Introduction

The substantially proliferative, multifocal, genomically divergent mutations harboring, and metastatic nature of muscle invasive bladder cancer (MIBC) underlies the poor prognosis of the disease [1]. Owing to the recurrence and frequent resistance to traditional therapies, developing new therapeutic options for bladder cancer patients is required.

Advances in tumor genomic analysis have unveiled dominant mutational signatures in bladder cancers. Bladder cancer has a highly enriched mutational load, with mutations particularly in chromatin regulatory genes encoding proteins such as histone demethylases (KDM6A), histone methyltransferases (KMT2C and KMT2D), histone acetyltransferases (CREBBP and EP300), and the SWI/SNF chromatin remodeling complex (ARID1A, SMARCA4) [2,3,4,5]. Remarkably, almost all these chromatin modifiers are involved in active chromatin organization, opposing Polycomb-mediated repression. The excessive mutational load of chromatin regulatory genes in bladder carcinoma is a hallmark of this disease and a potential weakness that epigenetic drugs could target.

Polycomb group (PcG) proteins function as epigenetic regulators that procure the pursuance of cell-specific transcriptional programs by playing a major role in the implementation of cellular identity and cell fate decisions [6]. Polycomb repressive complexes (PRCs) are composed of major chromatin modifiers that negatively regulate gene transcription by modulating the global epigenetic state [7]. The enhancer of zeste homolog 2 (EZH2) methyltransferase is the catalytic subunit of Polycomb repressive complex 2 (PRC2), functioning as a transcriptional repressor through trimethylation of lysine 27 of histone 3 (H3K27me3) [8]. This epigenetic alteration causes chromatin to be densely packed, rendering it hardly accessible to the transcriptional machinery, which consequently results in the silencing of genes [9]. It is established that EZH2 alters the tumor metastatic landscape, affects the tumor microenvironment, and influences cell fate decisions, propounding the highly context-dependent action of EZH2. Notably, EZH2 has been considered an oncogenic factor in a variety of cancers [10,11,12]. Converging lines of investigation have highlighted the role of EZH2 in regulating cell plasticity and encouraging intratumoral heterogeneity, driving the progression of cancer by promoting invasion and metastasis, and ultimately resulting in poor clinical outcomes [13, 14]. Orchestrating with other epigenetic modifiers, EZH2 represses genes linked to differentiation and cell cycle arrest, favoring stemness retention [12, 15, 16]. As observed by previous studies, increased expression of EZH2 contributes to the metastatic cascade of bladder cancers, and selective loss of H3K27me3 function of EZH2 dominates a notable hampering impact on cancer cell migration and tumor metastasis [17,18,19]. Therefore, given the proof of EZH2 gain of function becoming a cancer driver, therapeutic strategies have been developed to target EZH2 using specialized chemical inhibitors [20]. GSK-126, a potent EZH2 inhibitor, has been shown to originate loss of genome-wide H3K27 methylation, and reactivation of PRC2-repressed genes [12]. It has been demonstrated that knockdown of EZH2 decreases tumor growth and cell proliferation in a variety of cancers [21,22,23]. In addition, inhibiting EZH2 activity causes cancer cells to undergo selective apoptosis, but not normal cells, thereby portraying it as a promising anti-cancer therapeutic target [24].

The application of epigenetic drugs in a combined fashion with other drugs represents a principally rational strategy to sensitize cancer cells to the treatment and overcome acquired resistance mechanisms [25].

Retinoic acids (RAs) are signaling molecules that constitute genetic communication networks, having essential functions in embryonic development, organogenesis, organ homeostasis, as well as cell proliferation, differentiation, and death [26]. Through their driver roles in apoptosis and differentiation, retinoic acid derivatives—retinoids—have gained prominence in research for the development of innovative cancer therapeutics [27,28,29].

Retinoic acid receptor and retinoid X receptor (RAR/RXR) complexes, functioning as transcription factors, regulate transcriptional activation of RA target genes via the recruitment of certain co-regulators to the RA response elements [30]. In the absence of ligands for RAR-RXR dimers, recruitment of histone deacetylases (HDACs) leads to chromatin condensation and eventually repression of target genes. Conversely, upon RAR agonist binding, recruited histone acetyltransferase (HAT) complexes bring along derepression of the target genes; in that status, the PRC2 binding is reduced, thus the H3K27me3 is diminished at these regions, and eventually gene expression is activated, emphasizing PRC2’s potential crucial co-repressor role in RA-mediated gene transcription [31]. Therefore, Polycomb-group proteins function downstream of the RA signaling pathway; inhibiting PRC2 activity could potentially enhance RA action and overcome resistance to RA [32].

Considering the molecular and cellular complexity of bladder cancer, which seems unlikely to respond to single molecular pathway targeting, we envision that simultaneous targeting of the retinoic acid pathway and EZH2 could be a promising approach in bladder cancer therapy. In this study, we aim to elucidate the alterations in the phenotypic, transcriptomic, and epigenetic landscape of bladder cancer cells and to portray the molecular mechanisms linked to the co-treatment with a retinoid acid analog, fenretinide, and a potent EZH2 inhibitor, GSK-126. We report that combined treatment induces a synergistic reduction in proliferative potential correlated with increased apoptosis and modulation of a set of genes related to apoptosis and cell cycle in MIBC cells. We demonstrated that the endoplasmic reticulum stress response underlies the interactions of our coordinated targeting strategy.

Methods

Cell culture

5637 (DSMZ ACC 35) and HT1376 (DSMZ ACC 397) human bladder cancer cell lines were purchased from DSMZ. J82 and T24 human bladder cancer cell lines were purchased from ATCC. HT1376, J82, and T24 cells were cultured in DMEM (41965039, Thermo Fisher Scientific) medium, and 5637 cells were cultured in RPMI 1640 (21875, Thermo Fisher Scientific) medium supplemented with 10% fetal bovine serum (10500064, Thermo Fisher Scientific) and 1% penicillin/streptomycin (15140122, Thermo Fisher Scientific). The cells were cultured at 37 °C in cell culture dishes in a humidified 5% CO2/air atmosphere. Serial passages were carried out through treatment of sub-confluent monolayers with TrypLE Express Enzyme (12604013, Thermo Fisher Scientific). All cell lines were confirmed to be mycoplasma-free.

Drugs and reagents

GSK-126 (S7061, Selleck Chemicals, USA) and fenretinide (390900, Merck, Germany) were dissolved in dimethyl sulfoxide (DMSO, 276855, Sigma Aldrich) at a concentration of 10 mM and further diluted to different working concentrations.

Cell viability assays

Prior to cell viability assays, cell number titration is performed to determine the optimal cell seeding density. For viability assays, the cells were seeded on 384-well plates in 50 μL of culture medium in triplicates according to the optimal seeding density calculations. After 24 h of incubation, the cells were treated with the indicated serial dilution concentrations of drugs. DMSO and PAO were used for vehicle control and positive control, respectively. After the incubation period, 10 μl of 0.05 mg/mL Resazurin (R7017, Sigma-Aldrich) was added to each well. After 4 h of incubation at 37 °C, the plate was read on an Infinite 200 Pro microplate reader (TECAN) at an excitation wavelength of 570 nm and an emission wavelength of 600 nm. The values were normalized to vehicle control to calculate percent survival. The viability analyses were performed using the GraphPad Prism software.

Drug combination analysis

The cells were seeded on 384-well plates in 50 μL of culture medium in triplicate. After 24 h of incubation, the cells were treated with the indicated serial dilution concentrations of drugs, alone or in combination. Drug treatments were performed using the SPS High Performance D300 Digital Dispenser by HP, Inc. Specialty Printing Systems (1070 NE Circle Blvd., Corvallis, OR 97330). After 72 h incubation, 10 μL 0.05 mg/ml of resazurin (R7017, Sigma-Aldrich) was added to each well and incubated at 37 °C for 4 h, and plates were read using an Infinite 200 Pro microplate reader (TECAN). The synergy score was calculated using the SynergyFinder v2.0 web-based application with default parameters for calculating Bliss Independence.

Apoptosis assay

The cells were seeded into 60-mm dishes at the optimal density. After an overnight incubation, the cells were treated with single drugs or drug combination. The apoptosis assay was performed with the FITC Annexin V/Dead Cell Apoptosis Kit (V13242, Invitrogen). Briefly, after the incubation period, the cells were harvested, washed in cold phosphate-buffered saline (PBS), and then resuspended in 100 μL of 1X annexin-binding buffer. After adding 5 μL of Annexin FITC and 1 μL of PI (100 μg/mL) to each cell suspension, the cells were incubated at room temperature for 15 min. The stained cells were analyzed by flow cytometry, measuring the fluorescence emission at 530 nm and > 575 nm. Flow cytometer data were analyzed with the FlowJo software.

Cell cycle analysis

The cells were seeded into 60-mm dishes at the optimal density. After an overnight incubation, the cells were treated with drugs. The cells were harvested and washed in PBS (1X). The cells were fixed with 70% cold ethanol by adding dropwise to each sample and then incubated for at least 30 min at 4 °C. Fixed cells are washed with PBS and resuspended in 50 µl of RNase A (100 µg/ml), 200 µl of propidium iodide (50 µg/ml), and 0.1% Triton X-100 for staining. The cells were protected from light. The data was acquired on the flow cytometer. Cell cycle phases are analyzed using FlowJo software.

Scratch assay

The cells were cultured on 6-well plates and incubated overnight to generate a confluent monolayer. The cell monolayer was scratched with a P200 pipette tip. After washing with PBS, the cells were cultured in low-serum-containing media with or without treatments for different time intervals. Five reference points were randomly selected from a single well, and the percentage of scratch wound closure was analyzed by ImageJ software. Three independent experiments were performed for data representation.

RNA isolation, cDNA synthesis, and RT-qPCR

Total RNA was extracted using the Monarch Total RNA Miniprep Kit (T2010S, New England Biolabs) according to the manufacturer’s instructions. The RNA concentration was quantified with a NanoDrop 2000 spectrophotometer. 1 μg of RNA was transcribed using the Maxima First Strand cDNA Synthesis Kit for RT-qPCR (K1641, Thermo Scientific). Quantitative real-time PCR (RT-qPCR) was performed with the FastStart Essential DNA Green Master Kit (06402712001, Roche). Relative gene expression was assessed in three technical replicates of the samples. The results were normalized to the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) by using the comparative cycles of threshold (Ct) method (ΔΔCt) [33]. RT-qPCR assays were run on a 7500 fast real-time PCR system (Applied Biosystems). The primer sequences used are listed in Supplementary Table 1.

RNA-seq and RNA-seq data analysis

Total RNA was isolated using the Monarch Total RNA Miniprep Kit (T2010S, New England Biolabs) according to the manufacturer’s recommendations. The integrity of total RNA was checked using a 2100 Bioanalyzer. Paired-end libraries were prepared using the TruSeq Stranded Total RNA with Ribo-Zero H/M/R Sample Library Preparation Kit and sequenced on the Illumina Novaseq 6000.

The quality control of raw sequencing data was performed with FASTQC (v0.11.9; https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Adapter removal and quality trimming steps of sequencing reads were done with Trimmomatic (v0.36) [34]. The human reference genome (GRCh38) in FASTA format and related gene annotation in General Transfer Format (GTF) were manually downloaded from the GENCODE (Release 32) project web site (https://www.gencodegenes.org). RNA-seq data of each sample were aligned to the human reference genome with Rusbread v2.0.3package [35] of the R v4.1.2 statistical computing environment (https://www.r-project.org/) with the following settings: “align(index = {index file}, readfile1 = {input_1.fastq}, readfile2 = {input 2.fastq} type = “rna”, input_format = “gzFASTQ”, output_format = “BAM”, output_file = {output file}, nthreads=numParallelJobs)”. We utilized SAMtools v1.3.1 [36] to sort and index all BAM files created in the alignment step. In order to measure the expression levels of genes, we employed the featureCounts function [37] of the Rsubread package with the following command: “featureCounts (files = {infile.bam}, annot.ext = “{infile.gtf}”, isGTFAnnotationFile = T, GTF.featureType = “exon”, GTF.attrType = “gene_id”, useMetaFeatures = T, countMultiMappingReads = T, isPairedEnd = T, nthreads = numParallelJobs)” and FPKM values of each gene across samples were calculated with “fpkm” function of edgeR package (v3.24.3) [38]. All genes exhibiting a log2 fold change > 0.5 between groups were considered differentially expressed. The R statistical computation environment was also utilized for graphical representations of the study. The R-package clusterProfiler (version 3.18.0) [39] was used to identify and visualize the enrichment of Gene Ontology (GO) terms in the sets of genes of interest. Graphics associated with RNA-seq data were generated with the R-package ggplot2 (https://ggplot2.tidyverse.org/).

Western blot analysis

Cellular proteins were extracted using complete RIPA lysis buffer (0.15 M NaCl, 1% NP40, 0.5% DOC, 0.1% SDS, 50 mM Tris (pH 7.5), and 2 mM EDTA) with complete protease inhibitor cocktail (PIC). Histone proteins were extracted using the histone acid extraction method [40]. Briefly, the cell pellet was suspended in a hypotonic lysis buffer containing 10 mM Tris–Cl pH 8.0, 1 mM KCl, 1.5 mM MgCl2 and, 1 mM DTT with PIC. After incubating cells for 30 min at 4 °C, the intact nuclei were pelleted by spinning (10,000 g for 10 min at 4 °C) and suspended in 0.2 M H2SO4. After overnight incubation on the rotator, the supernatant containing histones was precipitated with trichloroacetic acid (TCA). After incubation on ice for 30 min, the histone pellet was washed twice with ice-cold acetone to remove acid from the solution. After drying, the histone pellet was dissolved in ddH2O. Protein lysates were cleared by centrifugation at maximum speed at 4 °C for 15 min. Protein concentrations of lysates were measured using the BCA assay (23227, Pierce, Thermo Fisher Scientific).

Western blot analysis was performed in accordance with standard protocols. Protein extracts were denatured with heat and 1x Laemmli sample buffer, separated on Acrylamide/Bis gel, and transferred to nitrocellulose membranes. The membranes were blocked in 5% nonfat dry milk in TBST for 1 h, and then membranes were incubated overnight at 4 °C with the following primary antibodies: EZH2 (5246, Cell Signaling), H3K27me3 (9733, Cell Signaling Technology), β-Actin mouse (3700, Cell Signaling), β-Actin rabbit (4970, Cell Signaling), and Histone H3 (39763, Active Motif). After washing with TBST, membranes were incubated with the following secondary antibodies: Anti-rabbit IgG (H + L) (5151, CST) and Anti-mouse IgG (H + L) (DyLight™ 680 Conjugate) (5470, Cell Signaling Technology). Protein bands were visualized using the Odyssey CLX imaging system (LICOR Biosciences).

Chromatin immunoprecipitation

Cells at 80% confluency were fixed with formaldehyde (1% final concentration) for 10 min, and crosslinking was stopped with 0.125 M glycine. To help to produce chromatin lysates, the cells were incubated with two lysis buffers containing a complete protease inhibitor cocktail (Roche, 11697498001) (Paro Rinse 1: 10 mM Tris pH 8.0, 10 mM EDTA pH 8.0, 0.5 mM EGTA, 0.25% Triton X-100 Paro Rinse 2: 10 mM Tris pH 8.0, 1 mM EDTA, 0.5 mM EGTA, and 200 mM NaCl). Then, the cells were lysed in sodium dodecyl sulfate (SDS) lysis buffer (0.1% SDS, 0.1% DOC, 1% Triton X, 1 mM EDTA, 500 mM NaCl, and 50 mM HEPES/KOH) containing a complete protease inhibitor cocktail (Roche, 11697498001) and homogenized, followed by incubation on ice for 15 min. The cell lysates were sonicated with Covaris sonicator (S220 High Performance Ultrasonicator) (15 cycles of 140 W Peak Power, 15 Duty Factor, 200 cycles/burst for 60 sec) to generate 200–500 bp long chromatin fragments. Following sonication, the lysates were centrifuged at 16,000 × g for 10 min at 4 °C to pellet the debris. In total, 50 μl of the sonicated lysate was stored as input control. RNase A (0.2 mg/ml) was added to clear lysates, and after Proteinase K (200 μg/ml) treatment, reversal of crosslink was performed with 1% SDS and 100 mM NaCl. Cross-linking was reversed by heating at 55 °C for 2 h and 65 °C for 16–18 h. After the sonication quality was confirmed on agarose gel, the DNA product was eluted using the Zymo DNA Clean and Concentrator Kit (D4034, Zymo Research) and stored as input DNA. Magnetic beads (11204D, Dynabeads M-280 sheep anti-rabbit IgG, Invitrogen, Thermo Fisher Scientific) were pre-blocked with BSA (1 mg/ml) and tRNA (1 mg/ml). Chromatin sample was pre-cleared with tRNA pre-blocked magnetic beads. The H3K27me3 antibody (9733, Cell Signaling Technology) was used for immunoprecipitation of chromatin and incubated overnight at 4 °C with overhead shaking. After adding corresponding pre-blocked magnetic beads to the AB-chromatin complex, the complex incubated for 3 h at 4 °C with overhead shaking. Followed by washing with lysis and high salt buffers (0.5% DOC, 0.5% NP-40, 0.25 M LiCl, 1 mM EDTA, 10 mM Tris pH 8.0), the immunoprecipitated DNA was eluted off the magnetic beads with elution buffer (1 M NaHCO3, 1% SDS) and reverse cross-linked by heating at 55 °C for 2 h and 65 °C for 16-18 h. The immunoprecipitated DNA was eluted using a ChIP DNA Clean and Concentrator Kit (D5205, Zymo Research). H3K27me3 content at the promoters of specified genes were assessed by RT-qPCR using primers designed from the Cistrome Data Browser resource [41]. ChIP primer sequences for different genes are enlisted in Supplementary Table 2. The enrichment of immunoprecipitated DNA with anti-H3K27me3 antibody was calculated using the following method: The average Ct of triplicates was calculated. Adjusted Ct value for input sample was calculated as (Ct10% Input-log (10,2)). ΔCt value of input sample was calculated (CtAdjusted Input -CtChIP Input), the percent input was calculated as (100∗2^(ΔCt)).

ChIP-seq analysis

A total of 50 ng of DNA was provided per ChIP sample to Macrogen for DNA library preparation using the TruSeq DNA sample preparation kit according to Illumina recommendations. The ChIP DNA libraries were sequenced as paired-end 150-bp reads using the NovaSeq 6000 S4 sequencer (Illumina).

ChIP-seq analysis was performed using the Nextflow (version 22.10.5) nf-core/chipseq (v2.0.0) pipeline (https://doi.org/10.5281/zenodo.3240506). This pipeline uses FastQC to read raw reads, Trim Galore to trim adapters, and Picard, SAMtools, BEDtools, and BAMtools to filter and prepare data for alignment. Then broad peaks were called by using MACS2 [42]. The featureCounts function was used to obtain the count values of consensus peaks [37]. The nf-core chip-seq pipeline was run using the following parameters: genome = GRCh38, macs_gsize = 2913022398, and p-value cutoff = 0.00001. Counts and annotation files, which were created by nf-core/chipseq, were used to obtain the bed files and for the visualization of ChIP-Seq results. The compute matrix was created by using the bigWig files for the desired gene lists in the created bedfiles with DeepTools’ computematrix function. The Bioconductor Gviz package was used to visualize ChIP-seq signal data (https://bioconductor.org/packages/release/bioc/html/Gviz.html).

Knock-down experiments

Silencer Select pre-designed and validated small interfering RNAs (siRNAs) targeting C/EBPβ (Assay ID:s2891, 4392420, Thermo Scientific), CHOP (Assay ID:s3995, 4392420, Thermo Scientific), and a negative control (4390843, Thermo Scientific) were used for knockdown experiments. siRNA transfection was conducted using Lipofectamine 3000. Briefly, cells were seeded in culture plates to be 80% confluent on transfection day. Lipofectamine 3000 was diluted with Opti-MEM medium. The master mix of siRNA was prepared by diluting siRNA in Opti-MEM medium at a 15 nM final concentration for transfection. Diluted siRNA and diluted Lipofectamine 3000 were mixed in a 1:1 ratio. After incubating for 10 min at room temperature, the siRNA-lipid complex was added to the cells. The medium was replaced with drug-containing medium after 24 h. Knockdown efficiency was confirmed by western blot and RT-qPCR at 96 h post-transfection.

Quantification, statistical analyses, and reproducibility

All data sets except RNA-seq data were analyzed with GraphPad Prism (version 7.0). All assays were performed in triplicate. Statistical differences between groups were calculated by applying the Mann-Whitney U test or Student’s t test, as indicated. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Results

Fenretinide and GSK-126 treatments impair bladder cancer proliferation

Overexpression of EZH2 is implicated in various cancer types and, in general, is correlated with a worse prognosis and a higher metastatic stage [16, 43,44,45]. Upon comparing the expression of EZH2 in non-muscle invasive bladder cancers (NMIBC) and muscle invasive bladder cancers (MIBC), we detected a significantly increased level of EZH2 expression in muscle invasive cases, both in primary tumors and cell lines (Supplementary Fig. 1). Therefore, we envisioned that using MIBC cell lines would be quite informative to test the effects of single and combinatorial treatments.

We evaluated the effects of fenretinide and GSK-126 on cell proliferation using resazurin survival assays. Treatment of the MIBC cell lines with varying doses of fenretinide or GSK-126 for 48 and 72 h resulted in a dose-dependent decrease in cell proliferation (Fig. 1, Supplementary Fig. 2). Since overall responses to drugs required lower doses for 72 h, we conducted our further analysis after 72 h of incubation with these drugs (Fig. 1, Supplementary Fig. 2).

Fig. 1: The effects of fenretinide and GSK-126 on MIBC cell viability.
figure 1

A Dose-response curves of fenretinide and GSK-126 in MIBC cells. The cells were treated with increasing concentrations of either fenretinide or GSK-126 for 72 h, and viability was determined in comparison to vehicle control cells. Each point represents the mean and standard deviation of the triplicates. B IC50 concentrations of fenretinide or GSK-126 in MIBC cell lines after 72 h of treatment.

Full size image

Fenretinide and GSK-126 combination have a synergic interaction in suppressing cell viability

Considering the PRC2 complex is involved in RA-mediated gene expression as a co-repressor, we hypothesize that EZH2 activity inhibition may improve retinoid action [31]. To investigate retinoid activation and EZH2 inhibition cooperation, we designed our experiments to study their possible interactions. We tested the fenretinide and GSK-126 combination over a range of concentrations to determine whether the drug interactions were additive, synergistic, or antagonistic. Subjecting MIBC cells to increasing concentrations of fenretinide along with GSK-126 showed that this combination suppressed proliferative capacity to a greater extent than either treatment alone. Based on the dose-response results of the Bliss Independence model, the synergy scores of this drug combination were calculated as 38.09; 11.46; 4.67; and 57.01 for 5637, HT1376, J82, and T24, respectively (Fig. 2). Since a synergy score of >10 is generally accepted as indicating a synergistic interaction [46, 47], we concluded that fenretinide and GSK-126 showed synergistic effects in decreasing MIBC cell viability (except for the J82 cell line) (Fig. 2). Overall, these results strongly argue for further testing of the combination of fenretinide and GSK-126 against MIBC.

Fig. 2: Synergistic effects of fenretinide and GSK-126 on cell viability.
figure 2

Dose-response matrices and Bliss synergy maps showing synergistic effects between fenretinide and GSK-126 in 5637 (A), HT1376 (B), J82 (C), and T24 (D) cells. Bliss Independence synergy scores for the combination of drugs in the indicated concentration range were shown. Cells were treated with either single drugs or a drug combination for 72 h. Cell viability was assessed by comparison to vehicle control cells following each treatment in triplicate. The color of the surface indicates the drug combination effect; green indicates no effect, whereas red indicates maximum effect.

Full size image

Fenretinide and GSK-126 combination induces anti-oncogenic effects

To define the phenotypic changes related to drug applications, we conducted a series of molecular analyses linked to apoptosis, cell cycle, and gene expression. We used 5 µM of each drug, which corresponds to IC25-30 values, alone or in combination for 72 h, to continue our further analysis.

To examine whether the reduction in cell viability was due to apoptotic cell death, we stained the cells with Annexin V-FITC/propidium iodide (PI) after 72 h of treatment with drugs and their combinations. We observed a significant induction of apoptosis with co-treatment with fenretinide and GSK-126, compared with single treatments (Fig. 3A). This finding was also validated by microscopy images, showing the presence of apoptotic cells as revealed by membrane blebbing (Supplementary Fig. 3). Further, the cell cycle distribution of MIBC cells was analyzed after treatments with single agents or their combination. Compared to vehicle control, treatment of T24 cells with drug combination resulted in a decrease in the S phase, a subtle increase in the G1 fraction, and an increase in the sub-G1 fraction, indicating the role of drugs in causing a moderate cell cycle arrest (Fig. 3B).

Fig. 3: The anti-oncogenic effects of fenretinide and GSK-126.
figure 3

A The effects of drugs on apoptosis. T24 cells were treated for 72 h with fenretinide (5 μM), GSK-126 (5 μM), or the drug combination. Cell death was assessed by flow cytometry using Annexin-V and PI staining (*p < 0.05, **p < 0.01, ***p < 0.001, two-tailed unpaired t-test, n = 3, means ± SD are shown). A representative example is on the left. The data summarized is on the right. B The effects of drugs on the cell cycle. Cell cycle progression was measured by flow cytometry using PI in T24 cells after single or combination treatment for 72 h. A representative image from one experiment is on the left. The histogram on the right shows the distribution of cells in G0/G1, S, G2/M, and sub-G1 phases of the cell cycle. Data represent mean ± SD (two-tailed unpaired t-test: *p < 0.05, **p < 0.01, ***p < 0.001). C The effects of drug combination on cell migration. Left: Representative phase-contrast images of the scratch wound migration assay at different time points performed on T24 cell monolayers treated with indicated drugs. The punctuated white lines indicate the boundaries of the wound. Right: Quantification of the percentage of scratch wound coverage in T24 cells treated with fenretinide (5 μM, 72 h), GSK-126 (5 μM, 72 h), or the drug combination at different time points. Data represent mean ± SD (two-tailed unpaired t-test: *p < 0.05, **p < 0.01, ***p < 0.001). D The effects of drug combination on gene expression. RT-qPCR analysis of mRNA levels of indicated genes in T24 cells following 72 h treatment with fenretinide (5 μM), GSK-126 (5 μM) or the drug combination. Gene-specific data were normalized to GAPDH expression and are represented as average relative expression compared to vehicle controls. Error bars specify the standard deviation between triplicates (two-tailed, unpaired t-test: *p < 0.1, **p < 0.01, ***p < 0.001, ****p < 0.0001).

Full size image

To test the effects of drugs on cell migration, we conducted the scratch assay (Fig. 3C). Our results showed that the recovered wound area slightly decreased upon GSK-126 treatment and strongly decreased upon fenretinide treatment 24 h after the scratch. Notably, co-treatment reduced migratory capacity in a time-dependent manner compared with single drugs, especially 24 h after the scratch (Fig. 3C).

To gain more insights into how the anti-tumor efficacy of fenretinide and GSK-126 in bladder cancer is reflected in gene expression changes, we performed RT-qPCR analysis. We evaluated the changes in gene expression levels of genes involved in cell proliferation, apoptosis, and cell cycle processes in T24 cells following treatment with fenretinide, GSK-126, and their combination (Fig. 3D). Confirming our results obtained with the apoptosis assays (Fig. 3A), expression levels of pro-apoptotic genes CASP4 [48] and GADD45A [49] were significantly increased after fenretinide alone or combinatorial treatment, suggesting the major role of fenretinide associated with apoptosis. Expression of CDKN1C, implicated in G1 phase cell cycle arrest [50], was significantly increased with GSK-126 and combinatorial treatment (Fig. 3D).

Our findings suggest that fenretinide and GSK-126 cooperate to regulate the expression of a subset of genes critical for cell proliferation, apoptosis, and cell cycle processes. Collectively, these findings emphasize that targeting EZH2 in conjunction with the retinoic acid pathway stimulates the anti-oncogenic profile of MIBC.

Single or combinational treatments result in a distinct set of differentially regulated genes

To investigate the effects on gene regulatory networks associated with drug treatments at a genome-wide scale, we performed RNA-seq. In our analysis, a total of 4308 common protein-coding genes were mapped and found to be differentially expressed. Hierarchical clustering analysis of the differentially expressed genes revealed five groups of differentially expressed genes after single and combinational treatments compared to control (Fig. 4A). The gene ontology (GO) analysis revealed that the first cluster, strongly downregulated by combination treatment (CoSDR), consisted of the genes involved in RNA processing and mitotic spindle organization (Supplementary Fig. 4). The second cluster, downregulated by combination or fenretinide-only treatments (CoFeDR), involved the genes regulating chromatin organization, DNA replication, and related processes (Supplementary Fig. 4). The third cluster, upregulated by GSK-126 treatment (GskUR), consists of the genes involved in developmental functions (Fig. 4B). The fourth cluster, upregulated by fenretinide-only or combination treatments (CoFeUR), was involved in apoptosis, unfolded protein response, and the endoplasmic reticulum (ER) stress response pathways (Fig. 4B). There were only a limited number of genes upregulated by fenretinide-only treatment (FeUR, cluster 5), and those genes were not identified as being involved in a specific pathway. Overall, gene expression changes induced by single or combinational treatments support the phenotypic responses, where apoptotic pathways are triggered while proliferation and cell cycle processes are attenuated.

Fig. 4: Hierarchical clustering of differentially expressed genes.
figure 4

A Heat map depicting hierarchical clustering of genes differentially expressed in T24 cells treated with fenretinide (5 μM) and GSK-126 (5 μM) alone or in combination for 72 h versus control using log 2 fold change > 0.5 cutoff. Red and blue indicate high and low expression of genes, respectively. B Gene Ontology (GO) enrichment analysis demonstrating the upregulated biological processes between clusters. p and q values < 0.05. C Identification of combination-enhanced genes in the CoFeUR cluster. D Chord diagram of predicted EZH2 targets and their associated GO Biological Terms.

Full size image

Identification of EZH2 target genes

We analyzed the CoFeUR cluster more thoroughly by dividing the upregulated genes as combination-enhanced and fenretinide-driven (Fig. 4C). The genes identified as combination-enhanced in comparison to fenretinide are most likely upregulated by additional effects facilitated by GSK-126. Gene-concept network analysis revealed that combination-enhanced genes were associated with unfolded protein response (UPR), cellular responses to stimuli, and some metabolic processes, consisting of the genes ATF3, DNAJB1, HSPA1A, HSPA1B, and HSPA5 (Fig. 4D). Importantly, among those genes, ATF3 was previously identified as being upregulated after GSK-126 treatment [51]. To test the theory that these genes are repressed by EZH2 function, we performed chromatin immunoprecipitation (ChIP) of H3K27me3 for these genes. Remarkably, ChIP-qPCR analysis revealed that GSK-126 treatment significantly decreased H3K27me3 levels at combination-enhanced genes AKR1C2, ATF3, HSPB8, IGFBP3, MEIS3, NDRG1, PTGS2, and SERPINE2 (Supplementary Fig. 5A). We next assessed the alterations in a subset of predicted target genes to determine whether EZH2 inhibition results in the derepression of these genes in other MIBC cell lines as well. As anticipated, EZH2 inhibition by GSK-126 or combination treatments significantly increased the RNA expression levels of these genes in 5637 cells (Supplementary Fig. 5B).

Collectively, our findings demonstrated that combination-enhanced genes, largely consisting of UPR-related genes, are directly targeted by EZH2-mediated H3K27me3 in MIBC, supported by robust changes in gene expression profiles.

Association of differentially expressed genes with H3K27me3 profiles

After determining H3K27me3 levels at combination-enhanced genes (Supplementary Fig. 5A), we proceeded with the identification of H3K27me3 levels at the genome-wide level by performing ChIP-seq. Initially, we confirmed that EZH2 inhibition causes a global reduction in H3K27me3 levels without changing EZH2 levels (Fig. 5A). Next, we characterized the genome-wide alterations in the H3K27me3 profile induced by drugs. As predicted, we observed a global decrease in H3K27me3 levels, particularly at the transcriptional start sites of genes, in cells treated with GSK-126 alone or in combination with fenretinide (Fig. 5B). ChIP-seq data indicated that the genes belonging to the GskUR cluster have the highest H3K27me3 signal in control cells (Fig. 5B), most likely reflecting the involvement of these genes in developmental processes (Fig. 4B). Genes associated with the CoSDR, CoFeDR, and FeUR clusters showed nearly constant H3K27me3 profiles under all examined conditions (Supplementary Fig. 6). Remarkably, analysis of H3K27me3 profiles at the CoFeUR cluster reveals that H3K27me3 levels are even more decreased by combination treatment, particularly downstream of the transcriptional start sites (TSS), suggesting that fenretinide further reduces H3K27me3 levels in addition to GSK-126 treatment (Fig. 5B). GO term analysis demonstrated the involvement of CoFeUR genes in UPR and ER stress processes (Fig. 4B), and those genes were identified as EZH2 target genes by H3K27me3 ChIP-qPCR (Supplementary Fig. 5B). In line with this data, we confirmed that GSK-126 and combination treatments resulted in a decrease in H3K27me3 levels at UPR-related genes by ChIP-seq as well (Fig. 5C).

Fig. 5: The effects of fenretinide and GSK-126 on H3K27me3 profiles.
figure 5

A Western blots performed on T24 cells assessing H3K27me3 and EZH2 levels after treatment with fenretinide (5 μM, 72 h), GSK-126 (5 μM, 72 h), or the drug combination. B Signal intensity heatmaps illustrating H3K27me3 ChIP-seq at genomic regions in ± 3.0 kb distance of the transcriptional start sites at the genes associated with the GskUR and CoFeUR clusters in T24 cells. C Snapshots showing H3K27me3 ChIP-seq signals at ATF3, FICD, HSPA9, and PPP2CB loci in T24 cells. The signal peak regions are highlighted.

Full size image

Association of single and combinational treatments with distinct transcription factor activities

To identify the factors potentially regulating distinct classes of differentially expressed genes after single or combinational treatments, we checked the association of differentially expressed genes with transcription factor activities using the information available in the EnrichR database.

For the genes belonging to the CoSDR and CoFeDR clusters, E2F family transcription factors were considerably enriched, explaining the downregulation of the genes involved in cell cycle-related processes in these clusters (Fig. 6A, B). Further, generally, expression levels of activator E2F1 and E2F2 transcription factors themselves were strongly downregulated, whereas canonical repressor (E2F3B, E2F5, and E2F6) levels were strongly upregulated after fenretinide-only or combination treatment (Supplementary Fig. 7A) [52].

Fig. 6: Transcription factor activities of differentially expressed genes.
figure 6

The association of differentially expressed genes with transcription factor activities in the CoSDR (A), CoFeDR (B), GskUR (C), and CoFeUR (D) clusters, which was determined using the information available in the EnrichR database. Adjusted P value < 0.05. Violin plots showing the comparison of expression levels of predicted target and non-target genes for C/EBPβ (E) and CHOP (F), which were determined by data obtained from the EnrichR database. **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by the Mann-Whitney U test.

Full size image

We identified the genes in the GskUR cluster as predominant Polycomb targets, confirming their status as PRC2 target genes (Fig. 6C). The genes belonging to the CoFeUR cluster were enriched for the C/EBP transcription factor family (Fig. 6D, Supplementary Fig. 7C). Among transcription factors regulating the expressions of the genes belonging to the CoFeUR cluster, C/EBPβ (CCAAT Enhancer Binding Protein Beta) and CHOP (C/EBP homologous protein, DDIT3) were of particular interest. Remarkably, in addition to identifying them as the top enriched transcription factors, RNA-seq analysis showed that C/EBPβ and CHOP were among the genes substantially upregulated after combination treatments (log2 FC = 2.285 and 2.660, respectively, for C/EBPβ and CHOP after combination treatment). Additionally, we confirmed the increased level of expression for C/EBPβ and CHOP at the protein level (Supplementary Fig. 7B). For the CoFeUR cluster, enrichment for other transcription factors such as USF1, USF2, NFE2, STAT3, and VDR was not considered important since our gene expression data did not show marked changes for these factors. As C/EBPβ and CHOP are transcription factors capable of autonomously regulating their own target genes, we proceeded to assemble a set of their target genes derived from the EnrichR database. The target genes of C/EBPβ and CHOP were significantly upregulated after fenretinide-only or combination treatments as compared to non-target genes of these two factors (Fig. 6E, F), strengthening the relevance of C/EBPβ and CHOP in the regulation of genes in response to fenretinide or combination treatments. The role of transcription factor CHOP is well defined in the ER stress response and UPR [53], suggesting that UPR plays an important role in the synergistic effect of fenretinide and GSK-126.

Unlocking the role of C/EBPβ and CHOP in anti-oncogenic programs stimulated by co-treatment

UPR is one major manifestation of ER stress. In this pathway, several branches of signaling result in the activation of CHOP protein. In turn, CHOP triggers the expression of genes involved in apoptosis (Fig. 7A). It is known that CHOP can form heterodimers with other members of the C/EBP family [54]. C/EBPβ was also identified as being involved in the ER stress response [55].

Fig. 7: Role of CHOP in the anti-oncogenic efficacy of combination treatment.
figure 7

A The UPR stress sensors stimulate the transcriptional activity of ATF3, ATF4, ATF6, and HSPA5. These factors induce the transcription of CHOP, which downregulates the anti-apoptotic protein BCL-2 and upregulates the pro-apoptotic proteins, thereby initiating apoptosis. Created with BioRender. B The changes in the expression of the genes involved in ER-stress, UPR, and regulation of the apoptotic signaling pathways after CHOP knockdown in T24 cells treated with GSK-126 (5 μM, 72 h), fenretinide (5 μM, 72 h), or the drug combination. Cells were transfected with negative control (siNC) or siRNA targeting CHOP (siCHOP) 24 h prior to treatments. Data represent mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by unpaired two-tailed t-test. C The changes in the dead cell fraction after CHOP knockdown in T24 cells treated with fenretinide (5 μM, 72 h), GSK-126 (5 μM, 72 h), or the drug combination. Cell death was assessed by flow cytometry using Annexin-V and PI staining and calculated as the sum of early apoptosis, late apoptosis, and necrosis (*p < 0.05, two-tailed unpaired t-test). D A proposed model highlighting the mechanisms of co-targeting of the EZH2 and RA pathways in bladder cancer. Coordinated targeting of EZH2 methyltransferase activity inhibition and RA pathway activation leads to CHOP-mediated activation of ER stress, which stimulates apoptotic programs. Created with BioRender.

Full size image

Identifying C/EBPβ and CHOP as key candidate transcription factors associated with the UPR and ER-stress response inspired us to interrogate whether these factors were involved in the anti-oncogenic effects of combination treatment in MIBC cells. To explore their role, we knocked down C/EBPβ and CHOP in T24 cells using specific siRNAs and validated their efficacy (Supplementary Fig. 8A). We confirmed that CHOP depletion decreased the expression levels of the genes involved in the UPR, ER stress, and regulation of the apoptotic signaling pathways in the CoFeUR cluster and increased the expression levels of the anti-apoptotic gene BCL-2 (Fig. 7B, Supplementary Table 3). Notably, transcriptional repression of CHOP also caused a decrease in the dead cell fraction for treated cells with drug combination (Fig. 7C). Therefore, these results reinforce our hypothesis that CHOP is a critical molecular component of the apoptotic effects of drug combination via activating the UPR pathway. However, the results obtained with C/EBPβ depletion were less clear. C/EBPβ depletion led to an increase in the expression of the genes linked to UPR and the ER stress pathways, as well as the anti-apoptotic gene BAD (Supplementary Fig. 8B). This might be because C/EBPβ is known to show both pre- and pro-oncogenic effects depending on dimerization partners and the expression status of its isoforms [56]. As we have determined both CHOP and C/EBPβ to upregulate upon combinatorial treatment, it might be speculated that C/EBPβ upregulation is most likely a reflection of CHOP-dependent upregulation.

Collectively, these data suggest that the simultaneous effects of retinoic acid pathway activation and EZH2 inhibition activate CHOP, ultimately triggering ER stress-mediated anti-oncogenic transcriptional programs.

Discussion

High-grade bladder tumors harbor considerable risks of recurrence, local invasion, and metastasis. Remarkably, the landscape of therapy regimens for patients with MIBC has not shifted beyond standard management, consisting of radical cystectomy and conventional chemotherapy [57]. Monotherapies largely have limited efficacy because of substantial tumor heterogeneity and rapidly engender drug resistance. Therefore, providing novel treatment options is urgently needed for bladder cancer patients.

In this study, we developed a novel approach to target MIBC using the combined action of EZH2 methyltransferase inhibition and the RA signaling pathway as a conceptual framework to potentiate the therapeutic response. Our data illustrated that the combined effects of EZH2 inhibition and retinoids synergistically inhibit bladder cancer proliferation and stimulate the expression of marker genes associated with apoptosis and cell cycle arrest. Our coordinated targeting strategy has anti-oncogenic consequences; thus, this approach might be a viable strategy for MIBC treatment. This phenomenon correlates with the existing literature highlighting the efficacy of simultaneous targeting of PRC2 and RA signaling in many cancers. According to research by Villa and colleagues, RA treatment can reverse most of the epigenetic silencing caused by PRC2 in leukemia patients [58]. Recently, Benoit and colleagues showed that PRC2 inhibition may overcome RA resistance [59].

Despite progress in the conception of the potent action of retinoids in the treatment or prevention of cancer, there is still a lack of understanding of downstream targets that highlight their biological effects. We demonstrated that the combination treatment induces C/EBPβ and CHOP expressions, thereby triggering anti-oncogenic transcriptional reprogramming and resulting in enhanced apoptosis in bladder cancer. In support of the previous studies, our analyses further showed that the CHOP transcription factor conducts anti-oncogenic functions by stimulating the ER stress response and apoptosis, as reflected by a decrease in the expression of the genes associated with the UPR and the ER stress pathways, increased expression levels of the anti-apoptotic gene BCL-2, as well as a decreased apoptosis rate after CHOP knockdown [60, 61]. Previous investigations have established the role of C/EBPβ and CHOP as the prominent retinoid-dependent target genes [62,63,64]. Our results, highlighting C/EBPβ and CHOP-mediated ER stress response as a regulatory node for the anti-oncogenic effects of fenretinide, support this conception. Another crucial direction of our results is the positioning of CHOP as one of the putative apoptotic regulators in MIBC. CHOP is widely regarded as a primary controller of ER stress-induced apoptosis.

Our hypothesis is that CHOP stimulates anti-oncogenic transcriptional reprogramming with the help of EZH2, reassembling accessible chromatin profiles in MIBC cells. Alterations in chromatin architecture conducting apoptotic transcriptional reorientation underlie the synergistic interactions of this combination. Therefore, we report that reconfiguration of the apoptotic program is coupled to EZH2. Other groups have reported that disruption of PRC2 activity can enhance apoptosis [24]. We further nominate C/EBPβ and CHOP as unique downstream targets of the EZH2/RA pathway interaction, which may reflect a molecular circuitry supporting apoptotic reprogramming. The output of EZH2-mediated transcriptional rerouting of CHOP results in increased apoptosis and may thereby contribute to an improvement in susceptibility to treatment for bladder cancer. We also attribute the anti-oncogenic effects of this combination to cell cycle arrest, reflecting the repression of the E2F1 and E2F2 family transcription factors implicated in cell cycle-related processes by combination treatment. Recently, Lee and colleagues identified E2F1 as a potential EZH2 regulator and a crucial mediator for bladder cancer progression [65]. We observed an increase in the canonical cell cycle repressors E2F3B, E2F5, and E2F6 [52]. Consistent with our findings, Poplineau and colleagues highlighted the key role of the E2F/EZH2 axis for RA resistance [66]. We observed an increase in repressors E2F7 and E2F8 with only EZH2 inhibitor treatment.

RNA-seq analyses and subsequent validation guide us in identifying certain genes as targets of PRC2. We specify a mechanism by which EZH2 regulates the repression of a subset of genes mainly functioning in response to topologically incorrect proteins, chaperone-mediated protein folding, and certain metabolic processes. Underpinning our observations, Arbuckle and colleagues demonstrated that EZH2 inhibition promotes the expression of inflammatory cytokines and mediators such as PTGS2 and transcription factors related to the ER stress response, including C/EBPβ, ATF4, and CHOP [67]. Supporting the previous research by other groups, cellular stress response genes including IGFBP3, NDRG1, PTGS2, and SERPINE2 were identified as EZH2 target genes in our study [68, 69] .

Several preclinical and clinical trials proved the superior efficacy and reduced therapeutic recurrence of EZH2 inhibitors, especially in lymphomas but not in solid tumors [70]. The overall increment in the efficacy of EZH2 inhibitors in hematologic versus solid malignancies is attributed to the existence of mutations that cause catalytically hyperactive EZH2, which renders them susceptible to EZH2 inhibitor treatment. Our observation of the strongest combination’s efficacy in T24 cells may be associated with the presence of KDM6A mutations. Cancers harboring loss-of-function mutations in the SWI/SNF complex and KDM6A have been considered to be sensitive to EZH2-based epigenetic therapy [71]. Ler and colleagues established that KDM6A mutations eventuate in prolonged repression of PRC2-regulated genes in urothelial bladder carcinoma, pointing to the therapeutic benefit of EZH2 inhibition in KDM6A-null or PRC2-enriched urothelial bladder carcinoma [72].

Collectively, in this study, we provide additional insights revealing that EZH2 inhibition in MIBC tumors cooperates with retinoids, causing activation of a network of genes providing anti-oncogenic properties. Intriguingly, the network of co-regulatory factors recruited by RA-EZH2 inhibition cooperation has a principal role in the response to ER stress. We hypothesize that PRC2 inhibition and retinoic acid pathway stimulation serve as a “second strike” that makes the anti-oncogenic effects of retinoids more penetrant. Consequently, sustaining apoptosis could provide a mechanistic explanation for the cooperative effect of RA gain and PRC2 loss. Substantially, our findings unveil a completely new aspect of the RA/EZH2 partnership.

Our observations regarding alterations in the transcriptome and epigenome may not reflect the actual clinical scenario. Based on its potent synergistic effects in cell culture, further validation of this novel drug combination regarding efficacy, tolerability, and pharmacokinetics in animal models is imperative for broadening our results for use in clinical settings.

Conclusion

Overall, our findings suggest a concerted mechanism between retinoic acid pathway activation and EZH2-mediated alterations in epigenetic signatures. Our data highlight the potential of incorporating RA/EZH2 targeting for reflection of the anti-oncogenic profile. Consequently, our strategy appears to be a rational approach for improving the therapeutic efficacy in bladder cancer and contributes to targeted treatment paradigms.