Metabolite Profiling and Dipeptidyl Peptidase IV Inhibitory Activity of Coreopsis Cultivars in Different Mutations

Coreopsis species have been developed to produce cultivars of various floral colors and sizes and are also used in traditional medicine. To identify and evaluate mutant cultivars of C. rosea and C. verticillata, their phytochemical profiles were systematically characterized using ultra-performance liquid chromatography time-of-flight mass spectrometry, and their anti-diabetic effects were evaluated using the dipeptidyl peptidase (DPP)-IV inhibitor screening assay. Forty compounds were tentatively identified. This study is the first to provide comprehensive chemical information on the anti-diabetic effect of C. rosea and C. verticillata. All 32 methanol extracts of Coreopsis cultivars inhibited DPP-IV activity in a concentration-dependent manner (IC50 values: 34.01–158.83 μg/mL). Thirteen compounds presented as potential markers for distinction among the 32 Coreopsis cultivars via principal component analysis and orthogonal partial least squares discriminant analysis. Therefore, these bio-chemometric models can be useful in distinguishing cultivars as potential dietary supplements for functional plants.


Introduction
Coreopsis species are annual or perennial plants belonging to the Asteraceae (Compositae) family [1]. Approximately 80 species of Coreopsis are native to North America and are currently widespread in America, Asia, and Oceania regions [2][3][4]. They are usually cultivated for ornamental purposes in gardens or on roadsides. The plants are in the range of 46-120 cm in height and the petals of the flowers are primarily yellow in color and are serrated [5,6]. The color and size of Coreopsis flowers have commercially important value and are the reason for Coreopsis breeding. In addition, the Coreopsis flower has been ethnopharmacologically used for the treatment of diarrhea, vomiting, and hemorrhage in North America, where the Coreopsis species originates [7,8]. It has also been used as a drink to control diabetes in China and Portugal, and as an herbal tea to eliminate toxins and fever from the body in China [8][9][10]. Nowadays, owing to scientific proof of its traditional use, several studies have been conducted on the phytochemical and biological Lemon candy 4418 -γ-Irradiated mutation 5 Shiny pink 4420 -γ-Irradiated mutation 6 Uri-dream 01 3993 -Herbicide-induced artificial mutation 7 Luckyten5 4411 -Herbicide-induced artificial mutation 8 Luckyten9 4413 -Herbicide-induced artificial mutation 9 Uri-dream red 6001 -γ-Irradiated mutation 10 Uri-dream 07 3998 -Herbicide-induced artificial mutation 11 Uri-dream 06 3997 -Herbicide-induced artificial mutation 12 Pink sherbet 4415 -γ-Irradiated mutation II (C. verticillata) 13 Citrine --Original cultivar 14 Golden Pumpkin Pie --Original cultivar 20 Gold ring 7523 -γ-Irradiated mutation 21 Golden ring 5994 -γ-Irradiated mutation 22 Mini ball yellow 6453 -γ-Irradiated mutation 23 Box tree 6462 -γ-Irradiated mutation 24 Orange ball 6005 -γ-Irradiated mutation IV (C. verticillata) 25 Route  As part of our investigation of the effects of mutation on metabolic changes between the original mutant cultivars and their biological functions, we analyzed metabolite profiling of the five original cultivars and each mutant cultivar. Given that Coreopsis species have been known to be effective for diabetes in folk medicine, 70% methanol extracts of 32 Coreopsis samples were evaluated for their inhibitory effect against dipeptidyl peptidase (DPP)-IV, a target of incretin-based therapies for the treatment of type 2 diabetes mellitus.

Subsection Identification of Metabolites in Coreopsis Cultivars Using UPLC-QTof-MS
Metabolites in Coreopsis cultivars were tentatively identified using UPLC-QTof-MS. The Metabolites were separated with high resolution within 10 min in the base peak ion (BPI) chromatogram. BPI chromatograms of the original Coreopsis cultivars are shown in Figure S1. The mass spectrum of each peak was carefully interpreted by analyzing its experimental and theoretical high resolution MS (the deprotonated molecular ion, [M -H] − ), error ppm, molecular formula, and MS/MS fragmentation. Additionally, these were compared with data from the literature of plants belonging to the same genus, such as C. tinctoria (known as snow chrysanthemum) and C. lanceolata [3,5,6,12,[26][27][28][29]. Moreover, its mass spectrum was compared to that in Waters Traditional Medicine Library that is built in UNIFI software (Waters, Milford, MA, USA) and MassBank available online (a public database for sharing mass spectral data) [30,31]. Forty compounds, including phenolic acids, flavonoids, and a polyacetylene were identified in methanol extracts of original and mutant cultivars of C. rosea and C. verticillata (Table 2). However, a peak observed in total ion chromatograms of all, or some Coreopsis cultivars could not be identified in this study.

. Flavanones and Flavanonols
It has been reported that a retro Diels-Alder reaction, as well as the loss of H 2 O, sugar (usually glucose), and carbonyl groups were observed in the ion fragmentation pathways of flavonoids [3]. Flavanones, chalcones, and their glycosides have been known as the major types of flavonoids found in Coreopsis species [28]. These compounds usually showed the loss of H 2 O caused by the disposition of hydroxyls at C-3 and C-4 in the flavanone structure or at C-3 and C-4 in the chalcone structure, and the loss of a glucose at C-7 in the flavone structure or at C-4 in the chalcone structure [3]. These phenomena were observed in mass spectra of flavanones and chalcones identified in this study. Thus, it was tentatively identified as flavanomarein [26]. The fragmentation pattern of peak 6 (t R 5.85 min) was identical to that of peak 5, except for the major molecular ion at m/z 595. . Therefore, peak 6 was predicted as isookanin-7-Orutinoside, which has been described previously [27]. Given that peak 16 (t R 6.64 min) also exhibited the same fragment ions with those of peaks 5 and 6, it was tentatively identified as their aglycone, isookanin [3,26].

Chalcones
Peak 23 has the same fragment rules as flavanomarein; however, it has been known that flavanones have shorter retention times than chalcones in chromatographic elution [29]. Therefore, peak 23 (t R 7.15 min) was identified as marein [3,26,27]. The aglycone of this compound, okanin (peak 32, t R 8.26 min), produced identical fragment ions with peak 23 [3,26,27]. Similarly, peak 30 (t R 8.02 min) showed the same molecular ion and fragment ions as peak 8, thus identified as coreopsin [3,26]. Butein (peak 38, t R 9.20 min), the aglycone of peak 30 O] − indicated the presence of a 3 ,4 -dihydroxyphenyl group for the B ring and the absence of a 3-hydroxy group in the C ring. Therefore, peak 33 was tentatively identified as eriodictyol chalcone-7-O-(glucosyl glucoside) or eriodictyol chalcone-O-diglucoside, which have not been described previously. Eriodictyol chalcone-7-O glucoside, which has one glucose, has been found in Antirrhinum majus [41]. Peak 34 (t R 8.73 min) produced a molecular ion at m/z 287.0553 [M -H] − and exhibited the same fragment pathway with that of peak 33, suggesting that it was an aglycone of peak 33, eriodictyol chalcone, which has been identified in Coreopsis species [42]. Peak 37 (t R  ] − . Therefore, peak 37 was tentatively identified as 4-methoxylanceoletin-4 -O-glucoside, which has been isolated from C. lanceolata [12], or lanceolein 2 -methyl ether, which has not been described previously.

Flavones and Flavanols
Flavone having a double bond between C-2 and C-3 exhibits a molecular ion that is 2 Da less than that of flavanone or chalcone and characteristic fragment ions by the loss of C 8  − , 285.0405) that were 162 Da and 324 Da less than that of peak 8, respectively, indicating that the sugar moiety was removed from C-2 and C-7 in luteolin-7-O-sophoroside (peak 7). Accordingly, peaks 22 and 36 were tentatively identified as luteolin-7-O-glucoside and luteolin, respectively [3,26,27]. In addition, a molecular ion at m/z 431.0978 [M -H] − (calculated for C 21 H 19 O 10 − , 431.0984) for peak 29 (t R 7.91 min) was 146 Da more than that of peak 36, indicating the addition of a rhamnose. Moreover, its fragment ions were similar to those of peaks 22 and 36. Thus, it was identified as luteolin-7-O-rhamnoside, which was first detected in Coreopsis species; however, it has been found in other plants, such as Glechoma grandis Kuprianova var. longituba, Rumex algeriensis, and Cornulaca monacantha [43][44][45]. Peak 15 (t R 6. This peak was tentatively identified as lobetyolinin, which was confirmed by the UNIFI local library and first detected in Coreopsis species; however, it has primarily been found in Lobelia species [54,55].

DPP-IV Inhibitory Effects of the 70% Ethanol Extract Obtained from Coreopsis cultivars
Type 2 diabetes mellitus is determined by several factors, including pancreas βcell dysfunction, insulin resistance, increased hepatic and intestinal glucose production, or deficient insulin secretion [56]. Recently, the incretin effect has been observed to be reduced in patients with type 2 diabetes mellitus, which is a symptom of increased insulin secretion induced by oral administration, such as eating a meal, compared to intravenous administration of glucose [56]. This effect is mediated by incretin hormones, glucagonlike peptide-1 (GLP-1), and glucose-dependent insulinotropic polypeptide (GIP), which stimulate insulin secretion from pancreatic β-cells and consequently increase the blood glucose level [57,58]. In the incretin system, an increase of the elimination of GLP-1 and GIP occurs primarily through enzymatic degradation of DPP-IV [59]. Thus, DPP-IV inhibition enhances the function of insulinotropic hormones. It improves glucose tolerance in patients with type 2 diabetes mellitus [58]. Hence, DPP-IV inhibitors have emerged as a new class of oral anti-diabetic agents, and synthetic compounds have mainly been used in current treatments with these inhibitors [59]. However, there have also been studies that show that DPP-inhibitors are derived from natural sources as promising candidates of functional foods or pharmaceuticals [12,[60][61][62][63].
In this study, the 70% ethanol extract of original and mutant cultivars of C. rosea and C. verticillata confirmed their anti-diabetic effect using an in vitro DPP-IV inhibitor screening assay. All extracts inhibited DPP-IV activity in a concentration-dependent manner with IC 50 values from 34.01 to 134.28 µg/mL ( Table 3) Among all Coreopsis cultivars samples, 'Orange sunlight (No. 30)' showed the best efficacy with an IC 50 value of 34.01 µg/mL; however, 'Uri-dream red (No. 9)' (IC 50 , 66.46 µg/mL) had the highest increase with 47% DPP-IV inhibitory activity compared to the original cultivar (IC 50 , 125.29 µg/mL). Therefore, 'Orange sunlight (No. 30)' had the potential to develop as a functional food, such as a tea ingredient or a food additive for the prevention or treatment of type 2 diabetes. The mutant cultivars with a greater increase in activity compared to the original cultivar, such as 'Uri-dream red (No. 9)' may be used for studies to identify metabolites changed by mutation using multivariate analysis, and for further research on genomic mutation mechanism.  1 Values are presented as the mean ± SD of three independent experiments. 2 Sitagliptin was used as the positive control.

Multivariate Analysis
Metabolite differences among original and mutant cultivars, C. rosea and C. verticillate, were examined based on the metabolite profiles analyzed by UPLC-QTof-MS. However, it was difficult to find differences among the samples in chromatograms. Therefore, PCA and OPLS-DA were used to provide an effective visualization for the classification and differentiation of a metabolome system.
To compare metabolites from different cultivars of C. rosea and C. verticillata, we performed PCA analysis on negative ion mode data obtained from UPLC-QTof-MS analysis. PCA analysis was performed with three principal components (PC1-PC3) describing variation explained, 0.66 of R 2 X and predictive capability, 0.366 of Q 2 . Eigenvalues for PC1 and PC2 were found to be 9.94 and 8.05, respectively, indicating these first two principal components explain a large amount of the variance in the data. PC3 showed a comparatively smaller eigenvalue of 3.13, which led us to choose only PC1 and PC2 for further analysis. As shown in Figure 2A, the first two principal components described 56.2% of the total variation (31.1% and 25.1% by PC1 and PC2, respectively), and 32 Coreopsis samples were clearly clustered into four groups. Group I and Group III were clustered together, indicating similar chemical profiles among samples, and these two groups were of the same species, C. rosea. This cluster also suggested that there is no distinct difference between original cultivars and other mutation cultivars induced from each original one. However, an exception was found in 'Luckyten 6 (No. 2)', which is one of the mutant cultivars artificially induced using herbicide from the original cultivar of Group I, 'Heaven's gate (No. 1)'. Alternatively, a distinct separation was observed from cultivars in Group II, Group IV, and Group V, although they were all included in C. verticallata. Group II demonstrated different chemical profiles compared with Group IV and Group V. However, Group II showed one clustering with no substantial deviations between the γ-irradiated mutant cultivars (No.14-No. 18) and the original cultivar 'Citrine (No. 13)'. In Group IV, the γ-irradiated mutant cultivar, 'Orange sunlight (No. 30)' was shown as the outlier, indicating that it had a different chemical profile than samples within the same group. Figure 2B, shows the derivation of markers primarily distributed among the four groups. However, this resulted in whole variability directions, with no distinction of variabilities among groups. Accordingly, we performed OPLS-DA analysis on the metabolite profiles between C. rosea cultivars (Group Cr) and C. verticallata cultivars (Group Cv) to find the differentiation and significant variances in these two species. Two clusters were clearly differentiated from each other according to species in the OPLS-DA model, with a cumulative R 2 Y value of 1.00 and a cumulative Q 2 value of 0.94 ( Figure 2C). However, 'Luckyten 6 (No. 2)' and 'Orange sunlight (No. 30)' were marginally out of each grouped sample area. The internal validation of OPLS-DA model was performed by a permutation test (n = 200). In permutation test, the intercept values of R 2 and Q 2 were 0.425 and −1.09 respectively. All permutations of the R 2 and Q 2 values to the left were lower than the original points to the right and the intersection of regression lines of the R 2 and Q 2 points on vertical axis was below 0.4 and −1.1, respectively ( Figure S43). These values indicated OPLS-DA model of this analysis was strongly validated without overfitting of the original model. As shown in Figure 2D, the corresponding OPLS-DA S-plot enabled the derivation of 13 potential marker compounds responsible for separating two groups by being far from the center. Eight marker metabolites which were shifted in the same direction as Group Cv from the OPLS-DA score plot were peaks 2, 8, 19, 21, 27, 30, 36, and 38, indicating the most abundant markers in Group Cv. Five marker metabolites, peaks 1, 5, 20, 23, and 34, were at the highest level in Group Cr. The variable importance plot (VIP) ( Figure 2E) confirms these 13 selected marker compounds are primarily responsible for the discrimination between Group Cr and Group Cv with high VIP values (VIP ≥ 1). Moreover, the variable average by group clearly shows differences of selected marker compounds ( Figure 2F) in these groups. shown as the outlier, indicating that it had a different chemical profile than samples within the same group. Figure 2B, shows the derivation of markers primarily distributed among the four groups. However, this resulted in whole variability directions, with no distinction of variabilities among groups. Accordingly, we performed OPLS-DA analysis on the metabolite profiles between C. rosea cultivars (Group Cr) and C. verticallata cultivars (Group Cv) to find the differentiation and significant variances in these two species. Two clusters were clearly differentiated from each other according to species in the OPLS-DA model, with a cumulative R 2 Y value of 1.00 and a cumulative Q 2 value of 0.94 ( Figure 2C). However, 'Luckyten 6 (No. 2)' and 'Orange sunlight (No. 30)' were marginally out of each grouped sample area. The internal validation of OPLS-DA model was performed by a permutation test (n = 200). In permutation test, the intercept values of R 2 and Q 2 were 0.425 and -1.09 respectively. All permutations of the R 2 and Q 2 values to the left were lower than the original points to the right and the intersection of regression lines of the R 2 and Q 2 points on vertical axis was below 0.4 and -1.1, respectively ( Figure S43). These values indicated OPLS-DA model of this analysis was strongly validated without overfitting of the original model. As shown in Figure 2D, the corresponding OPLS-DA S-plot enabled the derivation of 13 potential marker compounds responsible for separating two groups by being far from the center. Eight marker metabolites which were shifted in the same direction as Group Cv from the OPLS-DA score plot were peaks 2, 8, 19, 21, 27, 30, 36, and 38, indicating the most abundant markers in Group Cv. Five marker metabolites, peaks 1, 5, 20, 23, and 34, were at the highest level in Group Cr. The variable importance plot (VIP) ( Figure 2E) confirms these 13 selected marker compounds are primarily responsible for the discrimination between Group Cr and Group Cv with high VIP values (VIP ≥1). Moreover, the variable average by group clearly shows differences of selected marker compounds ( Figure 2F) in these groups.
(a) (b) The similarities in chemical composition and relative quantitative differences among different cultivars of C. rosea and C. verticallata were clearly visualized on a heatmap with a dendrogram, while a hierarchical cluster analysis exhibited the same pattern of clustering as observed in PCA analysis ( Figure 3). Heatmap is considered as one of the best tools for converting qualitative data into quantitative. Group I (No. 1-No. 12) and Group III (No. [19][20][21][22][23][24] were clustered as one big cluster with similar distribution of areas of peaks 1, 3,4,5,9,11,12,13,17,20,21,22,23,24,26,27,34,40 The results of multivariate analyses to verify the correlation between metabolites and DPP-IV activities of the 32 Coreopsis samples were similar to the chemometric patterns The similarities in chemical composition and relative quantitative differences among different cultivars of C. rosea and C. verticallata were clearly visualized on a heatmap with a dendrogram, while a hierarchical cluster analysis exhibited the same pattern of clustering as observed in PCA analysis ( Figure 3). Heatmap is considered as one of the best tools for converting qualitative data into quantitative. Group I (No. 1-No. 12) and Group III (No. [19][20][21][22][23][24] were clustered as one big cluster with similar distribution of areas of peaks 1, 3, 4, 5, 9, 11, 12, 13, 17, 20, 21, 22, 23, 24, 26, 27, 34, 40, and 41. 'Luckyten 6' (No. 2) was observed to have comparatively higher area values for peaks 1, 3, 20, and 34 than other cultivars in Group I, indicating the relatively high contents of these four peaks when compared to other samples in Group I. These four peaks could be responsible for making 'Luckyten 6' (No. 2) an outlier. Peaks 6,25,31,35, and 37 appear with intense color in heatmap representing high quantity in comparison with other samples, which was responsible for the clustering of group II (No. [13][14][15][16][17][18]. Group IV has peaks 2, 8, 10, 14, 19, 28, 30, and 36 in abundance, while 'Orange sunlight' (No. 30) is rich in peaks 8, 10, 14, 19, 28, and 30 among groups. These six peaks' composition and relatively higher content could turn 'Orange sunlight' (No. 30) into an outlier in this statistical study. The contents of eight peaks 7,14,15,29,32,33,36, and 39 determine the clustering of group V (No. [31][32], adjacent to group IV, sharing some similarities between them. The results of multivariate analyses to verify the correlation between metabolites and DPP-IV activities of the 32 Coreopsis samples were similar to the chemometric patterns between the two species. Given that DPP-IV inhibitory activities of C. verticallata cultivars appeared greater than that of C. rosea cultivars, distinguished metabolites between the active and inactive groups were almost identical to metabolites that showed differences between the two Coreopsis species presented in Table 3 (C. verticallata: IC 50 < 65 µg/mL and C. rosea: IC 50 > 65 µg/mL). Notably, 'Orange sunlight (No. 30)' and 'Luckyten 6 (No. 2)', which are outliers of Group Cv and Group Cr, respectively, were found to have the greatest and lowest DPP-IV inhibitory activity, respectively. These results suggested that the composition and relative content of distinguishable markers between C. rosea and C. verticallata cultivars were evaluated as key markers for the classification of species and contribution to the correlation of active and inactive cultivars.

Plant Material
Coreopsis cultivars were grown and collected from a wild cultivation field at Uriseed Group, Icheon-si, Gyeonggi-do, Republic of Korea and authenticated by Yeo Gyeong Jeon and Kong Young Park. These cultivars were selected according to their diverse phenotypic variants and exhibited a stable inheritance of these phenotypes for 4 years. Among them, five γ-irradiated mutants (Redfin, Lemone candy, Shiny pink, Uri-dream red, pink sherbet) of the original cultivar (Heaven's gate) and the series of γ-irradiated mutants of original cultivars (Citirne, Pumpkin pie, Route 66) were generated using γ ( 60 Co) irradiation (150 TBq capacity; AECL, Ottawa, ON, Canada). Six other mutant cultivars of 'Heaven's gate' (Luckyten 6, Uri-dream 01, Luckyten5, Luckyten9, Uri-dream 07, Uri-dream 06) were artificially mutated using an herbicide. 'Moonlight sonata' was selected as the phenotypic variation of the original cultivar 'Moonbeam'. Flowers used in this study were handpicked at the flowering stage in August 2018. These flowers were freeze-dried and stored at −20 • C for further analysis. Voucher specimens were deposited at the Uriseed Group Corporation.

Sample Preparation
Freeze-dried flowers of Coreopsis cultivars were ground into powder using a mixer. Extractions were performed with 200 mg of this powder in 20 mL of 70% methanol using an ultrasonic bath for 60 min, and subsequently evaporated to achieve a dry product. Thereafter, these dried extracts (1 mg each) were dissolved in 1 mL of 70% methanol and filtered through a 0.20 µm polyvinylidene fluoride filter. Samples (1000 ppm) were diluted with 70% methanol to a concentration of 200 ppm for further liquid chromatography-mass spectrometry (LC-MS) analysis. For the evaluation of bioactivity, methanol extracts were initially dissolved in dimethyl sulfoxide (DMSO) at a concentration of 10 mg/mL stock solution. All extraction and chromatographic solvents used in this study were of analytical grade (J. T. Baker, Phillipsburg, NJ, USA).

DPP-IV Inhibitor Screening Assay
DPP-IV activity of Coreopsis cultivars was analyzed using a DPP-IV inhibitor screening assay kit (Cayman Chemical, Ann Arbor, MI, USA) which provides a fluorescence-based method for screening DPP-IV inhibitors. The assay uses the fluorogenic substrate, Gly-Pro-Aminomethylcoumarine (AMC), to measure DPP-IV activity. Cleavage of the peptide bond by DPP releases the free AMC group, resulting in fluorescence that can be analyzed using an excitation wavelength of 350-360 nm and an emission wavelength of 450-465 nm. The tested samples dissolved in DMSO at a concentration of 10 mg/mL were subsequently diluted to a final concentration of 20 to 200 µg/mL using DMSO and were added to a 96-well plate to a final volume of 10 µL and a final concentration of 50 µM. The assay procedure is described in our previous studies [12,62,63]. Briefly, diluted assay buffer (30 µL) and diluted enzyme solution (10 µL) were added to the 96-well plate containing 10 µL of solvent (blank) or solvent-dissolved test samples. The reaction was initiated by adding 50 µL of a diluted substrate solution, and the plate was incubated for 30 min at 37 • C. Following incubation, fluorescence with an excitation wavelength of 350 nm and an emission wavelength of 450 nm was monitored using a plate reader (TECAN, Männedorf, Switzerland). The percent inhibition was expressed as ([DPP-IV level of vehicle-treated control − DPP-IV level of test samples]/DPP-IV level of vehicle-treated control) × 100. Subsequently, the 50% inhibitory concentration (IC 50 ) was determined using GraphPad Prism software (GaraphPad Software, La Jolla, CA, USA) via dose-response analysis.

Chemometric Data Analysis
Data management for the UPLC-QTof-MS analysis was performed using UNIFI software (Waters, Milford, MA, USA). MS data were processed using UNIFI to obtain a data matrix containing retention times, accurate masses, and normalized peak intensities. Parameters included retention time (t R , range of 0.0-15.0 min), mass-to-charge ratio (m/z, range of 100-1500 Da), and a mass tolerance of 0.04 Da. The resulting data were evaluated using SIMCA 15.0.2 (Umetrics, Umeå, Sweden) for multivariate statistical analysis. Unsupervised principal component analysis (PCA) was performed using UV (univariate)-scaled and supervised orthogonal partial least-squares discriminant analysis (OPLS-DA) was used to identify and compare different metabolite sizes of the 32 samples. The quality of the OPLS-DA model was evaluated with R 2 Y value and cumulative Q 2 value. The model was further validated with a permutation test (n = 200). Markers for the difference between groups were identified by analyzing the S-plot with pareto scaling, which were generated with covariance (p) and correlation (pcorr) data. These data sets were normalized by dividing with mean value to get a value between 0 and 10 and a heatmap with dendrograms was generated using OriginPro 2021 (OriginLab Corporation, Northampton, MA, USA) selecting ward for cluster method. Marker compounds were tentatively identified by comparison to published MS data in literature and databases such as Waters Local Library in UNIFI and Massbank [3,5,6,[26][27][28][29][30][31].

Conclusions
To the best of our knowledge, a comparative metabolomics approach to identify metabolite composition and DPP-IV inhibitory activities in various cultivars of C. rosea and C. verticillata, were demonstrated for the first time in this study. UPLC-QTof-MS techniques were used to identify several phenolic acids, flavonoids and a polyacetylene in mutant cultivars compared to original cultivars. PCA and OPLS-DA results showed that metabolites discriminate between the mutant and original cultivars and between the two species. In addition, significant changes in metabolite content were observed under different DPP-IV inhibitory activities of cultivars, and chlorogenic acid, butin-7-O-glucoside, sulfuretin-6-O-glucoside, maritimein, 3,5-dicaffeoylquinic acid, coreopsin, luteolin, and butein were abundant in the active extracts. Therefore, the DPP-IV inhibitory cultivars and the metabolites influencing their activities would be favorable for the development of functional foods and the information of the metabolites accumulated differently for each mutant cultivar would be useful as a scientific reference for further studies on plant mutation mechanisms.