Comparative bioactivity evaluation and metabolic profiling of different parts of Duhaldea nervosa based on GC-MS and LC-MS

Duhaldea nervosa (Wallich ex Candolle) Anderberg has been widely used as medicine and food additive in China for a long history. Its roots, known as Xiaoheiyao, are the mainly used medicinal part, while the other tissues of D. nervosa are ignored as non-medicinal parts despite their high biomass, resulting in a huge waste of resources. To mine and expand the medicinal values of different parts of D. nervosa, metabolic analysis by GC/LC-MS and bioactivity evaluation were performed. Based on the antioxidant activity and correlation analysis, a metabolite-related network was constructed. A total of 45 volatile and 174 non-volatile compounds were identified. Among them, caffeoylquinic acids and derivatives were more abundant in roots and flowers, while coumaroyltartaric acids and derivatives were mainly present in stems and leaves. By multivariate analysis, 13 volatile and 37 non-volatile differential metabolites were found, respectively. In the bioactivity evaluation of different parts, the order of antioxidant capacity was flowers > roots > leaves or stems. The flowers showed the highest FRAP value (354.47 μM TE/g DW) and the lowest IC50 values in the DPPH (0.06 mg/mL) and ABTS (0.19 mg/mL) assay, while higher inhibitory activity against α-glucosidase was exhibited by flowers and leaves. This study first established the similarities and differences of phytochemicals and bioactivities in D. nervosa, providing a scientific basis for developing non-medicinal parts and guiding the clinical application of this medicinal and edible herb.


Introduction
Duhaldea nervosa (Wallich ex Candolle) Anderberg is a medicinal and edible herb of Asteraceae family, and is mainly distributed in the Southwest of China.The roots of D. nervosa, known as Xiaoheiyao, have been long used as a traditional Chinese medicine (TCM) to treat stomachache and relieve rheumatism (1).Furthermore, the roots of D. nervosa have also been popularly used in cooking as food additives, such as being stewed with chicken in a soup (it is believed to relieve dizziness), which was officially approved as a new food material by the Ministry of Health of PR China in 2010 (2).With the extensive development of new plant-derived functional foods and dietary supplements, many edible plant sources that are rich in antioxidants (such as phenolic compounds) have begun attracting the public attention (3).Studies have shown that natural plants are rich in antioxidants, which are effective, easy to be absorbed, and almost have no side effect.Therefore, screening antioxidants with potential therapeutic effects from natural plants has a promising prospect.
Previous studies have demonstrated that D. nervosa contains steroids, terpenoids, flavonoids, polysaccharides, and phenolic acids (4), which have a wide range of bioactivities including antiinflammatory, neuroprotective, antioxidant, hepatoprotective, and anticancer properties (5).Some studies have revealed that the root extracts protect HepG2 cells from H 2 O 2 -induced oxidative stress by increasing the expression of Nrf2 and related antioxidant enzymes (6).
However, the focus of previous studies was only on the roots of D. nervosa, while the medicinal potential of other parts was unknown.Consequently, to highlight the medicinal, nutritional and edible importance of this medicinal and edible herb, it is of great significance to further develop the application potential of D. nervosa and fully explore the components and bioactivity in different parts of D. nervosa.
Metabolomics is aimed at identifying differences in total metabolite fingerprints, which has significant advantages in the detection of chemical components that are significantly different from each other.The UPLC-Q-Orbitrap HRMS (ultra-high-pressure liquid chromatography coupled with Orbitrap high-resolution mass spectrometry) based metabolomics is featured by its high sensitivity, high resolution, and specificity, And has been widely used to identify the major different chemical components in various plant samples (7).To date, there is no report about the metabolite comparison of the different parts of D. nervosa.
In this study, the metabolic profiles of roots (the traditional medicinal part) and other non-medicinal parts (including flowers, stems and leaves) of D. nervosa were analyzed by gas chromatographymass spectrometer (GC-MS) and UPLC-Q-Orbitrap HRMS.The differential metabolites between the traditional medicinal part and other non-medicinal parts of D. nervosa based on chemometrics were then screened.In addition, the DPPH, ABTS, FRAP and α-glucosidase bioassay of different parts were performed, and by correlation analysis the relationship between differential metabolites and the biological effects of different parts were determined.These results provided guidance for clinical application and quality control of the waste parts of D. nervosa as potential medicinal sources.

Chemicals and reagents
Duhaldea nervosa (Wallich ex Candolle) Anderberg was collected in Panzhihua, Sichuan, China.The whole plant was dried and parted into roots, stems, leaves, and flowers, separately.

UPLC-Q-Orbitrap HRMS analysis 2.2.1 Sample preparation
All dried samples from different parts of D. nervosa were finely ground and sieved, yielding approximately 0.5 g of pulverized powder.Ultrasonic extraction was performed by immersing the powder in 25 mL of methanol/water (7:3, v/v) for 1 h (8).Subsequently, the extracts were centrifuged at 12,000 rpm for 10 min, and then the supernatant was filtered through a 0.22 μm membrane for the UPLC-Q-Orbitrap HRMS analysis.

UPLC-Q-Orbitrap HRMS
Vanquish UPLC system coupled with Q Exactive Orbitrap highresolution MS (Thermo Fisher Scientific, Waltham, United States) was used for metabolite analysis.Instrument and data acquisition were performed by Xcalibur 4.1 software.Sample separation was performed on a Thermo Scientific Accucore™ C 18 column (100 mm × 3 mm, 2.6 μm) at 25°C.The flow rate was 0.3 mL min −1 , and the injection volume was 3 μL.The multi-step gradient program was beneficial to improve the separation efficiency, and the elution condition was conducted according to the references with some changes (9), the comparisons of chromatograms obtained under different experimental conditions were shown in Supplementary Figure S1, we decided to use the elution condition of Supplementary Figure S1C because it has a better separation.The mobile phase consisted of 0.1% (v/v) formic acid in water (A) and acetonitrile (B) with a gradient programme: 0-12 min, 2-30% B; 12-25 min, 30% B; 25-35 min, 30-32% B; 35-40 min, 32-34% B; 40-45 min, 34-70% B; 45-50 min, 70-95% B, 50-55 min, 95-2%.The source parameters were set as follows: spray voltage, 3.5 kV (+) /3.0 kV (−); capillary temperature, 320°C; heater temperature, 350°C; sheath gas flow rate, 35 arb; aux gas flow rate, 10 arb.The Orbitrap analyser scanned over a mass-to-charge ratio (m/z) range of m/z 100 to 1,500 Da with a resolution of 35,000 in full scan MS 1 and a resolution of 17,500 in dd-MS2.The mixed normalised collision energy was set at 20, 40 and 60 V. Quality control (QC) samples were injected every 3 samples throughout the run to monitor system stability.

Identification of compounds
Using Compound Discover v3.1 software, the raw MS data collected by UPLC-Q-Orbitrap HRMS were first screened and combined with online databases including mzCloud and mzVault (Thermo Fisher Scientific, Waltham, United States) and self-built databases, and unknown metabolites were identified based on concordance between MS 1 and MS 2 .

Sample preparation
Powdered samples (5 g) were sonicated in 50 mL of n-hexane for 40 min, then the supernatant was dried under a nitrogen stream to

GC-MS conditions and compounds identification
Analysis was conducted on an Agilent Technologies 7890A GC system and an Agilent Technologies 5975C inert MSD equipped with triple-axis detector (Agilent, United States).Sample separation was operated on a HP-INNOWax capillary column (30 m × 250 μm × 0.25 μm, Agilent, United States).The helium flow rate was controlled at 1 mL/min.A 10°C/min ramp was set from an initial temperature of 50°C to 150°C, then to 180°C at 5°C/min, and then to 250°C at 3°C/ min, for a total of 50 min.A sample volume of 1 μL was injected at a split ratio of 10:1.Spectra were recorded in the full scan range (from 35 to 1,000 m/z) with the EI source of positive ion mode, source temperature of 230°C, and quadrupole temperature of 150°C.MS Workstation v6.9.3 (Agilent, United States) was used for instrument control and data processing.Compounds were searched and identified by the NIST14.L database (NIST, United States).

Statistical analysis
The UPLC-Q-Orbitrap HRMS data were analyzed using Compound Discover software, which produced a matrix of features containing MS, retention time and peak area through peak extraction, deconvolution, peak alignment and other operations.The data of GC-MS were processed by peak alignment and gap filling, and a feature matrix was generated through peak area normalization.The two matrices of features were then imported into SIMCA 14.1 (Umetrics, Sweden) for principal component analysis (PCA), hierarchical cluster analysis (HCA), and orthogonal partial least squares discriminant analysis (OPLS-DA).R2(cum) and Q2(cum) values were used to validate the model.R2 represents the ability to explain the original data, and Q2 represents the predictive ability of the model.Data were analyzed using a combination of variable importance in the projection values (VIP, VIP > 1), fold changes (FC, FC > 2 or FC < 0.5) and p-values (p < 0.05) from the t-test for potential differential metabolites.Heatmap visualization was performed using MetaboAnalyst 5.0.

Evaluation of bioactivity
The supernatants were diluted to different concentrations for the DPPH, ABTS, FRAP (10) and α-glucosidase (11) bioactivity assay with slight modifications based on literatures.All experiments were carried out independently and repeated more than three times, and the experimental data were expressed as mean ± standard deviation.

UPLC-Q-Orbitrap HRMS
Compounds were identified by self-built and online databases (mzCloud and mzVault).The self-built database of 169 compounds was constructed from the phytochemical and pharmacological literature by searching the SciFinder, ChemSpider, Google Scholar and CNKI databases, which contained detailed information such as compound name, molecular formula and structural formula.Moreover, compound identification was conducted using MS/MS fragmentation patterns compiled from reference material along with considerations of retention time, characteristic ions, signal intensity, and relevant literature.
The hydroxycinnamic acids (mainly including caffeoylquinic acids, coumaroyltartaric acids and their derivatives) were more easily to be detected in the negative ion mode, whereas the flavonoids were detected in the positive ion mode.A total of 174 compounds, including 54 caffeoylquinic acids and derivatives, 13 feruloylquinic acids and derivatives, 8 p-coumaroylquinic acids and derivatives, 6 caffeoyltartaric acids and derivatives, 29 other hydroxycinnamic acids and derivatives, 16 hydroxybenzoic acids derivatives, 36 flavonoids, 5 coumarins, 4 lignan derivatives, and 3 other compounds were identified from the roots, flowers, stems and leaves of D. nervosa.The total ion chromatograms (TICs) of samples are shown in Supplementary Figures S2, S3.The specific fragment information is shown in Table 1.

Hydroxycinnamic acids and derivatives
The hydroxycinnamic acids and derivatives are generally divided into four types, namely caffeoylquinic acids (CQAs), p-coumaroylquinic acids (p-CoQAs), feruloylquinic acids (FQAs), and caffeoyltartaric acids (CTAs), The structures and explanation of fragmentation behaviors of mass spectra were given in Figure 1, most of which were reported for the first time in the species.

Feruloylquinic acids and derivatives
Compounds 55-60, with a precursor ion at m/z 367, were assigned as mono-FQA.Compounds 55-60 were identified as follows: (1) compounds 55 and 56 were identified as 3-FQA by the base peak at m/z 193 and the "demethylated" ion at m/z 134 [feruloyl-H-CO 2 -CH 3 ] − ; (2) compounds 58 and 59 were identified as 5-FQA by the base peak at m/z 163; and (3) with the fragment ion at m/z 173 (shown as base peak), compounds 57 and 60 were identified as 4-FQA (31).The cis or trans configuration can be judged from the intensity of these peaks, and the stability of cis compounds is more intense (12).Compound 67 had the same quasi-molecular ion [M-H] − at m/z 529, which was 162 Da more than FQA, indicating that they were the hexosides of FQA.(16).Peak 62 was assigned as 3F,5CQA based on the base peak at m/z 193 and the abundant ions at m/z 367 [M-H-caffeoyl] − and 134 as recorded in 3-FQA (9).For 66, a vicinal 4C, FQA was deduced from the base peak at m/z 173, and the fragment ions at m/z 179 and 135 due to the loss of feruloyl.

p-Coumaroylquinic acids and derivatives
Compounds 68 and 69, with characteristic ions at m/z 337 [pCoQA-H] − and 163 [pCoA-H] − , were assigned as pCoQA, and the remaining ions were temporarily attributed to fragment series similar to the CQAs (12).
The fragment ion at m/z 661 of compound 75 was similar to the MS 2 spectrum of pCoCQA, so it was tentatively assigned as pCo-diCQA (9).

Flavonoids
The fragmentation features of flavonoids involved the unique neutral removal of acetyl (42 Da), methyl (15 Da), and dimethyl (28 Da) groups, as well as the loss of sugar moieties such as 162, 146, 176, 308, and 324 Da which were, respectively, corresponding to hexose, deoxyhexose, glucuronic acid, rutinoside, and dihexose.Fragment ions, resulting from the neutral losses of A − and 1,3 B − obtained by RDA fragmentation at m/z 151 and 135, they were tentatively assigned to marein (158) and acetylmarein (159).Hispidulin (161) was determined from the base peak at m/z 284.To the best of our knowledge, these flavonoids were described for the first time in D. nervosa.
Coumarins derivatives included compounds 163-167.The mass spectrometric fragmentation of coumarin in negative ion mode showed the loss of neutral molecules such as CO (28 Da), CO 2 (44 Da), and CH 3 (15 Da) due to high energy collisions.Take compounds 164 and 165 for example, which have the same molecular formula C 9 H 6 O 4 with the characteristic ion [M-H-CO-H 2 O] − at m/z 133 and [M-H-2CO] − at m/z 105, and both were tentatively identified as dihydroxycoumarin (18).All the coumarins and lignans were identified for the first time in the species.

GC-MS analysis
Based on the NIST database and comparison with literature, the GC-MS data were analyzed and identified.There were 45 compounds with match degree greater than 80, the identified compounds could be classified into 9 types, including 11 monoterpenes, 10 sesquiterpenes, 2 diterpenes, 1 triterpene, 6 alkanes, 1 olefin, 12 fatty acids and their esters, 1 ketone, and 1 amide.A total list of compounds, The RDA cleavage mechanisms of the associated flavonoids.The relative peak areas of the metabolites were used to construct stack bar graphs of the distribution of metabolites in different parts of D. nervosa, as shown in Figure 3. Thymol-based monoterpenes were designated as the main volatile components in roots, flowers, stems, and leaves, with percentages of 58.47, 36.59,75.00, and 60.52%, respectively.Moreover, the content of fatty acids and their esters in each part is relatively rich.The identified non-volatile compounds were mainly dominated by hydroxycinnamic acids (main including caffeoylquinic acids, coumaroyltartaric acids and their derivatives) in the extracts of different parts, and the contents of the identified total hydroxycinnamic acid and derivatives were roots (85.95%), flowers (70.06%), stems (79.80%) and leaves (77.07%).It is worth noticing that flavonoids were present at low levels in the roots, while hydroxybenzoic acids were more abundant.Flower and roots were richer in caffeoylquinic acids.The content of different parts may be related to the biosynthesis and photosynthesis during flowering, which may affect the synthesis of large amounts of polyphenols (35).

Comparative chemical profiling of different parts of Duhaldea nervosa
Principal component analysis (PCA) was conducted to classify the different parts of LC-MS (R2X = 0.883) and GC-MS (R2X = 0.964).The samples that fall within 95% of the Hotelling T 2 ellipse and have no outliers are divided into four groups.As shown in Figures 4B, 5B, the PCA score plots demonstrated that the chemical profile of roots and flowers was significantly different from that of leaves, while the chemical profile of stems was similar to that of leaves.It is significantly indicated by the difference between the medicinal and non-medicinal parts of D. nervosa.Organ influence on chemical profiles was more pronounced in roots and flowers, as they showed greater chemical differences.The hierarchical clustering heat map intuitively visualized the degree of difference between chemical profiles in different parts (Figures 4A, 5A).The result of the HCA analysis also clarified it (Figures 4C, 5C), and the dendrogram showed three clusters.They were divided into the flowers cluster, the roots cluster, and finally into two subclusters, one representing the stems and one representing the leaves.
To further clarify the variations between different parts, the non-medicinal parts were compared with roots for OPLS-DA analysis to find marker compounds representing the difference between groups, and heat maps were generated from the relative peak area of different metabolites to visualize the differences in abundance between different parts of D. nervosa.
Heatmaps were generated from non-repetitive differential compounds (Figure 6A), consisting of 3 monoterpenes, 1 sesquiterpene, and 10 fatty acid and its esters, which were among the 13 differential metabolites identified in various parts.Specifically, the R/F comparative group had 3 differential metabolites, R/S had 2, and R/L had 8 (Supplementary Table S2).The monoterpenes were mainly distributed in stems, while the fatty acids were highly expressed in roots and flowers.
A total of 74 differential metabolites were identified using UPLC-Q-Orbitrap HRMS.The result included 23 differential metabolites in the R/F comparative group, 19 in R/S and 18 in R/L (Supplementary Table S3; Figures 7A-F).Thirty-two were non-repetitive differential compounds, including 13 caffeoylquinic acids, 3 feruloylquinic acids, 4 flavonoids, 4 hydroxybenzoic acids, 2 lignans, 1 coumarin, 9 other hydroxycinnamic acids and 1 other.Among them, 37 candidate marker compounds were screened out, which could suggest a remarkable discrimination capability between different parts of D. nervosa (Supplementary Table S3 and Figure 6B).

Biological activity of different parts
Excessive production of free radicals can result in oxidative stress, which is strongly linked to the onset of chronic inflammation and degenerative disorders such as cardiovascular diseases, cancers, and autoimmune diseases.Conversely, dietary antioxidants have the potential to diminish the likelihood of encountering these ailments (36).The present study suggested that extracts of D. nervosa may serve as natural dietary antioxidants for health promotion, and the    roots, flowers, stems and leaves may have significantly different radical scavenging properties.The DPPH, ABTS, and FRAP assays were utilized to assess the antioxidant activity of various parts.Significant variations were observed between the medicinal and non-medicinal sections, with the flowers and roots displaying the highest scavenging activity, followed by the stems, while the leaves exhibited the lowest activity (Table 2).The activity of DPPH was stems < leaves < roots< flowers, and the rise of ABTS and FRAP values were leaves < stems < roots< flowers, which may be explained by the different reaction mechanisms of the ABTS, DPPH, and FRAP.These findings indicated that the changes in biological activity were pronounced among various plant parts.These variations in bioactivity can be attributed to the differences in the chemical composition of the extracts obtained from different parts.Multivariable models can effectively demonstrate the predictive potential of metabolomic analysis for antioxidant activity and determine which group of metabolites is most responsible for this activity (37).To understand the correlation of constituents, samples were submitted to Pearson's correlation analysis (Figure 8A).The IC 50 values of DPPH, ABTS, and FRAP scavenging activities showed a mainly positive correlation with caffeoylquinic acids, feruloylquinic acids, p-coumaroylquinic acids, and their derivatives (since IC 50 values are inverse to antioxidant levels).Furthermore, Pearson's correlation analysis was used to show the contribution of the main differential metabolites (including diCQA and triCQA) to the antioxidant activity (Figure 8B).CQAs have gained recognition for their ability to perform as antioxidants, reduce inflammation and prevent diabetes.Previous research has shown that the unique molecular structures of these compounds, namely the presence of five active hydroxyl groups and one carboxyl group, contributed to their natural antioxidant properties.The phenolic hydroxyl structure readily reacted with free radicals to produce antioxidant hydrogen radicals that effectively scavenge hydroxyl radicals and superoxide anions, demonstrating their potent antioxidant activity (38).The contents of caffeoylquinic acids and flavonoids were more abundant in the flowers (Figure 3), which might explain the better antioxidant activity of the flowers.In addition, hydroxyl (OH) has a positive effect on the antioxidant properties of phenolic acids, and the relationship between the antioxidant properties of phenolic acids and the number of hydroxyl groups was listed as follows: trihydric phenolic acid > dihydroxyl > mono-hydroxyl (39), and the highest level of correlation between the main metabolites and the antioxidant activity was observed in triCQAs (Figure 6B), which were the most abundant in the flowers and roots.
α-Glucosidase inhibitors slowed the release of glucose from dietary carbohydrates, helping to lower postprandial blood glucose levels, and slow the development of diabetes (40).In our work, the hypoglycaemic activities of different parts were assessed by α-glucosidase inhibition assay.It was found that the flowers and leaves extracts exhibited more potent inhibitory activity than roots extracts (as medicinal parts) (Table 2).Particularly, apart from the quinic acids, the IC 50 values of α-glucosidase inhibition showed a positive relationship with the amount of flavonoids.Flowers and leaves are fragments at m/z 135 ([CA-H-CO 2 ] − ), and 161([CA-H-H 2 O] − ), and at least one quinic acid fragment at m/z 191 ([QA-H] − ), and 173 ([QA-H-H 2 O] − ).

FIGURE 1
FIGURE 1The possible fragmentation pathway of main hydroxycinnamic acids and derivatives.

FIGURE 3
FIGURE 3 Distribution of the components in different parts of D. nervosa.(A) Volatile components.(B) Non-volatile components.

FIGURE 4
FIGURE 4 Chemometric analysis for discrimination of non-volatile compounds in different parts (roots, flowers, stems, leaves) of D. nervosa.(A) Pearson's rank correlation coefficient of different parts.(B) Discriminated by PCA.(C) Discriminated by HCA.

FIGURE 5
FIGURE 5 Chemometric analysis for discrimination of volatile compounds in different parts (roots, flowers, stems, leaves) of D. nervosa.(A) Pearson's rank correlation coefficient of different parts.(B) Discriminated by PCA.(C) Discriminated by HCA.

FIGURE 6
FIGURE 6 Heatmap of the identified differential metabolites in the volatile compounds (A) and non-volatile compounds (B).Heatmap generated by hierarchical Pearson clustering of different metabolites in four subsamples based on mean values, rows represent different metabolites and columns represent samples.Color comparison plot based on relative amounts of metabolites.

FIGURE 7 (
FIGURE 7 (A-C) The OPLS-DA plot between different comparison groups.(D-F) The Volcano plots of metabolites of different comparison groups.
filtered through a 0.22 μm membrane to obtain the sample solutions.All solutions were stored at 4°C prior to analysis.

TABLE 2
Biological activity of different parts, results expressed in crude drug weight (DW) concentration.