Data collectionThe Cancer Genome Atlas (TCGA) PRAD

Due to the high heterogeneity of PRAD data, for the analysis, we selected only samples of patients that meet the following criteria: Gleason scores 6 through 8; no hormonal therapy; over 55-year old (Additional file 1: Table S1). Also, we did not include samples marked as contaminated with admixtures of other tissues in TCGA. After quality control and filtration steps, 225 samples of cancerous and 52 normal tissue samples remained.

Processing of RNA-seq dataTCGA and ICGC data

Level 3 data with mapped readings were used. To build correlation networks, TPM normalization was applied. Only those genes that were expressed in over 50% cancer samples (> 1TPM) with expression variability lying above the 25% percentile were included in further analysis. For differential analysis between cancer and not cancer, DESeq2 package was used. Genes were considered differentially expressed, if abs(logFC) > 0.5 and FDR < 0.05.

Taylor and fraser data

Normalized intensities were downloaded with GEO using GEOquery.

Data on knockdown and overexpression of TP63 and TRIM29RNA-seq data

Pre-processing of raw readings was performed with trimmomatic. Then, the reads were mapped onto the genome version hg19 using STAR software. To quantify gene coverage, the featureCounts software with gencode v37 annotation was used. For differential analysis, the DESeq2 package was used. Genes were considered differentially expressed, if |logFC|> 0.5 and FDR < 0.05.

Hybridization chip data

Normalized intensities were downloaded with GEO using GEOquery. Differential analysis of expression was performed with the Limma package.

Analysis of DNA methylation dataTCGA dataThe 450 K Methylation platform

Files containing raw intensity values (*.idat files) were downloaded from the TCGA portal. To process the data, RnBeads [56] program with the following setup parameters was used: methylumi.noob [57] + BMIQ [58] normalization; limma algorithm for differential analysis. CpG site was considered differentially methylated if |Δβ| > 0.2 and FDR < 0.05.

Data on TP63 overexpression in PC3EPIC Methylation platform

To process the data, RnBeads [56] program was used with the following setup parameters: ENmix [59] + BMIQ [58]; limma algorithm for differential analysis [60]. To search for motifs of transcription factors, HOMERv4 [61] software with the motif database HOCOMOCOv11 [62] were used.

Analysis of CNV data

Level 3 open data obtained on Affymetrix 6.0 platform were used. A region was considered to have an altered copy number if the absolute value of the ratio of intensities between sample and control exceeded 0.3. To describe chromosomal instability, chromosomal instability index (CIN), equal to the ratio of total length of CNV to the length of a chromosome, averaged across all chromosomes, was used.

Determination of clusters of co-expressed genes associated with epigenomic variability and chromosomal instability

To generate clusters, we used the WGCNA package [16] with standard parameters. We found 13 clusters of co-expressed genes in TCGA data. For each cluster, we determined its expression pattern, cluster eigengene (EG), which is the main component of PCA.

As it has been demonstrated in a previous work, the level of methylation of CpG sites can strongly depend on the ratio of various types of prostate cells in a sample [21]. We used the MuSic package [63] and prostate scRNA-seq data [22] to calculate the ratio of stromal, basal, and luminal epithelium cells in the prostate gland in TCGA samples. Calculated ratios of different epithelial cells were used as co-variants for further regression analysis. To determine the set of non-overlapping CpG sites associated with each cluster, we used the following regression model:

$$}}_ = }}_ + }}_*}}_ + }}_ + }}_,$$

where EGi is cluster i eigengene; ai, a constant; bi, coefficient by the vector of methylation of CpGi; and Cov, co-variants that include the ratio between various types of cells in a sample and the rest eigengenes of the cluster except for the studied one to determine only CpG sites uniquely associated with the cluster. We thus determined 14,443 differentially methylated CpG sites associated with eigengenes of the clusters.

We used linear regression to determine gene clusters associated with CIN:

$$}\hspace=\hspace}}_\hspace+\hspace}}_*}}_\hspace+\hspace}}_\hspace+\hspace}}_,$$

where CIN is chromosomal instability index, EGi, is cluster i eigengene; ai, a constant; bi, a coefficient at EGi; Cov, co-variants that include the ratio between various types of cells in a sample.

To determine association of clusters with differentially expressed genes between cancer and normal tissue, we used GSEA [20], where genes of each cluster were used as signatures.

Prediction of super-enhancer-associated genes

We used the ROSE framework [23] with DNAse-seq and H3K27ac ChIP-seq for normal prostate epithelial cell line PrEC for SEs prediction. For the prediction link between genes and enhancers and SEs, we used the Activity-by-contact framework [30].

Cell culture

Human RWPE-1 cells (ATCC® CRL-11609™) were cultured in Keratinocyte Serum-Free Media (Gibco, USA). Prostate cancer cell lines (PC3, LNCaP, 22Rv1) were kindly provided by Dr. M. Lagarkova (Federal Research and Clinical Center of Physical–Chemical Medicine, Moscow, Russia) and were cultured in corresponding media supplemented with 10% FBS and penicillin/streptomycin. Human cells were grown at 37 ℃ and 5% CO2 in a conventional humidified CO2 incubator.

Plasmids

The FLAG-tagged TRIM29 and its truncated forms were cloned into the LeGO-iG2 vector. pcDNA3-TP63-FLAG was obtained from addgene (cat. 26,979). FLAG-tagged TP63 was cloned штto LeGO-iT2. The truncated forms of TP63 were cloned into the pcDNA3.1 vector.

Preparation of lentivirus particles

One day prior to transfection, Phoenix cells were inoculated onto 10-cm Petri dishes covered with 0.1% gelatin 7 × 105 cells per dish. The cells were transfected with auxiliary plasmids containing the Rev (15.3% by mass to total DNA), RRE (29.8% by mass to total DNA), and VSV-F (5.6% by mass to total DNA) genes and target plasmid encoding TRIM29_FLAG. Transfection was performed using the TurboFect (ThermoFischer Scientific, US) transfection agent, 26.4 µL per 13.2 µg DNA. The procedure of transfection was performed in accordance with the manufacturer's recommendations. Virus-containing supernatant was collected 24, 48, and 72 h after transfection, filtered through a 0.45-µm filter and stored at – 70 ℃.

Generation cell lines with stable expression of TP63 and TRIM29

Two days before transduction, cells were seeded into 24-well plate 2 × 104 cells per well. One hour prior to transduction, 8 µg/mL polybrene was added. Cells were transduced with virus particles containing TP63-FLAG or TRIM29-FLAG. After 24 h, cell medium was replaced. Two days after transfection, cells were trypsinized and TP63/TRIM29-infected cells were isolated on a FACS BD Aria III cell sorter.

Cell treatment with TNFα

Prior to treatment with TNFα, cells were cultured to 50–60% confluence layer. TNFα was added into the medium to a final concentration of 100 ng/mL and cultured in an incubator for 48 h.

Chromatin immunoprecipitation

Chromatin immunoprecipitation was performed using the SimpleChIP Plus Enzymatic Chromatin IP Kit (Cell Signaling) according to the manufacturer's manual. In brief, 4 × 106 cells were fixed with formaldehyde, cell nuclei were isolated, and chromatin was fragmented with micrococcal nuclease. Then, chromatin was incubated with 5 µL anti-TP63 antibodies (Cell Signaling #13109) overnight and TP63-bound DNA was precipitated on magnetic beads. The DNA was purified from bound protein complexes and used for massive parallel sequencing using the Illumina HiSeq-2500 or for quantitative PCR.

The following algorithm was used to analyze chromatin immunoprecipitation followed by sequencing: pre-processing of raw reads was performed by trimmomatic tool [64]; then the reads were mapped onto human genome version hg19 using the bwa mem [65] program with standard parameters. The *.bam files thus obtained were sorted and indexed with the samtools program. Reliably determined regions of TP63 or histone binding with determined modifications were found using the macs2 program [66]. ChIP-seq data were visualized using the deepTools program [67].

Paired-end libraries were prepared according to the manufacturer’s recommendations using NEBNext Ultra II DNA Library Prep Kit (New England Biolabs, USA). The libraries were indexed with NEBNext Multiplex Oligos kit for Illumina (96 Index Primers, New England Biolabs, USA). Size distribution for the libraries and their quality were assessed using a high-sensitivity DNA chip (Agilent Technologies). The libraries were subsequently quantified by Quant-iT DNA Assay Kit, High Sensitivity (Thermo Scientific, USA). DNA sequencing was performed on the HiSeq 2500 platform (Illumina, USA) according to the manufacturer’s recommendations, using the following reagent kits: HiSeq Rapid PE Cluster Kit v2, HiSeq Rapid SBS Kit v2 (200 cycles), HiSeq Rapid PE FlowCell v2 and a 1% PhiX spike-in control.

TP63 and TRIM29 knockdown

TP63 and TRIM29 knockdown was performed using RNA interference (siTP63_1_F: CCGUGAGACUUAUGAAAUGTsT, siTP63_1_R: CAUUUCAuAAGUCUCACGGTsT; siTP63_2_F: UCACGACAGUCUUGUACAATsT, siTP63_2_R: UUGUACAAGACUGUCGUGATsT; siTRIM29_1_F: AUUGAUGAGCAAUUACUCUTsT, siTRIM29_1_R: AGAGUAAUUGCUCAUCAAUTsT; siTRIM29_2_F: ACCAAGUGAAGGUGAUCAUTsT, siTRIM29_2_R: AUGAUcACCUUcACUUGGUTsT). siRNA was transfected into RWPE-1 cells in a 24-well plate using HiPerFect Transfection Reagent (Qiagen, 301,704). siRNA, 68 ng, was mixed with 2 µL transfection reagent in 100 µL of Opti- MEM™ (Gibco) medium; the mixture was incubated at room temperature for 10 min. RWPE-1 cells were seeded onto a 24-well plate at 3.5 × 104 cells/cm2. Cells were introduced into wells in 400 µL Keratinocyte SFM (Gibco) medium and the mixture of siRNA and transfection reagent was added to the cells immediately. Therefore, total medium volume was 500 µL in a well; final siRNA concentration, 10 nM. After 24 h, the medium was partially replaced with fresh Keratinocyte SFM (Gibco). The efficiency of transfection was evaluated 48 h after the addition of siRNA by western blotting.

Cell lysate preparation for western blot analysis

Cell cultures were collected, and 1 × 106 cells were counted by trypan blue exclusion and washed with ice cold PBS. Cells were then resuspended in 100 μL PBS and lysed with 100 μL of 2 × Laemmli Sample Buffer supplemented with β-mercaptoethanol by boiling for 5–10 min.

Isolation of cytoplasm and nuclear fractions

Cells were washed with PBS and removed with a scraper in cool buffer A (0.35 M sucrose, 2 мM MgCl2, 0.1 mM EDTA, 0.1% Triton X-100, 10 mM Tris–HCl pH 8.0, 1 mM DTT, 0.1 mM PMSF, and protease inhibitors). Cell lysate was passed 10 times through a 21G needle syringe and a 23G syringe, 3 times. Then, it was centrifuged at 2000g for 10 min. The supernatant was collected as a cytoplasm fraction.

The precipitate containing nuclei was washed with 0.5 M sucrose and centrifuged at 10,000g for 10 min. The precipitate was resuspended in buffer C (400 mM NaCl, 20 mM HEPES pH 7.9, 25% glycerol, 1.5 mM MgCl2, 0.4 mM EDTA, 0.4 mM PMSF, 1 mM DTT, protease inhibitors) and passed 5 times through a 21G needle syringe. Then, it was incubated at + 4 ℃ for 40 min and centrifuged at 12,000g for 30 min. The supernatant was collected as the nuclear fraction.

Immunoprecipitation

Cells were washed with PBS solution. Cell lysis buffer (Cell Signaling Technology, Great Britain) was added and incubated for 5 min on ice. Cells were removed with a scraper. Cell lysate was homogenized on an ultrasound bath. Lysate was centrifuged at 10,000g for 10 min. The supernatant was incubated with antibodies overnight at + 4 ℃ while stirring. Lysate with antibodies was incubated with magnetic beads for 1 h while stirring at room temperature. Magnetic beads were washed 5 times with a cell lysis buffer.

For western blotting, the proteins were eluted with Laemmli buffer at 95 ℃ for 5 min. Total protein concentration in the sample was determined using Bradford assay.

To prepare samples for mass spectrometry analysis, proteins were eluted with an elution buffer (8 M urea, 2 M thiourea, 10 mM Tris/HCl pH 8.0) on a shaker at 25 ℃ for 2 h.

Western blotting

Electrophoretic separation of proteins was carried out in a Bio-Rad Mini Protean chamber. Then, semidry electrotransfer of proteins to a PVDF membrane was carried out in a Trans-Blot Turbo Transfer System (Bio-Rad, USA). The membrane was incubated in a blocking buffer (PBST, 5% nonfat dry milk). Then, it was incubated with primary antibodies against GAPDH (PAB932Hu02, Cloud-Clone Corp., USA), phospho-Histone H2A.X (Ser139) (JBW301, Sigma-Aldrich, USA), TP63 (D2K8X Cell Signaling Technology, USA) or TRIM29 (E1L4E, Cell Signaling Technology, USA) at+ 4 ℃ overnight, washed in PBST solution, and incubated with a solution of secondary antibodies conjugated with horseradish peroxidase (31460, Thermo Fisher Scientific, USA) for 1 h at room temperature. Proteins were visualized using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific, USA) at Chemidoc (Biorad, USA).

Sample preparation for mass spectrometric analysis

After immunoprecipitation, DTT was added to the eluted samples to a final concentration of 5 mM and incubated for 30 min on a shaker at room temperature to restore disulfide bonds. Freshly prepared iodoacetamide was added to a final concentration of 10 mM and incubated in the dark for 20 min to alkylate the thiol groups of cysteine. The mixture was diluted with 35 mM ammonium carbonate solution fourfold. Sequencing Grade Modified Trypsin (Promega, USA) was added 0.1 μg trypsin per 10 μg protein and incubated on a shaker at 37 ℃ overnight. Trypsin was neutralized with a fivefold volume of a 5% formic acid solution. Desalting was performed using reversed phase chromatography on homemade StageTips with SDB-RPS filter according to the protocol from Rappsilber et al [68]. The samples were concentrated in a vacuum centrifuge and redissolved in 3% acetonitrile with 0.1% trifluoroacetic acid solution. The amount of protein in the sample was approximately 10 μg.

LC–MS/MS analysis of tryptic peptides

After trypsinolysis, peptide fractions were loaded onto a column (diameter 75 μm, length 50 cm) with an Aeris Peptide XB-C18 2.6 μm sorbent (Phenomenex) in an aqueous solution containing 3% acetonitrile and 0.1% trifluoroacetic acid. Separation of peptides was performed on an Ultimate 3000 Nano LC System (Thermo Fisher Scientific), coupled to a Q Exactive HF mass spectrometer (Thermo Fisher Scientific) using a nanoelectrospray source (Thermo Fisher Scientific). Peptides were loaded onto a heated 40 ℃ column in buffer A (0.2% formic acid (FA) in water) and eluted with a linear (120 min) gradient 4 → 55% buffer B (0.1% FA, 19.9% water, 80% acetonitrile) in A at a flow rate of 350 nL/min. Before each new load, the column was washed with 95% buffer B in A for 5 min and equilibrated with buffer A for 5 min.

Mass spectrometry data were saved with automatic switching between MS1 scans and up to 15 MS/MS scans (topN method). The target value for the MS1 scan was set to 3 × 106 in the range of 300−1200 m/z with a maximum ion injection time of 60 ms and a resolution of 60,000. The precursor ions were isolated with a window width of 1.4 m/z and a fixed first mass of 100,0 m/z. The precursor ions were fragmented by high-energy dissociation in a C-trap with a normalized collision energy of 28 eV. MS/MS scans were saved with a resolution of 15,000 at m/z 400 and at a value of 1 × 105 for target ions in the range of 200−2000 m/z with a maximum ion injection time of 30 ms.

Analysis of LC–MS/MS data

The conversion of the "raw" mass spectrometric data from the instrument into MGF (Mascot Generic Format) mass sheets was carried out using the ProteoWizard msconvert utility with the following parameters: MS Levels 2–2, Peak Picking 2–2, Threshold Peak Filter Absolute intensity—Most intense—1, Zero Samples 2–2.

For identification and quantitative analysis of protein partners, the MaxQuant program (v1.5.3.30) was used with the Andromeda algorithm against the protein database UniProt Knowledgebase (UniProtKB), the human taxon, with the following parameters: the accuracy of determining the parent and daughter ions was 20 and 50 ppm, respectively; protease, trypsin; one missed cleavage per peptide is possible; variable modification of methionine, oxidation; fixed modification of cysteine, carbamidomethylation. The reliability of identification of both peptides and proteins was limited to 1% FDR, which was determined using the "target decoy" approach. For quantitative label-free analysis, LFQ values were calculated using the MaxQuant software.

Immunocytochemical analysis

Cells were washed with PBS solution 2 times for 5 min and fixed with 4% PFA for 30 min. The cells were washed 2 times for 5 min. Membranes were permeabilized with 0.1% Triton X-100 in PBS for 5 min. Nonspecific antigen adsorption was blocked by washing in 0.1% Tween 20 PBS solution 3 times for 5 min and then in a block solution (PBS, 0.1% Tween 20, 5% FBS, 5% goat serum) for 30 min. A solution of primary antibodies against GAPDH (PAB932Hu02, Cloud-Clone Corp., USA), phospho-Histone H2A.X (Ser139) (JBW301, Sigma-Aldrich, USA), TP63 (D2K8X Cell Signaling Technology, USA) or TRIM29 (E1L4E, Cell Signaling Technology, USA) was added in a block solution and incubated for 2 h at room temperature. The cells were washed from primary antibodies with a solution of 0.1% Tween 20 3 times for 5 min. A solution of secondary antibodies labeled with the Alexa Fluor 555 fluorophore was added in PBS and incubated in the dark for 1 h at room temperature. Then, the cells were washed from secondary antibodies with a solution of 0.1% Tween 20 3 times for 5 min. To stain the nuclei, the mixture was incubated with DAPI 100 ng/mL for 10 min in the dark at room temperature and then washed with PBS solution once. PBS was removed and glycerol was applied to the sample prior to covering with a coverslip. The stained proteins were visualized using a fluorescence microscope (Nikon Eclipse Ni-E, Japan). The number of γH2AX foci was calculated using ImageJ software with FindFoci plugins. 60–200 cells were analyzed in each sample.

DNA isolation and bisulfite conversion

Cells (~ 106) were resuspended in a lysis buffer (10 mM Tris pH = 8.0, NaCl 100 mM, 10 mM EDTA pH = 8.0, 0.5% SDS) and incubated with Proteinase K overnight. For DNA purification the phenol chloroform extraction was used. The DNA was then bisulfite converted using the EpiMark® Bisulfite Conversion Kit (E3318S, NEB) according to the manufacturer's protocol.

Infinium methylation EPIC Beadchip array

All DNA methylation experiments were performed according to Illumina manufacturer instructions for the Infinium Methylation EPIC 850K BeadChip Array (Illumina, USA). EPIC BeadChips were imaged using the Illumina iScan System (Illumina, United States).

Isolation of RNA and cDNA synthesis

Isolation of RNA from cells was performed using the Lira reagent (Biolabmix, Russia) according to the manufacturer's protocol. The quality of the isolated RNA was assessed using agarose gel electrophoresis. The concentration of the resulting RNA preparation was measured spectrophotometrically on an Infinite 200 Pro M Plex plate reader (Tecan, Switzerland) determining the absorbance at 260 nm. The purity of the preparation was assessed by the ratio of absorption at wavelengths of 260 and 230 nm.

A mixture was prepared containing 1 μg of RNA, 1 U DNase I (Thermo Fisher Scientific, USA), and a reaction buffer with MgCl2. Incubation was carried out for 30 min at 37 ℃. Then, 1 μL of 50 μM EDTA was added and the solution was heated to 65 ℃ to inactivate DNase.

The first strand of cDNA was synthesized from a single-stranded RNA template using the MMLV RT kit (Evrogen, Russia) according to the manufacturer's protocol.

RNA sequencing

Library preparation was performed with NEBNext Poly(A) mRNA Magnetic Isolation Module and NEBNext Ultra II Directional RNA Library Prep Kit (NEB) according to the manufacturer’s protocol. The library underwent a final cleanup using the Agencourt AMPure XP system (Beckman Coulter) after which the libraries’ size distribution and quality were assessed using a high-sensitivity DNA chip (Agilent Technologies). Libraries were subsequently quantified by Quant-iT DNA Assay Kit, High Sensitivity (ThermoFisher). Finally, libraries were sequenced by a high-throughput run on the Illumina HiSeq 2500 using 2 × 125 bp paired-end reads.

Real-time PCR

Real-time PCR was performed in a volume of 20 μL using the obtained cDNA as a template, qPCRmix-HS SYBR (Evrogen, Russia) containing HS Taq DNA polymerase, SYBR Green I dye, a mixture of deoxynucleoside triphosphates, Mg2+, reaction buffer, and primers for the target gene (TMPRSS2_ERG_fwd-CAGGAGGCGGAGGCGGA and TMPRSS2_ERG_rev-GGCGTTGTAGCTGGGGGTGAG; dNp63_fwd-GAAAACAATGCCCAGACTCAA and dNp63_re-TGCGCGTGGTCTGTGTTA; TRIM29_fwd-AAAGGCTATCCCTCCCTCAT and TRIM29_rev-TAGAATGGCCGGTAGTGAGA) and a reference gene (ACTB_fwd—CACCATTGGCAATGAGCGGTTC and ACTB_rev—AGGTCTTTGCGGATGTCCACGT). Each sample was prepared four times.

MSRE-PCR

The protocol was used from the article (Melnikov et al. 2005) with minor modifications. HpaII was used as a methyl-sensitive restriction enzyme. 500 ng genomic DNA was incubated with a restriction enzyme overnight. The treated DNA was purified by phenol–chloroform extraction. The purified DNA was used as a template for quantitative analysis. The region of the GAPDH gene in which there are no HpaII restriction sites was used as a control.

Statistics or data analysis

Results were analyzed using two-tailed Student’s t-test, Fisher’s exact test and two-way ANOVA. P value less than 0.05 was considered statistically significant.