Feature-Based Molecular Networking for the Exploration of the Metabolome Diversity of Common Egyptian Centaurea Species in Relation to Their Cytotoxic Activity

Centaurea is a genus compromising over 250 herbaceous flowering species and is used traditionally to treat several ailments. Among the Egyptian Centaurea species, C. lipii was reported to be cytotoxic against multidrug-resistant cancer cells. In this context, we aimed to explore the metabolome of C. lipii and compare it to other members of the genus in pursuance of identifying its bioactive principles. An LC-MS/MS analysis approach synchronized with feature-based molecular networks was adopted to offer a holistic overview of the metabolome diversity of the Egyptian Centaurea species. The studied plants included C. alexandrina, C. calcitrapa, C. eryngioides, C. glomerata, C. lipii, C. pallescens, C. pumilio, and C. scoparia. Their constitutive metabolome showed diverse chemical classes such as cinnamic acids, sesquiterpene lactones, flavonoids, and lignans. Linking the recorded metabolome to the previously reported cytotoxicity identified sesquiterpene lactones as the major contributors to this activity. To confirm our findings, bioassay-guided fractionation of C. lipii was adopted and led to the isolation of the sesquiterpene lactone cynaropicrin with an IC50 of 1.817 µM against the CCRF-CEM leukemia cell line. The adopted methodology highlighted the uniqueness of the constitutive metabolome of C. lipii and determined the sesquiterpene lactones to be the responsible cytotoxic metabolites.


Introduction
Centaurea is the fourth largest genus in the Asteraceae family [1] and has approximately 250 species (400 in an earlier classification) that are mostly centered in the Mediterranean region. The genus includes diverse biologically active metabolites, including sesquiterpene lactones, triterpenes, flavonoids, and lignans [2]. Owing to such metabolic diversity, many biological activities were reported for its members such as anti-inflammatory, antimicrobial, antioxidant, hepatoprotective, etc. [3].

Comparative Analysis of LC-MS/MS Profiles from Centaurea Species
The LC-MS/MS analysis of the selected Centaurea extracts showed clear qualitative and quantitative differences as observed in their respective base peak chromatograms (Supplementary Figure S1).
As C. lipii was the species with the most potent cytotoxicity, as indicated in a previous screening of Egyptian plant extracts [9], the identification of the contributing metabolites was our goal. Accordingly, a feature-based molecular network was constructed to better visualize the metabolome diversity of the selected Centaurea species and to accentuate the uniqueness of C. lipii.

MS/MS Molecular-Networking-Based Phytochemical Investigations
For a global overview, feature-based molecular networking (FBMN) was applied for the visual exploration of the discrepancy in the recorded metabolome of the studied species as well as for facilitating the metabolite annotation. The constructed FBMN was then analyzed using a MolNetEnhancer workflow which enhances the data annotation via combining outputs from different computational tools [10] (Figure 1a).
The constructed MN consisted of 977 nodes grouped in 77 clusters (with a minimum of 2 connected nodes) and 385 single nodes ( Figure 1). Then, the node and edge attributes were employed so that the color of a node corresponded to the name of the studied Centaurea species. The nodes are displayed as a pie chart to reflect the distribution of each ion among the 8 species (Figure 1).
The recorded metabolome encompassed unidentified clusters with no matches which were manually inspected and identified (i.e., clusters b and c corresponded to the sesquiterpene lactones, Figure 1b) which could be explained by their presence as either ammonia [M+NH 4 ] + or acetonitrile [M+ C 2 H 3 N +H] + adducts.  As delineated in Figure 1b and Supplementary. Table S1, a total of 81 metabolites were tentatively assigned belonging to different chemical classes. This includes 49 flavonoids, 15 sesquiterpene lactones, 10 lignans, four cinnamic acid derivatives, and two coumarins. Figure 2 displays represenative examples of compounds reported here in the genus Centaurea for the first time, and following is a discussion of the annotated metabolites in their elution order. The clusters of interest were as follows: cluster b: sesquiterpene lactones, cluster c: sesquiterpene lactone glycosides, cluster d: flavones, cluster e: flavonoid glycosides, cluster f: lignans, and cluster g: hydroxycinnamic acid derivatives (Figure 1b-g).
As delineated in Figure 1b and Supplementary Table S1, a total of 81 metabolites were tentatively assigned belonging to different chemical classes. This included 49 flavonoids, 15 sesquiterpene lactones, 10 lignans, 4 cinnamic acid derivatives, and 2 coumarins. Figure 2 displays representative examples of the compounds reported here in the genus Centaurea for the first time, and following is a discussion of the annotated metabolites in their elution order.

Hydroxycinnamic Acid Derivatives
Hydroxycinnamic acid derivatives were observed in the constructed FBMN (Figure 1g) exclusively as ferulic acid derivatives as confirmed by their shared daughter ions at m/z 177 and 145. This included feruloyl quinic acid ester (

Sesquiterpene Lactones
Unlike the cinnamic acid derivatives which showed no significant difference in distribution among Centaurea species, sesquiterpene lactones showed a different pattern. Sesquiterpene lactones were almost exclusively detected in C. lipii with few occurring in C. calcitrapa and C. eryngioides.
Sesquiterpene lactones are a group of secondary metabolites widely distributed in the Asteraceae family and are classified according to their carbocyclic skeletons into different classes, i.e., germacranolides, eudesmanolides, guaianolides, and pseudoguaianolides. Several sesquiterpene lactones were detected exclusively in C. lipii, belonging to the germacranolides, guaianolides, cadinanolides, elemanolides, and eudesmanolides ( Figure 3). Annotated sesquiterpene lactones were detected as adducts of acetonitrile [M+C 2 H 3 N+H] + while glycosidic derivatives were seen as ammonia adducts [M+NH 4 ] + (Supplementary  Table S1). Interestingly, the annotated sesquiterpene lactones were mostly reported previously in the genus except for the glycosidic ones which are reported here for the first time in Centaurea.
Hydroxycinnamic acid derivatives were observed in the constructed FBMN (Figure 1g) exclusively as ferulic acid derivatives as confirmed by their shared daughter ions at m/z 177 and 145. This included feruloyl quinic acid ester (1, m/z 369.1179 [M+H] + , C19H20O9) previously reported to occur in Centaurea [11], followed by its amide derivatives as N-feruloyl tyramine isomers (

Sesquiterpene lactones
Alike the cinnamic acid derivatives which showed no significant difference in distribution among Centaurea species, sesquiterpene lactones showed a different pattern. Sesquiterpene lactones were almost exclusively detected in C. lipii with few occurring in C. calcitrapa and C. eryngioides.
Sesquiterpene lactones are a group of secondary metabolites widely distributed in the Asteraceae family and are classified according to their carbocyclic skeletons into different classes, i.e., germacranolides, eudesmanolides, guaianolides, and pseudoguaianolides. Several sesquiterpene lactones were detected exclusively in C. lippi, belonging to the germacranolides, guaianolides, cadinanolides, elemanolides, and eudesmanolides ( Figure 3). Annotated sesquiterpene lactones were detected as adducts of acetonitrile [M+C2H3N+H] + while glycosidic derivatives were seen as ammonia adducts [M+NH4] + (Supplementary Table S1). Interestingly, the annotated sesquiterpene lactones were mostly reported previously in the genus except for the glycosidic ones which are reported here for the first time in Centaurea.  Among the annotated sesquiterpene lactones, germacranolides were the most abundant class. Germacranolide glycosides were tentatively assigned as dihydroparthenolide-Ohexoside isomers (2 and 6, m/z 446.2385 [M+NH 4 ] + , C 21 H 32 O 9 ), particularly in C. calcitrapa and C. eryngioides, and described for the first time in Centaurea.

Flavonoids
Similar to the bioactive sesquiterpene lactones discussed earlier, Centaurea species are well known for their high content of flavonoids [14].
In our investigation, methylated flavonols and flavones were the predominant species, occurring as diglycosides, monoglycosides, acylated monoglycosides, or as free aglycones. In total, 49 flavonoids were annotated, some of which were previously reported to exist in the genus (Supplementary Table S1 Likewise, the monoglycosides previously mentioned, the diglycosides are reported here for the first time in Centaurea, i.e., spinacetin-O-gentiobioside (9, m/z 671. 18  Lastly, 21 flavonoid aglycones were described in this study (Supplementary Table S1). Regarding the distribution of the flavonoids, no specific pattern was observed except a higher prevalence in C. alexandrina and C. pallescens (Figure 1e).

Lignans
In addition to sesquiterpene lactones and flavonoids, Centaurea is known to produce lignans [15], mainly as the dibenzylbutyrolactone type. Reported lignans include matairesinol and arctigenin along with their glycosides matairesionoside and arctiin, which were reported to exert anticancer effects against colorectal cancer [16].
In our study, lignan glycosides existed as ammonia adducts [M+NH 4 ] + as commonly detected in the positive ionization mode used [17]. Annotated lignans included previously reported ones such as matairesinol-O-glucoside (

Bioactivity-Guided Fractionation of C. lipii
According to the biological activity against CCRF-CEM cell lines that we previously reported, the methylene chloride/methanol (1: 1) fraction of C. lipii showed significant cytotoxic activity against CCRF-CEM with IC 50 4.30 µM [9]. Consequently, C. lipii extract was fractioned using a flash column to obtain five collective fractions. The cytotoxicity of these subfractions was evaluated against a drug-sensitive CCRF-CEM leukemia cell line. Fraction 1 (CL1) was found to be the most potent cytotoxic fraction with an IC 50 value 1.81 µM (Figure 4). tairesinol and arctigenin along with their glycosides matairesiono were reported to exert anticancer effects against colorectal cancer [ In our study, lignan glycosides existed as ammonia adducts [ detected in the positive ionization mode used [17]. Annotated l ously reported ones such as matairesinol-O-glucoside ( Additionally, the occurrence of secoisolariciresinol (32, C20H22O4) in Centaurea is reported here for the first time togeth lignan glycosides exemplified by acetyl matairesinoside (28, m C27H34O11) occurring exclusively in C. lipii.

Bioactivity-Guided Fractionation of C. lipii
According to the biological activity against CCRF-CEM cell li reported, the methylene chloride/methanol (1: 1) fraction of C. l cytotoxic activity against CCRF-CEM with IC50 4.30 µ M [9]. Conse was fractioned using a flash column to obtain five collective fracti these subfractions was evaluated against a drug-sensitive CCRF-C Fraction 1 (CL1) was found to be the most potent cytotoxic fract 1.81 µ M (Figure 4).

Discussion
In our study, an LC-MS/MS data analysis approach was ad metabolic diversity of Egyptian Centaurea species with the aid of m the in silico fragmentation trees generated by Sirius. The adopted vantageous in mapping the chemical space of Centaurea species t acids, sesquiterpene lactones, flavonoids, and lignans. Among the compounds are reported to occur in the genus Centaurea for the fir

Discussion
In our study, an LC-MS/MS data analysis approach was adopted to highlight the metabolic diversity of Egyptian Centaurea species with the aid of molecular networks and the in silico fragmentation trees generated by Sirius. The adopted methodology was advantageous in mapping the chemical space of Centaurea species that included cinnamic acids, sesquiterpene lactones, flavonoids, and lignans. Among the annotated features, 21 compounds are reported to occur in the genus Centaurea for the first time.
Additionally, the molecular networks delineated the uniqueness of the metabolic profile of C. lipii, being especially rich in sesquiterpene lactones which might explain its potent cytotoxic activity against multidrug-resistant cancer lines.
For instance, sesquiterpene lactones were detected solely in C. lipii (Figure 3) belonging to the germacranolides, guaianolides, cadinanolides, elemanolides, and eudesmanolides. Diverse biological activities were reported for sesquiterpene lactones, including anti-inflammatory, antiparasitic, antiviral, cytotoxic, and others [18]. Moreover, sesquiterpene lactones were recognized as potential candidates for cancer treatment owing to their selective inhibition of tumor and cancer stem cells [18]. Indeed, former investigations have highlighted that the biological activity of sesquiterpene lactones is attributed to the inhibition of enzymes, transcription factors, and/or functional proteins [18].
Since the late 1960s, the cytotoxicity of the sesquiterpene lactones has been investigated to understand the underlying structure-activity relationships. The exocyclic α-methyleneγ-lactone, together with the cyclopentenone and/or α, β-unsaturated ester, has a pivotal role in enhancing cytotoxicity [19]. Further comparisons of different scaffolds revealed that guaianolides and pseudoguaianolides possess the most potent activity [20]. These findings might explain the pronounced cytotoxic activity observed in C. lipii in comparison to the other Centaurea species.
Bioassay-guided fractionation confirmed such an assumption and led to the isolation of cynaropicrin from the cytotoxic fraction with an IC 50 of 1.817 µM against the CCRF-CEM leukemia cell line. Cynaropicrin has been formerly reported to exist in several Centaurea species, such as C. behen [21], C. ruthenica [22], and others. Additionally, its cytotoxic activity against the CCRF-CEM leukemia cell line was formerly documented with an IC 50 value of 0.473 µg/mL [23]. Its cytotoxic properties were correlated to its ability to diminish the generation of intracellular reactive oxygen species involved in carcinogenesis [24].
In conclusion, the described analysis proved efficient and competent for mapping and correlating the constitutive metabolome of the selected Centaurea species and simultaneously allowed for the rapid detection of the bioactive metabolites. The outcomes were further validated through bioactivity-guided isolation of the bioactive scaffold.

Plant Materials
Plant samples were collected from their respective locations as listed in Table 1  Methylene chloride, methanol, and acetonitrile were purchased from Sigma Aldrich (Steinheim, Germany). All the solvents used were of HPLC grade.

Preparation of the Extracts
The air-dried powdered aerial parts of the studied Centaurea species (100 g each) were macerated separately in 1 L CH 2 Cl 2 /MeOH (1:1) for 24 h at room temperature and then filtered. The filtrates were then evaporated under reduced pressure, lyophilized, and kept frozen at −20 • C for further analyses.

LC-MS/MS Data Acquisition
Dried CH 2 Cl 2 /MeOH (1:1) extract of each species was redissolved in MeOH (HPLC grade) to a final concentration of 2 µg/mL. Chromatographic separation was performed as described before [25].

Data Preprocessing, Molecular Networking, and Compound Dereplication
The feature-based molecular network (FBMN) was built from each species' HPLC-HRMS/MS data (in positive mode). Firstly, The MSConvert program was used to convert raw data files into 32-bit MzXML files, which were then loaded into Mzmine 2.53 for feature identification [26]. The mgf file from the Mzmine was transferred through WinSCP (https: //winscp.net accessed on 12 July 2021) to the Global Natural Products Social Molecular Networking platform (https://gnps.ucsd.edu accessed on 12 July 2021) to create an MN following the online protocol [27]. Subsequently, the constructed molecular network was enhanced with a MolNetEnhancer to boost the chemical structural annotation. For visualization of the resulting MN, Cytoscape (ver. 3.8.2.) was used.
Further data analysis was achieved by importing the mgf output file from Mzmine 2.53 to Sirius + CSI: Finger ID 4.4.29 for the molecular formula prediction (C, H, N, O, S, P) and searching the structure database with 10 ppm m/z tolerance using PubChem online database [28].

Cell Culture
The CCRF-CEM leukemia cells were kindly provided by Prof. Axel Sauerbrey (Department of Pediatrics, University of Jena, Jena, Germany) [29]. The cell lines were authenticated using Multiplex Cell Authentication (MCA) based on single-nucleotide polymorphism profiling by Multiplexion GmbH (Heidelberg, Germany) as previously detailed [30]. Those cell lines have been in culture for 14 years.

Resazurin Cytotoxicity Assay
The cytotoxicity of C. lipii fractions and the isolated compound was determined by the resazurin reduction assay using a modified protocol previously described [9].
NMR spectra were measured on a Bruker 500 NMR spectrometer (USA) (500 MHz for 1H and 125 MHz for 13C). All chemical shifts (δ) are given in ppm units with reference to TMS as an internal standard, and coupling constants (J) are reported in Hz.