Analysis of Plant–Plant Interactions Reveals the Presence of Potent Antileukemic Compounds

A method to identify anticancer compounds in plants was proposed based on the hypothesis that these compounds are primarily present in plants to provide them with an ecological advantage over neighboring plants and other competitors. According to this view, identifying plants that contain compounds that inhibit or interfere with the development of other plant species may facilitate the discovery of novel anticancer agents. The method was developed and tested using Magnolia grandiflora, Gynoxys verrucosa, Picradeniopsis oppositifolia, and Hedyosmum racemosum, which are plant species known to possess compounds with cytotoxic activities. Plant extracts were screened for growth inhibitory activity, and then a thin-layer chromatography bioautography assay was conducted. This located the major antileukemic compounds 1, 2, 4, and 5 in the extracts. Once the active compounds were located, they were extracted and purified, and their structures were determined. The growth inhibitory activity of the purified compounds showed a significant correlation with their antileukemic activity. The proposed approach is rapid, inexpensive, and can easily be implemented in areas of the world with high biodiversity but with less access to advanced facilities and biological assays.


Introduction
Natural products are an invaluable source of novel chemical structures for drug development. Approximately 32% of the small molecules approved for cancer treatment are natural or naturally derived from organisms such as plants, bacteria, and marine species [1]. Natural products have been evolutionarily optimized with many sp 3 -hybridized carbons, chiral centers, and complex ring structures [2,3]. It is very unlikely that complex naturally occurring compounds, such as homoharringtonine, paclitaxel, and vincristine, could have emerged from a drug design program [4].
The ability of a plant species to chemically inhibit a neighboring plant through the release of such secondary metabolites is called allelopathy [26,27]. The accumulation and release of these allelopathic secondary metabolites, called allelochemicals, may influence their survival, the dominance and succession of plants, community formation, the biodiversity of a region, and crop production [24,26,28]. Allelochemicals, which have a major role in chemical defense, must have a level of inherent bioavailability to successfully leave the donor plant, travel through the soil, and enter the receiving plant [29].
Although allelopathy is a process common to plants, it has been argued that allelopathy is more effective when and where plants are experiencing stress [30,31]. Notably, the concentrations of various clinically useful anticancer compounds in plants increase when the plant experiences some form of moderate stress. This has been reported for camptothecin [32], paclitaxel [33], vinblastine [34], podophyllotoxin [35], and several sesquiterpene lactones with known cytotoxic properties [36][37][38][39][40][41].
Identifying plants containing compounds that inhibit or interfere with the development of other plant species may increase the likelihood of discovering compounds with anticancer activity in a time-and cost-effective manner. To test this hypothesis, four species were selected: Magnolia grandiflora L, Gynoxys verrucosa Wedd., Picradeniopsis oppositifolia (Nutt.) Rydb. ex Britton, and Hedyosmum racemosum (Ruiz & Pav.) G. Don. These plant species have been reported to contain compounds with activity against cancer cell lines [42][43][44][45]. They are from three unrelated plant families (Magnoliaceae, Asteraceae, and Chloranthaceae), they grow in different plant habitats, and were available in sufficient amounts from previous collections.
The presence of compounds that inhibit or interfere with the development of other plant species on the extracts of the test plants was investigated using seeds of Lolium perenne L. ssp. multiflorum (Lam.) Husnota, a representative monocot, and Lactuca sativa L., a representative dicot. The use of plant seeds has several advantages over standard cytotoxicity assays with cell lines. Seeds are multicellular eukaryotic test systems that experience spatial and temporal changes in their cellular development. The seeds of these plants germinate uniformly and rapidly, and are commonly used for allelopathy studies [46][47][48]. The inhibitory effects of the extracts and compounds 1-5 ( Figure 1) isolated from the extracts were contrasted and correlated with their antileukemic activity using MV4-11 (acute myeloid leukemia, AML) cells. The ability of a plant species to chemically inhibit a neighboring plant through the release of such secondary metabolites is called allelopathy [26,27]. The accumulation and release of these allelopathic secondary metabolites, called allelochemicals, may influence their survival, the dominance and succession of plants, community formation, the biodiversity of a region, and crop production [24,26,28]. Allelochemicals, which have a major role in chemical defense, must have a level of inherent bioavailability to successfully leave the donor plant, travel through the soil, and enter the receiving plant [29].
Although allelopathy is a process common to plants, it has been argued that allelopathy is more effective when and where plants are experiencing stress [30,31]. Notably, the concentrations of various clinically useful anticancer compounds in plants increase when the plant experiences some form of moderate stress. This has been reported for camptothecin [32], paclitaxel [33], vinblastine [34], podophyllotoxin [35], and several sesquiterpene lactones with known cytotoxic properties [36][37][38][39][40][41].
Identifying plants containing compounds that inhibit or interfere with the development of other plant species may increase the likelihood of discovering compounds with anticancer activity in a time-and cost-effective manner. To test this hypothesis, four species were selected: Magnolia grandiflora L, Gynoxys verrucosa Wedd., Picradeniopsis oppositifolia (Nutt.) Rydb. ex Britton, and Hedyosmum racemosum (Ruiz & Pav.) G. Don. These plant species have been reported to contain compounds with activity against cancer cell lines [42][43][44][45]. They are from three unrelated plant families (Magnoliaceae, Asteraceae, and Chloranthaceae), they grow in different plant habitats, and were available in sufficient amounts from previous collections.
The presence of compounds that inhibit or interfere with the development of other plant species on the extracts of the test plants was investigated using seeds of Lolium perenne L. ssp. multiflorum (Lam.) Husnota, a representative monocot, and Lactuca sativa L., a representative dicot. The use of plant seeds has several advantages over standard cytotoxicity assays with cell lines. Seeds are multicellular eukaryotic test systems that experience spatial and temporal changes in their cellular development. The seeds of these plants germinate uniformly and rapidly, and are commonly used for allelopathy studies [46][47][48]. The inhibitory effects of the extracts and compounds 1-5 ( Figure 1) isolated from the extracts were contrasted and correlated with their antileukemic activity using MV4-11 (acute myeloid leukemia, AML) cells.

Results and Discussion
To determine whether the extracts inhibited or interfered with the development of other plant species, dose-response assays were performed for the plant extracts against Lolium perenne L. ssp. multiflorum and Lactuca sativa L., and the concentration that produced 50% growth inhibition (GI 50 ) was calculated. The antileukemic activity of extracts and compounds was measured using MV4-11 (AML) cells, and the concentration that produced a 50% reduction in cell viability (LD 50 ) was calculated.
Thin-layer chromatography (TLC) bioautography has been used extensively in the literature for isolating antibiotic or antifungal compounds from plant extracts [49][50][51][52]. The TLC bioautography assay works well with bacteria and fungi, since very little material is needed for activity and the resolution is high, with the organism covering the entire plate like a lawn. Plant seeds are larger than bacteria and fungi, which limits the number of seeds that can be used. Additionally, seedlings may shift the positions of their roots into other zones when growing. To address these issues, we developed a stainless-steel plant grid with symmetrical rows that was inserted into the agar after contact bioautography. The plant grid limited the diffusion of extracts in the agar and confined seeds to rows. This allowed for easy identification of the active fractions on the TLC plate since the position of the active row of the plant grid could be accurately translated into a measured zone on the TLC plate.
Magnolia grandiflora L. is a common evergreen tree that is native to the southeastern United States [53]. M. grandiflora contains compounds that have plant inhibitory activity and anticancer activity and was used as a positive control when developing the method.
Plants have difficulty growing under the canopy of Magnolia grandiflora, and the chipped mulch of these trees has been shown to reduce the germination and growth of Lactuca. sativa [54]. M. grandiflora has been reported to contain inhibitory sesquiterpene lactones in its leaves that contribute to its ability to suppress the growth of competing plants [55,56]. Marin and Mansilla [57] reported that M. grandiflora extracts had antileukemic activity. Parthenolide, a constituent of M. grandiflora, has been reported to have strong antileukemic effects on both leukemia stem and progenitor cells [37,58].
The aerial parts of Magnolia grandiflora were extracted with 70% methanol/water (MGME) or ethyl acetate (MGEA). These extracts were tested against Lolium perenne and Lactuca sativa to measure their effects on the germination and growth inhibition of seedlings, roots, and shoots (Figures 2A,C and S2). MGEA extract displayed GI 50 values for the germination and growth of L. perenne and L. sativa seedlings, roots, and shoots at concentrations under 3 mg/mL (Figures 2A and S2). The MGME extract was only effective at reducing the growth of L. perenne roots ( Figure S2). The antileukemic activity of the extracts was tested against MV4-11 leukemia cells at concentrations of 0 to 40 µg/mL. MGEA was found to be more potent against MV4-11 leukemia cells than the MGME extract ( Figure 2B,C).
TLC bioautography assays were carried out with Lactuca sativa seeds ( Figure 2D). The reason for this was that the seeds of L. sativa are smaller, have less size variability, and grow faster than the seeds of Lolium perenne. The overall effect on the seedling was used to determine the positions of the active compounds in the TLC plate. Data in different rows had unequal variances when the F test was employed. Thus, the Welch correction was applied to the t-test to analyze the data. The results are presented as percentage differences from the control. Zero represents the control, positive values represent stimulation, and negative values represent inhibition.
Magnolia grandiflora extracts were tested with 13.5 mg MGEA and 30 mg MGME for each assay ( Figures 2C and S2). These amounts were determined by running small TLC plates with different concentrations of extract to identify the maximum amount that produced distinct bands under UV inspection at both 254 and 365 nm. Compound 1, which was identified as parthenolide, was the major inhibitory component in both the MGEA extract ( Figure 2D) and the MGME extract ( Figure S3).  Growth is expressed as the % difference from the control value, and the chromatographic bands are divided by 1 cm rows from the application point to the elution limit on the TLC plate (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17). * The level of significance was p ≤ 0.05. Row 8 was identified as compound 1. All data are presented as the mean ± SEM.
The amount of compound 1 in the extracts was determined by gas chromatography/mass spectrometry (GCMS) analysis using the sesquiterpene lactone dehydroleucodine (DHL) as an internal standard. The MGEA and MGME extracts contained 157 and 46 mg of compound 1 per gram of dry extract, respectively, indicating an increase of approximately 3.4-fold in the amount of compound 1 in the ethyl acetate extract ( Figure 2C). These results showed that the activity level of the extracts on the plant seeds and on the MV4-11 cells paralleled the concentration of compound 1 in the corresponding extracts ( Figure 2A-C).
To test the validity and scope of the method, it was applied to Gynoxys verrucosa V.M. Badillo and Hedyosmum racemosum (Ruiz & Pav.) G. Don. These plant species were collected in the province of Loja in Ecuador, and we did not observe, nor have there been reports, of them having plant inhibitory activities. However, they have been reported to contain antileukemic compounds [43,44]. The method was also applied to Picradeniopsis oppositifolia (Nutt) Rydb. Ex Britt. This plant species was collected in the state of Montana in the United States. There are no reports of this species having plant inhibitory effects. However, we observed that in the hot and arid regions where the plant was collected, it grows in populations that are clearly separated from neighboring vegetation. P. oppositifolia has been reported to contain antileukemic compounds [45].
Gynoxis verrucosa was extracted with ethyl acetate (GVEA), Hedyosmum racemosum was extracted with ethyl acetate (HREA), and Picradeniopsis oppositifolia was extracted with 70% methanol/water (POME) or ethyl acetate (POEA). These extracts were tested against Lolium perenne and Latuca sativa to identify germination or growth inhibition of seedlings, roots, and shoots ( Figures 3 and S3). The GI 50 of all extracts against seedlings of L. perenne was under 2 mg/mL. Apart from POME, the GI 50 of the extracts against the seedling L. sativa was less than 4 mg/mL. Overall, L. perenne was found to be more sensitive than L. sativa. POEA was found to be more potent against both test seeds than POME.  . Experiments were conducted in triplicate. The concentrations that produced 50% growth inhibition (GI50) of germination and the growth of seedlings, roots, and shoots were calculated from a nonlinear regression of concentrationresponse models. All data are presented as the mean ±SEM (C).
Bioautography assays of Picradeniopsis oppositifolia with 70% methanol/H2O (POME) and ethyl acetate (POEA) extract, Gynoxys verrucosa ethyl acetate (GVEA) aerial extract, and Hedyosmum racemosum aerial extract were carried out with Lactuca sativa seeds ( Figure  4). The Gynoxys verrucosa extract (6.5 mg of GVEA) was submitted to the bioautography assay. Row 11 significantly inhibited seedling length ( Figure 4A). This inhibition was due to a reduction in the length of the shoot but not the root. Row 11 was separated and analyzed and shown to contain two compounds. These compounds were purified and recrystallized from ethyl acetate, and their structures were established to be dehydroleucodine (2) and leucodine (3) [43]. The Hedyosmum racemosum extract (HREA, 13.5 mg) was submitted to bioautography assays. Rows 9 to 16 significantly inhibited seedling, root, and shoot lengths ( Figure 4B). The inhibition was centered around row 13, with a broad band potentially indicating diffusion. Row 13 was separated, analyzed, and found to consist of . Experiments were conducted in triplicate. The concentrations that produced 50% growth inhibition (GI 50 ) of germination and the growth of seedlings, roots, and shoots were calculated from a nonlinear regression of concentrationresponse models. All data are presented as the mean ± SEM (C).
Bioautography assays of Picradeniopsis oppositifolia with 70% methanol/H 2 O (POME) and ethyl acetate (POEA) extract, Gynoxys verrucosa ethyl acetate (GVEA) aerial extract, and Hedyosmum racemosum aerial extract were carried out with Lactuca sativa seeds ( Figure 4). The Gynoxys verrucosa extract (6.5 mg of GVEA) was submitted to the bioautography assay. Row 11 significantly inhibited seedling length ( Figure 4A). This inhibition was due to a reduction in the length of the shoot but not the root. Row 11 was separated and analyzed and shown to contain two compounds. These compounds were purified and recrystallized from ethyl acetate, and their structures were established to be dehydroleucodine (2) and leucodine (3) [43]. The Hedyosmum racemosum extract (HREA, 13.5 mg) was submitted to bioautography assays. Rows 9 to 16 significantly inhibited seedling, root, and shoot lengths ( Figure 4B). The inhibition was centered around row 13, with a broad band potentially indicating diffusion. Row 13 was separated, analyzed, and found to consist of a single compound. This compound was purified and recrystallized from a mixture of methanol:chloroform by the slow evaporation method, and its structure was established to be onoseriolide (4) by X-ray crystallography ( Figure 5).   assays of (A) GVEA, (B) HRMEA, and (C) POEA extracts on the growth of L. sativa. The bioautography assay of Gynoxys verrucosa extract revealed dehydroleucodine as the major inhibitory component (row 11). The bioautography assay of Hedyosmum racemosum extract revealed onoseriolide as the major inhibitory component (row 13). The bioautography assay of Picradeniopsis oppositifolia extract revealed eucannabinolide as the major inhibitory component (row 6). Growth is expressed as the % difference from the control, and the chromatographic bands were divided by 1 cm rows from the application point to the elution limit on the TLC plate (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17). * Significance is indicated at p ≤ 0.05. All data are presented as the mean ± SEM. The extracts of Picradeniopsis oppositifolia (POEA 15 mg; POME 36 mg) were submitted for the bioautography assay ( Figures 4C and S4). The POEA extract inhibited seedling growth on rows 5 to 8 ( Figure 4C). Both the roots and shoots of the seedling showed significant growth inhibition in rows 5 to 8. The POME extract was shown to stimulate seedling growth in rows 5, 6, and 10 ( Figure S4). For both extracts, row 6 was separated and analyzed, and was found to consist of one compound. This compound was purified as an amorphous solid that was identified by NMR and LC/MS as eucannabinolide (5).
The effects of compounds 1-5 on the germination and seedling, root, and shoot growth of Lactuca sativa were tested in a Petri dish assay. Concentration-response curves were generated for all compounds ( Figure 6). All compounds except compound 3 were shown to inhibit 50% of the growth of L. sativa seedlings at concentrations of 1.52 mM or lower (Figure 7). The roots were more sensitive to pure compounds, with 50% of the growth being inhibited at concentrations of 1.11 mM or lower ( Figure 7). The shoots were less sensitive. They were not inhibited by compound 2 at concentrations of 20 mM. Compound 1, being the third most active growth inhibitor, inhibited 50% of the growth of seedlings and roots at 0.52 and 0.35 mM, but was much less effective against shoots, inhibiting 50% of their growth at 3.28 mM. The extracts of Picradeniopsis oppositifolia (POEA 15 mg; POME 36 mg) were submitted for the bioautography assay ( Figures 4C and S4). The POEA extract inhibited seedling growth on rows 5 to 8 ( Figure 4C). Both the roots and shoots of the seedling showed significant growth inhibition in rows 5 to 8. The POME extract was shown to stimulate seedling growth in rows 5, 6, and 10 ( Figure S4). For both extracts, row 6 was separated and analyzed, and was found to consist of one compound. This compound was purified as an amorphous solid that was identified by NMR and LC/MS as eucannabinolide (5).
The effects of compounds 1-5 on the germination and seedling, root, and shoot growth of Lactuca sativa were tested in a Petri dish assay. Concentration-response curves were generated for all compounds ( Figure 6). All compounds except compound 3 were shown to inhibit 50% of the growth of L. sativa seedlings at concentrations of 1.52 mM or lower (Figure 7). The roots were more sensitive to pure compounds, with 50% of the growth being inhibited at concentrations of 1.11 mM or lower (Figure 7). The shoots were less sensitive. They were not inhibited by compound 2 at concentrations of 20 mM. Compound 1, being the third most active growth inhibitor, inhibited 50% of the growth of seedlings and roots at 0.52 and 0.35 mM, but was much less effective against shoots, inhibiting 50% of their growth at 3.28 mM.  . Sigmoidal concentration-response models were used to calculate the growth inhibition at 50% (GI 50 ) for compounds 1-5 against L. sativa root, shoot, and seedling growth. The optim function in R was used to fit nonlinear regression curves. Error bars are ±SEM.
The growth inhibitory activity of the purified compounds showed a significant correlation with their antileukemic activity. The compounds were tested for antileukemic activity against MV4-11 leukemia cells. The GI 50 values for Lactuca sativa were converted into Log (1/GI 50 ) values, and the LD 50 values for MV4-11 were converted into Log (1/LD 50 ) values. These values were compared in a scatterplot, and correlations between seedling inhibition, antileukemic activity (Figure 7), and root leukemic activity ( Figure S5) were discovered.
Compounds 1, 4, and 5 were tested against normal peripheral blood mononuclear cells (PBMCs) isolated from healthy adult volunteers (Figure 8). At the highest concentration tested, 20 µM, which was 6 times higher than their LD 50 s against leukemia cells (Figure 7), the percentage of relative viability (treated cells/untreated cells ± SEM) for compounds 1, 4, 5 was 86.0 ± 1.53%, 89.0 ± 2.54%, and 99.8 ± 0.50%, respectively. Compound 3 has been reported to show significantly less toxicity against normal cells than against leukemia cell lines [43]. These results supported the notion that the proposed approach was capable of discovering compounds that target tumor cells without causing significant harm to noncancerous cells. The growth inhibitory activity of the purified compounds showed a significant correlation with their antileukemic activity. The compounds were tested for antileukemic activity against MV4-11 leukemia cells. The GI50 values for Lactuca sativa were converted into Log (1/GI50) values, and the LD50 values for MV4-11 were converted into Log (1/LD50) values. These values were compared in a scatterplot, and correlations between seedling inhibition, antileukemic activity (Figure 7), and root leukemic activity ( Figure S5) were discovered.
Compounds 1, 4, and 5 were tested against normal peripheral blood mononuclear cells (PBMCs) isolated from healthy adult volunteers (Figure 8). At the highest concentration tested, 20 μM, which was 6 times higher than their LD50s against leukemia cells (Figure 7), the percentage of relative viability (treated cells/untreated cells ±SEM) for compounds 1, 4, 5 was 86.0 ± 1.53 %, 89.0 ± 2.54 %, and 99.8 ± 0.50 %, respectively. Compound 3 has been reported to show significantly less toxicity against normal cells than against leukemia cell lines [43]. These results supported the notion that the proposed approach was capable of discovering compounds that target tumor cells without causing significant harm to noncancerous cells.

Plant Materials
The

Plant Materials
The

Extraction and Preparation of Extracts
Samples of the aerial parts of the plants studied were air-dried and ground into a powder and then macerated in an orbital shaker (25 • C and 250 rpm). The extracts were concentrated at a reduced pressure while keeping the temperature below 30 • C. Magnolia grandiflora samples (100 g) were extracted with 800 mL of either ethyl acetate (MGEA) or 70% methanol/water (MGME) to yield 5.98 g and 23.78 g of MGEA and MGME extracts, respectively. These extracts were used for the MV4-11 assays and to measure the concentration of compound 1. Another sample of M grandiflora (3.65 kg) from the same collection was extracted with ethyl acetate in the same manner as above to yield 372.4 g of crude extract. A portion of this extract was filtered through a reverse-phase C18 bed (LiChroprep Merck 25-40 µm) with 70% methanol/water to remove the chlorophyll, a process that helped to improve the tumor cells without causing significant harm to noncancerous cells. resolution of the TLC bioautography assays. MGME was redissolved in methanol and filtered to remove methanol insoluble compounds, which also improved the resolution of the TLC bioautography assay.
A sample of Picradeniopsis oppositifolia was extracted with ethyl acetate (a 260 g sample) or 70% methanol/water (a 50 g sample). The ethyl acetate extract was dissolved in methanol/water and filtered through a reverse-phase C18 bed, as described above, to yield 14.5 g of dried extract (POEA). The methanolic/water extract was suspended in methanol and filtered to yield 10.0 g of extract (POME).
A sample of Gynoxys verrucosa was extracted with ethyl acetate, dissolved, and filtered through a reverse-phase C18 bed, as indicated above, to yield 6.48 g of crude extract (GVEA).
A sample of Hedyosmum recemosum (500 g) was macerated (20-25 • C) for 2 h successively with hexane and ethyl acetate. The ethyl acetate extract was filtered and concentrated under reduced pressure at 35 • C to yield 9.47 g of dried extract (HREA).

Bioassay for Germination and Growth Studies
Seeds of ryegrass (Lolium perenne L. ssp. multiflorum (Lam.) Husnota) and lettuce (Lactuca sativa L.) were treated with 2% NaOCl for 15 min and 2 min, respectively. Seeds smaller than 5.0 mm for L. perenne or 3.0 mm for L. sativa and/or seeds that were discolored were discarded. All experiments were conducted in triplicate, and the seeds were incubated at 22-25 • C with a 12 h photoperiod (LED, 4000 K). The lettuce seeds were incubated in the dark for 48 h before starting the photoperiod. Lettuce seeds were incubated for 6 days, and ryegrass was incubated for 7 days. After incubation, plates were stored at 5 • C to avoid subsequent growth during the measurement process. Germination was determined for each treatment, and the root and shoot lengths were recorded using a computer-interfaced digital caliper system package (500-171-30, Gaging and Software Technologies, Inc., Colorado Springs, CO, USA) comprising a CD-6" AX Mitutoyo caliper, a 600-520-KB-USB computer interface, and a C-FS25-06 data send switch.
Pure compounds were dissolved in methanol and applied to filter paper circles (90 mm diameter) that were dried and placed in Petri dishes. Then, 1.5 mL of deionized water was added to each plate. Compounds 1, 4, and 5 were tested at 0, 0.01, 0.03, 0.11, 0.33, and 1 mg/mL, and compounds 2 and 3 were tested at 0, 0.02, 0.07, 0.22, 0.67, and 2 mg/mL against lettuce seeds using 20 seeds in each dish.
Data were analyzed statistically using Welch's test with the significance level set at 0.05. The results are presented as the percentage difference from the control, which was given a value of zero. Positive values represent stimulation, and negative values represent inhibition. The optim function in R [59] was used to fit logistic functions to root and shoot inhibition (% inhibited) from log-transformed (log10(x)) concentration values for each pure compound. GI 50 (growth inhibition at 50% of control growth) were calculated, and ±SEM was used for error bars.

Contact Bioautography Assay
Sterile agar sheets were prepared by pouring 100 mL of a 0.75% w/v agar solution buffered with MES (0.01 M pH 6.0) into 25 × 25 cm square petri dishes (Nunc 240835 Square Culture Dish, Thermo Scientific, Waltham, MA, USA). Plates were solidified at 4 • C and equilibrated at 25 • C for 30 min before contact bioautography.
Precoated glass (0.250 mm thick, 20 × 20 cm area) thin-layer chromatography (TLC) plates (Analtech Silica Gel GF F254, Newark, DE, USA) were eluted with ethyl acetatemethanol (2:1) to remove possible interference from contaminants in the plates. Dried plant extracts were dissolved in methanol (0.05-0.3 mg/mL) and applied to the plates with a glass capillary tube (50 µL). After sample application, the plates were dried at 40 • C for 1 h and then placed in a vacuum chamber for 30 min. MGEA was eluted with ethyl acetate/hexane (6:4). POEA was eluted with acetone/hexane (6:4). GVEA and HREA were eluted twice with ethyl acetate/hexane (7:3). After elution, the plates were dried for 1 h at 40 • C and then placed in a vacuum chamber for 30 min. The dried TLC plates were placed facedown onto the agar sheets to imprint for 45 min at 5 • C, after which the agar sheets were separated from the TLC plates and the plant grid was secured over the agar sheets, aligning the bottom of the grid with the application line on the TLC plate. Subsequently, 20 Lactuca sativa seeds were placed in each of the 17 rows of the plant grid. The agar sheets were incubated at 22-25 • C in a microbiological incubator. Seeds were grown using the same photoperiod. After 5 days, plants were stored at 5 • C to avoid growth during the measurement process. Germinated seeds were counted for each treatment, and the root and shoot lengths were recorded as described above.
The plant grid was custom built (Wescon Machining Inc.1325 Thomas G Wilson Dr, Conway, AR 72032, USA) from a solid 3/16-inch stainless-steel plate with a water jet cutter. The dimensions of the grid were 180 × 180 mm × 47.6 mm (L × W × H), and the thickness of the divider walls was 1.3 mm. The grid covered the TLC bioautography plate above the application line, dividing the plate into 18 rows with the grid walls tall enough to keep the plants from crossing over into other zones (Figure 9). Conway, AR 72032, USA) from a solid 3/16-inch stainless-steel plate with a water jet cutter. The dimensions of the grid were 180 × 180 mm × 47.6 mm (L × W × H), and the thickness of the divider walls was 1.3 mm. The grid covered the TLC bioautography plate above the application line, dividing the plate into 18 rows with the grid walls tall enough to keep the plants from crossing over into other zones (Figure 9). Figure 9. Plant grid used for the TLC bioautography assays. Figure 9. Plant grid used for the TLC bioautography assays.

Isolation and Characterization of Compounds 1-5
Duplicate TLC plates with MGEA, POEA, GVEA, and HREA extracts were eluted under the same conditions as the TLC plates used for the bioautography assays. Areas corresponding to the major inhibition zones identified in the bioautography assays (MGEA row 8, POEA row 6, GVEA row 11, and HREA row 13) were marked on the duplicate plates. Silica was scraped from the marked areas, transferred to a flask, and stirred with ethyl acetate (3×). Then, the ethyl acetate extracts were combined, filtered, and concentrated at reduced pressure. The isolated compounds (1)(2)(3)(4)(5) were tentatively identified by their mass spectra and 1D-NMR. Additional amounts of compounds 1-5 were isolated using silica gel column chromatography with hexane-ethyl acetate gradients, and with the samples separated by TLC used as references. The identities of compounds 1-3 and 5 were confirmed by MS and 1D-and 2D-NMR to be parthenolide [60], dehydroleucodine [43], leucodine [43], and eucannabinolide [45], respectively. MS, 1D-and 2D-NMR, and X-ray crystallography were used to determine the identity of compound 4 as onoseriolide [44].

Evaluation of Compounds in Normal Cells
Peripheral blood mononuclear cells (PBMCs) were isolated from the buffy coats of adult healthy volunteers (NPB 177, NPB 178, NPB 179). Cells were cultured in IMDM, 20% fetal bovine serum (FBS), and 10% penicillin-streptomycin (Life Technologies, Waltham, MA, USA) in 96-well plates. Cells were treated with compounds 1, 4, and 5 at concentrations of 2.5, 5, 10, and 20 µM, in triplicate. After 48 h of treatment, cell viability was determined using YO-PRO-1 and 7-AAD. At least 10,000 events were recorded per condition on an LSR-II flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). Data analysis was conducted using FlowJo 10.7.1 software for Mac OS X. Cells that were negative for YO-PRO-1 iodide and 7-AAD were scored as viable

Conclusions
We proposed an approach that facilitated the discovery of anticancer compounds from plants based on the hypothesis that those compounds were primarily present in a plant species to provide that species with an ecological advantage over other plant species. The approach was optimized using Magnolia grandiflora, a plant species known to have both antileukemic compounds and plant inhibitory activity. The method was used to correctly identify compound 1 as the major antileukemic compound in plant extracts, and it was also established that both the plant inhibitory activity and the antileukemic activity were correlated with the amount of compound 1 in the extracts. The method was applied to Gynoxys verrucosa, Picradeniopsis oppositifolia, and Hedyosmum racemosum, and in every case, the plant growth inhibitory activity of the extracts correlated with their antileukemic activity. The method was also able to rapidly locate the major antileukemic compounds (1, 2, 4, and 5) in every extract. Overall, the growth inhibitory activities of the purified compounds showed a significant correlation with their antileukemic activities.
Compound 3, which was inactive against leukemia cells ( Figure 6) but was structurally very similar to the potent antileukemic compound 2, had no plant growth inhibitory activity (Figure 7). The lack of plant growth inhibitory activity of compound 3 was likely due to the absence of an α,β-unsaturated exocyclic double bond in the lactone ring, as was the case in our previous study of its antileukemic activity [43].
Compounds 1 and 2, as well as other sesquiterpene lactones, are known to inhibit the NF-κB stress response in human cells, which is critical for the survival of leukemia cells, and it has been suggested that this protein is involved in their mechanism of action [37,61,62]. In this mechanistic explanation, the sesquiterpene lactones interact with NF-κB via a nucleophilic attack by way of an α,β-unsaturated exocyclic double bond [63]. Although no plant NF-κB equivalent has been found [64], other proteins that are critical to the maintenance of both plant and leukemia cells, such as the MYB proteins, have been identified [65,66]. MYBs are conserved across all eukaryotic species, and a conserved MYB in leukemia called cMYB has been considered a promising target for therapy [67]. MYB proteins can be classified according to the presence of one or more repeating ("R") MYB domains. Although the human 3R-MYB proteins regulate the G1/S transition, Feng et al. [68] discovered that the 3R-MYBs in plants functioned in both cell cycle regulation of the G1/S transition and the abiotic stress response, indicating that these proteins are critical to the adaptation and survival of sessile plants [68]. Compound 1 and related compounds have shown promising levels of activity against cMYB [69]. Given this information, it is plausible that 3R-MYB proteins in plants represent a target for growth inhibitory sesquiterpene lactones.
It is clear that the exocyclic α-methylene-γ-lactone is important for anticancer activity [29,43,62,70]. This was supported by the observation that compounds 1, 2, and 5 displayed activities against both plants and MV4-11 cells. Importantly, compound 3, which lacks the exocyclic α-methylene-γ-lactone, was found to be inactive in both systems. This finding suggested that the proposed method had the ability to distinguish structurally related mechanistic features.
Although no mechanism of action has been proposed for compound 4, the fact that it lacks the exocyclic α-methylene suggests that is likely to have anticancer and seed growth inhibitory effects via a different mechanism than compounds 1, 2, and 5. This finding suggested that the current approach can be used to screen for multiple mechanistic pathways. Importantly, the method was able to uncover growth inhibitory compounds that had activity against cancer cell lines but had low toxicity against normal cells.
Although plants have not developed bioactive molecules to cure human cancer, they have developed molecules that interfere with signaling pathways that are conserved across plants and humans. To our knowledge, the potential for the analysis of plant-plant interactions to discover antileukemic compounds or other types of anticancer compounds has not been demonstrated before. However, it has been suggested that the investigation of allelopathic interactions of marine organisms could be of value for the discovery of cytotoxic compounds [71]. A review of the literature, summarized in Table 1, suggests that the application of our approach could have helped to uncover many of the FDA-approved anticancer compounds of natural origin. Nevertheless, further experimentation is necessary to establish the generality and applicability of the method to discover compounds that are active against other types of cancer besides leukemia. The proposed approach can be implemented inexpensively in many parts of the world with rich biodiversity that may lack the infrastructure or resources necessary to implement the currently used high-throughput bioassay-directed isolation paradigms. The evidence presented in this paper shows that the investigation of plant-plant interactions may lead to the discovery of compounds with unique mechanisms of inhibition, and these discoveries may lead to future cancer therapies.

Anticancer Drug Anticancer Activity Plant Inhibition Activity
Paclitaxel Paclitaxel stabilized microtubes in cancer cells and arrested the replication of cancer cells [16].
Paclitaxel arrested onion and maize root cells from dividing by stabilizing microtubules [72,73].
Vinblastine bound to microtubules and created abnormal multipolar division in Allium cepa L. [73].
Podophyllotoxin inhibited onion (Allium cepa L.) root growth by affecting the formation of mitotic microtubular organizing centers [77]. Etoposide inhibited the division of polyploid cells of grass pea (Lathyrus sativus L.) seedlings. The presumed binding regions of etoposide to TopoII were conserved in plants, Drosophila melanogaster, yeast, and humans [78].
Camptothecin, a precursor to irinotecan and topotecan Camptothecin killed cancer cells by inhibiting topoisomerase 1 [16].
Camptothecin selectively caused the inhibition of young developing vascular tissues of the axillary buds of Nicotiana tabacum L. Camptothecin inhibited the sprouting of potatoes by interfering with cell division in the meristem [79]. Early reports showed this using a partially purified enzyme from barley seeds, and strong inhibition of the relaxation of supercoiled pBR322 DNA by the barley DNA enzyme was observed with camptothecin [79]. Later work showed that plants contained a conserved Topo1, and camptothecin-producing plants, including Camptotheca acuminata, Ophiorrhiza pumila, and Ophiorrhiza liukiuensis, had point mutations in Topo1 that conferred resistance to autotoxicity [80].

Homoharringtonine
Homoharringtonine was used for tyrosine kinase inhibitor-resistant chronic myelogenous leukemia (CML). It worked by binding to the A-site of the 80S ribosome and inhibiting translation [81].
Maytansine bound to β-tubulin and blocked the formation of longitudinal tubulin interactions in microtubules [83].
Ellipticine, a precursor to elliptinium Elliptinium is approved in France for the treatment of metastatic breast cancer. Elliptinium and ellipticine inhibited topoisomerase II [84].