to inflammation-induced gut dysbiosis.

The increased levels of hyocholate, 3-oxocholate, and cholate in the antibiotic-only group showed that bile acid dysmetabolism occurred, and these bile acids were positively correlated with three increased genera (g__Blautia, g__Parabacteroides, and g__Intestinibacter). Our results were consistent with the previous study. However, the increase in this study of g__Blautia, a kind of 7alpha-dehydroxylating bacteria, exhibited the opposite trend compared to the decrease in the secondary bile acids. According to previous studies, increased g__Blautia in feces has been found in Hashimoto’s thyroiditis, hemorrhagic diarrhea, and non-alcoholic liver diseases (Shen et al., 2017; Zeineldin et al., 2018; Zhao F. et al., 2018). Therefore, we hypothesized that the elevated level of g__Blautia was probably influenced by interaction between bile acid dysmetabolism and inflammation. Research has shown that the gut microbial composition of rats fed a high-fat diet was influenced by abnormal bile acid metabolism (Zheng et al., 2017). In our TACS group, the g__Blautia was restored, and the levels of certain secondary bile acids were also restored, which indicated that TACS may regulate the composition and amount of gut bacteria by modulating bile acid dysmetabolisms and inflammation, making it possible to improve gut self-stabilization.

N-acetylputrescine, which is metabolized by gut bacteria, is a polyamine. Polyamines play a vital role in animal, plant, and bacteria growth by acting as growth factors, which involves binding to nucleic acids and promoting protein and nucleic acid synthesis; if polyamine biosynthesis is disrupted, cell growth is inhibited (Matsumoto et al., 2012; Sugiyama et al., 2016). Recent studies have shown that polyamines have anti-inflammatory effects and promote autophagy, and they can also be absorbed from the intestinal lumen to repair and maintain the intestinal mucosal barrier (Eisenberg et al., 2009; Timmons et al., 2012; Kibe et al., 2014). Meanwhile, the gut bacteria are involved in modulating intestinal luminal polyamines levels (Matsumoto and Benno, 2007). Betonicine (also named hydroxy stachydrine), a derivative of proline metabolites, inhibits increased capillary permeability and inflammatory exudation induced by acute inflammation via the arachidonate pathway, decreasing the levels of pro-inflammatory factors and decreasing cell permeability (Wang and Wang, 2012). Additionally, betonicine is known to be an osmoprotectant and cryoprotectant; it can protect bacteria from extreme temperature and high-salt environments in vitro and enhance bacterial growth (Bayles and Wilkinson, 2000; Bashir et al., 2014).

After antibiotic administration in the antibiotic-only group, the elevated N-acetylputrescine in feces may have been caused by both the enhanced polyamine catabolism due to the increased proteolysis and the intestinal mucosa injury caused by the LPS released by the damaged gut bacteria, which leads to the malabsorption of intestinal luminal polyamines. The level of proline, which was positively correlated with g__Parabacteroides in the antibiotic-only group, was increased. This may be the result of enriched proteolysis and inflammation caused by the antibiotics. Together, the disorder of arginine and proline metabolism and intestinal bacteria dysbiosis may have influenced the inhibition of enzymatic activity related to proline catabolism, leading to decreased levels of N4-acetylaminobutanoate and betonicine, both of which were negatively correlated with the inflammation-relevant genus g__Blautia. Nevertheless, the fact that this disorder could be ameliorated by TACS treatment suggested that TACS could correct the perturbation of arginine and proline metabolism by inhibiting proteolysis and regulating the enzymatic activity of certain reactions, resulting in modulation of the gut microbiota dysbiosis. Furthermore, the elevation of betonicine plays a protective role in the permeability of the intestinal mucosa, reducing inflammation.

Hypoxanthine is produced during purine catabolism. According to recent research, hypoxanthine improves hepatic pathological conditions and reduces the absorption of fat by inhibiting the activity of superoxide dismutase (SOD) and malondialdehyde (MDA), which demonstrates that hypoxanthine could help to reduce fat deposition (Xiaoyu et al., 2016). The relationship between hyperuricemia and the gut bacteria composition has already been studied, and it was found that a high-purine diet induced structural shifts in the gut microbiota and hyperuricemia in quails; this led to elevated LPS and increased xanthine oxidase (XO) activity, resulting in decreased hypoxanthine (Huang et al., 2015). The reduced hypoxanthine in the feces of the rats with antibiotic-induced gut dysbiosis was probably due to the increased intestinal LPS released by the dying bacteria. XO activity may have been increased, leading to a lower level of hypoxanthine and an increased level of uric acid. Additionally, evidence shows that XO activity influences superoxide radical production, which leads to oxidative stress that may induce or aggravate hepatic diseases (Masarone et al., 2018). According to the correlation analysis, the inflammation-relevant genera were negatively correlated with hypoxanthine, indicating that TACS recovered the gut microbiome diversity and repressed XO activity by improving the abnormal purine metabolism.

Tryptophan metabolites, which are metabolites that are strongly associated with gut bacteria, can be metabolized into indole by tryptophanase from gut bacteria, particularly Escherichia coli (Botsford and Demoss, 1972). Antibiotic treatment rapidly decreased the gut bacteria community, and the amount of tryptophanase also decreased, leading to disruption of the tryptophan–indole pathway and elevated fecal tryptophan. After TACS treatment, the level of tryptophan was slightly reduced, which indicates that TACS mildly regulated the disordered tryptophan metabolism, increasing the levels of some of the gut bacteria. This led to the regulation of the disordered tryptophan metabolism, improving the abnormal metabolic profile.

N-acetyl-D-glucosamine, which is related to amino sugar and nucleotide sugar metabolism, recovers intestinal epithelial cells, reduces permeability (thereby stabilizing the intestine barrier); inhibits intestinal bacteria from entering the circulation; and has antibacterial, antitumor, and antioxidant effects (Krenek et al., 2007; Ying et al., 2011). Antibiotic treatment disrupts the balance of the rat gut microbial environment, causing gut bacteria dysbiosis and metabolism disorders and resulting in a lower level of N-acetyl-D-glucosamine. This weakens the stability of the gut barrier and increases its permeability. After TACS treatment, N-acetyl-D-glucosamine was mildly upregulated, demonstrating that TACS regulates the disordered amino sugar and nucleotide sugar metabolism. The effects of TACS maintained the function of the intestinal epithelial cells, reduced the permeability (thereby stabilizing the intestinal barrier), and improved the gut microbial environment, improving the abnormal metabolic profile.

Gut microbiota abnormalities, which were associated with changes in bile acids, were demonstrated. Of the 34 altered metabolites, 12 were bile acids, which indicated that the bile acid pathway could be one of the most relevant metabolic pathways underlying dysbiosis and the therapeutic effects of TACS in antibiotic-induced gut microbiota dysbiosis. Therefore, the dysbiosis in our study might have been caused by perturbations in the bile acid metabolic pathway. We hypothesized that the intervening effects of TACS on gut microbiota dysbiosis may act by influencing bile acid synthesis and metabolism. Thus, to simulate how ligands act in a complex molecular network, we performed a molecular docking analysis using the main alkaloid components of TACS as ligands, with CYP27A1 (PDB-ID: 1mfx) as the target protein, which is a key enzyme in the bile acid synthesis pathway.

There are three alkaloid components of TACS that can regulate gut microbiota dysbiosis (Deng et al., 2009; Zhou et al., 2017; Kokoska et al., 2018). We performed a molecular docking analysis to explore the alkaloid components that affect the key enzyme of the bile acid metabolism pathway, CYP27A1. The results showed that the docking scores of the five main alkaloid components, berberine, palmatine, chelerythrine, jatrorrhizine, and dehydrocavidine, from high to low, were as follows: 9.5164, 8.5358, 6.8842, 4.4558, and 4.4558. The binding sites of berberine, palmatine, chelerythrine, jatrorrhizine, and dehydrocavidine were ASN-370, ALA-314 and ARG-446, GLY-437, VAL-367 and GLY-437, and ARG-94, respectively. According to these results, we concluded that berberine (9.5164), palmatine (8.5358), and chelerythrine (6.8842) might be the major active compounds for TACS in reducing antibiotic-induced gut microbiota dysbiosis by influencing CYP27A1 during bile acid synthesis. Recently, a study by Satchidananda Panda’s group (Zarrinpar et al., 2018) found that gut microbiome depletion using antibiotics led to decreased bile acids. This reduction upregulated the expression of FXR and bile acid transport protein, and increased bile acid transport into the blood, resulting in increased bile acid synthesis in the liver, probably by increasing CYP7A1 expression. The different findings in our study may be caused by the therapeutic drug used (TACS) and the different antibiotic regimen. This is the first study to find that alkaloid components of TACS might modulate bile acid metabolism by affecting CYP27A1. Further research still needs to be carried out to investigate the mode of action of these components regarding their intervening effects in rats with antibiotic-induced gut microbiota dysbiosis.

This research aimed to explore the intervening effects of TACS against antibiotic-induced gut microbiota dysbiosis in rats. We explored the mechanism of these alkaloid components regarding the prevention of gut microbiota dysbiosis using sequencing combined with metabolomics. We focused on bile acid metabolism, which might be the main metabolic pathway, and identified five alkaloid components of TACS that may affect bile acid metabolism, leading to further influences on host co-metabolism and gut microbiota dysbiosis. However, whether these components could exactly modulate the bile acid pathway via the key enzyme of the bile acid synthetic pathways (CYP27A1) remains unknown. This is what we plan to investigate further in our next study.