Phytochemistry and Antioxidant Activities of the Rhizome and Radix of Millettia speciosa Based on UHPLC-Q-Exactive Orbitrap-MS

The root of Millettia speciosa Champ. (MSCP) is used in folk medicine and is popular as a soup ingredient. The root is composed of the rhizome and radix, but only the radix has been used as a food. Thus, it is very important to compare the chemical components and antioxidant activities between the rhizome and radix. The extracts were analyzed by UHPLC-Q-Exactive Orbitrap-MS and multivariate analysis, and the antioxidant activities were evaluated by 2,20-azinobis (3-ethylbenzothiazo-line-6-sulfonic acid) diammonium salt (ABTS) and 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging assays. Ninety-one compounds were detected simultaneously and temporarily identified. Ten compounds were identified as chemical markers to distinguish the rhizome from the radix. The antioxidant activities of the radix were higher than the rhizome. Correlation analysis showed that uvaol-3-caffeate, 3-O-caffeoyloleanolic acid, and khrinone E were the main active markers for antioxidant activity, which allowed for the rapid differentiation of rhizomes and the radix. Therefore, it could be helpful for future exploration of its material base and bioactive mechanism. In addition, it would be considered to be used as a new method for the quality control of M. speciosa.


Introduction
Millettia speciosa Champ. belongs to the Leguminosae family. It is a well-known food and medicine ingredient that is mainly found in Guangxi, Guangdong, and Hainan Provinces of China [1]. MSCP is called Niudali in southern China, which means that it has a strong power, like a bull. The MSCP is commonly used to make soups with pig bones, which can strengthen the functions of the immune system and promote anti-inflammatory and anti-tumor effects. It is also made into various functional products, such as MSCP powder, MSCP tea, and MSCP wine. In previous studies, the chemical constituents of MSCP mainly include polysaccharides, flavonoids, alkaloids, terpenoids, and phenylpropanoids [2][3][4]. Polysaccharides are major bioactive components in the aqueous extracts of MSCP [5,6]. Pharmacological studies showed that the aqueous extracts of MSCP exhibited antifatigue, immunomodulatory, antioxidative, anti-hepatitis, and analgesic activities [3][4][5][6][7].
However, the root of MSCP consists of the radix and rhizome. Usually, the radix is swollen and powdery, commonly known as Niudali potato, whereas the rhizome is almost lignified and fibrous ( Figure 1). Only the radix has been traditionally consumed [8]. Moreover, the price of the radix is higher than that of the rhizome in herbal markets. Therefore, the rhizome is often mixed into the radix before sale, which greatly affects the quality of the radix. Previous studies have mainly focused on the chemical composition and activity of MSCP [1,6,7,9]. A comparative analysis of alcohol soluble components Ultra-high performance liquid chromatography coupled with the quadrupole time of flight mass spectrometry (UHPLC-QTOF-MS/MS) has been widely used in the separation and structural analysis of complex systems, such as traditional Chinese medicines, due to its rapid separation, high efficiency, high sensitivity, high resolution, and molecular weight accuracy [12]. It can be widely used for the chemical identification of different base sources and different medicinal parts of herbal medicines [13][14][15]. Recently, UHPLC-Q-Exactive Orbitrap MS, with a high selectivity and effectivity, was widely used in the identification of chemical constitutions In this study, the characteristic chemical components in twenty-eight batches of MSCP were qualitatively analyzed by UHPLC-Q-Exactive Orbitrap-MS. Moreover, principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were used to distinguish the differences between the rhizome and the radix of MSCP. In addition, the antioxidant activities were evaluated by the scavenging rate of ABTS and DPPH. Gray relational analysis (GRA) and partial least squares (PLS) were used to identify potential bioactive markers and to assess the correlation between the characteristic compounds and antioxidant activities.

Optimization of Chromatographic Conditions
In our previous study, we compared the effectiveness of the two columns in the analysis of MSCP samples and found that the ACQUITY HSS T 3 column (100 mm × 2.1 mm, 1.8 µm) was more suitable for the analysis of MSCP because of the greater number of compounds detected and separated. The chromatographic separation was optimized by investigating parameters, such as organic phases (acetonitrile and methanol), with mobile phases (water, water-containing formic acid, and water-containing formic acid and ammonium), and temperatures (20,25,30,35, and 40 • C). The optimal mobile phase consisted of 0.1% aqueous formic acid-acetonitrile with 35 • C, due to a superior peak pattern, better response value, and peak shapes. The base peak ions (BPIs) of rhizome and radix samples obtained in the negative ion mode are shown in Figure 2.

Optimization of Chromatographic Conditions
In our previous study, we compared the effectiveness of the two columns in the analysis of MSCP samples and found that the ACQUITY HSS T 3 column (100 mm × 2.1 mm, 1.8 μm) was more suitable for the analysis of MSCP because of the greater number of compounds detected and separated. The chromatographic separation was optimized by investigating parameters, such as organic phases (acetonitrile and methanol), with mobile phases (water, water-containing formic acid, and water-containing formic acid and ammonium), and temperatures (20,25,30,35, and 40 °C). The optimal mobile phase consisted of 0.1% aqueous formic acid-acetonitrile with 35 °C, due to a superior peak pattern, better response value, and peak shapes. The base peak ions (BPIs) of rhizome and radix samples obtained in the negative ion mode are shown in Figure 2.

Identification of Components in the Rhizome and Radix Based on UHPLC-Q-Exactive Orbitrap MS
Under the optimal chromatographic and MS conditions, 91 compounds were unambiguously identified, including 71 flavones, 10 phenolic acids, 7 triterpenes, 3 alkaloids, and 10 fatty acid derivatives. Among them, 7 compounds were identified with corresponding reference standards (peaks 10, 28, 40, 55, 70, 72, and 81); however, the remaining 84 compounds were tentatively assigned in accordance with previously published MS data and literature. The retention times, MS data, and fragment ions of all detected compounds are shown in Table 1.     The flavonoid compounds reported in M. speciosa were mainly divided into several types, including flavones, isoflavones, flavanones, and chalcones. The fragmentation pathways of flavonoids were followed by Retro-Diels-Alder (RDA) [16]. In this study, a total of 30 flavones, isoflavones, and glycosides were identified. O] + , and it yielded RDA fragment ions at m/z 137 ( 1,3 A + ) in positive mode, indicating that one hydroxy group was attached to ring A and one hydroxy group and one methoxy group were attached to ring B. Compound 37 was tentatively identified as 5,4 -dihydroxy-3 -methoxy-isoflavone [1]. Compounds 41, 45, 60, and 65 were tentatively identified as isoprunetin, 2 -hydroxyformononetin, calycosin, and glycitein, respectively, based on data reported in the literature [9] and the RDA fragmentation pathway. Similarly, compound 61 was deduced to be 3 ,4 -dihydroxy-7-methoxyisoflavone [9]. therefore, it was tentatively identified as calycosin-7-O-beta-D-glucoside [19]. However, compound 35 gave characteristic fragment ions at m/z 151 ( 1,3 A − ) and 132 ( 1,3 B − ) by the C-ring RDA fragment, which showed that two hydroxy groups were attached to the Aring and one methoxy group was attached to the B-ring. Compound 35 was tentatively identified as sissotrin. Similarly, according to the reported literature and MS data in Table 1, compound 31 was tentatively identified as yuankanin [20].
Compounds 48, 54, 58, 73, and 74 produced an [M-H] − ion at m/z 297.07 with the same formula of C 17 H 14 O 5 . Their MS 2 spectra produced fragment ions at m/z 282,269, indicating that there were two methoxy groups. Compounds 48, 54, 58, and 73 yielded fragment ions at m/z 167 ( 1,3 A − +H) and 132 (1,3B − ) by RDA reactions, indicating that one methoxy group and hydroxy group were assumed to attach to ring A and one methoxy group was assumed to attach to ring B. Compared with MS data and literature data [1,9], compounds 48, 54, 58, and 73 were tentatively determined to be 8-O-methylretusin, afrormosin, alfalfa, and 7-O-methylbiochanin A, respectively. Compound 74 generated the RDA fragment ion at m/z 135 ( 1,3 A − ), indicating that one hydroxy group was attached to ring A. Compound 74 was tentatively identified as cladactin [9].
Compounds 49 and 72 exhibited a molecular ion at m/z 269.0799 [M+H] + with the same formula of C 16 H 12 O 4 . Compound 49 was observed at m/z 254.0555, with the loss of a CH 3 (−15Da). The fragment ions at m/z 151.0413 ( 1,3 A + ) and 118.0411 ( 1,3 B + ) were generated by the RDA reaction, which indicated that one methoxy group was on ring A and one hydroxy group was on ring B. Compound 49 was tentatively identified as 4 -hydroxy-7-methoxyisoflavone [1]. Compared with the reference standard, compound 72 was confirmed as formononetin [1,9].
Compounds 52, 59, and 62 displayed an [M+H] + ion at m/z 301 and a product ion at m/z 286, indicating the loss of CH 3 (−15Da). Compounds 52 and 62 produced RDA fragment ions at m/z 153 ( 1,3 A + ) and 134 ( 1,3 B + -CH 3 ) in positive mode, indicating that two hydroxy groups were attached to ring A and hydroxy and methoxy groups were attached to ring B. Compounds 59 and 62 produced a fragment ion at m/z 286 [M-H-CH 4 ] − , which demonstrated that the position of methoxy is ortho to hydroxy. Therefore, compounds 52 and 62 were tentatively assigned as 2 -hydroxybiochanin A [1] and pratensein [9], respectively. Compound 59 exhibited a fragmentation ion at m/z 93, with the loss of ring B by the breakage of the C3-C9 bond. This indicated that ring B was substituted by one hydroxy group. Compound 59 was tentatively identified as tectorigenin [9,21].

Identification of Flavonone, Isoflavonone, and Flavonone Glycosides
As shown in Table 1

Identification of Pterocarpans and Other Flavonoids
Usually, pterocarpan cannot produce RDA reactions because of its tight structure. The proposed fragmentation pathways of pterocarpan compounds are summarized in Figure 3A. The cleavage of the C-C bonds of pterocarpan compounds mainly occurred on the side of C 6a and C 11a to generate ion fragments a, b, and c.  (Figure 3B), indicating that ring I was substituted by one hydroxy group and ring IV was substituted by one methoxy group. The hypothesized fragmentation pattern of compound 50 is demonstrated in Figure 3C. Thus, compound 50 was tentatively identified as medicarpin.

Identification of Pterocarpans and Other Flavonoids
Usually, pterocarpan cannot produce RDA reactions because of its tight structure. The proposed fragmentation pathways of pterocarpan compounds are summarized in Figure 3A. The cleavage of the C-C bonds of pterocarpan compounds mainly occurred on the side of C6a and C11a to generate ion fragments a, b, and c.  (Figure 3B), indicating that ring I was substituted by one hydroxy group and ring IV was substituted by one methoxy group. The hypothesized fragmentation pattern of compound 50 is demonstrated in Figure 3C. Thus, compound 50 was tentatively identified as medicarpin.   Figure 4. Thus, compound 33 was tentatively identified as sulfuretin, belonging to aurone.

Identification of Alkaloids
Alkaloid compounds tend to produce signals in negative ion mode. In this work, three alkaloid compounds were rapidly identified. The fragment ion at m/z 163 was obtained, and compound 89 was tentatively assigned as 2,2 -methylenebis.

Discrimination of Chemical Profiles of Rhizome and Radix
As shown in Figure 2, a total of 91 chemical constituents were identified in the rhizome and radix samples. It is worth noting that all of the intensities of the chemical constituent peaks of the rhizome were stronger than those of the radix under the same analytical conditions. In total, the intensities of the peaks were comparatively low, within 6.2-8, 8.5-10, and 13.5-14 min in the radix, and the intensities of peaks 5, 9, 13-15, 17, 18, 20, 21, 23, 24, 28, 33-35, 43, 44, 48, and 49 were almost undetectable in the radix samples. To establish representative chromatographic fingerprints, the established method was applied to analyze 16 batches of rhizome and 12 batches of radix. There were 91 common compounds identified in the MS spectra fingerprints from the rhizome and radix. However, considering many common peaks and a large amount of peak area data, it was difficult to distinguish rhizome and radix samples using this information. Multivariate analysis, including PCA and OPLS-DA, was performed on mass spectral data sets using SIMCA-P14.1 software. The loading plot from OPLS-DA together with the variable importance in the projection (VIP) were applied to reveal potential markers [14][15][16]. Thus, it is necessary to use multivariate analysis to reduce the dimensionality of the primal data.

Principal Components Analysis (PCA)
To efficiently visualize the differences between 16 batches of rhizomes and 12 batches of radix from MSCP, PCA was applied to analyze the MS spectral data using SIMCA-P14.1 software. The parameters R 2 (cum) and Q 2 (cum) are generally used to explain the quality and reliability of the models. The PCA results are presented in two score plots ( Figure 5A), and 28 batches of MSCP samples were divided into two types containing rhizome or radix (R 2 (cum) = 0.235 and Q 2 (cum) = 0.377). The samples from different parts were clearly distinguished into two groups according to the PCA model. As shown in Figure 5A, the radix (A1-A16) of MSCP is displayed on the left side of the score plot, whereas rhizome (B1-A12) is displayed on the right side of the plot, indicating a difference between the rhizome and radix in terms of chemical composition. Therefore, it is possible to distinguish the rhizome and radix samples based on UHPLC-Q-Exactive Orbitrap MS fingerprint analysis with PCA. These results revealed the distinctive differences in chemical composition between the rhizome and radix samples of MSCP.

Chemical Markers to Distinguish the Rhizome from Radix with OPLS-DA
To identify chemical markers unique to the rhizome and radix samples, OPLS-DA was utilized to further process the MS spectral data by SIMCA-P14.1 software. The loading S-plot was used to characterize the chemical difference between the rhizome and the radix. In Figure 5B, the spots located at the end of the plot indicate that the contribution of the variable to the differentiation is higher. In this S-plot, every spot represents an ion tR-m/z pair. The x-axis presents the alterable contributions of the variables, whereas the yaxis presents the alterable confidence levels of the variables. Therefore, when the distance between the ion tR-m/z pair spots and zero increases, the confidence level of the difference between the rhizome and radix also increases. Thus, those spots located at the end of the plot were tentatively regarded as potential chemical markers, leading to differences between the rhizome and radix samples.
In addition, the VIP value was employed to confirm the potential markers, which represent the differentiation between the rhizome and radix. When the VIP is higher, the variables are more important to the model. The S-plot, together with the variables of VIP ≥ 1, suggests the influence of the differentiation of samples using OPLS-DA.

Chemical Markers to Distinguish the Rhizome from Radix with OPLS-DA
To identify chemical markers unique to the rhizome and radix samples, OPLS-DA was utilized to further process the MS spectral data by SIMCA-P14.1 software. The loading S-plot was used to characterize the chemical difference between the rhizome and the radix. In Figure 5B, the spots located at the end of the plot indicate that the contribution of the variable to the differentiation is higher. In this S-plot, every spot represents an ion t R -m/z pair. The x-axis presents the alterable contributions of the variables, whereas the y-axis presents the alterable confidence levels of the variables. Therefore, when the distance between the ion t R -m/z pair spots and zero increases, the confidence level of the difference between the rhizome and radix also increases. Thus, those spots located at the end of the plot were tentatively regarded as potential chemical markers, leading to differences between the rhizome and radix samples.
In addition, the VIP value was employed to confirm the potential markers, which represent the differentiation between the rhizome and radix. When the VIP is higher, the variables are more important to the model. The S-plot, together with the variables of VIP ≥1, suggests the influence of the differentiation of samples using OPLS-DA.

Antioxidant Activity Test Results
The UHPLC-Q-Exactive Orbitrap MS results show that MSCP extracts mainly contained flavonoids, phenolic acids, and steroid saponins. Previous studies have confirmed that plants with antioxidant activity are closely related to their chemical compounds, such as flavonoids, phenolic acids, and steroid saponins [28]. Antioxidants are closely associated with human health. The damaging effects of free radicals can directly or indirectly lead to diseases and cancers [29]. The antioxidant activities of different extracts between the rhizome and radix of MSCP were determined by ABTS and DPPH antioxidant assays. As shown in Table 2, different batches of MSCP showed different antioxidant abilities, compared with the IC 50 values. In addition, a significant difference in the IC 50 values between rhizomes and radixes was observed using ABTS and DPPH antioxidant tests. The IC 50 values of the rhizome and radix extracts were 5.05-10.13 µg/mL and 2.09-4.56 µg/mL in the ABTS test, respectively. The IC 50 values of the rhizome and radix extracts were 5.86-11.86 µg/mL and 2.76-5.77 µg/mL in the DPPH test, respectively. The lower IC 50 values indicate a higher antioxidant activity [30]. Therefore, in both the ABTS and DPPH antioxidant assays, the antioxidant activity of radix samples was obviously higher than that of rhizome samples. The results indicate that both rhizome and radix extracts of MSCP presented remarkable antioxidant activities in vitro. However, it is still unknown whether the variability in the chemical components is related to differences in antioxidant activity and whether the variability influences the ability to discriminate between the rhizomes and radix. Therefore, further study is necessary to reveal the relationship between chemical components and antioxidant efficacy via statistical analysis.  PLS analysis was further used to predict the correlation between the characteristic chemical components and antioxidant activity. Figure 6 was drawn to reflect the correlation between 10 peaks and antioxidant activity. The regression equations obtained by the PLS model are as follows: Molecules 2022, 27, x FOR PEER REVIEW 19 of 23 Y2(DPPH) = −0.039Xa + 0.065Xb + 0.056Xc + 0.066Xd + 0.056Xe + 0.067Xf + 0.058Xg + 0.037Xh − 0.014Xi + 0.053Xj (2) As shown in Figure 6A and Equation (1), the areas of peaks b, d, e, f, g, i, and j showed a clear positive correlation with ABTS antioxidant activity, whereas peaks a, c, and h were negatively correlated with antioxidant activity. Figure 6B and Equation (2) show that the areas of peaks b, c, d, e, f, g, h, and j are positively correlated with DPPH antioxidant activity, whereas peaks a and i are negatively correlated with antioxidant activity. Therefore, 2,2′-methylenebis (peak 89), uvaol-3-caffeate (peak 91), 3-O-caffeoyloleanolic acid (peak 88), khrinone E (peak 76), 3-β-O-trans-caffeoyl betulinic acid (peak 87), and 27-pcoumaroyloxyursolic acid (peak 86) represent compounds that significantly contribute to the pharmacological effects of MSCP. Khrinone E (peak 76) showed the highest regression coefficient and was considered to have the highest contribution to antioxidant activity.
Therefore, correlation analysis shows that uvaol-3-caffeate (peak 91), 3-O-caffeoyloleanolic acid (peak 88), and khrinone E (peak 76) are the main active markers for the antioxidant activity of MSCP. As shown in Figure 6A and Equation (1), the areas of peaks b, d, e, f, g, i, and j showed a clear positive correlation with ABTS antioxidant activity, whereas peaks a, c, and h were negatively correlated with antioxidant activity. Figure 6B and Equation (2) show that the areas of peaks b, c, d, e, f, g, h, and j are positively correlated with DPPH antioxidant activity, whereas peaks a and i are negatively correlated with antioxidant activity. Therefore, 2,2 -methylenebis (peak 89), uvaol-3-caffeate (peak 91), 3-O-caffeoyloleanolic acid (peak 88), khrinone E (peak 76), 3-β-O-trans-caffeoyl betulinic acid (peak 87), and 27-p-coumaroyloxyursolic acid (peak 86) represent compounds that significantly contribute to the pharmacological effects of MSCP. Khrinone E (peak 76) showed the highest regression coefficient and was considered to have the highest contribution to antioxidant activity.

Reagents and Chemicals
HPLC-grade acetonitrile and methanol were obtained from Fisher Scientific Co. (Loughborough, UK). Deionized water was used. High-purity (>98.0%) hypaphorine, salicylic acid, liquiritigenin, naringin, isoliquiritigenin, formononetin, and maackiain were purchased from the National Institutes for Food and Drug Control (Beijing, China). All other chemicals were of analytical grade. The ABTS and DPPH were purchased from Sigma-Aldrich (St. Louis, MO, USA).

Sample Collection
The MSCP was collected from different locations in Guangxi Province and authenticated by Professor Zhifeng Zhang (Institute of Qinghai-Tibetan Plateau, Southwest Minzu University). The sample information is shown in Table 2. The fresh sample was sectioned and dried in the sun. The samples were kept in the herbarium of Qin Zhou Provincial Health School (Qinzhou, China).

Preparation of Standard and Sample Solutions
Standard stock solutions of hypaphorine, salicylic acid, liquiritigenin, naringin, isoliquiritigenin, formononetin, and maackiain were prepared in methanol/water (50% v/v), and the final concentration was 0.1 mg/mL. The stock solutions were further diluted and stored in a refrigerator at −20 • C until UHPLC-Q-Exactive Orbitrap MS analysis.
The dried sample was ground to a powder and saved in desiccators at normal temperature for future use. Subsequently, the sample powder (0.3 g) was extracted with 10 mL of 70% methanol in an ultrasonic water bath for 30 min at room temperature. The solution was filtered through a 0.22 µm microfiltration membrane.

Instrumentation and Chromatographic Conditions
The solutions were measured using a Thermo Scientific™ Vanquish™ Flex UHPLC (Thermo Fisher Scientific Inc., Waltham, MA, USA) equipped with a binary solvent system, autosampler, and full UV wavelength spectrophotometer. The chromatographic conditions were set and modified according to our previous work [32]. The chromatographic separation was performed with an ACQUITY HSS T 3 column (100 mm × 2.1 mm, 1.8 µm). Mass spectrometry was performed on a Thermo UHPLC-Q-Exactive Orbitrap mass spectrometer equipped with an electrospray ionization (ESI) source. Both positive and negative ionization modes were applied to acquire a scanning range from 100 to 1000 Da with a scanning time of 0.2 s and a 30 min detection period. The MS parameters were set as follows: the capillary voltage was set to 3.5 kV (positive mode) and 3.2 kV (negative mode); the source and desolvation temperatures were 100 and 350 • C, respectively; the drying gas flow rate was 600 L/h; and the cone flow rate was 50 L/h. Finally, processing and analysis of the data were carried out using Xcalibur 2.1 software (Thermo Fisher Scientific, Bremen, Germany).

ABTS Activity Assay
The ABTS assay was performed according to a previously described study by Wang et al. [11] with a few modifications. The ABTS aqueous solution (7 mM) was mixed with K 2 S 2 O 8 (2.45 mM) and protected from light at room temperature for 12 h. The configured ABTS + solution was diluted with anhydrous ethanol, and an absorbance of 0.70 ± 0.02 was measured at 734 nm for ABTS + analysis. Five different concentrations of sample solutions were prepared by diluting with anhydrous ethanol. Then, 0.4 mL of the sample extract was mixed with 4 mL of the diluted ABTS + solution at 25 • C for exactly 5 min. The absorbance was determined at 734 nm. The ABTS activity was calculated using the following Equation.
ABTS activity (%) = (1 − A sample /A blank ) × 100% (3) where A sample is the absorbance of 0.4 mL of sample extract combined with 4 mL of the diluted ABTS + solution, and A blank is 0.4 mL of anhydrous ethanol combined with 4 mL of the diluted ABTS + solution. The half inhibition concentration (IC 50 ) value was evaluated using a regression equation from serial concentrations of the sample extract.

DPPH Activity Assay
The DPPH scavenging capacity was determined as previously described by Wang et al. [11] with a small modification. The extracts were concentrated to five different concentrations by distillation to prepare the sample solutions. A 4 mL DPPH (0.04 mg/mL) solution was mixed with 1 mL of the sample and kept in the dark at 25 • C for 30 min. The absorbance was recorded at 517 nm, and DPPH activity was calculated via the following Equation.
DPPH activity (%) = (1 − A sample /A blank ) × 100% (4) where A sample is the absorbance of 1 mL of sample extract combined with 4 mL of the diluted DPPH solution, and A blank is 1 mL of anhydrous ethanol combined with 4 mL of the diluted DPPH solution. Regression analysis of the data was used to estimate the IC 50 values.

Statistical Analysis
The processed data with accurate mass were exported from Xcalibur 2.1 software. The match factor and retention time window of the peaks were set to 0.3 ppm and 0.05 min, respectively.
The spectral information was imported into SIMCA software (SIMCA-P 14.1, Umetrics Inc., Umea, Sweden) for multivariate statistical analysis. Principal component analysis (PCA) and orthogonal partial least-squares discriminate analysis (OPLS-DA) were used to distinguish between rhizome and radix samples. The S-plot and the importance in the projection (VIP) were used for predicting potentially characteristic chemical compounds between the rhizome and radix. Molecular formulas were determined using fragment information obtained from the workstation, as well as the established database and information from the literature. The cleavage process of each compound was predicted based on the MS and MS/MS information. Data were compared by one-way ANOVA followed by Turkey's test for multiple comparisons. A value of p < 0.05 was considered to represent a significant difference.

Gray Relational Analysis
GRA was performed using the DPS software (DPS 9.5, China) to calculate the correlation degree between the characteristic chemical compound peak area of MSCP and the antioxidant activities (ABST and DPPH radical scavenging activity). The gray relational grade was calculated with a distinguishing coefficient of 0.5.

Partial Least Squares Analysis
PLS analysis was performed using SIMCA software (SIMCA-P 14.1, Umetrics Inc., Umea, Sweden) for regression analysis between the peak area of characteristic components and the antioxidant activities. In the PLS model, the areas of characteristic component peaks were the independent variables (X), and the scavenging activities of ABTS and DPPH were the dependent variables (Y).

Conclusions
In this study, the UHPLC-Q-Exactive Orbitrap MS with multivariate analysis approach was established to identify and compare the chemical profiles of rhizome and radix MSCP. A total of 91 compounds were tentatively identified, and 10 compounds were selected as quality control markers to distinguish them. Additionally, the ABTS and DPPH assays were used to evaluate their antioxidant efficiency. GRA and PLS analysis indicated that uvaol-3-caffeate (peak 91), 3-O-caffeoyloleanolic acid (peak 88), and khrinone E (peak 76) are the primary bioactive markers for their antioxidant activity. The radix of MSCP is a rich source to be considered as a natural antioxidant reagent. The study could be helpful for future exploration of its material base and bioactive mechanism. In addition, it would be considered to be used as a new method for the quality control of M. speciosa.