1 Introduction

Lysine acetylation is regulated by two groups of enzymes: Histone acetyltransferases (HATs) mediate the acetylation event [[1]], and histone deacetylases (HDACs) regulate the deacetylation event [[2]]. Lysine acetylation occurs on many proteins and therefore influences pathways with diverse functional roles [[3]]. Significantly, aberrant protein acetylation is recognised to take on an important role in driving the malignant phenotype [[4]]; thus, deregulation of HDAC activity occurs in different types of cancer and HDAC as a cancer drug target has been validated in many preclinical models [[5]]. Therapeutically, however, clinical success in human disease has been surprisingly limited [[6]]; most clinical activity has been observed in haematological malignancies [[7]] where recent HDAC inhibitor drug approvals include panobinostat for multiple myeloma and chidamide for T-cell lymphoma [[8]]. Generally speaking, HDAC inhibitors have met with limited success in solid cancer clinical trials [[7]]. For example, in clinical trials on colorectal cancer (CRC), negligible activity was observed in treated individuals [[9]]. This may reflect our limited insights into the key molecular mechanisms and cancer-relevant pathways upon which HDAC inhibitors act. Having this information at hand could allow for a more scientifically driven and rational clinical development plan.

CXD101 is a promising second-generation inhibitor with selective activity towards class 1 HDAC subunits [[10]]. It is a potent antiproliferative agent, which in human clinical studies demonstrated a favourable safety profile [[10]]. In addition, encouraging durable clinical activity was seen in a phase I clinical trial in patients with T-cell lymphoma, follicular lymphoma and Hodgkin lymphoma (including postallogenic stem cell transplantation), with tumour reduction evident in 63% of patients [[10]]. Although efficacious in haematological malignancy, we wanted to develop a scientific rationale for deploying CXD101 in the solid cancer setting [[11]].

With this objective in mind, we have sought to explore the mechanisms through which CXD101 acts. By performing a genome-wide expression analysis on human CRC cells, we identified a diverse set of differentially expressed genes (DEGs) upon treatment with CXD101. Functional profiling of the gene expression data highlighted biologically enriched concepts related to immune recognition, specifically antigen presentation (AP) and natural killer (NK) cell activity. Similar concepts were apparent in gene expression data derived from the murine CRC cell line colon26 treated with CXD101 in vitro and in colon26 syngeneic tumours growing in vivo. The enriched immune recognition concepts reflected changes in the tumour microenvironment (TME), where a marked effect on the population of tumour-infiltrating lymphocytes and other immune relevant cells was observed upon treatment. These results led us to test the therapeutic impact of CXD101 in combination with agents that act through the immune system, such as the immune checkpoint inhibitors (ICIs) anti-PD-1 and anti-CTLA4 [[12]]. Under conditions where there was minimal effect of the ICI monotherapy, enhanced antitumour activity was observed in the CXD101 combination treatment, suggesting that the gene expression changes and the subsequent impact on antigen presentation in the TME act to enhance the antitumour effects of ICIs. Our results have important implications for understanding the mechanisms, which underpin HDAC inhibitor-based therapies and provide a powerful rationale for testing the combined effect of CXD101 with ICIs in human solid malignancies.

2 Materials and methods 2.1 Cell culture and compound treatment

Human colorectal adenocarcinoma SW620 (ATCC® CCL-227; RRID:CVCL_0547) and HCT116 (ATCC® CCL-247; RRID:CVCL_0291), human breast cancer MCF7 (ATCC® HTB-22; RRID:CVCL_0031), human lung cancer A549 (ATCC® CCL-185; RRID:CVCL_JK07) and the mouse colon carcinoma cell line colon26 (ATCC® CRL-2639; RRID:CVCL_7255) were obtained from ATCC (Manassas, VA, USA). Human cell lines were cultured in Dulbecco's modified Eagle medium (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% FBS (Labtech, Heathfield, UK) and 1% penicillin/streptomycin (Gibco, Life Technologies, Carlsbad, CA, USA), while colon26 cells in RPMI (Sigma-Aldrich). All cell lines were tested for mycoplasma contamination before use. CXD101 was used as described [[10]].

2.2 MTT assay

Cells were seeded onto 96-well plates overnight and the next day dosed with CXD101 and incubated for 72 or 120 h. Next, 100 µL of thiazolyl blue tetrazolium bromide (MTT; Sigma-Aldrich) was added into a well (final concentration 5 µm) and incubated for 2 h at 37 °C. After that medium was discarded and formazan crystals were dissolved in 100 µL DMSO (VWR International, Radnor, PA, USA) by shaking for 15 min. Absorbance was read by Omega FLUOstar plate reader (BMG Labtech Ltd, Ortenberg, Germany) at the 584 nm wavelength. Data were analysed, and IC50 doses were calculated in graphpad prism 8 (GraphPad Software, San Diego, CA, USA; RRID:SCR_002798).

2.3 RNA extraction library preparation and RNA-seq analysis

SW620 and colon26 cells were treated as described with CXD101 or DMSO as a negative control. Total RNA (triplicates unless otherwise stated) was isolated using Direct-zol RNA MiniPrep Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer's instructions. RNA sequencing was performed by BGI Genomics (Beijing, China). Briefly, an Agilent 2100 Bioanalyzer (Agilent RNA 6000 Nano Kit; Santa Clara, CA, USA) was used for RNA sample quality control purposes (RNA concentration, RIN value, 28S/18S and the fragment length distribution). mRNAs were isolated from total RNA using the oligo(dT) method. Then, the mRNAs were fragmented, and first-strand/second-strand complementary DNAs (cDNAs) were synthesised. cDNA fragments were purified and resolved with EB buffer for end reparation and single nucleotide A (adenine) addition. Subsequently, the cDNA fragments were linked with adapters. Those cDNA fragments with suitable size were selected for the PCR amplification. Agilent 2100 Bioanalyzer and ABI StepOnePlus Real-Time PCR System were used in quantification and qualification of those libraries. The RNA sequencing was carried out using Illumina HiSeq Platform (SW620) or BGI500 platform (colon26 in vitro and in vivo).

2.4 RT-qPCR

RNA was isolated from cells using TRIzol (Thermo Fisher Scientific, Waltham, MA, USA) or the Direct-zol RNA MiniPrep Kit (Zymo Research) according to the manufacturer's instructions. One microgram of total RNA was used for cDNA synthesis. Reverse transcription with oligo(dT)20 primer (Invitrogen, Carlsbad, CA, USA) was performed using SuperScript III Reverse Transcriptase (Invitrogen) as per the manufacturer's instructions. Quantitative reverse transcription PCR (qRT-PCR) was carried out in technical triplicate using the indicated primer pairs and the Brilliant III Ultra-Fast SYBR® Green qPCR Master Mix (Agilent) on an AriaMX Real-Time qPCR Instrument (Agilent). Results were expressed as average (mean) fold change compared with control treatments using the ΔΔCt method from three biological repeat experiments. Glyceraldehyde phosphate dehydrogenase primer sets were used as an internal calibrator. Error bars represent SE unless otherwise indicated.

2.5 Genome-wide expression analysis in human and mouse cell lines and tissues

FASTQ files for CXD101 and DMSO-treated SW620 and colon26 cells were generated. Sequencing reads were trimmed to remove adapters and low-quality bases with trimgalore v.0.4.3 (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) (RRID:SCR_011847). Likewise, FASTQ files for experiments addressing CXD101 treatment in mice were generated in four biological replicates and reads were trimmed as described above. The expression data for all three experiments have been further processed as follows; the trimmed reads were aligned to the human and mouse reference genomes (builds hg19 and mm10, respectively) with star aligner v.2.7 (RRID:SCR_015899) [[13]] with two mismatches allowed. Differential gene expression analysis was done with deseq2 r bioconductor package (v.1.22) (RRID:SCR_000154; RRID:SCR_006442) [[14]], using read count data provided by the aligner. Genes were considered differentially expressed if the adjusted P-value, calculated using the Benjamini–Hochberg method in order to minimise the false discovery rate, was less than 0.01. We further filtered significantly DEG sets to select for genes expressed at high levels using twofold change in absolute expression levels for both human and mouse data sets.

Gene expression data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series (RRID:SCR_005012) Accession Number GSE158164.

2.6 Gene set enrichment analyses

Genes significantly differentially expressed upon CXD101 treatment of SW620 cells were further subjected to enrichment analyses in the pi [[15]] and xgr software packages (v.1.12 and v1.1.6, respectively) [[16]] to reveal signalling and metabolic pathways over-represented in the DEG sets. For pathway-based gene enrichment analyses, we used Reactome Knowledge Base [[17]] pathways and, more specifically, superpathways that correspond to domains of biology such as immune system and signal transduction. P-values for pathway enrichment analyses were calculated using the formula for hypergeometric distribution, reflecting the probability for a pathway to arise by chance. Significantly enriched pathways were identified using a threshold FDR of 0.05.

2.7 Parametric GSEA

Parametric gene set enrichment analysis was performed with r pgsea package (v. 1.58) [[18]] on the collections of curated gensets (c2) derived from the KEGG pathway database (RRID:SCR_012773) available from the Broad Institute's Molecular Signatures Database (MSigDB v. 6.2). The expression matrix used in these analyses was normalised and rlog-transformed with deseq2 r package. Gene sets with less than 10 genes and more than 10 000 genes were excluded from the analyses. A linear model was applied employing the limma package (RRID:SCR_010943) (v.3.44.0) [[19]] followed by empirical Bayesian analysis to determine concepts associated with significant differences between treated and untreated samples. Differences were considered significant if the adjusted P-value, calculated using the Benjamini–Hochberg method [[20]] in order to minimise false discovery rate, was less than 0.005.

To perform parametric gene set enrichment analyses in mouse experiments, we used annotations provided by gskb r bioconductor package [[21]], which contains molecular signature databases for pathway analysis in the mouse. The procedure for discovery of significantly over-represented biological concepts (i.e. pathways) was the same as above; however, differences were considered significant if the adjusted P-value was less than 0.01.

2.8 Functional genomics analysis

For the analysis of ‘antigen processing and presentation’ and ‘natural killer cell-mediated cytotoxicity’ pathway gene expression levels in human cancers, Xena Browser (University of California, CA, USA) was used (https://xena.ucsc.edu/) [[22]]. The TCGA TARGET GTEx data set was selected, which contained transcript expression data from TCGA (https://portal.gdc.cancer.gov/; cancer tissue) and Genotype-Tissue Expression (GTEx; https://gtexportal.org/home/; healthy tissue) samples. For subsequent detailed analysis of microsatellite stability and staging, TCGA colon, stomach and oesophageal cancer data sets collected from cBioPortal (www.cbioportal.org) were used [[23, 24]]. Data were presented as heat maps generated using morpheus software (Broad Institute, Cambridge, MA, USA).

2.9 Genevestigator analysis

For the analysis of ‘antigen processing and presentation’ and ‘natural killer cell-mediated cytotoxicity’ pathway gene expression levels in different cell lines treated with HDAC inhibitors, the Genevestigator tool (Nebion AG, Zurich, Switzerland) (RRID:SCR_002358) was used. Data from Subramanian et al. [[25]], Kubo et al. [[26]] and Ghandi et al. [[27]] were collected and presented as graphs or as heat maps generated using morpheus software (Broad Institute, Cambridge, MA, USA).

2.10 Western blotting

Cell pellets were lysed in radio-immunoprecipitation assay buffer [50 mm Tris/HCl (pH 8), 150 mm NaCl, 1% Igepal CA-630, 0.5% sodium deoxycholate, 0.1% SDS, 0.2 mm sodium orthovanadate and protease inhibitor cocktails], for 30 min on ice and centrifuged for another 30 min at maximum speed at 4 °C. Protein concentration was assessed by Bradford assay (Quick Start™ Bradford 1× Dye Reagent; Bio-Rad Laboratories, Hercules, CA, USA). After gel electrophoresis, proteins were transferred onto the PVDF or nitrocellulose membrane by means of Trans-Blot® Turbo™ Transfer System (Bio-Rad Laboratories) and blocked by 1-h incubation in 5% skimmed milk (Merck Group, Darmstadt, Germany) in PBST at room temperature. The following antibodies were used in immunoblotting: anti-H3 (ab1791; Abcam, Cambridge, UK; RRID:AB_302613), anti-H3AcK9 (ab10812; Abcam; RRID:AB_297491), anti-H3AcK14 (#7627; Cell Signaling, Danvers, MA, USA; RRID:AB_10839410) and anti-β-Actin (#3700; Cell Signaling; RRID:AB_2242334), all overnight at 4 °C. Next, membranes were washed and treated with secondary antibody for 1 h at RT. Chemiluminescent signals were detected by LI-COR C-Digit (LI-COR Biosciences, Lincoln, NE, USA), and the data were quantified using imagej software (National Institutes of Health, Bethesda, MD, USA) (RRID:SCR_003070).

2.11 Immunohistochemistry

Tumours were harvested at day 14, embedded in paraffin blocks and cut into 5-µm sections. FFPE slides were washed for 5 min with HistoChoice (Sigma-Aldrich), followed by two times of 3-min washing in 100% ethanol, 3 min in 70% ethanol and 5 min in tap water. Next, samples were incubated with antigen retrieval solution (e.g. sodium citrate buffer or Tris/EDTA – depending on used antibody) at 99 °C in water bath for 20 min. After 3× washing with purified water, samples were incubated in freshly made 6% methanol/H2O2 for 15 min and washed in tap water. In the next steps, slides were washed in 1% PBST for 5 min, blocked in blocking serum solution (VECTASTAIN ABC Kit; Vector Laboratories, Burlingame, CA, USA) for 20 min., washed again in 1% PBST for 5 min and incubated overnight at 4 °C (staining with anti-H3AcK9 was performed for 8 min at room temperature) with primary antibody: anti-H3AcK9 (ab10812; Abcam; RRID:AB_297491), anti-CD8 (ab203035; Abcam), anti-CD4 (ab183685; Abcam; RRID:AB_2686917), anti-CD68 (ab125212; Abcam; RRID:AB_10975465), anti-CD163 (ab182422; Abcam; RRID:AB_2753196), anti-NKp46 (ab224703; Abcam), PD-L1 (ab233482; Abcam; RRID:AB_2811045) and anti-FoxP3 (14208S; New England Biolabs, Ipswich, MA, USA). Samples were further stained with secondary antibody (VECTASTAIN ABC Kit) at room temperature. In the next step, ABC solution (VECTASTAIN® ABC-HRP Kit, Peroxidase, Rabbit IgG, PK-4001; Vector Laboratories) was added for 30 min, and slides were washed in 1% PBST and incubated with DAB solution (Vector DAB) for another 10 min. Sections were counterstained with haematoxylin (Sigma-Aldrich). Results were analysed using Leica DM2500 optical microscope (Wetzlar, Germany) and presented as semi-quantitative using imagej software (National Institutes of Health).

2.12 Evaluation of CXD101 monotherapy

All experiments and protocols were approved by the animal welfare body at Charles River Discovery Research Services Germany (where experiment was performed) and the local authorities, and were conducted according to all applicable international, national and local laws and guidelines. Twenty female Balb/c mice (RRID:IMSR_CRL:547) at 6–8 weeks of age (10 mice per group: control and CXD101 treated) (Charles River Laboratories, Freiburg, Germany) received unilateral subcutaneous injections of 5 × 105 colon26 cells in PBS in a total injection volume of 100 µL/mouse. Upon reaching individual tumour volumes of 50–150 mm3, mice were assigned to treatment groups based on tumour volumes aiming at comparable group mean/median tumour volumes. Within 24 h of randomisation, mice were daily treated by oral administration (gavage) with 50 mg·kg−1 (dosing volume 10 mL·kg−1) of CXD101 using 5% DMSO/PBS as a vehicle. Body weights and tumour volume [mm3] were performed by calliper measurement twice weekly. Termination of individual mice was conducted at day 14 of the experiment or at > 1000 mm3 (unilateral), in case of tumour ulceration or body mass loss at < 70% of initial weight. From each group, four snap-frozen tumours were collected for RNA isolation and four formalin-fixed samples were prepared for immunohistochemical staining.

2.13 Evaluation of CXD101 in combination therapy

Animal welfare for this study complied with the U.S. Department of Agriculture's Animal Welfare Act (9 CFR parts 1, 2 and 3). All experimental data management and reporting procedures were in strict accordance with applicable Crown Bioscience San Diego Guidelines and Standard Operating Procedures where the study was performed.

The tumour model colon26 was implanted subcutaneously in immunocompetent BALB/c mice (Charles River Laboratories; RRID:IMSR_CRL:547). The experiment comprised six groups of six mice each, the first of which was a vehicle control group treated only with the vehicle for CXD101. The second group received CXD101 monotherapy, administered orally at a dose level of 50 mg·kg−1 with a 5-day on/2-day off schedule. Groups 3 and 4 were treated with anti-mPD-1 or anti-mCTLA4 monotherapy (both Bio X Cell; RRID:SCR_004997), respectively. Both agents were administered intraperitoneally, the former twice weekly at a dose level of 5 mg·kg−1 and the latter on day 1 (5 mg·kg−1) and days 3 and 6 (2.5 mg·kg−1). Groups 5 and 6 received a combination of CXD101 with anti-mPD-1 or anti-mCTLA4, respectively. Body weights and tumour volume [mm3] by calliper measurement were performed twice weekly. The treatment phase was followed by a dose-free observation period of varying length (max. until day 38) depending on antitumour efficacy and/or condition of the mice.

The tumour model MC38 was implanted subcutaneously in immunocompetent 6–8 weeks of age female C57BL/6 mice (RRID:MGI:5658456). The experiment comprised four groups of eight mice each (unless stated otherwise), the first of which was a vehicle control group treated only with the vehicle for CXD101. The second group received CXD101 monotherapy, administered orally at a dose level of 50 mg·kg−1 with a 5-day on/2-day off schedule. Group 3 was treated with anti-mPD-1 monotherapy (Bio X Cell, Lebanon, NH, USA) administered intraperitoneally twice per week at a dose level of 10 mg·kg−1 (days 1, 6, 11, 16, 21 and 26). Group 4 received a combination of CXD101 with anti-mPD-1. Body weights and tumour volume [mm3] by calliper measurement were performed twice weekly. The treatment phase was followed by a dose-free observation period of varying length (max. until day 38) depending on antitumour efficacy and/or condition of the mice.

2.14 Statistical analysis

Statistical analyses were performed using two-tailed, unpaired Student's t-test and one-way ANOVA test with graphpad prism 8 Software (GraphPad Software; RRID:SCR_002798). Data are shown as means with SD displayed unless otherwise indicated. P-values lower than 0.05 were considered significant and are labelled by asterisks (*) for P < 0.05, (**) for P < 0.01, (***) for P < 0.001 and (****) for P < 0.0001.

3 Results 3.1 CXD101 treatment causes global effects on gene expression in human cells

We assessed the effect of CXD101 on a variety of human CRC cell lines and chose SW620 cells for further analysis because of the sensitivity to CXD101 at 72 h of treatment and coincidental increase in the level of acetylation on histone H3 lysine (K) 14 (Fig. S1A,B) [[28]]. In the RNA-seq analysis, we used treatment conditions where there was minimal effect on cell viability but where an increased acetylation mark was apparent (1 µm for 48 h; Fig. S1A,B). We performed RNA-seq on polyA-enriched RNA prepared from CXD101-treated SW620 cells and compared the data to the vehicle-alone (DMSO) treatment. Sequencing reads (FASTQ format) were aligned to the reference human genome (hg19) with star aligner and analysed for differential expression using deseq2 r package [[29]]. The sequencing data were of high-quality with on average 92% of the reads able to be mapped to the genome. Genes were considered to be significantly differentially expressed if the adjusted P-value, calculated using the Benjamini–Hochberg method in order to minimise the false discovery rate, was less than 0.01. We further filtered significantly DEG sets to select for genes expressed at moderately high levels using a twofold change in absolute expression level. Thus, mining the RNA-seq data for genes differentially regulated upon CXD101 treatment revealed a large number (over 1000) of candidates (Fig. 1A). The majority of them were upregulated although a significant proportion was downregulated (70% compared with 30%, respectively) (Fig. 1A). We then applied parametric gene set enrichment analyses to the normalised rlog-transformed gene expression matrix, which disclosed enrichment of several immune-related KEGG concepts, including natural killer (NK) cell-mediated cytotoxicity (Fig. 1B). Inspection of DEGs for enriched KEGG concepts identified sets of genes associated with immune system (Fig. 1A; shown in violet on the separate lane on the left of the heat map).

Genome-wide analysis on CXD101-treated SW620 cells. (A) Heat map of differential gene expression. The heat map shows 1192 significantly DEGs between CXD101-treated cells (1 µm for 48 h; n = 2) and DMSO control (n = 3). Normalised rlog-transformed gene expression values corresponding to significantly expressed genes (FDR  1 were mean-centred by rows. Each row of the heat map represents transformed expression values of one DEG across all samples (blue, low expression; red, high expression). Genes associated with immune system-related KEGG (in.kegg), pathways revealed with PGSEA (see panel B) are indicated in violet on a separate panel on the left of the heat map (see also Dataset S1). (B) Immune system-related KEGG concepts encompassing pathways associated with significant differences in expression change upon CXD101 treatment in SW620 cell line. Heat maps show PGSEA statistic (Z-score), which characterises how much the mean of the fold changes for genes in a certain pathway deviates from the mean observed in all the genes between CXD101 treatment (n = 2) and the control (DMSO, n = 3) groups. Blue indicates gene sets with decreased expression, while red corresponds to those with increased. (C) Gene enrichment analyses (pi/xgr r package) on preranked significantly DEG lists showing enriched Reactome pathways, and specifically Reactome immune system (reactome.is) and Reactome signal transduction (reactome.st) pathways. Gene ranking was performed using data sets provided and strategies implemented in xgr package. (D) Heat map showing significantly DEGs associated with Reactome immune system, referred to as the AP signature. Gene expression values were calculated as described above (see panel C). (E) Heat map of differential expression showing significantly DEGs associated with ‘Natural Killer Cell-Mediated Cytotoxicity’ KEGG pathway. Gene expression values were transformed as described above (see panel A). (F) qRT-PCR validation of genes identified in panels D and E (i and ii, respectively) in SW620 cells treated for 2 days with 1 µm CXD101 or DMSO control (Student's t-test; *P S1A(i).

Next, we ran a targeted gene enrichment analyses on preranked lists, which allowed us to address the relevance of the findings to immune system function and dysfunction (Fig. 1C). We found highly enriched Reactome pathway descriptors and specifically Reactome superpathways associated with the immune system and cytokine signalling in the immune system (Fig. 1C). We constructed a heat map from significant DEGs within the Reactome immune system descriptor, focussing on major histocompatibility complex (MHC) class I and class II genes, for further analysis (Fig. 1D); we refer to this as the antigen presentation (AP) signature. We did the same with significant DEGs associated with the natural killer cell-mediated cytotoxicity KEGG pathway (Fig. 1E), from which we selected for further analysis highly expressed genes (Fig. S2A); we refer to this signature as the natural killer (NK) signature.

It was important to validate the results from the RNA-seq and the genes assigned to the aforementioned ontologies. We therefore measured the expression of individual transcripts by qPCR in cells. Many of the MHC class I and class II genes within the AP signature were upregulated at the single gene level in SW620 cells treated with CXD101 (Fig. 1Fi). When the same set of AP signature genes was analysed in other cell lines, including the CRC HCT116 cells, breast cancer MCF7 cells and lung cancer A549 cells, CXD101 treatment caused a similar increase in gene expression (Fig. S2B–G). We also measured genes within the NK signature where many, at the single gene level, exhibited increased expression in treated cells (Fig. 1Fii). These results highlight the ability of CXD101 to regulate immune-relevant gene expression.

3.2 CXD101 upregulates genes involved with immune recognition

The KEGG and Reactome analysis of DEGs in the human cell line identified gene signatures associated with antigen presentation and natural killer cells. We reasoned therefore that CXD101 in situ may have a wider impact perhaps on immune response to the tumour, in addition to a direct antiproliferative effect on tumour cells. We investigated this idea using the murine syngeneic colon cancer colon26 model, regarded to be MSS genomic status [[30]] where, initially, we studied genome-wide expression changes in colon26 cells growing in vitro upon treatment with CXD101. We performed RNA-seq on polyA-enriched mRNA under treatment conditions where there was a CXD101-dependent increase in histone acetylation (Fig. 2Fiii and Fig. S1C,D). Subsequently, the RNA-seq data were aligned to the reference Mus musculus genome (mm10) with STAR aligner and analysed for differential expression. Over 90% of the reads could be mapped to the murine genome.

Genome-wide analysis on CXD101-treated colon26 cells. (A) Heat map of differential gene expression observed in colon26 cells. The heat map shows 2514 significantly DEGs between CXD101-treated (2.7 µm for 72 h) and control (DMSO) colon26 cells. Normalised rlog-transformed gene expression values corresponding to significantly expressed genes (FDR  1 were mean-centred by rows. Each row of the heat map represents transformed expression values of one DEG across all samples (blue, low expression; red, high expression). Genes associated with immune system-related KEGG pathways (see panel (B)) are indicated in violet on a separate panel (in.mm kegg) on the left of the heat map (see also Dataset S2); n = 3. (B) Significantly over-represented KEGG concepts encompassing pathways associated with significant differences in expression change upon CXD101 treatment in colon26 cells. Heat maps show PGSEA statistic (Z-score), which characterises how much the mean of the fold changes for genes in a certain pathway deviates from the mean observed in all the genes between CXD101 treatment and the control (DMSO) groups. Blue indicates gene sets with decreased expression, while red corresponds to those with increased; n = 3. (C) Significantly over-represented GO terms encompassing biological processes or cellular components associated with significant differences in expression change upon CXD101 treatment in colon26 cells. Heat maps show PGSEA statistic (Z-score), which characterises how much the mean of the fold changes for genes in a certain pathway deviates from the mean observed in all the genes between CXD101 treatment and the control (DMSO) groups. Blue indicates gene sets with decreased expression, while red corresponds to those with increased; n = 3. (D) Gene expression heat map showing significantly DEGs associated with ‘Antigen processing and presentation’ KEGG pathway, referred to as AP signature. Gene expression values were transformed using approach outlined above (see panel A). (E) Heat map of differential expression showing significantly DEGs associated with ‘Natural Killer Cell-Mediated Cytotoxicity’ KEGG pathway, referred to as NK signature. Gene expression values were transformed using approach outlined above (see panel A). (F) qRT-PCR of genes identified in panels D and E (i and ii, respectively) in colon26 cells treated for 3 days with 2.7 µm CXD101 or DMSO control (Student's t-test; *P n = 3.

We constructed heat maps from the CXD101 and vehicle (DMSO)-treated colon26 cell data, which indicated that in mouse (as in human) cells a large proportion of genes were upregulated upon CXD101 treatment (1891 upregulated and 611 downregulated; Fig. 2A). Genes found to be associated with ‘Antigen Processing and Presentation’ and ‘Natural Killer Cell-Mediated Cytotoxicity’ KEGG pathways (Fig. 2B; referred to as AP and NK, respectively) as a result of PGSEA were attributed to ‘in.mm_kegg’ category and portrayed alongside the heat map. We could further support this statement with PGSEA-derived GO term enrichment results, which similarly exhibited MHC-associated gene set enrichment (Fig. 2C, and for comparison a general GO analysis, Fig. S7C). In a side-by-side comparison of the AP and NK KEGG pathways, the majority of genes within these categories were upregulated upon CXD101 treatment (Fig. 2D,E). For the AP category, we focussed on class I and class II genes within the murine MHC H2 gene locus, which were induced (Fig. 2D). Similarly, for the NK pathway the expression levels of many genes within the KEGG pathway were upregulated in CXD101-treated cells (Fig. 2E). We further evaluated the expression of highly expressed genes at the single gene level by qPCR (Fig. S3A). For the AP pathway, MHC H2 class I and class II genes were induced (Fig. 2Fi). For the NK pathway, many genes were upregulated in treated cells (Fig. 2Fii).

3.3 Genome-wide effects of CXD101 during tumorigenesis

To assess gene expression in the TME, we evaluated the effect of CXD101 in the syngeneic colon26 carcinoma model in tumours grown subcutaneously in Balb/c mice, with CXD101 given orally for two consecutive 5-day periods. RNA-seq was performed on polyA-enriched RNA purified from the tumours. Formalin-fixed paraffin-embedded samples were prepared in parallel to assess by immunohistochemistry (IHC) any change in the cellular content of the TME.

CXD101 treatment caused a significant inhibition of tumour growth with minimal effect on body weight (Fig. 3A). The tumour RNA-seq data were aligned to the reference M. musculus genome (mm10) with STAR aligner where over 90% of the reads mapped to the mouse genome. We analysed DEGs (|log2 FC| >1 and FDR <1%) that were upregulated and downregulated and created a heat map of the expression changes where many genes were upregulated with a smaller group downregulated upon CXD101 treatment (Fig. 3B). Similarly, PGSEA revealed a proportion of the DEGs in the tumour RNA-seq data associated with KEGG pathways with immune-related functions (Fig. 3B). The same two KEGG pathways namely ‘Antigen Processing and Presentation’ (AP) and ‘Natural Killer Cell-Mediated Cytotoxicity’ (NK) were found to be enriched in the CXD101 tumour expression data (Fig. 3C). In the GO term, enrichment analysis performed with PGSEA, MHC-associated genes and other immune-related terms were enriched upon CXD101 treatment (Fig. 3D). In a side-by-side comparison of the AP and NK signatures, a majority of genes within each pathway were upregulated upon CXD101 treatment (Fig. 3E,F).

Genome-wide analysis on CXD101-treated colon26 tumours. (A) Schematic representation of the experiment with CXD101 in colon26 tumours (i). Balb/c mice were treated with orally administrated CXD101 at 50 mg·kg−1 for 14 days with respect to vehicle-only control; n = 10 per group; (ii) relative tumour growth volume in CXD101-treated and nontreated Balb/c mice presented as a mean value (Student's t-test of the values at day 14; *P t-test; *P n = 4) and control (n = 3) colon26 syngeneic tumour samples. Normalised rlog-transformed gene expression values corresponding to significantly expressed genes (FDR  1 were mean-centred by rows). Each row of the heat map represents transformed expression values of one DEG across all samples (blue, low expression; red, high expression). Genes associated with immune system-related KEGG pathways (see panel (C)) are indicated in violet on a separate panel (in.mm kegg) on the left of the heat map (see also Dataset S3). (C) Significantly over-represented KEGG concepts encompassing pathways associated with significant differences in expression change upon CXD101 treatment of the tumours. Heat maps show PGSEA statistic (Z-score), which characterises how much the mean of the fold changes for genes in a certain pathway deviates from the mean observed in all the genes between CXD101 treatment (n = 4) and the control (DMSO, n = 3) groups. Blue indicates gene sets with decreased expression, while red corresponds to those which were increased. (D) Significantly over-represented GO terms encompassing biological processes or cellular components associated with significant differences in expression change upon CXD101 treatment of the tumours. Heat maps show PGSEA statistic (Z-score), which characterises how much the mean of the fold changes for genes in a certain pathway deviates from the mean observed in all the genes between CXD101 treatment (n = 4) and the control (DMSO, n = 3) groups. Blue indicates gene sets with decreased expression, while red corresponds to those with increased. (E) Gene expression heat map showing significantly DEGs associated with ‘Antigen processing and presentation’ KEGG pathway, the AP signature. Gene expression values were transformed using approach outlined above (see panel B). (F) Heat map of differential expression showing significantly DEGs associated with ‘Natural Killer Cell-Mediated Cytotoxicity’ KEGG pathway, the NK signature. Gene expression values were transformed using approach outlined ab