HFD (17.7% sucrose, 17.7% fructose, 19.4% protein and 40% fat) was obtained from Beijing Huafukang Bioscience Co., Ltd. (Beijing, China). Triglyceride (TG), total cholesterol (TC), aspartate aminotransferase (AST), alanine aminotransferase (ALT), superoxide dismutase (SOD), methane dicarboxylic aldehyde (MDA), and glutathione peroxidase (GSH-Px) assay kits test kits were obtained from Nanjing Jiancheng Biological Engineering Institute (Nanjing, China). Oil Red O Staining kit was purchased from Solarbio Biotechnology Co., Ltd. (Beijing, China). Rat insulin enzyme-linked immunosorbent assay (ELISA) kit was obtained from Multi Science Biotechnology Co., Ltd. (Hangzhou, China). Reference standards of paeoniflorin, baicalin, saikosaponin A, saikosaponin D, emodin, synephrine, succinic acid, 6-gingerol and oleanolic acid were obtained from Sichuan Weikeqi Biological Technology Co., Ltd. (Sichuan, China).

The dosage of each drug in the DCH preparation used in this study was in accordance with the record in “Treatise on Febrile Diseases”, written by Zhang Zhongjing in 200 C.E.–210 C.E.. DCH contained: 12 g of Bupleurum chinense DC. (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1907026), 9 g of Scutellariae baicalensis Georgi (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1906002), 9 g of Paeonia lactiflora Pall. (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1901016), 9 g of Pinellia ternata (Thunb.) Makino (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1812006), 9 g of Citrus × aurantium L. (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1901016), 15 g of Zingiber officinale Roscoe (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1903002), 12 g of Ziziphus jujuba Mill. (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1907007), and 6 g of Rheum officinale Baill. (Tianjin traditional Chinese Medicine prepared pieces Co., Ltd, Tianjin, China, Batch number: 1902005). All herbs were authenticated by Pharmacist Li Wang in Department of Pharmacy of the Tianjin Second People’s Hospital. The above herbs were then soaked in 300 ml water for 30 min and decocted for 30 min to obtain the aqueous extract of DCH. The aqueous extract of DCH was filtered and concentrated to a density of 0.8 g crude herb/ml.

Quality control of DCH was performed using high performance liquid chromatography (HPLC; UltiMate 3,000, Thermo Scientific^™, USA) coupled with mass spectrometer (MS; Q Exactive^™, Thermo Scientific^™, USA). The chromatographic conditions were as follows: The chromatographic column was an Eclipse Plus C₁₈ RRHD column (2.1 × 100 mm, 1.8 µm). The column temperature was maintained at 40°C and the flow rate was 0.3 ml/min. 0.1% formic acid aqueous solution (A) and acetonitrile (B) were used as the mobile phases and the injection volume was 5 μl. The mobile phase conditions were: 0 min, 5% B; 1 min, 5% B; 9 min 80% B; 12 min 100% B; 14 min 100% B; 14.1 min 5% B; 16 min 5% B. A mass spectrometer equipped with an electrospray ionization (ESI) source was used for both positive and negative ionization scan modes (m/z ranges from 100 to 1,500). The detailed parameters of MS were: spray voltage of 3,500 V (positive mode) and 3,000 V (negative mode), capillary temperature at 320°C (both positive and negative modes), sheath gas flow rate of 30 arbitrary units (both positive and negative modes), and auxiliary gas flow rate of 10 arbitrary units (both positive and negative modes).

6-8 weeks old male Sprague-Dawley (SD) rats weighing 190–210 g, were purchased from Huafukang Animal Co., Ltd. (Beijing, China). They were acclimated in a controlled environment (12 h light/dark cycle, 21 ± 2°C with a relative humidity of 45 ± 10%) with ad libitum access to food and water. All animal experiments were approved by the Animal Ethics Committee at Tianjin University of Traditional Chinese Medicine.

After the acclimatization for 1 week, all animals were randomly divided into four groups (n = 10): control, model, positive control and DCH. Rats in the control group were fed with standard laboratory chow, while rats in the model, positive control, and DCH groups received HFD for 12 weeks to induce NAFLD (Li et al., 2020). The ingredients of standard chow and HFD were showed in Table 1. After 4 weeks of HFD feeding, rats in the positive control and DCH groups were orally treated with metformin (200 mg/kg rat weight) (Zheng et al., 2014) and DCH (8 g/kg rat weight), respectively, once per day for 8 weeks. Whereas, rats in the control and model groups received an equivalent volume of saline. At the end of 8 weeks of metformin and DCH treatment, rats were sacrificed and their livers weighed. Liver index was calculated based on the percentage of liver to body weight.

Serum samples were collected at the end of 8 weeks of metformin and DCH treatment for biochemical analysis. Briefly, rats were anaesthetized and blood was harvested by a syringe from the aorta abdominalis. The blood was then centrifuged at 3,000 rpm for 15 min to isolate the serum. The levels of TG, TC, ALT, and AST in serum were analyzed according to the manufacturer’s instructions provided by Nanjing Jiancheng Biological Engineering Institute (Nanjing, China) and the absorbance value was detected using a microplate reader (Varioskan Flash, Thermo Fisher, Massachusetts, USA).

After euthanasia, the rat’s livers were immediately removed and fixed in 10% formalin, dehydrated, and embedded in paraffin wax. The tissues were then cut into 5 µm sections using a microtome (RM2125, Leica, Buffalo Grove, USA) and were subsequently stained with hematoxylin and eosin (H&E), as has been previously described (Cui et al., 2018). The pathological severities of steatosis, lobular inflammation, and hepatocyte ballooning were determined using the NAFLD activity score (NAS) as described in our previous publication (Cui et al., 2020).

Livers were sectioned into 20 μm thick coronal sections using a microtome-cryostat (CM3050S, Leica, Buffalo Grove, USA). The sections were then stained with Oil Red O, following the manufacturer’s instructions. The staining of lipid drops by Oil Red O was quantified using Image J to obtain the integrated optical density (IOD). The mean optical density (MOD) was calculated based on the ratio of IOD to the sum area.

OGTT was conducted at the end of 8 weeks of metformin and DCH treatment as has been previously described (Cummings et al., 2011). Briefly, rats were fasted for 16 h and the levels of fasting blood glucose (FBG) were determined. Then, rats received 50% glucose solution (1 g/kg) intragastrically and the blood glucose levels were measured at 30, 60 and 120 min post glucose solution treatment. The area under the curve (AUC) of OGTT was then calculated.

Determination of Homeostatic Model Assessment of Insulin Resistance (HOMA-IR).

Rats were fasted as detailed above and the levels of fasting insulin (FINS) in the serum were measured using ELISA according to the manufacturer’s instructions (Multi Science Biotechnology Co., Ltd., China). Additionally, the level of FBG was measured. The HOMA-IR was calculated using the following formula: HOMA-IR = FBG (mmol/L) × FINS (μU/ml)/22.5.

0.1 g of liver tissues were immersed in 900 µl normal saline followed by ultrasonic trituration to obtain liver tissue homogenates. The homogenates were then centrifugated at 3,000 rpm for 15 min and their supernatants were used to measure the MDA level as well as the SOD and GSH-Px activities according to the manufacturer’s instructions provided by Nanjing Jiancheng Biological Engineering Institute (Nanjing, China).

At the end of 8 weeks of DCH treatment, feces from the control, model, and DCH groups were simultaneously obtained under sterile conditions in a laminar flow hood. Fecal total DNAs were extracted and their purities and concentrations were measured by agarose gel electrophoresis. The DNA samples were then diluted to 1 ng/μl and polymerase chain reaction (PCR) was conducted to amplify the V4 region of 16S rRNA of DNA samples using specific primers with the barcode (forward: GTGCCAGCMGCCGCGGTAA reverse: GGACTACHVGGGTWTCTAAT). The PCR amplification mixture of each sample consisted of 15 μl of Phusion® High-Fidelity PCR Master Mix (New England Biolabs), 0.2 μM of forward and reverse primers, and 10 ng of template DNA. The PCR products were then quantified by 2% agarose gel electrophoresis and purified using Qiagen Gel Extraction Kit (Qiagen, Germany). The sequencing libraries were generated using TruSeqRDNA PCR Free Sample Preparation Kit (Illumina, United States) and then sequenced on the NovaSeq6000 platform to generate paired-end reads.

Paired-end reads obtained from 16S rRNA sequencing were assigned to samples, truncated by cutting off the barcode and primer sequence, and merged by the FLASH V1.2.7 analysis tool (http://ccb.jhu.edu/software/FLASH/) to obtain the raw tags. The raw tags were then filtered to generate the clean tags according to the QIIME V1.9.1 quality controlled process (http://qiime.org/scripts/split_libraries_fastq.html). The chimera sequences in clean tags were detected and removed to obtain the effective tags using the UCHIME algorithm (http://www.drive5.com/usearch/manual/uchime_algo.html). The sequences of effective tags with ≥97% similarity were then assigned to the same OTUs using the Uparse V7.0.1001 software (http://drive5.com/uparse/). Representative sequences for each OTU were then screened for further annotation via the Silva database (http://www.arb-silva.de/). The relative abundances of OTUs were normalized using a standard of sequence number corresponding to the sample with the least sequences. The normalized data were then used for alpha diversity and beta diversity analysis. The gene family abundance of 16S rRNA sequencing data was predicted according to the Phylogenetic Investigation of Communities by Reconstruction of Unobserved States database (PICRUSt).

At the end of 8 weeks of DCH treatment, serum samples from the control, model and DCH groups were collected for metabolomics analysis. The changes in serum metabolites were screened using liquid chromatography–mass spectrometry (LC–MS) and the data were analyzed using a method described in our previous study (Cui et al., 2020).

Spearman’s correlation analysis was conducted to analyze the relationship between physiological data (body weight, liver index, TG, TC, AST, ALT, AUC of OGTT, FINS, and HOMA-IR), differential serum metabolites, and gut microbiota at genus level in the control, model and DCH groups.

All data were reported as the mean ± standard deviation (mean ± SD) for the independent experiments. Statistical differences between the experimental groups were examined using the analysis of variance (ANOVA) and SPSS software, version 20.0. A p-value < 0.05 was considered statistically significant. Curve fitting was performed using the GraphPad Prism 5 software.

HPLC-MS analysis was conducted to investigate the chemical profiles of DCH. Synephrine, succinic acid, paeoniflorin, baicalin, saikosaponin A, 6-gingerol, saikosaponin D, emodin and oleanolic acid were used as the reference standards to validate the main bioactive compounds in DCH. The molecular formulas and chemical structures of these reference standards were shown in Supplementary Figure S1. The typical based peak intensity (BPI) chromatograms of DCH and the reference standards were analyzed in both positive and negative modes (Supplementary Figure S2) and their characteristic fragment ions were shown in Supplementary Table S1. Paeoniflorin in Paeonia lactiflora Pall., baicalin in Scutellariae baicalensis Georgi, saikosaponin A and saikosaponin D in Bupleurum chinense DC., emodin in Rheum officinale Baill., synephrine in Citrus × aurantium L., succinic acid in Pinellia ternata (Thunb.) Makino, 6-gingerol in Zingiber officinale Roscoe and oleanolic acid in Ziziphus jujuba Mill. were identified as the preeminent bioactive compounds in DCH.

After 8 weeks of metformin and DCH treatment, the body weight (p < 0.01, Figure 1A), liver index (p < 0.01, Figure 1B) and serum levels of TG, TC, AST, and ALT (p < 0.01, respectively, Table 2) were significantly increased in model group compared with the control group. Whereas, metformin and DCH treatment significantly decreased the body weight (p < 0.01, respectively, Figure 1A), liver index (p < 0.05, respectively, Figure 1B), and serum levels of TG, TC, AST, and ALT (p < 0.01, respectively, Table 2) compared to the NAFLD model rats. There were no significant differences in body weight, liver index and levels of serum ALT and TG between the positive control and DCH groups (Table 2), however, the levels of serum TC and AST were significantly higher in DCH group compared with those in positive control group (p < 0.01, respectively, Table 2). Notable steatosis of hepatocytes accompanied by hepatocytes ballooning and lobular inflammation was revealed by H&E staining in the model group, while metformin and DCH treatment improved the hepatic steatosis, hepatocytes ballooning, and lobular inflammation (Figure 1C). Likewise, the NAS score was significantly higher in the model group compared with the control (p < 0.01, Figure 1E) and H&E staining of the positive control and DCH groups showed a lower NAS score compared with the model group (p < 0.01, respectively, Figure 1E). There were no significant differences in NAS between the positive control and DCH groups (Figure 1E). Oil Red O staining also confirmed hepatic steatosis and increased lipid deposition in the model group (p < 0.01, Figures 1D,F), whereas the numbers of lipid-loaded hepatocytes were significantly decreased in metformin and DCH treated rats (p < 0.01, respectively. Figures 1D,F). There were no significant differences in the numbers of lipid-loaded hepatocytes between the positive control and DCH groups (Figure 1F).

IR is also an important clinical outcome of NAFLD and as such we measured the effects of DCH on IR. The AUC of OGTT was significantly increased in model group compared with the control group (p < 0.01, Figures 1G,H), whereas metformin and DCH treated rats displayed a lower AUC of OGTT compared with the NAFLD model rats (p < 0.01, respectively, Figures 1G,H). The AUC of OGTT was lower in the DCH group compared with the positive group (p < 0.05, Figures 1G,H). Additionally, the levels of FINS and HOMA-IR were higher in the model group compared with the control group (p < 0.01, Table 3) and were lower in the positive control and DCH groups compared to the model group (p < 0.01, respectively, Table 3). There were no significant differences in FINS and HOMA-IR between the positive control and DCH groups (Table 3).

We also investigated the anti-oxidative effects of DCH on NAFLD. Our results showed significantly lower activities of SOD, GSH-Px, and higher levels of MDA in the NAFLD model rats compared to rats in the control group (p < 0.01, respectively, Table 4). Compared with the model group, the activities of SOD (p < 0.05, respectively, Table 4) and GSH-Px (p < 0.01, respectively, Table 4) were increased and the level of MDA (p < 0.01, respectively, Table 4) was decreased in the positive control and DCH groups.There were no significant differences in SOD, GSH-Px and MDA between the positive control and DCH groups (Table 4).

16S rRNA sequencing was performed to detect gut microbiota changes in NAFLD model rats post DCH treatment. Overall, 2,196,489 useable reads and 1,174 OTUs were obtained from 30 samples. There were no significant differences in Chao1 index and observed species number between control and model groups, indicating that the richness of gut microbiota community was not changed after HFD treatment (Figures 2A,B). The Chao1 index was lower in DCH group than that in the model group, whereas there was no significant differences in observed species number between model and DCH groups (Figures 2A,B). The Shannon and Simpson indexes were higher in the model group compared with the control group and were lower in DCH group compared with the model group, indicating that the alpha diversity of gut microbiota community was increased in the model group compared with the control group and DCH treatment decreased the alpha diversity of gut microbiota community in NAFLD model rats (Figures 2C,D). Principal co-ordinates analysis (PCoA) showed significant variations of gut microbiota in each group, with a shorter distance between the control and DCH groups than between the model and DCH groups (Figure 2E). Likewise, system clustering tree analysis indicated that the distance from the control group to the DCH group was closer than either to the model (Figure 2F).

We further investigated the changes in the relative abundances of gut microbiota species. A venn diagram of the three groups demonstrated that 660 OTUs overlapped among all groups; 714 OTUs were present in the control and model groups; 775 in the control and DCH groups; and 789 in the model and DCH groups (Figure 2G). At the phylum level, Firmicutes and Bacteroidetes were the most abundant phyla among all samples (Figure 2H). The Firmicutes to Bacteroidetes (F to B) ratio was higher in the model group than in the control group (p < 0.01, Figure 2I), whereupon the F to B ratio was decreased after DCH treatment (p < 0.05, Figure 2I). At the genus level, the relative abundances of Romboutsia, Bacteroides, Lactobacillus, Akkermansia, and Turicibacter (p < 0.01, p < 0.01, p < 0.01, p < 0.01 and p < 0.05, respectively, Figure 2J) were signifiacntly lower and the relative abundances of Lachnoclostridium, unidentified_Enterobacteriaceae, Allobaculum, and Enterococcus (p < 0.05, p < 0.05, p < 0.01 and p < 0.01, respectively, Figure 2J) were significantly higher in the model group than that in the control group. Compared with the model group, DCH treatment significantly increased the relative abundances of Romboutsia, Bacteroides, Lactobacillus, Akkermansia, and Turicibacter (p < 0.01, p < 0.05, p < 0.01, p < 0.05 and p < 0.05, respectively, Figure 2J) and significantly decreased the relative abundances of Lachnoclostridium and unidentified_Enterobacteriaceae (p < 0.01 and p < 0.05, respectively, Figure 2J). Additionally, we predicted the possible pathways related to the differential gut microbiota at the genus level by PICRUSt analysis. The top 10 terms of metabolic pathways with the highest proportion and a p-value < 0.05 were listed in Figure 2K (control group vs. model group) and Figure 2L (model group vs. DCH group). Proportions of metabolic pathways that were increased in the model group but decreased in DCH group, or vice versa, were considered as differential pathways. The abundances of pentose phosphate, glycine/serine/threonine and arachidonic acid (AA) metabolic pathways as well as the valine, leucine and isoleucine biosynthesis pathways were all increased in the model group compared with the control group (Figure 2K). Conversely, the abundance of glycerophospholipid metabolism pathway was decreased in model group compared with the control group (Figure 2K). For the DCH group, the abundances of pentose phosphate, glycine/serine/threonine, and AA metabolic pathways along with the valine, leucine and isoleucine biosynthesis pathways were lower than the model group and the abundance of glycerophospholipid metabolism pathway was higher (Figure 2L).

The changes of metabolites in serum were investigated using untargeted metabolomics. According to the principle component analysis (PCA) model, a clear group separation could be observed between the control and model groups, while the distinction between the model and DCH groups was unclear (Figures 3A,B). Therefore, we performed orthogonal partial least squares discriminant analysis (OPLS-DA) to further visualize the metabolic alterations occurring between the control group and model group as well as between the model group and the DCH group. The OPLS-DA models showed significant distinctions of metabolomic data between the control group and the model group as well as between the model group and the DCH group (Figures 3C,E). The over-fitting in the OPLS-DA model was controlled using seven-round cross validation and 200 repetitions of RPT based on the R² and Q² values. The R² and Q² values of OPLS-DA model in the comparison of control and model groups were 0.686 and −1.08, respectively. The R² and Q² values in the OPLS-DA model comparing the model group with the DCH group were 0.22 and −0.624, respectively. These results indicated that the OPLS-DA models were robust (Figures 3D,F). Metabolites with a VIP >1 and p < 0.05 between control and model groups or between DCH and model groups were considered to be differential metabolites. The numbers of differential metabolites between the control group and the model group as well as between the model group and the DCH group were visualized in Figure 3G. Compared with the control group, the levels of phosphatidylcholine (PC), rumenic acid, linoleic acid, eicosapentaenoic acid (EPA), L-threonine, gluconic acid, and lacto-N-tetraose were decreased and the levels of L-proline, L-lysine, L-isoleucine, L-valine, L-arginine, L-leucine, glycocholic acid, uric acid, creatinine, stearic acid, ursodeoxycholic acid, phosphatidylethanolamine (PE), L-tryptophan, 12(R)-HETE, 5-HPETE, and glycine were increased in the model group. For the DCH group compared with the model group: the levels of linoleic acid, PC, L-threonine, and rumenic acid were increased and the levels of L-isoleucine, L-valine, L-arginine, L-leucine, stearic acid, indoleacrylic acid, THTC, 12(R)-HETE, 5-HPETE, glycine, uric acid, and PE were decreased (Table 5).

In addition, differential metabolites with a fold change (FC) greater than 1.2 or a FC of less than 0.8 were analyzed using MetaboAnalyst software to screen for significant metabolic pathways (p < 0.05, impact value >0.10). Tryptophan, AA, linoleic acid, glycerophospholipid, and glycine/serine/threonine metabolisms were identified to be significant metabolic pathways between the control and model groups (Figure 3H). Between the model and DCH groups, AA, linoleic acid, glycerophospholipid, and glycine/serine/threonine metabolism were identified to be significant metabolic pathways (Figure 3I). The same pathways obtained from both PICRUSt analysis of 16S rRNA sequencing and pathway analysis of untargeted metabolomics, including AA, glycerophospholipid, and glycine/serine/threonine metabolic pathways, were visualized in Figure 3J and discussed in detail.

As shown in Figure 4A, Lactobacillus and Romboutsia have shown significant negative correlations with the pathological changes in NAFLD rat models, whereas Enterococcus, Allobaculum, and Lachnoclostridium have shown positive correlations with the pathological changes in NAFLD rat models.

Additionally, Lactobacillus showed positive correlations with PC, rumenic acid, and lacto-N-tetraose and negative correlations with the majority of the amino acids and metabolites related to AA metabolism. Enterococcus and Lachnoclostridium showed positive correlations with most of the amino acid metabolites (Figure 4B).

The dysfunction of AA metabolism has been previously implicated in the course of metabolic disorder and inflammatory diseases (Sonnweber et al., 2018). Our results showed that the levels of 12(R)-HETE and 5-HPETE, which are metabolic products of AA, were increased in NAFLD model rats. Whereupon, DCH treatment decreased the levels of 12(R)-HETE and 5-HPETE. The serum levels of 12(R)-HETE have been positively correlated to the severity of NAFLD (Maciejewska et al., 2015). A possible mechanism of action is that increased levels of 12(R)-HETE and 5-HPETE induce the inflammatory response by recruiting neutrophil cells (Han and Corey, 2000). Dietary intervention has been shown to decrease the levels of 12-HETE and ameliorate hepatic steatosis in NAFLD patients (Dominika et al., 2015). Furthermore, negative correlations between Lactobacillus, Romboutsia, and 12(R)-HETE, 5-HPETE levels were found by our spearman’s analysis. Likewise, other studies showed that excessive dietary polyunsaturated fatty acids, such as linoleic acid (LA), could augment the adipose inflammatory response in obese mice through inducing AA metabolism. Colonization of Lactobacillus could inhibite AA metabolism to ameliorate adipose inflammatory response and obesity through converting excessive dietary LA to 10-hydroxy-cis-12-octadecenoic acid (HYA) (Miyamoto et al., 2019). Up to now, Romboutsia has not been demonstrated to associate with AA metabolism. The detailed effects and mechanisms of Romboutsia on AA metabolism still need to be further investigated.

The dysfunction of glycerophospholipid metabolism has been shown to disturb the energy metabolism of hepatocytes (Mesens et al., 2012). Our results showed that the level of PC was decreased and the level of PE was increased in NAFLD model rats, with the inverse effect following DCH treatment. PCs comprise 40–50% of total phospholipids in mammalian cells (Vander Veen et al., 2017). The generation of PC includes two avenues. Choline is transformed into CDP-choline, an intermediate product in PC generation, and then CDP-choline interacts with diacylglycerol to form PC. PC can also be transformed from PE directly. PEs account for 15–25% of total phospholipids in mammalian cells (Vander Veen et al., 2017). PE is derived from the synthesis of diacylglycerol and CDP-ethanolamine. The decreased ratio of PC to PE, observed in NAFLD patients, could impair the permeability of cellular membrane and induce damage to hepatocytes (Li et al., 2006). Moreover, modulating the balance of PC and PE has shown beneficial effects on NAFLD (Calzada et al., 2016). Studies have demonstrated that orally treatment of Lactobacillus-containing probiotic formulation could increase the levels of PC in HFD-induce obese rat model (Shin et al., 2018). Negative correlations of PC (18:3), PC (20:2) and Lachnoclostridium have been found in obese mice (Liu et al., 2018). Our spearman’s analysis also showed a positive correlation between Lactobacillus and PC level, and negative correlations of Lachnoclostridium, Enterococcus and PC. Additionally, we observed a negative correlation between PE and Romboutsia. Given these correlations, its likely the modulatory effects of DCH on glycerophospholipid metabolism might occur through affecting the abundances of Lactobacillus, Lachnoclostridium and Romboutsia.

The dysfunction of glycine, serine and threonine metabolism has also been associated with the NAFLD-related metabolic disorders (Huang et al., 2018). Increased levels of glycine and decreased levels of L-threonine were observed in NAFLD model rats, with DCH treatment reversing the trend. L-threonine is an essential amino acid in mammal cells and has shown beneficial effects on NAFLD through inducing the synthesis of phospholipid and oxidation of fatty acids (Zhang et al., 2020). Unlike other amino acids, L-threonine can be transformed into different amino acids, such as glycine, through specific enzymatic reaction rather than through the catalysis of dehydrogenase or transamination. Glycine can be transformed into serine, which can be subsequently transformed into pyruvate through dehydration and deamination. Pyruvate can then enter the citrate cycle to generate ATP. Clinical studies showed an increase of plasma glycine levels in patients with NAFLD (Melania et al., 2018). Excessive glycine levels can disrupt the metabolism of amino acids and the citrate cycle. Spearman’s correlation analysis results indicated that Lactobacillus, Romboutsia, and Bacteroides exhibited positive correlations with L-threonine, while Akkermansia, Enterococcus, Enterobacteriaceae, and Lachnoclostridium exhibited negative correlations. Likewise, positive corelations have been demonstrated of Lactobacillus, Romboutsia and L-threonine in mice after fed with Ganoderma lucidum spores oil (Wu et al., 2020). No bacteria examined showed significant correlations with glycine. The changes in glycine levels after DCH treatment might be caused by the transformation of L-threonine.