Rapid Screening of Forskolin-Type Diterpenoids of Blumea aromatica DC Using Ultra-High-Performance Liquid Chromatography Tandem Quadrupole Time-Of-Flight Mass Spectrometry Based on the Mass Defect Filtering Approach

The discovery of new active compounds of natural products tends to be increasingly more challenging due to chemical complexity and unpredictable matrices. Forskolin is an active natural labdane-type diterpenoid ingredient widely used worldwide for the treatment of glaucoma, heart failure, hypertension, diabetes, and asthma, and is expected to be a promising anticancer, anti-inflammation, and anti-HIV agent. In recent years, demand for forskolin in the medicine market has increased dramatically. However, natural forskolin originates exclusively from traditional Indian herb medicine Coleus forskohlii (Willd.) Briq. In a previous study, we isolated a series of diterpenoids including an 8,13-epoxy-14ene labdane carbon skeleton from Blumea aromatica DC. In order to identify alternative plant resources, a novel and effective strategy was proposed for the screening of potential forskolin-type diterpenoids (FSKD) compounds obtained from B. aromatica, using the mass defect filtering (MDF) strategy via ultra-high-performance liquid chromatography tandem quadrupole time-of-flight mass spectrometry (UHPLC-QTOF/MS) approach. Within a narrow, well-defined mass defect range, the strategy developed could significantly improve the detection efficiency of selected FSKD compounds by filtering out certain major or moderate interference compounds. Additionally, the MS/MS cleavage behavior and the characteristic diagnostic ions of the FSKD compounds were proposed to be used in aiding structural identification of the filtration compounds. As a result, a total of 38 FSKD of B. aromatica were filtered out and tentatively identified. To the best of our knowledge, it was the first time that these forskolin-type diterpenoids were identified in B. aromatica, which significantly expands our understanding of the chemical constituents of Blumea species, and allows B. aromatica to be used as a potential alternative plant resource that contains these forskolin-type active compounds. The strategy proposed was proven efficient and reliable for the discovery of novel compounds of herbal extracts.


Introduction
It is well known that natural products, particularly those with novel chemical structures, are recognized as playing an important role in the discovery of drugs that are beneficial for human health. However, it is still a great challenge to explore active constituents of natural sources due to the In order to find more FSKDs and fully profile the FSKDs in B. aromatica, identify other alternative plant resources containing forskolin active ingredients, and understand the effective material basis and pharmacological activity of crude herbs, it is urgently required to develop a sensitive and reliable method to characterize the FSKD in B. aromatica. In this study, we describe an effective and facile approach for the overall capture of the FSKDs in B. aromatica based on UHPLC-QTOF/MS coupled with the MDF approach. For the first time, a total of 38 FSKDs were filtered out and tentatively identified from B. aromatica. It is expected that the Method and Results of the present study would expand our understanding of the chemical constituents of B. aromatica, and provide a helpful strategy to identify natural plant resources containing FSKD active components, as well as to be used as an effective tool to screen for chemical targets for future phytochemical and pharmacological activity studies on these important herbal medicines.

Results
According to the proposed screening strategy below, first, the MDF filtration method was established based on the FSKDs constituents which have been reported in Coleus forskohlii [21,22]. Secondly, the extract of B. aromatica sample was analyzed by UHPLC-QTOF-MS/MS at both positive and negative modes. Then, the developed MDF method was performed to pick out the potential precursor ions of the FSKDs. Finally, the structures of the screened compounds were further elucidated based on MS/MS spectra and fragmentation pathways were summarized using the reference substance (see Figure 1). a few detailed reports are available on the chemical properties of B. aromatica [20]. In our previously study, a total of five diterpenoid compounds were isolated from B. aromatica in our lab. It was notable that the structures of the diterpenoid isolated from B. aromatica shared a core skeleton which was similar to that of forskolin. As far as we know, this was the first time that these forskolin-type diterpenoid (FSKD) compounds were found in B. aromatica.
In order to find more FSKDs and fully profile the FSKDs in B. aromatica, identify other alternative plant resources containing forskolin active ingredients, and understand the effective material basis and pharmacological activity of crude herbs, it is urgently required to develop a sensitive and reliable method to characterize the FSKD in B. aromatica. In this study, we describe an effective and facile approach for the overall capture of the FSKDs in B. aromatica based on UHPLC-QTOF/MS coupled with the MDF approach. For the first time, a total of 38 FSKDs were filtered out and tentatively identified from B. aromatica. It is expected that the Method and Results of the present study would expand our understanding of the chemical constituents of B. aromatica, and provide a helpful strategy to identify natural plant resources containing FSKD active components, as well as to be used as an effective tool to screen for chemical targets for future phytochemical and pharmacological activity studies on these important herbal medicines.

Results
According to the proposed screening strategy below, first, the MDF filtration method was established based on the FSKDs constituents which have been reported in Coleus forskohlii [21,22]. Secondly, the extract of B. aromatica sample was analyzed by UHPLC-QTOF-MS/MS at both positive and negative modes. Then, the developed MDF method was performed to pick out the potential precursor ions of the FSKDs. Finally, the structures of the screened compounds were further elucidated based on MS/MS spectra and fragmentation pathways were summarized using the reference substance (see Figure 1).
The base peak ion (BPI) chromatogram of the extraction of B. aromatica before and after MDF filtration is shown in Figure 4. In this study, a total of 38 FSKDs of the B. aromatica sample were selected and characterized (Table 1).  The base peak ion (BPI) chromatogram of the extraction of B. aromatica before and after MDF filtration is shown in Figure 4. In this study, a total of 38 FSKDs of the B. aromatica sample were selected and characterized (Table 1).

Optimization of Sample Preparation and UHPLC-QTOF-MS Analysis Conditions
The extract parameters were optimized in order to obtain a higher extraction efficiency, including the following: the type of extraction method (ultrasonic and refluxing); extraction solvent (different methanol/water ratio solvent of 70:30, 80:20, and 85:15, v/v); and extraction time (20,30,45, and 60 min). The chromatograms of B. aromatica extract with different extraction conditions were available in Supplementary Materials (online at www.mdpi.com/1420-3049/24/17/3073/s1). The results were evaluated by the number of detected peaks after UPLC-QTOF-MS analysis. In the end, ultrasonic extraction conducted for 45 min with 80% methanol (v/v) as a solvent was selected as the optimal extraction conditions for crude sample preparation.
The UHPLC-Q-TOF-MS conditions, such as the column type, the composition of mobile phase, ionization mode, capillary voltage, and collision energy, were also evaluated to get a shorter analytical time and satisfactory separation effect. The mobile phase additives may have a significant influence on signal response. It was found that the aqueous mobile phase containing formic acid (0.1% v/v) could improve the peak shape and give higher ionization efficiency, especially that of trace peaks. Thus, a gradient elution system composed of acetonitrile-0.1% aqueous formic acid was chosen as the eluting solvent system to give the optimal chromatographic peak shapes and resolution. In order to shorten analytical time and obtain satisfactory separation, two RP-C 18 columns, Waters ACQUITY UPLC BEH C18 (100 × 2.1 mm, 1.7 µm) and ACQUITY UPLC HSS T3 (100 × 2.1 mm, 1.8 µm), flow rate (0.3 mL/min, 0.4 mL/min, 0.5 mL/min, and 0.6 mL/min) and column temperature at 30 • C, 35 • C, and 40 • C were evaluated. Finally, ACQUITY UPLC HSS T3 column (100 × 2.1 mm, 1.8 µm) with the temperature at 35 • C and flow rate of 0.5 mL/min were chosen to obtain optimal chromatographic results. MS spectra were scanned at both positive and negative modes. The positive scanning mode was selected due to its higher ionization efficiency, and higher abundance fragment ions, even at low collision energy. Meanwhile, the precursor ions prone to form the [M + HCOO] − quasimolecular ions were scanned under a negative scanning mode, which helped to confirm the molecular formula. The collision energy was optimized in the range of 20-60 V to obtain better MS/MS spectra results for the FSKD. In general, low collision energy is applied to ions with a low molecular weight to produce satisfactory fragment ions. In contrast, large molecules require high collision energy to obtain more fragment ions.

Establishment of the MDF Filtration Method
The MDF strategy has been applied to greatly improve the efficiency of the identification of bioactive constituents in TCM [23][24][25]. The components share the same core structure but have different substituent groups, which result in relatively narrow and defined mass defect shifts. Hence, in order to follow the MDF strategy, it was essential to define the mass defect value based on the combination of core structure and substituents.
On the basis of the concept of MDF described above, the first step was to ensure filtering reference skeleton based on published forskolin analogues [19,21,22]. The next step was to establish the mass defect range based on the core structure and substituent group combination. Then, filtration was applied to the chromatogram based on the setting expressed as central formula ± mass defect tolerance. Finally, background ions were automatically removed, and characteristic ions remained visible after screening. In this study, the filtering parameters were established using the "accurate mass filter" function in MassLynx 4.1 software (Waters Corp., Milford, MA, USA). On the basis of the information published on these compounds, the filter reference was determined as 8, 13-epoxy-14 ene labdane diterpenes (C 20 H 34 O) and is illustrated in Figure 2. The substituents of the core structure mainly included carbonyl-function at the C-11 position, hydroxylation at C-1, C-6, C-7, and C-12 positions, acetylation or deacetylation, and other oxidized derivatives. The mass defect values were different between various substituents, with both carbonylation and acetylation increasing the mass defect value while hydroxylation decreased the mass defect value. On the basis of extract mass summaries of published forskolin analogues and mass defect value of the various substituents, the exact mass value was set on the X-axis and the mass defect value on the Y-axis. The mass defect distribution diagram obtained is shown in Figure 3. It is interesting to note that all the components were found to be clustered in an elliptical range. Thus, the calculated minimum and maximum mass defect values of the labdane-type diterpenoids were found to be in the range of 0.2171-0.2788 Da, while the mass ranged from 280-460 Da. Therefore, the filtering template used for FSKD screening was set as follows: mass range 280-460 Da, MDF value set at 0.25 Da, with an upregulation of 30 mDa and downregulation of 40 mDa. The mass error of the predicted chemical formula should be less than 10.00 ppm, while the element composition was set at C (18)(19)(20)(21)(22)(23)(24)(25), H (28-40), and O (2-7).

FSKD Analogue Identification After Filtration Using the MDF Approach
After filtration using the optimized MDF approach, all interfering ions with mass defects that were significantly not within the filter window were excluded. The base peak intensity (BPI) chromatograms before and after filtration are shown in Figure 4. The noise level of the filtered chromatogram was much lower than that of the chromatogram without MDF filtration. As a result, a total of 38 peaks of the potential FSKD were filtered out within 28 min of chromatographic running time (Table 1). Similar types of chemical structures always possess similar MS/MS fragmentation behaviors and lead to common characteristic ions [8,9,23]. The fragment ions of each filtered peak are listed on Table 1. The accurate mass and element composition of the precursor ions and fragment ions were calculated using the tool of "elemental composition" of MassLynx V 4.1 software. The mass error of the predicted chemical formula of less than 10.00 ppm were selected as candidate compounds. Meanwhile, MassFragmentTM software was used to assist with the elucidation of chemical structure based on a scoring system. The structural identification of the FSKDs were further confirmed using reference compounds, published literature, the MS/MS fragmentation pathway, key product ions filtration, SciFinder, ChemSpider, and other online databases. All related data of the identified components are summarized in Table 1. The cleavage pathways of FSKDs of B. aromatica are proposed based on MS/MS data of the reference compounds. In previous research, diterpene compounds with a similar skeleton were found in B. aromatica, of which the structure belongs to 8,13-epoxy-14 ene labdane type diterpenes, and it is similar to the structure of forskolin. Until now, five FSKDs have been isolated from B. aromatica by our group. Three of them, acromatin A, acromatin B and acromatin C, together with another two commercial standard compounds, forskolin and isoforskolin, were employed as template structures to study the fragmentation patterns of the FSKDs, which can subsequently enhance the comprehensive identification of other FSKDs in B. aromatica with the same core skeleton.

Structural Elucidation of Acromatin A, Acromatin B, and Acromatin C
On the basis of previous studies, compounds acromatin A, acromatin B, and acromatin C extracted from B. aromatica were isolated and characterized by our team and were identified using UV, IR, HRMS, and NMR. Their chemical structures are shown in Figure 2. The structures of acromatin A, acromatin B, and acromatin C were found to share the same core skeleton of 8,13-epoxy-14 ene labdane diterpenes, with only differences in chemical configurations of the substituent group at the C-4 position, for which the substitutions groups were carboxyl, aldehyde, and methylene hydroxyl for acromatin A, acromatin B, and acromatin C, respectively. Unlike forskolin, no carbonyl group was substituted at the C-11 position and no acetyl group substitution at the C-6 and C-7 positions was found in these FSKDs extracted from B. aromatica. Their main fragmentation pathway was the cleavage of COOH, CHO, and CH 2 OH at C-4 and the neutral losses of H 2 O groups, along with the cleavage of the C 9 -C 11  under the negative scanning mode, however, it is difficult to break these down to obtain fragment ions even under high collision energy. In contrast, the fragmentation ions easily formed under the positive ionization mode, even at low collision energy (6 V). Therefore, in our study, both negative and positive scanning modes were used to obtain more information for structural elucidation of SFKDs in B. aromatica.
As shown in Figure  .1543 corresponded to the successive loss of the formic acid group, the cleavage of C 9 -C 11 , and C 11 -C 12 , respectively. As for the low mass fragments, such as that at m/z 145.1005 (C 11 H 13 ) and m/z 133.1010 (C 10 H 13 ), they were derived from the A-ring and B-ring segments. In brief, the proposed fragmentation pathways of the SFKDs were mainly derived from the neutral losses of COOH, CHO, and CH 2 OH at C-4 and H 2 O, the cleavage of C 8 -O-C 13 , and then the cleavage of A and B rings. Notably, ions at m/z 273, 255, 221, and 175 were found to be the characteristic fragment ions of these types of SFKDs. The prominent fragmentation pathway proposed for acromatin C is shown in Figure 4, which is expected to facilitate the characterization of SFKDs after filtration using the MDF method.  Figure 6). The characteristic fragments for this type of SFKDs were also observed at m/z 275, 257, and 177, which could be used for the identification of ions in this type of FSKD structure.

Structural Elucidation of Forskolin
Forskolin shares a common core frame structure with acromatin A, acromatin B, and acromatin C, which belong to the 8,13-epoxy-14ene labdane type of diterpenes. The differences are in the substituent group of the structure. There is usually a carbonyl substitution at the C-11 position of forskolin, dimethyl substitution at C-4, and acetyl substitution at the C-6 or C-7 position. Additionally, no hydroxyl group is substituted in the C-12 position (

Accuracy Evaluation of the MDF Filtration Method
The effectiveness of the MDF approach using the core substructure and substitution groups as a filter reference has been well demonstrated in the B. aromatica extract. After removal of the interfering compounds in B. aromatica extract using the above UHPLC-QTOF-MS and MDF filtration approach, the background noise of the BPI chromatograph ( Figure 4) was significantly reduced and detection sensitivity was successfully enhanced. The efficiency of the MDF approach was also evaluated based on the number of FSKD compounds identified using UHPLC-QTOF-MS/MS platform divided by the number of compounds selected by MDF as potential FKSDs. Theoretically, all ions with a mass defect value within the setting range should be selected for further structural elucidation. However, some of these ions were not identified due to a weak response intensity, leading to the lack of MS/MS data and unavailability of fragment ions of the selected precursor ions under the MS conditions selected. In order to increase screening efficiency, these ions need to have a MS response intensity above 2.0 e 4 (positive mode) or 6.0 e 4 (negative mode) to be suitable for use in further research.
After filtering the FSKDs in B. aromatica extract using MDF, a total of 45 peaks were selected within the defined range. Then, UHPLC-QTOF-MS/MS was applied to characterize both target and non-target components in B. aromatica extract. Finally, a total of 38 potential FSKD precursor ions were selected based on accurate mass, fragment behaviors, and diagnostic ions, described above. All identified FSKDs were confirmed to be distributed in the established mass defect range. The MS/MS fragmentation patterns of the 38 FSKDs are summarized in Table 1. The BPI chromatogram of the FSKDs under the positive ion mode is shown in Figure 4. On the basis of the results of structural characterization, the accuracy of the MDF filtration method for FSKDs in B. aromatica was found to be 84.44%.
It is notable that, many FSKD isomers were detected in the B. aromatica extract. These include peaks at 14, 17, 20, 23, 25, /z 337, 319, 301, 289, 273, 255, 221, and 175, which indicates that these compounds share the same or similar substructure. At the same time, the diagnostic fragment ions of 273, 255, 221, and 175 are consistent with that of acromatin C. Therefore, compound 9, 10, 12, and 13 were deduced to be FSKD analogues. Using the same strategy, other diterpenoids in B. aromatica extract were characterized as FSKDs.
It is noteworthy that the diagnostic fragment ions at m/z 297 and 201 were not detected in the filtered peaks and neither was the characteristic neutral loss of acetyl (-42) observed. Therefore, it was assumed, that they are different from the structure of forskolin, with no carbonyl group substituted at the C-11 position and no acetyl group substitution at C-6 and C-7 positions in these types of FSKDs detected in B. aromatica extract.

Chemicals and Materials
The authentic standard of forskolin and isoforskolin were purchased from the National Institutes for Food and Drug Control (Beijing, China). Acromatin A, acromatin B and acromatin C were isolated from B. aromatica in our laboratory and their structures were identified using ESI-MS and NMR. The purity was more than 98% (determined by HPLC). Acetonitrile (HPLC-grade) and formic acid with a purity of >98% were purchased from Merck KGaA (Darmstadt, Germany), and ultrapure water purified with a Milli-Q Academic system (Millipore, Bedford, MA, USA) was used for the LC/MS analysis. The structures of the reference compounds are summarized in Figure 2.
The herbal material was collected from Fengshan county, Hechi, Guangxi Province of China and authenticated by S. X. Feng as Blumea aromatica DC. and the voucher specimens (No. 153603) were deposited at the Herbarium Centre, Guangxi Botanical Garden of Medicinal Plants, Nanning, China.

Preparation of Sample and Standard Solutions
The air-dried sample of B. aromatica was pulverized into a powder and sieved through an 80-mesh screen. An aliquot of 0.5 g of dried B. aromatica powder was suspended in 20 mL of 80% methanol (v/v) in a 25 mL volumetric flask and sonicated at 100 Hz for 45 min. After reaching room temperature, the suspension was made up to 25 mL by adding 80% methanol (v/v). The solutions obtained were centrifuged at 13,000 rpm for 10 min, then filtered through 0.22 µm filter units, before being injected into UHPLC-Q-TOF/MS systems.
Stock solutions: A specified amount of forskolin, isoforskolin, acromatin A, acromatin B, and acromatin C were dissolved in 80% methanol-water (v/v) to obtain initial stock solutions at a concentration of 1.0 mg/mL, and then stored at under 4 • C in a refrigerator. All solutions were filtered using a 0.22 µm PTFE syringe filter before being injected for LC/MS analysis. MS/MS analysis was performed on a high-resolution quadrupole time-of-flight (QTOF-MS) system (Xevo G2-S, Waters Corporation, Manchester, UK), with a performance of resolution >30,000, mass accuracy <1 ppm, and coupled with an electrospray ionization (ESI) interface. All samples were determined at both positive and negative ionization mode with mass ranging from 100 to 1200 Da over a 28 min analysis time. LockSpray™ was used to ensure MS data accuracy and reproducibility. The other QTOF-MS operating parameters were optimized as follows: nitrogen was used as the source of gas, while ultrahigh purity argon was used as the collision gas for collision induced dissociation (CID). The desolvation gas flow rate was set at 800 L/h with a temperature of 350 • C, and the cone gas was set at 20 L/h and the cone voltage was set at 40 V, the source temperature was set at 120 • C and the capillary voltage was set at 2.7 kV and 3.0 kV at the negative and positive scanning modes, respectively. For the MS/MS experiments, an MS E experiment with a dynamic collision energy range of 30-50 V was conducted to obtain both precursor ions and fragment ions using one injection of the sample. The MS E parameters were set as follows: function 1: m/z 100-1200, 0.2 s scan time, 0.02 s inter-scan delay, 40 V sample cone, low collision energy set at 6 V; function 2: m/z 50-1200, 0.2 s scan time, 0.02 s inter-scan delay, 40 V sample cone, 30-50 V ramp high collision energy.

Data Process
The MDF method was followed using the "accurate mass filter" function of Masslynx 4.1 software. The molecular formulas were obtained using the "elemental composition" of Masslynx 4.1 software. The mass error of the predicted chemical formula lower than 10.00 ppm was selected as the candidate compound. Meanwhile, MassFragmentTM software, which can automate structural assignment based on a particular scoring system, was applied to assist the elucidation of chemical structure and to ensure the accuracy and efficiency of MDF. The MS/MS fragment analysis using MassFragment TM was performed at the positive scanning mode. The key parameters were set as follows: maximum mass error, double-bond equivalent DBE 0-10; electron count, odd; maximum H deficit, 6; and fragment number of bonds, 4.

Conclusions
In the present study, a powerful strategy was used to rapidly obtain the characteristic forskolin analogues of B. aromatica using a method of UHPLC-QTOF/MS coupled with MDF data filtration approach. Compared with the conventional method, the MDF strategy enables a majority of the interfering ions to be automatically filtered, allowing the original data to be analyzed more effectively, accurately, and rapidly. It could be used for the rapid analysis of other homologous families used in TCM, especially when there is a lack of literature and inaccessibility of standards of authentication.
A total of 38 FSKD precursor ions were characterized, or tentatively identified, in B. aromatica extract, for the first time. The newly discovered FSKDs significantly expand our understanding of the chemical constituents of B. aromatica, which is a potential alternative plant resource that contains forskolin-type diterpenoid active compounds. Accurate structural identification of chemical compounds in TCM requires integration of NMR, IR, UV, and MS data. The current DMF filtration strategy based on MS data is able to provide structural information of interested compounds used in TCM. Further research will be conducted at our lab to target the phytochemical and pharmacological activity of the FSKDs of important herbal medicines.
Author Contributions: UHPLC method development and validation, L.H. and Z.Z.; sample preparation, C.Y. and B.Y.; data curation, L.W.; conceived and designed the experiments, Z.S., S.W. and J.M.; project administration and analyzed the data: L.H. and Z.S.; paper writing-review and editing, Z.Z. and L.P. All authors read and approved the final manuscript.