Achillea fragrantissima (Forssk.) Sch.Bip Flower Dichloromethane Extract Exerts Anti-Proliferative and Pro-Apoptotic Properties in Human Triple-Negative Breast Cancer (MDA-MB-231) Cells: In Vitro and In Silico Studies

The aggressive triple-negative breast cancer (TNBC) is a challenging disease due to the absence of tailored therapy. The search for new therapies involves intensive research focusing on natural sources. Achillea fragrantissima (A. fragrantissima) is a traditional medicine from the Middle East region. Various solvent extracts from different A. fragrantissima plant parts, including flowers, leaves, and roots, were tested on TNBC MDA-MB-231 cells. Using liquid chromatography, the fingerprinting revealed rich and diverse compositions for A. fragrantissima plant parts using polar to non-polar solvent extracts indicating possible differences in bioactivities. Using the CellTiter-Glo™ viability assay, the half-maximal inhibitory concentration (IC50) values were determined for each extract and ranged from 32.4 to 161.7 µg/mL. The A. fragrantissima flower dichloromethane extract had the lowest mean IC50 value and was chosen for further investigation. Upon treatment with increasing A. fragrantissima flower dichloromethane extract concentrations, the MDA-MB-231 cells displayed, in a dose-dependent manner, enhanced morphological and biochemical hallmarks of apoptosis, including cell shrinkage, phosphatidylserine exposure, caspase activity, and mitochondrial outer membrane permeabilization, assessed using phase-contrast microscopy, fluorescence-activated single-cell sorting analysis, Image-iT™ live caspase, and mitochondrial transition pore opening activity, respectively. Anticancer target prediction and molecular docking studies revealed the inhibitory activity of a few A. fragrantissima flower dichloromethane extract-derived metabolites against carbonic anhydrase IX, an enzyme reported for its anti-apoptotic properties. In conclusion, these findings suggest promising therapeutic values of the A. fragrantissima flower dichloromethane extract against TNBC development.


Introduction
Breast cancer is the first-rank diagnosed cancer worldwide and the second leading cause of cancer mortality in the female population, responsible for about 685,000 deaths in 2020 [1]. Due to its heterogeneity in morphological, molecular, and biochemical features, breast cancer is one of the most challenging diseases among all types of cancer [2]. A highly aggressive subtype of breast cancer is triple-negative breast cancer (TNBC), which is mainly characterized by the lack of the expression of the estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 [3]. This TNBC subtype, accounting for 10-20% of all breast cancer subtypes, is mostly diagnosed at the advanced stage of the cancer, and contributes to the metastatic disease, which is the most critical condition reducing patient survival [4,5]. The treatment of this heterogeneous TNBC presents a clinical challenge due to the lack of conventional targeted therapies, requiring the development of tailoring treatment, including immune-checkpoint inhibitors and antibody-drug conjugates [5,6]. Although both chemotherapy and surgery can be beneficial, many studies demonstrate that surgical removal of the cancer generates tumor relapse and metastatic recurrence [7]. Thus, chemotherapy remains the first-line standard therapeutic option, paving the way for different systemic therapies, including the development of novel anticancer natural product-based drugs [8][9][10][11].
Traditional herbal medicine and derivative natural compounds have been widely studied for their anticancer properties via different mechanisms leading to cancer cell death, including programmed cell death (i.e., apoptosis) [12,13]. Apoptosis is a complex series of proteolytic network reactions distinguished into two molecular pathways, death receptor-dependent (extrinsic) and mitochondria-dependent (intrinsic) apoptosis involving initiators caspase-8 and caspase-9, respectively [14]. Mitochondria play a regulatory role in the intrinsic apoptosis pathway through the expression of mitochondrial anti-apoptotic proteins, such as Bcl-2, and pro-apoptotic protein members of the Bcl-2 family such as pro-apoptotic proteins Bak (Bcl-2 antagonist or killer) and Bax (Bcl-2-associated X protein). Bax proteins generate oligomerization or heteromerization between Bax/Bak proteins to form large pores in the outer membrane of the mitochondria leading to the increase in outer-membrane permeabilization, which subsequently release cytochrome c, a main component of caspase-9 activation in the apoptosome [14]. The cleavage of these initiator caspases results in the cleavage of executioner caspases (caspase-7, caspase-3), which leads to membrane lipid profile reversion characterized by phosphatidylserine exposure, cytoskeletal and nuclear protein degradation, and deoxyribonucleic acid (DNA) degradation, causing the apoptotic body formation [15,16].
The Achillea genus belongs to the family Asteraceae (i.e., Compositae), which contains over 130 perennial species found in temperate climates in dry or semi-dry habitats, including the species of Achillea fragrantissima (Forssk.) Sch.Bip (A. fragrantissima) (Synonym, Santolina fragrantissima Forssk.), which are known as traditional herbal medicine consumed by Middle Eastern populations [17,18]. It is commonly referred to as Qaisoom (in Arabic, ) in the Middle East region [19]. This traditional herbal medicine is used to treat numerous illnesses including respiratory diseases, eye infections, smallpox, fever, gastrointestinal disturbances, dysmenorrhea, headache, fatigue, and diabetes [18,20]. In this study, we hypothesize that this understudied local traditional herbal medicine might contain phytochemical(s) that could be potential therapeutic agent(s) against TNBC.

Screening of the Anti-Proliferative Potential of A. fragrantissima Extracts Derived from Flowers, Leaves, and Roots against TNBC Cells
In the search for the discovery of new anticancer natural compound-based drugs for the treatment of TNBC, different parts (i.e., flowers, leaves, and roots) of the plant A. fragrantissima were extracted using various solvents (i.e., methanol, ethanol, dichloromethane, and chloroform). Each aqueous A. fragrantissima extract was then tested for endotoxin contaminants, which was confirmed by the low endotoxin unit (EU) levels reaching 0.0032 EU/µg herb weight. Concerning the chemical analysis, the fingerprinting for each extract was analyzed using high-performance liquid chromatography and ultra-violet detector (HPLC-UVD), which displayed different chromatograms, as seen in Figure 1.

Screening of the Anti-Proliferative Potential of A. fragrantissima Extracts Derived from Flowers, Leaves, and Roots against TNBC Cells
In the search for the discovery of new anticancer natural compound-based drugs for the treatment of TNBC, different parts (i.e., flowers, leaves, and roots) of the plant A. fragrantissima were extracted using various solvents (i.e., methanol, ethanol, dichloromethane, and chloroform). Each aqueous A. fragrantissima extract was then tested for endotoxin contaminants, which was confirmed by the low endotoxin unit (EU) levels reaching 0.0032 EU/µg herb weight. Concerning the chemical analysis, the fingerprinting for each extract was analyzed using high-performance liquid chromatography and ultraviolet detector (HPLC-UVD), which displayed different chromatograms, as seen in Figure  1. Fingerprinting chromatograms of A. fragrantissima flower-, leaf-, and root-derived solvent extracts using HPLC-UV. The chromatograms are based on a gradient method of methanol and water, and UV set at 275 nm with 5 µL injection volume. The separation of the dimethyl sulfoxide (DMSO)-dissolved compounds was performed using a Kinetex ® C-18 column (250 × 4.6 mm, 5 µm) and revealed diverse polar and non-polar compositions for the different solvent extracts of the herbal medicine.
The measurement of half-maximal inhibitory concentrations (IC50) was performed using an ATP-based viability assay. The dichloromethane extract of the A. fragrantissima flowers displayed the lowest IC50 value followed by the chloroform extract. This dichloromethane extract was used for the remaining in vitro investigations for anticancer properties. Based on the ATP generated by living untreated and treated TNBC MDA-MB-231 cells, the IC50 value reflecting the anti-proliferative potential of each extract was determined. All the A. fragrantissima extracts significantly inhibited MDA-MB-231 cell viability in a dose-dependent manner ( Figure 2). The mean IC50 values ranged from 32.4 to 161.7 µg/mL, including the dichloromethane and chloroform extracts derived from the flowers showing the lowest mean IC50 values, followed by the leaf-derived extracts, as seen in Table 1. The dichloromethane extract of the A. fragrantissima flower was shown to be the most potent leading to MDA-MB-231 cell growth inhibition and, thus, was selected for further experiments in this study. Fingerprinting chromatograms of A. fragrantissima flower-, leaf-, and root-derived solvent extracts using HPLC-UV. The chromatograms are based on a gradient method of methanol and water, and UV set at 275 nm with 5 µL injection volume. The separation of the dimethyl sulfoxide (DMSO)-dissolved compounds was performed using a Kinetex ® C-18 column (250 × 4.6 mm, 5 µm) and revealed diverse polar and non-polar compositions for the different solvent extracts of the herbal medicine.
The measurement of half-maximal inhibitory concentrations (IC 50 ) was performed using an ATP-based viability assay. The dichloromethane extract of the A. fragrantissima flowers displayed the lowest IC 50 value followed by the chloroform extract. This dichloromethane extract was used for the remaining in vitro investigations for anticancer properties. Based on the ATP generated by living untreated and treated TNBC MDA-MB-231 cells, the IC 50 value reflecting the anti-proliferative potential of each extract was determined. All the A. fragrantissima extracts significantly inhibited MDA-MB-231 cell viability in a dosedependent manner ( Figure 2). The mean IC 50 values ranged from 32.4 to 161.7 µg/mL, including the dichloromethane and chloroform extracts derived from the flowers showing the lowest mean IC 50 values, followed by the leaf-derived extracts, as seen in Table 1. The dichloromethane extract of the A. fragrantissima flower was shown to be the most potent leading to MDA-MB-231 cell growth inhibition and, thus, was selected for further experiments in this study.     In vitro investigations were performed to determine if the inhibition of MDA-MB-231 cell growth was mediated via the caspase-dependent extrinsic apoptotic pathway. The treatment of MDA-MB-231 cells by the A. fragrantissima flower dichloromethane extract increased observed cell shrinkage in a dose-dependent manner, suggesting apoptosis as the mechanism of cell death ( Figure 3). The caspase-dependent apoptosis was confirmed in A. fragrantissima flower dichloromethane extract-treated MDA-MB-231 cells presenting cleaved caspase-3/7; the active form indicated in red fluorescence and was determined under confocal fluorescence microscopy ( Figure 4). The confocal images clearly display an increase in caspase-3/7 activity in the cells treated with the protein kinase inhibitor staurosporine (STS, a potent apoptosis inducer) and A. fragrantissima flower dichloromethane extract (25 to 200 µg/mL), in comparison with the untreated and DMSO-treated cells, the negative controls. The status of apoptosis was also determined using the FACS analysis based on the Annexin V/propidium iodide (PI) stain, which clearly showed early and late apoptotic status of A. fragrantissima flower dichloromethane extract-treated cells, in a dose-dependent manner, compared to DMSO-treated cells ( Figure 5). The mitochondria potential membrane assay detected an increase in the number of apoptotic whole-cell green fluorescence indicating the involvement of the intrinsic apoptotic pathway in STS-and A. fragrantissima flower dichloromethane extract-treated cells, compared to the untreated and DMSO-treated cells ( Figure 6).

Induction of Apoptosis in MDA-MB-231 Cells by A. fragrantissima Flower Dichloromethane Extract
In vitro investigations were performed to determine if the inhibition of MDA-MB-231 cell growth was mediated via the caspase-dependent extrinsic apoptotic pathway. The treatment of MDA-MB-231 cells by the A. fragrantissima flower dichloromethane extract increased observed cell shrinkage in a dose-dependent manner, suggesting apoptosis as the mechanism of cell death ( Figure 3). The caspase-dependent apoptosis was confirmed in A. fragrantissima flower dichloromethane extract-treated MDA-MB-231 cells presenting cleaved caspase-3/7; the active form indicated in red fluorescence and was determined under confocal fluorescence microscopy ( Figure 4). The confocal images clearly display an increase in caspase-3/7 activity in the cells treated with the protein kinase inhibitor staurosporine (STS, a potent apoptosis inducer) and A. fragrantissima flower dichloromethane extract (25 to 200 µg/mL), in comparison with the untreated and DMSOtreated cells, the negative controls. The status of apoptosis was also determined using the FACS analysis based on the Annexin V/propidium iodide (PI) stain, which clearly showed early and late apoptotic status of A. fragrantissima flower dichloromethane extract-treated cells, in a dose-dependent manner, compared to DMSO-treated cells ( Figure 5). The mitochondria potential membrane assay detected an increase in the number of apoptotic whole-cell green fluorescence indicating the involvement of the intrinsic apoptotic pathway in STS-and A. fragrantissima flower dichloromethane extract-treated cells, compared to the untreated and DMSO-treated cells ( Figure 6).

PASS Online Anticancer Predictions
The anticancer activity of the identified metabolites in A. fragrantissima flower dichloromethane extract was predicted using the prediction of activity spectra for substances (PASS) online webserver. If the Pa value is greater than 0.3, it indicates a high probability that the metabolite will be active experimentally. As summarized in Table 2, all the 14 identified metabolites demonstrated high probabilities that were greater than 0.3 (M5, M7, M8, M10, M12, and M13), and some metabolites (M1-M4, M6, M9, M11, and M14) were higher than 0.7 indicating a high probability that these metabolites would

PASS Online Anticancer Predictions
The anticancer activity of the identified metabolites in A. fragrantissima flower dichloromethane extract was predicted using the prediction of activity spectra for substances (PASS) online webserver. If the P a value is greater than 0.3, it indicates a high probability that the metabolite will be active experimentally. As summarized in Table 2, all the 14 identified metabolites demonstrated high probabilities that were greater than 0.3 (M5, M7, M8, M10, M12, and M13), and some metabolites (M1-M4, M6, M9, M11, and M14) were higher than 0.7 indicating a high probability that these metabolites would be active experimentally. Moreover, the inactivity scores determined for the metabolites were near to zero suggesting their promising anticancer activity.

Molecular Target Predictions
The possible molecular targets for the identified metabolites were predicted using the Molinspiration webserver. The webserver gives an activity prediction involving main five targets important for biological effects including the G protein-coupled receptor (GPCR), ion channel, kinase, nuclear receptor, protease, and enzyme. Our results showed that most of the metabolites demonstrated high bioavailability scores ranging from 0.13 to 0.81 as enzyme inhibitors (Table 3). Moreover, some of the metabolites exhibited high bioavailability scores as nuclear receptor ligands ranging from 0.01 to 0.81. These results suggest that the observed biological anticancer activity could involve several interactions and mechanisms.

Molecular Docking
The fourteen identified metabolites from A. fragrantissima flower dichloromethane extract were docked into Tubulin [30][31][32] and Carbonic anhydrase (CA) IX enzyme [33,34], two main substrates reported to be inhibited by flavonoids, to investigate the molecular interactions using Maestro Schrödinger software. The crystal structures of Tubulin and CA IX enzyme were optimized and minimized, followed by grid generations for both binding pockets. For Tubulin docking, Colchicine (native ligand) was used as a positive control and the docking protocol was validated by docking the Colchicine into the binding site with a low root-mean-square deviation (RMSD) value relative to crystal structure. The docking scores for the fourteen identified metabolites range from −4.23 to −9.89 kcal/mol. As shown in Table 4, metabolite 11 (M11, Quercetin) demonstrated the highest docking score (−9.89 kcal/mol), followed by M1 (Cosmosiin, −9.59 kcal/mol) and M2 (Centaureidin, −9.16 kcal/mol). Moreover, M11 exhibited several interactions with the binding site including Cys241, Val181, and Asn350, and occupied a similar binding pose relative to Colchicine as illustrated in Figure 8. All the 2-D interaction maps for the metabolites with Tubulin are presented in Figure 8. Moreover, the fourteen metabolites were also docked into the CA IX crystal structure and docking scores are summarized in Table 5. The docking scores ranged from −1.30 to −5.46 kcal/mol, and zinc coordination was maintained in the most docked metabolites. Interestingly, M11 demonstrated the highest docking score (−5.46 kcal/mol) followed by M4 (−5.25 kcal/mol) and M5 (−5.21 kcal/mol). Additionally, several hydrophilic and hydrophobic interactions were formed at the binding site as depicted in Figure 9. All the 2-D interaction maps for the metabolites with CA IX are depicted in Figure 9.

Molecular Docking
The fourteen identified metabolites from A. fragrantissima flower dichloromethane extract were docked into Tubulin [30][31][32] and Carbonic anhydrase (CA) IX enzyme [33,34], two main substrates reported to be inhibited by flavonoids, to investigate the molecular interactions using Maestro Schrödinger software. The crystal structures of Tubulin and CA IX enzyme were optimized and minimized, followed by grid generations for both binding pockets. For Tubulin docking, Colchicine (native ligand) was used as a positive control and the docking protocol was validated by docking the Colchicine into the binding site with a low root-mean-square deviation (RMSD) value relative to crystal structure. The docking scores for the fourteen identified metabolites range from −4.23 to −9.89 kcal/mol. As shown in Table 4, metabolite 11 (M11, Quercetin) demonstrated the highest docking score (−9.89 kcal/mol), followed by M1 (Cosmosiin, −9.59 kcal/mol) and M2 (Centaureidin, −9.16 kcal/mol). Moreover, M11 exhibited several interactions with the binding site including Cys241, Val181, and Asn350, and occupied a similar binding pose relative to Colchicine as illustrated in Figure 8. All the 2-D interaction maps for the metabolites with Tubulin are presented in Figure 8.  Moreover, the fourteen metabolites were also docked into the CA IX crystal structure and docking scores are summarized in Table 5. The docking scores ranged from −1.30 to −5.46 kcal/mol, and zinc coordination was maintained in the most docked metabolites. Interestingly, M11 demonstrated the highest docking score (−5.46 kcal/mol) followed by M4 (−5.25 kcal/mol) and M5 (−5.21 kcal/mol). Additionally, several hydrophilic and

Absorption, Distribution, Metabolism, and Excretion (ADME) Predictions
Several pharmaceutical properties important for drug discovery were calculated for the identified metabolites using the SwissADME web server and QikProp tool. These parameters include Molecular Weight (g/mol), Log Po/w (lipophilicity), Log S (solubility), blood-brain barrier (BBB) Permeability, and gastrointestinal (GI) Absorption. As shown in Table 6, all the metabolites exhibited acceptable pharmaceutical parameters ( Figure 10) that were within Lipinski's rule of five (ROF), except for metabolites M1 and M10 that

Absorption, Distribution, Metabolism, and Excretion (ADME) Predictions
Several pharmaceutical properties important for drug discovery were calculated for the identified metabolites using the SwissADME web server and QikProp tool. These parameters include Molecular Weight (g/mol), Log Po/w (lipophilicity), Log S (solubility), blood-brain barrier (BBB) Permeability, and gastrointestinal (GI) Absorption. As shown in Table 6, all the metabolites exhibited acceptable pharmaceutical parameters ( Figure 10) that were within Lipinski's rule of five (ROF), except for metabolites M1 and M10 that violated the hydrogen bond donor/acceptor rule and molecular weight, respectively. Moreover, most of the metabolites were predicted to act peripherally with no central nervous system effects, except for M6, M7, and M13. Additionally, the majority of the identified metabolites were predicted to possess a high GI absorption, except for M1 and M12.

Cytochrome P450 (CYP) Enzyme Inhibition Profiling
The effects of the identified metabolites of A. fragrantissima flower dichloromethane extract on CYP isoenzymes were reported in Table 7 using the SwissADME webserver. Six out of fourteen of the evaluated metabolites were predicted to inhibit CYP1A2 and CYP3A4, while five metabolites were predicted to likely inhibit CYP2D6 and CYP2C9, and only one metabolite was predicted to inhibit CYP2C19, as summarized in Table 7.

Organ and Endpoint Toxicity Predictions
The toxicity of the identified metabolites was assessed using the ProTox-II webserver against hepatotoxicity, carcinogenicity, immunotoxicity, mutagenicity, and cytotoxicity. As summarized in Table 8, all the metabolites were inactive as hepatotoxic and cytotoxic. While six (M2-M4, M6, M8, M10, and M13) out of fourteen were predicted to be immunotoxic, fewer were predicted to exhibit mutagenicity (M1, M3, M9, and M11) and carcinogenicity (M6, M9, and M11).

Discussions
Natural products, including bioactive molecules from plant, animal, fungal, and microbial origins, remain a vital source for therapeutic drug discovery. Since the last century, approximately more than fifty percent of pharmaceutical drugs for humans have been derived from natural sources [35,36]. The importance of natural products as a source of cancer treatment is well recognized and documented. It is estimated that forty-nine percent of approved small molecules for cancer treatment are derived from natural products in the time period between the 1940s and 2014 [37]. Endemic species of plants in the Middle Eastern region are among the least studied species and could potentially reveal novel lead compound(s). The medicinal plant species belonging to the Achillea genus found in the Middle East region are popular treatments for several illnesses, including respiratory diseases, fever, gastrointestinal disturbances, headache, fatigue, and diabetes [17,18,20]. This study explored one local species of A. fragrantissima for new therapies against TNBC using a series of in vitro experiments to determine its potential anti-proliferative, cytotoxic, and pro-apoptotic properties.
The investigation of this current study was initiated by chemical fingerprinting for the three plant parts of A. fragrantissima (i.e., flowers, leaves, and roots) using methanol, ethanol, dichloromethane, and chloroform. The solvents were chosen to represent the polar (i.e., methanol and ethanol), moderately polar (i.e., dichloromethane), and non-polar (i.e., chloroform) extractions that were reflected in the chromatograms. The chromatograms of methanol and ethanol extracts displayed more polar phytochemical constituents, which eluted first due to the separation using the reverse-phase C18 column while the least polar constituents appeared more abundant and eluted last in the chromatograms of dichloromethane and chloroform extracts. The numerous and abundant phytochemical constituents in the DMSO-reconstituted extracts also indicated the efficiency of the extraction method using the high-power probe sonicator, and that the abundant phytochemical constituents were retained after reconstitution with DMSO solvent. The rich phytochemical compositions observed in the chromatograms were consistent with studies that isolated and characterized numerous phytochemical constituents of A. fragrantissima plant's parts using various techniques and identified several families of phytochemicals, including phenolic acids, flavonoids, lignans, terpenic lactones, and alkamides [18].
The anti-proliferative effects of A. fragrantissima plant parts-derived extracts on TNBC MDA-MB-231 cells were determined using the CellTiter-Glo™ assay. The bioassay determined the viability of cell populations by measuring the amount of adenosine triphosphate (ATP) generated by metabolically active cells. It is known that the bioassay is among the most sensitive cell viability bioassay in comparison to tetrazolium reduction, resazurin reduction, and protease activity bioassays [38]. The performance of the bioassay in this study was shown to be sensitive enough to establish the concentration-inhibition curves efficiently despite the relatively low cell seeding count. Furthermore, the variation in the inhibition concentrations values did reflect the chromatograms of the polar extracts displaying higher IC 50 values in comparison to the least polar extracts, suggesting that the least bioactive polar compounds contributed to the weight of the extracts, resulting in higher inhibition values. The inhibition concentrations were used as a guideline to select the most potent anti-proliferative extract. The A. fragrantissima flower dichloromethane extract presenting the lowest mean IC 50 value was chosen for further cell death-related molecular mechanism investigation. Furthermore, additional confirmatory experiments were required to confirm apoptosis as the main mechanism of cell death [39]. Further experiments in this study include the observation of apoptosis-related cell morphological changes, the detection of caspase-3/7 activation, and abnormal exposure of phosphatidylserine on the outer cell membrane in treated MDA-MB-231 cells. The treatment of MDA-MB-231 cells with A. fragrantissima flower dichloromethane extract clearly led to a gradual increase in cell shrinkage, in a dose-dependent manner, with the highest concentrations (i.e., 100 and 200 µg/mL) displaying similar observations as the apoptosis inducer, STS, used as a positive control. For caspase-3/7 activation, the detection was based on the fluorescent inhibitor of caspases (FLICA) methodology. The method was based on the covalent binding of fluoromethyl ketone (FMK) moiety to the center of the activated caspase by a recognition sequence for capase-3/7 (i.e., aspartic acid-glutamic acid-valine-aspartic acid) [40]. The FLICA method allowed direct visualization of activated caspase-3/7 in treated MDA-MB-231 cells as redfluorescence cells while cells without activated caspases display the nucleus only after use of the Hoechst 33342 (blue) staining. The A. fragrantissima flower dichloromethane extract clearly increased the number of activated caspase-3/7 cells to levels comparable to the protein kinase inhibitor, STS. The FACS analysis using the fluorescein isothiocyanate (FITC)labeled Annexin V/phycoerythrin (PE)-labeled PI dual staining allowed the detection of viable, early apoptosis, late apoptosis, and dead/necrotic cell populations in MDA-MB-231 cells treated with STS or different A. fragrantissima extract concentrations. The experiment confirmed the occurrence of the early and late apoptotic status by Annexin V binding to the exposed phosphatidylserine on the outer leaf of the cell membrane, which is a hallmark characterization of programmed cell death [41]. Furthermore, the PI binding to nucleic acids is used to detect cells with lost cell membrane integrity as seen in late apoptotic and dead/necrotic cell populations, which is also considered an irreversible cell death status [42].
After confirming that caspase-dependent apoptosis as a cell death mechanism occurred in the MDA-MB-231 cells exposed to A. fragrantissima flower dichloromethane extract, a deeper investigation was conducted to determine the involvement of the intrinsic apoptosis pathway using the mitochondrial permeability transition pore activation assay.
Indeed, apoptosis occurs in two pathways referred to as intrinsic and extrinsic. The intrinsic apoptosis pathway involves mitochondrial permeability transmembrane pore opening forming the apoptosome leading to caspase-9 activation, then subsequently to executioner caspases-3/7 activation [43]. The bioassay directly measured the mitochondrial permeability transition pore opening activity using acetoxymethyl ester of calcein dye, which accumulated in the cytosolic compartments and liberated into a highly-fluorescent calcein. In pro-apoptotic conditions, the cells with activated mitochondrial permeability transition pore opening appeared brightly green due to the redistribution of calcein from the cytosol into the mitochondria. The addition of the quencher, CoCl 2 , removed the remaining cytosolic fluorescence, while the mitochondrial fluorescence was maintained visible under the confocal microscopy. The treatment of MDA-MB-231 cells with A. fragrantissima flower dichloromethane extract increased the number of cells displaying activated mitochondrial permeability transition pore opening in a dose-dependent manner, confirming the involvement of the mitochondria-dependent intrinsic apoptosis pathway. Interestingly, our findings characterizing the main hallmarks of apoptosis, including caspase-3/7 detection, apoptosis status determination by FACS, and mitochondrial membrane transition pore opening activation, exhibited a similar pattern. The results clearly displayed an induction of apoptosis in TNBC MDA-MB-231 cells, which slightly and gradually increased in response to the 25 and 50 µg/mL treatment to reach the highest pro-apoptotic changes in response to 100 and 200 µg/mL of A. fragrantissima flower dichloromethane extract.
Numerous studies have reported the anticancer potencies of many species belonging to the Achillea genus, including the widely-consumed native North American species of A. millefolium [44]. Only a few studies reported A. fragrantissima crude extracts with anticancer properties using cancer cell lines. For instance, Sathiyamoorthy et al. tested A. fragrantissima arial part aqueous extract on sensitive and resistant melanoma cell lines and reported a significant cytotoxicity [45]. Alenad et al. tested A. fragrantissima leaves methanolic extract against K562 (myeloid leukemia), Jurkat (T-cell leukemia), and HepG2 (hepatocarcinoma) cells and reported drastic morphological changes and viability reduction in K562 cells [46]. Binbreak et al. treated HepG2, A549 (lung carcinoma), HCT-116 (colorectal carcinoma), and MCF-7 (hormone-dependent breast cancer) cells with methanolic extract of A. fragrantissima leaves and reported cytotoxic bioactivity of the extract against A549 cells with induction of apoptosis, including activation of caspase-3, increased p53 and Bax expression, and down-regulation of Bcl-2 expression [47]. Several studies reported the isolation of A. fragrantissima phytochemicals and performed further experiments for the evaluation of their anticancer properties using several cell lines. For example, Awad et al. isolated several phenolic compounds, including piceol, eupatilin 7-methyl ether, chrysosplenol D, cirsiliol, and cirsimaritin, from A. fragrantissima leaves methanolic extract tested on MCF-7, HepG2, A549, HeLa (cervix adenocarcinoma), and PC3 (prostatic adenocarcinoma) cell lines and presented IC 50 values ranging from 3.2 to 28.3 µg/mL [48]. To the best of our knowledge, the present study is the first to determine the in vitro cytotoxicity of different A. fragrantissima plant parts (i.e., flowers, leaves, and roots) crude extracts against TNBC using MDA-MB-231 cell line and to investigate the pro-apoptotic effect of A. fragrantissima flower dichloromethane extract.
Our experimental observations suggested that A. fragrantissima possesses promising anticancer activity against TNBC. After chemical analysis of the A. fragrantissima flower dichloromethane extract, we attempted to correlate the observed experimental findings with the identified metabolites. Several computer-aided drug discovery (CADD) approaches were utilized to provide accurate and fast predictions of several biological and pharmaceutical properties required for the further development of novel drug entities [49,50]. Therefore, PASS online webserver was utilized to validate the anticancer activity of A. fragrantissima identified metabolites. Interestingly, all fourteen identified metabolites demonstrated moderate to high anticancer activity confirming the observed experimental results. Moreover, the most probable molecular targets for the identified metabolites were predicted and the analysis suggested that certain enzymes and several other targets could contribute to the observed activity.
An additional molecular docking study was performed to assess the possibility of the identified metabolites including flavonoids, to bind and inhibit Tubulin [30][31][32] and CA IX enzyme [33,34]. The selection of these two targets was based on the previously reported in vitro studies that confirmed the inhibition by flavonoids, particularly, Quercetin. It is worth noting that the CA IX enzyme is highly expressed in the TNBC MDA-MB-231 cell line, which is reported to contribute to the aggressiveness and dysfunction of this type of cancer [51][52][53]. Our docking results were consistent with the previously reported studies in which Metabolite 11 (M11, Quercetin) possesses the highest docking score among the other metabolites against the two targets, suggesting the validity of the protein selection. Moreover, it has been demonstrated by several studies that inhibition of Tubulin polymerization and CA IX enzyme serve as a new and novel mechanism for promising anticancer activity [54][55][56][57][58][59][60][61][62].
One of the main aspects of drug design and development is to assess and evaluate the pharmacokinetics or drug-likeness properties relative to Lipinski's rule of five required for orally active drugs [63]. Twelve out of fourteen metabolites passed the rules indicating a good pharmaceutical profile. Additionally, one of the main mechanisms for drug-drug interactions has resulted from the inhibition of one or several CYP isoenzymes involved in the metabolism of drugs. Therefore, the inhibition profile for the fourteen metabolites was evaluated and our results demonstrated that not all the metabolites inhibited the CYP isoenzymes, except for a few that were predicted to inhibit CYP1A2, CYP3A4, CYP2D6, and CYP2C9. The toxicity of the identified metabolites demonstrated immunotoxicity for six metabolites, while none were predicted to exhibit hepatotoxicity and cytotoxicity.

Collection and Authentication of A. fragrantissima
The traditional herbal medicine of A. fragrantissima flowers, leaves, and roots was wildcrafted at Albatin (in Arabic, ) dam area near Jalajil (in Arabic, ) city  Figure S1).

Chromatographic Fingerprinting of A. fragrantissima Extracts
The fingerprinting of A. fragrantissima extracts was performed using our previously published method with modifications [65]. The major components of A. fragrantissima were separated using an Agilent (Agilent Technologies Inc., Santa Clara, CA) 1260 infinity HPLC-UVD and a C-18 Kinetex ® (250 × 4.6 mm, 5 µm particle size) column (Phenomenex Inc., Torrance, CA, USA). The multi-wave UVD was set at wavelengths ranging from 225 to 375 nm at 25 nm segments without reference. The gradient programming of methanol and ultra-pure water was as follows: 5 to 100% methanol, (0 to 35 min), 100% methanol, (35 to 40 min), and 100 to 5% methanol, (40 to 45 min) at flow rate of 1 mL/min. The injection volume was 5 µL at ambient temperature and the total run time was 45 min.

TNBC MDA-MB-231 Cell Culture and Treatment
The TNBC MDA-MB-231 (catalog number, HTB-26) cell line was provided by American Type Culture Collection Inc. (Manassas, VA, USA). The cells were cultured in complete DMEM Glutamax-I medium containing 10% heat-inactivated FBS, 1% penicillin G (100 IU/mL), and streptomycin (100 µg/mL) solution. The incubation was set at 37 • C in 5% CO 2 and 95% relative humidity. The culture medium was changed every two to three days and sub-cultured when the cell population density reached 70 to 80% confluency. The cells were exposed to various concentrations of A. fragrantissima extracts and the appropriate controls; including 1% DMSO (negative control) and 1 µM STS (potent apoptosis inducer) were used.

Determination of MDA-MB-231 Cell Viability Using the CellTiter-Glo™ Assay
The cell viability was determined using Promega (Madison, WI) CellTiter-Glo™ assay according to the manufacturer's instructions as described in [66]. Briefly, the MDA-MB-231 cells (400/cm 2 ) were seeded in opaque white bottom 96-well plates (Costar, Thermo Fisher Scientific, Waltham, MA, USA). The next day, the untreated cells and the cells treated with various concentrations (0.05-2000 µg dry extract weight/mL) of each solvent extract were incubated for 48 h. Then, the plate was left at room temperature for 30 min and 100 µL of CellTiter-Glo™ reagent were added to each well and mixed on a shaker for 2 min. Luminescence was read using the Perkin Elmer ® (Waltham, MA, USA) EnVision ® multilabel plate reader.

Evaluation of the Hallmarks of Apoptosis
The shrinkage of cells was observed using a ZEISS inverted phase-contrast microscope (Peine, Germany). The activity of pro-apoptotic executioner caspase-3/7 and of the mitochondrial permeability transition pore opening in the untreated and treated cells were assessed using Image-iT™ LIVE Red caspase-3/7 detection kit and Image-IT™ LIVE Mitochondrial Transition Pore Assay Kit (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA), respectively, according to the manufacturer's instructions and as previously described in [67]. The cells with activated caspase-3/7 (red fluorescence) and apoptotic cells (whole green fluorescence) with mitochondrial outer membrane permeabilization were observed under the confocal microscopy. The abnormal phosphatidylserine exposure was determined using the Becton Dickinson (BD) Annexin V apoptosis detection kit (Becton Dickinson, San Jose, CA, USA) as described in [68]. The apoptotic status of 10,000 cells based on the dual staining of Annexin V conjugated to FITC and PI conjugated to PE was analyzed using BD FACSDiva™ 6.0 software on the BD FACSCanto™ II flow cytometer.

Metabolite Identification Using ESI-LC-QTOF
The method of separation and identification of A. fragrantissima flower metabolites in dichloromethane extract was performed using an Agilent (Santa Clara, CA, USA) 1260 Infinity HPLC system coupled to Agilent 6530 quadrupole time-of-flight (Q-TOF) as described earlier in [69]. Briefly, the separation was performed using an Agilent SB-C18 column (4.6 mm × 150 mm, 1.8 µm) with the following elution gradient; 0 to 2 min, 5% B; 2 to 17 min, 5 to 100% B; 17 to 21 min, 95% B; 21 to 25 min, 5% B, using mobile-phase A (0.1% formic acid in water) and mobile-phase B (0.1% formic acid in methanol). The flow rate was set at 250 µL/min and injection volume was 10 µL. The scanning range was set at 50-800 (m/z), gas temperature at 300 • C gas flow of 8 L/min, nebulizer pressure of 35 psi, sheath gas temperature at 350 • C, and sheath gas flow rate of 11 L/min. The data were generated by the Agilent MassHunter (version B. 06.00) qualitative analysis software.

Anticancer Activity Predictions
The activity spectra for substances (PASS) online service provides online predictions of a variety of biological activities including anticancer predictions (http://www.way2drug. com/passonline/info.php, accessed on 25 April 2020). The Simplified Molecular Input Line Entry System (SMILES) of the chemical structures of the identified metabolites was utilized to generate the predictions using a database containing 250,000 biological activities. The generated results are represented as P a (probability of active molecule) and P i (probability of inactive molecule), ranging from 0.00 to 1.00. The higher P a value for a certain biological activity indicates the molecule is predicted to be highly active experimentally P a > 0.7), while P a ranging from 0.5 to 0.7 suggests a moderate probability for experimental activity, and P a lower than 0.5 indicates negligible experimental activity.

Molecular Target Predictions
The biological targets for the identified metabolites were predicted using the Cheminformatics tool Molinspiration webserver (https://www.molinspiration.com/, accessed on 25 April 2020). Each metabolite was evaluated as GPCR Ligand, Ion Channel modulator, Kinase Inhibitor, Nuclear Receptor Ligand, Protease Inhibitor, and Enzyme Inhibitor. The bioactivity scores were classified into three main classifications, including active (score above 0.00), moderately active (−0.50 to 0.00), and inactive (score below −0.50).

Molecular Docking
The molecular docking study was performed using Maestro Schrödinger software (Release 2022-1). The crystal structures of the Tubulin-Colchicine complex (Protein Data Bank identifier, PDB ID 4O2B) and human CA IX in complex with 5-[1-(4-Methylphenyl)-1H-1,2,3-triazol-4yl]thiophene-2-sulfonamide (PDB ID Y0R) were downloaded from the protein data bank (https://www.rcsb.org/, accessed on 25 April 2020). The proteins were prepared using the protein preparation wizard followed by receptor grid generation using the Glide tool. The 2D structures of the identified metabolites were converted into 3D and minimized using the LigPrep tool. All the prepared metabolites were docked using the Glide tool and the 2-D interaction map with docking scores was used to rank the docked metabolites into the two targets.

ADME Properties Predictions
The online accessible webserver SwissADME (http://www.swissadme.ch/, accessed on 25 April 2020) and commercial QikProp (Schrödinger software, Release 2022-1) tools were utilized to predict the ADME properties for the identified metabolites. Several important pharmaceutical properties were selected for evaluation including Molecular Weight (g/mol), Log Po/w (lipophilicity), Log S (solubility), BBB Permeability, GI Absorption, and violations of ROF.

CYP Enzyme Inhibition Predictions
To evaluate if the identified metabolites are substrates for the CYP enzymes, the SwissADME tool was used to evaluate the inhibition profile against five main isoenzymes (CYP1A2, CYP2C19, CYP2C9, CYP2D6, and CYP3A4). The predicted results could enrich the literature regarding the possible drug-drug interactions, toxicity, or unwanted side effects of the metabolites that result from the inhibition of CYP enzymes.

Organ Toxicity and Safety Predictions
The prediction of various toxicity endpoints was performed using a freely accessible webserver ProTox-II (https://tox-new.charite.de/protox_II/, accessed on 25 April 2020). For each metabolite, several toxicity endpoints were evaluated including hepatotoxicity, cytotoxicity, carcinogenicity, mutagenicity, and immunotoxicity. The ProTox-II predicted results were based on reported in vitro and in vivo assays.

Data and Statistical Analyses
For the cell viability experiments, the extract concentration versus inhibition curves were obtained from plotting the log 10 treatment concentrations versus the percentage inhibition of MDA-MB-231 cell viability. The IC 50 values were extrapolated from 50% maximum viability inhibition. The inhibition and effective doses were measured using the log-inhibition variable-slop (three or four-parameter) model in GraphPad Prism (San Diego, CA, USA) software version 5.04. For the FLICA caspase-3/7 and mitochondrial membrane transition pore opening activation assay, several images were counted for each individual treatment and averaged. Student's t-test and one-way ANOVA statistical analysis were used to determine if the treatment is statistically significant to no treatment when p value is less than 0.05. The measured variables in this study are expressed as mean ± SD.

Conclusions
This study demonstrated the anticancer properties of A. fragrantissima flower dichloromethane extract through apoptosis induction using a series of in vitro apoptosis-related experiments on TNBC MDA-MB-231 cells. The findings revealed caspase-dependent apoptosis as the mechanism of cell death with the involvement of the intrinsic mitochondrial apoptosis pathway. Using in silico approaches, fourteen metabolites were extensively evaluated for their anticancer activity, which could be correlated to several mechanisms involving CA IX and Tubulin polymerization inhibition. Molecular docking investigations revealed that Quercetin was the most active hit compared to other evaluated metabolites that were consistent with previously reported studies with moderate CYP enzyme inhibition and toxicity. These findings suggest that A. fragrantissima flower dichloromethane extract is a promising source containing anticancer agents that could be further structurally modified and optimized to discover novel drug entities. Further in vitro and in vivo studies are required to isolate and characterize the phytochemicals and determine the potential bioactive compound(s) against TNBC.