Loss of p53 epigenetically modulates epithelial to mesenchymal transition in colorectal cancer

Highlights

• •Loss of p53 confers aggressive mesenchymal phenotypes upon EMT induction in CRCs.
• •SATB1 shows agonistic relationship with EMT TFs SNAI1, SNAI2 and VIM with loss of p53.
• •Lysine Demethylase KDM6B epigenetically regulates EMT via SNAI1.
• •SATB1 expression in CRC patients correlates with enhanced levels of ZEB1 and KDM6A.

Abstract

Epithelial to Mesenchymal transition (EMT) drives cancer metastasis and is governed by genetic and epigenetic alterations at multiple levels of regulation. It is well established that loss/mutation of p53 confers oncogenic function to cancer cells and promotes metastasis. Though transcription factors like ZEB1, SLUG, SNAIL and TWIST have been implied in EMT signalling, p53 mediated alterations in the epigenetic machinery accompanying EMT are not clearly understood. This work attempts to explore epigenetic signalling during EMT in colorectal cancer (CRC) cells with varying status of p53. Towards this, we have induced EMT using TGFβ on CRC cell lines with wild type, null and mutant p53 and have assayed epigenetic alterations after EMT induction. Transcriptomic profiling of the four CRC cell lines revealed that the loss of p53 confers more mesenchymal phenotype with EMT induction than its mutant counterparts. This was also accompanied by upregulation of epigenetic writer and eraser machinery suggesting an epigenetic signalling cascade triggered by TGFβ signalling in CRC. Significant agonist and antagonistic relationships observed between EMT factor SNAI1 and SNAI2 with epigenetic enzymes KDM6A/6B and the chromatin organiser SATB1 in p53 null CRC cells suggest a crosstalk between epigenetic and EMT factors. The observed epigenetic regulation of EMT factor SNAI1 correlates with poor clinical outcomes in 270 colorectal cancer patients taken from TCGA-COAD. This unique p53 dependent interplay between epigenetic enzymes and EMT factors in CRC cells may be exploited for development of synergistic therapies for CRC patients presenting to the clinic with loss of p53.

Graphical abstract

Wild type p53 binding sites are associated with promoters of SATB1 and with the loss of p53, these promoters are occupied by repressive histone marks H3K27me3 correlating with decreased expression of SATB1 (unpublished work from the lab). Expression of SATB1 antagonistically relates with EZH2 which correlates with our findings. Upon EMT induction with TGF-β treatment, the expression of SATB1 increases in p53 null CRC cells. The higher expression of SATB1 also correlates with increased level of SNAI1, SNAI2 and VIM suggesting that chromatin regulation plays a vital role in regulating EMT.

Introduction

Epithelial to Mesenchymal transition is a highly dynamic and programmed event occurring in a broad range of tissue types, developmental stages and cancer metastasis. The EMT program enables the epithelial cells to acquire mesenchymal state through a series of intermediate phenotypic changes along the E-M axis [1]. During this transition, epithelial cells with cobble stone morphology lose cell-cell adhesions and reorganize the cytoskeleton transiting into mesenchymal states with spindle-like morphology and exhibit increased motility, invasiveness and metastasis. EMT is associated with transcriptional reprogramming of core EMT-activating transcription factors (EMT-TFs) which include Snail (also known as SNAI1), Slug (also known as SNAI2), Twist-related protein 1 (TWIST1), zinc-finger E-box-binding homeobox 1 (ZEB1) and ZEB2 [2], [3], [4], [5], [6]. These EMT-TFs promotes metastasis by suppressing the expression of key epithelial genes such as CDH1 which encodes the adherens junction protein E-cadherin and by enhancing the expression of key mesenchymal genes N-cadherin and Vimentin [7], [8], [9]. Various knockout and over expression studies of these transcription factors have shown aggressive phenotypes highlighting the inherent role of EMT-TFs in metastasis [10], [11], [12]. Multiple signalling events like TGFβ and growth factor signalling, Notch ligand and Wnt signalling, and hypoxia are known to induce and regulate EMT [13].

Though several signalling pathways have been studied extensively to understand the molecular mechanisms of EMT, epigenetic regulation of EMT has not been completely understood. The dynamic process of EMT and transcriptional reprogramming of gene expression involved during this process highlights that reversible epigenetic regulation plays an important role in this process. Chromatin regulators play a key role in governing this cellular plasticity.

Global mapping of the accessible chromatin regions using FAIRE seq (Formaldehyde-assisted isolation of regulatory element sequencing) and ATAC Seq have shown that EMT is regulated through alteration of chromatin accessibility by Ras-induced transformation and TGFβ signaling [14,15]. TGFβ-induced EMT in various cancers have shown enhancer rewiring and functional alterations during EMT transition [16,17]. Genetic mutations associated with various chromatin regulators function as driver mutations in metastatic cancer. Gain-of-function mutations and overexpression of PRC2 methyl transferase Enhancer of Zeste Homolog 2 (EZH2) have been associated with a higher risk of cancer progression or recurrence [18]. EZH2 cooperates with SOX4, a mesenchymal gene in breast cancer cell EMT, whereas PRC2 inhibition by EED deletion or through pharmacological intervention promotes EMT in kRas-driven lung cancer cells [19,20]. Although genetic mutations associated with these chromatin regulators are known to drive cancer progression, the crosstalk between epigenetic modifiers and EMT-signalling factors in promoting invasion has not been well investigated.

Tumour suppressor p53 modulates several intercellular cascades to directly or indirectly modulate EMT processes. Hence the mutational state of p53 in cancer cells is a major determinant of phenotypic changes associated with EMT in cancer. Single amino acid substitutions on the DNA binding domain of p53 leads to loss of its tumour suppressor functions [21,22]. These mutations occurring in the DNA binding domain of p53 gain oncogenic functions, enhanced metastasis and drug resistance. Mevalonate and etoposide resistance pathways have been implicated in the altered function of p53 regulation with other transcription factors.

Loss of p53 activity correlates with increased genome instability, cell proliferation and higher prevalence of cancer [23]. Studies using in-vivo and in-vitro model systems have shown that mutations in the TP53 gene are associated with enhanced Missense mutations account for more than 70 % of TP53 mutations in human cancers. Most of these are clustered as hotspots in the DNA Binding Domain (DBD) regions of TP53. These DNA binding domain mutations can directly or indirectly affect the DNA-binding ability of p53. Early stages of tumorigenesis on the DBD regions show dominant-negative effects by inactivating the function of wild-type p53. Late stages of cancer progression accompanied by the loss of wild-type TP53 allele, exhibit oncogenic GoF activities that enhance metastasis [24].

Loss of p53/Gain of function mutation in p53 activates the EMT programme, accompanied by an increased stem cell population via modulating the expression profile of various EMT-TFs including ZEB1, SNAI1. Crosstalk of p53 mediated suppression of EMT-TFs is also facilitated by non-coding RNA such as miR200 family [25], [26], [27], [28]. Transforming growth factor-beta (TGF-β) regulates invasion and metastasis during cancer progression through loss of epithelial markers and gain of mesenchymal markers [29].

Mutant p53 generally subverts tumour suppressive TGFβ responses through interaction with Smads. Smads and mutant p53 complex further block p63-mediated activation of metastasis suppressing genes to promote tumour progression suggesting a crosstalk mechanism between the two in cancer metastasis [30]. Epigenetic regulation mediated by TGF β signalling enables cells to respond in a context dependent manner. This epigenetic regulation is largely mediated by interaction of various chromatin regulators with SMAD complex [31]. Jumonji domain containing-3 (JMJD3 also called KDM6B) has also been shown to promote TGF-β-mediated Smad activation and EMT across various cancers [32]. Although direct and antagonistic relation of various epigenetic regulators and EMT has been shown across various cancers, the role of p53 in mediating the crosstalk of various epigenetic modifiers and EMT genes during the transition or cancer progression has not been explored. Evidences of cross talk between gain of function mutations of p53 and epigenetic regulation in the context of EMT have not been explored in detail.

This work attempts to explore the chromatin pathways and epigenetic events that regulate p53 mediated EMT in colorectal cancer. Towards this, we have used 4 colorectal cancer cells lines with varying p53 status and have induced EMT in each of the cell lines with TGF-β. Our experiments imply that significantly aggressive EMT phenotypes emerge with the loss of p53 and are mediated by alterations in epigenetic modifications on the promoters of EMT factors. We observed that the epigenetic regulation showed highest correlation with EMT-TFs with loss of p53 as compared to wild type and mutant p53 suggesting a synergistic approach of EMT-TFs and epigenetic factors in driving cancer metastasis upon p53 loss. Interestingly, the EMT Factor SNAI1 was highly correlated with KDMs such as KDM6A and KDM6B and with SATB1 and was antagonistically related to methyltransferase SETDB1. This unique observation of epigenetic modulation of SNAI1 by histone demethylases associated with loss of p53 suggests a p53 mediated epigenetic crosstalk with EMT induction. Our findings provide a novel framework for developing synergistic therapies with epigenetic inhibitors and conventional cancer therapeutic drugs for colorectal cancer patients presenting to the clinic with loss of p53. This work confers a specific and hitherto unaddressed role for p53 in regulating EMT plasticity and in cancer metastasis.

Materials and methods

Cell culture

Colorectal cancer cell lines with different p53 status were considered for the study. HCT116WT (p53 Wild Type), SW480 (p53 mutant: R273H, P309S), HT29 (p53 mutant: R273H) cell lines were purchased from NCCS, Pune, India and HCT116 p53-/- (Colon origin, p53 null) cell line was a kind gift from Vogelstein Lab through Jayanta Sarkar Lab at CDRI, Lucknow, India. All four CRC cell lines were cultured in Dulbecco Modified Eagle's Medium (DMEM-F12 (#12,500,062 Invitrogen-Gibco) supplemented with 10 % Fetal Bovine Serum (FBS) (#16,000,044 Invitrogen Gibco) and 1 % Penicillin/Streptomycin (#1,507,070,063 Invitrogen-Gibco) and were maintained at 37 °C in a 5 % CO₂ incubator. The cells were cultured for 24 h for adherence and to attain around 60 % confluency before they were subjected to pharmacological intervention with TGFβ.

Induction of epithelial to mesenchymal transition (EMT) in CRC cell lines

TGFβ (#240-B-002, R&D Systems), a potent EMT inducer was reconstituted at 20 μg/mL in sterile 4 mM HCl containing 1 mg/mL Bovine Serum Albumin (#160,069-Albumin, Mp-Bio) as per manufacturer's instructions. Final concentrations of 10 ng/ml were made using DMEM-F12. The plates treated with TGFβ along with the relevant non-treated controls were taken for further experimentation after 48 h of treatment.

Scratch assay

Colorectal cancer cells with different p53 status were plated at a density of 0.05 million cells/ml on 35 mm petri dishes. The cells were incubated overnight and left for 24 h for adherence. The culture media was supplemented with 10 ng/ml of TGFβ to induce epithelial to mesenchymal transition (EMT). This was followed by creation of a scratch or wound with a 200 μl pipette tip. Cells located at the edge of the wound undergo polarization and migrate into the wound space [33]. The migration of the cells was an indicator of the proliferation kinetics. The percentage of wound healing was calculated using the following formula:

$$\text{Wound}\text{Healing}\text{Percentage}=\left[\frac{{\mathrm{A}}_0-{\mathrm{A}}_{\mathrm{t}}}{{\mathrm{A}}_0}\right]\times 100$$

where, A₀=0 h, represents the scratch area measured at 0 h and A_t represents the scratch area measured at 24, 48 or 72 h.

Transwell migration assay

2 × 10⁵ control and TGFβ treated colorectal cancer (CRC) cells were seeded on Matrigel (#354,234 Corning) in media without FBS in the ratio of 1:1 ratio and coated on upper Boyden chambers (24-well inserts; pore size 8um) (# 354,578 Corning.) in serum-free media and were incubated for 48 h. The inserts accommodate porous membranes made up of matrigel in media without FBS that allow migration of the cells towards the bottom chamber. The bottom chamber was filled with complete media supplemented with 10 % FBS, which acted as a chemo attractant to stimulate invasion. The inserts were removed and fixed using pre-chilled 70 % ethanol. Matrigel was removed and the membrane with invaded cells was stained with 0.2 % crystal violet solution (#TC510-25 G Hi-Media) for 10 min. The number of invaded cells was counted in five representative fields of the membrane under a light microscope (20X objective). Cell migration was quantified using ImageJ software [34].

Immunoblotting

The control cells and the TGFβ treated CRC cells were harvested after 48 h, and cell lysates were prepared using RIPA buffer supplemented with Protease Inhibitor Cocktail (#11,697,498,001 Sigma Aldrich). The protein concentration in the lysates was determined using the Bradford method. Subsequently, 40 µg of protein from each sample was resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) [35]. The separated proteins were then transferred onto a polyvinylidene fluoride (PVDF) membrane (#1,620,177- BioRAD), followed by blocking using 1 % BSA and primary antibody incubation at 4 °C overnight. The following antibodies were used in the study β-actin (# 4970S Cell Signaling, Technology 1:1000), E- cadherin (#3195S Cell Signalling Technology,1:1000) and Vimentin (5741S Cell Signalling Technology,1:1000). Primary antibodies were incubated with the appropriate goat anti-rabbit IgG secondary antibody conjugated with horseradish peroxidase (HRP) (#7074P2 Cell Signalling Technology) (1:2000), before signal detection using the enhanced chemiluminescence (ECL) system (#170-5060 Biorad). The protein bands were quantified using ImageJ software [34].

Flow cytometry

In order to confirm EMT induction with TGFβ treatment, the levels of E-cadherin and Vimentin on the CRCs cells were measured using, flow cytometry. Following TGF β treatment, cells were harvested after 48 h. 2 × 10⁶ cell aliquots per condition were fixed by adding 1 % formaldehyde for 15 min at room temperature. Cells were then centrifuged at 2500 rpm for 5 min at 4 °C to pellet cells and remove formaldehyde and washed twice with 1 X PBS. Cells were permeabilized and blocked with 0.1 % saponin supplemented with 1 %BSA and 1XPBS. The cells were incubated for 15 min at room temperature. Cells were washed twice with 1X PBS. Flow cytometry for the detection of E-cadherin and Vimentin population in CRCs was performed by double labelling of single-cell suspensions using a combination of PE-conjugate rabbit anti-human Vimentin (1:50, #12020S Cell Signalling Technology) and Alexa Fluor 647 labeled mouse anti-human E-cadherin (1:50, #77381S Cell Signalling Technology). All antibody incubations were performed in PBS with 1 % BSA supplemented with 0.1 % saponin for 1 h at room temperature. Cells were washed with 1X PBS and prepared for flow cytometry analysis.

Clinical samples

Three paired tumor and adjacent normal tissues from patients with colorectal cancer were collected from the HCG hospitals Cancer Centre Bengaluru immediately after ressection. All tissue samples were immediately stored in RNAlater and stored at 4 °C until RNA extraction. The approval of the institutional ethics committee was obtained before starting the work (Institutional Ethics Committee (IEC) number IBABIEC-04/PR01/1,281,219

RNA isolation from patient tissues

Three CRC cancer tissues and three matched normal tissues were chosen from 3 different CRC patients for RNA isolation. Fresh tumors were collected in RNA later (Thermoscientific). Diethyl Pyrocarbonate (DEPC) was used for removing RNase from all experimental setups. 80–100 mg of each tissue was crushed and homogenize using mortar pestle in 500μl of TRIzol reagent (RNA iso) (#9108 Takara). 1/5th volume of chloroform was added to each sample for phase separation. The aqueous phase was carefully transferred to fresh tubes and equal volume of isopropyl alcohol was used for precipitation. The samples were centrifuged at 12,000 rpm for 10 min at 4 °C. The pellets obtained were washed with 80 % ethanol to remove contaminants and impurities and RNA obtained were quantified using nanodrop.

RNA isolation from cell lines

10⁵ Cells were seeded in each of the wells of six well plates. Cells were treated with TGFβ for 48 h prior isolation. Cells were lysed with 500 µl of RNA isoreagent (#9108 Takara) to extract total RNA from TGFβ treated and control CRC cells. Following phase separation, the RNA was precipitated using equal volume of isopropyl alcohol and washed with ethanol to remove contaminants and impurities. The purified RNA was then dissolved in nuclease-free water, and its quality and quantity were assessed using nanodrop.

cDNA synthesis and qRT-PCR analysis

The RNA concentration was detected using a NanoDrop 2000 (Thermo Fisher Scientific, USA). For gene expression analysis, 5ug of total extracted RNA from cell lines, while 2ug of total RNA from patient tissues were used to generate cDNA. RNA obtained was treated with DNase (NEB- M0303S) to remove the traces of contaminating DNA. cDNA synthesis was performed using random hexamer primers (NEB #S1330S) and M-MuLV Reverse Transcriptase (NEB M0253S) as per manufacturer instructions. Real-time-PCR analysis was performed using Universal SYBR master mix (CST # 88989S) on the Applied Biosystems StepOnePlus Real-Time PCR System with the following conditions: 3 min at 95 °C Denaturation, Annealing and Extension (15 s at 95 °C, 1 min 60 °C) for 40 cycles followed by melt curve (15 s 95 °C,1 min 60 °C, 15 s 95 °C). Relative gene expression was calculated using the 2^−△△Ct method after normalizing C_t-values to those of the housekeeping gene GAPDH. All primers used in the study are listed in Supplementary Table S1.

Library preparation and RNA sequencing

500 ng of isolated RNA was used for library preparation. RIN Values for all the samples were calculated and found to be above 9.5. RNA-sequencing libraries were prepared with Illumina-compatible NEBNext® Ultra™ II Directional RNA Library Prep Kit (New England BioLabs). 500 ng of total RNA was taken for mRNA isolation, fragmentation and priming. Fragmented and primed mRNA was further subjected to first strand synthesis followed by second strand synthesis. The double stranded cDNA was purified using NEB Next sample purification beads. Purified cDNA was subjected to end-repair and adenylation and was ligated to Illumina adapters as per NEBNext® UltraTM II Directional RNA Library Prep protocol followed by second strand excision using USER enzyme at 37 °C for 15 mins. Adapter ligated cDNA was purified using JetSeq Beads and was subjected to 11 cycles for indexing-(98 °C for 30 s, cycling (98 °C for 10 s, 65 °C for 75 s) and 65 °C for 5 min) and enrich the adapter-ligated fragments. Final PCR products (sequencing library) were purified with JetSeq Beads, followed by library quality control check. Illumina-compatible sequencing library were quantified by Qubit fluorometer (Thermo Fisher Scientific, MA, USA) and fragment size distribution was analyzed on Agilent 2200 Tape Station. The libraries were sequenced on Illumina Next Seq platform to generate 40–50 million paired end reads/sample.

RNA sequencing data analysis

The quality of the paired-end raw reads generated from RNA sequencing using the Illumina NextSeq platform was assessed using FastQC(v0.1.1), a widely used tool for evaluating sequencing data quality [36]. Adapter trimming was performed using Trimmomatic [37] for relevant sequences. The paired end RNA sequencing reads were mapped to GRCh38/hg38 reference genome using STAR (v2.5.3a) aligner [38]. To perform differential expression analysis, raw read counts matrix was estimated by HTseq count. Differential expression analysis was carried out using R/Bioconductor package DESeq2 (v1.22.1) [39]. Genes with FDR less than 0.05 and absolute fold change more than |1| were considered as differentially expressed genes. DESeq2 employs a negative binomial distribution model to assess differential expression, taking into account biological variability and sample-specific effects. Heatmaps were generated using log of normalized count values with pheatmap (v1.0.10) R package.

Calculation of EMT scores of CRC cell lines upon EMT induction

Three unique methods were employed to determine the EMT status of CRC cells with varying p53 status upon TGFβ treatment 76GS (76 Gene Signatures) [40,41], KS (Kolmogorov-Smirnov) [42], and MLR (Multinomial Logistic Regression) [43]. The 76GS score was calculated based on the expression levels of 76 specific genes. Higher scores are indicative of the indicate epithelial state of the cell. A 76GS score greater than 0 typically signifies an epithelial phenotype, while 76GS score less than 0 indicates a mesenchymal phenotype. The score for each sample was computed by weighing the expression values of the 76 genes, with the weight determined by the correlation of each gene expression with that of CDH1 in the given dataset.

The KS score was determined through a statistical Kolmogorov-Smirnov two-sample test. Using a set of 218 genes expression levels, cumulative distribution functions were estimated for mesenchymal and epithelial signatures. The statistic for the two-sample KS test was derived from the maximum difference between these cumulative distribution functions (CDFs) [42] KS scores range from −1 to 1, where negative scores represent epithelial phenotypes and positive scores represent mesenchymal phenotypes.

MLR scores were provided on a scale of 0 to 2, with higher scores indicating a higher prevalence of mesenchymal samples [43]. The scoring was achieved through an ordinal multinomial logistic regression, which incorporates an ordered structure. The MLR score encompassed a hybrid epithelial/mesenchymal signature that lies between the epithelial and mesenchymal phenotypes. The scores were calculated based on the probability for each sample to belong to one of the three phenotypes.

ssGSEA (single sample gene set enrichment analysis) on CRC cells for EMT scoring

ssGSEA (single sample Gene Set Enrichment Analysis) was done to calculate the individual enrichment scores for each pair of CRC cell lines (TGFβ untreated and TGFβ treated) and gene sets [44]. In ssGSEA, the enrichment score for a specific gene set reflects the extent to which the genes within that set are collectively up or down-regulated within a particular sample. For this analysis, the HALLMARK_EMT, gene set obtained from the Molecular Signatures Database (MSigDB) was utilised. The enrichment scores were computed using the Python package gseapy (https://github.com/zqfang/gseapy). All statistical analyses were performed using R version 4.2.0. The ggplot2 function was employed to generate the plots for visualization.

Gene set enrichment analysis (GSEA)

To identify the significantly enriched biological processes or pathways associated with a set of differentially expressed genes (pval<0.05 and log2FC|1|), GSEA [44] was performed. The gene sets used in the analysis were obtained from the set of differentially expressed genes. The enrichment analysis was performed using the default settings and statistically significant enrichment results were determined based on the false discovery rate (FDR) adjusted p-value threshold of 0.05.

Determining agonistic and antagonistic relationship between EMT and epigenetic factors

RNA sequencing of CRC cells with varying p53 status treated with TGFβ were analysed to obtain raw counts. Raw counts were obtained using HTseq-count and were normalized for gene length and were transformed to calculate TPM (transcripts per million) values. A list of epithelial-mesenchymal transition (EMT) and epigenetic factors was compiled from the EpiFactor database [45], [46], [47]. The TPM counts corresponding to these factors were extracted from the normalized TPM values. To explore the potential agonistic and antagonistic relationships between EMT and epigenetic factors, pairwise Spearman correlation analysis was conducted using corrplot R package. To visualize the agonistic and antagonistic relationships between EMT and epigenetic factors, correlation heatmaps were constructed in R using corrplot function.

Validation of the observed markers using TCGA-COAD cohort

Raw expression counts for 479 primary colon cancers analyzed as part of the TCGA-COAD program was obtained from the TCGA data portal website (http://tcga—data.nci.nih.gov/tcga) using TCGA-biolink [48]. The count matrix was normalized (using TCGAanalyze_Normalization function) and genes with low expression count were filtered out. We then classified the 479 primary colon cancer obtained from TCGA-COAD into two groups, High SATB1 and Low SATB1 based on the median value of SATB1. Differential gene expression analysis was carried out between these two groups using limma method through TCGAanalyze_DEA function(with method = glmLRT).

Identification of p53null and p53 mutant CRC patients from TCGA and gene expression analysis

The mutations were classified into p53null mutation(N = 27), p53R273H(N = 12) and p53 WT(N = 12) based on the somatic mutations reported for colon cancer (TCGA-COAD). The p53 null group contained frame-shift, nonsense and splice-site mutations, as a consequence of intragenic nucleotide insertions or deletions and/or single base-pair substitutions, which are expected to affect protein encoding reading frames and have been associated with unstable transcripts and lack of protein. Only samples with known TP53 mutations were used in our analyses. The raw expression counts of protein-coding genes for colon cancer samples were obtained from the GDC portal using TCGAbiolinks. Further, the count matrix was normalized (using TCGAanalyze_Normalization function) and filtered out genes with low expression count (using TCGAanalyze_Filtering with method = quantile, and qnt.cut = 0.25).The differential expression analysis was carried out using limma method through TCGAanalyze_DEA function(with method = glmLRT). Genes with |log2foldchange| > 0.5 and FDR (q-value) < 0.1 were significant and differentially expressed. The difference in expression in the TCGA samples was computed using t-test.

Survival analysis for TCGA COAD patients

To investigate the clinical significance of the uniquely expressed epigenetic enzymes in the HCT116p53-/- CRC cell line and combination of EMT-Epigenetic factors, survival analysis was performed on the TCGA-COAD (The Cancer Genome Atlas - Colon Adenocarcinoma) cohort [49] for 270 colorectal cancer patients. The Kaplan-Meier curves were generated using GEPIA2 [50] to screen the hazard ratio of list of epigenetic factors and the combination of EMT-Epigenetic factors in colorectal cancer, which were selected with p-value<=0.05 and the “median” as the group cutoff.

Results

Colorectal cancer cells with deficient p53 exhibit higher migration and invasion on EMT induction

In order to understand how the varying mutational status of p53 influences EMT induction in colorectal cancer cells, a wound healing assay was performed with HCT116WT (p53 WT), HCT116p53-/- (p53 null), SW480 (p53 double mutant: R273H, P309S) and HT29 (p53 single mutant: R273H) cells. These cells were treated with TGFβ for 6,12,24, 48 and 72 h). Wound healing assay revealed that TGF-β increased migration of CRC cells towards the center of the scratched area (Supplementary Fig. S1A). Since significant closure was observed in 48 h, it was set as a time point for further downstream experiments. As expected, the proliferation was found to increase with time in all the cell lines investigated, leading to wound closure. However, with loss of p53, we observed the wound closure to be faster as compared to other CRC cell lines correlating with percentage decrease in the area of the wound as depicted in Fig. 1a and Fig. 1b. The kinetics of wound healing in the other cell lines at different time points is depicted in Supplementary Fig.S1B. It was also observed that the HT29 colorectal cells (with a mutation at R273H on p53) show the least migration correlating more to an epithelial phenotype. Hence further downstream experiments were performed on HCT116 p53 WT cells and on HCT116 p53 null cells.

Fig. 1.

Confirmation of EMT induction upon TGFβ treatment in CRC cells with varying status of p53 (wound healing assays and western blotting). Wound healing assay on all 4 CRC cells (HCT116WT, HCT116p53-/-, SW480, HT29) before treatment (without TGFβ) and upon EMT Induction (with 10 ng/ml TGFβ) at time points of 0, 24, 48, 72 h. The figures demonstrate increased cell mobility to the center of the scratched area on TGFβ treatment in CRC cells with the highest mobility observed in the case of HCT116p53-/- cells. B. Analysis of the wound healing assay upon EMT induction performed at 24,48,72 h depicting percentage change in area of all CRC cell lines.The wound closure is significant and is the fastest in HCT116p53-/- CRC cell line. (One sample t-test was performed for the statistical analysis, pvalue>0.05,ns; p value< = 0.05,*; p value< = 0.01,**; p value< = 0.001,***) C. Protein level expression analysis of EMT markers using Western Blots upon EMT induction. Decreased levels of E-cadherin and increased levels of Vimentin following TGF-β treatment in the wild type and p53 null cell lines confirms EMT induction (both HCT116WT and HCT116p53-/- CRC cell line). D and E . Quantification of protein expression in HCT116WT and HCT 116 p53-/- upon TGF β treatment showing decrease in expression of E-cadherin and increase in Vimentin upon TGFβ treatment. Data are represented as the mean ± standard deviation of three independent experiments. Student's t-test has been performed compared to the Control group.

Next, we measured the protein expression levels of EMT marker genes: E-Cadherin and Vimentin in HCT116 WT and HCT116 p53 null CRC cell line through western blotting to confirm EMT induction upon TGFβ treatment. Western blotting revealed that the loss of p53 increases the levels of Vimentin by 2-fold while decreasing the levels of the epithelial marker E-Cadherin, as compared with p53WT (Fig. 1c, 1d and 1e) establishing the mesenchymal phenotype acquired by the HCT116 p53 null cells on EMT induction.

A transwell invasion assay was also performed to understand the capability of these cells to invade a transwell membrane supplemented with a matrigel. We observe that with TGFβ treatment, there is an increase in invasive potential in both WT and null CRC cells (Fig. 2a, 2b, Supplementary Fig 2A-D). Since p53 null CRC cells by themselves show strong invasive capability as shown in Fig 2b,. the change in the invasion kinetics due to EMT induction by TGFβ did not show notable differences. These observations were also confirmed by the Flow Cytometry assay with high expression of PE Conjugated Vimentin and low expression of Alexa Flour 647 E-Cadherin in p53 null CRC cells (Fig. 2c, d).

Fig. 2.

Characterisation of EMT upon TGFβ induction in CRC cell lines (transwell migration assay and flow cytometry) A. Transwell migration assay on HCT116WT and HCT116p53-/- cell lines before and after EMT induction with TGFβ for 48 h. It can be observed that EMT induction accelerates the invasion ability of CRC cells. B. Quantification of relative number of invaded cells in both HCT116 and HCT116 p53-/- cells. It is observed that HCT116p53-/- cells show maximum invasion with EMT induction. Data are represented as the mean ± standard deviation of three independent experiments. Student's t-test has been performed compared with respect to the control group. C. Flow cytometry analysis of EMT induction in CRC cells with wild type and p53 null states. Cells were stained with E-Cadherin and Vimentin and the population statined for each of these markers were measured. Vimentin (PEConjugate) (X-axis) was plotted against E-cadherin (Alexa Flour 647) (Y-axis) after TGFβ treatment in HCT116WT and HCT116p53-/-. The figure shows maximum concentration of Vimentin with EMT induction. D. Quantification of flow cytometry analysis. It is observed that EMT induction is accompanied by maximum number of Vimentin (PE Conjugate) cells in HCT116p53null CRC cells as compared with HCT116 WT cells confirming higher EMT induction than its wild type counterpart. HCT116 WT showed maximum number of E-cadherin (Alexa Flour Conjugate) cells.

EMT induction on diverse colorectal cancer cells shows a p53 dependent regulation of key genes involved in progression and metastasis

To investigate how the varying status of p53 influences key genes after EMT induction, we performed mRNA sequencing of 4 CRC Cell lines (HCT116WT, HCT116p53-/-, SW480, HT29) with various status of p53 as described in the Materials and Methods section of the manuscript. The RNA sequencing data illustrates that several key genes are differentially expressed (DEGs) across all the 4 CRC cell lines under study (significance level of pvalue<=0.05 and log2FC>=1 (upregulated DEGs); log2FC<=1 (down regulated DEGs) (Fig. 3a, 3b). Interestingly, p53 null cells showed differential regulation of maximum number of genes (5832 upon TGFβ treatment, primarily comprising 2929 upregulated genes and 2903 downregulated genes). The log2FC values of the most highly expressed upregulated genes in HCT116 p53-/- CRC cells upon TGFβ treatment were as high as +9.

Fig. 3.

Transcriptome Profiling of 4 CRC Cell Lines with varying p53 status (HCT116WT, HCT116p53-/-, SW480, HT29) upon EMT induction A. Common and unique upregulated genes expressed in CRC cell line with varying p53 status upon EMT induction (p value< = 0.05 and log2FC> = 1) B. Identification of common and unique downregulated genes expressed in CRC cell line with varying p53 status upon EMT induction (p value< = 0.05 and log2FC< = 1) C. Gene set enrichment analysis (GSEA) performed to examine the differential enrichment of gene signatures which shows that kRas signalling is positively enriched in TGFβ treated HCT116WT CRC cell line (positive enrichment score with a pvalue of < = 0.05) D. Gene set enrichment analysis (GSEA) performed to examine the differential enrichment of gene signatures which shows that G2M checkpoint is negatively enriched in TGFβ treated HCT116WT CRC cell line (negative enrichment score and with a p value of < = 0.05) E. Gene set enrichment analysis performed to examine the differential enrichment of gene signatures which shows that Epithelial to Mesenchymal Transition is positively enriched in HCT116p53-/- CRC cell line (positive enrichment score with a p value of < = 0.05) F. Gene set enrichment analysis performed to examine the differential enrichment of gene signatures which shows that DNA repair is negatively enriched in HCT116p53-/- CRC cells (enrichment score with a p value of <=0.05).

HCT116WT shows the second highest number of DEGs (1390) governing the kRas signalling pathway and the G2M checkpoints (3C, 3D). Gene Set Enrichment Analysis further indicated that the upregulated genes in p53 null CRC cells govern the EMT pathway and downregulated genes govern DNA repair pathways on induction (Fig. 3e, 3f).

The top 25 differentially expressed genes from the transcriptomic profile of 4 CRC cells after EMT induction are presented in Supplementary Fig. S3A-H. The overall transcriptomic profiles of the 4 CRC cells highlight the increase in the expression of mesenchymal genes and decrease in the expression of epithelial genes upon TGFβ treatment confirming the induction of EMT in all CRC cells (Fig. 4a). It is observed that the loss of p53 confers a higher expression of a majority of mesenchymal genes suggesting that the absence of p53 accelerates the transition of CRC cells towards EMT upon TGFβ induction (Fig. 4a). The enrichment of epithelial and mesenchymal signatures in all the 4 CRC cell lines were also quantified using single-sample Gene Set Enrichment Analysis (ssGSEA) (Fig. 4b). Of all the 4 CRC cell lines, HCT116 p53-/- upon TGFβ treatment exhibited the highest mesenchymal score as evident through EMT score. Among the untreated cell lines, HT29 shows a highest degree of epithelial characteristics and HCT116 p53-/- exhibits the highest level of mesenchymal features.

Fig. 4.

Differential expression profiles of epithelial, mesenchymal genes, and epigenetic factors with TGFβ treatment in CRC cells. A. Gene expression prolife of epithelial and mesenchymal genes represented as log2 FoldChange (Yaxis) in all CRC cells with TGFβ treatment. Median expression value of mesenchymal genes was found to be highest in HCT116 p53-/- upon EMT induction as compared to other CRC cells [p value < = 0.05 and log2FC|1|]

B. ssGSEA plot representing EMT Scoring Xaxis: Epithelial Score and Y axis: Mesenchymal Score) of each sample upon TGFβ treatment which denotes that loss of p53 shows maximum mesenchymal score as compared to other p53 status of CRC cells C. Heatmap representing top 25 upregulated epigenetic factors in HCT116 p53-/- on EMT induction with TGFβ. SATB1, SETDB2, HDACs and KDMs are upregulated upon EMT induction with p53 loss in these cells (p value< = 0.05 and log2FC> = 1) D. Heatmap representing top 25 uniquely upregulated epigenetic factors in HCT116 p53-/- cell line on EMT induction with TGFβ . KAT2B, KDMs, SMYD3, SETDB2 and ARID2 are uniquely expressed upon EMT induction with p53 loss (pvalue< = 0.05 and log2FC> = 1) in these cells.

Loss of p53 regulates EMT through regulation of key cytoskeletal, ECM and hippo signalling genes

Since the CRC cells with loss of p53 show distinct phenotypes of EMT as compared to their p53 counterparts, we attempted to focus on the EMT signalling on these cells and to investigate if the loss of p53 impinges on the cytoskeletal and extracellular matrix function and if it influences key signalling pathways associated with EMT.

Extra cellular matrix and cytoskeletal proteins are upregulated with the loss of p53 and EMT induction in CRC cells

The Extracellular Matrix (ECM) is a key player in tumourigenesis and metastatic events [51,52]. The ECM stiffness, composition and organization enable migration and invasion of cancer cells through the process of EMT [53,54]. Primary components of the extracellular matrix such as collagen and integrin regulate progression of several cancer types. It has been shown that TGFβ mediated crosslinking of collagen1 facilitates the migration and invasion of metastatic hepatocellular carcinoma [55]. Therefore we investigated the alteration and expression of ECM proteins after TGFβ treatment in CRC cells HCT116 with deficient p53. Interestingly, we observed an upregulation of integrin alpha isoforms ITGA1 and ITGA9 and integrin beta isoform ITGB8, Matrix Metalloproteases (MMP1, MMP28), collagen-alpha genes (COL3A1, COL13A1, COL4A3, COL5A3), metalloproteases ADAMTS4 and ADAMTSL4 in colorectal cancer cells with p53 null phenotype.

ITGA1 (Integrin alpha1) has been reported to be upregulated in pancreatic ductal adenocarcinoma and correlates with poor patient prognosis [56]. ITGA9 is a metastatic marker whose levels correlate with mortality in a few subtypes of cancer [57]. ITGA9 is correlated to high tumour grade, metastasis and poor survival in breast cancer [58,59]. ADAMTSL4 is an immune related biomarker for primary glioblastoma multiform and is an indicator of poor prognosis [60].

Collagen proteins are involved in cell polarity and migration besides regulation of the tumour microenvironment. Recent work from several labs have shown an association of collagen overexpression with invasion, drug resistance and poor survival [61].

The remodeling events succeeding tumour growth are marked by substitution of normal collagens in the ECM with tumour specific collagen. Collagen-alpha genes (COL3A1, COL13A1, COL4A3, COL5A3) are emerging as novel diagnostic markers in CRC. COL3A1 and COL13A1 are novel diagnostic and prognosis marker of CRC and bladder cancer [62]. These observations correlate well with the mesenchymal phenotype observed in our experiments onp53 null CRC cells.

Recent findings suggest that Autophagy and EMT (Epithelial to Mesenchymal Transition) are linked in a complex relationship.Cells that underwent EMT require autophagy activation to survive during the metastatic spreading [63]. The higher expression of Autophagy Related Gene 4A (ATG4A) in HCT116p53null CRC cells on EMT induction indicates a probable metastatic cascade triggered by autophagy leading to EMT in CRC cells.

Hippo signalling is activated with the loss of p53 in EMT induced CRC cells

The Hippo/Yes-Associated Protein (YAP) signalling pathway is proven to control organ size and maintain tissue stemness, promoting cancer metastasis via EMT [64]. We examined how EMT induction with TGFβ affects the genes in the Hippo signalling pathway to see if there are unique expression profiles that CRC cells exhibit with the loss of p53. We identified upregulation of Wnt4 and Wnt11, PARD3 and PARD6B, Amphiregulin AREG, SMAD2 and SMAD4, Ferm Domain-Containing Protein 6 (FRMD6) and Serine Threonine Kinase STK3 and the downregulation of WNT9A and SAV1.

Wnt protein acts as a potential prognostic biomarker in CRC and promotes tumour progression by activating the Wnt/β-catenin pathway [65]. Upregulation of Wnt genes (Wnt4, Wnt11) promotes tumour progression in CRC. Overexpression of Par-3 Family Cell Polarity Regulators like PARD3 and PARD6B plays a crucial role in tumourigenesis in CRC and predicts poor survival in Hepatocellular carcinoma patients [66]. AREG (Amphiregulin) is a chemosensitive potential biomarker to predict clinical outcomes after cetuximab-based chemotherapy [67]. Its upregulation in CRC tumour promotes metastasis and mediates EMT via the EGFR/ERK/NFκB signalling pathway [68]. Overexpression of FRMD6 (Ferm Domain-Containing Protein 6) promotes cell migration and invasion via the mTOR signalling pathway in lung cancer and can be used as a surrogate marker for EMT [69]. These observations imply that the loss of p53 activated EMT through regulation of Wnt genes, mTOR signalling pathway and through cell polarity regulation.

Loss of p53 alters the expression profile of epigenetic factors upon EMT induction

To investigate the differential regulation of epigenetic factors and their probable crosstalk with EMT - factors upon TGFβ induction in CRC cell lines with varying p53 status, we screened for a list of 820 epifactors through Epifactor database and analysed their expression profiles from the RNA sequencing data of all the 4 CRC cell lines under study. We observed the maximum number of differentially expressed epigenetic factors to be associated with loss of p53 CRC upon TGFβ induction as compared to wild type and mutant p53.

Out of the total 1390 differentially expressed genes in HCT116 CRC cell line (WT p53 status), 56 epigenetic factors were differentially expressed (with a log2FC|1|,pvalue<0.05) upon TGFβ induction, while in HCT116 p53-/-, out of 5832 differentially expressed genes, 254 epigenetic factors were differentially expressed (99 upregulated and 155) upon TGFβ induction (log2FC|1|,pvalue<0.05), suggesting that the loss of p53 contributes to EMT by dysregulating expression of epigenetic factors. Epigenetic writers and erasers including KAT2B, SETDB2 and SMYD3 were upregulated upon TGFβ induction in p53 null CRCs as represented in the heat map (Fig. 4c).

Interestingly, special AT-rich binding protein 1 (SATB1), a nuclear matrix protein which helps in mediating chromatin looping was found to be highly upregulated upon TGF-β treatment in p53 null CRC cells (log2FC=4.6) and weakly upregulated in p53 mutant CRC cell line SW480. Several studies have demonstrated that elevated levels of SATB1 correlate with aggressive metastatic phenotype and poor prognosis [70,71]. This suggests that EMT induction alters the genomic organization upon loss of p53. We understand from our unpublished work that WT p53 binding sites are associated with promoters of SATB1 and with the loss of p53, these promoters are occupied by repressive histone marks H3K27me3 correlating with decreased expression of SATB1. Furthermore, studies using ChIP assays has shown epigenetic regulation of SATB1 with EZH2 [72] which correlates with our findings. However, upon EMT induction via TGF-β treatment, the expression of SATB1 increases in p53 null CRCs. The higher expression of SATB1 also correlates with increased level of SNAI1, SNAI2 and VIM suggesting that SATB1 plays an important role in regulating EMT process via chromatin regulation.

On further analysis, a unique set of 175 differentially expressed epigenetic factors associated with loss of p53 and TGFβ induction were identified. These gene sets comprise 65 upregulated and 110 downregulated epigenetic factors as represented by the heat map in Figure (Fig. 4d) which shows the top 25 uniquely upregulated epigenetic factors. Interestingly, we found various histone deacetylases including HDAC 4,8 and 9, histone lysine demethylases such as KDM3A, KDM5A,5B and 7A to be upregulated with an overall downregulation of KAT7, SUV39H1, SETD1B, SETDB1 and PRMTs suggesting an overall decrease in both acetylation and methylation of histones upon TGFβ induction with p53 loss. This suggests that the loss of p53 and EMT induction exert a significant impact on both the euchromatin and heterochromatin signatures.

SETDB2 and SMYD3 were found to be upregulated in p53 null CRC cells upon TGFβ induction. SETDB2 is a histone H3 lysine 9 (H3K9) tri-methyltransferase involved in transcriptional gene silencing, the overexpression of which is associated with advanced stages of cancer and poor prognosis [73]. High SMYD3 expression is significantly associated with advanced tumor stage and is an independent prognostic factor of poor survival in patients with CRC [74]. SET and MYND domain containing 3 (SMYD3) is an oncogenic driver which controls Wnt induced activation of the ASCL2 gene, a master regulator of stem cell maintenance [75]. SMYD3 has also been shown to bind to SLUG gene promoter with ANKHD1 in a manner correlating with elevated levels of H3K4me3, H3K9ac and H3K14ac suggesting an epigenetic regulation of key EMT-TF associated with SMYD3 upregulation [75].

These findings emphasize the active role of epigenetic modifications to regulate EMT and cancer metastasis. The upregulation of various KDMs on TGFβ induction on the HCT116 p53 null CRC cell line opens up a clear possibility of epigenetic regulation of EMT transcription factors. Upregulation of KDM5A suggests an overall decrease in H3K4me3 upon TGFβ induction with p53 loss. KDM5A removes trimethyl (me3) marks from lysine 4 of histone 3 (H3K4) and serves as a general transcriptional corepressor. KDM5A and KDM5B perform context-dependent oncogenic or tumor-suppressive roles in the tumorigenesis of various cancers including breast cancer and prostate cancer. Furthermore KDM5A is known to have an important role in suppressing p53 function in neuroblastoma and other p53 wild-type-expressing tumors, suggesting that KDM5A is a regulator of the p53 signaling pathway in most cancers [76]. This correlates with our observation that KDM5A is upregulated with the loss of p53 in CRC cell lines on EMT induction. KDM5A expression associates with decreased E‑cadherin and increased N-cadherin expression in ovarian cancers [77] suggesting that it functions as a critical regulator of EMT-TF. Furthermore, significant positive correlation between KDM5B and ZEB family expressions in human lung cancer sample was observed [78]. Knockdown of KDM5B associates with increase CDH1 expression and decreased expression of ZEB1 and ZEB2 via microRNA-200 family. This leads us to hypothesize that the loss of p53 upregulated KDM5A which removes H3K4me3 marks from the colorectal cancer genome. The reduction in levels of H3K4me3 leads to reduced expression of epithelial marker E-Cadherin and increased expression of N—Cadherin, a typical signature of EMT, suggesting an epigenetic signaling cascade regulating the epithelial to mesenchymal transition. This further suggests that KDMs could be used as a potential target in metastatic patients with truncating p53 mutations.

We also observed an increase in the level of Chromobox (CBX) family proteins including CBX1 and CBX3 which are crucial components of heterochromatin protein 1γ (HP1) and Polycomb (Pc) complexes which are involved in transcriptional regulation, chromatin structural modification, and cell development process. Although interaction of CBX proteins with p53 has not been well studied, CBX3 has been shown to promote colon cancer progression via the CDK6/p21 pathway [79]. High level of CBX1/CBX3 family members associates with poor outcomes in CRC patients.

The elevated levels of SMARCA2 and HCFC2 and lower expression of PRDM4, PRMT1 and MAPKAPK3 observed on EMT induction in p53 null CRC cells correlate with low survivability of CRC patients (pvalue<0.05). PRMT1 triggers asymmetric di-methylation of histone H4 on arginine 3 (H4R3me2a) and its low expression is associated with poor prognosis in cancer patients [80]. Moreover, PRMT1 directly binds to p53 and inhibits the transcriptional activity of p53 in an enzymatically dependent manner, resulting in an alteration in the expression levels of several key downstream targets of the p53 such as CDKN1A and GADD45A [81]. Downregulation of PRMT1 correlates with upregulation of key p53 target genes GADD45A and CDKN1A as observed in our dataset suggesting it to be a potential biomarker and target for cancer therapeutics. Although the potential oncogenic role of SMARCA2 is not well studied in tumors [82], SMARCA2/Brm has been shown to be necessary for p53-mediated induction of p21 in Brg1-deficient cells, suggesting that Brm and Brg1 differentially regulate a subset of p53 target genes through chromatin remodeling [83].

The observations described above strongly validate an epigenetic signalling event associated with p53 which triggers the higher expression of EMT TFs in colorectal cancer cells with a p53 deletion.

EMT transcription factors show agonistic/antagonistic relationship with epigenetic factors

It is well established that several epigenetic factors play crucial roles in EMT regulation [84]. Therefore, it is crucial to comprehend if the EMT - epigenetic factor interaction is vital for EMT regulation. A group of EMT-inducing transcription factors (EMT-TFs) including Zinc-finger E-box-binding homeobox 1 (ZEB1) and ZEB2, Snail (SNAI1), Slug (SNAI2), and TWIST1/2, mediate EMT leading to dysfunction of cellular adhesion, loss of epithelial phenotype, metastasis, stemness and survival [3] but their interactions with the epigenetic machinery are less understood.

Towards this we investigated the agonist and antagonist relationship among EMT-TFs and between the EMT and epigenetic factors in CRC cells, pairwise Spearman correlation analysis was done on the normalised TPM (Transcripts Per Million) counts in all the four CRC cells upon EMT Induction.

A positive correlation was observed between the EMT- TF ZEB1 and SMAD proteins (SMAD2 and SMAD4) in HCT116p53 WT CRC cells (Fig. 5a). With the loss of p53, EMT-TFs SNAI1 and SNAI2 show positive correlation with SMAD proteins (SMAD2 and SMAD4). miR200 shows a positive correlation with SMAD7 (Fig. 5b). The p53 mutant cell line SW480 shows antagonistic relationship between SNAI1 and TWIST2, miR200 (Fig. Supplementary Fig S4A). In the p53 mutant cell line HT29, SNAI1 shows an antagonistic relationship with SMAD7 and CDH1 (Supplementary Fig. S4B). The pairwise correlations among the EMT and epigenetic factors in HCT116 and HCT116 p53 null cell lines are depicted in Fig. 5c and 5d while those of SW480 and HT29 cell lines are depicted in Supplementary Fig S4C and S4D. Since a significant number of such correlations were observed in the HCT116 p53 null cell line, relationships observed in this cell line were considered for further analysis on the epigenetic factor interaction with EMT machinery.

Fig. 5.

Correlation heatmaps representing agonistic and antagonistic relationship between various EMT factors and epigenetic factors in HCT116 WT and HCT116p53-/- CRC cells upon EMT induction (Pairwise Spearman correlation, p value< = 0.05,*; p value< = 0.01,**; p value< = 0.001,***) A. Agonistic correlation was observed between EMT-TF (ZEB1) and SMAD protein (SMAD2 and SMAD4) in HCT116WT CRC cell line B. Positive correlation was observed between EMT-TFs (SNAI1 and SNAI2) and SMAD proteins (SMAD2 and SMAD4) in HCT116 p53-/- cell line with the loss of p53 C. Positive correlation was observed between Lysine Demethylase (KDM6A) and EMT-TF (ZEB1) in HCT116WT CRC cells upon EMT Induction. D. Correlation of EMT-TF (SNAI1) and epigenetic factors (KDM6B,SATB1 and SETDB1) in HCT116 p53-/- CRC cells upon EMT Induction.

Since EMT-TF SNAI1 (Snail) shows both agonistic and antagonistic correlation with other epigenetic factors as compared to other EMT-TFs upon TGFβ treatment in p53 null CRC cells, we further investigated the epigenetic regulation of SNAI1 upon TGFβ treatment in HCT116p53null CRC cells.

SNAI1 and SETDB1 shows antagonistic correlation in HCT116 p53 null CRCs upon TGFβ treatment

SNAI1 shows antagonistic relationship with SETDB1 (R = 1, p< = 0.05) in HCT116 p53null CRC cell line upon TGFβ treatment with a strong significant negative correlation value of 1 (Fig. 5d). The histone methyltransferase SET domain bifurcated 1 (SETDB1, also known as ESET or KMT1E) is known to be involved in di/tri methylation of Histone 3 Lysine 9 and have a role in transcriptional repression. These are involved in establishing larger domain of repressive chromatin states. We observed that SETDB1 was downregulated upon TGFβ treatment in HCT116 p53 null CRC cells. SETDB1 epigenetically suppresses the expression of SNAI1 through H3K9 methylation at their promoter region of SNAI. Downregulation of SETDB1 leads to overexpression of SNAIL1 leading to EMT. Although previous studies have reported the epigenetic regulation of SNAIL1 through SETDB1, loss of p53 contributing to this regulation have not been explored yet. We observed that SETDB1 (log2FC=−0.85, p < 0.05) was uniquely downregulated upon TGFβ induction in HCT116 p53null cell lines while SNAIL1 is significantly upregulated (log2FC=1.4, p < 0.05) suggesting the role of p53 in establishing this antagonistic correlation of SETDB1 and SNAI1.

SNAI1 and SATB1 shows agonistic correlation in HCT116 p53 null CRCs upon TGFβ treatment

We identified a positive correlation between Special AT-rich binding protein 1 (SATB1) and SNAIL1 (R = 0.6, p < =0.05) in HCT116 p53 null cells CRC upon TGFβ treatment (Fig 5d). SATB1 is a nuclear matrix protein which mediates chromatin looping and has been shown to influence EMT process in breast, colorectal and prostate cancers [71].

In vivo experiments using mice models have shown that SATB1 induced EMT correlated with increased expression of Snail1, Slug, Twist and vimentin and decreased expression of E-cadherin, tight junction protein ZO-1 and desmoplakin [71]. Furthermore, SATB1 expression was shown to be positively correlated with the expression of Snail, Slug and Twist1 in non‑small cell lung cancer (NSCLC) clinical samples [71]. These findings correlate with our observation where SATB1 and SNAIL1 show positive correlation in p53 loss CRC on EMT induction. The expression levels of SNAI2 (Slug) and Vimentin have also been found to increase with SATB1 regulation. These findings suggest chromatin mediated regulation of EMT with the loss of p53 in CRC with the help of chromatin organiser.

The chromatin organizer SATB1 shows lower expression in p53 null CRC cells with higher enrichment of H3K27me3 in its promoter region as observed in our ChIP sequencing data (unpublished results). When treated with TGFβ, the expression of SATB1 is upregulated (transcriptomic experiments). EZH2 plays an important role in epigenetically regulating the expression of SATB1 [72], we see an agonistic relationship of EZH2 and SATB1 with loss of p53. We hypothesize that the treatment with TGFβ may help remove EZH2 mediated repression of H3K27me3 via KDM6A from the promoters of SATB1 contributing to the elevated expression.

SNAI1 and KDM6B show agonistic correlation in HCT116p53null CRC cell line upon TGFβ treatment

In CRC cell lines with loss of p53, Lysine demethylase KDM6B was found to agonistically regulate SNAI1 (R = 1; p < 0.05). KDM6B is a histone demethylase that activates gene expression through removal of repressive histone mark H3K27me3 from the chromatin [85]. p53 is also known to interact with KDM6B and to modulate the expression of H3K27me3 in cancer cells [85]. Overexpression of KDM6B has been shown to induce the expression of mesenchymal genes and promotes EMT in breast cancer. Chromatin immunoprecipitation (ChIP) assays revealed that KDM6B promoted SNAI1 expression by removing histone H3 lysine trimethylation(H3K27me3) repressive marks from the promoter region of SNAI1 (Snail) [86]. In this case, we hypothesize that the loss of p53 in HCT116p53 null CRC cells brings down the binding of KDM6B around p53 target genes thereby increasing the expression of SNAI1.

Differential regulation of EMT-TFs and epigenetic enzymes upon EMT induction in p53 deficient CRC cells correlates with expression and survival outcomes in CRC patients (TCGA-COAD)

To explore the expression profiles of the identified epigenetic signatures and EMT-TFs, SATB1, SNAI1, KDM6B and EZH2 in colorectal tumors we performed qRT-PCR from the RNA isolated from the tumors and their adjacent normals. A total of three colorectal tumors with three adjacent normals were collected. The transcriptional expression of SATB1 and SNAI1 was higher in tumor tissues than that in the adjacent normal tissues (Fig 6A,Supplementary Figure Fig S5A-B). This finding correlates with the increased expression of SATB1 and SNAI1 in CRC cell line with varying p53 status upon TGFβ induction as shown in supplementary figure FigS5C-E with highest expression of expression in p53 null CRC cells, suggesting p53 mediated regulation of SATB1 and SNAI1 upon EMT induction. SATB1 expression has been shown to be positively correlated with vimentin in colorectal tumors [87] and its overexpression is associated with poor prognosis in colon cancer [88] suggesting that SATB1 plays an important role in EMT. This increased expression of various EMT-TF correlates with increased expression of epigenetic factors including KDM6B.

Fig. 6.

Validation of the observed epigenetic-EMT cross talk signatures in colorectal cancer patients. A. Alteration in the expression profile of SATB1, epigenetic regulators and EMT-TFs in CRC patients. qRT-PCR analysis was performed to evaluate the expression profile of SATB1, KDM6B, EZH2 and SNAI1. Graph showing increase in relative expression of SATB1 and SNAI1 in CRC patient tumor tissue as compared to normal. RT-qPCR data analysis was based on 2-∆∆CT and GAPDH was used as housekeeping gene for RT-qPCR data normalization. Results show mean ± SD B. Volcano plot representing differentially expressed genes in Low SATB(N = 261) vs High SATB1(N = 218) with logFC|0.5| and padj<=0.05 C. High SATB1(N = 218) TCGA-COAD cohort is associated with significantly higher expression profile of EMT-TFs ZEB1 and SNAI1 and epigenetic signatures KDM6A and KDM6B(pvalue>0.05,ns; p value< = 0.05,*; p value< = 0.01,**; p value< = 0.001,***) D. High median expression value of SNAI1 associates with poor survival in TCGA-COAD (n = 270) patients [HR =1.5 and p value < =0.1] E.High median expression values of both SNAI1/SATB1 associates with poor survival in TCGA-COAD (n = 270) patients [HR = 1.5 and pvalue < = 0.1] F.High median expression values of both SNAI1/KDM6B associate with poor survival in TCGA-COAD (n = 270) patients [HR =1.5 and p value < = 0.1].

To further investigate the impact of SATB1 on colon cancer and its crosstalk with various epigenetic factors, the expression pattern of SATB1 was examined in TCGA-COAD cohort. Based on the median expression of SATB1(1.56), the TCGA-COAD cohort was divided into high-(N = 218) and low-SATB1 (N = 261) expression groups (Fig 6b). We observed that higher expression of EMT-TF ZEB1, VIM, SNAI1 and epigenetic factor KDM6A, KDM6B and EZH2 were associated with high SATB1 group as compared to low SATB1 group (Fig 6C), suggesting that SATB1 promotes oncogenesis in colon cancer through a cross talk mechanism via various epigenetic factors and EMT-TF. We then wanted to understand the expression pattern of these differentially regulated EMT-TF and epigenetic factors in p53R273H mutant(N = 12) and p53 null(N = 27) colon cancer patient cohort through TCGA-COAD (Fig S6A). Toward this we found that SATB1 and SNAI1 showed higher expression in p53 null mutation patients as against p53WT(N = 12) and p53R273H mutant colon cancer cohort. This suggests that the loss of p53 and not R273H mutant p53 in colon cancer patients increases the expression of SATB1 and thus drives the cancer towards metastasis. The increased expression of SATB1 correlates with expression profiles of various epigenetic factors including KDM6A and KDM6B and EMT-TFs SNAI1 and ZEB1(Supplementary FigS6A) in p53 null mutation cohort suggesting a crosstalk between various EMT-TFs, SATB1 and epigenetic factors which drives the cancer towards metastasis associated. Next, in order to understand clinical significance/prognosis of the epigenetic and EMT factors differentially regulated in TGFβ induced p53 null CRC cells and through TCGA-COAD cohort, we analysed the survival outcomes of 270 colorectal cancer patients (TCGA-COAD) from GEPIA2. High and low expression groups were defined in the cohort using the median value for each gene. It was interesting to observe that patients with tumors exhibiting high levels of expression of SNAI1 and SMARCA2 (upregulated in p53 null CRC cells) showed worse prognosis (Fig 6d and Supplementary FigS6B). Patients who had tumors with low expression of PRDM4 (Supplementary Fig.S6C) showed worse prognosis. It was observed that the key EMT factor SNAI1 has a significant prognostic value and its upregulation is related to a significant (p< = 0.05) decrease in survival of CRC patients. Since the SNAI1 higher expression and its relation to low survival is well established, we wanted to explore if EMT regulation by SNAI1 is facilitated by an epigenetic event. We therefore wanted to investigate the clinical implication of the agnostic or antagonistic correlation between epigenetic factors and EMT–TFs identified from our transcriptomic studies.

We found that highly expressing SNAIL showed a positive correlation with KDM6B (the lysine demethylase for H327me3) upon TGFβ treatment in p53null CRCs and that a combination of high expression of SNAI1/SATB1 and SNAI/KDM6B was associated with poor clinical outcome in CRC patients (Fig 6e and 6f) This suggests that SATB1 may be a potential target for CRC patients. These data suggest that chromatin modifiers play a key role in regulation of expression profiles of EMT-TF thus contributing towards metastatic states in cancer. This work provides new insights into epigenetic therapies in clinical oncology, initiated by a crosstalk between the EMT and epigenetics machinery leading to cancer progression. This opens a new frontier of devising synergistic therapies using epigenetic drugs for these targets in metastatic CRC patients with p53 loss.

Discussions

Epithelial to Mesenchymal transition is a dynamic regulatory process in which epithelial cells lose their cell-cell junctions and gain mesenchymal characteristics acquiring more invasiveness. The metastatic cascade involves several signalling pathways and somatic mutations in the tumour suppressors and oncogenes, all of which promote EMT. Although the mutational events leading to metastasis have been studied, the epigenetic regulatory mechanisms involved in EMT have not been well established [89], [90], [91].

Since p53 driver mutation is one of the late events in colorectal cancer progression, (driving adenoma to adenocarcinoma), we investigated if epigenetic players function in concert with other genetic events and p53 signalling to facilitate the transformation in colorectal cancer. Towards this, we employed an EMT induction model with TGFβ on 4 CRC cell lines with varied status of p53 (HCT116 p53WT, HCT116 p53-/-, HT29 p53R273H, SW480 p53R273H, P309S) in this work. We have attempted to explore how epigenetic signalling regulates EMT in p53 null or mutant CRC cell lines and to validate the observed epigenetic signatures in patient tumours.

Our transcriptomic analysis on the 4 CRC cell lines harboring p53 WT, p53 null and p53 mutation upon TGFβ induction have identified that the extracellular matrix and the cytoskeleton play active roles in the transition from the epithelial to mesenchymal state. Upregulation of integrin alpha isoforms ITGA1 and ITGA9, matrix metalloproteases (MMP1 and MMP28, ADAMTS4 and ADAMTSL4), collagen alpha genes were observed only in p53 null CRC cells on EMT induction. The p53 null CRC cell line exhibits the most significant alterations as demonstrated by its invasion kinetics and expression changes favouring a mesenchymal phenotype. It is well known that miR-34a/b/c downregulates Snail expression through a p53 mediated signaling pathway and subsequently protects cells from epithelial to mesenchymal transition. The activation of the Wnt-β catenin pathway in the wild type p53 CRC cells (HCT116) used in the study leads us to hypothesise that the TGFβ mediated EMT observed in p53 null cells may be due to activation of Snail with the loss of p53.

It is well established that ECM proteins produced by the tumour and stromal cells undergo alterations in integrin signalling to enable invasion and metastasis. The integrins ITGA1, ITGA9 have been correlated with poor prognosis in several cancers. Serum concentration of ITGA1 was found to be higher in 50 CRC patient tissues and was associated with metastatic TNM stages of colorectal cancer. It was also observed that ITGA1 promotes the invasion and migration of CRC cells through the activation of Ras/Erk signalling pathways [92]. ITGA9 also interacts extensively with the proteins of the ADAM family and ITGA9 inhibition stimulates a vital interaction domain of ADAM proteins and controls metastasis [93].

TGFβ controls the expression of MMPs through Smads and several other response elements in MMP promoters. Wild type p53 inhibits p300-mediated induction of the MMP1 promoter via Activator protein 1 (AP1). Loss/mutation on p53 stimulates p300 mediated histone acetylation at promoters of MMP1 leading to increase in expression of MMP1 [94]. Since TGFβ is known to enhance DNA binding activities of AP1 we hypothesise that the loss of p53 upon TGFβ signalling increases the binding of AP1 around the promoter regions of MMP1 resulting in enhanced expression of the MMPs and other collagen genes.

We also found that the expression of kinases like CASK (Calcium/calmodulin-dependent serine protein kinase), DAPK1(Death Associated Protein Kinase 1), MYLK (Myosin Light Chain Kinase), NEK2 (NIMA-related Kinase 2) were upregulated with EMT induction on HCT116 p53 null cells. These kinases play important roles in tumour cell proliferation, migration, invasion and promotes metastasis via EMT signalling. Upregulation of DAPK1 promotes the stemness of cancer cells and EMT process by activating ZEB1 in CRC. This establishes that the loss of p53 activates EMT through signalling kinases by promoting cancer stemness.

Chromatin regulation has a pivotal role in establishing and maintaining cell state identity through transcriptional regulation. Recent work from Weinberg's group using CRISPR screen on HMLER cells has identified chromatin-regulatory complexes PRC2 and KMT2D-COMPASS as key regulators of epithelial to mesenchymal plasticity (EMP) [95]. Large-scale CRISPR interference (CRISPRi) screen for genes involved in RTK signaling and epigenome regulation have identified a large number of chromatin regulators (SWI/SNF chromatin remodeling complex (ARID1A, SMARCB1, and SMARCA1) writer complexes such as KMT2A, DOT1L, and EPC1 and reader complexes such as BRD2 and ZMYND8, whose loss affects epithelial to mesenchymal transition [96].

In our experiments, the loss of p53 in CRC cells shows a highly mesenchymal and invasive phenotype upon TGFβ treatment with differential regulation of significant number of chromatin regulators. Our transcriptomic profiles observed that upregulation of eraser machinery such as KDMs and HDACs correspond to the down regulation of various writers including SUV39H1, SETDB1 pointing to key epigenetic alterations upon p53 loss.

Our analyses on the correlation of the expression of epigenetic factors with relevant EMT transcription factors establish a significantly strong agonistic relationship between (SNAI1 and KDM6B) and (SATB1 and KDM6B/KDM6A) and an antagonistic relation between SNAI1 and SETDB1. We have proposed a model based on the signaling events observed in our experiments and the expression profiles obtained through transcriptome sequencing. The presence of p53 in p53 wild type CRC cells prevents the deposition of H3K27me3 marks on the promoter of SATB1. When p53 is lost, the promoter of SATB1 is occupied by repressive histone marks H3K27me3 and its methyl transferase EZH2, which aid the downregulation of SATB1 as shown in Figure 7. This clearly demonstrates a p53 mediated rearrangement of the chromatin in CRC cells.

EMT induction through TGFβ in HCT116 p53 null CRC cells removes H3K27me3 and its methylase EZH2 from the promoter region of SATB1 through KDM6A. The upregulation of SATB1 corresponding to the upregulation of KDM6A and the removal of repressive facultative heterochromatin mark H3K27me3 from the promoter region of SATB1 upon EMT induction implies a reorganization of the higher chromatin structure and the 3D genome architecture. The higher expression of Snail correlates with higher expression of SATB1, KDM6A, KDM6B and SETDB1 confirming alterations in the epigenomic landscape around key mesenchymal genes on p53 loss upon EMT induction. Unpublished work from our group has found that the promoters of SATB1 accommodate repressive H3K27me3 marks which are significantly downregulated upon p53 loss. In accordance with this, we observe that the expression of SATB1 increases with KDM6A upon TGFβ treatment suggesting that KDM6A might play an important role in demethylating the promoters to activate the expression of SATB1 upon p53 loss. High levels of KDM6B are also known induce the expression of Snail and Slug, promoting TGFβ-induced EMT [32]. These observations point to the pivotal role of SNAIL as a major player in the aggressive phenotype induced by p53 loss in colorectal cancer cells.

Our survival analysis on TCGA - COAD cohort using a combination of differentially regulated epigenetic factors and EMT drivers (SNAIL1/SATB1, SNAI2/KDM3A) shows a strong correlation of the high expression profiles with poor survival outcomes in the patients. The epigenetic and EMT factors identified from the sequencing data have also been validated in colorectal patient tumours. These CRC patients show upregulation of the chromatin modulator SATBI and EMT Factor SNAI1 as observed in colorectal cancer cells and in the TCGA cohort expressing high levels of SATB1. This clearly establishes a SATB1 mediated epigenetic control of EMT factors during metastasis.

The number of patients considered for validation of the observed epigenetic EMT cross talk is less and statistically limits the further interesting findings from patients. We have now expanded the recruitment of patients with specific status of p53 for exome and RNA sequencing.

In summary, we report an interesting aggressive phenotype of EMT regulation in p53 null CRC cells driven by cytoskeletal, ECM and epigenetic markers in concert with EMT transcription factors and a SATB1 mediated epigenetic control of EMT phenotype. The transcriptomic profile of TGFβ mediated EMT induction in colorectal cancer has been reported for the first time in this work and it opens up interesting possibilities of exploring signaling cross talk between EMT TFs, non coding RNAs and the epigenetic machinery in colorectal cancer cells and in patient tumours.

The modulation of chromatin organizer SATBI by lysine demethylases during the loss of p53 implies alterations in chromatin structure and 3D organization as a critical event in EMT. This novel hitherto unexplored existence of agonistic and antagonistic relations between the epigenetic enzymes and Snail (SNAI1) reported for the first time and the alterations in 3D genome architecture open up newer possibilities of developing synergistic therapies with an epigenetic inhibitor and an EMT inhibitor for CRC patients with loss of p53.

Approval from institutional ethics committee (IEC)

The approval from institutional ethic committee (IEC) for using tumor samples of CRC patients from HCG hospitals Bangalore was obtained (IBABIEC-04/PR01/1,281,219)

CRediT authorship contribution statement

Shreya Sharma: Data curation, Formal analysis, Investigation, Writing – original draft. Harsha Rani: Data curation, Formal analysis, Investigation, Writing – original draft. Yeshwanth Mahesh: Formal analysis. Mohit Kumar Jolly: Investigation, Methodology, Supervision. Jagannath Dixit: Resources. Vijayalakshmi Mahadevan: Data curation, Conceptualization, Funding acquisition, Investigation, Methodology, Supervision, Writing – original draft.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.