The herbal of Huang Lian (origin: Coptis chinensis Franch,; source: Mianyang, Sichuan) was purchased from Anhui Wugeng Traditional Chinese Medicine Pieces Co., Ltd. (Bozhou, China) and identified by Jilin Provincial Academy of Traditional Chinese Medicine according to the Chinese Pharmacopeia method. HL was stored in a ventilated and dry place. And HL extract was prepared as follows. The small pieces of HL were refluxed with 5-fold of 70% ethanol (1:5, w/v) twice for 2 h. Two batches of filtrates were combined, and pH value was adjusted to 1.5-2.0 with hydrochloric acid. The mixture was subsequently dissolved using NaCl. After being refrigerated for 48 h, precipitation was dried by vacuum concentration to obtain HL extract. Then extract was grinded into powder and stored in seal at 4°C. The freeze-dried powder was supplied by Jilin Provincial Academy of Traditional Chinses Medicine and was dissolved in distill water when used in the experiment.

The contents of marker components within HL extract were analyzed with high performance liquid chromatography (HPLC) method. 25 mL solution (methanol: hydrochlori acid = 100: 1) was added to 40 mg extract powder and the mixture was ultrasound for 20 min. Then mixture was diluted 10 times with solution (methanol: hydrochlori acid = 100: 1) and well vortexed. After centrifugation, the supernatant was analyzed using HPLC system to evaluate the content of epiberberine, coptidis, palmatine, and berberine. HPLC was performed on LC-Q10 AT system, consisting of a binary solvent delivery manager, an auto-sampler, and a UDV detector. Chromatographic separations were performed on an Agilent C18 column (4.6 × 250 mm,5 μm). Flow rate and column temperature were set at 1 ml/min and 25°C, respectively. The injection volume was 10 µL, and the solvent was filtered through 0.45 µm Millipore filter and degassed prior to use. Briefly, acetonitrile (A) and 0.05% potassium dihydrogen phosphate mixed with 0.4% sodium dodecyl sulfate (pH was regulated to 4.0 by phosphoric acid) (B) (50:50) were used as mobile phase. Ultraviolet detection with the wavelength at 345 nm and was used for the determination of epiberberine, coptidis, palmatine, and berberine.

Male C57BL/6 mice (5 weeks old) were purchased from Slake experimental Animal Co., Ltd. (Shanghai, China). All mice were housed in a temperature-controlled room (22 ± 3°C) under specific pathogen-free (SPF) conditions with a standard 12 h light/dark cycle. All animal experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Shanghai Jiao Tong University (No. A2020030). Because there is a certain risk of death in STZ injection, we used additional three and two mice to DM group and DM-HL group, respectively. When we performed the randomly grouping of the mice, the mice were weighted and sorted in order and divided into multiple groups according to body weight. During the experiment, none of mice died with STZ or HL treatment. Eventually, mice were divided into the NC group (n = 8), DM group (n = 11), and DM-HL group (n = 10). The mice in the NC group were fed a normal chow diet (NC, 12450J, 10% calories from fat, 70% calories from carbohydrate, 20% calories from protein, 3.85 kcal/g; produced by Fanbo Biotechnology Co., Ltd., Shanghai, China); other mice were fed a high-fat diet (HFD, D12492, 60% calories from fat, 20% calories from carbohydrate, 20% calories from protein, 5.24 kcal/g; produced by Fanbo Biotechnology Co., Ltd., Shanghai, China). Two weeks after dietary intervention commenced, the mice in the DM-HL group were orally gavaged daily with a dosage of 50 mg/kg HL extract, which continued until the end of the experiment. The mice in the NC and DM groups were gavaged daily with an equivalent volume of distilled water as a control. After the 4-week diet intervention commenced, the mice in the DM and DM-HL groups were injected intraperitoneally with a single dose of streptozotocin (STZ, 75 mg/kg) (S0130, Sigma-Aldrich, Germany) to construct the T2DM model. At the same time, the mice in the NC group received the equivalent volume of citric buffer to eliminate the effects of the injection procedure. All mice were allowed ad libitum access to water and food during the trial. The trial lasted for 6 weeks in total. Body weight was monitored weekly. Fecal samples were collected in week 6 and stored at −80°C for gut microbiota analysis.

All mice were fasted for 6 h before sampling. Blood samples were collected from the orbital vascular plexus. Serum samples were isolated from blood samples after centrifugation for 15 min at 4°C at 3000 g and then stored at −80°C. Liver, adipose tissues (epididymal, retroperitoneal, perirenal), and colon contents were weighed and collected immediately and stored at −80°C or in 4% paraformaldehyde for further analyses.

After 6 h of fasting, all mice were administered a glucose solution (2 g·kg^-1 body weight) through oral gavage. The glucose levels at 0 min (which is fasting glucose) and 15, 30, 60, 90, and 120 min after the glucose challenge were measured using a blood glucometer (ACCU-CHEK Performa, Roche, USA). Blood samples at 0 min, 15 min, and 60 min after the glucose challenge were collected from the tip of the tail vein. Serum samples were isolated for insulin detection after centrifugation at 4°C at 3000 rpm for 15 min. The insulin concentration was measured using an ultrasensitive mouse insulin ELISA kit (90080, Crystal Chem, USA).

Fresh epididymal fat and liver were fixed with 4% paraformaldehyde and embedded in paraffin. Then, 4-μm sections were stained with hematoxylin and eosin (H&E) (G1003, Wuhan Servicebio Technology Ltd., China). The adipocyte area and nonalcoholic fatty liver disease (NAFLD) activity score were analyzed using Image Pro Plus v6.0 software (Media Cybernetics Inc., Silver Springs, USA). For epididymal fat, at least 300 adipocytes of each mouse were required to assess the mean area of adipocytes under ×100 magnification. For the liver, three discontinuous scans of each mouse were used to assess the NAFLD activity score under ×100 magnification as previously described (29).

Epididymal fats were fixed overnight with 4% paraformaldehyde. In accordance with the standard process, the tissue was sectioned and stained with Oil Red O. The slices were imaged under microscope.

Frozen livers were homogenized using a tissue homogenizer in pro-cooled 1× RIPA buffer (ab156034, Abcam, Cambridge), containing protease and phosphatase inhibitors (P1045, Beyotime, China) (30). Then, homogenates were lysed for 5 min at 25 Hz/s using a TissueLyser (Qiagen, Germany). Protein was isolated from homogenates after centrifugation at 4°C at 12000 rpm for 10 min. A BCA protein Assay Kit (P001P, Beyotime, China) was used to quantify the protein concentration according to the manufacturer’s instructions. The protein was boiled at 100°C for 10 min after an equal volume of 2× SDS loading buffer (P0015B, Beyotime, China) was added. An equal amount of protein across all samples was separated on 10% SDS-PAGE gels (180-9117HA, Tanon, China) and diverted immediately to PVDF membranes (G6044-0.45, Servicebio, China). After 2 h of blocking in 8% skimmed milk at room temperature, the membranes were incubated with the primary antibodies overnight at 4°C. After washing 3 times with 0.1% TBST solution, the membranes were incubated with secondary antibodies for 2 h at room temperature. Eventually, the bands were visualized using the ECL stain kit (180-501, Tanon, China), and the intensity of the bands was determined by Image Pro Plus v6.0 software. Primary antibodies were as follows: IRS1 (1:2000, ab131487, Abcam), p-IRS1 (1:3000, 2386, cell signaling), Akt (1:2000, 9272, cell signaling), p-Akt (1:1000, 9271, cell signaling), Glut4 (1:2000, ab33780, Abcam), and GAPDH (1:3000, ab9485, Abcam).

Total RNA was extracted from liver tissues using a RNeasy minikit (74804, Qiagen, Germany). Then, reverse transcription was performed to synthesize cDNA using a SuperScript kit (18080-051, Invitrogen, USA). According to the manufacturer’s instructions, cDNA was then amplified using real-time PCR (qPCR) to determine the expression of Cyp27a1, Cyp7b1, Cyp7a1, and Cyp8b1 in a 20-μL reaction system containing cDNA template, forward primer, reverse primer, ddH₂O, and SYBR Green I PCR Supermix (Bio-Rad). The reaction conditions of forty cycles were as follows: 95°C for 20 s, 56°C for 30 s, and 72°C for 30 s, followed by plate reads for 5 s. Gene expression levels were calculated using the ΔΔC_T method and normalized to β-actin. The forward and reverse primer sequences of target genes and β-actin were listed as follows (31, 32). Cyp27a1 gene, F: 5'-CCAGGCACAGGAGAGTACG-3'; R: 5'-GGGCAAGTGCAGCACATAG-3'. Cyp7b1 gene, F: 5'-GGAGCCACGACCCTAGATG-3'; R: 5'-TGCCAAGATAAGGAAGCCAAC-3'. Cyp7a1 gene, F: 5'-AGCTCTGGAGGGAATGCCAT-3'; R: 5'-GAGCCGCAGAGCCTCCTT-3'. Cyp8b1 gene, F: 5'-CCTCTGGACAAGGGTTTTGTG-3'; R: 5'-GCCATCAAGGACGTCAGCA-3'. β-actin gene, F: 5'-GGCTGTATTCCCCTCCATCG-3'; R: 5'-CCAGTTGGTAACAATGCCATGT-3'.

Total DNA was extracted from week 6 fecal samples according to a procedure published previously (33). DNA concentration was determined by a NanoPhotometer and expressed in micrograms (μg). The DNA concentration across all samples was diluted to 10 ng/µL for subsequent analysis.

A plasmid containing the 16S full-length Roseburia inulinivorans strain was diluted successively to 10⁹,10⁸, 10⁷, 10⁶, 10⁵, 10⁴, 10³, and 10² copies/μL. According to the manufacturer’s instructions, qPCR was performed in a 20-µL reaction system containing a DNA template, Uni331F primer (5'-TCCTACGGGAGGCAGCAGT-3'), Uni797R primer (5'-GGACTACCAGGGTATCTAATCCTGTT-3') (34), ddH₂O, and SYBR Green I PCR Supermix (Bio-Rad). The reaction conditions of forty cycles were as follows: 95°C for 10 s, 65°C for 60 s, and 80°C for 5 s, followed by plate reads for 5 s. A standard curve was constructed according to the linearity between the copy number of plasmids and the C_T value. The copy number of DNA samples was calculated according to the standard curve.

An amplicon library was prepared by amplifying the V3-V4 region of the 16S rRNA gene according to previously published procedures (35). To profile the composition of microbial communities, high-throughput sequencing of the prepared amplicon library was performed on an Illumina Miseq platform (Illumina, Inc., USA).

Raw paired-end reads were analyzed with Quantitative Insights into Microbial Ecology2 software (QIIME2, v2021.4). After the removal of the adapters and primers with the “Cutadapt” plugin, DADA2 was used to obtain amplicon sequence variants (ASVs) through denoising, merging, and filtering. Then, a rooted phylogenetic tree was constructed with the “FastTree” plugin based on representative sequences of ASVs. Subsequently, the taxonomy of ASVs was identified based on the Silva138 16S rRNA database. In combination with the quantified total bacterial load, the relative abundance of ASVs was converted into absolute abundance for subsequent analysis (36, 37).

Redundancy analysis (RDA) was performed using Canoco 5 software (Microcomputer Power, USA) to identify key variations of ASVs responding to the T2DM model and HL treatment, which drive the segregation of the gut microbiota among groups. A heatmap representing the abundance variations of the key ASVs generated from the RDA model was constructed using the “Complexheatmap” R package. Pairwise comparisons (DM vs. NC, DM-HL vs. DM, DM-HL vs. NC) of ASVs were assessed using a Mann-Whitney U test. Microbial functional genes were predicted using PICRUSt2 based on the Greengenes database of 16S rRNA sequences (38) and the Kyoto Encyclopaedia of Genes and Genomes (KEGG) database.

The colon content was homogenized in 600 µL 2-chlorophenylalanine (4 ppm) methanol and vortexed for 30 s. Then, homogenates were sonicated at room temperature for 10 min. The supernatants were collected after centrifugation (4°C, 12000 rpm, 10 min) and filtration with a 0.22-µm membrane for high-performance liquid chromatography-mass spectrometry analysis (HPLC-MS/MS). Quality control (QC) samples were prepared with the same procedures. HPLC-MS/MS was performed by Suzhou PANOMIX Biomedical Tech Co., Ltd. (Suzhou, China). Chromatographic separation was performed with an ACQUITY UPLC® HSS T3 (150 × 2.1 mm, 1.8 μm, water) column maintained at 40°C. Data-dependent acquisition in the MS/MS experiment was operated with an HCD scan. The raw data of metabolites were converted into mzXML format with Proteowizard software. Then, perk identification, filtration, and alignment were performed using the “XCMS” R package. The output metabolites were identified based on the Metlin metabolites database (metlin.scripps.edu), MoNA metabolite database (mona.fiehnlab.ucdavis.edu), and in-house MS2 database.

Variable importance in projection (VIP) scores calculated by the orthogonal projection to latent structure-discriminant analysis (OPLS-DA) model were used to estimate the importance of each variable to drive the segregation of the metabolome among the two groups. Through pairwise comparison between groups (DM vs. NC, DM-HL vs. DM, DM-HL vs. NC), differential metabolites were selected based on the following criteria: 1) VIP > 1.5 in the OPLS-DA model, 2) fold-change > 2, and 3) p < 0.05 (including methods used for testing and p value correction). MetaboAnalyst was used for the analysis of pathway enrichment (MetaboAnalyst).

Statistical analysis was performed using GraphPad Prism version 8.0 (GraphPad Software, Inc., USA). Gut microbiota data were exhibited by a box plot, and significance was analyzed using one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons. Physiological data, metabolites, and functional genes were expressed as the mean ± SEM, and significant differences were evaluated by one-way ANOVA or an unpaired t test with Welch’s correction. Spearman correlations were constructed with R software (v4.0.5). P < 0.05 is considered to be statically significant.

The chromatographic peaks of the main chemical constituents within HL extract were shown in Supplementary Figure S1 . The contents of epiberberine, coptidis, palmatine, and berberine were determined to be 1.18%, 6.71%, 5.34%, and 44.34%, respectively. In short, the content of total alkaloids within HL extract was determined to be 69.43% using spectrophotometry.

We used HFD with one streptozotocin (STZ, 75 mg/kg) injection to induce T2DM in C57BL/6 mice. One group of these mice was administered HL (50 mg/kg·day) by oral gavage after 2 weeks of feeding with HFD, which continued for an additional 4 weeks (DM-HL group), and the others were given distilled water as a disease control (DM group). At the end of the trial, fasting glucose and glucose levels during the oral glucose tolerance test (OGTT) were significantly higher in the DM group compared to that in the healthy mice (NC group) ( Figures 1A, B ). Fasting insulin, the area under the curve (AUC) of insulin during OGTT, HOMA-IR, and the insulin resistance index were also significantly higher in the DM group than in the NC group ( Figures 1C–F ). It was indicated that the mice in the DM group suffered glucose tolerance impairment and insulin resistance. Notably, although HL treatment had no effect on fasting glucose, fasting insulin, HOMA-IR, and insulin levels during OGTT, the mice in the DM-HL group had significantly reduced insulin resistance index and blood glucose levels (from the 30^th min during OGTT) compared to the DM group ( Figures 1A–F ), suggesting that HL treatment may increase insulin sensitivity but not insulin secretion. In the insulin signaling pathway, binding of insulin to the insulin receptor would induce the phosphorylation of insulin receptor substrates (p-IRS1), phosphorylate the downstream signaling substance Akt (p-Akt), and then increase the level of glucose transporter (Glut4), leading to glucose uptake (39). Thus, we measured the levels of p-IRS1, p-Akt, and Glut4 in the liver to evaluate the impact of HL on insulin sensitivity. We found that HL treatment reversed the down-regulation of the phosphorylation of IRS1 and Akt induced by HFD/STZ, and the level of Glut4 was dramatically higher in the DM-HL group than in the DM group ( Figure 1G ). These results suggested that HL intervention alleviated glucose tolerance by stimulating insulin sensitivity and glucose uptake.

In addition to impaired glucose tolerance, we also observed that the mice in the DM group had significantly higher body weight gain and over-accumulation of fat ( Figure 2 , Supplementary Figure S2 ). Across the experiment, HL treatment effectively inhibited excess body weight gain and successfully blunted lipid accumulation, as reflected by significantly lower body weight gain from week 3 during the experiment, as well as reduced adipocyte size and lipid droplets in epididymal fat and the NAFLD activity score in the DM-HL group than in the DM group ( Figure 2 , Supplementary Figure S2 ). These results suggest that HL treatment effectively inhibited obesity and fat accumulation induced by HFD/STZ.

To determine the effects of HL treatment on the gut microbiota in T2DM mice, we collected fecal samples from all mice at the end of the trial for gut microbiota analyses. First, the total bacterial loads were determined by qPCR of the 16S rRNA gene. We found that the total bacterial loads of mice in the DM group and DM-HL group decreased by 3.8 times and 3.3 times, respectively, compared to the NC group, indicating significantly lower total bacterial loads in all HFD/STZ-treated mice ( Supplementary Figure S3 ). Next, we sequenced the V3-V4 region of the 16S rRNA gene to analyze the structure and composition of the gut microbiota. Due to significant differences in the total bacterial loads among the three groups, the relative abundance of ASVs could not accurately reflect the alteration of the gut microbial structure (37). Therefore, for further analyses, we converted the relative abundance to the absolute abundance of ASVs based on the total bacterial loads of each sample. The richness and diversity of the gut microbiota, represented by the number of species and Shannon index, were similar between the DM group and NC group, while the mice in the DM-HL group had markedly reduced richness and diversity of gut microbiota, suggesting that HL may have antimicrobial activity against specific bacteria ( Figure 3A ). Principal coordinate analysis (PCoA) based on the Bray-Curtis distance at the ASV level showed obvious discriminations of the gut microbial structure among the three groups ( Figure 3B ). The structure of the gut microbiota in the DM group significantly separated from that in the NC group along PC1 (explained 34.78% of the total variance) and PC2 (explained 27.14% of the total variance). Specifically, compared to the DM group, the samples in the DM-HL group were more distinct from those in the NC group along PC1, while the samples in the DM-HL group were closer to those in the NC group along PC2. Notably, the Bray-Curtis distance of the gut microbiota in the mice between the DM-HL group and NC group was significantly higher than that between the DM group and NC group ( Figure 3C ). In short, HL did not reverse the gut microbiota disrupted by HFD/STZ treatment but constructed a strikingly distinct gut microbiota.

Redundancy analysis (RDA) based on the Monte Carlo permutation procedure (MCPP, p = 0.0001) also showed a significant effect of the T2DM model and HL treatment on the gut microbiota, which was consistent with the PCoA results. Then, we applied RDA to identify gut microbes responding to the T2DM model and/or HL treatment. Eventually, we selected 101 ASVs as key variations that accounted for at least 46.8% of the variability explained by both axes ( Supplementary Figure S4 ). Compared to the NC group, the DM group significantly altered 61 ASVs (10 ASVs increased and 51 ASVs decreased). Among these 61 ASVs altered by the DM group, HL treatment changed 42 ASVs (4 ASVs increased and 38 ASVs decreased) and about one-third of them (13 out of 42 ASVs) were reversed to the same direction as the ASVs observed in the NC group. Of the 13 reversed ASVs, four were enriched (belonging to Bacteroides, Enterococcus, Lachnospiraceae GCA-900066575) and nine were reduced (belonging to Lachnospiraceae, Ruminococcaceae, Lactobacillus, Clostridia UCG-014, Oscillibacter) in the DM-HL group. The rest of the HL-changed ASVs (29 out of 42 ASVs) were significantly inhibited by HL treatment and most of them were eliminated in the DM-HL group except for ASV42 (belonging to Enterornabdus). In addition, the abundance of 40 ASVs belonging to various taxa were affected only by HL treatment, including the enrichment of 15 ASVs and the elimination of 25 ASVs in the DM-HL group. Overall, 76 out of 101 identified key ASVs were almost undetectable in the DM-HL group, including 34 ASVs belonging to the Lachnospiraceae family. Only 25 out of 101 identified key ASVs were predominant in the DM-HL group that accounted for about 48.6% of the total bacterial loads (12.1% in the NC group; 12.9% in the DM group), of which the top five abundant ASVs belonged to the Bacteroides (ASV1, ASV5, ASV8), Parasutterella (ASV9), and Clostridium innocuum (ASV17) groups. Moreover, we noticed that some ASVs in the same family or genus showed different behaviors. For example, among the 44 ASVs in the Lachnospiraceae family, 13 ASVs were increased while 31 ASVs were decreased in the DM-HL group compared with the DM group. Similar results were also found in the genera Bacteroides and Roseburia, indicating that the response of bacteria is species or even strain-specific ( Figure 3D ). These results showed a significantly distinct gut microbiota with lower total bacterial loads and diversity that developed in the DM-HL group.

To determine how the T2DM model and HL treatment impact the metabolites of the gut microbiota in mice, the untargeted metabolites of the colon of all mice were tested by HPLC-MS/MS. In total, we identified 541 metabolites under polar ionic mode. The principal-component analysis (PCA) and orthogonal projection to latent structure-discriminant analysis (OPLS-DA) exhibited obvious discriminations of metabolites among the three groups ( Figure 4A , Figure S5A ). Subsequently, we determined a total of 119 differential metabolites through pairwise comparisons between groups (VIP > 1.5 in the OPLS-DA model, fold-change > 2, and p < 0.05) ( Supplementary Figure S5B ; Supplementary Table S1 ). The number of differential metabolites in the mice between the DM-HL group and NC group was much higher than that between the DM group and NC group (71 vs. 37).

Among 119 differential metabolites, we recognized multiple herbal ingredients based on the TCMID database (http://www.megabionet.org/tcmid/), including BBR, tyrosol, yamogenin, 26-hydroxyecdysone, and 2,4-dihydroxyacetophenone, which may be present in the HL extract that was significantly enriched in the DM-HL group compared with the DM group ( Supplementary Figure S6 ). Pathway enrichment analysis revealed that the differential metabolites between the DM and NC groups were mainly associated with nutrient and energy metabolisms, such as carbohydrate, lipid, amino acid, nucleotide, and cofactor and vitamin metabolisms. Notably, taurine and hypotaurine metabolism, and primary BA biosynthesis, which were highly related to microbial BA metabolism, were among the top 5 pathways enriched in the DM-HL group compared to the DM and NC groups ( Figures 4B, C ; Supplementary Figure S7 ). It has been reported that members of the gut microbiota can deconjugate the conjugated primary BAs to release free primary BAs and amino acids (taurine or glycine) and subsequently modify the free primary BAs to produce a broad range of secondary BAs (40). To assess the impact of HL on BA metabolism, we identified 12 metabolites in the gut that were involved in microbial BA metabolism ( Figure 4D ; Supplementary Figures S8C, D ). For the conjugated primary BAs, the levels of GCA and GCDCA were similar between the DM group and DM-HL group, while the HL treatment resulted in significantly higher levels of TCA and TCDCA compared to the DM group. As a result of microbial deconjugation, the mice in the DM-HL group had higher levels of taurine and UDCA compared with the DM group, in combination with a significantly decreased level of CDCA. For the secondary BAs, the levels of THCA, DCA, and ACA were significantly enriched in the DM-HL group compared to the DM group. No significant difference was observed in LCA between groups, but the downstream metabolite of iso-LCA was reduced after HL treatment. These results showed that HL treatment dramatically changed the composition of the BA pool, leading to increased BA-related metabolites except for CDCA, LCA, and iso-LCA, which were the downstream metabolites of TCDCA.

To investigate the potential contribution of BA-related metabolites to the benefits of HL, we analyzed the correlations between BA-related metabolites and T2DM-related parameters using Spearman’s correlation ( Figure 4E ). We found that 11 out of 12 BA-related metabolites were significantly correlated with at least one parameter. Among these 11 metabolites, five metabolites enriched in the DM-HL group, including ACA, UDCA, TCDCA, taurine, and TCA, were negatively correlated with both body weight gain and adipocyte size. Moreover, ACA was also negatively correlated with the AUC of glucose during OGTT and insulin resistance. Two metabolites that markedly decreased in the DM-HL group, i.e., CDCA and iso-LCA, were positively correlated with body weight gain and adipocyte size. These correlations suggested that BA-related metabolites may be key mediators of the microbiota-host interaction during HL treatment, leading to the protective effect of HL on HFD/STZ-induced T2DM.

To further determine the role of the altered gut microbiota in mediating the composition of BAs that are potentially protective against T2DM, we investigated the absolute abundance of putative functional genes of gut microbiota involved in BA metabolism using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) analysis. We observed that the abundance of the cbh gene, which is responsible for the deconjugation of BAs, increased markedly following HL treatment, and the abundance of bai operons (baiB, baiCD, baiH) that encoded several essential proteins for BA dehydroxylation was also higher in the DM-HL group than in the DM group ( Figures 5A, B ). It is suggested that HL treatment might improve the capability of microbial BA metabolism. Then, Spearman’s correlation analysis was performed to identify the potential relationships between key microbes and BA-related metabolites ( Figure 5C ). In total, 11 out of 12 metabolites were correlated with at least one ASV. Among these 11 metabolites, metabolites of ACA, taurine, THCA, TCA, and TCDCA, showed a significantly positive correlation with ASVs that were enriched in the DM-HL group (mainly belonging to Marvinbryantia, Clostridium, Coprobacillus, Roseburia, Parasutterella, Bacteroides) and negative correlations with ASVs that were suppressed by HL treatment (mainly belonging to unclassified Lachnospiraceae, Oscillibacter). On the contrary, CDCA and iso-LCA showed the opposite trends of correlations with the gut microbiota than the above-mentioned five metabolites. Altogether, the results indicated that the enriched gut microbes in the DM-HL group might be essential for the alteration of BA composition, possibly through the improvement of microbial BA metabolism.