Human HCT116 colorectal carcinoma, OE33 esophageal adenocarcinoma of the lower esophagus (Barrett's metaplasia) and KYSE70 poorly differentiated invasive esophageal squamous cell carcinoma resected from middle intra-thoracic esophagus cell lines, were purchased from ATCC. Cells were maintained in RPMI-1640 medium containing 10% FBS and 1% Glutamine in a humidified chamber containing 5% CO₂.

Cotyledon orbiculata plants were purchased from Van den Berg Garden Village (Western Cape Province, Stellenbosch, South Africa). The leaves of the plants were removed, washed with water, cut into small pieces and dried in a ventilated oven at 40°C for 120 h. The dried leaves were crushed in a blender. Two and half (2.5) liters of boiling water were added to 100 g crushed leaf material and the mixture was stirred for 16 h at room temperature. This water extract was filtered through Whatman filter paper (0.45 μm pore size), frozen in liquid nitrogen and subsequently freeze-dried. The dried extract was kept in a desiccator until further use. Stock solutions of the extract was prepared in DMSO at 10 mg/ml, which was stored at −20°C.

The effects of the crude extract of C. orbiculata on viability of HCT116 colorectal cancer cells, OE33 and KYSE77 esophageal cancer cells were determined using WST-1 assay. Briefly, a sub-confluent culture of HCT116, OE33, and KYSE70 were harvested using trypsin-EDTA and centrifuged at 200 × g for 5 min and re-suspended in growth medium to 5 × 10³ cells/ml. A total of 200 μl of the cell suspension was pipetted into each well of columns 2–11 of a 96 well culture plate. The same amount of the growth medium was added to wells of column 1 and 12 to maintain humidity and minimize the edge effect. The plates were incubated at 37°C in a 5% CO₂ incubator overnight until the cells were in the exponential phase of growth. After incubation, cells were then treated with 6 concentrations of the extract (100–3.125 μg/ml). Each dilution of the test sample was tested in quadruplicate in each experiment and the experiments were repeated three times. The plates were again incubated for 48 h at 37°C in a 5% incubator. A negative control (untreated cells) and positive control (cells treated with different concentrations of cisplatin) were included. After incubation, 20 μL of WST-1 reagent was added to each well and the plates were incubated for a further 1 h at 37°C. After incubation with WST-1, the plates were gently shaken and the amount of WST-1 reduction was measured immediately by detecting the absorbance using a microplate reader at a wavelength of 570 nm. The wells in column 1 and 12, containing medium and WST-1 but no cells were used to blank the microplate reader. The percentage of cell viability was calculated using the formula below:

The LC₅₀ values (lethal concentration at which 50% of the cells are killed) were calculated as the concentration of the test sample that resulted in 50% reduction of absorbance compared to untreated cells.

RNA was submitted to Deepseq for library preparation and next generation sequencing was completed on an Illumina NextSeq (Queen Mary University of London). Raw sequence reads in fastq format were processed for quality (phred scores <30 retained) and adapter trimmed using the Trimgalore wrapper for FastQC and Cutadapt (https://github.com/FelixKrueger/TrimGalore). The resultant quality-controlled reads were aligned to the GRCh38.83 genome build using STAR (28). Gene expression was quantified using FeatureCounts (29) and significantly differentially expressed genes identified using EdgeR and DESeq2 (30, 31). The rMATS tool (32) was used to determine the effect of the extract differential canonical and de novo splicing. Gene ontologies and gene set enrichments were determined using Webgestalt (33) and GSEA (34), respectively.

To assess the effects of C. orbiculata on splicing of hnRNPA2B1 and BCL2L1 in HCT116, OE33, and KYSE70 cancer cell lines, cells were treated with varying concentrations of the crude extract of C. orbiculata (LC₅₀/2, LC₅₀, LC₅₀ × 2) for 48 h. Following treatment, total RNA was isolated from treated and untreated cells using TRI Reagent^® (Sigma). RNA quality and quantity were measured using a Spectrophotometer (NanoDrop2000, ThermoFisher Scientific) and samples were stored at ×80°C. Reverse transcription polymerase chain reaction (RT-PCR) was carried out using 1 μg RNA per sample to obtain a constant amount of cDNA using a Takara PrimeScript™ RT Reagent Kit (RR037A). Samples were denatured at 65°C for 10 min, followed by addition of 1 μl PrimeScript RT enzyme. Thereafter, samples were incubated at 25°C for 10 min, then 37°C for 60 min followed by enzyme inactivation at 85°C for 1 min. The cDNA was stored in −20°C for future use.

The cDNA samples were then subjected to PCR using GoTaq^® G2 Green Master Mix (Promega, M7822) and primers which amplify the different splice variants of hnRNPA2B1 and Bcl-xl/s (Table 1). GAPDH was used as a control. Primers were used at final concentration 0.4 μM with 1 μl cDNA (~50 ng). Samples were denatured at 96°C for 5 min, followed by multiple cycles (Table 1) of denaturation at 96°C for 30 s, annealing at primer specific temperature (Table 1) for 30 s, and extension at 72°C for 1 min. There was then a final extension step at 72°C for 10 min. Thereafter, the PCR products were subjected to electrophoresis. Samples were run on 2–3% agarose gels containing 50 ng/ml ethidium bromide for approximately 1 h at 90 V. Gels were visualized using a BioRad Gel Doc™ EZ System.

To detect and quantify activation of caspase-3 in HCT116, OE33, and KYSE70 cancer cell lines following treatment with crude extract of C. orbiculata, cells were subjected to immunofluorescence staining. Cells were seeded at 5,000 cells per well in a black-walled, glass-bottomed 96 well plate (Corning) overnight and treated with varying concentrations of the plant extract based on the calculated LC₅₀ value (LC₅₀/2, LC₅₀, LC₅₀ × 2) for each cell line. Staurosporine and cisplatin were used as positive controls. The cells were washed with sterile 1 X PBS and fixed with 4% paraformaldehyde (PFA). After fixing, the cells were permeabilized with 0.2% TritonX, followed by blocking using 1% BSA for 1 h at room temperature. The cells were probed for cleaved caspase-3 (Cell Signaling Technology #9661). The cells were then washed with sterile 1 × PBS, then labeled with 4 μg/mL Alexa Fluor^® 488 goat anti-rabbit IgG secondary antibody (Product # A-11008, ThermoFisher Scientific, US) for 30 min in the dark. The cell nuclei and cytoskeleton were identified by counterstaining with DAPI and phalloidin. Samples were imaged using a Leica TCS SPE confocal microscope. The images were analyzed using Fiji (Image J 1.0) image analyzer software to calculate the amount of cleaved caspase-3 in cells, and estimate cell numbers. The red and blue channels were threshholded independently and the number of positive CC3 spots counted and divided by the number of DAPI positive nuclei. An analysis was done blind and automated.

Pladienolide B, a known mRNA splicing inhibitor and antitumor compound was used as a positive control for RT-PCR and immunocytochemistry experiments. Cells were treated with varying concentrations of Pladienolide B (0, 2.5, 5, and 10 nM) for 48 h.

The hnRNPB1 specific siRNA was designed using Sigma MISSION Custom siRNA. The sequence [hnRNPB1 anti sense oligo # 8811245892-000030, UUCCUCUCC AAAGGAACAG[dT][dT]]. Transfection was performed using Lipofectamine 2000 (Invitrogen) following the manufacturers instruction. Briefly, 2 μg of siRNA was mixed with the transfection reagent and added to HCT116 cells. Non-target (mock siRNA) and transfection reagent (alone) were included as controls. A time course of 24, 48, and 72 h was conducted to establish the time of optimal knockdown then 48 h taken forward for subsequent experiments. After 48 h, cells were collected for RNA extraction to be used for RT-PCR, fixed and stained for cleaved-caspase 3 or tested for effects on viability via cell8/WST-1 as described above. Over expression of GFP-tagged hnRNPB1 (Origene, Rockville MD) in HCT116 cells was achieved through plasmid transfection with FuGENE^® HD Transfection Reagent (Promega). HCT116 cells were seeded in a 6 well, tissue culture plate (150,000 cells/ml) and were allowed to attach. After attachment, the media was removed, cells were washed in PBS and trypsinised. Six (6) plates were used, 3 for RNA extraction at 24, 48 and 72 h after seeding and the remaining 3 for protein extraction at 24, 48, and 72 h after seeding. They were incubated at 37 degrees at 5% CO₂ during this time. For each plate the cells in 2 wells were treated with medium, 2 were treated with vector DNA and 2 with the hnRNPB1 plasmid.

To confirm whether C. orbiculata induces apoptosis in HCT116, OE33, and KYSE70 cancer cells, the FITC Annexin V/Dead cell Apoptosis kit (Invitrogen, ThermoFisher Scientific) was used. Cells (15 × 10⁴ cells/mL) were seeded in 6 well plates and treated with varying concentrations of the crude extract of C. orbiculata for 48 h. Cells were harvested, washed two times with ice cold PBS and adjusted at a density of 1 × 10⁶ cells/sample (500 μL). The cells were stained with FITC Annexin-V/PI staining kit according to the manufacturer's instructions. The cells were analyzed using Becton Dickinson and Company (FC500) flow cytometer. Results were analyzed using Kaluza analysis 1.5 software.

An RNA-binding protein immunoprecipitation (RIP) assay was conducted using the Magna RIP RNA-Binding Protein Immunoprecipitation Kit (Millipore) according to the manufacturer's protocol. Briefly, HCT116 colorectal cancer cells were treated with 70 μg/ml of C. orbiculata crude extract, collected and lysed by RIP lysis buffer. Untreated cells were used as a negative control. Subsequently, 100 μl cell extract was incubated with RIP buffer containing magnetic beads conjugated with human anti-hnRNPA2B1 antibody or negative control normal mouse IgG. Proteinase K was used to digest the protein and the immunoprecipitated RNA was purified. The isolated RNA was used for quantitative RT-qPCR analysis of the products of the BCL2L1, BCL2, MDM2, cMYC, CD44, CDK6, cJUN, and TXNP genes using LightCycler^® SYBR Green (Roche Life Science) qPCR kit following the manufacturer instructions. The selected mRNAs are involved in one way or another in apoptosis.

Data obtained from all experiments were analyzed using Graphpad Prism version 7.0 statistical software and the values are expressed as mean ± standard error of mean (SEM). Statistical significance was considered at p < 0.05 using the One-way ANOVA with post-hoc Dunnett's multiple comparisons test. The asterisk (^*), (^**), and (^***) show p < 0.05, p < 0.01, and p < 0.001, respectively.

The in vitro antiproliferative activity of the crude leaf extract of C. orbuculata against HCT116 colorectal carcinoma, OE33 esophageal adenocarcinoma and KYSE70 esophageal squamous carcinoma cell lines was evaluated using WST-1 cell viability assays at concentrations ranging from 100 to 3.1 μg/ml. The assay was carried out to determine the effective concentrations for use in subsequent experiments. The percentage cell viability results are shown in Figure 1. The WST-1 assay showed that the crude extract of C. orbiculata dose dependently decreased the viability of HCT116, OE33 and KYSE70 cancer cell lines with LC₅₀ values of 64.9, >100, and 36.9 μg/ml, respectively (Figure 1A). The KYSE670 esophageal cancer cells were most susceptible to the extract, followed by HCT116 colorectal cancer cells and lastly OE33 esophageal cancer cells.

To determine the mechanism of cell death induced by C. orbiculata extract, we used Annexin-V/PI flow cytometry (Figure 1B), and found that programmed cell death (apoptosis, detected by an increase in the proportion of AnnexinV positive, PI negative cells), was induced after treatment with 35 μg/ml or above of the extract. Staurosporine at 10 nM was used as a positive control. To confirm that this was due to apoptosis, cleaved caspase-3, a key mediator of programmed cell death was stained for in HCT116, OE33, and KYSE70 cancer cells by immunocytofluorescence in cancer cells treated with varying concentrations of the plant extract (LC₅₀ × 2, LC₅₀, and LC₅₀/2) (Figures 1C–E). Quantification showed an increase in the amount of cleaved caspase-3 in all three-cancer cell lines when comparing treated and untreated cells (Figure 1F, Supplementary Figure 1). Quantification of the images show a significant dose dependent increase in caspase-3-cleavage/cell number with up to a 12-fold increase in caspase-3-cleavage in HCT116 cancer cells treated with the plant extract when compared to the untreated cells (p < 0.0001 at 140 μg/ml) and up to a 2-fold increase in both OE33 and KYSE77 (Supplementary Figure 1).

To determine the mechanisms through which C. orbiculata drives apoptosis, we subjected HCT116 cells to 48 h treatment with 50 μg/ml of the extract, followed by RNA extraction, and RNASeq and differential gene expression (Supplementary Table 1) and differential isoform expression determined (data available on request). Gene ontology analysis of the RNASeq results based on expression showed that there was significant upregulation of various pathways, including movement of cell or subcellular components, inflammatory response, cell motility, localization of cell, and small molecule metabolic process at a significance of >10⁻⁴, but none of these pathways were linked to apoptosis or cell death, and the significance was not strong (gene ontology data available upon request). GSEA on Hallmarks of Cancer identified apoptosis and TNFα through NFκB as being significantly altered (Figure 2A). There was no obvious upregulation of genes involved in a coordinated manner to induce apoptosis. However, when we ran an analysis of splicing using rMATS, we identified by gene ontology analysis that RNA processing events were the most common pathways activated by C. orbiculata (Figure 2B, Supplementary Table 2). We examined the top 25 pathways ranked by False Discovery Rate (FDR) for each of the five types of splicing event—retained introns (RI), alternative 5′ and 3′ splice sites (A5SS and A3SS), mutually exclusive exons (MXE) and skipped exons (SE). Thirty-two (32) different pathways fell into this category. Of the 9 pathways that were significantly altered for four types of splicing event–5 were pathways involved in RNA regulation (Figure 2B). There were significant changes in splicing seen in all 5 categories of splicing. Of the 10,459 alternatively spliced events, 56% were skipped exons (Figure 2C). There were substantially fewer exons where exon inclusion was enhanced by C. orbiculata (45% of altered splicing events, Figure 2C), than exons inclusion was decreased by C. orbiculata (55%, Figure 2C), suggesting that C. orbiculata preferentially causes exon skipping. Retained introns were also significantly enhanced by C. orbiculata (from 7.6 to 18.4%).

Two of the highly regulated genes detected in the splicing environment that have been linked to apoptosis and RNA splicing, respectively, were, B Cell Lymphoma 2 like protein 1 (BCL2L1), and hnRNPA2/B1. BCL2L1 encodes for two splice variants of the protein Bcl-x, a large isoform (Bcl-xl) and short isoform (Bcl-xs) differentiated by alternative 5′ splice site selection in exon 2. Inclusion of the distal splice site (i.e., BCl-xl) was reduced from 97 ± 1.5% to 89 ± 0.5% inclusion (p = 0.0055). hnRNPA2/B1 codes for two isoforms, hnRNPA2 and hnRNPB1. hnRNPA2B1 exon 2 exclusion (resulting in B1 expression) was induced by treatment with C. orbiculata, inclusion of the exon fell from 22 ± 2.4 to 16 ± 1.6% (p = 0.00182). Neither gene was statistically significantly altered when analyzed by EdgeR, but were significantly altered by DESeq2 analysis (treated was 67% of control, p = 0.004 for BCL2L1 and treated was 54% higher than control for hnRNPA2B1, p = 4.3 × 10⁻¹⁸). Total transcript levels (in counts) were BCL2L1 8,128 ± 1,214 control, 3,692 ± 327 treated, hnRNPA2B1 16,252 ± 2,783 vs. 28,585 ± 1,494). To verify this effect we carried out PCR for these two genes on cells treated with C. orbiculata. As shown in Figure 3, treatment of HCT116, OE33, and KYSE70 cancer cells with varying concentrations of C. orbiculata extract results in modulation of splicing of both hnRNPA2B1 and BCL2L1 (Figure 3A). The extract of C. orbiculata increased alternative splicing of hnRNPA2B1 in a dose dependent manner resulting in reduced expression of hnRNPB1 in all 3-cell lines (Figure 3A). The splicing of BCL2L1 was also affected resulting in an increase in Bcl-xs (Figure 3A). In all cases, the hnRNPB1 splice variant was found to be downregulated in a concentration dependent manner whilst Bcl-xs was upregulated in a concentration dependent manner in cells treated with the crude extract of C. orbiculata (Figure 3B), consistent with the RNASeq data. This switch in splicing resulted in a decrease in the ratio of hnRNPB1 to hnRNPA2 and an increase in the ratio of Bcl-xs to Bcl-xl in all 3-cancer cell lines (Figure 3B). These results suggest that C. orbiculata could be acting as a splicing modulator to induce apoptosis through hnRNPA2/B1 and BCL2L1.

To determine whether this specific switch in splicing is a common mechanism when splicing is altered, we used pladienolide B, a naturally occurring macrolide with both antitumor activity and mRNA splicing inhibition activities and measured the effect on splicing of these two genes. Pladienolide B dose dependently (0–10 nM) induced apoptosis in HCT116 colon cancer cells as evident by immunocytochemical analysis of caspase-3-cleavage (Figure 4A). The macrolide significantly increased caspase-3-cleavage/cell number up to 3-fold (p < 0.01, Figure 4B). Additionally, pladienolide B significantly decreased viability of HCT116 cancer cells at all tested concentrations ranging from 0 to 10 nM (p < 0.0001, Figure 4C). RT-PCR evaluation shows that pladienolide B switched splicing of hnRNPA2B1 and BCL2L1 (Figure 4D). Quantification of splice variant expression demonstrates a significant dose dependent decrease in hnRNB1 (p < 0.1), and a significant dose dependent increase in Bcl-xs (p < 0.1). Interestingly, pladienolide B, unlike C. orbiculata resulted in an increase in hnRNPA2 and a decrease in Bcl-xl shown in Figures 4E,F. These results suggest that the switch in splicing seen by C. orbiculata is consistent with it containing a naturally occurring splicing factor modulator, but that it could modulate a different pathway from Pladienolide B. hnRNPA2/B1 is a well-described splicing factor, having been shown to regulate multiple genes, including VEGF, Nav1.6, ANXA7 and the activity of kinases such as MEK. However, it is not clear whether the induction of the pro-apoptotic splice variant of BCL2L1 is independent of a splice form specific reduction in hnRNPB1 or a direct consequence of it. We therefore set out to determine whether hnRNPA2B1 could regulate BCL2L1 splicing.

As C. orbiculata reduced the B1 isoform of hnRNPA2B1, we mimicked this by siRNA knockdown specifically of the hnRNPB1 isoform in HCT116 colon cancer cells. Cells were transfected with 2 μg of siRNA. The most effective silencing was at 48 h post transfection (Figures 5A,B). The knockdown of hnRNPB1 also caused a switch in splicing of BCL2L toward Bcl-xs (Figures 5A,C). Quantification of splice variant expression shows a decrease in hnRNB1 and an increase in Bcl-xs when comparing untransfected cells with cells transfected with 2 μg of hnRNPB1 siRNA (Figures 5B,C). This indicates that hnRNPB1 modulates splicing of BCL2L1. The transfection experiment also showed that induction of apoptosis in HCT116 colon cancer cells is replicated by isoform specific knockdown of hnRNPB1 using siRNA for 24, 48, and 72 h (Figure 5D). Silencing hnRNPB1 resulted in a significant increase in caspase-3-cleavage when comparing transfected cells to wild type/untransfected cells (p < 0.01, Figure 5E). There was up to a 3-fold increase in caspase-3-cleavage/cell number in transfected cells. Additionally, using WST-8/cell 8 cell viability assay, we found that hnRNPB1 siRNA significantly inhibited cell viability in transfected cells (p < 0.01, Figure 5F).

To confirm that the C. orbiculata extract was inducing apoptosis through switching splicing from hnRNPB1 we measured the effect of treatment after over-expression of hnRNPB1 isoform in HCT116 cells. Fourty eight (35) hours after transfection with hnRNPB1 plasmid, RNA was extracted, DNAse treated and subjected to RT-PCR. Figure 6A shows that cell transfected with the hnRNPB1 plasmid had increased hnRNPB1 expression as expected. When cells were treated with C. Orbiculata extract there was a small, significant reduction in the percent of viable cells as measured by PI-Annexin V flow cytometry, but this was much less than that induced in the wild type (Figure 6B). This protection from C. orbiculata extract was due to a decrease in apoptosis as measured by annexin V positivity in the transfected cells compared with the wild type (Figure 6C). These results clearly demonstrate that C. orbiculata extract induces apoptosis through reduction in the hnRNPB1 isoform.

Switching splicing of hnRNPA2B1 from the B1 to the A2 isoform should be able to alter the specific RNAs that the protein can interact with. We therefore determined the effect of C. orbiculata crude extracts on association of hnRNPA2B1 with selected mRNAs known to be either regulated, or involved in apoptosis (Table 2), by RNA- immunoprecipitation (RIP) assay. We evaluated the association of hnRNPA2B1 with BCL2L1, BCL2, MDM2, cMYC, CD44, CDK6, cJUN, and TXNP in HCT116 colon cancer cells treated with 70 μg/ml of the crude extract. Treatment of HCT116 colon cancer cells with 70 μg/ml of C. orbiculata extract resulted in an increase in the co-precipitation of hnRNPA2B1 with BCL2L1 (Figure 7A), BCL2 (Figure 7B), MDM2 (Figure 7C), cMYC (Figure 7D), CD44 (Figure 7E), CDK6 (Figure 7F), and cJUN (Figure 7G) mRNA. The highest fold enrichment was observed for CDK6 (6,333-fold), followed by cJun (5,947-fold), CD44 (817-fold), cMYC (595-fold), MDM2 (416-fold), and BCL2L1 (120-fold) in HCT116 cells treated with C. orbiculata. There was no change in the binding of TXNP (Figure 7H). These results clearly show that a reduction in the hnRNPB1 isoform by C. orbiculata extract changes the interaction with multiple RNAs of genes associated with apoptosis, indicating that hnRNPA2B1 is a key regulator of apoptosis through its RNA binding activity and effect on alternative splicing.