Comparative Analysis of Coumarin Profiles in Different Parts of Peucedanum japonicum and Their Aldo–Keto Reductase Inhibitory Activities

Peucedanum japonicum (Umbelliferae) is widely distributed throughout Southeast Asian countries. The root of this plant is used in traditional medicine to treat colds and pain, whereas the young leaves are considered an edible vegetable. In this study, the differences in coumarin profiles for different parts of P. japonicum including the flowers, roots, leaves, and stems were compared using ultra-performance liquid chromatography time-of-flight mass spectrometry. Twenty-eight compounds were tentatively identified, including three compounds found in the genus Peucedanum for the first time. Principal component analysis using the data set of the measured mass values and intensities of the compounds exhibited distinct clustering of the flower, leaf, stem, and root samples. In addition, their anticancer activities were screened using an Aldo–keto reductase (AKR)1C1 assay on A549 human non-small-cell lung cancer cells and the flower extract inhibited AKR1C1 activity. Based on these results, seven compounds were selected as potential markers to distinguish between the flower part versus the root, stem, and leaf parts using an orthogonal partial least-squares discriminant analysis. This study is the first to provide information on the comparison of coumarin profiles from different parts of P. japonicum as well as their AKR1C1 inhibitory activities. Taken together, the flowers of P. japonicum offer a new use related to the efficacy of overcoming anticancer drug resistance, and may be a promising source for the isolation of active lead compounds.

Drug resistance is a major obstacle to the overall effectiveness of chemotherapy. Aldoketo reductases (AKRs), which consist of the AKR1C1-AKR1C4 members, play a pivotal role in nicotinamide-adenine dinucleotide phosphate (NADPH)-dependent reduction and are involved in biosynthesis, intermediate metabolism, and detoxification [24,25]. With respect to the catalytic-dependent function of AKR1C isoforms in cancer chemotherapy, AKRs mediate the reduction of carbonyl groups in anticancer drugs (e.g., daunorubicin, doxorubicin, and oracin) to their inactive metabolites, thus contributing to drug resistance [26][27][28][29]. In addition, AKR1C1 is associated with cisplatin-resistance in colon cancer cell, head and neck squamous cell carcinoma, nasopharyngeal carcinoma, and paediatric T-cell acute lymphoblastic leukemia [30][31][32][33]. It has been recently reported that AKR1C3 mediates chemotherapy resistance in breast cancer and esophageal adenocarcinoma [34,35]. Thus, AKR1C isoforms may be a potential target for preventing the development of anticancer drug resistance. Moreover, AKR1C1 plays a role in regulating the cellular concentration of progesterone by catalyzing the reduction of progesterone to its inactive form, 20-alphahydroxy-progesterone, thus it is also considered a drug target for the prevention of pre-term birth throughout the inhibition of progesterone metabolism mediated by AKR1C1 [36].
In our work to identify AKR1C1 inhibitors from natural products to overcome cisplatinresistance in human non-small lung cancer cells (NSCLC), the chemical composition of the extracts from four parts of P. japonicum and their AKR1C1 inhibitory activities in NSCLC (A549) were examined. Ultra-performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UPLC-QTof MS) and multivariate analysis was used to characterize the metabolomic differences between the four extracts.

Identification of Coumarins in the Different Parts of P. japonicum Using UPLC-QTof MS
Four parts of P. japonicum (flowers, roots, leaves, and stems) were prepared with six samples from each source that were designated as follows: S1-S6 for flower, S13-S18 for root, S7-S12 for leaf, S19-S24 for stem parts. Using UPLC-QTof MS with a gradient solvent system of acetonitrile (containing 0.1% formic acid) and water (containing 0.1% formic acid), the metabolites of the 24 P. japonicum samples were separated at high resolution within 22 min in the base peak ion (BPI) chromatogram. The representative BPI chromatograms of the methanol extract of flower, root, leaf, and stem of P. japonicum are shown in Figure 1. In these BPI chromatograms, the flower extract exhibited the most peaks and was fractioned using dichloromethane to obtain a chromatogram with increased intensity of coumarins ( Figure S1). The mass spectrum of each peak in the BPI chromatogram of the dichloromethane fraction of the P. japonicum flowers (Figures S2-S29) was carefully interpreted by analyzing its experimental and theoretical high-resolution MS (the positive mode), error ppm, molecular formula, and MS/MS fragmentation (Table 1), as well as by comparing with existing data for coumarins of the Peucedanum species reported in the literature [3,6,[37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52][53]. In addition, UPLC-PDA chromatograms of all the peaks present in the methanol extract of four parts of P. japonicum and their UV-Vis spectra between 210-500 nm were shown in Figures S30 and S31. The coumarin structures, which were tentatively identified from the analysis of their MS data, are shown in Figure 2.   Peaks 1-4 and 8 had a common fragment ion for 4-hydroxy-7H-furo [3,2-g][1]benzopyran-7one (5-hydroxypsoralen) or 9-hydroxy-7H-furo [3,2-g][1]benzopyran-7-one (8-hydroxypsoralen) at m/z 203.0344 [M − R + H] + , which resulted from the loss of different substituents at the C-5 oxygen or the C-8 oxygen [37]. In a review of coumarins isolated from Peucedanum species [1], the isolation of linear furanocoumarins, such as bergapten, cnidilin, isoimperatorin, and oxypeucetanin hydrate from P. japonicum was reported. Therefore, the linear furanocoumarins in this study were tentatively identified by giving preference to the structure based on the 5-hydroxypsoralen backbone. Peak 1 (t R 1.831 min) produced a major molecular ion at m/z 305.1027 [M + H] + (calculated for C 16 H 17 O 6 + , 305.1025) and yielded a fragment ion at m/z m/z 203.0344 [M − R + H] + , indicating that a 2,3-dihydroxy-3-methylbutyl group was substituted at the 5-OH of 5-hydroxypsoralen. Therefore, peak 1 was tentatively identified as oxypeucedanin hydrate, which has been previously reported in P. japonicum [38]. Peak 2 (t R 4.712 min) produced a molecular ion at m/z 319.1223 [M + H] + (calculated for C 17 H 18 O 6 + , 319.1223), which was 14 Da less than that of peak 1, indicating the presence of a methoxy group instead of a hydroxyl group. The fragment ions at m/z 203.0366 [M − R + H] + represent a 2-hydroxy-3-methoxy-3-methylbutyl at the 5-OH of 5-hydroxypsoralen. Thus, it tentatively was identified as oxypeucedanin methanolate, which has been previously reported in P. ostruthium [39], although it was first detected in P. japonicum. Peaks 3 (t R 4.895 min) and 4 (t R 5.889 min) produced a molecular ion at m/z 287.0919 [M + H] + (calculated for C 16 H 15 O 5 + , 287.0920), which is 18 Da less than that of peak 1, indicating the loss of a hydroxy group. The fragment ion at m/z 203.0340 [M − R + H] + indicated the presence of a 2-hydroxy-3methyl-3-butenyl group or a 33-hydroxy-3-methyl-1-butenyl group at the 5-OH of 5-hydroxypsoralen. Therefore, these two peaks were tentatively identified as 5-[(2-hydroxy-3methyl-3-butenyl)oxy]psoralen (pabulenol) and 5-[(3-hydroxy-3-methyl-1-butenyl)oxy]psoralen, respectively, but they are interchangeable. 5-[(2-Hydroxy-3methyl-3-butenyl)oxy]psoralen (pabulenol) has been identified in various species of Peucedanum [40], but not in P. japonicum, whereas 5-[(3-Hydroxy-3-methyl-1-butenyl)oxy]psoralen has not been reported previously. Peak 8 (t R 10.015 min) produced a molecular ion at m/z 293.0759 [M + Na] + (calculated for C 16 H 14 O 4 Na + , 293.0790), which is 16 Da less than that of peak 4, indicating the absence of a hydroxyl group. The fragment ions at m/z 203.0344 [M − R + H] + represent the presence of a 2-hydroxy-3-methoxy-3methylbutyl group at the 5-OH of 5-hydroxypsoralen. Thus, it was tentatively identified as isoimperatorin, which has been reported previously in P. japonicum [41].    [37]. In a review of coumarins isolated from Peucedanum species [1], the isolation of linear furanocoumarins, such as bergapten, cnidilin, isoimperatorin, and oxypeucetanin hydrate from P. japonicum was reported. Therefore, the linear furanocoumarins in this study were tentatively identified by giving preference to the structure based on the 5-hydroxypsoralen backbone. Peak 1 (tR 1.831 min) produced a major molecular ion at m/z 305.1027 [M + H] + (calculated for C16H17O6 + , 305.1025) and yielded a fragment ion at m/z m/z 203.0344 [M − R + H] + , indicating that a 2,3-dihydroxy-3-methylbutyl group was substituted at the 5-OH of 5-hydroxypsoralen. Therefore, peak 1 was tentatively identified as oxypeucedanin hydrate, which has been previously reported in P. japonicum [38]. Peak 2 (tR 4.712 min) produced a molecular ion at m/z 319.1223 [M + H] + (calculated for C17H18O6 + , 319.1223), which was 14 Da less than that of peak 1, indicating the presence of a methoxy group instead of a hydroxyl group. The fragment ions at m/z 203.0366 [M − R + H] + represent a 2-hydroxy-3-methoxy-3methylbutyl at the 5-OH of 5-hydroxypsoralen. Thus, it tentatively was identified as oxy- In the ion fragmentation pathways of angular dihydropyranocoumarins, such as khellactone (3,4,5-trihydroxy-2,2-dimethyl-6-chromanacrylic acid δ-lactone), the MS fragment peak of the molecules resulting from a loss of the C-4 substituent was observed as the main peak and the molecular ion peak due to fragmentation of the C-3 substituent was small or undetectable [6]. Based on this phenomenon, we concluded that the position of the C-3 and C-4 substituents could be determined by MS fragmentation. In this study, molecular ions for angular dihydropyranocoumarins also exhibited characteristic presence of a propionate group at C-4 and an acetate group at C-3 . Thus, peak 5 was identified as 3 -O-acetyl-4 -O-propanoylkhellactone, which was first detected in P. japonicum; however, it has also been isolated from P. prearuptorum [41]. Peak 6 (t R 8.261 min) produced a major molecular ion at m/z 369.1314 [M + Na] + (calculated for C 19 H 22 O 6 Na + , 369.1314) and yielded fragment ions at m/z 329.1384 [M-R 2 OH+H] + and 245.0814 [M-R 2 OH-(R 1 -H)+H] + , indicating the presence of a hydroxyl group at C-4 and a 2-methyl butanoate or isovalerate group at C-3 . Therefore, peak 6 was tentatively identified as 3 -O-(2-methyl-butyryl)-4 -hydroxy khellactone, which was isolated from P. japonicum [6] or 3 -O-(isovaleryl)-4hydroxy khellactone, which was isolated from P. turgeniifolium [42], but not P. japonicum.  [49], respectively, but they are interchangeable. Both structures have not been found in P. japonicum; however, they were reported in P. prearuptorum [45,48,49].  [52] has not been found in P. japonicum; however, it has been identified in P. prearuptorum [52]. Peak 26 (t R 18.799 min) exhibited the same molecular ion at m/z 451.1731 [M + Na] + as peak 25, which corresponded to the molecular formula C 24

Aldo-Keto Reductases Inhibitory Effects of the Different Parts for P. japonicum
To confirm the ability of methanol extracts of four different parts of P. japonicum, flower (S1-S3), root (S13-S15), leaf (S7-S9), and stem (S19-S21), to circumvent anticancer drug resistance, they were tested using an in vitro AKR1C1 activity assay. Among the four extracts, the flower extract exhibited AKR1C1 activity inhibition of 11.74 ± 0.48% at a concentration of 50 µg/mL without causing cytotoxicity, whereas the other extracts increased AKR1C1 activity (Figure 3a). By observing optical density over time, the flower extract also showed a lower value compared with the control (Figure 3c). Thus, the flower extract was considered a potential natural source for compounds that prevent or overcome anticancer drug resistance in cancer. In the current reports of AKR1C1 inhibitors, non-competitive inhibitors include benzodiazepines, such as medazepam, phthalimide, pyrimidine, and anthranilic acid derivatives [54]. The representative competitive inhibitor of AKR1C1 is 3-bromo-5-phenylsalicylic acid [55]. Efforts for the development of AKR1C1 inhibitors are continuing through molecular docking and in vitro enzymatic studies using a number of synthetic compound derivatives [56][57][58]; however, there have been few reports of AKR1C1 inhibitors developed from natural products [59,60]. Therefore, the flower extract could be a promising source for the isolation of active lead compounds that inhibit AKR1C1 activity.

Multivariate Analysis
This study compared the differences in coumarin composition between the flowers, roots, leaves, and stems of P. japonicum. Using UPLC-QTof MS, the extracts from four different parts were analyzed. Twenty-four samples, six from each part containing 26 coumarins, were included in the multivariate analysis comparison study (peaks 2 and 5 were not included because they were not detected in methanol extracts). Principal component analysis (PCA) clearly distinguished flower (S1-S6), root (S13-S18), leaf (S7-S12), and stem (S19-S24) samples. As shown in Figure 4, leaf (S7-S12) and stem (S19-S24) samples clustered into one large group. PCA was conducted with three principal components (PC1-PC3) that described the variation, 0.913% R2X, and predictive capability, 0.826% Q2. Eigenvalues for PC1 and PC2 were 15.9 and 4.98, respectively, indicating that these first two principal components account for a substantial amount of the data variance. The relatively smaller eigenvalue of PC3 (0.996) led us to select only PC1 and PC2 for further examination. The first two principal components accounted for 87.2% of the variance (66.4% and 20.8% by PC1 and PC2, respectively). The corresponding PCA loading plot (Figure 4b) revealed that most of the peaks were coumarins, except for peaks 9, 10, and 19-21, which were chemical markers responsible for separating the flower sample from the root, leaf, and stem samples. Based on the PCA values, the coumarin profiles for the four parts of P. japonicum were distinct. Moreover, a variable average by different parts (Figure 4c) also clearly showed that most identified coumarins are abundant in the flower part compared with the root, leaf, and stem parts. Finally, leaves and stems share a chemical profile that is distinct from that of roots and flowers.
was considered a potential natural source for compounds that prevent or overcome anticancer drug resistance in cancer. In the current reports of AKR1C1 inhibitors, non-competitive inhibitors include benzodiazepines, such as medazepam, phthalimide, pyrimidine, and anthranilic acid derivatives [54]. The representative competitive inhibitor of AKR1C1 is 3-bromo-5-phenylsalicylic acid [55]. Efforts for the development of AKR1C1 inhibitors are continuing through molecular docking and in vitro enzymatic studies using a number of synthetic compound derivatives [56][57][58]; however, there have been few reports of AKR1C1 inhibitors developed from natural products [59,60]. Therefore, the flower extract could be a promising source for the isolation of active lead compounds that inhibit AKR1C1 activity.

Multivariate Analysis
This study compared the differences in coumarin composition between the flowers, roots, leaves, and stems of P. japonicum. Using UPLC-QTof MS, the extracts from four different parts were analyzed. Twenty-four samples, six from each part containing 26 coumarins, were included in the multivariate analysis comparison study (peaks 2 and 5 were To further identify marker compounds responsible for the potency of flower samples, OPLS-DA with a VIP-plot and S-plot were used to compare the groups. Flower samples (S1-S6) were designated as Group 1, while root (S13-S18), leaf (S7-S12), and stem samples (S19-S24) were designated as Group 2. In the OPLS-DA model, flowers versus roots, leaves, and stems were clearly differentiated into two clusters (Figure 5a). The OPLS-DA model had a cumulative R2Y value of 1.00 and a cumulative Q2 value of 0.968. The internal validation of the OPLS-DA model was conducted by a permutation test (n = 200).
The intercept values of R2 and Q2 in the permutation test were -0.0066 and -0.399, respectively. All permutations to the left of the R2 and Q2 values were less than the original points to the right ( Figure S32). These values indicated that the OPLS-DA model for the analysis was robustly validated and not overfit. Several potential marker compounds capable of distinguishing the two groups were located apart from the center of the corresponding OPLS-DA S-plot (Figure 5b). Seven coumarins, peaks 8, 11, 23, 25-28, were shifted in the same direction as the Group 1 samples on the score scatter plot. The variable importance for the projection plot (VIP-plot, Figure 5c) indicates that these seven coumarins are the primarily responsible markers for discriminating between Group1 and Group 2 with significant VIP values (VIP ≥ 1). This indicates that the Group 1 samples contained higher concentrations of these coumarins compared with the Group 2 samples. Therefore, the composition and relative content of key markers that differentiate between flowers and other parts contribute to distinguishing active and inactive plant parts. clustered into one large group. PCA was conducted with three principal components (PC1-PC3) that described the variation, 0.913% R2X, and predictive capability, 0.826% Q2. Eigenvalues for PC1 and PC2 were 15.9 and 4.98, respectively, indicating that these first two principal components account for a substantial amount of the data variance. The relatively smaller eigenvalue of PC3 (0.996) led us to select only PC1 and PC2 for further examination. The first two principal components accounted for 87.2% of the variance (66.4% and 20.8% by PC1 and PC2, respectively). The corresponding PCA loading plot (Figure 4b) revealed that most of the peaks were coumarins, except for peaks 9, 10, and 19-21, which were chemical markers responsible for separating the flower sample from the root, leaf, and stem samples. Based on the PCA values, the coumarin profiles for the four parts of P. japonicum were distinct. Moreover, a variable average by different parts (Figure 4c) also clearly showed that most identified coumarins are abundant in the flower part compared with the root, leaf, and stem parts. Finally, leaves and stems share a chemical profile that is distinct from that of roots and flowers.  for the projection plot (VIP-plot, Figure 5c) indicates that these seven coumarins are the primarily responsible markers for discriminating between Group1 and Group 2 with significant VIP values (VIP ≥ 1). This indicates that the Group 1 samples contained higher concentrations of these coumarins compared with the Group 2 samples. Therefore, the composition and relative content of key markers that differentiate between flowers and other parts contribute to distinguishing active and inactive plant parts.

Sample Preparation
Freeze-dried parts of P. japonicum (flowers, roots, leaves, and stems) were ground into powder using a mixer. Six samples were prepared from each part and designated as follows: S1-S6 for flower, S13-S18 for root, S7-S12 for leaf, S19-S24 for stem parts. Then, 1 g of each sample was extracted with 20 mL of methanol in an ultrasonic bath for 1 h and then concentrated. The dried extract (10 mg) was dissolved in 1 mL of methanol and filtered using a 0.20 µm polyvinylidene fluoride filter. The samples were diluted with methanol to a concentration of 500 ppm for further LC-MS analysis. For the evaluation of bioactivity, each dried extract was initially dissolved in dimethyl sulfoxide (DMSO) at a concentration of 50 mg/mL. All extraction and chromatographic solvents used in this study were of analytical grade (J. T. Baker, Phillipsburg, NJ, USA).

Aldo-Keto Reductases Activity Assay
The AKR1C1 activity of the extracts was measured using the EZDetect TM Aldo-keto Reductase Activity Assay Kit (Colorimetric) (BioVision Inc., Waltham, MA, USA), based on the manufacturer's instructions. In this assay, AKR uses a general substrate to convert nicotinamide-adenine dinucleotide phosphate (NADP+) to dihydronicotinamide-adenine dinucleotide phosphate (NADPH). NADPH reacts with the AKR probe and generates a color which is proportional to the activity of AKR in the sample. To determine the effect of the extract on AKR enzyme activity in A549 non-small-cell lung cancer cells (American Type Culture Collection, Rockville, MD, USA), the cells were cultured in a 6-well plate at a density of 5 × 10 5 cells/mL for 24 h. They were subsequently treated with the extracts at a concentration of 50 µg/mL for an additional 24 h. After the incubation period, the cells were harvested according to the Kit manual for measuring AKR1C1 activity. The assay was performed using a single experiment and the samples were evaluated in triplets.

Statistical and Multivariate Analysis
The data set of the measured mass values and intensities (peak area) for each sample was exported to SIMCA 17.0.2 (Umetrics, Umeå, Sweden) and analyzed by multivariate statistical analysis including PCA and orthogonal partial least-squares discriminant analysis (OPLS-DA). An S-plot was used to identify the potential marker compounds that are responsible for the differentiation of the flower part from the root, leaf, and stem parts. Statistical analysis of the biological assay was performed to identify significant differences between the control and test groups (unpaired t-test) using GraphPad Prism9 software (GraphPad Prism version 9.0.0 (121) for Mac, San Diego, CA, USA).

Conclusions
A comparative study was done to identify coumarin composition and AKR1C1 inhibitory activities in different parts of P. japonicum. UPLC-QTof MS analysis was used to identify 28 compounds, including three coumarins which have not been reported in the genus Peucedanum. Among the methanol extracts of different parts of P. japonicum, the flower extract exhibited AKR1C1 activity inhibition in A549 cells at a concentration of 50 µg/mL without cytotoxicity. PCA results showed that the metabolites discrimi-nate among the flower, leaf, stem, and root samples of P. japonicum and the OPLS-DA results revealed the responsible markers for discriminating between the active flower extract and the other three inactive parts. These markers were tentatively identified as isoimperatorin (peak 8), 3 -O-acetyl-4 -O-(2-methyl butanoate)khellactone (or 3 -O-acetyl-4 -O-isovalerylkhellactone; peak 11), 3 -O-(2-methyl butyryl)-4 -O-isobutyrylkhellactone (or 3 -O-isovaleryl-4 -O-isobutyrylkhellactone; peak 23), and compounds having a khellactone structure, in which four substituents, angeloate, senecioate, 2-methyl butyrate, and isovalerate groups, were frequently substituted at C-3 and C-4 (peaks [25][26][27][28]. Therefore, these markers have the potential to have the ability to inhibit AKR1C1 activity, however further study involving the isolation of these compounds, the determination of their structures and stereochemistry, and the evaluation of their AKR1C1 activity inhibitory effects is required. Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/molecules27217391/s1, Figure S1: LC-MS base peak ion chromatograms of the flowers of P. japonicum at positive ion mode (6 eV, ESI+), Figure S2: ESI-QTof-MS spectrum of oxypeucedanin hydrate (peak 1), Figure S3: ESI-QTof-MS spectrum of oxypeucedanin methanolate (peak 2), Figure S4: ESI-QTof-MS spectrum of pabulenol (peak 3), Figure