Due to the complexity of traditional Chinese medicine (TCM) and human diseases, there is no effective mode for revealing the relationships between the components and efficacy of herbal prescriptions. Serum pharmacochemistry is a powerful tool for elucidating the in vivo compounds of herbal prescriptions, in comparison with one-to-one separation in conventional methods, because only ingredients taken from the blood have the probability of becoming potential medicinal substances [1]. Network pharmacology is considered to be an efficient technique for a thorough understanding of the integrity and systemicity of herbal prescriptions. Its core idea is to clarify the biological language among active compounds, molecular targets, and disease features. Network pharmacology can even be combined with omics technology to improve the accuracy of unknown and known target identification [2]. Metabolomics, as an advanced system biology approach, was mainly developed to explore disturbed biomarkers and biological pathways throughout the biological system, and especially to characterize biochemical phenotypes in organisms. Thus, metabolomics has great potential to influence new drug discovery [3]. The integrative network analysis of serum pharmacochemistry, metabolomics, and network pharmacology can link active compounds and targets to endogenous metabolites, which is conducive to the overall target identification of the beneficial effects of active compounds [4].

Cholestatic jaundice is a clinical disorder caused by the abnormal accumulation of bile acid induced by obstruction of the bile duct inside and outside the liver or by damage to hepatocytes, which is mainly characterized by jaundice [5]. Yinchenhao Tang (YCHT), a classic herbal prescription for the treatment of cholestatic jaundice, was first described in the Treatise on Febrile Diseases written by Zhongjing Zhang from the Eastern Han Dynasty, and consists of Rheum officinale Baill (DH), Artemisia capillaries Thunb. (YC), and Gardenia jasminoides Ellis (ZZ) [6]. Numerous clinical and animal studies have shown that YCHT has many functions, including choleretic de-yellowing, treatment of hepatic injury, treatment of hepatic fibrosis, anti-inflammatory activity, pancreatic protection, and anti-tumor activity[6–7]. Although the animal and clinical efficacy of YCHT for the treatment of cholestatic jaundice has been widely recognized and determined to be without obvious toxicity or side effects, some problems with the complexity and diverse characteristics of TCM hinder the exploration of its active ingredients and hepatoprotective mechanism. The potential active molecules, core targets, and action mechanisms of YCHT have not yet been systematically clarified, which greatly restricts the discovery of novel active compounds.

In this study, we established a novel and potentially widely applicable strategy based on serum pharmacochemistry, metabolomics, network analysis, and experimental verification in order to systematically decipher the active ingredients and core targets of an herbal prescription. A detailed overview of the study design was provided in Fig. 1. First, we recruited 226 human subjects and 60 mice in order to reveal disease biomarkers and evaluate the formula efficacy from both clinical and animal aspects through untargeted metabolomics analysis. Second, active components in vivo originating from YCHT were discovered by means of serum pharmacochemistry, and their potential targets were determined by means of network analysis including text mining, MetaboAnalyst pathway analysis (MetPA), and ingenuity pathway analysis (IPA). Given the strong similarities between mouse and human biological processes, clinical and animal data were further anlayzed in order to discover active compounds, core biomarkers, and drug targets. Finally, the crucial targets were experimentally validated to demonstrate the potential of the current strategy. This approach could be beneficial in elucidating bioactive compounds and discovering emerging therapeutic targets in herbal prescriptions.

Fig. 1. The integrated strategy of serum pharmacochemistry, metabolomics, and network pharmacology. ROC: receiver operating characteristic; IPA: ingenuity pathway analysis; UGT1A1: uridine diphosphate glucuronosyl transferase 1A1; CYP7A1: cholesterol 7α-hydroxylase; FXR: farnesoid X receptor; ABCC: multidrug-resistanceassociated protein; TCMSP: Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform. Disease gene search engine evidence sentences (DigSee), GeneCards, and Therapeutic Target Database (TTD) are databases.

All crude drugs were obtained from Harbin Tong Ren Tang Drugstore (China). YCHT is composed of ZZ, YC, and DH. α-Naphthyl isothiocyanate (ANIT) was provided from Sigma-Aldrich (USA). Alcohol and olive oil were procured from Sinopharm Chemical Reagent Beijing Co., Ltd. (China). Assay kits for multidrugresistance-associated protein 2 (ABCC2), farnesoid X receptor (FXR), cholesterol 7α-hydroxylase (CYP7A1), multidrugresistance-associated protein 3 (ABCC3), and uridine diphosphate glucuronosyl transferase 1A1 (UGT1A1) were purchased from Jingmei Biotechnology Co. Ltd. (China). Assay kits for direct bilirubin (DBIL), alkaline phosphatase (ALP), total bilirubin (TBIL), malondialdehyde (MDA), total superoxide dismutase (T-SOD), and aspartate aminotransferase (AST) were purchased from BioSino BioTechnology & Science Inc. (China). Assay kits for total bile acid (TBA), -glutamyl transpeptidase (-GT), glutathione peroxidase (GSH-Px), alanine aminotransferase (ALT), and others were provided by Nanjing Jiancheng Biotechnology Institute.

Specific pathogen free (SPF) male Balb/c mice (weight (20 ± 2) g) were provided from Shanghai SLAC Laboratory Animal Co., Ltd. (China). Sixty animals were divided into two categories, including Dataset 1 containing the control group and jaundice group, and Dataset 2 containing the control, jaundice, and YCHT groups, with 12 mice in each group. The jaundice model was duplicated as follows: a Rhizoma Zingiberis extracting solution (decocted in distilled water to prepare the orally administered solution at a concentration of 0.013 g·mL^–1 ) in the morning and alcohol (12.5% (v/v), dissolved in distilled water) in the afternoon with quantities of 0.1 mL per 10 g bodyweight were given to the jaundice and YCHT groups via oral gavage for 14 consecutive days. On the 15th and the 16th days, different concentrations of ANIT solution (1.5 and 1.0 mg·mL^–1 , dissolved in olive oil) with quantities of 0.1 mL per 10 g bodyweight were given to the jaundice and YCHT groups once a day via oral gavage. Moreover, distilled water at a dose of 0.1 mL per 10 g bodyweight was given via oral gavage every day in the control group. On Day 17, members of the YCHT group were given an oral gavage of YCHT solution (5 g·mL^–1 , 10 mg per 1 kg bodyweight, dissolved in distilled water) for seven days, whereas the control group was given an oral gavage of distilled water (0.1 mL per 10 g bodyweight) [8]. All experimental procedures were approved by the Animal Ethics Committee of Heilongjiang University of Chinese Medicine, and the experiments were conducted according to the principles of the Declaration of Helsinki.

A total of 106 jaundice patients and 120 healthy subjects from the First Affiliated Hospital of Heilongjiang University of Chinese Medicine were enrolled between 2016 and 2018. Case inclusion criteria were as follows: Patients with the typical characteristics of a serum TBIL level of about 171 μmol·L^–1 ; clinically diagnosed yellow staining of skin tissue; and elevated serum ALP, -GT, and DBIL levels were included in the study. Subjects that were conceiving or breast feeding, participating in other drug trials affecting efficacy and safety judgments, exhibiting severe primary diseases in the kidney or hematopoietic system, or exhibiting mental dysfunction were excluded from the study. YCHT was decocted and packaged by the First Affiliated Hospital of Heilongjiang University of Chinese Medicine. YCHT-treated patients were assigned to take YCHT (100 mL per pack) orally two times a day for 15 consecutive days. The subjects were divided into two cohorts containing Dataset 1, comprising 120 healthy participants and 72 jaundice patients, and Dataset 2, comprising 120 healthy participants, 72 jaundice patients, and 34 YCHT-treated patients. All patients signed written informed consent before the research was conducted, and the study was approved by the Ethics Committee of Heilongjiang University of Chinese Medicine.

Mouse blood samples were collected by picking the eyeball in mice, whereas human serum samples were collected from the vein, and the blood samples were immediately centrifuged at 3000 revolutions per minute (rpm) for 15 min at 4 C. Supernatant blood samples were collected and stored at –80 °C prior to analysis [9]. A mixture of 600 μL methanol and 200 μL thawed serum was vortexed for 30 s and centrifuged (13 000 rpm, 4 °C) for 5 min. Two microliters of the supernatant was injected for metabolomics analysis. For constituent analysis in human, the mixure of 2 mL serum and 40 μL of phosphoric acid was assigned to the preactivated and prebalanced solid-phase extraction (SPE) column after ultrasonication and vortexing procedure. The column was eluted with 1 mL of water and 2 mL of 100% methanol, and 100% methanol eluates were dried under nitrogen stream at 45 °C. The residue was redissolved in 0.15 mL 100% methanol and centrifuged at 13 000 rpm 15 min at 4 °C. Five microliters of the supernatant was injected for constituent analysis. The methods used for constituent analysis in mouse were consistent with those used for above human.

According to the kit manufacturer’s instructions, clinical biochemistry indexes including the activity levels of TBA, T-SOD, ALP, DBIL, TBIL, -GT, ALT, MDA, UGT1A1, ABCC2, CYP7A1, AST, FXR, ABCC3, and GSH-Px were quantitatively measured. After the blood samples were obtained, the liver tissues were stained with hematoxylin–eosin (HE) for histopathological observation under an optical microscope.

A ultra-performance liquid chromatography (UPLC) system equipped with a quadrupole time-of-flight mass spectrometry (Q-ToF-MS; Waters, USA) was used for serum sample detection and analysis. The mice and human serum samples were separated on an ACQUITY UPLC HSS T3 chromatographic column (100 mm × 2.1 mm, 1.8 μm) and an ACQUITY UPLC HSS C18 chromatographic column (100 mm × 2.1 mm, 1.8 ，μm) (Waters, USA), respectively. The metabolomics analysis parameters, including chromatographic conditions such as column temperature and flow rate, gradient eluting procedure, and injection volume, as well as MS conditions such as the temperature of the source and desolvation, sample cone voltages, and capillary voltages, as presented in Tables S1–S3 in Appendix A, were responsible for the stability, accuracy, and repeatability of the ultimate results. The column temperatures for mouse and human sample analysis were set at 40 and 45 °C, respectively, and both injection volumes were modulated at 2 μL.

The procedures used for metabolite identification were consistent with those reported in previous research [9]. Significant variables were identified by means of the multivariate statistical method using precise molecular mass, molecular formula (isotope fitting pattern (i-FIT) value close to zero, a range of less than 5 parts per million (ppm) deviation), tandem MS (MS/MS) fragment in combination with the reported procedure, and databases including Human Metabolome Database (HMDB), ChemSpider, Kyoto Encyclopedia of Genes and Genomes (KEGG), and Metlin. To map the most relevant pathways of identified metabolites from jaundiceassociated research, a visualization of the enrichment analysisbased metabolic pathways was generated and annotated by means of MetPA.

The blood samples were separated on an ACQUITY UPLC HSS C18 chromatographic column at a flow rate of 0.4 mL·min^–1 and a column temperature of 45 °C. The other UPLC and MS parameters are listed in Tables S4 and S5 in Appendix A [9]. The data were processed by Progenesis QI software coupled with the elemental composition tool module and MS/MS fragment information for the identification of serum constituents.

First, we introduced potential effective components into the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP), SwissTargetPrediction, and SuperPred databases in order to manually search for molecular targets according to the obtained candidate serum components that originated from YCHT. Potential targets associated with cholestatic jaundice disease were then retrieved from various databases such as Online Mendelian Inheritance in Man (OMIM), Therapeutic Target Database (TTD), and GeneCards[10,11]. Duplicate and false-positive results having nothing to do with jaundice were removed by screening during the process of data mining. The shared targets associated with the active component and jaundice disease were considered to be functional molecules. As a result, an integrated component–target– disease interaction network was constructed using Cytoscape 3.0. To facilitate the reasonable interpretation of the identified functional targets involved in biological processes, molecular functions, and cellular components, gene ontology (GO) enrichment analysis and KEGG pathway enrichment analysis were conducted by means of the search tool for recurring instances of neighboring genes (STRING) database. The comprehensive metabolomics data analysis platform MetaboAnalyst 3.0 and IPA were applied to enrich the metabolic pathway of the identified biomarkers (P < 0.05) and to analyze the interaction network of various small molecules including biomarkers, proteins, and genes. Enrichment pathways combined with network pharmacology and serum pharmacochemistry were applied in order to deeply capture the biological network of the core targets and metabolic pathways regulated by the active compounds; subsequently, a component–target–metabolite interaction network was constructed.

The data were analyzed by means of Statistical Product and Service Solutions (SPSS) software (IBM, USA) and reported as mean ± standard deviation (SD). The differences between the two groups were analyzed using Student’s t-test, and differences between multiple groups were analyzed by means of one-way analysis of variance (ANOVA). A P value of less than 0.05 was considered to be statistically significant.

The baseline biochemical and demographic data on the jaundice patients in Dataset 1 are shown in Table S6 and Fig. S1(a) in Appendix A. Pre-albumin (PA), globulin (GLB), albumin (ALB), and total protein (TP) levels were not significantly different between the two groups, whereas ALP, ALT, AST, TBIL, DBIL, cholinesterase (CHE), -GT, and indirect bilirubin (IBIL) were significantly associated with cholestatic jaundice (P < 0.01). Dataset 2 contained 120 healthy subjects, 72 patients, and 34 YCHT-treated patients. The ALT, AST, TBIL, DBIL, and IBIL levels of the jaundice group showed a significant downward trend (P < 0.01 or P < 0.05), and partial remission of the yellow staining of skin tissue was observed after YCHT treatment. Very consistent results were observed upon the establishment of cholestatic jaundice mice. The serum contents of AST, DBIL, TBIL, ALP, ALT, -GT, and TBA in the jaundice group were markedly elevated (P < 0.01 or P < 0.05), and the level of TSOD was significantly down-regulated (P < 0.05) compared with the control group. In the liver tissue, the amounts of MDA were not markedly up-regulated, and the contents of GSH-Px were significantly down-regulated (P < 0.01; Fig. S1(b) and Table S7 in Appendix A). The histopathological results showed that highly edematous hepatocytes and hepatic lobular structural disorders were observed, and normal cells were infiltrated by inflammatory cells in the model group, as shown in Fig. S1(c) in Appendix A. Compared with the model group, the necrotic area and edema degree were decreased in the YCHT group; furthermore, the outline of the liver tissues including the lobule, liver plate, sinusoid, central lobular vein, bile duct, and so on were clear without obvious abnormality. Occasionally, a small amount of inflammatory cells was distributed in the YCHT group, as shown in Fig. S1(c). These results confirm that the use of YCHT resulted in prominent improvement in the treatment of cholestatic jaundice.

EZinfo software was typically employed to conduct unsupervised principal component analysis (PCA) of serum data in order to acquire intuitive clustering information on the metabolic trajectory in various groups. From the two-dimensional (2D) PCA and three-dimensional (3D) PCA score plots of mice and humans, evident separation was observed of the metabolic profiles between the jaundice group and the control group, as shown in Fig. 2, which was confirmed by the base peak intensity (BPI) serum chromatograms shown in Fig. S2 in Appendix A. In order to further evaluate the therapeutic effects of YCHT, focusing on PCA score plot analysis, we found that, regardless of animal or human, are markable reverse tendency appeared in the jaundice group after intervention by YCHT, indicating that YCHT can restore the pathological progress of jaundice disease, as shown in Fig. S3 in Appendix A.

Fig. 2. Multivariate analysis in the untargeted serum metabolomics of mice and humans. (a)–(d) The 2D and 3D PCA score plots of all the analysis data in mice were carried out in (a, c) ESI^– and (b, d) ESI⁺ ; (e)–(h) the 2D and 3D PCA score plots of all the analysis data in humans were carried out in (e, g) ESI^– and (f, h) ESI⁺ . PC: principal component.

To discover the potential serum biomarkers responsible for the differences between the two groups, we used UPLC-Q-ToF-MS integrated with multivariate data analysis to decipher variables in positive ion mode in the electrospray ionization (ESI⁺ ) and negative ion mode in the electrospray ionization (ESI^– ). The variable influence on projection (VIP) value (>1.0) of the orthogonal projections to latent structures discriminant analysis (OPLS-DA) and a probability value (P < 0.05) of the Student’s t-test-based biomarker selection were used for the screening of potential variables. Of all the variables, we finally characterized 12 biomarkers including four and eight biomarkers from the ESI⁺ and ESI^– modes in mice, respectively (Table S8 in Appendix A), and 14 biomarkers including seven and seven biomarkers from the ESI⁺ and ESI^– modes in humans, respectively (Table S9 in Appendix A), based on the available databases and MS/MS fragment information. A scatter plot and a clustering heat map of the relative intensities of the identified biomarkers were simultaneously made in order to intuitively interpret how YCHT affects metabolic changes and clearly ascertain vital call back biomarkers, as shown in Figs. 3 and 4. To examine the classification power of the identified metabolites, we utilized these metabolites as a basis for a metabolic profile analysis from 60 clinical samples, which were randomly selected from 192 subjects under the premise that we did not know whether the selected sample was from a jaundice patient or a healthy subject. As a result, the datasets were divided into two clusters, in which 32 samples were arranged in Group 1 and 28 samples were arranged in Group 2. Upon observing the information of each sample in the two groups, it was found that the samples in Group 1 and Group 2 were the subjects of the healthy group and the jaundice group, respectively. Based on this, we initially verified the suitability of the screened markers, as shown in Fig. S4 in Appendix A.

Fig. 3. Serum metabolites were further verified by Dataset 1 from humans. (a) Scatter plot of the relative intensity of identified biomarkers between healthy participants, jaundice patients, and YCHT treatment; (b) heatmap of serum metabolites between healthy participants, jaundice patients, and YCHT treatment patients. Results are expressed as mean ± SD. ^*P < 0.05 and ^**P < 0.01 vs the healthy group; ^#P < 0.05 and ^##P < 0.01 vs the jaundice group. S1P: sphingosine-1-phosphate; PC: phosphatidylcholine; TG: triglyceride.

Fig. 4. Serum metabolites were further verified by Dataset 1 from mice. (a) Scatter plot of the relative intensity of identified biomarkers between the control, jaundice, and YCHT groups; (b) heatmap of serum metabolites from the control, jaundice, and YCHT groups. Results are expressed as mean ± SD. ^*P < 0.05 and ^**P < 0.01 vs the control group; ^#P < 0.05 and ^##P < 0.01 vs the jaundice group. PA: phosphatidic acid.

To reduce the number of metabolites for clinical diagnostic and functional analysis applications, we performed an index reduction as an evaluation model on 14 biomarkers in human patients and 12 serum biomarkers in mice following the screening principle of the receiver operating characteristic (ROC) curves with area-underthe-curve (AUC) values greater than 0.9 for the metabolites. Using the order of AUC values from human or mouse subjects, taurocholic acid, bilirubin glucuronide, bilirubin, and biliverdin were identified as the potential biomarkers of the clinical jaundice patients (Table S10 and Fig. S5(a) in Appendix A), and phosphatidylcholine (PC) (16:0/16:0), bilirubin, sulfoglycolithocholate (2–), LysoPC(18:1(9Z)), biliverdin, and taurocholic acid were regarded as the potential biomarkers of the jaundice mice (Table S11 and Fig. S5(b) in Appendix A). Coincidentally, of all these metabolites, three common markers—namely, bilirubin, biliverdin, and taurocholic acid—were found in both the human patient and mouse samples. By focusing on a pathway topology analysis of the clinical core biomarkers including bilirubin, biliverdin, bilirubin glucuronide, and taurocholic acid based on MetPA, we found that they mainly referred to porphyrin and chlorophyll metabolism and primary bile acid biosynthesis. The relevant network between the metabolic pathways and the biomarkers was then further constructed using the KEGG pathway database for the next core target analysis, as shown in Fig. 5.

Fig. 5. Correlation pathways of potential metabolites that are differentially expressed in jaundice patients and mice. The main disordered pathways (a) lipid metabolism and (b) bile acid metabolism in jaundice mice are shown inside the blue frame. The main disordered pathways (c) lipid metabolism and (d) bile acid metabolism in jaundice patients are shown inside the orange frame. A comparative overview of MetPA metabolic pathways existing in (e) jaundice mice and (f) patients is illustrated. The Identity document (ID) inside the box refers to the KEGG number of the metabolic markers tentatively detected in this study; among them, the ID inside the red box was differentially expressed, and that inside the other box was not identified in the current research. Symbols on the arrows connecting the boxes present key enzymes provided by KEGG; box plots next to the red box demonstrate the relative differential intensities between the jaundice group (red) and the control group (green). LacCer: lactosylceramider; CoA: coenzyme A.

To further excavate the potential mechanisms of functional changes in jaundice disease, the IPA omics platform was exploited. With the assistance of IPA for the disease, the canonical pathway, and the functional changes in jaundice mice and patients (Tables S12 and S13 in Appendix A), we identified the selected metabolites. Of these, taurocholic acid, cholic acid, bilirubin glucuronide, bilirubin, and biliverdin were involved in cholestasis associated with bile acid biosynthesis, heme degradation, interleukin (IL)-10 signaling, and FXR/retinoid X receptor (RXR) activation, and then serum biomarker predictive networks were constructed. According to the canonical pathways and interaction network analysis (Figs. S6 and S7 in Appendix A), we found that the downstream proteins containing UGT1A1, ABCC3, and so forth, had a relationship with bilirubin related to heme degradation. Downstream ABCC3, FXR, and so forth, and upstream proteins such as CYP7A1, and so forth, had a relationship with taurocholic acid involved in primary bile acid biosynthesis and FXR/RXR activation. After YCHT treatment, the activation of FXR, ABCC3, and UGT1A1 and the inhibition of CYP7A1 suggested that the three pathways played a key role in modulating the bile acid balance (Figs. S8 and S9 in Appendix A). From the perspective of metabolomics analysis, it can be inferred that the enriched pathways of the pharmacodynamic metabolites included the same canonical pathways as the jaundice biomarkers. The IPA indicated that YCHT regulated the primary bile acid biosynthesis, heme degradation, and FXR/RXR pathway to prevent the progress of jaundice disease in mice and humans.

Multivariate pattern-recognition analysis with PCA and OPLSDA was used for screening the variables by means of Progenesis QI software (Fig. 6). Of all the variables, ions that were not present in the model group but were present in the dosed group were regarded as potential blood components. Based on the previous constituent analysis of YCHT, a total of 26 prototype components and three metabolites were tentatively characterized in human blood, and 33 prototype components and three metabolites were tentatively characterized in mice (Tables S14 and S15 in Appendix A).

Fig. 6. UPLC–MS chromatograms and multivariate analysis in serum pharmacochemistry analysis. (a) Chromatograms of the YCHT group and the jaundice group in mice; (b) S-plot between the jaundice group and the YCHT group in mice; (c) trend plot of the identified compound 15.55_283.0253 in mice; (d) PCA score plot between the jaundice group and the YCHT group in mice; (e) chromatograms of the YCHT group and jaundice group in humans; (f) S-plot between the jaundice group and YCHT group in humans; (g) trend plot of the identified compound 15.55_283.0253 in humans; (h) PCA score plot between the jaundice group and YCHT group in humans. p[1]: the first principal component in s-plot; p(corr)[1]: the partial correlation coefficient of the first principal component in s-plot.

A combination strategy of network pharmacology and serum pharmacochemistry provided new insight into the common targets of jaundice disease and the serum components originating from YCHT. A total of 79 common targets in both mice and humans were identified from serum-component-related targets and jaundicedisease-related targets by means of a database search. To illustrate the potential pharmacological effects of YCHT and the key roles of its common targets, formula–herbs–compound–target–jaundice disease networks were constructed, as shown in Fig. 7; detailed information is provided in Tables S16 and S17 in Appendix A. We extracted the common targets and introduced them into a string database search and transformation operation for GO and KEGG pathway enrichment analysis, which were sorted to distinguish differences according to the descending order of –lgP. The top rankings were visualized using the online drawing site Omishare Tool (Fig. S10 in Appendix A). According to the results of the GO enrichment analysis across mice and humans, common molecular targets mainly occurred in biological processes, such as the response to chemicals, the cellular response to chemical stimulus, and the response to lipids. Among the cell components, cell part and extracellular region part ranked the highest in their –lgP score and gene numbers. In the molecular function analysis, enzyme binding and protein binding were prominent. According to the bubble map of KEGG pathway enrichment analysis, expanded correlative pathways including bile secretion, porphyrin and chlorophyll metabolism, and primary bile acid biosynthesis were highlighted. Both the animal and clinical trials implied that the identified targets were distributed in different pathways, which were modulated by the YCHT components in a cooperative manner.

Fig. 7. Shared targets of serum ingredients and jaundice disease after YCHT treatment. (a) Unique and common targets for jaundice and serum ingredients in mice. The network map includes 36 target ingredients (orange) and three herbs (light blue) associated with YCHT (red) and 323 unique targets (blue) associated with cholestatic jaundice (red), with 79 targets shared by both (green). (b) Unique and common targets for jaundice and serum ingredients in humans. The network map includes 29 target ingredients (orange) and three herbs (light blue) associated with YCHT (red) and 323 unique targets (blue) associated with cholestatic jaundice (red), with 79 targets shared by both (green).

Based on a cross-talk between serum pharmacochemistrybased network pharmacology and serum metabolomics analysis, we speculated that porphyrin and chlorophyll metabolism, bile secretion, and primary bile acid biosynthesis may serve as the crucial underlying mechanisms in the pathogenesis of jaundice disease and YCHT efficacy. Focusing on the correlation of biomarkers, possible upstream targets, and active components across humans and mice, we extracted core targets including ABCC2, ABCC3, UGT1A1, FXR, and CYP7A1, which were confirmed by the IPA interaction network for subsequent target validation. The top eight components linked to core targets were geniposide, scoparone, isorhamnetin, quercetin, naringenin, rhein, chlorogenic acid, and kaempferol. The key metabolites containing bilirubin, biliverdin, bilirubin glucuronide, and taurocholic acid may be regarded as the core biomarkers. A functional mechanism diagram of YCHT was constructed, as shown in Fig. 8. Internal validation of four clinical key metabolites in humans including bilirubin, biliverdin, bilirubin glucuronide, and taurocholic acid—three of which also presented in mice—was performed on an independent cohort with 60 random participants extracted from 226 subjects. Unsupervised clustering analysis demonstrated that the model with four metabolites could distinguish between healthy subjects and jaundice patients (Fig. S11(a) in Appendix A). Moreover, ROC analysis showed that the AUC values were equal to 0.996, indicating that the model of the selected metabolites may serve as a complementary diagnostic method (Fig. S11(b) in Appendix A).

Fig. 8. Discovery of potential targets and active compounds driven by metabolomics, serum pharmacochemistry, and network pharmacology. (a) A network illustration of the mechanism of YCHT in the regulation of jaundice was constructed; (b) a correlative relationship of the targeting efficacy between the novel targets and the potential active compound was constructed.

For in-depth exploration and predictive target validation by the integrated analysis of small molecules, the effects of YCHT on biliary excretion-associated proteins including ABCC2, ABCC3, and UGT1A1, and on bile biosynthesis-associated proteins including CYP7A1 and FXR, were assessed by means of an enzyme-linked immunosorbent assay (ELISA) kit. Compared with healthy subjects, the levels of ABCC2, ABCC3, UGT1A1, and FXR in jaundice subjects were significantly decreased, and the level of CYP7A1 was markedly elevated (P < 0.01 or P < 0.05). YCHT treatment can attenuate the decrease of ABCC2, ABCC3, UGT1A1, and FXR and the increase of CYP7A1 (Fig. S12 in Appendix A). In summary, YCHT possesses a favorable therapeutic effect on jaundice disease via the regulation of CYP7A1, ABCC2, ABCC3, UGT1A1, and FXR to promote bile secretion and inhibit bile acid biosynthesis.

As is well known, the bile flow in the hepatocyte is impaired for cholestatic jaundice patients, which leads to the retention and accumulation of toxic bile acid, enhancing the aggravation of jaundice disease. Hepatocyte function improvements boosted by inhibiting bile acid production and provoking biliary secretion in jaundice disease play an indelible role in the bile acid metabolism of the body[12–15]. Considering the above results, modulation of bile acid production and secretion may facilitate the therapeutic effects of YCHT on jaundice disease. Numerous observations found obvious remission of jaundice symptoms after YCHT treatment, along with callback of its relevant biomarkers. The potential targets and active ingredients of YCHT were discovered via an IPA platform and plotting of correlation between marker metabolites and serum constituents; nevertheless, the underlying mechanisms of how YCHT targets the pathological process of cholestatic jaundice are far from being clearly revealed[16,17]. These discoveries have only been deduced in the animal model and need to be further verified by reliable evidence combined with clinical data. To thoroughly decipher these problems, the first comprehensive strategy integrating serum pharmacochemistry, network pharmacology, and metabolomic characterization across an animal model and humans was established. We assumed that the phenotype relevant to the disease should be associated with the systemic effects of abnormal endogenous disorders, thus, identification of the phenotype required comprehensive multi-angle analysis at the systemic level for deep data mining, which was transformed into point-to-point target validation of individual genes.

Using multiplex analysis containing the aforementioned clinical and animal bidirectional findings, we established a schematic model depicting the most potential molecular network shared in both animal and human (Fig. 8). Interestingly, the same candidate target molecules, including ABCC2, ABCC3, CYP7A1, UGT1A1, and FXR, were recognized as functional targets of YCHT by the enrichment analysis of two types of subjects. We were convinced that the bioactive components of YCHT may jointly interact with multiple signaling pathways, and consequently exhibit synergistic effects in the treatment of jaundice disease. These findings provide a novel mechanistic connection to elucidate how YCHT can prevent the pathological progression of cholestatic jaundice. Furthermore, a systematic study of untargeted metabolomics, multiple subjects, and multi-step verifications was performed to identify bilirubin, biliverdin, and taurocholic acid associated with cholestatic jaundice disease, which are indicative of the most promising biomarkers of altered liver function for cholestatic jaundice. Taurocholic acid, a crucial signaling molecule of biliary secretion, is recognized as a diagnostic marker of impaired biliary excretion during cholestatic jaundice and is believed to have immunoregulatory and antiinflammatory effects [18]. Of these identified biomarkers, bilirubin and biliverdin also deserve to play a significant role in clinical jaundice diagnosis [19]. Bilirubin is a metabolite synthesized from biliverdin by means of biliverdin reductase. During cholestatic jaundice, the accumulation of bile acid results in mitochondrial dysfunction and impaired bilirubin glucuronidation, which stimulates the conversion of biliverdin to bilirubin in order to exert antioxidant and cytoprotective effects [20]. Several studies have shown that metabolites related to jaundice disease, including kynurenic acid and D-glucuronic acid, are abnormally expressed in animal serum and urine samples[14,21]. Therefore, the present work highlights the potential utility of a novel model with bilirubin, taurocholic acid, bilirubin glucuronide, and biliverdin in the prediction and diagnosis of clinical jaundice disease. Other associated metabolites such as L-homocysteic acid, cholic acid, LysoPC (18:1(9Z)), and LysoPC (16:1(9Z)) were differentially presented between humans and mice, which may be attributed to intrinsic differences between species.

Integrating serum pharmacochemistry, network pharmacology, and metabolomics investigation not only further verified the therapeutic effects of YCHT, but also assisted in developing a better understanding of its functional nature. An exciting result of our experimental study is the discovery of a novel metabolismmediated therapeutic mechanism of YCHT. The most differentiated affected metabolites from jaundice mice and human patients included bile acids and lipids associated with bile acid biosynthesis and biliary secretion. A dataset analysis revealed a higher bile acid level of jaundice patients or mice than that of healthy participants or mice, and bile acid biosynthesis and biliary secretion appeared to be in close relationship with cholestatic jaundice. In fact, abundant findings demonstrate that an improvement in bile acid synthesis and the inhibition of biliary secretion are beneficial to the amelioration of cholestatic jaundice disease [22]. Moreover, network pharmacological analysis indicated that YCHT could participate in bile acid secretion and synthesis by a potential means of regulation of bile acid, illustrated by its effects on ABCC2, ABCC3, and CYP7A1. The multidrug-resistance-associated protein (MRP) family, a class of organic anion transporters belonging to the adenosine triphosphate-binding cassette (ABC) transporter superfamily including ABCC2, ABCC3, and so forth, is responsible for the transportation of glucuronide[23,24]. ABCC2, also known as multidrug-resistance-associated protein 2 (MRP2), participates in the secretion of various amphiphilic anions such as bilirubin and bile acid sulfate in hepatocytes in order to maintain bile acid homeostasis, which plays a pivotal role in bile secretion [25]. CYP7A1, a rate-limiting enzyme in the classical pathway of bile acid biosynthesis, has the functions of modulating cholesterol homeostasis and bile acid synthesis. In line with extensive jaundice model evidence, the expression level of ABCC2 in the jaundice group was significantly decreased and the expression level of CYP7A1 was significantly increased in comparison with the healthy group, which was reversed by YCHT treatment, as further verified by solid experimental animal and clinical data [26]. Due to the activation of ABCC2, the secretion capability of bivalent bile salts in the lateral membrane cells of bile canaliculi was reinforced, which could explain why the level of bile acid in the YCHT group was down-regulated. ABCC3 is usually not expressed in normal hepatocyte membranes; once cholestasis occurs, the expression level of ABCC3 in the basal membrane of hepatocytes is significantly increased. Furthermore, since the ABCC3 transport substrate has extensive homology with that of ABCC2, it is generally believed that the high expression level of ABCC3 compensates for ABCC2 to alleviate choline-induced hepatotoxic damage during cholestatic jaundice [27]. This supports the vital role of ABCC3 in the protection of cholestatic jaundice, as was firmly verified by our experiments from the comparison of ABCC3 protein expression level among various groups.

Due to the abrupt increase in bile acid volume that follows bile formation, secretion, and excretion disorders, intrahepatic cholestasis forms. Correspondingly, bile flow cannot be transported normally into the duodenum and flows back into the bloodstream, increasing the level of bile acids in the blood. This results in the activation of FXR, which participates in suppressing bile acid biosynthesis by lowering CYP7A1 activity[28,29]. FXR, a member of the hormone nuclear receptor superfamily, plays a central role in bile acid synthesis, transport, and secretion via inducing the expression of special target genes. Activation of FXR inhibits bile acid biosynthesis and activates secretion into the bile duct, thereby protecting against the toxic accumulation of bile acid in hepatocytes[30,31]. We have observed that YCHT can ameliorate the symptoms of cholestatic jaundice and restore bile acid disorders, as regulation of FXR activation and inhibition of CYP7A1 were deeply involved in the synergetic efficacy of YCHT. Thus far, a considerable amount of evidence has implied that the alteration of bile acids could initiate FXR activation or inhibition in the modulation of CYP7A1 expression[32,33]. Moreover, compared with the jaundice group, the opposite changes of down-regulated CYP7A1 and up-regulated FXR were observed in the jaundice subjects with YCHT intervention, supporting the close relationship between the two targets.

Of the targeted pathways, many are deeply involved in the evolution of jaundice disease. By improving porphyrin and chlorophyll metabolism, YCHT can alleviate the severity of jaundice in both mice and humans, as the pathological changes in cholestatic jaundice disease are mainly caused by bilirubin dysfunction. Further analysis found that YCHT simultaneously altered bilirubin, biliverdin, and bilirubin glucuronide in porphyrin and chlorophyll metabolism. UGT1A1, a key metabolomic enzyme of the porphyrin and chlorophyll metabolism, plays a catalytic role in the formation of bilirubin glucuronides in the liver to promote ABCC2- or ABBC3- mediated bile secretion under good conditions[34,35]. From this perspective, the implication of UGT1A1 and ABC transporters in the functional regulation of the aberrant metabolism of bilirubin in the liver hints at bilirubin’s crucial role in cholestatic jaundice. Inhibition of UGT1A1 could prevent the binding of bilirubin and glucuronic acid, thus leading to an elevated bilirubin level. Agents targeting porphyrin and chlorophyll metabolism will block their excessive expression, and may be regarded as therapeutic candidates for the treatment of jaundice diseases [36]. For these reasons, UGT1A1 may serve as an effective potential target for jaundice drug discovery. Treatment with YCHT significantly increased the activity of UGT1A1, which could contribute to the downregulated bilirubin and up-regulated bilirubin glucuronide, implying that the production and elimination of bilirubin were normalized by YCHT. Meanwhile, improvements in ABCC2- or ABCC3- mediated bilirubin glucuronide secretion were essential to the alleviation of jaundice disease, while also yielding some solid benefits.

In summary, YCHT exerts therapeutic effects on cholestatic jaundice by enhancing the activity of FXR, UGT1A1, ABCC2, and ABCC3, and suppressing the activity of CYP7A1. Further research needs to be conducted in order to evaluate the efficacy of YCHT’s active ingredients, including geniposide, scoparone, isorhamnetin, quercetin, naringenin, rhein, chlorogenic acid, and kaempferol, on the core target pathways. Among these ingredients, the therapeutic effects of geniposide and scoparone on jaundice-related disease have been strongly validated[37–39]. This research was valuable for conducting integration research on drug discovery within a TCM formula and on its effective therapeutic targets.

This study established an innovative strategy for discovering the potential targets and active compounds of YCHT. We reported a comprehensive strategy for identifying relevant disorders in circulating metabolites in cholestatic jaundice disease across human and mouse biosamples. Notably, we found that eight active components in YCHT could modulate CYP7A1, ABCC2, ABCC3, UGT1A1, and FXR as well as regulating the primary bile acid biosynthesis, porphyrin and chlorophyll metabolism, and biliary secretion pathways. The integrated strategy was demonstrated to be useful for the effective discovery of active compounds and therapeutic targets from herbal medicine.

This work was supported by grants from the Key Program of National Natural Science Foundation of China (81430093, 81830110, and 81861168037) and Heilongjiang Touyan Innovation Team Program.

Hui Xiong, Ai-Hua Zhang, Ya-Jing Guo, Xiao-Hang Zhou, Hui Sun, Le Yang, Heng Fang, Guang-Li Yan, and Xi-Jun Wang declare that they have no conflict of interest or financial conflicts to disclose.

to this article can be found online at https://doi.org/10.1016/j.eng.2020.12.016.