Identification of Oxygenated Fatty Acid as a Side Chain of Lipo-Alkaloids in Aconitum carmichaelii by UHPLC-Q-TOF-MS and a Database

Lipo-alkaloid is a kind of C19-norditerpenoid alkaloid usually found in Aconitum species. Structurally, they contain an aconitane skeleton and one or two fatty acid moieties of 3–25 carbon chains with 1–6 unsaturated degrees. Analysis of the lipo-alkaloids in roots of Aconitum carmichaelii resulted in the isolation of six known pure lipo-alkaloids (A1–A6) and a lipo-alkaloid mixture (A7). The mixture shared the same aconitane skeleton of 14-benzoylmesaconine, but their side chains were determined to be 9-hydroxy-octadecadienoic acid, 13-hydroxy-octadecadienoic acid and 10-hydroxy-octadecadienoic acid, respectively, by MS/MS analysis after alkaline hydrolysis. To our knowledge, this is the first time of the reporting of the oxygenated fatty acids as the side chains in naturally-occurring lipo-alkaloids. In order to identify more lipo-alkaloids, a compound database was established based on various combinations between the aconitane skeleton and the fatty acid chain, and then, the identification of lipo-alkaloids was conducted using the database, UHPLC-Q-TOF-MS and MS/MS. Finally, 148 lipo-alkaloids were identified from A. carmichaelii after intensive MS/MS analysis, including 93 potential new compounds and 38 compounds with oxygenated fatty acid moieties.


Introduction
Lipo-alkaloid is a kind of C19-norditerpenoid alkaloid usually found in Aconitum species. Structurally, they consist of an aconitane skeleton and one or two fatty acid moieties of 3-25 carbon chains with 1-6 unsaturated degrees [1]. So far, more than 200 lipo-alkaloids have been reported from plants [1], semisynthesis [2] and biotransformations [3][4][5]. Because naturally-occurring lipo-alkaloids are so close structurally, it is very difficult to purify them from the mixture. Their characterization is mainly conducted by high sensitive mass spectrometry (MS) or liquid chromatography-mass spectrometry (LC-MS) techniques, such as ion trap-MS [6], MALDI-TOF-MS [7], ESI-Fourier Transform Ion Cyclotron Resonance FTICR-MS [8], LC-ESI-Ion TrapIT-MS [8][9][10], and LC-Linear Ion TrapLTQ-Orbitrap-MS [11]. To our knowledge, only one pure lipo-alkaloid, 8-O-azeloyl-14-benzoylaconine, was isolated and elucidated by NMR from plants [12]. In this research, the separation of naturally-occurring lipo-alkaloids was optimized using column chromatography (CC), and six pure lipo-alkaloids (A1-A6) were obtained from a natural source for the first time. At the same time, a mixture of oxygenated fatty acid-containing lipo-alkaloids (A7) was also obtained, and the structures were determined by MS/MS analysis after alkaline hydrolysis. Furthermore, a compound database of lipo-alkaloids was established based on the basic skeletons and fatty acids groups, and by the combination of the compound database, UHPLC-Q-TOF-MS and MS/MS analysis, 148 lipo-alkaloids, including 93 potential new ones, were identified from A. carmichaelii.

Isolation and Structural Elucidation of Lipo-Alkaloids A1-A7
Structurally, lipo-alkaloids usually contain one or two long fatty acid moieties; therefore, they have low polarity and can be extracted by n-hexane. After column chromatography separation on silica gel and ODS, a lipo-alkaloids-rich fraction was obtained. Compared to water and methanol as mobile phases, 0.01% diethylamine in water and methanol gave better separation of lipo-alkaloids on a preparative HPLC C18 column ( Figure 1A). Seven compounds (A1-A7) were determined to be lipo-alkaloids by UHPLC-Q-TOF-MS and NMR techniques.
The molecular formulae of Compounds A3 and A4 were C 47 H 73 NO 11 and C 49 H 75 NO 11 , which were two carbons less and two hydrogens more than that of Compound A1. The same MS/MS spectra (Table S2) (Table S1) suggested that the side chains for A3 and A4 were palmitic acid and oleic acid, respectively. Therefore, their structures were determined to be 8-O-palmitoyl-14-benzoylmesaconine (A3) and 8-O-oleoyl-14-benzoylmesaconine (A4).
The molecular formula of Peak A2 was determined as C 50 H 75 NO 11 from the high resolution MS, which was one carbon and two hydrogens more than that of A1. The pattern of fragmentation ions of A2 was very similar to that of A1 with most of the peaks moved 14 Da (CH 2 ) to the right side ( Figure 1B and Table 1). The 1 H-NMR spectrum of A2 showed the presence of NCH 2 CH 3 (Table S1); therefore, the basic skeleton should be 14-benzoylaconine (BA), and it was determined to be 8-O-linoleoyl-14-benzoylaconine. With a similar comparison, Compounds A5 (C 48 H 75 NO 11 ) and A6 (C 50 H 77 NO 11 ) were determined to be 8-O-palmitoyl-14-benzoylaconine and 8-O-oleoyl-14-benzoylaconine, respectively, based on their MS/MS and 1 H-NMR spectra.   Peak A7 was obtained as a single peak in the UHPLC-MS chromatogram and with a quasi-molecular ion at m/z 868.5228, which is in agreement with the molecular formula of C 49 H 73 NO 12 that is one oxygen more than that of A1. The MS/MS spectrum showed that A7 was a BMA derivative with C 18 H 32 O 3 as the side chain ( Figure 1C). However, the 1 H-NMR spectrum of A7 was unexpectedly complicated. Except for BMA signals, the unconjugated vinyl protons at δ 5.371 (4H, m), bis-allylic methylene at δ 2.790, allylic methylene at δ 2.051 and methylene at around δ 1.3, there were three proton signals at δ 4.413, 4.189 and 4.127 with unproportioned integrations to other signals. Therefore, A7 might be a mixture containing different side chains. After alkaline hydrolysis of A7, the released fatty acids were analyzed by UHPLC-MS and MS/MS, and the results are shown in Figure 2 indicate the presence of a hydroxyl group at C-13 [14] and the unconjugated vinyl bonds should be between C-2 and C-12, i.e., 13-hydroxyoctadecadienoic acid. The fragmentation ions in the MS/MS spectrum for the third peak at 5.5 min were related to the neutral losses of H 2 O, C 8 H 16 and CO, just like the MS/MS spectrum of 10-hydroxyoctadecadienoic acid [15]. The fourth peak was identified as 9-hydroxyoctadecadienoic acid based on the ions of [M´H´H 2 O]´and [M´H´C 9 H 16 ]´ [14]. Although these compounds were identified by LC-MS and/or semi-synthetic methods before, this is the first time that they have been purified from nature. Besides, oxygenated fatty acid-containing lipo-alkaloids were firstly obtained, and they shared the same aconitane skeleton of 14-benzoylmesaconine, while their side chains were determined to be 9-hydroxy-octadecadienoic acid, 13-hydroxy-octadecadienoic acid and 10-hydroxy-octadecadienoic acid, respectively. Their structures are shown in Figure 3.

Establishment of Lipo-Alkaloids Database
In the structures of the previously-reported naturally-occurring lipo-alkaloids [1,11], there were 13 basic aconitine skeletons and 51 fatty acid chains. Oxygenated fatty acids are widely present in plants; thus, it is possible that these fatty acids connect with aconitane skeletons to form the lipo-alkaloids, as reported above. All possible lipo-alkaloids can be hypothesized by the following formula: Herein, MF lipo-alkaloid and MF basic skeleton are the molecular formulae of potential lipo-alkaloids and the 13 reported aconitane skeletons, while MF fatty acid is the molecular formula of possible fatty acids, which have 3-25 carbons and 2-6 oxygen atoms with the unsaturated degrees from 1-7. Then, the names and molecular formulae of the hypothesized lipo-alkaloids were input into an Agilent MassHunter database file to establish an in-house lipo-alkaloids database with a total of 484 molecular formulae.  [14]. For the first peak, the fragmentation ions from the neutral losses of H2O and C5H12O were observed. The cleavage of hydroxyl group usually gives an unsaturated carbonyl group, which is different from the neutral loss of C5H12O, so the fatty acid at 4.4 min needs further investigation. Finally, Peak A7 was determined as the mixture of the oxygenated fatty acids-containing BMA derivative by the combination of 1 H-NMR and MS/MS analysis after alkaline hydrolysis.  As a result, six pure lipo-alkaloids were obtained by optimized column chromatography and characterized as 8 14-benzoylaconine (A5) and 8-O-oleoyl-14-benzoylaconine (A6), respectively. Although these compounds were identified by LC-MS and/or semi-synthetic methods before, this is the first time that they have been purified from nature. Besides, oxygenated fatty acid-containing lipo-alkaloids were firstly obtained, and they shared the same aconitane skeleton of 14-benzoylmesaconine, while their side chains were determined to be 9-hydroxy-octadecadienoic acid, 13-hydroxy-octadecadienoic acid and 10-hydroxy-octadecadienoic acid, respectively. Their structures are shown in Figure 3.

Establishment of Lipo-Alkaloids Database
In the structures of the previously-reported naturally-occurring lipo-alkaloids [1,11], there were 13 basic aconitine skeletons and 51 fatty acid chains. Oxygenated fatty acids are widely present in plants; thus, it is possible that these fatty acids connect with aconitane skeletons to form the lipo-alkaloids, as reported above. All possible lipo-alkaloids can be hypothesized by the following formula:
Compounds 105, 120, 131 and 137 shared the same fragment ions at m/z 524.30, 492.27, 464.28, 460.25 and 432.25 ( Figure 6H). The fragmentation patterns were very similar to those compounds with 3-DMDBA as the basic skeleton ( Figure 6E), but with one oxygen less. Because there was no ion produced from the loss of FA + methanol + benzoic acid, the absence of 13-OH was indicated; therefore, the basic skeleton should be demethoxy-3,13-dideoxy-14-benzoylaconine (DMDDBA).    Figure 6E). Except for the ion at m/z 354.21, all other ions contained one OCH 2 group less than that of 3-deoxy-14-benzoylaconine (3-DBA) derivatives ( Figure 6C); therefore the basic skeleton was determined to be demethoxy-3-deoxy-14-benzoylaconine (3-DMDBA). The ion at m/z 354.21 was derived from the loss of two methanols rather than three methanols, which in turn further confirmed that the basic skeleton had one methoxyl group less than 3-DBA with 1-OCH 3 or 6-OCH 3 missing (Scheme S4).
The isomers of Compounds 103, 117 and 116 were observed at the retention times of 24. Based on the finding of oxygenated fatty acids as the side chains of lipo-alkaloids, the possible lipo-alkaloids were predicted and included in an in-house database. By the combination of the database, UHPLC-MS and MS/MS analysis, not only more oxygenated fatty acid-containing lipo-alkaloids were determined, but also four aconitane skeletons not reported in lipo-alkaloids before were detected. Finally, 148 lipo-alkaloids, including 93 potential new ones, were identified ( Table 2). Although most of previous reports showed that the contents of lipo-alkaloids usually increased after processing, no significant difference was detected when using heat reflux extraction or ultrasonic extraction in our preliminary research (data not shown).

MS/MS Characterizations of Aconitane Skeletons in Lipo-Alkaloids
In this study, we reported 13 aconitane skeletons (including four new ones) in the lipo-alkaloids with their main fragmentation ions from the neutral losses of MeOH, H2O, CO and BzOH ( Figure 6 and Table 1). Based on structures and MS/MS spectra, the relationship between the substitutions and the fragmentation ions can be summarized as follows. (1) The ions produced from the neutral loss of MeOH have higher abundance, and the numbers of methoxy group substituted on aconitane skeleton usually are determined from the corresponding ions. For instance, ions with the loss of three  Figure 6G and Scheme S3). The fragmentation patterns were very similar to those of 3-DMDBA derivatives ( Figure 6E), but with one CH 2 less. The ions corresponding to the loss of two molecules of methanol at m/z 462.23 ([M + H´FA´2CH 3 OH] + ) and 340.19 ([M + H´FA´2CH 3 OH´benzoic acid] + ) indicated that the differences were the substitution groups on the N atom, and it should be N-CH 3 rather than N-C 2 H 5 in these three compounds, i.e., the basic skeleton should be demethoxy-14-benzoylhypaconine (DMBHA).
Compounds 105, 120, 131 and 137 shared the same fragment ions at m/z 524.30, 492.27, 464.28, 460.25 and 432.25 ( Figure 6H). The fragmentation patterns were very similar to those compounds with 3-DMDBA as the basic skeleton ( Figure 6E), but with one oxygen less. Because there was no ion produced from the loss of FA + methanol + benzoic acid, the absence of 13-OH was indicated; therefore, the basic skeleton should be demethoxy-3,13-dideoxy-14-benzoylaconine (DMDDBA).
Based on the finding of oxygenated fatty acids as the side chains of lipo-alkaloids, the possible lipo-alkaloids were predicted and included in an in-house database. By the combination of the database, UHPLC-MS and MS/MS analysis, not only more oxygenated fatty acid-containing lipo-alkaloids were determined, but also four aconitane skeletons not reported in lipo-alkaloids before were detected. Finally, 148 lipo-alkaloids, including 93 potential new ones, were identified (Table 2). Although most of previous reports showed that the contents of lipo-alkaloids usually increased after processing, no significant difference was detected when using heat reflux extraction or ultrasonic extraction in our preliminary research (data not shown).

MS/MS Characterizations of Aconitane Skeletons in Lipo-Alkaloids
In this study, we reported 13 aconitane skeletons (including four new ones) in the lipo-alkaloids with their main fragmentation ions from the neutral losses of MeOH, H 2 O, CO and BzOH ( Figure 6 and Table 1). Based on structures and MS/MS spectra, the relationship between the substitutions and the fragmentation ions can be summarized as follows. (1) The ions produced from the neutral loss of MeOH have higher abundance, and the numbers of methoxy group substituted on aconitane skeleton usually are determined from the corresponding ions. For instance, ions with the loss of three molecules of methanol were detected for the aconitine skeletons with tetramethoxy substitution, while ions with neutral loss of two molecules of methanol were observed for the trimethoxy-substituted skeletons. Due to the higher bond energy between C 18 and the methoxy group [22], it is difficult to detect the fragment ions from the loss of C 18

Fatty Acid Side Chains in Lipo-Alkaloids
Besides common long chain fatty acids, medium and long chain oxidized fatty acids were detected as the side chains of lipo-alkaloids in plants for the first time, e.g., C 9 H 16 (Table 3). These oxygenated fatty acids might occur as hydroxyl-, oxo-, epoxy-, hydroperoxy-type or diacid [23]. However, due to the limitation of LC-MS data, it is difficult to determine in which form they exist in the lipo-alkaloids. In this study, three oxygenated fatty acids in a lipo-alkaloids mixture were determined by 1 H-NMR, alkaline hydrolysis and MS/MS analysis, but other oxygenated fatty acid groups could not be determined due to the limited amount of sample available. Considering the polarity and occurrence of fatty acids in nature, the most possible structures were proposed in Table 2 by searching the lipid maps [24] and comparing retention times of the lipo-alkaloids to common fatty acid side chains.
Plant oxylipins are involved in the stress responses, and some of them have anti-microbial and anti-insecticidal activities [25]. Some oxylipins, e.g., 2-hydroxyoleic acid (C 18 H 34 O 3 ), were found to have anti-cancer activity [26], while some oxylipins have anti-inflammatory activity [27]. When these oxidized fatty acids connect to aconitane alkaloids to form the lipo-alkaloids, the bioactivity and toxicity of aconitane alkaloids might change. Thus, the occurrence, bioactivity and toxicity of these oxygenated fatty acid-containing lipo-alkaloids are worth further investigations.

Plant Materials
The roots of Aconitum carmichaelii Debx. (ChW-02) were obtained from Hehuachi Medicinal Materials Market in Chengdu, Sichuan Province of China, and authenticated by Ying Liu, Chengdu University. A voucher specimen was deposited in Macau University of Science and Technology.

Separation of Lipo-Alkaloids
The air-dried roots of A. carmichaelii (7.2 kg) were powdered and soaked in methanol (12 L) at room temperature for one week and then extracted with methanol at reflux 3 times (3ˆ12 L, 1 h for each extraction). The combined methanol extracts were evaporated under vacuum to give 356 g of residue, which was suspended in distilled water (3 L) followed by the participation with n-hexane (3ˆ3 L), ethyl acetate (3ˆ3 L) and n-butanol (3ˆ3 L), successively. The n-hexane extract (32 g) was subjected to silica gel CC (6ˆ60 cm) using CH 2 Cl 2 /CH 3 OH as the eluate to provide four fractions (A-D). Fraction D (5 g) was further subjected to ODS CC (4.5ˆ50 cm) using water-containing methanol (0%-100%) as the eluate to produce five subfractions (D1-5). Subfraction D4 (1 g) was divided into 10 parts by another ODS CC with an increasing gradient of water-containing methanol (40%-100%). Preparative HPLC separation of the eighth part (D4-8, 134 mg) on an ODS column (10ˆ250 mm, 5 µm) produced 7 compounds, 8 mg A1, 6 mg A2, 3 mg A3, 9 mg A4, 7 mg A5, 3 mg A6 and 4 mg A7. The mobile phases were 0.01% diethylamine-containing water (A) and methanol (B) with the following gradient: 0-40 min, 70%-95 B%; 40-120 min, 95% B. The flow rate was 2 mL/min, and the detection wavelength was set at 230 nm. The structures were characterized by NMR and mass spectrometry.

Alkaline Hydrolysis of Peak A7
One milligram of A7 was dissolved in 400 µL of KOH-saturated methanol solution and then heated to 75˝C for 15 min and 60 min. The reaction solution was neutralized with 800 µL of 5 M HCl-MeOH and participated with ethyl acetate, respectively. The ethyl acetate layer was analyzed by UHPLC-Q-TOF-MS.

Preparation of Methanol Extracts of Herbal Sample
One gram of powdered herbal sample was extracted with 6 mL methanol for 60 min with the aid of an ultrasonicator and then centrifuged at 13,000 rpm for 10 min. The supernatant was collected and diluted 10-times, then followed by the acquisition of UHPLC-Q-TOF-MS data.

UHPLC-Q-TOF-MS Analysis
Agilent 1290 UHPLC system (UHPLC, Agilent Technologies, Santa Clara, CA, USA) consisting of an autosampler, thermostated column compartment and binary pump and equipped with an Agilent Eclipse C18 column (2.1ˆ100 mm, 1.8 µm, Agilent Technologies) was applied for the separation of components. The mobile phases were 0.1% formic acid in water (A) and 0.1% formic acid in acetonitrile (B). Method 1 was applied for the determination of lipo-alkaloids, and the mobile phase gradient was set as follows, 0-0.5 min, 20% B; 0.5-30 min, 20%-98% B; 30-33 min 98% B; 33-33.1 min, 98%-20% B, and then maintained for 2 min. Method 2 was used for the analysis of fatty acids, and the gradient was 0-11 min, 25% B, 11-11.1 min, 25%-95% B, and then maintained for 2 min. The flow rate was 0.3 mL/min, and the injection volume was 2 µL. The mass spectrometry was conducted on a 6550 UHD Accurate-Mass Q-TOF/MS system (Agilent Technologies) with a dual Agilent Jet Stream electrospray ion source (dual AJS ESI). The mass parameters were optimized using the standards of aconitine, mesaconitine and hypaconitine and set as follows: dry gas temperature and flow were 250˝C and 15 L/min; sheath gas temperature and flow were 300˝C and 11 L/min; nebulizer at 20 psi; the capillary and nozzle voltages were 4000 and 500 V, respectively. The fragmentor was 380 V, and the collision cell energies were set at 50 eV for lipo-alkaloids in positive mode and 30 eV for fatty acids in negative mode, respectively.

Establishment of the Lipo-Alkaloids Database
Based on the possible fatty acid chains and known aconitane skeletons reported in Aconitum plants, the possible lipo-alkaloids were hypothesized and input into Agilent MassHunter database file ("Compound Formula Database") to establish an in-house lipo-alkaloids database. Then, the potential lipo-alkaloids in A. carmichaelii were extracted using the function of "Find Compounds by Formula (FBF)" and determined by MS/MS analysis.

Conclusions
In this study, the separation method of lipo-alkaloids was optimized, and using this method, oxygenated fatty acids-containing lipo-alkaloids were obtained for the first time. A lipo-alkaloids database was established based on the known basic aconitane skeletons and possible fatty acid side chains. By using the database, potential lipo-alkaloids were first extracted from UHPLC-Q-TOF-MS, and then, the structures were determined from the comprehensive analysis and deduction of MS/MS spectra, resulting in successful identification of 148 lipo-alkaloids. Among them, 38 compounds contain medium or long chain oxidized fatty acids as side chains that were not reported previously. The combination of database and LC-MS dramatically speeds up the finding of potential new compounds and is confirmed to be a powerful tool in the study of natural product chemistry. The new finding of oxygenated fatty acids as side chains of lipo-alkaloids provides a kind of possible structures, which accounts for the bioactivities of A. carmichaelii, a widely-used traditional medicine.