1521-009X/44/6/800–808$25.00 http://dx.doi.org/10.1124/dmd.115.068189DRUG METABOLISM AND DISPOSITION Drug Metab Dispos 44:800–808, June 2016Copyright ª 2016 by The American Society for Pharmacology and Experimental Therapeutics

An Integrated Approach for Studying Exposure, Metabolism, andDisposition of Multiple Component Herbal Medicines Using

High-Resolution Mass Spectrometry and Multiple DataProcessing Tools s

Caisheng Wu, Haiying Zhang, Caihong Wang, Hailin Qin, Mingshe Zhu, and Jinlan Zhang

State Key Laboratory of Bioactive Substance and Function of Natural Medicines, Institute of Materia Medica, Chinese Academy ofMedical Sciences and Peking Union Medical College, Beijing, China (C.Wu., C.Wa., H.Q., J.Z.); Department of Biotransformation,

Bristol-Myers Squibb Company, Princeton, New Jersey (H.Z., M.Z.)

Received November 2, 2015; accepted March 23, 2016

ABSTRACT

A typical prescription of traditional Chinese medicine (TCM) containsup to a few hundred prototype components. Studying their absorption,metabolism, distribution, and elimination (ADME) presents greatchallenges. The objective of this study was to develop a practicalapproach for investigating ADMEof individual prototypes in TCM. Anactive fraction of Xiao-Xu-Ming decoction (AF-XXMD) as a modelTCMprescription was orally administered to rats. AF-XXMD–relatedcomponents in plasma, urine, bile, and feces were detected usinghigh-resolution mass spectrometry and background subtraction, anuntargeted data-mining tool. Components were then structurallycharacterized on the basis of MSn spectral data. Connection ofdetected AF-XXMD metabolites to their precursor species, eitherprototypes or upstream metabolites, were determined on the basisof mass spectral similarity and the matching of biotransformation

reactions. As a result, 247 AF-XXMD–related components weredetected and structurally characterized in rats, 134 of which weremetabolites. Among 198 AF-XXMDprototypes dosed, 65 were fully orpartially absorbed and 13 prototypes and 34 metabolites were foundin the circulation. Glucuronidation, isomerization, and deglycosylationfollowed by biliary and urinary excretions and direct elimination ofprototypes via kidney and liver were themajor clearance pathways ofAF-XXMD prototypes. As an example, the ADME profile of H56, thesingle major AF-XXMD component in rat plasma, was elucidatedon the basis of profiles of H56-related components in plasma andexcreta. The results demonstrate that the new analytical ap-proach is a useful tool for rapid and comprehensive detectionand characterization of TCM components in biologic matrix in aTCM ADME study.

Introduction

The bioactive substances of traditional Chinese medicine (TCM) andunderstanding their action mechanisms are of great interest to drugdiscovery scientists and clinicians. The effectiveness of TCM isgenerally recognized to be associated with chemical constituents ofTCM in the circulation (Wang et al., 2011; Zhang et al., 2011, 2013).Concentrations and duration of individual TCM components in thecirculation depend on their absorption, distribution, metabolism, andexcretion (ADME), processes that are often mediated by metabolizingenzymes and transporters. Inhibition or induction of the involvedenzymes or transporters by a coadministered TCM component or apharmaceutical drug could significantly change exposure levels ofbioactive components, leading to drug-drug interactions among herbalcomponents and between an herbal component and a pharmaceutical

drug (Posadzki et al., 2013; Cheng et al., 2014; Ma et al., 2014; Jia et al.,2015). In addition, the study of ADME of herbal medicines is veryimportant for the elucidation of mechanisms of TCM-induced toxicity.Herbal medicine–mediated organ toxicity is often related to highexposure and accumulation of certain toxic components in the organs(Qiu et al., 2000; Zhu, 2002; Hu et al., 2004; Yue et al., 2009; Yuan et al.,2011; Ding and Chen, 2012; Xiong et al., 2014).ADME studies of a drug in human and animal are often carried out

using a radiolabeled drug. However, it is not practical to use multipleradiolabeled TCM prototype compounds in a TCM ADME study.Therefore, the success of ADME study of a TCM prescription relies onliquid chromatography (LC)/mass spectrometry (MS) technology (Songet al., 2014). In the past ten years, many TCM research groups havemade significant efforts in the development and application of a vari-ety of LC/MS approaches for detection and characterization of TCMcomponents in the circulation and excreta (Yang et al., 2012; Wu et al.,2012; Yan et al., 2013; Geng et al., 2014; Zuo et al., 2015). The firstanalytical challenge faced in studying the ADME of herbal medicinesusing high-resolution mass spectrometry (HRMS) is to sensitively andcomprehensively detect TCM components in plasma, urine, feces, and/or bile samples. The task is often accomplished by processing accurate

This work was supported by the Beijing Natural Science Foundation[Grant 7133252] and the National Natural Science Foundation of China[Grant 81302740].

dx.doi.org/10.1124/dmd.115.068189.s This article has supplemental material available at dmd.aspetjournals.org.

ABBREVIATIONS: ADME, absorption, distribution, metabolism, and elimination; AF-XXMD, active fraction of Xiao-Xu-Ming decoction; EIC,extracted ion chromatography; GI, gastrointestinal; HRMS, high-resolution mass spectrometry; LC, liquid chromatography; MDF, mass defect filter;MS, mass spectrometry; MTSF, mass spectral trees similarity filter; NLF, neutral loss filter; PATBS, precise-and-thorough background-subtraction;TCM, traditional Chinese medicine; TIC, total ion chromatogram.

800

http://dmd.aspetjournals.org/content/suppl/2016/03/24/dmd.115.068189.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

MS and MS/MS datasets using targeted data-mining tools (Wu et al.,2012; Yang et al., 2012), including mass defect filter (MDF), extractedion chromatography (EIC), product ion filter (PIF), neutral loss filter(NLF), and isotope pattern filter. These HRMS-based data-miningtechnologies were developed originally for the detection and identifica-tion of drug metabolites in complex biologic systems (Bateman et al.,2007; Ma and Zhu 2009; Zhu et al., 2011; Ma and Chowdhury, 2012;Geng et al., 2014; Du et al., 2015). In addition, mass spectral treessimilarity filter (MTSF) technology (Jin et al., 2013) is employed for thedetection and identification of TCM metabolites on the basis of thesimilarity of their product ion spectra to those of their precursor species.These targeted data-mining approaches are capable of searchingfor metabolites of individual TCM prototypes on the basis of themetabolite’s predicted mass defect (MDF), fragmentation pattern(NLF and PIF), or molecular weight (EIC). However, since an herbalmedicine can contain up to a few hundred parent components, searchingfor their metabolite components on the basis of their masses, massdefects, or product ion spectra predicted from individual TCMprototypes is truly time-consuming and labor-intensive. More impor-tantly, many major TCM metabolite components in plasma or excretaare formed via multiple steps of biotransformation so that their detectionby targeted data-mining tools may fail. The second analytical challengefaced in a TCM ADME study is determination of metabolic pathways ofindividual TCM parent components, especially when dealing with a fewhundred in vivo TCM components.The main objective of the current study was to develop a practical and

integrated approach for in vivo ADME study of TCM medicine. Toevaluate the utility and effectiveness of this approach, metabolism anddisposition of an active fraction of Xiao-Xu-Ming decoction (AF-XXMD) as a model TCM prescription were determined in rats. Theprescription of Xiao-Xu-Ming decoction is used for the treatment oftheoplegia and the sequelae of theoplegia. The formula of XXMDconsists of 12 crude herbal medicines (Supplemental Table 1).AF-XXMD was a kind gift from Professor Hailin Qin and showssimilar pharmacological effects as XXMD. About 68 prototypecomponents were previously identified in AF-XXMD (Wang et al.,2014). However, to date there has been no report on the detection andcharacterization of XXMD or AF-XXMD metabolites in animals andhumans.

Materials and Methods

Materials.Methanol and acetonitrile ofMS grade and formic acid of analyticalgrade were purchased from Mallinckrodt Baker Co. (Phillipsburg, NJ). Purifiedwater used in the study was provided by Wahaha Co., Ltd. (Hangzhou, China).Wistar rats (200 6 20 g) were purchased from Beijing Vital River ExperimentalAnimal Co. Ltd. (Beijing, China).

Animal Experiment. Rats were kept in an environmentally controlledbreeding room for 3 days before the experiment and then fasted (water only)for 12 hours in metabolic cages prior to the dosing of AF-XXMD (0.5 g/kg). Therats [24 intact rats and three bile duct–cannulated rats (BDC)] were divided to ninegroups (three rats per group). Bile samples were collected from the group of BDCrats 2 hours prior to dosing (control bile samples) and 0–4, 4–12, and 12–24 hourspostdosing (test bile samples). Another group of rats were kept in metabolic cages,and control samples of urine and feces were collected 0–4 hours prior to theadministration. Urine and feces samples for testing were collected 0212 hoursand 12224 hours postadministration. Blood samples were collected at 0 (controlsamples), 0.5, 1.25, 3, 8, 12, and 18 hours from the abdominal artery of theremaining seven groups of the rats (one time point per group of rats). Plasmasamples prepared from collected blood samples were pooled at equal volumes inindividual time points across rats and then placed in 15-ml plastic centrifuge tubes.The centrifuge tubes were then vigorously vortexed for 30 seconds prior to storageat 280�C until use. The urine, bile, and feces samples were pooled in equalvolumes or weights for each time period across rats, then they were kept in a

refrigerator at –80�C. The experiment was approved by the Animal Care andWelfare Committee of the Institute of Materia Medica, Chinese Academy ofMedical Sciences and Peking Union Medical College (Beijing, China).

Sample Preparation. Pooled bile (1.0 ml), urine (2.0 ml), or plasma samples(2.0 ml) were mixed with 5 volumes of methanol in test tubes and then vigorouslyvortexed for 30 seconds. After centrifugation at 1721g for 10 minutes,supernatants were evaporated to dryness under a stream of nitrogen at 40�C.The residues were dissolved in 0.1 ml methanol. Each dissolved sample was thencentrifuged at 15,493g for 10 minutes, and 5 ml of supernatant was injectedinto the LC-HRMS system. Pooled feces samples were powdered, weighed,and added to an equal volume of saline. A 10� volume of methanol was thenmixed with the fecal solutions. After vortexing, ultrasonic extraction for 30minutes, and centrifugation at 1721g for 10 minutes, supernatants were thencollected and filtered through a 0.45-mm nylon filter film, and 5-ml aliquotswere injected into the LC-HRMS system.

Chromatography/Mass Spectrometry Analysis. Analyses of AF-XXMDcomponents in the dosing solution and rat plasma, urine, bile, and feces sampleswere carried out using an LC-HRMS system. A Surveyor LC plus system(Thermo Fisher Scientific, San Jose, CA) equipped with a Surveyor MS pumpplus, a Surveyor autosampler, a Thermo BDS HYPERSIL C18 column (150 �2.1 mm, 3 mm), and an Agilent SB-C8 guard column (12.5� 2.1 mm, 5 mm) wasemployed for separation. The mobile phase consisted of water containing 0.1%formic acid (A) and acetonitrile (B) delivered at a flow rate of 0.2 ml/min using agradient program as follows: 0–5 minutes, A: 95–95%, B: 5–5%; 5–25 minutes,A: 95–70%, B: 5–30%; 25–35 minutes, A: 70–60%, B: 30–40%; 35–45 minutes,A: 60–20%, B: 40–80%; 45–50 minutes, A: 20–20%, B: 80–80%; 50–51minutes, A: 20–95%, B: 80–5%; 51–60 minutes, A: 95–95%, B: 5–5%. Thecolumn temperature was maintained at 30�C and the sample injection volumewas 5 ml.

An LTQ FT mass spectrometer (Thermo Fisher Scientific) was coupled to theLC system via an electrospray ionization interface. Ultrahigh-purity helium (He)was used as collision gas and high-purity nitrogen (N2), as nebulizing gas. Theoperating parameters in the positive ion mode were as follows: ion spray voltageat 4.0 kV, capillary temperature at 250�C, capillary voltage at 40 V, sheath gasflow rate of 40 (arbitrary units), auxiliary gas flow rate of 10 (arbitrary units),sweep gas flow rate of 3 (arbitrary units), and tube lens at 90 V. Compounds weredetected by full-scan mass analysis from m/z 100 to 1200 at a resolving power of50,000 with data-dependent MS/MS analysis triggered by the two most abundantions from full-scanmass analysis, followed byMSn analysis on the most abundantproduct ions. Collision-induced dissociation was conducted with an isolationwidth of 2 Da. The collision energy was set to 35%. Dynamic exclusion wasconducted by utilizing a repeat count of one prior to exclusion. Each mass-to-charge (m/z) value resided on the dynamic exclusion list for 30 seconds afterperforming a data-dependent MSn experiment to generate MS2 and MS3 spectra.Data acquisition was performedwith Xcalibur version 2.0 SR 2 software (ThermoFisher Scientific).

Background Subtraction Data Processing. Background subtraction wasperformed using an in-house developed precise-and-thorough background-subtraction (PATBS) algorithm, which was previously described and applied tothe detection of drug metabolites (Zhang and Yang, 2008; Zhang et al., 2008). Inthe PATBA processing, a specified time window was set at60.5 minutes arounda chromatographic time point of a full MS dataset of a dosed sample. Masstolerance window around the same ions present in the full MS dataset of thedosing sample was set to 610 ppm and a specified scaling factor that wasmultiplied with the highest intensity of the identified ion in the spectra of thecontrol sample was set to 2.

Data Processing by MTSF. MTSF was applied to generate information onthe spectral similarity between two TCM components. LC-HRMS data acquiredwith Xcalibur version 2.0 SR2 software were processed using the Mass Frontier7.0 software (ThermoFisher Scientific) to construct mass spectral trees. Toconvert HRMS and multiple-stage mass spectrometric data (including MS2 andMS3 data) of all detected compounds to mass spectral trees data, a total extractioncomponent detection algorithm was used to detect components with the setting asfollows: Tree-branching began at the second MS stage with a tree match factor of90%. The library of mass spectral trees was built by importing mass spectral treesdata for template compounds consisting of the parent components in the dosingsolution. The MTSF technique with the search type setting of “similarity” wasapplied to screen the potential TCM-related compounds on the basis of similarity

ADME of TCM Components by HRMS and Data Processing 801

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

by comparing the mass spectral trees of detected compounds with those oftemplate compounds in the library. The similarity score thus obtained, amongwhich the highest was 1000, was connected with the deviation of the parent ionand daughter ion mass, as well as with the matching of daughter ion categories. Ahigher score indicated a stronger structural correlation between two compoundstested.

Biotransformation Matching. Once a spectral similarity score between twocomponents was determined to be higher than 200, the shift of molecular weightfrom one to another component was searched automatically against mass shiftsfrom the parent drug to a metabolite via common biotransformation reactionslisted in an Excel spreadsheet. If the shift matched a biotransformation reaction(mass error less than 10 ppm), the metabolic relationship between the two TCMcomponents was established. Furthermore, the established metabolite formationpathways were confirmed on the basis of the interpretation of their accurate massMSn spectral data.

Results

Integrated Analytical Strategy for Study of ADME of MultipleComponents of TCM in Vivo. A new integrated approach fordetermining exposure, metabolism, and disposition of multiple TCMprototype components in animal species and human was developed andevaluated (Fig. 1). In this study, dosed plasma, urine, feces, and bilesamples (dosed samples) were collected from rats after the administra-tion of a TCM prescription that contained more than two hundredprototype components. Control samples were collected from the samerats prior to administration. As an alternative, control samples could alsohave been collected from a control group in which dosing formulationwithout TCM had been administered. The dosed and control samples as

well as the dosing solution were subjected to analysis by LC/HRMS. Thefirst step of the analytical approach was to acquire MS and MS/MSspectral data sets for TCM chromatographic components using a data-dependent MS/MS acquisition method on an Fourier transform ioncyclotron resonance instrument. The second step was to discover bothTCM-prototype and -metabolite components by processing collectedfull-scan MS datasets using a background-subtraction tool (PATBS). Ifa TCM component was uncovered by the PATBS, but its MS/MSspectrum was not acquired by the data-dependent method, an additionalinjection to record its MS/MS orMSn spectral data would also need to beperformed. The third step was to connect TCM metabolite componentsto their parents or metabolic precursors (upstream metabolites) onthe basis of the similarity of their MS/MS spectra and matching ofbiotransformation reactions. The MS/MS spectral similarity amongAF-XXMD components was determined using MTSF. The biotransfor-mation reaction matching was carried out using an Excel spreadsheet inwhich common metabolic reactions were listed. The fourth step was todetermine metabolite structures on the basis of their MS/MS spectraldata and their formation pathways. The spectral data of the parentcomponents were used to facilitate metabolite structural elucidation.Finally, an ADME profile of a key individual TCM prototype wasdetermined on the basis of its biotransformation network established inplasma, urine, bile and feces.Profiling and Characterization of AF-XXMD Prototype

Components in the Dosing Solution. Figure 2A displays a total ionchromatogram (TIC) of a high-resolution, accurate MS dataset of theAF-XXMD dosing solution, in which a total of 198 AF-XXMDprototypes were detected. Molecular formulas andMS2 andMS3 spectraldata of these AF-XXMD prototypes are summarized in SupplementalTable 1. Among the 198 prototypes, 68 components were structurallycharacterized (Supplemental Fig. 1) on the basis of their MSn spectra andcomparisons with those of 14 reference standards. The same prototypeswere previously identified in an AF-XXMD dosing solution (Wanget al., 2014). Additionally, 14 new, minor AF-XXMD prototypes werestructurally characterized on a basis of their msn spectra (SupplementalFig. 1). Structures of the remaining 116 prototype components in theAF-XXMD dosing solution were not characterized, although theirMS/MS spectra were recorded by HRMS. Large scale isolation ofthese unknowns followed by NMR analysis would be an ideal way todetermine their structures.Detection and Characterization of AF-XXMD–Related

Components in Rat Bile, Urine, and Feces. Therewere only three AF-XXMD components displayed in the TIC of full MS dataset of a pooledbile sample (0–4 hours) at retention times between 24–30minutes (Fig. 2Band Supplemental Fig. 2A). A majority of the AF-XXMD componentswere buried under high levels of background noise or coeluted withintense endogenous components. After background subtraction againsta full MS dataset collected from a pooled, predose rat bile sample(control sample), most endogenous chromatographic ion componentswere removed (Fig. 2C). As a result, 116 AF-XXMD–related compo-nents were found in bile samples (0–4 hours, 4–12 hours, and12–24 hours; Fig. 2C, Supplemental Fig. 2B and Supplemental Table 2).To differentiate metabolite components from AF-XXMD prototypecomponents, the processed dataset was further subtracted by the full MSdataset of the dosing solution (Fig. 2A). The resultant ion chromato-gram of the dosed bile sample only exhibited AF-XXMD metabolitecomponents without the AF-XXMD prototype components (Fig. 2D).Thus, about 27 AF-XXMD prototypes and 89 metabolites wereconfirmed in the bile samples (0–24 hours) (Table 1). Similarly,urine and feces samples were processed using the PATBS process(Supplemental Figs. 3 and 4). As a result, about 101 AF-XXMD–relatedcomponents, including 70 metabolites, were found in urine samples

Fig. 1. An integral analytical strategy for detection, structural characterization, andmetabolic pathway identification of TCM components in animals and human usingHRMS and data-processing tools.

802 Wu et al.

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

(24 hours) and about 121 AF-XXMD–related components including 22metabolites were detected in the fecal samples (0–24 hours) (Table 1).Detection and Characterization of AF-XXMD–Related Compo-

nents in Rat Plasma. A total ion chromatogram of a full MS analysisof a pooled, dosed plasma sample (1.25 hours postdose) showed onlyone AF-XXMD component, H56, with multiple intense endogenous

components and high background noises were also present (Fig. 3A).After PATBS processing, many minor AF-XXMD components wererevealed in the processed TIC (Fig. 3B). About 21 components weredisplayed in zoomed area (15–30 minutes) of the ion chromatogram(Fig. 3C) that were absent or present in significantly lower abundancein the control plasma. On the basis of structures and formation pathwaysof the metabolites determined, 17 components were confirmed as AF-XXMD–related components in this plasma sample (1.25 hours post-dose), 11 ofwhichwereAF-XXMDmetabolites (Supplemental Table 3).LC/UV profiles of plasma samples showed that H56 was a single AF-XXMD component detected by UV (data not shown), suggesting thatH56 was the single major AF-XXMD component in rat plasma.Furthermore, the structure of H56 was determined to be cimicifugin(data not shown). Similarly, other plasma samples (0.5, 3, 8, 12, and 18hours postdose) were processed using the same method. As a result, atotal of 13 AF-XXMD prototypes and 34 metabolites were found in thedosed plasma samples (Tables 1 and Fig. 3C and Supplemental Table 2).Characterization of Structures and Formation Pathways of

AF-XXMD Metabolite Components. As stated above, a total of 339components, which were either absent or present at significantly lowerlevels in control samples, were revealed in dosed plasma, bile, urine,and feces samples (Table 1 and Supplemental Table 2). Among them,247 were identified AF-XXMD components, including 113 prototypesand 134 metabolites. The rest (92) of the detected components wereeither unidentified AF-XXMD metabolites or elevated endogamouscomponents. None of the identified AF-XXMD metabolites had been

Fig. 2. Untargeted analysis of multicompo-nents of AF-XXMD in a bile sample (postdose0–4 hours). (A) Zoomed area (24–30 minutes)of TIC of the AF- XXMD dosing solution. (B)Zoomed area (24–30 minutes) of TIC of apooled rat bile sample. (C) Zoomed area (24–30minutes) of TIC of the same bile sample afterPATBS process against an LC/MS dataset froma pooled control rat bile. (D) Zoomed area (24–30 minutes) of TIC of the same bile sampleafter sequential PATBS processes againstLC/MS datasets from the pooled control ratebile sample and the dosing solution. Red =:AF-XXMD–parent components. Blue =: AF-XXMD–related components displayed in theTIC of the rat bile sample without data processGreen =: metabolites of the AF-XXMDprototype components in the rat bile samplerevealed and confirmed via PATBS and MTSFprocesses, respectively. Yellow =: unknowncomponents displayed in the ion chromatogramof the rat bile sample after sequential PATBSprocesses. Labeled peaks: H stands for AF-XXMD prototype components presented in thedosing solution, M stands for metabolites of theAF-XXMD prototype components, and U standsfor unknown components.

TABLE 1

The total number of individual AF-XXMD components and unknowns detected inrat samples by HRMS and background subtraction

SampleaAF-XXDMPrototypesb

AF-XXMDMetabolitesc

Unknownsd

Plasma (0.5,1.25,3.8,12,18 h) 13 34 15Urine (0–12, 12–24 h) 31 70 30Bile (0–4, 4–12, 12–24 h) 27 89 39Feces (0–4, 4–12, 12–24 h) 99 22 26All of samples abovee 113 134 92

aRats were dosed with AF-XXMD (0.5 g/kg) and plasma samples (6 time points), urine, bileand feces were collected.

bProposed structures and mass spectral data of AF-XXMD prototypes are shown inSupplemental Table 1 and Supplemental Figure 1.

cProposed structures and mass spectral data of AF-XXMD metabolites are shown inSupplemental Table 2.

dUnknown components included: 1) AF-XXMD–related components that were eitherunknown AF-XXMD prototypes or unknown metabolites of AF-XXMD prototypes and 2)endogenous components whose levels were significantly elevated in the dosed samples.

eAll of AF-XXMD prototypes and metabolites found in all dosed samples are shown inSupplemental Table 2.

ADME of TCM Components by HRMS and Data Processing 803

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

previously reported (Wang et al., 2009). To build the connectionsbetween a metabolite and its precursor species, which could be either aprototype or an upstream metabolite, MTSF was employed to processMS2 and MS3 spectral data to determine structural similarity among allof detected AF-XXMD components. First, the 198 parent compounds inthe dosing solution were used as template compounds to set up a massspectral trees library. Then, the mass spectral trees were established forthe 226 components (134 identified AF-XXMD metabolites and 92unidentified components) discovered by PATBS. Finally, a similarityscore between two components was calculated. A metabolite that had asimilarity score greater than 200with respect to anAF-XXMDprototypecomponent or another metabolite was considered to be related toeach other structurally. Furthermore, a mass difference between thetwo components was calculated and then matched with mass shiftsof common metabolites from its parent drug using biotransformationmatching to confirm their metabolic relationship. In the end, theformation pathway and structure of the metabolite were determined onthe basis of the structural similarity, matching of the biotransformationreaction, and interpretation of their MS2 and MS3 spectra. For example,M29 was detected in plasma, bile, and urine by PATBS (Fig. 3C,Supplemental Figs. 2B and 3B, and Supplemental Table 2), and its MS2

andMS3 were retrieved fromMSn dataset (Fig. 4A). Searches for similarstructures ofM29 on the basis of its spectral tree against theMSn spectrallibrary of AF-XXMD components led to the identification of H35, H25,and H56, each of which had a Similarity Score (against M29) greaterthan 650 (Fig. 4B). Thus, M29 was determined to have a structuresimilar to those of H35, H25, and H56 (Fig. 4C). The follow-upbiotransformation matching found that the molecular ion of M29 was atm/z 483.1493 (Supplemental Table 2), 176 greater than that of H56 and amatch for a glucuronidation reaction. Therefore, M29 was identified as aglucuronide conjugate of H56 (Figs. 4D). Likewise, H33 and H35 wereidentified as linked to H56. In addition, M17,M19,M27,M29, andM42

were determined to be associated with H56 (Supplemental Tables 2 and3). On the basis of results from processing by MTSF and biotransfor-mation matching, the biotransformation network of H56 was determinedand showed the connections of H56 with various AF-XXMD compo-nents via metabolism, (Fig. 5).Metabolism and Disposition of H56 in Rat. The ADME profile

of H56 in rat was determined (Fig. 6) on the basis of the proposedbiotransformation network of H56 (Fig. 5) and the AF-XXMDcomponent profiles determined in plasma, urine, bile, and feces (Figs. 2and 3, Supplemental Figs. 2B, 3B, 4B, and Supplemental Tables 2 and3). H56 was a prototype in the dosing solution (Supplemental Table 1)and observed in feces, bile, plasma, and urine (Supplemental Table 2),suggesting that a part of H56 was absorbed via the gastrointestinal (GI)track and unabsorbed H56 was eliminated directly into feces (Fig. 6).Additionally, H33 (a prototype) can be converted to H35 (a prototype)via the loss of the xylose moiety, and then H35 can be converted to H56via deglycosylation in the GI track (Fig. 5). Absorbed H56 went tothe liver and underwent extensive hepatic metabolism. A majority ofmetabolites (M17, M19, M29, M42) and precursor species (H33 andH35) of H56 were excreted into bile (Supplemental Fig. 2 andSupplemental Table 2), and H56 and some of its metabolites, M19and M29, entered the circulation and then were excreted into urine viakidney. M17 and M27, two metabolites of H56, were not seen in thecirculation (Figs. 3 and 6) but were observed in urine (SupplementalFig. 3 and Supplemental Table 2), suggesting these metabolites werequickly eliminated via kidney after forming in the liver and quicklypassing through the circulation (Fig. 6).

Discussion

In this study, we took advantages of PATBS, a unique backgroundsubtraction algorithm, for sensitively and comprehensively detecting

Fig. 3. Untargeted analysis of multi-components of AF-XXMD in a pooled rat plasma (postdose 75 minutes). (A) TIC of the rat plasma sample without the data processing.(B) TIC of the same sample from background subtraction using PATBS. (C) Zoomed area (15–30 minutes) of the TIC displayed in Figure 3B. Blue =: AF-XXMD–relatedcomponents detected in the plasma sample without data processing. Green =: AF-XXMD–related components in the plasma sample revealed by background subtraction.Yellow =: unknown components displayed in the TIC of the rat bile sample after sequential PATBS processes. Red: endogenous components. H stands for AF-XXMDprototype components presented in the dosing solution. M stands for metabolites of the AF-XXMD prototype components. U stands for unknown components.

804 Wu et al.

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

TCM components in a complex biologic matrix (Fig. 1). PATBS isoriginally developed for untargeted detection of drug metabolites, suchas in vitro drug metabolites (Zhang and Yang, 2008), in vivo drugmetabolites (Zhang et al., 2008; Zhu et al., 2009), and modified peptideshydrolyzed from protein-drug adducts (Zhang et al., 2015). In addition,the combination of PATBS and targeted data mining (EIC, MDF,isotope pattern filter) significantly reduces false positives and improvesdetection sensitivity (Zhang et al., 2008; Xing et al., 2015). In the current

study, TCM component profiles in bile (Fig. 2), plasma (Fig. 3), urine(Supplemental Fig. 3), and feces (Supplemental Fig. 4) samples werequickly generated by PATBS, revealing any components that werepresent solely in a dosed sample or had concentrations significantlyhigher in a dosed sample than those in a control sample. A majority ofthese detected components, shown in Figs. 2C, 2D, 3B, 3C, Supple-mental Figs. 2B, 3B, and 4B, were AF-XXMD–related components.Additionally, endogenous metabolites significantly elevated owing to

Fig. 4. Identification of AF-XXMD components that have similar structures to M29 using MTSF. (A) Mass spectral tree of M29; (B) Detection of H35, H25, and H56 usingM29 as a template and MTSF; (C) Structures of H35, H25, and H56; (D) Structure of M29 that was determined by comparing spectral trees of H56 and M29.

Fig. 5. Proposed biotransformation network of H56 in rats. The term of “-glc” stands for loss of glucose. The term of “-xyl” stands for loss of xylose.

ADME of TCM Components by HRMS and Data Processing 805

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

the exposure to AF-XXMD components were detected by the process(Table 1). The detection of elevated endogenous biomarkers by PATBSwas reported previously (Zhang et al., 2010).To overcome the limitations of targeted data-mining techniques, a

few untargeted analysis methods, including metabolomics (Xie et al.,2012; Yan et al., 2013) and in-house-developed background subtractiontechniques (Gong et al., 2012; Yan et al., 2013), have been applied to thedetection of TCM components in vivo. In this study, PATBShas demonstrated several distinct advantages over targeted and otheruntargeted LC/MS approaches in detecting TCM components incomplex biologic samples. First, PATBS was capable of rapid andunbiased detection of TCM components regardless of their molecularweights, mass defects, and fragmentation patterns. Additionally, de-tected TCMmetabolite (Fig. 2D) can be immediately differentiated fromTCM prototypes via the subtraction of prototypes in the dosing solution(Fig. 2A) from the detected total TCM-related components (Fig. 2C).Second, PATBS had superior sensitivity and coverage in finding TCMcomponents in complex biologic samples. The background subtractionprocessing not only completely removed intensive endogenous chro-matographic components to reveal overlapped TCM components in ratplasma (Fig. 3C), bile (Fig. 2C), urine (Supplemental Fig. 3B), and feces(Supplemental Fig. 4B) but also significantly reduced background levelsso that minor TCM components were revealed. Third, PATBS had goodselectivity in detecting TCM components in complex biologic samples.All of the significant components displayed in PATBS-processedchromatograms of the dosed-bile (Fig. 2C) and plasma (Fig. 3C)samples were those that either did not exist (TCM components orother xenobiotics) or had much higher abundance (elevated endogenousmetabolites) in these dosed samples compared with those in thecorresponding control samples.Among 339 components detected in the plasma, urine, bile, and feces

by background subtraction (Table 1 and Supplemental Table 2), 247components were identified to be AF-XXMD–related components,including 134 newly characterized metabolites of individual AF-XXMDprototypes. It was demonstrated previously that MTSF can quantita-tively evaluate structural similarity among multiple unknown compo-nents by comparing their fragmentation patterns and product ions(Sheldon et al., 2009; van der Hooft et al., 2011; Ridder et al., 2012;

Rojas-Cherto et al., 2012). Furthermore, a biotransformation-matchingmethod was employed to define the relationship between two TCMcomponents after high structural similarity was determined by MTSF.The utility of the MTSF and biotransformation-matching processesin analyzing metabolic pathways of individual prototypes was demon-strated in the determination of the biotransformation network of H56 inrats (Fig. 5).Determination of absorption of individual TCM components after oral

administration of a TCM prescription to animals and humans is one ofthe key tasks of a TCM ADME study. A majority of TCM componentprofiling studies reported in the literature focused on the detection andidentification of TCM components in plasma using HRMS-basedtargeted data-mining tools (Xue et al., 2011, 2014; Sun et al., 2013;Tao et al., 2015). However, such an experimental approach cannot fullyevaluate which TCM prototypes were absorbed, since absorbed TCMprototypes may not be present in the circulation owing to fast metab-olism, high affinity to certain tissues, or direct elimination via bile. TCMcomponent profiles in plasma, urine, bile, and feces provided compre-hensive information on the absorption of individual prototypes in a TCMprescription. For example, AF-XXMD component profiles of all dosedsamples found 68 AF-XXMD prototypes and one metabolite of anotherprototype in feces but not in plasma, urine, or bile, suggesting that the69 AF-XXMD prototypes were not absorbed via the GI track inrat (Supplemental Table 2). About 26 AF-XXMD prototypes werecompletely absorbed, since they were absent in feces. Additionally, 39prototypes were partially absorbed because they were present in feces(Supplemental Table 2). The number of absorbed AF-XXMD prototypecomponents determined on the basis of plasma profiles was only 20%of the total of absorbed AF-XXMD prototypes determined on the basisof information collected from profiling plasma, urine, bile, and feces(Supplemental Table 2), suggesting that TCM component profiles inplasma did not provide accurate information on the absorption. TheADME data clearly demonstrate that a few of the absorbed AF-XXMDprototypes (such as H32) were directly eliminated via bile beforeentering the circulation and a few of the absorbed AF-XXMDprototypes(such as H101) were extensively metabolized in the GI tract followed bydirect biliary excretion. Several absorbed AF-XXMD components (suchas H71) or their metabolites (such as M57) were present in urine but not

Fig. 6. Proposed ADME profile of H56 in rats.

806 Wu et al.

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

found in plasma because these components were quickly eliminated viakidney after entering the circulation. These TCM components canexpress biologic effects in kidney, even though they were not detected inplasma because of low abundances or absence.To further demonstrate the utility and effectiveness of the approach

(Fig. 1) for studying ADME of individual TCM prototypes in vivo,we constructed the formation, metabolism, and elimination pathwaysof H56 (Fig. 6). H56, identified as cimicifugin, was the single majorAF-XXMD component in rat plasma. Its role in expressing thepharmacological effects of AF-XXMD is currently under investiga-tion. In addition to direct absorption via the GI track, H56 in plasmacould be formed from H35 via deglycosylation (Fig. 5). H33 could bealso metabolically converted to H56 via the formation of H35. BothH35 and H56 were major prototype components in the AF-XXMDdosing solution (Wang, et al., 2014). Good oral absorption of H56,rapid conversion fromH33 and H35 to H56, and slowmetabolism andurinary excretion of H56 in rats could be the reasons why the H56level in rat plasma was very high compared with other AF-XXMDprototype components or metabolites (Fig. 3C). Results from theADME profiling experiment also suggest that H56 underwent threemajor metabolic reactions: glucuronidation to M19, sulfation to M17and M29, and hydroxylation to M27. These metabolites were mainlyeliminated via biliary and urinary excretions. M17 and M27 were alsofound in the feces, suggesting that the two metabolites may passthrough intestinal membrane into feces or were formed in the GI track(Fig. 6).In summary, a new and integrated approach (Fig. 1) for study of

exposure, metabolism, and disposition of multiple herbal prototypesin vivo was developed and applied to ADME study of AF-XXMDin rats after an oral administration. The approach used HRMS toacquire accurate full MS and MSn data in plasma and excreta.Confirmation of TCM prototypes and detection of unknown TCMmetabolites were accomplished using PATBS, a unique backgroundsubtraction algorithm. As a result, over 247 AP-XXMD–relatedcomponents were detected and structurally characterized in ratplasma, urine, bile, and feces (Table 1 and Supplemental Table 2).It was evident that this untargeted data-mining technique hassignificant advantages over targeted data-mining technologies withrespect to sensitivity, selectivity, analytical speed, and comprehen-siveness in finding TCM components in complex biologic ma-trixes. Among 198 AF-XXMD prototype components dosed in rats(Supplemental Table 1), 26 prototypes were completely absorbedand 39 prototypes were partially absorbed (Supplemental Table 2).In the rat plasma, 13 AF-XXMD prototypes and 34 metabolites weredetected and structurally characterized (Fig. 3 and SupplementalTable 3), among which H56 was determined to be the singledominant component in the circulation. Glucuronidation, isomeri-zation, and deglycosylation followed by biliary and urinary excre-tions played major roles in the clearance of a majority of AF-XXMDprototype components in rats. About 31 and 27 prototype compo-nents were found in urine and bile, respectively (SupplementalTable 2), suggesting that direct elimination of absorbed prototypesvia kidney and liver were significant clearance pathways for some ofthese prototypes. Furthermore, an approach that combined MTSFand biotransformation matching was applied to connect metabolitesto their metabolic precursors, either prototypes or upstream metab-olites, which enabled the rapid establishment of metabolic pathwaysof individual TCM prototypes. As an example, ADME profile ofH56 was determined (Figs. 5 and 6). These results demonstrate thatthe integrated approach is a useful tool for qualitative study ofADME of multiple components in a TCM prescription in animalsand humans.

Acknowledgments

We thank Xin Wang for the development, and implementation of the precise-and-thorough background-subtraction (PATBS) data processing software.

Authorship ContributionsParticipated in research design: Wu, Zhu, J. Zhang.Conducted experiments: Wu, H. Zhang, Wang.Contributed reagents: Qin.Performed data analysis: Wu, H. Zhang.Wrote or contributed to the writing of the manuscript: Wu, Zhu, J. Zhang.

References

Bateman KP, Castro-Perez J, Wrona M, Shockcor JP, Yu K, Oballa R, and Nicoll-Griffith DA(2007) MSE with mass defect filtering for in vitro and in vivo metabolite identification. RapidCommun Mass Spectrom 21:1485–1496.

Cheng ZY, Tian X, Gao J, Li HM, Jia LJ, and Qiao HL (2014) Contribution of baicalin on theplasma protein binding displacement and CYP3A activity inhibition to the pharmacokineticchanges of nifedipine in rats in vivo and in vitro. PLoS One 9:e87234.

Ding YJ and Chen YH (2012) Developmental nephrotoxicity of aristolochic acid in a zebrafishmodel. Toxicol Appl Pharmacol 261:59–65.

Du LY, Qian DW, Shang EX, Liu P, Jiang S, Guo JM, Su SL, Duan JA, Xu J, and Zhao M (2015)UPLC-Q-TOF/MS-based screening and identification of the main flavonoids and their metab-olites in rat bile, urine and feces after oral administration of Scutellaria baicalensis extract.J Ethnopharmacol 169:156–162.

Geng JL, Dai Y, Yao ZH, Qin ZF, Wang XL, Qin L, and Yao XS (2014) Metabolites profile ofXian-Ling-Gu-Bao capsule, a traditional Chinese medicine prescription, in rats by ultra perfor-mance liquid chromatography coupled with quadrupole time-of-flight tandem mass spectrometryanalysis. J Pharm Biomed Anal 96:90–103.

Gong P, Cui N, Wu L, Liang Y, Hao K, Xu X, Tang W, Wang G, and Hao H (2012) Chemicalomeand metabolome matching approach to elucidating biological metabolic networks of complexmixtures. Anal Chem 84:2995–3002.

Hu SL, Zhang HQ, Chan K, and Mei QX (2004) Studies on the toxicity of Aristolochia man-shuriensis (Guanmuton). Toxicology 198:195–201.

Jia W, Du F, Liu X, Jiang R, Xu F, Yang J, Li L, Wang F, Olaleye OE, and Dong J, et al. (2015)Renal tubular secretion of tanshinol: molecular mechanisms, impact on its systemic exposure,and propensity for dose-related nephrotoxicity and for renal herb-drug interactions. Drug MetabDispos 43:669–678.

Jin Y, Wu CS, Zhang JL, and Li YF (2013) A new strategy for the discovery of epimediummetabolites using high-performance liquid chromatography with high resolution mass spec-trometry. Anal Chim Acta 768:111–117.

Ma L, Zhao L, Hu H, Qin Y, Bian Y, Jiang H, Zhou H, Yu L, and Zeng S (2014) Interaction of fiveanthraquinones from rhubarb with human organic anion transporter 1 (SLC22A6) and 3(SLC22A8) and drug-drug interaction in rats. J Ethnopharmacol 153:864–871.

Ma S and Chowdhury SK (2012) Application of LC-high-resolution MS with ‘intelligent’ datamining tools for screening reactive drug metabolites. Bioanalysis 4:501–510.

Ma S and Zhu M (2009) Recent advances in applications of liquid chromatography-tandemmass spectrometry to the analysis of reactive drug metabolites. Chem Biol Interact 179:25–37.

Posadzki P, Watson L, and Ernst E (2013) Herb-drug interactions: an overview of systematicreviews. Br J Clin Pharmacol 75:603–618.

Qiu Q, Liu ZH, Chen HP, Yin HL, and Li LS (2000) Long-term outcome of acute renal injuryinduced by Aristolochia manshuriensis Kom in rats. Acta Pharmacol Sin 21:1129–1135.

Ridder L, van der Hooft JJ, Verhoeven S, de Vos RC, van Schaik R, and Vervoort J (2012)Substructure-based annotation of high-resolution multistage MS(n) spectral trees. Rapid Com-mun Mass Spectrom 26:2461–2471.

Rojas-Cherto M, Peironcely JE, Kasper PT, van der Hooft JJ, de Vos RC, Vreeken R, Hankemeier T,and Reijmers T (2012) Metabolite identification using automated comparison of high-resolutionmultistage mass spectral trees. Anal Chem 84:5524–5534.

Sheldon MT, Mistrik R, and Croley TR (2009) Determination of ion structures in struc-turally related compounds using precursor ion fingerprinting. J Am Soc Mass Spectrom20:370–376.

Song YL, Jing WH, Yan R, and Wang YT (2014) Metabolic characterization of (6)-praeruptorin Ain vitro and in vivo by high performance liquid chromatography coupled with hybrid triplequadrupole-linear ion trap mass spectrometry and time-of-flight mass spectrometry. J PharmBiomed Anal 90:98–110.

Sun H, Wu F, Zhang A, Wei W, Han Y, and Wang X (2013) Profiling and identification ofthe absorbed constituents and metabolites of schisandra lignans by ultra-performanceliquid chromatography coupled to mass spectrometry. Biomed Chromatogr 27:1511–1519.

Tao JH, Zhao M, Wang DG, Yang C, Chen GT, Zhao X, Pu XL, and Jiang S (2015) UPLC-Q-TOF/MS-based screening and identification of two major bioactive components and theirmetabolites in normal and CKD rat plasma, urine and feces after oral administration ofRehmannia glutinosa Libosch extract. J Chromatogr B Analyt Technol Biomed Life Sci1001:98–106.

van der Hooft JJ, Vervoort J, Bino RJ, Beekwilder J, and de Vos RC (2011) Polyphenol identi-fication based on systematic and robust high-resolution accurate mass spectrometry fragmenta-tion. Anal Chem 83:409–416.

Wang CH, Wu CS, Qin HL, and Zhang JL (2014) Rapid discovery and identification of 68compounds in the active fraction from Xiao-Xu-Ming decoction (XXMD) by HPLC-HRMS andMTSF technique. Chin Chem Lett 25:1648–1652.

Wang X, Sun H, Zhang A, Jiao G, Sun W, and Yuan Y (2011) Pharmacokinetics screening formulti-components absorbed in the rat plasma after oral administration traditional Chinesemedicine formula Yin-Chen-Hao-Tang by ultra performance liquid chromatography-electrosprayionization/quadrupole-time-of-flight mass spectrometry combined with pattern recognitionmethods. Analyst (Lond) 136:5068–5076.

ADME of TCM Components by HRMS and Data Processing 807

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from

Wang Y, Ding C, Du K, Xiao Y, Wu C, Zhang J, Qin H, and Du G (2009) Identification of activecompounds and their metabolites by high-performance liquid chromatography/electrosprayionization Fourier transform ion cyclotron resonance mass spectrometry from Xiao-xu-mingdecoction (XXMD). Rapid Commun Mass Spectrom 23:2724–2732.

Wu L, Hao H, and Wang G (2012) LC/MS based tools and strategies on qualitative andquantitative analysis of herbal components in complex matrixes. Curr Drug Metab 13:1251–1265.

Xie G, Zhao A, Zhao L, Chen T, Chen H, Qi X, Zheng X, Ni Y, Cheng Y, and Lan K, et al. (2012)Metabolic fate of tea polyphenols in humans. J Proteome Res 11:3449–3457.

Xing J, Zang M, Zhang H, and Zhu M (2015) The application of high-resolution massspectrometry-based data-mining tools in tandem to metabolite profiling of a triple drugcombination in humans. Anal Chim Acta 897:34–44.

Xiong A, Fang L, Yang X, Yang F, Qi M, Kang H, Yang L, Tsim KW, and Wang Z(2014) An application of target profiling analyses in the hepatotoxicity assessmentof herbal medicines: comparative characteristic fingerprint and bile acid profilingof Senecio vulgaris L. and Senecio scandens Buch.-Ham. Anal Bioanal Chem 406:7715–7727.

Xu J, Qian D, Jiang S, Guo J, Shang EX, Duan JA, and Yang J (2014) Application of ultra-performance liquid chromatography coupled with quadrupole time-of-flight mass spec-trometry to determine the metabolites of orientin produced by human intestinal bacteria.J Chromatogr B Analyt Technol Biomed Life Sci 944:123–127.

Xue C, Jiang S, Guo J, Qian D, Duan JA, and Shang E (2011) Screening for in vitro metabolites ofAbelmoschus manihot extract in intestinal bacteria by ultra-performance liquid chromatography/quadrupole time-of-flight mass spectrometry. J Chromatogr B Analyt Technol Biomed Life Sci879:3901–3908.

Yan GL, Zhang AH, Sun H, Han Y, Shi H, Zhou Y, and Wang XJ (2013) An effective method fordetermining the ingredients of Shuanghuanglian formula in blood samples using high-resolutionLC-MS coupled with background subtraction and a multiple data processing approach. J Sep Sci36:3191–3199.

Yang M, Cheng C, Yang J, and Guo DA (2012) Metabolite profiling and characterization formedicinal herbal remedies. Curr Drug Metab 13:535–557.

Yuan SY, Yang CR, Cheng CL, Hsu SL, Liao JW, Lin CC, Chou YY, and Cheng YW (2011)Comparative nephrotoxicity of aristolochic acid and tetrandrine in vitro and in vivo. Int J Toxicol30:35–46.

Yue H, Chan W, Guo L, and Cai Z (2009) Determination of aristolochic acid I in rat urine andplasma by high-performance liquid chromatography with fluorescence detection. J ChromatogrB Analyt Technol Biomed Life Sci 877:995–999.

Zhang A, Sun H, Qiu S, and Wang X (2013) Advancing drug discovery and development fromactive constituents of yinchenhao tang, a famous traditional chinese medicine formula. EvidBased Complement Alternat Med 2013:257909.

Zhang A, Sun H, Yuan Y, Sun W, Jiao G, and Wang X (2011) An in vivo analysis of thetherapeutic and synergistic properties of Chinese medicinal formula Yin-Chen-Hao-Tang basedon its active constituents. Fitoterapia 82:1160–1168.

Zhang H, Ma L, He K, and Zhu M (2008) An algorithm for thorough background subtraction fromhigh-resolution LC/MS data: application to the detection of troglitazone metabolites in ratplasma, bile, and urine. J Mass Spectrom 43:1191–1200.

Zhang H, Patrone L, Kozlosky J, Tomlinson L, Cosma G, and Horvath J (2010) Pooled samplestrategy in conjunction with high-resolution liquid chromatography-mass spectrometry-basedbackground subtraction to identify toxicological markers in dogs treated with ibipinabant. AnalChem 82:3834–3839.

Zhang H and Yang Y (2008) An algorithm for thorough background subtraction from high-resolution LC/MS data: application for detection of glutathione-trapped reactive metabo-lites. J Mass Spectrom 43:1181–1190.

Zhang H, Gan J, Shu YZ, and Humphreys WG (2015) High-resolution mass spectrometry-basedbackground subtraction for identifying protein modifications in a complex biological system:detection of acetaminophen-bound microsomal proteins including argininosuccinate synthetase.Chem Res Toxicol 28:775–781.

Zhu YP (2002) Toxicity of the Chinese herb mu tong (Aristolochia manshuriensis). What historytells us. Adverse Drug React Toxicol Rev 21:171–177.

Zhu M, Zhang H, and Humphreys WG (2011) Drug metabolite profiling and identification by high-resolution mass spectrometry. J Biol Chem 286:25419–25425.

Zhu P, Ding W, Tong W, Ghosal A, Alton K, and Chowdhury S (2009) A retention-time-shift-tolerant background subtraction and noise reduction algorithm (BgS-NoRA) for extraction ofdrug metabolites in liquid chromatography/mass spectrometry data from biological matrices.Rapid Commun Mass Spectrom 23:1563–1572.

Zuo R, Ren W, Bian BL, Wang HJ, Wang YN, Hu H, Zhao HY, and Si N (2016) Metabolic fateanalysis of Huang-Lian-Jie-Du Decoction in rat urine and feces by LC-IT-MS combining withLC-FT-ICR-MS: a feasible strategy for the metabolism study of Chinese medical formula.Xenobiotica 46:65–81.

Address correspondence to: Dr. Mingshe Zhu, Department of Biotransforma-tion, Bristol-Myers Squibb Company, Princeton. E-mail: mingshe.zhu@bms.com.Dr. Jinlan Zhang, State Key Laboratory of Bioactive Substances and Functions ofNatural Medicines, Institute of Materia Medica, Chinese Academy of MedicalSciences and Peking Union Medical College, 2 Nanwei Road, Beijing 100050,China. E-mail: zhjl@imm.ac.cn

808 Wu et al.

at ASPE

T Journals on M

arch 16, 2021dm

d.aspetjournals.orgD

ownloaded from