Cytotoxic and Anti-Plasmodial Activities of Stephania dielsiana Y.C. Wu Extracts and the Isolated Compounds

Natural products remain a viable source of novel therapeutics, and as detection and extraction techniques improve, we can identify more molecules from a broader set of plant tissues. The aim of this study was an investigation of the cytotoxic and anti-plasmodial activities of the methanol extract from Stephania dielsiana Y.C. Wu leaves and its isolated compounds. Our study led to the isolation of seven alkaloids, among which oxostephanine (1) is the most active against several cancer cell lines including HeLa, MDA-MB231, MDA-MB-468, MCF-7, and non-cancer cell lines, such as 184B5 and MCF10A, with IC50 values ranging from 1.66 to 4.35 μM. Morever, oxostephanine (1) is on average two-fold more active against cancer cells than stephanine (3), having a similar chemical structure. Cells treated with oxostephanine (1) are arrested at G2/M cell cycle, followed by the formation of aneuploidy and apoptotic cell death. The G2/M arrest appears to be due, at least in part, to the inactivation of Aurora kinases, which is implicated in the onset and progression of many forms of human cancer. An in-silico molecular modeling study suggests that oxostephanine (1) binds to the ATP binding pocket of Aurora kinases to inactivate their activities. Unlike oxostephanine (1), thailandine (2) is highly effective against only the triple-negative MDA-MB-468 breast cancer cells. However, it showed excellent selectivity against the cancer cell line when compared to its effects on non-cancer cells. Furthermore, thailandine (2) showed excellent anti-plasmodial activity against both chloroquine-susceptible 3D7 and chloroquine-resistant W2 Plasmodium falciparum strains. The structure–activity relationship of isolated compound was also discussed in this study. The results of this study support the traditional use of Stephania dielsiana Y.C. Wu and the lead molecules identified can be further optimized for the development of highly effective and safe anti-cancer and anti-plasmodial drugs.

a unique characteristic phenotype such as a transient G2/M arrest and chromosome segregation failure that may eventually lead to aneuploidy [34]. Although the aporphine alkaloids isolated from Stephania dielsiana Y.C. Wu [37] and the cytotoxicity of some of these alkaloids were previously reported [38], their mechanism of action for the antiproliferative and antimalarial activity has not been elucidated.
In this study, we systematically evaluated the anti-proliferative and anti-plasmodial effects of the crude methanol extracts of Stephania dielsiana Y.C. Wu (S. dielsiana) leaves, its fractions and isolated compounds on the human cancer and non-cancer cells as well as chloroquine-susceptible 3D7 and chloroquine-resistant W2 P. falciparum parasites. A bioassay-guided fractionation led to the isolation of seven alkaloids, among which oxostephanine (1) and thailandine (2) show strong anticancer and/or anti-plasmodial activities. The mechanism of action of oxostephanine (1) on human cancer cell line HeLa was investigated. This is the first time that the functional analysis of oxostephanine identified Aurora kinase as a putative cellular target, and we determined that oxostephanine (1) interacts in the ATP binding pockets of Aurora A and B kinases. Moreover, the anti-plasmodial activity of isolated alkaloids was also evaluated in this study. Our data from an in vitro study showed that thailandine (2) possess excellent anti-plasmodial activity against both chloroquine-susceptible 3D7 and chloroquine-resistant W2 P. falciparum strains. To our knowledge, this is the first report that oxostephanine, thailandine and other alkaloid compounds from Vietnamese S. dielsiana Y.C. Wu possess both anticancer and anti-plasmodial activities.

Bioassay-Guided Fractionation and Isolation of S. dielsiana Y.C. Wu
The methanol crude (MB2L) extract showed good activity against HeLa, MDA-MB-231, MDA-MB-468, and MCF7 cancer cells lines (Table 1). Further fractionation of the methanol extract showed that the dichloromethane (MB2L-CH) and butanol (MB2L-B) fractions possess high activity against cancer cells with IC 50 values in the range of 0.57 to 14.35 µg/mL, while the hexane (MB2L-H) fraction was not active (Table 1). In particular, MB2L-CH fraction was effective on all of the cell lines examined, although it did not show any selectivity between four cancer cell lines and two non-cancer cell lines (184B5 and MCF10A). The MB2L-B fraction was effective against only MDA-MB-468; however, it is very selective against the cancer cell line when compared with its cytotoxicity against 184B5 (3.2-fold) and MCF10A (7.8-fold) non-cancer cells (Table 1). In addition to cancer cells, the MB2L-CH fraction was also quite potent against both chloroquine-susceptible 3D7 (IC 50 , 4.5 µg/mL) and chloroquine-resistant W2 (IC 50 , 5.8 µg/mL) P. falciparum strains ( Table 1).

In Vitro Antiproliferative and Structure-Activity Relationship (SAR) of Isolated Compounds
All seven isolated pure compounds were evaluated for their anti-proliferative activity against four cancer (HeLa, MDA-MB231, MDA-MB-468 and MCF7) and two non-cancer (184B5 and MCF10A) cell lines (Table 2A). The selectivity index (SI) of cancer versus non-cancer cell lines was calculated (Table 2B) for each of the pure compounds with a calculated IC50. The selectivity of most of the compounds, as well as the control compounds chloroquine and paclitaxel is in the range of 1-2, with some selectivity seen in MCF7 and MDA-MB-468 in comparison with MCF10A (Table 2B).

In Vitro Antiproliferative and Structure-Activity Relationship (SAR) of Isolated Compounds
All seven isolated pure compounds were evaluated for their anti-proliferative activity against four cancer (HeLa, MDA-MB231, MDA-MB-468 and MCF7) and two non-cancer (184B5 and MCF10A) cell lines (Table 2A). The selectivity index (SI) of cancer versus non-cancer cell lines was calculated (Table 2B) for each of the pure compounds with a calculated IC 50 . The selectivity of most of the compounds, as well as the control compounds chloroquine and paclitaxel is in the range of 1-2, with some selectivity seen in MCF7 and MDA-MB-468 in comparison with MCF10A (Table 2B). A study by structure-bioactivity relationship on these isolates showed that aporphine alkaloids (1-5) showed anti-proliferative activities against cancer and non-cancer cell lines in the range of 0.7-74 µM, whereas protoberberine (6) and tetrahydroprotoberberine (7) did not show any effect ( Figure 1, Table 2A). This finding suggests that the activity of the butanol and dichloromethane fractions can be attributed to the aporphine alkaloids present in each fraction, which is consistent with previous reports on the cytotoxicity on this type of alkaloids isolated from Stephania verosa [31,34]. Of five aporphine alkaloids (1)(2)(3)(4)(5), oxostephanine (1) is the most active against all of the cell lines examined with IC 50 values ranging from 1.66 to 4.35 µM ( Figure 2, Table 2A), although thailandine (2) and stephanine (3) also show strong activities (Table 2A). These three active aporphine alkaloids have similar structures ( Figure 1). Oxostephanine (1) maintains most of the stephanine (3) structure with the addition of a 7-oxo group and a double bond between carbon 6a and nitrogen ( Figure 1). Comparing the two compounds, it is clear that these changes increase the effectiveness of oxostephanine (1) by 2-fold (HeLa, 3.33 µM vs. 1.73 µM) to almost 4-fold (184B5, 6.25 µM vs. 1.66 µM) against the cell lines. However, these changes result in less selectivity against cancer cells when compared to its cytotoxicity on non-cancer 184B5 and MCF10A cells (Table 2A,B). An addition of a methyl group linked to the nitrogen atom of oxostephanie (1) (which forms a quaternary nitrogen atom and subsequent addition of a positive charge) shown in the thailandine (2) reduces the effectiveness of the molecule against most cancer and non-cancer cells, which corroborates with previous report by Makarasen [31]. However, thalandine (2) is very effective against MDA-MB-468 with the IC 50 of 0.78 µM, where it is 3.9-fold (0.78 µM vs. 3.02 µM) and 6.4-fold (0.78 µM vs. 5.01 µM) more selective over 184B5 and MCF10A non-cancer cells, respectively (Table 2A). This raises the possibility that the selectivity shown by MB2L-B (Table 1) may be the effect of thailandine (2). The effectiveness of stephanine (3) against certain cancer cell types may be substantially better with the addition of a 7-oxo group or the modification of nitrogen without positive charge at the position of carbon-6 to a quaternary nitrogen atom (forming a positive charge). Thailandine (2) is electron-deficient, and the positive charge of the molecule may make it more effective against a certain sub-types of cancer cells, potentially in a chloroquine-like weak base manner [39]. According to Boyd [40], a compound can be considered as active if its IC 50 is less than 100 µM. In that sense, crebanine (4) and O-methylbulbocapnine (5) are also considered weakly active because their IC 50 values are in the ranges of 17-48 µM and 39-73 µM, respectively (Table 2A), consistently with previous results [34]. The antiproliferative activities of oxostephanine (1), thailandine (2) and crebanine (4) against several cancer cell lines have been reported by Makarasen [31]; however, the MCF-10A and 184B5 non-cancer cell lines were not included in their study. The antiproliferative of the alkaloids isolated from S. dielsiana Y.C. Wu was investigated for the first time on two human cancer cell-lines of BT174 (breast) and HCT116 (colon) [38]. They mentioned thailadine (2), stephanine (3) and crebanine (4) are active on BT474 cell lines. However, the structure-activity relationship of these isolated compounds was not discussed [38].

Oxostephanine Leads to A Transient G2/M Arrest and Apoptotic Cell Death
Since oxostephanine (1) was the most active compound against all of the cancer cell lines examined, its mode of action was investigated. The effects of oxostephanine on a population level was examined using flow cytometry. An initial study of cell cycle distribution showed that the majority of cells (HeLa) were arrested at G2/M or accumulated as multiploidy after 48 h postoxostephanine (1) treatment ( Figure 3A). By 72 h post-treatment, a substantial portion of the cell population contained sub-G1 DNAs with profile that is typical for apoptotic cells. These data suggest that oxostephanine (1)

Oxostephanine Leads to A Transient G2/M Arrest and Apoptotic Cell Death
Since oxostephanine (1) was the most active compound against all of the cancer cell lines examined, its mode of action was investigated. The effects of oxostephanine on a population level was examined using flow cytometry. An initial study of cell cycle distribution showed that the majority of cells (HeLa) were arrested at G2/M or accumulated as multiploidy after 48 h post-oxostephanine (1) treatment ( Figure 3A). By 72 h post-treatment, a substantial portion of the cell population contained sub-G1 DNAs with profile that is typical for apoptotic cells. These data suggest that oxostephanine (1) induces a transient G2/M arrest and apoptotic cell death, reminiscent of previous results for the structurally similar stephanine (3) [34].
To confirm the mode of cell death, we carried out an Annexin V staining assay with live cells. As shown in Figure 3B-D, MDA-MB-231 cells treated with oxostephanine started to undergo apoptosis within 16 h post-treatment. The number of apoptotic cells rapidly increased until 48 h, from which time it remained relatively static until 72 h post-treatment. Compared to camptothecin, a well-known apoptosis inducer (as a positive control in this study), oxostephanine (1) showed a much stronger apoptotic effect at 10 µM ( Figure 3C, 1-way ANOVA, p < 0.05). This finding supports the previous study [41], which showed oxostephanine (1) possess a selective toxicity in HeLa cells and a marked reduction in the mitotic index for tumoral cells. However, the mechanism of action of this compound was not investigated in the above study.

Oxostephanine Inhibits Aurora Kinase Activity
Inactivation of Aurora kinases A and B by specific inhibitors leads to the induction of spindle assembly checkpoint (SAC), arresting cells at mitosis. However, the SAC checkpoint is not permanent, as it is soon overridden and cells exit mitosis without proper chromosome segregation [42]. The inhibition of Aurora A or Aurora B may eventually induce apoptosis or DNA reduplication without undergoing cytokinesis [42]. Data shown in Figure 3 are reminiscent of the previous observations in human cells with ablated Aurora kinases. Therefore, we examined whether Aurora kinases are the targets of oxostephanine (1) using Western blot analysis. As shown in Figure 4A,B, HeLa cells treated with oxostephanine (1) for 12 h dramatically downregulated the phosphorylation of histone H3 on Ser-10, a known molecular target of Aurora kinases [43]. The effect of oxostephanine on autophosphorylation of all three Aurora kinases in HeLa cells arrested in mitosis was also investigated ( Figure 4C). Data from our Western blot also showed that upon treatment of cells, Aurora A, B, and C phosphorylation was reduced, showing that oxostephanine (1) inhibits the auto-phosphorylation of all three Aurora kinases ( Figure 4C). These data indicate that Aurora kinases are indeed a target of oxostephanine.
Molecules 2020, 25, x FOR PEER REVIEW 8 of 17 kinases are the targets of oxostephanine (1) using Western blot analysis. As shown in Figure 4A,B, HeLa cells treated with oxostephanine (1) for 12 h dramatically downregulated the phosphorylation of histone H3 on Ser-10, a known molecular target of Aurora kinases [43]. The effect of oxostephanine on autophosphorylation of all three Aurora kinases in HeLa cells arrested in mitosis was also investigated ( Figure 4C). Data from our Western blot also showed that upon treatment of cells, Aurora A, B, and C phosphorylation was reduced, showing that oxostephanine (1) inhibits the autophosphorylation of all three Aurora kinases ( Figure 4C). These data indicate that Aurora kinases are indeed a target of oxostephanine. (A) Histone H3 phosphorylation on Ser-10 (p-H3 S10 ) is substantially downregulated in response to oxostephanine 1 for 12 h. (B) Quantification of histone H3 S10 phosphorylation (p-H3 S10 ). P-H3 S10 was normalized to total histone H3 protein (H3). (C) The autophosphorylation of Aurora kinases A, B and C is inhibited in the cells treated with 20 μM oxostephanine for 12 h. Total histone H3 protein was used as a control.

Oxostephanine Inhibits Aurora Kinase Activity in Vitro and is Predicted to Bind to the ATP Binding Pocket of both Aurora A and B
In order to further characterize the activity of oxostephanine (1) and determine if there is activity of oxostephanine (1) against aurora kinases, we carried out an in vitro kinase assay against only Aurora A and B kinase to determine if oxostephanine (1) can inhibit either of the kinases. As expected, data from an in vitro kinase assay showed that one half of the kinase activities were inhibited (IC50) by oxostephanine (1) at 422 ± 130 nM to 952 ± 467 nM for Aurora kinases A and B, respectively ( Figure  5A). Employing in silico modeling, we found that oxostephanine would interact with the ATP binding pocket of Aurora A (crystal structure 4O0U) when docked blindly to the entire structure of Aurora A (Figure 5B-top). Oxostephanine would 'lock' in the hydrophobic pocket, where it would interact with the fluorophenyl Leu-210 residue. As well, oxostephanine is also likely to form hydrogen bonds with Glu-211 and Ala-213, the two amino acids that interact with the adenine of ATP (A) Histone H3 phosphorylation on Ser-10 (p-H3 S10 ) is substantially downregulated in response to oxostephanine 1 for 12 h. (B) Quantification of histone H3 S10 phosphorylation (p-H3 S10 ). P-H3 S10 was normalized to total histone H3 protein (H3). (C) The autophosphorylation of Aurora kinases A, B and C is inhibited in the cells treated with 20 µM oxostephanine for 12 h. Total histone H3 protein was used as a control.

Oxostephanine Inhibits Aurora Kinase Activity In Vitro and Is Predicted to Bind to the ATP Binding Pocket of Both Aurora A and B
In order to further characterize the activity of oxostephanine (1) and determine if there is activity of oxostephanine (1) against aurora kinases, we carried out an in vitro kinase assay against only Aurora A and B kinase to determine if oxostephanine (1) can inhibit either of the kinases. As expected, data from an in vitro kinase assay showed that one half of the kinase activities were inhibited (IC 50 ) by oxostephanine (1) at 422 ± 130 nM to 952 ± 467 nM for Aurora kinases A and B, respectively ( Figure 5A).
Employing in silico modeling, we found that oxostephanine would interact with the ATP binding pocket of Aurora A (crystal structure 4O0U) when docked blindly to the entire structure of Aurora A ( Figure 5B-top). Oxostephanine would 'lock' in the hydrophobic pocket, where it would interact with the fluorophenyl Leu-210 residue. As well, oxostephanine is also likely to form hydrogen bonds with Glu-211 and Ala-213, the two amino acids that interact with the adenine of ATP in the pocket. This suggests that oxostephanine would occupy much of the same space as one ATP molecule and, thus, effectively compete with it for binding to the pocket. Similarly, oxostephanine also interacts with the hydrophobic ATP binding pocket of Aurora kinase B (4AF3) (Figure 5C-top). Oxostephanine is predicted to bind to the ATP binding pocket of Aurora kinases A. Two dimensional representations generated by LigPlot+ are shown of the best binding energy poses of oxostephanine based on Autodock results. One-letter and three-letter codes are elements and amino acids, respectively. Green dashed lines denote H-bonds, and number is the distance in angstrom. Red half circles denote hydrophobic interactions. Predicted binding energy is indicated in blue. Aurora A is based on crystal structure 4O0U. (C) Oxostephanine is predicted to bind to the ATP binding pocket of Aurora kinase B. The representation is the same as above and Aurora B is based on crystal structure 4AF3.
Because of their anti-proliferative activity and structural similarity to oxostephanine (1), we also examined how stephanine (3) binds to Aurora A kinase. As shown in Figure 5B, they both bind favorably to the ATP binding site, although not as favorably as oxostephanine (1). This is likely due to the inability of stephanine (3) to form a hydrogen bond with any residues in the pocket, and in particular the Glu-211 and Ala-213 ATP interacting residues ( Figure 5B-bottom). Stephanine (3) does interact with the ATP pocket through hydrophobic interactions, and notably with the Leu-210 in the fluorophenyl pocket. When blind docking was carried out to the human Aurora B kinase (4AF3), we found that the two aporphine alkaloids interact with the hydrophobic ATP binding pocket ( Figure 5C). Oxostephanine's binding was more energetically favorable than stephanine. In part, the carbonyl group and aromatic nitrogen can form a hydrogen bond with Ala-157 ( Figure 5C-top). However, as with the Aurora A ATP pocket, stephanine (3) does not form a hydrogen bond ( Figure 5C-bottom). The predicted binding of oxostephanine in the ATP pockets of both Aurora A and B, and by extension Aurora C (since it shares much similarity with Aurora B, but does not have a published crystal structure that we are aware of), could be a reason why it has higher efficacy against cancer cells than stephanine. Based on the cell cycle phenotypes, in vitro kinase assay and the in-silico modeling, it appears that the two aporphine alkaloids extracted from S. dielsiana Y.C. Wu are active in large part because they are Aurora kinase inhibitors.
Although oxostephanine (1) is more effective than stephanine (3), it also is cytotoxic towards proliferating non-cancer models MCF10A and 184B5 at similar concentrations, suggesting that oxostephanine may have off-target cytotoxic effects or may have a narrow therapeutic window. Alternatively, MCF10A and 184B5 are still proliferating cells, even though they are not tumorigenic, and thus would require Aurora kinases for proper cell division. Thus, the low therapeutic index observed may reflect the inhibition of Aurora kinase function in those cell lines as well. It does highlight the challenge within the Aurora kinase inhibitor development field. Since no Aurora kinase inhibitor has made it past Phase III clinical trials as of now, there may be limitations in targeting Aurora kinases in general, or specifically Aurora A or B as a single agent. In order to optimize the efficacy and selectivity of oxostephanine against cancer cells, a combinatorial approach could be employed based on what has been found for other Aurora kinase inhibitors. For example, the inhibition of Aurora A in combination with the PI3Kα inhibitor BYL719, enhances apoptosis in breast cancer models [44]. Taken together, our data suggests that oxostephanine (1) provides a lead structure that could be optimized as a novel therapeutic in breast cancer and potentially other cancers.

In Vitro Anti-Plasmodial Activity of Isolated Compounds from S. dielsiana Y.C. Wu
Both MB2L-CH and MB2L-B fractions showed similar activity with IC 50 values in the range of 4.5 to 7.9 µg/mL and 5.8 to 7.1 µg/mL against chloroquine-sensitive 3D7 and chloroquine-resistant W2 strain, respectively (Table 1). Of the seven alkaloid isolates, thailandine (2), crebanine (4), O-methylbulbocapine (5), and palmatine chloride (6) show strong activity against both 3D7 and W2 (Table 3). Thailandine (2) is the most active alkaloid on both the 3D7 and W2 P. falciparum strains with IC 50 of 0.24 ± 0.04 µM and 0.22 ± 0.02 µM, respectively. Thailandine acted equally against chloroquine-susceptible and chloroquine-resistant P. falciparum parasites, suggesting that its antimalarial activity was independent on the mechanism of resistance to chloroquine. In contrast, stephanine (3) is not effective on chloroquine-resistant W2 strain, albeit very effective on chloroquine-sensitive 3D7. Thailandine (2) showed a low selectivity index (SI) ranged from 12 to 23 between cells and parasites (Table 3). When compared to its cytotoxicity against human cells, thailandine's SI against P. falciparum is in the range of 12.6-fold (3D7 vs. 184B5) to 22.8-fold (W2 vs. MCF10A) ( Table 3), indicating that it may be relatively safe. Our data thus show clearly that thailandine (2) is more active against P. falciparum parasites than stephanine (3), although the strong anti-malarial activity of stephanine (3) was recognized previously [32,34,45]. Crebanine (4), O-methylbulbocapine (5) and palmatine chloride (6) also show decent anti-plasmodial activities against both 3D7 and W2 with a low cytotoxic activity against 184B5 and MCF10A cells and, thus, with good selectivity (especially selective index SI > 20 on 3D7 strain, Table 3). Therefore, these compounds can also be good leads for further development. Oxostephanine (1) and tretrahydraoplamatine (7) were not active against P. falciparum parasites with IC 50 value in the range of 63.9 to 275.5 µM against 3D7 and 215.5 to 226 µM against W2, respectively (Table 3). Human Aurora kinase inhibitors, including hesparadin and its analogs, have already displayed excellent potency (from 0.01 to 1.6 µM µM) against P. falciparum parasites (chloroquine susceptible D6 strain) [8]. Oxostephanine (1), which shows the best anti-proliferative activity against cancer cells (HeLa, MCF7, MDA-MB-231, and MDA-MB-468) and which inhibits Aurora kinase activity, demonstrated poor activity against 3D7 and W2 strains. The modes of action between anti-proliferative and anti-plasmodial activities seem to be different. Thailandine (2), which shows the best anti-plasmodial activity, also displays significant cytotoxicity against cancer cells. The target of thailandine is still unknown in P. falciparum. But, the lysosomotropic properties of thailandine, due to quaternary nitrogen, could be involved in anti-plasmodial activity. Thailandine could inhibit endocytosis in parasites, like chloroquine, mefloquine, or artemisinin [46].

Thailandine as A Potential Anticancer Agent
Thailandine (2) shows significant cytotoxicity against cancer and non-cancer cell lines with IC 50 values ranging from 0.78 to 7.11 µM (Table 2A). This compound is the most active compound on both P. falciparum strains with IC 50 of 0.22 µM against 3D7 and 0.24 µM against W2 (Table 3). However, we were unable to isolate enough thailandine (2) from the MB2L-B fraction to pursue mechanistic studies. Thailandine remains of interest as both an anti-malarial and anti-cancer agent. Since thailandine (2) is electron-deficient and charged, it is possible that the positive charge of the molecule makes it more effective against certain sub-types of cancer [39]. Based on our work and others [31], thailandine may have an additional target that is specific to cells that have high expression of EGFR and/or are dependent on endocytic signaling. Since endocytic signaling depends on maintaining different pH within the endosome and lysosome, this pathway is sensitive to drugs that can act as weak bases and altering proton concentration. The quaternary nitrogen of thailandine (2) may be lysosomotropic, like ammonium chloride and other weak bases, and prevent lysosomal and endocytic signaling [47]. Thailandine's cell killing action could be through the inhibition of EGFR signaling through the inhibition of endocytosis. Intriguingly, this may mean that thailandine can be utilized to target uniquely sensitive cancer cells (like MDA-MB468 and A549) using the high level of EGFR as a biomarker [31].

General Experimental Procedures
The column chromatography (CC) was performed on silica gel Merck 60 (0.040-0.063 µm) and Sigma-Aldrich Sephadex LH-20. All thin layer chromatography (TLC) analyses were carried out on DC-AlufolienKieselgel 60 F254 (Merck, New Jersey, USA) and detected with a UV at 254 and 365 nm, followed by spraying with Dragendorff's reagent for alkaloid detection. The 1 H-and 13 C-NMR spectra of isolated compounds in CDCl 3 and/or CD 3 OD, DMSO-d 6 solvent were recorded on Bruker AM 500 MHz spectrometer (Bruker, Fällanden, Switzerland) using TMS as the internal reference. Accurate, high resolution, and MS/MS mass measurements were carried out with an LTQ-Orbitrap mass spectrometer (Thermo Fisher Scientific, Inc., Bremen, Germany).

Plant Material
The leaves of S. dielsiana Y.C. Wu were collected in Ba Vi, Hanoi, Vietnam in March, 2015 and the identity was confirmed by Dr. Nguyen Quoc Huy, Hanoi University of Pharmacy. A voucher specimen, Nr. SD01/2015 is deposited in Institute of Chemistry, Vietnam Academy of Science and Technology, Vietnam.

Extraction and Isolation
The extraction and isolation of alkaloid compounds from the methanol extracts of S. dielsiana leaves were conducted as described previously [34,38]. Briefly, the dried and powdered leaves (3.0 kg) of S. dielsiana Y.C. Wu were macerated with methanol/water (85:15 v:v), at room temperature, three times (5 L, 24 h each time). The combined methanol extracts were evaporated under reduced pressure at 50 • C, and the methanolic extract (MB2L) was further fractioned with solvents of increasing polarity: n-hexane, dichloromethane, and n-butanol, successively. The organic solution was then concentrated under vacuum to give MB2L-H, MB2L-CH, and MB2L-B fractions. All these fractions were evaluated for their anti-proliferative activity. The MB2L-CH and MB2L-B fractions, the most effective fractions (Table 1), were further purified by repeated chromatography on silica gel columns to get the pure compounds [38]. The compounds 1, 2, and 6 and compounds 3, 4, 5, and 7 were isolated from the MB2L-B and MB2L-CH fraction, successively. The chemical structures of these seven alkaloids, such as oxostephanine (1), thailandine (2), stephanine (3), crebanine (4), O-methylbulbocapnine (5), palmatine (6), and tetrahydropalmatine (7) were elucidated and confirmed by spectroscopic methods, including MS, 1 H and 13 C-NMR and compared to the reported references [34,38] and standard compounds (Supplementary Tables S1 and S2). The structures of isolated compounds 1-7 are illustrated in Figure 1.

Cell Lines and Cell Culture
In vitro anticancer study: The in vitro anti-proliferative tests were carried out as described previously [15], using aliquots of the cell lines used therein. Briefly, HeLa, MDA-MB-231 and MDA-MB-468, MCF7, and 184B5 and MCF 10A cell lines were purchased from ATCC. The identity of the cell lines presented in this study were verified by short tandem repeat profiling by Genetica DNA Laboratories (www.celllineauthentication.com) (April 8, 2015). MCF10A was grown in DMEM F12 medium supplemented with 0.02 µg/mL epidermal growth factor, 10% FBS, 10 µg/mL insulin and 0.5 g/mL hydrocortisone. MDA-MB-468 were grown in DMEM supplemented with 10% FBS and antibiotics cocktail (100 µg/mL streptomycin, 100 units/mL penicillin and 250 ng/mL amphotericin B). SRB cell proliferation assays were carried out as previously described [34,48,49], with some minor modifications. The concentration range for all tested cell lines were from 100-0.39 µg/mL for extracts and 100-0.39 µM for pure compounds. Paclitaxel (Cayman Chemical, Ann Arbor, MI, USA) (range: 100-0.39 nM) and chloroquine (Sigma Aldrich, Oakville, ON, Canada) (range: 100-1.56 µM) were used as positive controls. All experiments were carried out at least twice.
Cell cycle analysis: Propidium iodide staining and flow cytometry analysis were carried out as described previously [34,48] with minor modifications. Briefly, 250,000 HeLa cells (50,000 cells/mL) were treated with 10 µM oxostephanine (1) and harvested at the indicated time points. All experiments were carried out at least twice.
Western Blot analysis: Western blotting was carried out as described previously [48]. The primary antibodies used were from Cell Signaling (Histone H3; p-Histone H3 S10 and Phospho-Aurora A (Thr288)/Aurora B (Thr232)/Aurora C (Thr198)) and used at recommended dilution. For cell synchronization at prometaphase, HeLa cells were treated with 50 ng/mL nocodazole (NZ) (Santa Cruz Biotechnology) for 12 h. All experiments were carried out at least twice.
Annexin V staining: Cell death mechanism and cell proliferation assays were analysed using the IncuCyte Apoptosis Assay method using the Annexin V Green reagent (Essen Biosciences, Ann Arbor, MI, USA) and the IncuCyte S3 and IncuCyte ZOOM software (Version 2018B, Bohemia, NY, USA) according to supplier's protocol. Statistical analysis was carried out in Graph Pad Prism 5. All experiments were carried out at least twice.
Molecular modeling: Docking was carried out using PyRx (0.9.8), utilizing Autodock 4.2 and the default PyRx settings [50]. All small molecules were energy minimized before docking. The docking box was drawn around the whole protein to allow for blind docking. Docking representations were exported as PDB files and displayed in 2D using LigPlot+ [51].
In vitro kinase assay: The Aurora A and B kinase assays were carried out using the Promega ADP-Glo based Kinase Activity Profiling system. The protocols for the Aurora A/(Myelin Basic Protein (MBP) and Aurora B/MBP assay kits were according to those recommended by the suppliers. The concentration range used was from 20 µM-39.1 nM in 384-well plates. The luminescent output from the kinase assay was read on a Biotek Synergy H4 plate reader.
In vitro anti-plasmodial assay: The two P. falciparum strains 3D7 and W2 were synchronized twice with sorbitol before use [52]. The clonality of these strains was verified both in our laboratory and in an independent laboratory from the Worldwide Anti-malarial Resistance Network (WWARN) by PCR genotyping of the polymorphic genetic markers msp1 and msp2 and microsatellite loci [53,54]. The compounds were re-suspended in methanol or DMSO 5% (v/v) and then diluted in methanol to obtain final concentrations ranging from 0.008 to 500 µg/mL (final concentration of 0.5% of methanol or DMSO). Chloroquine, mefloquine, and dihydroartemisinin were prepared and distributed into 96-well plates as described previously [55] and used for P. falciparum activity comparison. For the in vitro microtests, 200 µL of parasitized red blood cells (final parasitemia, 0.1%; final hematocrit, 1.5%) were aliquoted into 96-well plates pre-dosed with anti-malarial drugs (chloroquine, mefloquine and dihydroartemisinin) or natural compounds from S. dielsiana. The plates were incubated for 72 h at 37 • C in controlled atmospheric conditions 5% CO 2 , 10% O 2 , and 75% N 2 . After thawing the plates, haemolysed cultures were homogenized by vortexing the plates. Both the success of the drug susceptibility assay and the appropriate volume of haemolysed culture to use for each assay were determined for each clinical isolate during a preliminary HRP2 ELISA. Both the pre-test and subsequent ELISA assays were performed using a commercial kit (Malaria Ag Celisa, ref KM2159, Cellabs PTY LTD, Brookvale, Australia) according to the manufacturer's recommendation. The optical density (OD) of each sample was measured with a spectrophotometer (Safire 2, Tecan, Lyon, France). The concentration at which the drugs were able to inhibit 50% of parasite growth (IC 50 ) was calculated with the inhibitory sigmoid Emax model, with estimation of the IC 50 through non-linear regression using a standard function of the R software (IC Estimator version 1.2). IC 50 values were validated only if the OD ratio (OD at concentration 0/OD at concentration max) was greater than 1.6 and the confidence interval ratio (upper 95% confidence interval of the IC 50 estimation/lower 95% confidence interval of the IC 50 estimation) was less than 2.0. The anti-plasmodial in vitro assay was performed five times for each product. The IC 50 values represented the mean value calculated from five experiments. According to Hout [56], a selectivity index (SI) was based on the ratio between anti-parasitic and cytotoxic activities. In the case of our study, a SI was calculated between non-tumorigenic cell line 184B5 and parasites 3D7 or W2 and/or MCF10A and these two parasites.

Conclusions
In conclusion, this is the first report on the cytotoxicity and anti-plasmodial activities of the methanol extract of S. dielsiana Y.C. Wu leaves, its fractions, and isolated compounds. The most active compounds isolated from this extract are two aporphine alkaloids, oxostephanine (1) and thailandine (2). Oxostephanine proves to be the most effective against most cancer cells, although its selectivity for cancer cells is low. Thailandine is effective only on sub-types of cancer cells including MDA-MB-468 and MCF7. However, it shows a good selectivity (SI > 12). The cytotoxic effect of oxostephanine is likely due to its inhibition of Aurora kinases activity, probably through its binding to the ATP-binding pocket. As for anti-plasmodial activity, thailandine is the most desirable among the seven isolates as it is effective on both chloroquine-susceptible 3D7 and chloroquine-resistant W2 P. falciparum strains. Furthermore, substantially high doses of thailandine are required to kill human cells, suggesting that it could be relatively safe. Together, our data suggest that oxostephanine and thailandine can be excellent lead compounds towards developing effective and safe anticancer and anti-malarial drugs. Further studies are needed to investigate the mode of action of thailandine on the human cancer cells as well as on P. falciparum parasites and its potential activity on hepatic stages and gametocytes.  (Table S1) and 13 C-NMR data (Table S2)