The Canadian Prairie Plant Thermopsis Rhombifolia Contains Luteolin, a Flavone that Inhibits Cyclin Dependent Kinase 9 and Arrest Cells in the G1-Phase of the Cell Cycle

INTRODUCTION: Plant species within the prairie ecological zone of Canada are a source of natural products with important bio-activities. Investigation of these plants and the secondary metabolites that they produce will provide insight into their biology, and identify sources of natural products that may become new medicines or scientific tools. METHODS: We investigated a G1 phase arrest activity in extracts from the prairie plant, Thermopsis rhombifolia (Buffalo bean) by biology-guided fractionation and isolated luteolin. Cell based assays and CETSA were used to identify a cyclin-dependent kinase 9 inhibitory activity. RESULTS AND DISCUSSION: Luteolin treated cells showed decreased phosphorylation of the carboxy terminal domain of RNA polymerase II, and low levels of Mcl-1. Plant extracts or luteolin inhibited Cdk9 (cyclin dependent kinase) in tests in vitro, stabilized Cdk9 as determined by the cellular thermal shift assay (CETSA), and arrested cells in the G1 phase of the cell cycle. CONCLUSIONS: Luteolin joins an increasing number of flavonoid inhibitors that make convenient cell biology tools and contribute to our understanding of natural product biology in plants.


Introduction
Plant species of the Canadian prairie ecological zone have evolved under a competition regime comprised of grazing herbivores and a short reproductive period limited by climatic conditions. To compete successfully under this regime, some plants synthesize secondary metabolites that affect mammalian physiology [1]. Both anecdotal and scientific evidence demonstrate that prairie plants produce chemicals that are toxic to grazing herbivores or humans [2]. Only in a handful of cases, however, has the toxicity been investigated to the level of the identification of the chemical and its biological target. It is noteworthy that natural products are the major source of modern medicines, and that these chemicals are invaluable tools with which to investigate cellular pathways [3].
Phenotypic assays are an experimental approach in which extracts or chemicals are screened in cells or organisms for effects upon broad biological phenomena such as proliferation, cell death, or morphology, amongst others [4]. This approach is in contrast to that of targeted assays in which a specific molecular target, such as a protein kinase, is screened. The two approaches are complementary, although phenotypic assays have been recognized as the more successful of the two in identifying biological effects of inhibitors [5]. Some of the reasons for the success of phenotypic assays include that live biological systems (cells or organisms) capture more of features required for measuring meaningful inhibition, such as multiple competing targets, and structural differences of a protein in the native cellular environment. Once an effect of a chemical has been observed, one can use cell biology data to evaluate results from phenotypic experiments. Finally, hypotheses that arise from phenotypic assays can be tested by targeted approaches, such as the cellular thermal shift assay (CETSA).
We have previously reported that extracts prepared from the prairie plant species, T. rhombifolia, were toxic to human cells and induced a cell cycle arrest [10]. Here, we identified that extracts specifically inhibited Cdk9 in biochemical assays and decreased levels of the anti-apoptotic protein, Mcl-1, in human cells. By biology-guided fractionation we isolated the flavonoid, luteolin, which downregulates Mcl-1 by inhibition of Cdk9 in vitro and subsequent inhibition of RNA polymerase II phosphorylation.

Plant Collection and Extract Preparation
Thermopsis rhombifolia aerial parts were collected at undisturbed sites near Lethbridge, AB, Canada. The taxonomical classification was confirmed by the University of Lethbridge Herbarium and using botanical criteria [6], and voucher specimens (#672), #Golsteyn020 and #Golsteyn120 have been deposited in the University of Lethbridge Herbarium. Aerial parts were cleaned, dried at 40°C and stored in dark paper bags at room temperature until used. Extractions were prepared by grinding dried material to a fine powder and suspending it in 75% ethanol (10x w/v) with stirring overnight. The solute fraction was collected by vacuum filtration and dried under reduced pressure. The dried extracted material was weighed, named PP-020 and stored in darkness at room temperature. For testing in biological assays, samples of dried extracts were dissolved in DMSO to 50 mg/mL or 100 mg/mL. Extracts in DMSO were stored at -20°C.

Extraction and Subfractionation of Thermopsis Rhombifolia
Dried aerial parts of T. rhombifolia (100 g) were finely powdered and macerated overnight with a solution of ethanol/H 2 O (75:25) at room temperature. After vacuum filtration on paper filter, the solute was concentrated under reduced pressure. The crude extract (28.6 g) was further subjected to silica column chromatography on prepacked cartridge Kieselgel (40 g, 40 × 100 mm, 40-60 µm) using a CH 2 Cl 2 -MeOH gradient [1:0 to 0:1], to give 10 fractions (20 mL each). All fractions were analyzed by TLC on silica gel 60F 254 (Merck) using the solvent mixture CH 2 Cl 2 -MeOH (90:10) and submitted in parallel to a bio guided selection. Spots were visualized by heating after spraying with 3% H 2 SO 4 + 1% vanillin.

Cytotoxicity Assays
The cytotoxicity of T. rhombifolia extracts (100 mg/mL) or subfractions was measured by (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) tetrazolium (MTT) reduction assay (Sigma; M2128-1G). HT-29, M059K and WI-38 cells were plated at 10,000, 4000 and 4000 cells per well with 200 uL medium per well in 96 well culture plates, respectively, and cultured for 72 h prior to treatment. At 96 h after treatment, 20 μL MTT solution (5 mg/mL MTT in PBS (137 mM NaCl, 3 mM KCl, 100 mM Na 2 HPO 4 , 18 mM KH 2 PO 4 )) was added to the media in each well and the plates were incubated at 37°C for 3.5 h. The media was then aspirated and 100 μl MTT solvent (4 mM HCl, 0.1% (v/v) octylphenoxypolyethoxyethanol, in isopropanol) was added to each well. Plates were shaken for 30 min in the dark, and absorbance was measured at 590 nm using a BioTek Epoch microplate spectrophotometer operated by Gen 5.0 software. Results were expressed as IC50 concentrations; the concentration of the compound that reduced the absorbance of MTT by 50%. The percent absorbance was determined by normalization to 0.1% (v/v) DMSO-treated cells. The log concentrations of the compound were plotted against the normalised percent absorbance using Microsoft Excel 2016 software. Analysis was performed with GraphPad Prism 5 software, using non-linear regression (log(inhibitor) versus normalised response), to estimate the IC50 concentrations. Standard curves were plotted using the equation: Y = maximum + (maximum -minimum)/ (1+10 (x-LogIC50) ), where maximum is the percentage of viable cells after treatment with 0.1% DMSO, minimum is the percentage of viable cells after treatment with the highest concentration of the treatment and x is the log 10 value of the treatment concentration. All measurements were performed in triplicate and experiments were performed three times.

Flow Cytometry
HT-29 cells were plated at 1 × 10 6 cells/75 cm 2 flask and incubated at 37°C for 72 h prior to treatment. Treated cells were collected by trypsinization, washed with 1% (w/v) BSA in PBS and fixed in ice-cold 90% ethanol for at least 24 h. Fixed samples were stored at −20°C until analyzed. For analysis, samples were centrifuged at 750 g for 5 min at 4°C and washed with ice-cold PBS. Cells were then washed twice with ice-cold wash buffer (1% (w/v) BSA in PBS) then incubated with labelling buffer (1% (w/v) BSA (Fisher; AAJ6465522), 20 μg/mL propidium iodide (PI; Life Technologies; P1304MP) and 200 μg/mL RNAse A (Sigma; R6513-250MG in PBS) for 30 min. Samples were analysed by a FACSAria fusion flow cytometer (BD Biosciences) using BD FACSDiva software (BD Biosciences). Gating was set using cells treated with DMSO or nocodazole. Experiments were performed three times.

Identification of Luteolin from Thermopsis Rhombifolia
Analytical HPLC was performed on a Merck-Hitachi apparatus equipped with an L-7200 automated sample injector, a L-7100 pump, a L-7450 diode array detector, a D-7000 interface and EZChrom software. Luteolin isolated as solid (9.7 mg). Its elemental formula was established by high resolution mass spectrometry as C 15  Analysis of luteolin isolated from T. rhombifolia and commercially purchased luteolin was performed using a Symmetry C18 (Waters) column (100Å, 3.5 µm, 4.6 mm × 75 mm) at room temperature on a Waters 1525 Binary HPLC System coupled to a UV/Visible detector operated by Breeze 2 software. Samples were dissolved to 100 mg/mL and 5 µL was loaded with a loop injector. The samples were eluted using formic acid (0.1%) as solvent A and acetonitrile as solvent B, 0-15 min from 15% B to 60% B and a flow rate of 1.0 mL/min. A 10 min re-equilibration was used between runs. UV-Vis spectra were recorded at a detection wavelength of 360 nm and retention time was determined by Breeze 2 software.

Statistical Analysis
Data were analyzed using Microsoft Excel 2016 and GraphPad Prism 5 software. Data were plotted as means ± standard error of the means. Statistical significance was calculated by one-way ANOVA with Tukey HSD coupled anti-rabbit IgG (Promega; PRS3731; 1:2500). The membranes were washed with TBS-T and developed using an alkaline phosphatase conjugate substrate kit (BioRad; 172-1063). Development was stopped with Tris-EDTA (diaminoethane tetra-acetic acid) buffer (10 mM Tris base, 1 mM EDTA, pH 8.0). Western blot analyses were performed three times and anti-actin analysis was used to compare sample loading.

Protein Kinase Assays
Kinase activities were quantified according to the radiometric assay described in [7]. Kinase assays were performed in the presence of 15 µM [γ-33 P] ATP (3000 Ci/ mmol; 10 mCi/mL) with either protein or peptide as substrate and using the appropriate buffer (described hereafter). Controls were performed with appropriate dilutions of dimethylsulfoxide (DMSO). The peptides were obtained from Proteogenix, Oberhausbergen, France. To perform the experiments described in this article, we used the product provided by GE Healthcare

Cellular Thermal Shift Assay
Cellular thermal shift assays were performed on HT-29 cells using the method described previously [8,9]. Briefly, (honest significant difference) post-hoc analysis. Values were considered significantly different when p < 0.05.

Extracts Prepared from Thermopsis Rhombifolia are Cytotoxic to Human Cells
We previously identified an extract (PP-003) from the prairie plant species, Thermopsis rhombifolia (Fabaceae) as a source of natural products that arrest cells in the cell cycle [10]. We proceeded to investigate the chemicals responsible this effect by preparing T. rhombifolia 75% ethanol extractions named PP-020 and confirming its toxicity against human cell lines. Colorectal adenocarcinoma (HT-29), malignant glioblastoma (M059K) and normal lung fibroblast cell lines (WI-38) were treated with either DMSO or with different concentrations of PP-020 for 96 h and viability was measured as IC 50 values using the MTT assay ( Figure 1). The M059K cells were the most sensitive, with an IC 50 of 90 ± 14 µg/mL.
The least sensitive cells were the normal diploid fibroblasts, WI-38 with an IC 50 of 240 ± 34 µg/mL, whereas HT-29 cells displayed a value of 130 ± 8 µg/mL. These data indicated that extract PP-020 was toxic to human cultured cells and there was a period of 96 hours in which pathways might be inhibited by PP-020 to cause cell death.

T. Rhombifolia Extract PP-020 Inhibits Cdk9 and Reduces Mcl-1 Protein Levels
We had previously demonstrated that T. rhombifolia extracts arrest cells in the G1 phase of cell cycle without engaging a DNA damage response [10]. This type of cell cycle arrest is a characteristic of cyclin dependent kinase (Cdk) inhibitors, therefore, we sought to investigate whether extract PP-020 could inhibit a panel of protein kinases that included several members of the CMGC group (Cdks, mitogen-activated protein kinases, glycogen synthase kinases and CDK-like kinases) (Figure 2A). PP-020 inhibited Cdk9/cyclin T at an IC 50 of 0.5 µg/mL whereas Cdk1/cyclin B, Cdk2/ cyclin A, Cdk5/p25, or the glycogen synthase kinase 3 a/b (GSK-3 α/β), were not inhibited at concentrations up to 10 µg/mL of PP-020.
We reasoned that if extract PP-020 could inhibit the Cdk9/cyclin T complex in cells then the levels of short-lived proteins, such as Mcl-1, would be reduced in treated cells [11]. To test this, HT-29 cells were either not-treated (NT), or treated with different concentrations of PP-020. In addition, we tested HT-29 cells with representative compounds to provide insight into the specificity of cellular responses. These included camptothecin (CPT) to damage DNA [12]; CR8, a pan-Cdk inhibitor that induces cell cycle arrest and reduces Mcl-1 levels [11]; quercetin a flavonoid common to Fabaceae [13]. Cells were collected and examined by Western blotting ( Figure 2B). We observed that NT cells contained Mcl-1 and CR8-treated cells had low levels of Mcl-1, as expected. Strikingly, cells treated with 500 µg/mL PP-020 also had low levels of Mcl-1, whereas CPT, quercetin or a 150 µg/mL concentration of PP-020 had little effect. CPT treated cells, however, revealed a histone gamma H2AX signal characteristic of damaged DNA, whereas only a weak signal was detected in 500 µg/mL PP-020 treated cells.
We next compared PP-020 treated cells to cells engaged in apoptosis by staurosporine (STS) treatment ( Figure 2C). We monitored cells by western blot analysis for the Poly (ADP-ribose) polymerase (PARP) cleavage product, which is present in apoptotic cells [14]. STS treated cells revealed a PARP cleavage product whereas no signals were observed for PP-020 treated cells or other treatments. Equal amounts of cell proteins were loaded for each treatment as confirmed by anti-actin antibodies. These data indicated that PP-020 causes a reduction in Mcl-1 levels in cells that was consistent with an inhibition of Cdk9/cyclin T in biochemical assays. In addition, apoptotic and DNA damage pathways were not activated after treatment with PP-020 for 24 h.

Luteolin Isolation from PP-020 Extracts by Biology-Guided Fractionation Using Mcl-1 Activity
The capacity of PP-020 to reduce Mcl-1 levels in treated cells provided us with an assay to identify the active chemical. We first determined which aerial plant part contained the anti-Mcl-1 activity ( Figure 3A). T. rhombifolia plants were separated into leaves (L.E.), flowers (F.E.) or woody stem parts (S.E.) and equal amounts of each were extracted under similar conditions. In addition, leaves from T. rhombifolia collected at a different year from that of PP-020, named PP-029 were also extracted and tested. Cells were either not-treated (NT), treated with CR8, or with extracts prepared from different T. rhombifolia parts and plants. Mcl-1 was present in NT cells and in reduced amounts in CR8 treated cells, as expected. L.E., S.E. and PP-029 extracts reduced Mcl-1 levels in treated cells, whereas F.E. treated cells contained Mcl-1. By actin western blotting, we confirmed that similar amounts of total cellular proteins were tested under each treatment.
We undertook several rounds of biology-guided purification in which HT-29 cells were treated with various subfractions of increasing purity of chemicals to determine their capacity to reduce Mcl-1 levels. In a representative experiment, we observed that cells treated by fraction PF-17 had reduced Mcl-1 levels ( Figure 3B). A pure chemical was isolated (PF-36) and identified as the flavone, luteolin ( Figure 3C). We tested other flavonoid chemicals, including commercially available luteolin, the isoflavones daidzein, genistein, and the flavonol quercetin, at similar concentrations and showed that luteolin was the most effective of these ( Figure 3B). To confirm further that luteolin isolated from T. rhombifolia was the source of anti-Mcl-1 activity, we compared its elution profile by HPLC to a commercially available luteolin and found that the two profiles were identical ( Figure 3D).

Luteolin is Cytotoxic, Reduces RNA Polymerase II Phosphorylation, and Arrests Cells in the G1 Phase of the Cell Cycle
Mcl-1 downregulation can sensitize cells to cell death and luteolin was previously reported to be toxic to human cells [15]. Therefore, we confirmed that luteolin reduced cell viability in HT-29 cells. We treated HT-29 cells with 0.01-100 µg/mL of luteolin for 96 h and measured cell viability by the MTT assay ( Figure 4A). We found that luteolin was toxic to HT 29 cells with an IC 50 of 7.0 ± 0.7 µg/mL at 96 h. We then tested if another anti-apoptotic protein, Bcl-xL might also be downregulated ( Figure 4B).  We treated cells with 10, 50, or 100 µg/mL of luteolin and compared Mcl-1 and Bcl-xL levels to cells either not treated (NT), treated with CR8 or with PP-020. Nottreated cells maintained a readily detected amount of Mcl-1, and CR8 or PP-020 treated cells had relatively low levels of Mcl-1. We observed a striking reduction in Mcl-1 levels with 50 and 100 µg/mL of luteolin, and a minor reduction with 10 µg/mL at 24 h treatment. By contrast, the levels of the related Bcl-2 protein, Bcl-xL, did not change. We reasoned that if T. rhombifolia extract or luteolin treated cells inhibited Cdk9 and reduced Mcl-1 protein levels, then treated cells might have reduced levels of RNA polymerase II (RNAPII) phosphorylation, a known substrate of Cdk9. HT-29 cells treated with CR8 or PP-020 had little RNAPII phosphorylation when compared to NT cells ( Figure 4B). Furthermore, cells treated with 50 or 100 µg/mL of luteolin showed little RNAPII phosphorylation, whereas 10 µg/mL luteolin had RNAPII phosphorylation levels comparable to NT. Total RNAPII levels were similar in either NT cells or all treatments.
We have previously reported that cells treated with PP-020 arrest in the G1 phase of the cell cycle. We therefore tested whether cells treated with luteolin arrested in the G1 phase. HT-29 cells either not-treated (NT) or treated for 24 hours with PP-020 or luteolin and analyzed for DNA content by flow cytometry ( Figure 4C). NT cells had 54% of cells in the G1 phase, 14% in S phase and 30% in G2/M-phase. By contrast, PP-020 and luteolin treated cells showed a similar cell cycle distribution with 83% of cells in the G1 phase, less than 3% of the population in S phase and less than 15% in the G2/M-phase. These data indicated that luteolin treated cells arrest in the G1 phase of the cell cycle.

Luteolin Protects Cdk9 from Thermodenaturation as Shown by the Cellular Thermal Shift Assay (CETSA)
Knowing that PP-020 inhibited Cdk9 in biochemical assays and that PP-020 or luteolin could reduce RNAPII phosphorylation levels and Mcl-1 protein levels in cells, we tested if one could detect Cdk9 inhibition by the cellular thermal shift assay (CETSA) [8]. We first confirmed that Cdk4, a member of the Cdk family could be stabilized by the Cdk4 inhibitor PD-0332991 using this method ( Figure 5A). Using temperatures between 39 and 65°C, we were able to detect greater amounts of Cdk4 in solution in the presence of the Cdk4 inhibitor. By contrast, the quantity of Cdk4 in solution did not change in the presence of luteolin when compared to non-treated cells ( Figure 5B). We next tested Cdk9 and luteolin, and were able to detect greater amounts of Cdk9 in solution in the presence of luteolin compared to non-treated cells across a temperature range of 44-50°C ( Figure 5C). These data strongly supported our observations that luteolin inhibits Cdk9 dependent events in treated human cells.

Discussion
We are investigating Canadian prairie plants for natural products that inhibit vital cellular pathways. This ecological zone harbours a number of plant species that have limited global distribution and have not been explored extensively for natural products. In this study, we expanded on our previous work that showed that extracts prepared from Thermopsis rhombifolia were toxic to human cultured cells and induced a G1 phase arrest [10].
Here we show extracts from this plant inhibit Cdk9 and downregulate the anti-apoptotic protein Mcl-1. By bioassay-guided fractionation, we then purified the natural product flavone, luteolin, and show that it is cytotoxic to cultured cells. Furthermore, we provide evidence that luteolin inhibits Cdk9 in cells, as it reduces phosphorylation of the RNA polymerase II C-terminal domain and induces a G1 phase arrest.
Thermopsis rhombifolia is known by the common names as the buffalo bean or golden bean [16]. It is a prominent plant in Canadian prairies that blooms in May with bright yellow flowers and produces beanlike fruits by mid-summer. If consumed, the plant has been reported to be a source of poisoning in children or livestock [17][18][19]. The precise toxic ingredients have not yet been identified; however, the abundance of flavones, such as luteolin may contribute to toxicity as we detect here. A number of alkaloids have been previously reported to contribute to toxicity in domestic animals [20]. Extracts prepared from related Thermopsis species present in Turkey have been shown to have potential genotoxic activity or antimicrobial activity [21,22].
The cytotoxicity and cell cycle phase arrest induced by T. rhombifolia extracts led to further investigations into the biological pathways inhibited in treated cells. We found that extracts targeted Cdk9, the catalytic subunit of the positive-transcription elongation factor b (P-TEFb) [23,24]. Cdk9 phosphorylates the carboxy-terminal domain (CTD) of RNA polymerase II (RNAPII) at Ser5 and Ser2. Previous studies have shown that either inhibition of Cdk9 by small molecules, such as flavopiridol, or knockdown of Cdk9 by shRNA, inhibits CTD phosphorylation, which is required for transcription initiation and elongation [25,26]. Hence, Cdk9 inhibition reduces transcription, therefore proteins with high turn-over rates, such as Mcl-1, are depleted under these conditions. Mcl-1 is an anti-apoptotic protein with N-terminal motifs that signal proteasomal degradation, unlike other Bcl-2 family proteins such as Bcl-xL [27,28]. Accordingly, we found that treating cells with T. rhombifolia extracts led to reduced Mcl-1 levels but not reduced Bcl-xL levels.
We used Mcl-1 reduction in a biology-guided fractionation of the leaves of T. rhombifolia to identify the chemical responsible for loss of Mcl-1 protein. We discovered the flavone luteolin as the bioactive compound and confirmed the result by comparing the isolated luteolin to that of commercially available luteolin in bioassays and by HPLC. Consistent with the cellular activity of Thermopsis extracts, luteolin reduced CTD phosphorylation of RNAPII and downregulated Mcl-1 but not Bcl-xL protein. We also found that luteolin induced an arrest in the G1 phase of the cell cycle, confirming reports by others [29][30][31]. We extended the previous observations by investigating luteolin inhibition by CETSA. Since luteolin did not thermostabilize Cdk4 in the CETSA assay, it is unlikely that the G1 phase arrest was due to direct Cdk4 inhibition and likely due to decreased levels of cyclin D1 [31]. We found by CETSA, however, that luteolin thermostabilizes Cdk9 in cells. Notably, Cdk9 contains a malleable ATP binding site and a flexible C-terminal tail that is not present in cell cycle Cdks [32,33]. Thus, selective inhibition of Cdk9 over other Cdks by certain flavones may result from the malleability of the Cdk9 ATP binding site, as previously suggested in structural and functional studies with Cdk9 inhibitors, such as the flavone-derived compound flavopiridol [34].
The arrangement of the core structure rings and the functional groups of flavonoids has been shown to affect their biological activities. Accordingly, the isoflavones daidzein or genistein did not affect Mcl-1 levels, whereas the flavonol quercetin had relatively minor activity. Polier et al. 2011 demonstrated that certain flavones inhibit Cdk9 activity and downregulate Mcl-1 [35], whereas Echalier et al. 2014 showed that daidzein or genistein induce an 8-fold lower thermal shift of Cdk9 in vitro when compared to the flavone apigenin [36]. These data suggest that anti-Mcl-1 activity of some flavonoids is related their capacity to inhibit Cdk9.
We draw several conclusions from the isolation of luteolin from T. rhombifolia and its inhibition of Cdk9. First, some Canadian prairie plant species produce secondary metabolites that can affect mammalian cells and these chemicals can be isolated by biology-guided fractionation. We report that Thermopsis rhombifolia, a toxic plant, produces sufficient amounts of luteolin that can be detected in cell based assays as an anti-Mcl-1. This flavone is widely studied and is reported to affect a multitude of pathways when applied to cells, including toxicity, anti-oxidant, inflammation, and proliferation [37]. The sum of its activities has led to studies suggesting roles in chemopreventive cancer treatments [38]. G1 phase inhibitors are scarce as cell biology tools, and we propose that luteolin can be used as an inexpensive and convenient chemical inhibitor of the cell cycle, when used at relatively low concentrations.

Conclusions
The cell cycle arrest and anti-Mcl-1 activity in extracts of the prairie plant, Thermopsis rhombifolia, was used to isolate the natural product, luteolin. This molecule inhibits the protein kinase Cdk9 biochemically and in cells.