Discriminative Metabolomics Analysis and Cytotoxic Evaluation of Flowers, Leaves, and Roots Extracts of Matthiola longipetala subsp. livida

Matthiola longipetala subsp. livida is an annual herb in Brassicaceae that has received little attention despite the family’s high reputation for health benefits, particularly cancer prevention. In this study, UPLC-HRMS-MS analysis was used for mapping the chemical constituents of different plant parts (i.e., flowers, leaves, and roots). Also, spectral similarity networks via the Global Natural Products Social Molecular Networking (GNPS) were employed to visualize their chemical differences and similarities. Additionally, the cytotoxic activity on HCT-116, HeLa, and HepG2 cell lines was evaluated. Throughout the current analysis, 154 compounds were annotated, with the prevalence of phenolic acids, glucosinolates, flavonol glucosides, lipids, peptides, and others. Predictably, secondary metabolites (phenolic acids, flavonoids, and glucosinolates) were predominant in flowers and leaves, while the roots were characterized by primary metabolites (peptides and fatty acids). Four diacetyl derivatives tentatively assigned as O-acetyl O-malonyl glucoside of quercetin (103), kaempferol (108 and 112), and isorhamnetin (114) were detected for the first time in nature. The flowers and leaves extracts showed significant inhibition of HeLa cell line propagation with LC50 values of 18.1 ± 0.42 and 29.6 ± 0.35 µg/mL, respectively, whereas the flowers extract inhibited HCT-116 with LC50 24.8 ± 0.45 µg/mL, compared to those of Doxorubicin (26.1 ± 0.27 and 37.6 ± 0.21 µg/mL), respectively. In conclusion, the flowers of M. longipetala are responsible for the abundance of bioactive compounds with cytotoxic properties.


Introduction
Brassicaceae (=Cruciferae) is one of the economically important angiosperm families, commonly known as the crucifers, cabbage, or mustard family, containing over 372 genera and approximately 4636 species [1].Plants of the family Brassicaceae have been an interesting research subject for years due to their economic and agricultural importance.Many species have been valued as food crops; some are vegetables, others are sources of industrial and cooking oils, forage, and condiments and others are grown as ornamental species for their showy flowers and significant numbers as medicinal herbs [2].Additionally, certain wild cruciferous plants are rich in secondary metabolites, especially glucosinolates, phenolic acids, and flavonoids, which have many biological activities and, therefore, numerous nutritional and medicinal benefits [3,4].Matthiola longipetala subsp.livida (Delile) Maire is one of the common wild medicinal cruciferous herbs growing mainly in the Egyptian Mediterranean region, and it is locally known as "Manthor" [5].Although some previous phytochemical studies have been conducted on M. longipetala subsp.livida [6][7][8][9], the reported compounds represent only a small portion of the species' chemical composition.
Similarly, certain biological activities such as antibacterial, antifungal, and anticancer effects have been reported for the investigated species [8][9][10].
Lately, metabolomics platforms have been widely used to map the metabolome of plants, among which ultra-performance liquid chromatography coupled with highresolution tandem mass spectrometry (UPLC-HRMS/MS) as the most extensively adopted for mapping the secondary metabolome space.UPLC-HRMS/MS offers the advantages of high efficiency, reproducibility, and shorter analysis [11].Additionally, advances in the data analysis tools, such as molecular networks through the Global Natural Products Social Molecular Networking (GNPS) [12], allow for the visual display of the constitutive metabolome among samples and the propagation of metabolites annotation [13].
Over the last few decades, most new therapeutic interventions involving plant secondary metabolites and their derivatives have been aimed at combating cancer.In this regard, cruciferous plants have been previously reported to lower the risk of developing various cancers [14].Our previous research reported the moderate cytotoxic potential of the alcoholic extract of the aerial part of M. longipetala subsp.livida against cervix (HeLa) and colon (HCT116) cell lines [8].Moreover, another report assessed the extract's activity against HepG2 cells in vitro using MTT, DNA fragmentation, and cell proliferation cycle measurements, and it demonstrated significant activity [15].
In continuation of our previous study, the present work aimed to map the underexplored chemistries of different organs (i.e., flowers, leaves, and roots) of M. longipetala using UPLC-HRMS-MS analysis that recruited for a holistic overview of the plant's constitutive chemistries, coupled with spectral similarity networks through the GNPS [12]; this was in addition to evaluating the cytotoxic activities of the three organs on HCT-116, HeLa, and HepG2 cell lines to suggest the one responsible for this potential.

Chemicals and Reagents
All chemicals for chemical analysis were obtained from Sigma-Aldrich (Merck, Kenilworth, NJ, USA).

Plant Material and Preparation of the Extracts
M. longipetala subsp.livida (650 g fresh weight) was collected from Alexandria-Marsa Matruh Road, 31 • 04 15.3 N 27 • 58 10.4 E, Egypt, in February 2018.The identity of the plant was authenticated by Prof. Dr. Mona M. Marzouk.A voucher specimen (ML_28_2_18) was placed in the herbarium of the National Research Centre (CAIRC), Cairo, Egypt.The flowers, leaves, and roots (117, 175, and 152 g fresh weight, respectively) were washed thoroughly with bi-distilled H 2 O, dried in shade, and ground finely.Fifteen grams of each dried powdered organ was separately extracted using 70% methanol (500 mL) by sonication (2 h, 60 • C) and filtered over charcoal to yield three aqueous methanolic extracts [16,17].The flowers, leaves, and roots extracts were concentrated under reduced pressure at 50 • C to produce three dried extracts (2.761, 1.140, and 1.832 g, respectively).

Sample Preparation for UPLC-HRMS-MS Measurement
The dried extracts were prepared for UPLC-HRMS/MS analyses following a previously described protocol [16].The extracts (50 mg each) were dissolved in 70% MeOH (HPLC-grade) with sonication (10 min), then centrifuged.Aliquots were then evaporated under reduced pressure, followed by freeze-drying for 48 h.For MS analysis, 1 mg in 250 µL MeOH (MS-grade) were prepared consuming 5 µL as an injection volume in the UPLC-MS analysis.

UPLC-HRMS-MS Analysis
The HRMS/MS analysis was conducted on a MaXis 4G instrument (Bruker Daltonics ® , Bremen, Germany) coupled with an Ultimate 3000 UPLC (Thermo Fisher Scientific ® , Waltham, MA, USA).A UPLC method was applied as described by [17] as follows: (with 0.1% formic acid in H 2 O as solvent A and 100% ACN as solvent B), an isocratic gradient of 10% B for 10 min, 10% to 100% B in 30 min, 100% B for an additional 10 min, using a flow rate of 0.3 mL/min; 5 µL injection volume and UV detector (UV/VIS) wavelength monitoring at 210, 254, 280, and 360 nm.The separation was conducted on a Nucleoshell RP 18 column, 2.7 µm, 150 × 2 mm (Macherey-Nagel ® , Düren, Germany), and the range for MS acquisition was 50-1800 Daltons (Da).A capillary voltage of 4500 V, nebulizer gas pressure (nitrogen) of 2 (1.6) bar, ion source temperature of 200 • C, dry gas flow of 9 L/min, and spectral rates of 3 Hz for MS 1 and 10 Hz for MS 2 , were utilized.For acquiring MS/MS fragmentation, the 10 most intense ions per MS 1 were selected for subsequent CID, with stepped CID energy applied.The employed parameters for tandem MS were applied as previously detailed [18].

Data Analysis and Preprocessing
Raw data inspection was performed using Compass Data Analysis 4.4 (Bruker Daltonics ® ).A Metaboscape 3.0 (Bruker Daltonics ® ) was utilized for feature detection, grouping, and alignment, employing the T-ReX 3D (Time aligned Region Complete eXtraction) algorithm [19].Bucketing was performed with an intensity threshold of 1 × 10 5 and a retention time range from 0 to 40 min with a restricted mass range m/z from 130 to 1800.

Feature-Based Molecular Networking (FBMN) and Metabolites Dereplication
The produced MGF file and the feature quantification table (CSV file) were used in the feature-based molecular networking (FBMN) following the online workflow in GNPS platform (http://gnps.ucsd.edu),accessed on 28 December 2019 [20].The parameters applied for the construction of the FBMN via the GNPS platform as follows: a parent mass tolerance (0.05 Da), a fragment ion tolerance (0.05 Da), a cosine score (0.7), and minimum shared fragments (6).To avoid misinterpretation of artifacts, the blank run was uploaded as a distinct sample on GNPS workflow and excluded from the networks.Cytoscape version 3.9.1 (https://cytoscape.org/),accessed on 28 February 2022, was used for the network visualization.
The metabolites' dereplication was based on the chromatographic performance, chemical formula, and fragmentation pattern compared to those of MS 2 data from literature and spectra from MS reference database (MoNA, NIST14, and Respect) (Table 1).Sirius plus CSI:FingerID 5.5.4 were used for the manual putative structures identification [21], assisted by the molecular formula prediction (C, H, N, O, S, and P) and candidate search with m/z tolerance set to 20 ppm connected to online Pubchem.The proposed in silico fragmentation trees are the impetus for further support for identification.

Cell Viability by MTT Assay
The samples were prepared by dissolving stock solution in DMSO to give operating concentrations of each sample range from 100 to 0.78 µg/mL, and the cells were incubated with these concentrations in triplicate (37 • C, 72 h) in a CO 2 environment.Control wells were treated with the same amount of complete growth media only.For all treatments and untreated control groups, complete growth media without cells were added as a blank to reduce the background absorbance values.Separately, each experiment were conducted three times.MTT assay was performed by removing the medium quietly and adding MTT solution (10 µL) with a last concentration (5 mg/mL) per well then incubating (37 • C, 4 h) until the purple crystals were shaped.Then, the MTT solution was discarded from every well and DMSO (100 µL) was subjected to dissolve the crystals.The 96-well plate was shaken (15 min) using a microplate shaker until totally dissolved of the crystals.For each well, the absorbance value was assayed (595 nm wavelength) using a microplate multi-well reader [63].The cell viability (CV) percentage after treatment with M. longipetala subsp.livida extracts were considered as follows: CV (%) = (absorbance of the treated cells − absorbance of blanks)/(absorbance of control cell − absorbance of blanks) × 100.The lethal concentration of the samples caused the death of 50% of cells (LC 50 ) which was also calculated at 48 h.Doxorubicin, the anticancer drug, was used as a positive control.

UPLC-HRMS/MS Metabolites Profiling of the Extracts
The current study aimed to comparably chart the metabolic composition of different organs (i.e., flowers, leaves, and roots) of M. longipetala via UPLC-PDA-ESI-HRMS/MS analysis in both positive and negative ionization modes.The overlaid BPC (base peak chromatograms) of the three extracts exhibited some differences, especially at the Rt range of (10-25 min) in the positive ionization mode and (6-15 min) in the negative ionization mode (Supplementary Figure S1), suggesting that the three extracts could be of different biological relevance.

UPLC-HRMS/MS Metabolite Annotation Aided with Molecular Networking
The UPLC-HRMS/MS data were mined employing the GNPS platform, in which feature-based molecular networks (FBMNs) were generated to visually display the existing chemical space and the metabolites distribution in the different plant parts of M. longipetala.
Two FBMNs were laid out from the acquired MS/MS data for both ionization modes.The negative FBMN constituted 188 nodes grouped into 19 clusters (with a minimum of two connected nodes) and 130 singletons.The significant dereplicated sets of the negative FBMN were the secondary metabolites clusters: A (flavonoid glycosides), B (glucosinolates), C (hydroxycinnamic acid derivatives), D (hydroxybenzoic acid derivatives), and E (biflavones) (Figure 1).These metabolites are distributed in the flowers and leaves organs with their abundance in flowers, which could be responsible for the current cytotoxic assessment and guidance for further biological activities.Similarly, the positive FBMN constituted 257 nodes in 41 clusters and 104 discrete nodes, in which the classes of interest are cluster A (flavonoid glycosides and hydroxylated flavonoid aglycones) and B (methoxylated flavonoid aglycones); besides, cluster C (peptides) is presented as a primary metabolites class which characterized the roots organ and ionized in the positive ionization mode only (Figure 2).In general, nodes were portrayed as a pie chart to reflect the relative abundance of each ion in the three plant parts.
In total, 154 compounds were annotated belonging to different chemical classes (i.e., glucosinolates, phenolic acids, flavonoids, etc.).Almost all the annotated features are reported for the first time to exist in M. longipetala subsp.livida (Table 1).The classes and/or subclasses of compounds were preformed manually guided by the literature [2,28,33,80,81] and automatically through the ClassyFire webserver at http://classyfire.wishartlab.com/(accessed on 27 June 2023) [82].Following is a detailed discussion of the detected metabolites according to their chemical class.In total, 154 compounds were annotated belonging to different chemical classes (i.e., glucosinolates, phenolic acids, flavonoids, etc.).Almost all the annotated features are reported for the first time to exist in M. longipetala subsp.livida (Table 1).The classes and/or subclasses of compounds were preformed manually guided by the literature [2,28,33,80,81] and automatically through the ClassyFire webserver at http://classyfire.wishartlab.com/(accessed on 27 June 2023) [82].Following is a detailed discussion of the detected metabolites according to their chemical class.

Glucosinolates
Glucosinolates are one of the main bioactive metabolites of the Brassicaceae species and are thought to play a significant role in the health benefits of such species [2,28,80].Their fragmentation behavior involves the cleavage of the sugar-sulfur bond, giving the

Glucosinolates
Glucosinolates are one of the main bioactive metabolites of the Brassicaceae species and are thought to play a significant role in the health benefits of such species [2,28,80].Their fragmentation behavior involves the cleavage of the sugar-sulfur bond, giving the fragment ion m/z 259 and the sulfur-aglycone showing fragment ions at m/z 195 and m/z 275.The intramolecular rearrangements of the attachment of aglycone and sulfate to the glucose moiety give the fragment ion m/z 241 after water cleavage from m/z 259 [28,80].
Eight of the nine identified glucosinolates are grouped in cluster B of the negative FBMN, occurring in the three plant parts (Figure 1).This includes isomers of glucoraphanin ).Lastly, one glucosinolate was observed in the positive FBMN as a self-looped node and was identified as raphenin (10, m/z 176.0201 [M − H] + ) (Table 1).

Phenolics
Besides the glucosinolates, members of the Brassicaceae are well recognized for their high content of phenolic metabolites, with qualitative and quantitative differences among species and varieties, within the same species, and plant parts [33].In the present study, phenolic metabolites showed the highest accumulation in the flowers extract and the least in the roots.The major phenolic classes identified were phenolic acids and flavonoids.

Phenolic Acids and Derivatives
Detected phenolic acids included hydroxybenzoic acid and hydroxycinnamic acid (coumaric, ferulic, and sinapic acids) derivatives, which are widely distributed in numerous members of the Brassicaceae family, commonly as glycosylated descendants [2,33].
The negative FBMN delineated the abundance of glycosylated hydroxycinnamic acids in the flowers and grouped in cluster C (Figure 1 Similarly, glycosylated hydroxybenzoic acids were distributed in the three organs, and were observed in the negative FBMN (Figure 1

Flavonoids
Flavonoids protect plants from various biotic and abiotic stresses by acting as natural antioxidants, unique UV filters, signal molecules, allelopathic compounds, and antimicrobial defensive compounds [81].Additionally, their impressive biological effects have made them excellent candidates as nutraceutical supplements for human intake, disease prevention, and health promotion [2,81].
Throughout the current analysis, around 40% of the detected constituents are flavonoids (64 metabolites) (Table 1) delivered as cluster A and some as self-looped nodes in FBMN of the negative ionization mode (Figure 1) and clusters A and B in the positive one (Figure 2), being more abundant in the flowers.
Only one flavone-O-glycoside was detected exclusively in the roots and was assigned as apigenin 7-O-glucoside (92, m/z 431.0981 [M − H] − ) based on the main fragment ion at m/z 269 which corresponds to apigenin aglycone and the loss of a glucose moiety [M − H-162] − .
Flavonol-O-glycosides The predominant annotated flavonol glycosides were mainly glycosides of kaempferol, isorhamnetin, and quercetin with little presence of rhamnocitrin, based on our former studies through acid hydrolysis and NMR data [8,46].The quercetin glycosides in both FBMN were directly linked to their kaempferol correspondences by a difference of 16 Da (-O-), and with the isorhamnetin correspondences by a difference of 14 Da (-CH 2 ).The direct attachment of the isorhamnetin glycosides to those of the kaempferol correspondents with a mass difference of 30 Da suggests possible OCH 3 expansion.
Acylated flavonol-O-glycosides A total of 16 acylated flavonol mono-glycosides were also observed in group A of the positive and negative FBMNs and connected with their O-glucoside analogs with MS differences of either 42 Da (acetyl) and/or 86 Da (malonyl).Whereas the acetylated and malonylated counterparts were correlated with each other with a 44 Da (CO 2 ) difference (Figures 1 and 2).S2).Acylated monoglycoside derivatives of quercetin, kaempferol, and isorhamnetin have already been found in some cruciferous species [3,39], while reported for the first time from the genus Matthiola.
Additionally, two biflavone-structure were detected as a cluster (E) in a negative FBMN and elucidated as two isomers of methylamentoflavone (120 and 121, at m/z 551.09 [M − H] − ), confirmed by their fragmentation pattern and GNPS library (Figure 1).Rare biflavone derivatives were reported before for some species of Brassicaceae [48].

Iridoids and Diterpenes
Only one iridoid compound was found for the first time in the investigated species and concentrated in the flower parts.The iridoid is identified as loganic acid (30)  Both compounds showed a wide range of activities including anti-cancer, anti-inflammatory, and antioxidant effects [36,71].

Coumarin
Coumarins are another vital class of secondary metabolites and were mainly observed in the positive ionization mode (Table 1).Hydroxy coumarin (38,

Fatty Acids and Derivatives
Eight fatty acids and one fatty acid ester were detected, in the case of compound (116), the fragmentation patterns were matched with 9,12,13-trihydroxy-octadecadienoic acid, the molecular ion at m/z 327.2178 [M − H] − and the fragments at m/z 229 and m/z 171 pointed to the positions of hydroxyl groups of fatty acids (that is, at 12 and/or 13, 9 and/or 10th carbon) but it was not easy to assign the functional groups and double bonds depending on our data.Therefore, this compound was identified as trihydroxy-octadecadienoic acid.

Peptides
Eleven polypeptides were detected in the positive ionization mode (cluster D), thoroughly characterized for the root organ (Table 1, Figure 2).They were tentatively identified according to MS differences and fragmentation patterns, then further sequenced corresponding to [72].

Cytotoxicity
As expected, flower extract that showed the highest abundance of secondary metabolites revealed a significant cell viability inhibition of HCT-116 and HeLa cell lines growth, with LC 50 values (24.8 ± 0.45 and 18.1 ± 0.42 µg/mL), compared to those of Doxorubicin (37.6 ± 0.21 and 26.1 ± 0.27 µg/mL), respectively.Similarly, the leaf extract inhibited the propagation of the HeLa cell line with an LC 50 value of 29.6 ± 0.35 µg/mL.The three methanolic extracts did not show any cytotoxic effect on the HepG2 cell line (Supplementary Table S1).These findings summarize the relationships between the cytotoxic assessment of the three examined organs and the concentration of secondary metabolites, particularly flavonoids.Consequently, the present data indicates that the flower organ is responsible for activities reported before for aerial parts on the same species [8].

Table 1 .
Metabolites identified in the aqueous methanol extracts of flowers, leaves, and roots from Matthiola longipetala subsp.livida via UPLC-HRMS-MS in negative and positive ionization modes.