Introduction

Metabolite identification is an integral part of drug discovery and development. High-performance liquid chromatography-mass spectrometry (HPLC-MS) has become a routine tool for detecting and identifying drug metabolites. Although analytical sensitivity and the processing of data for HPLC-MS have been improving, the detection of drug metabolites, especially for minor metabolites in complex biological matrices, continues to be a challenge.

Recently, with the development of various data acquisition and mining technologies, high-resolution mass spectrometry (HRMS) has exhibited the potential to greatly increase the speed, selectivity, sensitivity, accuracy, and comprehensive nature of metabolite detection [1]. The HRMS analytical strategy is different from the traditional multiple-step or multiple-instrument LC-MS approaches, such as unity or integration of full mass multiple reaction monitoring, precursor ion, neutral loss, and product ion scans. With HRMS, metabolite detection and identification can be accomplished in one or only a few injections [1–3]. In HRMS systems, the full mass scan and product ion spectral datasets for the metabolites are acquired first. Data mining technologies are then used to find drug metabolite ions. Finally, metabolite structures can be elucidated based on the processed data [1]. There have been many studies on the development and application of data mining technologies, such as product ion filter (PIF) [4], neutral loss filter (NLF) [4], mass defect filter (MDF) [5, 6], and isotope pattern filter [7]. In contrast, data acquisition research is relatively rare. On-line data acquisition and off-line data mining are equally important, and the former is the foundation of the latter, which needs more research.

The MSE acquisition called the “all in one” scan was used on Q-TOF and LTQ Orbitrap instruments, which has been proven to streamline the complicated task studying in vitro and in vivo metabolism [8–11]. However, the product ion spectra of drug metabolites and the endogenous components may not be chromatographically discernible, rendering the correct interpretation of the observed spectra difficult, though the MSE scan is capable of fragmenting all components [12]. The intensity-dependent MS/MS acquisition method followed by data processing has been shown to be effective in the analysis of in vitro metabolites. Nevertheless, because excess quantities of endogenous components are co-eluted, the ions of the metabolites are likely to be overwhelmed with interferences from the matrix, so the MS/MS acquisition of the metabolites is not be triggered [13]. The MDF technique has been used as a valid data mining method for detecting both predicted and unexpected drug metabolites [14–16]. Different bio-transformations have distinct mass defect values. Almost all of the mass defect values of metabolites fall within a defined narrow window around that of the parent drug. With a suitable window setting, ions corresponding to metabolites can be distinguished by excluding ions outside the window, and a significant number of background interference ions can be removed [12, 17]. Accordingly, MDF-based data-dependent acquisition makes selective triggering of the MS/MS acquisition for minor metabolites possible. In addition, multiple mass defect windows over multiple mass ranges can be used to acquire MS/MS spectra for various types of drug metabolites. To the best of our knowledge, there is no detailed report on MDF-dependent acquisition and its application.

The linear-type furocoumarin, IMP (9-(3-methylbut-2-enyloxy)-7H-furo[3,2-g]chromen-7-one; Fig. 1) is an active natural component originating from many traditional Chinese herbal medicines, such as Angelica Dahurica [13], Radix Glehniae [14], and Radix Saposhnikoviae [15]. It possesses anti-inflammatory [16], antiviral [17], antitumor [18], and anticonvulsive properties [19]. Recent pharmacological research demonstrated that IMP participated in hypertension treatment by inhibiting voltage-dependent calcium channels and receptor-mediated Ca2+ influx and release [20]. With this development, IMP is expected to be pursued as a novel vasorelaxant agent. Pharmacokinetic profiles of IMP in rats indicate rapid absorption and extensive metabolic elimination after administration [21]. Less than 0.01 % of the dose was recovered from rat bile and urine as the parent drug IMP within 24 and 72 h, respectively [22]. All of these results indicate that IMP is rapidly and thoroughly metabolized in vivo and is excreted mainly as metabolites. By now, there has been a report on the IMP and its two metabolites pharmacokinetic profiles study. In this study, two metabolites of IMP were identified for the first time in dog plasma [23]. In addition, for the in vitro studies, five metabolites were prepared from the Aspergillus flavus transformation of IMP [24], and ten transformed products were isolated by the incubation of IMP with Penicillium janthinellum [25].

Fig. 1
figure 1

Chemical structure of IMP

Full size image

In the present study, an approach combining MMDF and DBS-dependent on-line data acquisition followed by comprehensive data processing techniques, including XIC (extracted ion chromatography), MDF, PIF, and NLF, was developed on a Triple TOF 5600 system. Only one injection was needed for the analysis. Herein, the data acquisition method is described in detail. The method was utilized for the detection and structure characterization of IMP metabolites in a rat urine sample. Based on accurate spectra data, relevant drug bio-transformation knowledge and previously investigated fragmentation regulations of furocoumarins [27], a total of 51 compounds were identified or tentatively assigned. The in vivo metabolic pathway of IMP and the fragmentation rules of IMP metabolites were also proposed. This was the first metabolic study of IMP using HPLC-Triple TOF-MS and multiple data-mining techniques for the rapid identification of IMP metabolites in rat urine.

Experimental

Chemicals and materials

IMP was purchased from Shanghai Sunny Biotech Co., Ltd., People’s Republic of China. Scopoletin, psoralen, isopsoralen, and isoimperatorin were purchased from the China Institute for Control of Pharmaceutical and Biological Products. Bergapten, xanthotoxin, and xanthotoxol were obtained from Shanghai Tauto Biotech Co., Ltd., People’s Republic of China. The purities of all ingredients were greater than 98 % according to HPLC analysis.

Methanol (J.T. Baker, USA) and ammonium acetate (Diamond, USA) were of HPLC grade. De-ionized water was produced by a Heal Force-PWVF Reagent Water System (Shanghai CanRex Analyses Instrument Corporation Limited, People’s Republic of China). Acetic ether and other chemicals were of analytical grade (Tianjin Chemical Corporation, People’s Republic of China).

HPLC-Triple TOF-MS conditions

HPLC experiments were conducted on a Shimadzu (Kyoto, Japan) HPLC system consisting of an LC-20AD quaternary solvent delivery system, an SIL-20AC auto sampler, and a CTO-20A column oven. Chromatographic separation was conducted on an Agilent Eclipse XDB-C18 (150 × 4.6 mm, 5 μm) column with a Security Guard C18 guard column (4.0 × 3.0 mm i.d., 5 mm; Phenomenex, Torrance, CA, USA). The column temperature was maintained at 25 °C. The mobile phase consisted of methanol (A) and water containing 1 mmol/L ammonium acetate (B). The following gradient condition was used: initial 0–3 min, linear change from A–B (20:80, v/v) to A–B (40:60, v/v); 3–25 min, linear change from A–B (40:60, v/v) to A–B (65:35, v/v); 25–30 min, linear change from A–B (65:35, v/v) to A–B (95:5, v/v); and 30–40 min, isocratic elution A–B (95:5, v/v). The mobile phase flow rate was set at 1.0 mL/min, and the injection volume was 10 μL.

A Triple TOF™ 5600 system with a DuoSprayTM source operating in the positive ESI mode was used for detection (AB SCIEX, CA, USA). The following parameter settings were used: ion spray voltage, 5.5 kV; ion source heater, 600 °C; curtain gas, 30 psi; ion source gas 1, 60 psi; and ion source gas 2, 60 psi. For the full MS-IDA (information dependent acquisition)-4MS/MS analysis, the full MS, the survey scan, and the MS/MS experiments were run in positive mode with a scan range from m/z 100 to m/z 800 and a 250-ms accumulation time for the Full MS and a 100-ms accumulation time for the MS/MS experiments. The CE was 35 eV, and the CES was 15 eV in the MS/MS experiment. The IDA was used to trigger the acquisition of MS/MS spectra for ions matching the IDA criteria. A real-time MMDF and DBS were used to fulfill the IDA criteria.

Metabolite identification was performed with Metabolite Pilot 1.5 (AB SCIEX, CA, USA) based on accurate measurements of the m/z values and processing of the post-data obtained from the XIC, MDF, PIF, and NLF screening of the probable metabolites. In addition, the elemental compositions and chemical formulae could be calculated.

Animals and drug administration

Six male Sprague-Dawley rats (divided into two groups, three in experimental group and three in blank group; 12–14 weeks of age; 200–240 g body weight) were provided by the Laboratory Animal Center of Hebei Medical University. The animal experiment protocols were approved by the Animal Center of Hebei Medical University. The animals were maintained at ambient temperature (22–24 °C) and 60 % relative humidity with a 12-h light/dark cycle. The animals were kept in an environmentally controlled breeding room for 3 days and fasted 12 h before the experiments. IMP powder was dissolved in 5‰ sodium carboxyl methyl cellulose (CMC-Na) solution and administered by oral gavage at a dose of 80 mg/kg body weight to the medication group. An equivalent CMC-Na solution with no IMP was given to the blank group.

Sample collection and pretreatment

Following the oral administration of IMP, urine samples were collected 72 h after dosing (n = 3). The urine samples were stored at −20 °C until analysis. IMP and its metabolites were extracted using a liquid–liquid extraction procedure. For the urine samples, 40 mL of urine was transferred to a 500-mL separatory funnel and mixed with 120 mL acetic ether while shaking for 10 min. Each urine sample was independently extracted three times. The extract was evaporated to dryness in a rotary evaporator, dissolved in 1 mL of methanol, then the dissolved methanol of three experiment rats were mixed and centrifuged (14,000 rpm) for 10 min, a 10-μL aliquot was then injected for LC/MS analysis. Blank urine was obtained from the rats without oral administration and was treated as urine samples.

Analytical strategy

A strategy was developed for the rapid profiling and identification of the urinary metabolites of IMP on a Triple TOF instrument with MMDF and DBS-dependent on-line data acquisition and multiple post data-mining tools (Fig. 2). The first step was on-line data acquisition. A full mass scan was performed, and accurate MS/MS data sets were obtained using a unique and novel MMDF and DBS-dependent data acquisition method. The application of DBS could intelligently differentiate the background and matrix-related MS/MS ions from drug-related MS/MS ions. Thus, the lower level of target compounds can be captured clearly. The second step was post-acquisition data mining using a combination of various data mining tools, including XIC, MDF, PIF, and NLF, to identify the IMP metabolites. The last step was structure elucidation based on accurate spectra data, relevant drug bio-transformation knowledge, and previously investigated fragmentation regulations of furocoumarins.

Fig. 2
figure 2

Analytical strategies for IMP metabolites detection and identification (XIC extracted ion chromatography, MDF mass defect filter, PIF product ion filter, NLF neutral loss filter)

Full size image

Results and discussion

A total of 51 IMP metabolites, including 44 phase I and 7 phase II metabolites (Drug metabolism often occurs in two phases. Phase I is usually an oxidation, a reduction, or a hydrolysis. Phase II is always a synthetic reaction, such as a conjugation [26].), were detected in rat urine using MMDF and DBS on-line data acquisition methods combined with multiple post data-mining methods. Glutathione (GSH) and N-acetylcysteine conjugation metabolites were not detected in our study, though the two conjugated MMDF templates of IMP were also set. The detected metabolites of IMP are shown in Table. 1.

Table 1 Summary of phase I and phase II metabolites of imperatorin in rat urine
Full size table

Establishment of MMDF on-line data acquisition methods

The bottleneck in metabolite identification is the detection of low levels of unpredicted metabolites because the metabolite ions of interest are often masked by background noise and endogenous components. In the present study, MMDF technology was used for on-line data acquisition to trace all probable metabolites when running the full mass scan. Setting the metabolite templates is a key part of the MMDF method. Customarily, the templates are set according to the properties of the prototype drug and general metabolite bio-transformation rules. For IMP, the isopentenyl group was easily removed for the formation of the ion at m/z 203 [27]. Ten templates were utilized in parallel to screen IMP metabolites based on the core structure of IMP, the product ion at m/z 203, and five metabolite bio-transformation pathways (Table. 2). Each individual MMDF template was set to approximately ±25 mDa around the mass defect of an applied filter template over a mass range of ±50 Da around the mass of the filter template. Ten MMDF templates were divided into three types: parent drug filter, substructure filter, and conjugate filter. The parent drug filter template was based on the location of IMP (C16H14O4) with the mass defect shift of 89.2091 mDa. The m/z 202 (C11H6O4) was derived from the neutral loss of C5H8 from the parent drug, and the mass defect shift of 26.6088 mDa was designed for the substructure filter. The conjugate filters used IMP and its typical fragment (m/z 202, C11H6O4) as filer templates and were designed to detect different classes of conjugated metabolites. Four types of conjugate filters, glucuronide, sulfate, GSH, and N-acetylcysteine, were established in this method, including IMP + glucuronide with a mass defect shift of 121.2973 mDa, C11H6O4 + glucuronide with a mass defect shift of 58.6970 mDa, IMP + sulfate with a mass defect shift of 46.0248 mDa, C11H6O4 + sulfate with a mass defect shift of 16.5755 mDa, IMP + GSH with a mass defect shift of 173.0165 mDa, C11H6O4 + GSH with a mass defect shift of 110.4161 mDa, IMP + N-acetylcysteine with a mass defect shift of 103.8741 mDa, and C11H6O4 + N-acetylcysteine with a mass defect shift of 41.2738 mDa.

Table 2 Mass shift, mass defect shift, and MMDF templates of IMP metabolites
Full size table

Using the designed MMDF templates, the IMP metabolites could be identified without endogenous substance interference. Despite the higher signal intensity of the biological matrices compared to the metabolites, the templates could trigger the MS/MS acquisition of minor metabolites in complex biological matrices. The representative spectra of IMP metabolites (M10, M15, and M50) obtained from MMDF on-line data acquisition mode are shown in Fig. 3. As shown in the MS1 spectrum of M15, the intensity of the quasi-molecular ion of M15 at m/z 303.0857 was much lower than that of other ions existed in the mass spectra. But the expected signal was not embedded by the higher noise signal, the characteristic product ions at m/z 285, 219, 217, 201, 189, 173 generated from the quasi-molecular ion of M15 still could be monitored. Overall, the great virtue of MMDF-dependent on-line data acquisition is that the minor metabolites in complex biological matrices can be detected in one injection cycle.

Fig. 3
figure 3

MS1 and MS/MS spectra of minor metabolites (M10, M15, M50) in complex matrices using MMDF-dependent data acquisition method

Full size image

Combined post-acquisition data mining methods

A combination of various data mining tools, including XIC, MDF, PIF, and NLF, were chosen to identify the IMP metabolites. The metabolites can be deduced on the basis of the mass shifts from the parent drug, elemental compositions derived from the accurate mass measurements, and the interpretation of accurate MS/MS spectra. The MDF technique provides a simple and effective way to identify unknown drug metabolites from full-scan data regardless of their MWs or fragmentation patterns [28]. The chromatographs of IMP metabolites filtered by MDF technique are shown in Fig. 4. By employing the XIC and MDF methods, the metabolite molecular weights and elemental compositions derived from the accurate mass measurements can be readily predicted, and accurate MS/MS spectra can also be interpreted. The PIF technique was applied to extract the precursors of selected product ions and IMP from the accurate MS/MS data. The processing of the MS/MS data using the PIF technique with 203 Da uncovered 15 IMP metabolites [see Fig. S1 in the Electronic supplementary material (ESM)]. The NLF process was an effective way to screen phase II metabolites of IMP. In Fig. S2 (described in the ESM), seven metabolites are tentatively identified.

Fig. 4
figure 4

MDF chromatograms for IMP and its metabolites in rat urine, for 0–20 min (A) and 21–40 min (B)

Full size image

Structural assignment of product ions of IMP

IMP was eluted at 26.7 min with a detected molecular ion at m/z 271.0956. The accurate mass measurement gave a theoretical elemental composition of C16H15O4 [M+H]+ ions. Subsequent product ions were observed at m/z 203 (C11H7O4; neutral loss of side chain C5H8), which was followed by the first loss of H2O, CO, or CO2 and then successive losses of CO, yielding the typical product ions at m/z 185, 175, 157, 147, 131, and 91, among which the fragment ion at m/z 185 allowed the identification of the substituent group at C-8. Moreover, the fragment ion at m/z 69 originating from the ion m/z 271 was another diagnostic ion present in the spectrum. The cleavage pathway and MS/MS spectrum of IMP are illustrated in Fig. 5.

Fig. 5
figure 5

MS/MS spectrum of IMP and its major MS fragmentation pathway

Full size image

Phase I metabolites of IMP in rat urine

M1-M6

In the extracted chromatogram from m/z 287.0789 to 287.1039 [M+H]+ (molecular ion of mono-oxidized IMP under positive mode), six major peaks (M1-M6) were observed from urine samples collected from IMP-treated rats (Fig. 6). Three metabolites, M1-M3, were detected from 15.0 to 19.0 min and showed characteristic product ions similar to those of IMP. The diagnostic product ion at m/z 85 [C5H9O]+ was obtained from the cleavage of the side chain and had a mass shift of 16 Da for the isopentenyl group, coupled with the ion at m/z 203, indicating that the position of oxidation occurred in the side chain of IMP. This indicated that metabolites M1-M3 underwent two probable mono-oxidation reactions: hydroxylation and epoxidation reaction at the side chain of IMP. In previous studies, a hydroxylation metabolite and an epoxidation metabolite of IMP were identified in dog plasma [23] and a pair of isomer hydroxylated metabolites produced by transformation of IMP with A. flavus [24]. Yet, owing to the insufficiency of mass spectrometry information in our study, the exact structures of M1-M3 couldn’t be deduced. The cleavage pathway of M1-M3 and the MS/MS spectrum of M2 are shown in Fig. 7A. Moreover, three metabolites, M4-M6, were also detected from 22.0 to 27.5 min. The product ion spectra of their quasi-molecular ions [M+H]+ mainly displayed product ions at m/z 219 (C11H7O5), which was followed by the first loss of H2O or CO and the successive loss of CO, yielding the typical product ions at m/z 201, 191, 173, 163, 145, 135, and 107. The product ions at m/z 219 and 69 suggested that oxidation occurred in the linear-type furocoumarin nucleus of IMP. The cleavage pathway of M4-M6 and the MS/MS spectrum of M5 are shown in Fig. 7B. There are five potential positions of hydroxyl groups in the structure of IMP, in which five potential mono-oxidized metabolites of IMP can be theoretically formed. Because the resulting product ions are similar, the location of hydroxyl group was not determined.

Fig. 6
figure 6figure 6figure 6

Representative XIC chromatograms of IMP and its metabolites in rat urine, C1 XIC from m/z 287.0789 to 287.1039; C2 XIC from m/z 303.0738 to 303.0988; C3 XIC from m/z 319.0687 to 319.0937; C4 XIC from m/z 289.0946 to 289.1196; C5 XIC from m/z 305.0895 to 305.1145; C6 XIC from m/z 321.0844 to 321.1094; C7 XIC from m/z 301.0582 to 301.0832; C8 XIC from m/z 233.0320 to 233.0570; C9 XIC from m/z 203.0214 to 203.0464; C10 XIC from m/z 187.0265 to 187.0515; C11 XIC from m/z 217.0370 to 217.0620; and C12 XIC from m/z 271.0840 to 271.1090; C13 XIC from m/z 351.0408 to 351.0658; C14 XIC from m/z 367.0357 to 367.0607; C15 XIC from m/z 463.1110 to 463.1360; C16 XIC from m/z 479.1059 to 479.1309

Full size image
Fig. 7
figure 7

MS/MS spectra and cleavage pathway of mono-oxidized metabolites (A M1-M3, B M4-M6, phase I metabolites)

Full size image

M7-M15

The quasi-molecular ion of m/z 303.0865 (M7), eluted at 8.24 min (Fig. 6), was 32 Da more than the quasi-molecular ion of IMP and showed a series of distinctive fragment ions at m/z 203, 185, 175, 147, and 131, suggesting that the oxidation positions occurred only in the side chain of IMP. In addition to the peak of M7, eight major peaks (M8-M15) were observed in the extracted chromatograph from m/z 303.0738 to 303.0988 (Fig. 6). The characteristic product ions at m/z 285, 219, and 217 of deoxidized IMP all appeared in the MS/MS spectrum of M8-M15, which was in close agreement with the fragmentation pathway of mono-oxidized IMP (M3-M6). The characteristic product ion at 219 Da and the neutral loss ion at 85 Da (C5H9O) suggested that oxidation occurred both in the side chain and the linear-type furocoumarin nucleus of IMP. Because the deoxidized metabolites proposed above all have similar mass spectra of MS1 and MS2, the locations of hydroxyl groups were not determined. The simultaneous appearance of characteristic product ions at m/z 219 and 217, 191 and 189, and 163 and 161 in the MS/MS spectra of M8-M15 indicated that the location of the hydroxyl and isopentenyl group was at C-5 and C-8 [29]. Here, the product ion at m/z 219 can easily lose two hydrogen atoms to form the product ion at m/z 217, which has a more stable conjugated structure. Thus, the molecular ion at m/z 303 was the metabolite of IMP resulting from the formation of di-oxidation. The different cleavage pathways and MS/MS spectrum of di-oxidized metabolites are illustrated in Fig. S3 described in the ESM.

M16-M21

Six major peaks (M16-M21) were observed in the extracted chromatogram from m/z 319.0687 to 319.0937 ([M+H]+ molecular ion of tri-oxidized IMP under positive mode; Fig. 6). The quasi-molecular ions of M16-M21 were all increased by 48 Da compared with the quasi-molecular ion of IMP. We determined that the tri-oxidized metabolites possessed three oxidation pathways: first, all of the oxidation reactions occurred at the side chain of IMP; second, the di-oxidation reaction occurred at the side chain of IMP, and the hydroxylation reaction occurred at the linear-type furocoumarin nucleus of IMP; third, all of the oxidation reactions occurred at the linear-type furocoumarin nucleus of IMP (three kinds of cleavage pathways and MS/MS spectrum of tri-oxidized metabolites are illustrated in Fig. S4 described in the ESM). Based on the daughter ions at m/z 203, we proposed that the compounds M16-M17 had the skeleton of hydroxypsoralen, and on the basis of the distinctive ion at m/z 219 that compounds M18-20 possessed the skeleton of dihydroxypsoralen similar to the cleavage pathway of mono-oxidized IMP (M4-M6). However, related data showed that in addition to the similar product ions of mono-oxidized IMP, the MS/MS spectra of M18-M20 possessed daughter ions at m/z 101 (C5H9O2), 83 (C5H9O2-H2O), and 55 (C5H9O2-H2O-CO) from the isopentenyl moiety of tri-oxidized IMP. Correspondingly, M16-M17 possessed daughter ions at m/z 117 (C5H9O3), 99 (C5H9O3-H2O), 81 (C5H9O3-H2O-H2O), and 53 (C5H9O3-H2O-H2O-CO). The different fragmentation behaviors of the side chains from tri-oxidized IMP further verify our above conclusion. According to the product ions at m/z 251.0535 (C16H15O7-C5H8) and 69.0699 (C5H9), the side chain of IMP was not substituted, but all of the oxidation reactions occurred at the linear-type furocoumarin nucleus of IMP. Based on the product ions at m/z 219 and 191 in the MS/MS spectrum, we deduced that M21 possessed the characteristic fragment of linear-type furocoumarins.

M22-M24

Three chromatographic peaks from m/z 289.0946 to 289.1196 were detected in the extracted ion chromatogram (Fig. 6). M24 eluted at 18.2 min with a detected molecular ion at m/z 289.1073, which was 18 Da more than the predominant quasi-molecular ion of IMP. The MS/MS spectra of M24 yielded product ions at m/z 271, 203, 175, 147, and 69. The spectra showed a cleavage pathway and MS/MS spectrum (see Fig. S5-A in the ESM) similar to that of IMP. For the metabolites M22 and M23, in addition to the characteristic ion at m/z 205, another distinctive ion at m/z 215 was present in the MS/MS spectra. The product ion at m/z 215 was followed by the successive loss of CO, yielding the typical product ions at m/z 187 and 157. All of these features allowed us to deduce that M22-M24 were the internal hydrolysis metabolites of IMP.

M25-M28

Four metabolites, M25-M28, at m/z 305 ([M+H]+; C16H17O6) were detected from 9.40 to 19.0 min (Fig. 6). The MS/MS spectrum of M25 (see Fig. S5-B in the ESM), obtained from the precursor ion at m/z 305.1020, showed eight major product ions at m/z 287.1715 (C16H17O6-H2O), 237.0402 (C16H17O6-C5H8), 219.0330 (C16H17O6-H2O-C5H8), 180.1025 (C16H17O6-C5H9-2CO), 166.0866 (C16H17O6-C5H8-2CO-CH3), and 69 (C5H9). The fragment at m/z 219 and a series of product ions at m/z 191, 173, and 163 showed characteristic product ions similar to those of mono-oxidized IMP. On the basis of the elemental compositions of the product ions and the bond connectivity present in the parent molecule, the most likely hydrogenation reaction occurred on the side chain of mono-oxidation metabolites (M4-M6).

M29-M31

M29-M31 were detected with the precursor ion at m/z 321 ([M+H]+; C16H17O7). Their protonated molecule ions mainly displayed product ions at m/z 303 ([M+H-H2O]+; C16H15O6). It was therefore deduced that the three peaks at 8.07, 8.95, and 14.0 min were the hydrogenation adding products of di-oxidized IMP (Fig. 6). The product ions at m/z 219 (C16H17O7-H2O-OC5H8) and 85 (OC5H9) suggested that the di-oxidized positions occurred both in the side chain and the linear-type furocoumarin nucleus of IMP (see Fig. S5-C in the ESM). Therefore, we concluded that M29-M31 were the hydrolysis adding products of the di-oxidized metabolites (M8-M15), but the position of hydrolysis addition could not be defined.

M32-M35

Four metabolites M32-M35 eluted at 5.25, 5.93, 7.06, and 7.63 min were detected in the extracted ion chromatogram from m/z 301.0582 to 301.0832 (Fig. 6). The quasi-molecular ion of each of them was 30 Da more than that of IMP. We determined that M32-M35 were the demethylation to carboxylic acid metabolites of IMP. The distinctive product ions at m/z 203 (C16H13O6-H2O-C5H4O) and m/z 201 and the product ions at m/z 175, 173,147, and 145 were also observed in the MS/MS spectrum. Combined with the product ions at m/z 99 (C5H7O2) and 81 (C5H5O), these ions indicated the presence of the carboxyl group (see Fig. S5-D in the ESM). The literature also reported that IMP could generate mono-oxidative metabolites (M1-M3) and then further generate the carboxylic metabolites of IMP [24].

M36-M44

The extracted chromatograph at m/z 233.0436 showed two peaks eluted at 11.4 and 13.2 min (Fig. 6). In addition, similar fragmentation behavior and characteristic ions at m/z 218, 172, 162, and 134 were present in the MS/MS spectra of M36 and M37. Based on previous data from our lab, M36 and M37 were identified as 5-methoxy-8-hydroxypsoralen and 8-methoxy-5-hydroxypsoralen [29], a pair of isomers substituted with the same methoxy and hydroxyl groups. By comparing the molecular weights and fragment ions with standards, the peaks of M38, M40, M41, M42, M43, and M44 were identified as xanthotoxol, psoralen, isopsoralen, xanthotoxin, bergapton, and isoimperatorin (Fig. 6), respectively. The peak of M39, with retention time of 10.1 min, had a quasi-molecular ion at m/z 203. The product ion spectrum that contained unchanged xanthotoxol product ions at m/z 203, 147, 131, 129, 119, 101, and 91 confirmed the isomers of xanthotoxol.

Phase II metabolites of IMP in rat urine

M45

Sulfate conjugation is one type of conjugating reactions. Phenolic compounds are able to combine with sulfate. Other compounds, such as alcohols, amines, and thiols, can also be conjugated with sulfate [26]. M45, eluted at 13.2 min (Fig. 6), showed the predominant quasimolecular ion [M+H]+ at m/z 351.0525 (C16H15O7S), 80 Da more than the parent compound of IMP, suggesting that it was a sulfated product of IMP (Fig. 8A). We determined that a tautomerized reaction took place between the ketone and the diluted alcohol.

Fig. 8
figure 8

MS/MS spectra and cleavage pathway of M45-M51 (A sulfate conjugation metabolites; B oxidation and sulfate conjugation metabolites; C oxidation and glucuronide conjugation metabolites; D di-oxidation and glucuronide conjugation metabolites, phase II metabolites)

Full size image

M46-M47

M46 and M47 eluted at 13.8 and 14.6 min (Fig. 6) and showed the same predominant quasi-molecular ions [M+H]+ at m/z 367.0426 (C16H15O8S), 80 Da more than the mono-oxidized metabolites (M4-M6), suggesting that they were sulfated products of M4-M6. The MS/MS spectrum showed an ion at m/z 298.9847, corresponding to the loss of C5H8 and a fragment ion m/z 69.0700 (C5H9), all suggesting that the side chain of IMP was not substituted. The neutral loss of m/z 79.9564 (SO3) from the ion at m/z 298.9827 led to the formation of an ion at m/z 218.0205, indicating that it was a sulfated product of M4-M6. Due to the identical skeleton to phase I metabolites, similar fragment ions at m/z 191.0300 and 173.0232 were also expected for the phase II metabolites. The phase II MS/MS spectra are shown in Fig. 8B.

M48-M49

M48 and M49 eluted at 10.0 and 10.4 min and showed the predominant quasi-molecular ion [M+H]+ at m/z 463.1223 and 463.1221 (Fig. 6). The product ion at m/z 287.0902 (C16H15O5), the loss of a glucose (Glu), is one of the major fragments that was 16 Da more than that of IMP. The diagnostic fragment ions at m/z 219.0287 and 69.0629 indicated that the side chain of IMP was not substituted (Fig. 8C). Thus, M48 and M49 could be the glucose conjugated product of M4-M6, and the glucose conjugation reaction occurred on the hydroxyl group of the linear-type furocoumarin ring of M4-M6.

M50-M51

M50 and M51 were detected with the precursor ion at m/z 479.2484 ([M+H]+; C22H23O12). Their protonated molecule ions mainly displayed product ions at m/z 303.0835 ([M+H-C6H8O6]+; C16H15O6). It was therefore determined that the two peaks at 8.15 and 5.60 min were the glucose conjugated products of the di-oxidized IMP (Fig. 6). The MS/MS spectra (Fig. 8D) obtained from M51 showed the major product ion at m/z 203.0327, indicating the glucose conjugation reaction occurred on the side chain of IMP. Likewise, M50 showed the major product ion at m/z 219.0273, in agreement with M48 and M49 with the same conjugation reaction position.

Fragmentation rules deduced from IMP metabolites

To provide sufficient and useful structural information for the analysis of other linear-type furocoumarins metabolites by using MS, the fragmentation behaviors of the IMP metabolites were determined. The postulated fragmentation pathways of IMP metabolites are shown in Fig. 9.

Fig. 9
figure 9

Fragmentation pathway of IMP metabolites according to MS/MS spectra by Triple-TOF MS

Full size image
  1. 1.

    All of the IMP metabolite fragmentation pathways are consistent with the linear-type furocoumarins’ distinctive fragmentation pattern: the successive loss of CO [27]. Moreover, both the nature of the substituent and the position have an effect on the fragmentation patterns of IMP metabolites.

  2. 2.

    These metabolites are classified into two groups based on the different linear-type furocoumarin nucleus ions: (1) if the distinctive fragment ion at m/z 203 is present, the skeleton of IMP is not substituted. (2) The presence of the characteristic fragment ions at m/z 219, and a series of fragment ions at 201, 191, 173, 163, and 145 generated from it, indicates that the skeleton of IMP is substituted.

  3. 3.

    The appearance and relative abundances of [M+H-C5H8]+ and C5H9 + ions also play a significant role in the differentiation of the oxidation position. If the two fragment ions can be found at the same time, then the side chain of IMP is not changed (still isopentenyl group).

  4. 4.

    The characteristic ions via functional group cleavage reactions and the product ions of [M+H-SO3]+ and [M+H-C6H8O6]+ provide valuable evidence for the phase II metabolite identification of IMP.

Metabolite pathway of IMP

The proposed major metabolic pathway of IMP in rat urine is shown in Fig. S6 (described in the ESM). These pathways indicated that the side chain and the skeleton of IMP were the metabolic sites in vivo, which were accordance with the previous results [23]. Additionally, IMP mainly underwent oxidation metabolism, i.e., mono-oxidation, di-oxidation, tri-oxidation, and hydroxylation. Mono-oxidation was the primary metabolic action, followed by further transformation through di-oxidation, tri-oxidation, and internal hydrolysis reaction. Conjugation with sulfate and glucuronidase groups was the predominant phase II metabolite reaction. The IMP metabolites with xenobiotic molecules had altered pKa values [26], and after metabolic reaction, the polarity of the IMP metabolites became stronger than that of the parent drug. The increase in polarity allowed the metabolites to be more easily excreted from the body. Comparing with the metabolism of IMP in vitro [24, 25], it was found that hydroxylation and hydrolysis reactions in the prenyloxy side chain could occurred both in vivo and in vitro, but lactone ring-opening reaction of furocoumarin skeleton was not involved in vivo.

Conclusions

The use of MMDF-dependent on-line data acquisition, combined with multiple post-data mining acquisition, was a powerful integrated approach for fast and reliable screening and identification of 51 metabolites of IMP in rat urine. A Full MS-IDA-4MS/MS scan was used for high-throughput acquisition of potential metabolites. The metabolite fragmentation behavior and the metabolic pathway were also summarized. This is the first report of the identification of metabolites of IMP. The results lead to a better understanding of the bio-transformations and pharmaceutical applications of IMP. Furthermore, the application of this approach to the screening and identification of metabolites provided essential data for further pharmacological studies of other linear-type furocoumarins.