Integrated Solid-Phase Extraction, Ultra-High-Performance Liquid Chromatography–Quadrupole-Orbitrap High-Resolution Mass Spectrometry, and Multidimensional Data-Mining Techniques to Unravel the Metabolic Network of Dehydrocostus Lactone in Rats

Dehydrocostus lactone (DL) is among the representative ingredients of traditional Chinese medicine (TCM), with excellent anticancer, antibacterial, and anti-inflammatory activities. In this study, an advanced strategy based on ultra-high-performance liquid chromatography–quadrupole-Orbitrap high-resolution mass spectrometry (UHPLC–Q-Orbitrap HRMS) was integrated to comprehensively explore the metabolic fate of DL in rats. First, prior to data collection, all biological samples (plasma, urine, and feces) were concentrated and purified using solid-phase extraction (SPE) pre-treatment technology. Then, during data collection, in the full-scan (FS) data-dependent acquisition mode, FS-ddMS2 was intelligently combined with FS-parent ion list (PIL)-dynamic exclusion (DE) means for targeted monitoring and deeper capture of more low-abundance ions of interest. After data acquisition, data-mining techniques such as high-resolution extracted ion chromatograms (HREICs), multiple mass defect filters (MMDFs), diagnostic product ions (DPIs), and neutral loss fragments (NLFs) were incorporated to extensively screen and profile all the metabolites in multiple dimensions. As a result, a total of 71 metabolites of DL (parent drug included) were positively or tentatively identified. The results suggested that DL in vivo mainly underwent hydration, hydroxylation, dihydrodiolation, sulfonation, methylation, dehydrogenation, dehydration, N-acetylcysteine conjugation, cysteine conjugation, glutathione conjugation, glycine conjugation, taurine conjugation, etc. With these inferences, we successfully mapped the “stepwise radiation” metabolic network of DL in rats, where several drug metabolism clusters (DMCs) were discovered. In conclusion, not only did we provide a refined strategy for inhibiting matrix effects and fully screening major-to-trace metabolites, but also give substantial data reference for mechanism investigation, in vivo distribution visualization, and safety evaluation of DL.


Introduction
Dehydrocostus lactone (DL), extracted and isolated from the roots of Aucklandia lappa Decne., Vladimiria souliei (Franch.) Ling, Saussurea costus (Falc.) Lipsch. and other plants of the Asteraceae family [1][2][3], is a natural sesquiterpene lactone ( Figure 1). In recent years, DL has been proven to exhibit broad and outstanding biological activities (such as anticancer, anti-inflammatory, antitumor, and antibacterial properties) [4][5][6][7], which makes it a potential candidate against lung cancer, laryngocarcinoma, gastrinoma, ulcerative colitis, airway allergic inflammation, and acute lung injury [8][9][10][11][12][13]. Reportedly, DL could Any form of transformation and presence of the drug after absorption into a body may be an essential component towards its efficacy. Hence, recognizing all the metabolites produced by a drug in vivo and defining their structures is an indispensable step for drug development. Ultra-high-performance liquid chromatography-high-resolution mass spectrometry (UHPLC-HRMS), a powerful analytical technique that has been developed over the past years, is widely used for the metabolic analysis of various Chinese herbal compounds and monomers with its superiority of high sensitivity, high selectivity, and high mass accuracy [16][17][18]. Yet, the signals of numerous trace metabolites are highly susceptible to being overwhelmed by the matrix effects and background noise generated during sample detection. In addition, although many post-acquisition data processing methods have been manifested [19][20][21], such as isotope pattern filtering (IPF), background subtraction (BS), and mass defect filtering (MDF), there is still a lack of effective multimethod integration in the practical application process. Consequently, an advanced strategy of "purification-high quality acquisition-multidimensional data-mining" was developed to address the shortcomings of the existing "pre-acquisition", "in-acquisition" and "postacquisition" processes for LC-MS/MS-based drug metabolism analysis, which was verified and expected to be expanded for the in vivo metabolic profiling of complicated ingredients of traditional Chinese medicine (TCM).
In fact, a refined strategy based on UHPLC-Q-Orbitrap HRMS analysis was established to comprehensively capture and characterize the major-to-trace metabolites of DL Any form of transformation and presence of the drug after absorption into a body may be an essential component towards its efficacy. Hence, recognizing all the metabolites produced by a drug in vivo and defining their structures is an indispensable step for drug development. Ultra-high-performance liquid chromatography-high-resolution mass spectrometry (UHPLC-HRMS), a powerful analytical technique that has been developed over the past years, is widely used for the metabolic analysis of various Chinese herbal compounds and monomers with its superiority of high sensitivity, high selectivity, and high mass accuracy [16][17][18]. Yet, the signals of numerous trace metabolites are highly susceptible to being overwhelmed by the matrix effects and background noise generated during sample detection. In addition, although many post-acquisition data processing methods have been manifested [19][20][21], such as isotope pattern filtering (IPF), background subtraction (BS), and mass defect filtering (MDF), there is still a lack of effective multimethod integration in the practical application process. Consequently, an advanced strategy of "purification-high quality acquisition-multidimensional data-mining" was developed to address the shortcomings of the existing "pre-acquisition", "in-acquisition" and "post-acquisition" processes for LC-MS/MS-based drug metabolism analysis, which was verified and expected to be expanded for the in vivo metabolic profiling of complicated ingredients of traditional Chinese medicine (TCM).
In fact, a refined strategy based on UHPLC-Q-Orbitrap HRMS analysis was established to comprehensively capture and characterize the major-to-trace metabolites of DL in rat plasma, urine, and feces: (1) solid-phase extraction (SPE) methodology was adopted for the fast purification and concentration of pre-acquisition biological samples to minimize matrix effects; (2) in the full-scan (FS) data-dependent acquisition (DDA) mode, FS-ddMS 2 was rationally coupled with FS-parent ion list (PIL)-dynamic exclusion (DE) for more complete and in-depth targeting analysis; (3) several technologies including diagnostic product ions (DPIs), neutral loss fragments (NLFs), high-resolution extracted ion chromatograms (HREICs), and multiple mass defect filters (MMDFs) were consolidated for data mining in multiple dimensions. With these methodological improvements, the biotransformation forms of DL in rats were successfully probed in detail for the first time and based on which the metabolic network of DL in vivo was mapped.

The Construction and Interpretation of Analysis Strategy
In this study, the whole analytical workflow was divided into four modules: animal experiment, sample purification, instrumental analysis, and data processing ( Figure 2). First, all raw biological samples (i.e., plasma, urine, and feces) from rats orally administered DL were collected. Second, all the samples were quickly enriched and purified using the SPE method in order to remove a majority of impurities and interferents such as salts, proteins, and lipids, so that the matrix effects and background noise could be both reduced. Then, by utilizing the unique DDA mode of Q-Orbitrap HRMS, FS-ddMS 2 and FS-PIL-DE were innovatively coordinated to enlarge the width and depth of the trapped ions, which could indirectly achieve the target recognition of small-molecule metabolites. Lastly, multiple data-mining techniques such as DPIs, NLFs, HREICs, and MMDFs were synthesized to screen and affirm the candidate metabolites in diverse dimensions, with preliminary identification based on chromatographic retention times, exact mass measurements, and regular fragmentation patterns, from which the metabolic network of DL in rats was further mapped. The elaborate scheme of the developed analytical strategy was presented in Figure 2.

Establishment of the MMDF Screening Method
After obtaining the multidimensional LC-MS/MS datasets of all the biological samples, it is undoubtedly crucial to effectively and accurately determine the metabolite candidates from the enormous and redundant database. Taking into account the literature survey, metabolic pathway speculation, and HREIC verification, the MDF templates were set up as listed in Table 1. The mass error for small molecules generated by mass defects was set to ± 50 mDa around the mass defect of the filter template over a mass range of 50 Da above and below the accurate mass of each individual MDF template. The settings of the upper several filter templates adequately accounted for the involvements of drug filters, substructure filters, and conjugation filters, so that endogenous substances could be maximally eliminated with some low-level or unpredictable metabolites then picked out.

Establishment of the MMDF Screening Method
After obtaining the multidimensional LC-MS/MS datasets of all the biological samples, it is undoubtedly crucial to effectively and accurately determine the metabolite candidates from the enormous and redundant database. Taking into account the literature survey, metabolic pathway speculation, and HREIC verification, the MDF templates were set up as listed in Table 1. The mass error for small molecules generated by mass defects was set to ± 50 mDa around the mass defect of the filter template over a mass range of 50 Da above and below the accurate mass of each individual MDF template. The settings of the upper several filter templates adequately accounted for the involvements of drug filters, substructure filters, and conjugation filters, so that endogenous substances could be maximally eliminated with some low-level or unpredictable metabolites then picked out.

Analysis of the Fragmentation Behaviors of DL in the Positive Ion Mode
In the post-acquisition handling phase of the analytical procedure, DPIs and NLFs were ascertained with primary reference to the characteristic fragmentation behaviors of DL. As illustrated in Figure 3 Figure 4. Given the similarity between the substructure or parent nucleus of the metabolites and the original drug, the DPIs and NLFs determined so far certainly would serve as significant aids for structure qualification.

Analysis of the Fragmentation Behaviors of DL in the Positive Ion Mode
In the post-acquisition handling phase of the analytical procedure, DPIs and NLFs were ascertained with primary reference to the characteristic fragmentation behaviors of DL. As illustrated in Figure 3 Figure 4. Given the similarity between the substructure or parent nucleus of the metabolites and the original drug, the DPIs and NLFs determined so far certainly would serve as significant aids for structure qualification.

Structural Identification of DL Metabolites
After data acquisition by UHPLC-Q-Orbitrap HRMS, the total ion chromatograms (TICs) presented by blood, urine and feces samples of the drug-treated group must be, respectively, compared to the control group in the Xcalibur 4.3 visualized datasets so that interference from endogenous substances or impurities could be excluded for more accurately identifying metabolites.

Structural Identification of DL Metabolites
After data acquisition by UHPLC-Q-Orbitrap HRMS, the total ion chromatograms (TICs) presented by blood, urine and feces samples of the drug-treated group must be, respectively, compared to the control group in the Xcalibur 4.3 visualized datasets so that interference from endogenous substances or impurities could be excluded for more accurately identifying metabolites.

Summary and Generalization of All Metabolites
Facilitated by the optimized strategy, a total of 71 DL metabolites including the parent drug (DL) were recognized and identified in the biological samples from rats. Among these metabolic products, 48 metabolites were attributed to plasma, 68 metabolites were discovered in urine, and just 24 metabolites were contained in feces. The chromatographic and mass spectrometric information from all the metabolites was summarized in Table 2, whereas the corresponding HREICs were visualized in Figure 5. Furthermore, the ESI-MS 2 spectra of several representative metabolites were presented jointly in Figure 6.

Mapping of the Metabolic Network for DL in Rats
Putting together all the metabolites detected and characterized, as well as analyzing their corresponding biotransformation reactions in rats, it was noted that the metabolic process of DL was hierarchical and progressive. At first, the parent drug (DL) was metabolized in rats to produce a few "core metabolites", which were then continuously "branched out" in a "bifurcate-and-diverge" manner, until finally "flourishing" to integrate the special "stepwise radiation" metabolic network of DL. The network diagram was visualized in Figure 7.

Mapping of the Metabolic Network for DL in Rats
Putting together all the metabolites detected and characterized, as well as analyzing their corresponding biotransformation reactions in rats, it was noted that the metabolic process of DL was hierarchical and progressive. At first, the parent drug (DL) was metabolized in rats to produce a few "core metabolites", which were then continuously "branched out" in a "bifurcate-and-diverge" manner, until finally "flourishing" to integrate the special "stepwise radiation" metabolic network of DL. The network diagram was visualized in Figure 7.  Most certainly, the "core metabolites" including M11, M16, M26, M33, M36, M54, and M68, which were produced through hydroxylation, hydration, dehydration, dihydrodiolation, deoxidization, dehydrogenation, and N-acetylcysteine conjugation reactions, were crucial and might even be the necessary aspects for DL exerting its therapeutic effect, which was also in line with the drug metabolism clusters (DMCs) described in the available publications [23][24][25].

Discussion
In this study, a refined and optimal strategy was developed by integrating SPE technology, UHPLC-Q-Orbitrap HRMS, and various data-mining techniques, which was successfully applied to the metabolic study of an active ingredient of TCM-dehydrocostus lactone (DL) in rats. Compared with the previous reports [22], our results have greatly enlarged the metabolite database of DL and successfully mapped the corresponding metabolic network in vivo using the available analytical means, which will undoubtedly contribute to the current research regarding mechanisms and drug safety monitoring of DL.
As is well known, the most challenging problem in drug metabolism study is trace metabolite signal loss as a result of matrix effects and interference of endogenous substances in the analytical sessions. This unfavorable barrier may lead to the possibility that certain small-molecule metabolites with low-level but excellent pharmacological activities are unfortunately neglected. For example, acetaminophen (paracetamol), derived from the rapid de-ethylation of phenacetin in the liver, has an amazing antipyretic-analgesic effect [26]. As such, the suppression of background noise, wide-range and efficient data acquisition, and improvement of data processing techniques are always the directions of our endeavors.
Solid-phase extraction is a sample pre-treatment technology that is being developed quickly during recent years, which combines liquid-solid extraction (LSE) and columnliquid chromatography (CLC) techniques for sample separation, concentration, and purification [27]. When the samples pass through the stationary phase, the impurities and endogenous interferents can be effectively removed by adsorption, wash-out, and elution procedures, which will ultimately result in a purified collection containing the desirable target compounds. In contrast to the traditional protein precipitation and liquid-liquid extraction (LLE) approaches, SPE can be regarded as a "micro-scale chromatography", which can minimize the matrix effects and ionization interferences with the integration of LC-MS/MS to achieve a "pseudo-two-dimensional chromatography", particularly suitable for the selective enrichment of low-level small-molecule metabolites [28]. Additionally, the simplicity and rapidity of its operation make SPE destined to become a preferred tool for the optimization of data pre-acquisition strategies in drug metabolism analysis.
For UHPLC-MS/MS analysis, the conventional data-dependent acquisition mode has been broadly utilized in the majority of food and drug analysis studies because of its large-scale database acquisition capability with full scan followed by intensity-biased secondary (MS/MS) triggered successive acquisition patterns [29][30][31]. Nevertheless, the structural information obtained in this mode is redundant and contains a considerable number of unnecessary ion pairs, as well as a limited amount of MS 2 information per data point, which may result in the absence of essential information and recognition difficulties for some compounds. Herein, based on the superior properties of Q-Orbitrap HRMS, the PIL-DE acquisition tactics coupled with FS-ddMS 2 enabled either targeted monitoring of those ions under our interest and provoked MS 2 collection (as literature findings and biotransformation reactions deduced), or dynamic exclusion on top of that to activate fragmentation for more low-abundance parent ions as well as to extend the coverage of sample constituents screening. As reported, PIL-DE could actually acquire more MS/MS information of trace ingredients [32], with the integration of FS-ddMS 2 effectively realizing the high sensitivity and selectivity of HRMS for target constituents data acquisition while applying to drug metabolism analysis.
After obtaining the datasets by instrumental analysis, the mining of useful data from the wide-volume database up to high-efficiency screening and precise characterization of trace contents is a highly challenging topic. Previous studies mostly focused on the application of one or two methodologies, so that the screening and designation of trace metabolites were not accurate and complete enough. Mass defect filtering technology is designed to rapidly screen metabolites by imposing predetermined criteria around the mass defect of certain chosen template compounds (e.g., parent drug). However, with our practice, it was still difficult to cover numerous components in complex mixtures from just a single template, so a screening process that combined multiple templates (original drug, substructures, conjugates) was necessary. For this reason, a means of postacquisition data handling merging MMDFs and HREICs was proposed, which allowed to conveniently exclude the majority of disturbing ions in the defined window from complicated matrices with further accurately pinpointing relevant metabolites at the ESI-MS 1 level. Moreover, compounds with the same parent nucleus and similar backbones display analogous fragmentation behaviors under high-energy collisional dissociation (HCD), thus providing the common DPIs and NLFs for structural elaboration, which could further serve as potent auxiliary tools for metabolites identification.
In general, the above-mentioned refined strategy was mainly aimed at the signals overwhelmed, screening omission, and structures mis-definition of trace metabolites caused by severe matrix effects, narrow acquisition coverage, and massive data processing difficulties in drug metabolism LC-MS/MS analysis. In fact, in the present study, the entire three phases of the drug metabolism experiment-"pre-acquisition, in-acquisition, and post-acquisition"-were improved, resulting in an advanced strategy that facilitated the high-efficiency and high-velocity collection for the targeted recognition of major-to-trace metabolites as well as the precise and comprehensive screening to enrich the small-molecule metabolites library. Following our verification, the strategy was promising to be well suited for in vivo metabolic process profiling studies of any Chinese medicine prescriptions or monomeric compounds.

Reagents and Chemicals
The dehydrocostus lactone (DL) reference standard was purchased from Chengdu Must Bio-technology Co., Ltd. (Sichuan, China). Its structure was well demonstrated by spectral matching with published literature (ESI-MS, 1 H, and 13 C NMR), of which the purity was no less than 98% according to HPLC-UV analysis. MS grade acetonitrile and methanol were obtained from Thermo Fisher Scientific (Fair Lawn, NJ, USA), while HPLC grade formic acid (FA) was purchased from Merck KGaA (Darmstadt, Germany). All additional chemicals of analytical grade were accessible at the work station, Beijing Chemical Works (Beijing, China). The deionized water (ddH 2 O) used throughout the analytical workflow was prepared in purification with the Milli-Q Synthesis System (Millipore, Billerica, MA, USA). Sep-Pak ® Vac C18 solid-phase extraction cartridges utilized for the pre-treatment of all biological samples were obtained from Waters (Milford, MA, USA).

Animals and Drug Administration
Six Sprague Dawley (SD) rats were acquired from SPF Biotechnology Co. Ltd. (Beijing, China). All animals were living under stable controlled conditions in the barrier environment (temperature: 23 ± 2 • C; relative humidity: 55±10%; alternating day/night time: 12/12 h) with free access to food and water. Following one week of continuous acclimatization to the environment, the rats were randomly divided into two groups: the control group (n = 3) and the drug-treated group (n = 3). In this case, the drug-treated group was given DL, which was suspended in 0.5% sodium carboxymethyl cellulose (CMC-Na) by gavage at a dose of 300 mg/kg, while the control group was orally administered an equal amount of 0.5% CMC-Na solution. The overall animal experimental program and details were authorized by the Institutional Animal Care and Use Committee at the Beijing University of Chinese Medicine, as well as all procedures were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (USA National Research Council, 1996).

Biological Samples Collection
All the rats were fasted for 12 h but were allowed to drink water freely before drug administration. After gastric administration, approximately 0.8 mL of blood was gathered from the orbital venous plexus in each rat at 0.5, 1, 2, 4, and 6 h followed by transferring to heparin sodium anticoagulation EP tubes, which were then centrifuged for 15 min (3500 rpm, 4 • C) to separate plasma.
Further, all urine and feces excreted by every individual rat within 24 h post-dosing were also collected. Urine was likewise centrifuged for 15 min (3500 rpm, 4 • C) to accumulate supernatant, whereas the feces were dried, ground, and crushed, of which 0.5 g was taken from each group and plunged into 70% ethanol (m/v = 1:10), and the supernatant was extracted by ultrasonication for 30 min, followed by 15 min centrifugation (3500 rpm, 4 • C).
Eventually, homogeneous biological samples (i.e., plasma, urine, and feces) from the same group were co-mingled into an aggregate set and stored at −80 • C.

Samples Pre-Treatment-The SPE Method
Pending instrumental analysis, SPE technology was adopted for rapid separation, purification, and concentration of all biological samples. At first, the SPE cartridges were sequentially flushed with 5 mL of ddH 2 O and 5 mL of methanol for activation and equilibration. Next, 1 mL of biological samples (plasma, urine, and feces) were separately loaded into the cartridges allowing natural adsorption. After that, the cartridges were washed again with 5 mL of ddH 2 O to remove most of the matrix interferents. Last, the cartridges were eluted twice using 1 mL of methanol so that a total of 2 mL eluate (target analytes contained) was collected, which was further evaporated by blow-drying with nitrogen under room temperature. The final residues were re-solubilized in 2% acetonitrile solution (100 µL) with centrifugation for 15 min (12,000 rpm, 4 • C) after which the supernatants were collected for UHPLC-HRMS analysis.

Instrument and Conditions
All LC-MS/MS analytical procedures were accomplished on a UHPLC-Q-Orbitrap high-resolution mass spectrometer (Thermo Scientific, Bremen, Germany) equipped with an ESI source. The chromatographic separation was carried out using a Waters ACQUITY UPLC HSS T3 column (2.1 mm × 100 mm i.d., 1. The operational parameters for Q-Orbitrap HRMS in the positive ion mode were optimized as follows: spray voltage, 3500 V; sheath gas flow rate, 50 arb; auxiliary gas flow rate, 10 arb; ion transfer tube temperature, 350 • C. All objective compositions were detected and analyzed with a resolution of 60,000 in the scan range of m/z 100-1000. Collision Energy Type was chosen as "Normalized" with HCD collision energies (CE) being set to 30%, 50%, and 70%. The dynamic exclusion (DE) assignments were configured as follows: the repeat count was 5, the dynamic repeat time was 30 s, and the dynamic exclusion duration was 60 s.

Data Processing
All LC-MS/MS data were extracted and processed on an Xcalibur 4.3 workstation. For the purpose of securing as much information as possible concerning ions fragmentation of small-molecule metabolites, we selected the peaks with an intensity threshold exceeding 50,000 for screening and identification. In the prediction interface for the formulated molecular formula, the parameters of "elements in use" were set up as: In addition, MetWorks (version 1.3) software (Thermo Scientific, Waltham, MA, USA) was employed to implement the MMDF function along with the NLF capability for extensive screening of metabolites, which combined with Mass Frontier (version 8.0) software (Thermo Scientific, Waltham, MA, USA) to predict the fragmentation behaviors of metabolites and attribute fragment ions, ultimately allowing for precise structural characterization of metabolites.

Conclusions
In the current study, an advanced strategy integrating solid-phase extraction technology, UHPLC-Q-Orbitrap HRMS analysis, and multidimensional data-mining techniques was developed, which successfully achieved rapid and in-depth targeted recognition and accurate identification of 71 metabolites from DL. Furthermore, based on the step-by-step profiling of biotransformation reactions (i.e., hydration, hydroxylation, dehydrogenation, dehydration, N-acetylcysteine conjugation, etc.), the "stepwise radiation" metabolic network of DL in rats was resoundingly mapped with the discovery of drug metabolic cluster centers such as M11, M16, M26, M33, M36, M54, and M68, which will hopefully accelerate the mechanism exploration and exploitation of DL in the medical community. Most importantly, to the best of our knowledge, this study was the first report on the comprehensive metabolism study of DL in a bio-organism, and the proposed refined strategy is prospectively and widely applicable to the metabolic profile elucidation of active ingredients from TCM in vivo.

Data Availability Statement:
The research data in this study are available from the corresponding author upon reasonable request.

Conflicts of Interest:
The authors declare no conflict of interest.