Molecular cloning

For CRISPR/Cas9-mediated genome editing, sgRNAs were subcloned into pX330 or pLentiCRISPR v2 according to Feng Zhang protocol.51 Briefly, BbsI-digested pX330 or pLentiCRISPR v2 were ligated to the annealed and phosphorylated sgRNA. All constructs were subjected to Sanger sequencing with human U6 primer before used in the study.

For shRNA cloning, potent shRNAs were predicted using the algorithm by Pelossof et al 52 and cloned into the MLPe vector (MSCV-LTR-miR-E-PGK-Puro-IRES-GFP) as described before.53 The hairpins were ordered as 97-mer oligos and were PCR amplified with miRE-XhoI and miRE-EcoRI primer. The PCR product was then digested with EcoRI-HF and XhoI and purified before ligated to the EcoRI-HF/XhoI-digested MLPe backbone. To evaluate the knockdown efficiency of the shRNAs, target cells were transduced with a multiplicity of infection (MOI) <0.7 to achieve single copy integration. After puromycin selection cells were harvested and subjected to western blot. Knockdown efficiency was compared with cells transduced with Renilla luciferase, a non-targeting control. Two most potent shRNAs against KDM6A were chosen and further cloned into pT3-TRE-tRFP-miR-E and pCol-TGM. Both plasmids were also digested with EcoRI-HF and XhoI, and ligated to the EcoRI-HF/XhoI-digested PCR amplicons (hairpin). All plasmids were sequenced with miR-E primer before used in the study. Oligonucleotides sequences of sgRNA, shRNA and primers are listed in online supplemental table 5.

For overexpression plasmid, pT3-EF1a-MYC-IRES-Deptor was constructed by PCR amplification of Deptor from Addgene plasmid 21 334 and was cloned into pT3-EF1a-MYC-IRES-rtTA3, which was digested with MscI and XmaI with NEBuilder HiFi DNA Assembly (NEB) according to the manufacturer’s protocol.

Animal experiments

The group size for individual animal experiments was determined on the basis of our experience with previous similar experiments. For hydrodynamic tail vein injections, 8 weeks old female C57Bl/6 animals were purchased from Envigo. 5 µg DNA of pT3-EF1a-MYC, 20 µg of pX330 expressing indicated sgRNAs together with CMV-SB13 Transposase (1:5 ratio); 20 µg of pT3-TRE-tRFP-shRNA, 5 µg of pT3-myc-IRES-rtTA3 and CMV-SB13; 5 µg of pT3-EF1a-MYC-IRES-Deptor, 20 µg of pX330 sgRNA and CMV-SB13 were prepared in a sterile 0.9% Sodium chloride (NaCl) solution and injected into the lateral tail vein with a total volume corresponding 10% of body weight in 5–7 s as described before.

All animals were monitored daily and animal experiments were performed in compliance with all relevant ethical regulations determined in the animal permit. On euthanasia, relevant organs from experimental mice were visually inspected, harvested and photographed. Tumour samples were taken to obtain genomic DNA, RNA, protein and the rest were incubated in 4% paraformaldehyde for at least 24 hours for further use. All animal experiments were approved by the regional board Karlsruhe, Germany.

ESC-based pancreatic cancer

Embryonic stem cells (ESCs) harbouring disease predisposing alleles as described before were used to introduce the conditional KDM6A knockdown by generating pCol-TGM plasmids containing the shRNA. Two shRNAs against KDM6A and one control shRNA against Renilla luciferase were cloned into pCol-TGM plasmid and were electroporated into the ESCs together with pCAGs-FLPe to mediate recombinase-mediated cassette exchange. After electroporation ESCs were selected for hygromycin resistance and only ESCs with successful integration of the targeting construct at the ColA1a locus confer hygromycin resistance. After selection, resistant clones were picked and expanded. Targeted clones were used for blastocyst injections (in cooperation with the DKFZ transgenic core facility) to generate cohorts of chimeric mice, which were directly used for further experiments. Mice received Dox-containing diet at the age of 4 weeks in order to activate shRNA expression and thus KDM6A knockdown. Disease onset was monitored by weekly palpation.

MRI

MRI was carried out by our small animal imaging core facility in DKFZ using a Bruker BioSpec 1Tesla (Ettlingen, Germany). For the imaging, mice were anaesthetised with 3% sevoflurane in air. T2-weighted imaging were performed using a T2_RARE_ sequence axial: TE=84 ms, TR=4806.1582 ms, FOV 30×30 mm, slice thickness 1 mm, averages=4, Scan Time 461.39 s, echo spacing 8 ms, rare factor 10, slices 20, image size 192×192, flip angle 180. If liver lesions can be observed in T2, contrast-enhanced T1 measurement (80 µl ProHance, 0.5 mmol/kg, Bracco, intraperitoneal injection) were carried out to visualise and quantify tumour growth. Unfortunately, the liver tumours did not accumulate the contrast reagent and thus a volumetric size determination with T1 was not possible. The size determination was then performed using T2-weighted MRI images. The region of interest (ROI) were drawn manually in each layer and total volume of the lesion from the individual ROI was calculated with ParaVision software (Bruker). The evaluation was carried out by the same person throughout the study.

Immunohistochemistry

Samples were fixed in 4% paraformaldehyde for at least 72 hours, embedded in paraffin and sliced into 2 µm thick sections for immunohistochemistry (IHC) staining.

For Kdm6a and GFP staining in mouse tissue, slides were deparaffinised with xylene, rehydrated through a descending alcohol series and washed in water. For antigen retrieval, slides were put in sodium citrate buffer (pH 6.0) and boiled in a pressure cooker for 8 min followed by cooling down under running water for 5 min. Subsequently slides were blocked with 3% hydrogen peroxide for endogenous peroxidase activity, washed for 1 min under running water and twice with phosphate-buffered saline (PBS) and blocked with 5% bovine serum albumine (BSA) in PBS with 0.05% Triton X-100 at room temperature for 1 hour. Slides were incubated with the rabbit monoclonal anti-KDM6A overnight at 4°C. Slides were then washed three times with PBS+0.05% Triton X-100, incubated with ImmPRESS goat anti rabbit or mouse IgG Polymer Kit, peroxidase reagent (Vector Laboratories #MP-7451) for 30 min at room temperature and washed further twice with PBS+0.05% Triton X-100. Staining was visualised using ImmPACT DAB Peroxidase Substrate Kit (Vector Laboratories #SK-4105) according to the manufacturer’s protocol and were counterstained with haematoxylin. Finally, the slides were washed through ascending alcohol series that ended with xylol and mounted with Surgipath Micromount Mounting Medium (Leica # 3801731).

For c-casp3 and both pS6RP staining in mouse tissue, the BOND-MAX (Leica Biosystems) was used to carry out automated IHC staining of slides. Antigen retrieval was carried out with BondTM citrate solution (AR9961, Leica) or BondTM EDTA solution (AR9640, Leica). Thereafter sections were incubated with the specific antibodies against antigens in BondTM primary antibody diluent (AR9352, Leica Biosystems). Primary antibody exposure was followed by secondary antibody (Leica Biosystems) and staining using the Bond Polymer Refine Detection Kit (DS9800, Leica Biosystems). For quantification of stainings, slides were scanned using a SCN400 slide scanner (Leica Biosystems) at 20× magnification.

For human tissues, liver specimens were fixed overnight in 4% paraformaldehyde and embedded in paraffin. Sections were done at 5 µm in thickness. For immunohistochemical staining, slides were deparaffinised in xylene, rehydrated through a graded alcohol series and rinsed in PBS. Antigen retrieval was performed in 10 mM sodium citrate buffer (pH 6.0) by placement in a microwave oven on high for 10 min, followed by a 20 min cool down at room temperature. After a blocking step with the 5% goat serum and Avidin-Biotin blocking kit (Vector Laboratories, Burlingame, California, USA), the slides were incubated with the primary antibodies overnight at 4°C. Slides were then subjected to 3% hydrogen peroxide for 10 min to quench endogenous peroxidase activity and, subsequently, the biotin conjugated secondary antibody was applied at a 1:500 dilution for 30 min at room temperature. The immunoreactivity was visualised with the Vectastain Elite ABC kit (Vector Laboratories) and Vector NovaRed (Vector Laboratories) as the chromogen. Slides were counterstained with haematoxylin. Slides were evaluated semi-quantitatively by comparing each tumour with its surrounding non-neoplastic counterpart, thus defining ‘high’ and ‘low’ the levels of a given protein in a HCC sample when compared with the corresponding non-tumourous counterpart.

Derivation of primary liver tumour cell lines

Liver tumours were resected with sterile instruments and washed in sterile PBS prior to digestion. Then tumour tissue was minced and resuspended in a mix of 4 mg/mL collagenase intravenous and dispase (w/v in sterile, serum free Dulbecco’s Modified Eagle Medium (DMEM, Sigma)) at 37°C for 30 min with gentle shaking. The dissociated tumour cells were then washed with complete DMEM (supplemented with 10% (v/v) fetal bovine serum and 1% penicillin/streptomycin) and plated on collagen-coated plates (PurCol, Cell Systems; 0.05 mg/mL). Primary cultures were passaged until visibly free from other contaminating cell types.

Derivation of primary pancreatic tumour cell lines

Pancreatic tumours were dissected with sterile instruments, washed in sterile Hanks balanced salt solution (HBSS) and were minced with a blade until chunks were about 1–2 mm. Tumour tissues were resuspended in 1 mg/mL collagenase V (Sigma) (w/v in sterile, serum free HBSS with Ca²⁺/Mg²⁺) and incubated at 37°C for 30 min with gentle shaking. The dissociated tumour cells were then washed with PBS, resuspended in 0.25% trypsin and incubated at 37°C for 5 min to break up some extracellular matrix. Trypsin was then neutralised with complete DMEM. Before plated on collagen-coated plates, cells were further washed twice with complete medium. Primary cultures were passaged until the GFP positive population >90%.

Cell culture

All cell lines were maintained in complete DMEM at 37°C with 5% carbon dioxide. Liver and pancreatic cancer cells were split twice per week at a ratio of 1:30–40 and 1:5 correspondingly using collagen-coated plates. Myc;shKdm6a and KrasG12D;shKdm6a cells were cultivated in complete DMEM, with Dox (VWR; 1 µg/mL) when applicable.

Virus production

For lentivirus production, HEK293T cells were plated 1 day before transfection into 10 cm plates and transfected when nearly full confluence was reached using a plasmid mix of 2.5 µg pMD.2G, 8 µg psPAX2 (Addgene plasmid #12 259 and # 12260, both were a gift from Didier Trono) and 10 µg pLenti CRISPR v2 harbouring respective guides in 1000 µl serum-free DMEM and 60 µl polyethylenimine (PEI, 1 µg/µl). The plasmid mix was then vortexed for 5 s, incubated at room temperature for 30 min incubation and added drop-wise to cells. Twenty-four hours following the transfection, medium was exchanged and lentiviral supernatant was harvested 48 hours post-transfection using 0.45 µm Cellulose Acetate Membrane filters (VWR) and stored at −80°C until use.

For retrovirus production, HEK-gp-cells were also seeded out as for lentivirus production and transfected using a plasmid mix of 2.5 µg pMD.2G and 20 µg retroviral plasmid such as MLPe vector, pSIN or pMSCV-rtTA3-PGK-Puro in 1000 µl serum-free DMEM and 60 µl PEI (1 µg/µl). The viruses were harvested as lentivirus.

Transduction

Target cells were plated on 10 cm plate and 1 day following the plating, cells were transduced with viral supernatants in the presence of polybrene (4 µg/mL). Two days post-transduction cells were selected with puromycin (2 µg/mL), blasticidin (10 µg/mL) or G418 (neomycin; 1 mg/mL) dependent on the plasmid.

Immunoblotting

Cells were harvested and lysed in cell lysis buffer (Cell Signaling Technology) supplemented with both protease (Complete Mini; Roche) and phosphatase inhibitors. To ensure lysis, cells were sonicated for 5 min in ice and subsequently centrifuged at 4°C at 13 000 rpm to collect protein lysates. Furthermore, protein lysates were equalised using BCA protein assay (Thermo Scientific), equal amount of protein were mixed with Laemmli buffer (100 mM Tris-HCl pH 6.8, 5% glycerol, 2% sodium dodecyl sulfate (SDS), 5% 2-mercaptoethanol) and boiled at 95°C for 5 min. Proteins were separated by SDS-PAGE, transferred onto polyvinylidene fluoride (PVDF) membrane, and detected by immunoblotting using the appropriate antibodies. Image detection was performed with AlphaView software (ProteinSimple) using the Clarity Western ECL substrate Solution (Bio-Rad). The list of antibodies and their sources can be found in online supplemental table 6.

Mutation detection by T7 assay

CRISPR/Cas9-mediated mutations were detected using the T7 Endonuclease I (New England Biolabs). Briefly, genomic DNAs was isolated using Gentra Puregene Tissue Kit (Qiagen) in accordance to manufacturer’s protocol. An approximately 700 bp region surrounding the CRISPR/Cas9-targeted site was amplified using the Q5 Hot Start DNA Polymerase (New England Biolabs), column-purified (Qiagen) and subjected to a series of melting and annealing cycles with the annealing temperature gradually lowered in each successive cycle. T7 Endonuclease I was then added to selectively digest heteroduplex DNA. Digest products were visualised on a 2%–3% agarose gel.

Alternatively, Sanger nucleotide sequencing analysis was performed on PCR products using a T7 primers to detect mutations.

Clonogenic assay

To assess cell survival and proliferation, 500 cells were plated in 6-well plates as triplicate. Both tetracycline inducible-shKdm6a or Kdm6a-cDNA expressing cells were grown in the presence or absence of doxycycline, and cells were fixed with methanol and stained with 0.05% crystal violet after 10 days.

For quantification, depending on confluency of the plates, two methods were used. Either, the amount of crystal violet taken up by the cells were dissolved in solubilisation buffer (50% methanol, 5% acetic acid and 0.1% SDS) and quantified in a spectrophotometer by reading the absorbance at 570 nm or with an ImageJ Plugin.54

Competition assay

To evaluate the effect of Kdm6a overexpression in cell proliferation/viability, Myc;sgKdm6a cells expressing rtTA3 (GFP negative) were mixed with either GFP-positive Kdm6a overexpressing or control cells with 30:70 ratio. The GFP⁺/GFP⁻ ratio were evaluated with Guava easyCyte benchtop flow cytometer (Merck Millipore) over time in the presence or absence of Dox.

Apoptosis assay

Apoptosis was assessed in liver cancer cell lines via CellEvent Caspase-3/7 Green Detection Reagent (Invitrogen) according to manufacturer’s protocol. Briefly, 10 000 cells were grown with and without Dox for 6 days in 6-well plate, trypsinised and resuspended in complete DMEM. About 100 000 cells were incubated with 2 µM final concentration of CellEvent Caspase-3/7 Green Detection Reagent for 45 min at 37°C and subsequently analysed with Guava flow cytometer.

Drug treatment and cell viability assay

For assessment of cell sensitivity towards Torin-1 (Cayman Chemical), cells were plated in 96-well plate 1 day prior to treatment and cell viabilities were assessed with CellTiter-Blue (Promega) 72 hours following the drug treatment. The reagent was added into each well of 96-well plate with 1:10 dilution, incubated for 4 hours at 37°C and fluorescent signal was recorded (560_Ex/590_Em) using FLUOstar Omega (BMG Labtech).

Quantitative reverse transcription PCR

Total RNA was isolated using RNeasy Mini Kit (Qiagen) and RNase-Free DNase Set (Qiagen) in accordance to manufacturer’s protocol. Purified RNA 1 µg was reverse transcribed using TaqMan Reverse Transcription Reagents (Thermo Fisher Scientific) and diluted 1:20 before subjected to qPCR. For the qPCR reaction, cDNA was mixed with Power SYBR Green Master Mix (Thermo Fisher Scientific) and target-specific primers and performed in triplicate. Transcript levels were normalised to the levels of tubulin mRNA expression and calculated using the deltaCt (ΔCt) method. qPCR was carried out using QuantStudio 3 Real-Time PCR system (Applied Biosystems).

For human patient quantification, Gene Expression Assays for human UTX/KDM6A (ID# Hs00253500_m1), Deptor (Hs00961900_m1) and β-Actin (ID # 4333762T) genes were purchased from Applied Biosystems (Foster City, California, USA). Quantitative values for each gene were calculated by using the PE Biosystems Analysis software and expressed as number target (NT). NT=2−ΔCt, wherein ΔCt value of each sample was calculated by subtracting the average Ct value of the target gene from the average Ct value of the β-Actin gene.

RNA sequencing and differential expression analysis

For RNA sequencing, total RNA from three independent tumour-derived cell lines (Myc;sgMll3 and Myc;sgTp53) was isolated using RNeasy Mini Kit (Qiagen), QIAshredder Columns and RNase-Free DNase Set (Qiagen). RNA-seq library construction and sequencing were performed according to protocols used by the integrated genomics operation Core at Memorial Sloan Kettering Cancer Center (MSKCC). From each replicate sample 5–10 million reads were acquired. After removing adaptor sequences with Trimmomatic,55 RNA-seq reads were aligned to GRCh37.75(hg19) with STAR.56 Genome-wide transcript counting was performed by tool for the analysis of high-thoughput sequencing data (HTSeq) to generate a fragments per kilobase per million mapped reads (FPKM) matrix.57 Differentially expressed genes were identified by DESeq2 (V.1.8.2, package in R) and plotted in the volcano plot. The complete data set is available at NCBI Gene Expression Omnibus GSE155630.

CUT&RUN

CUT&RUN was performed as previously described by Skene et al.58 Briefly 250 000 Myc;TREshKdm6a cells grown in the presence or absence of Dox were harvested and immobilised on activated concanavalin A—coated beads at room temperature for 10 min, permeabilised with 0.025% digitonin and incubated with antibody with rotation overnight at 4°C. All the antibodies used in this study were diluted 1:100 for CUT&RUN experiment and are listed in online supplemental table 3. Following the incubation, cells were incubated with the pA-MNase to a final concentration of 700 ng/mL at 4°C for 1 hour, washed and digested on Ca²⁺ addition at 0°C for 30 min. To release the DNA-protein complex, cells were further incubated at 37°C for 10 min and the supernatant was collected. The DNA fragments in the supernatant were then extracted using the spin column (Qiagen) and libraries were prepared using NEBNext Ultra II DNA Library Kit for Illumina (New England Biolabs) following the manufacturer’s protocol. The libraries were sequenced using HiSeq2500 (Illumina) in 25 bp paired-end rapid mode (library concentration of flowcell: 12pM, 1% PhiX spike-in) and HiSeq Rapid SBS Kit v2 (50 cycles; FC-402–402) was used for the HiSeq PE 25 R sequencing type. The complete data set is available at European Nucleotide Archive with accession ID PRJEB39876.

CUT&RUN processing and analysis

Adapter and quality trimming of raw sequencing reads was performed using Trim Galore V.0.4.459 in conjunction with Cutadapt V.1.1460 and the non-default parameters ‘--length_1 35’, and ‘--length_2 35’, ‘--paired’, and ‘--illumina’. Bowtie2 with the ‘--very-sensitive’” flag61 was deployed to separately align trimmed reads to both the Genome Reference Consortium Mouse Build V.38 and the Saccharomyces cerevisiae R64 reference genome. Removal of PCR duplicates was performed by means of Picard MarkDuplicates V.2.17.4.62 Unpaired alignments as well as mappings with a quality below 20 on the Phred scale were filtered out using SAMtools view V.1.5.63 Filtered alignments to the R64 genome were counted. The minimal yeast alignment count observed among all libraries of a specific antibody target was determined. Library-specific scaling factors were calculated by dividing minimal yeast alignment counts by the library-specific count. Coverage tracks were generated by deploying the bamCoverage functionality of deepTools V.3.1.164 with the non-default parameters ‘--effectiveGenomeSize 2652783500’ and ‘--ignoreForNormalization chrM chrY chrX’. Additionally, the scaling factors were included via the ‘--scaleRatio’ option. For peak calling, MACS2 callpeak V.2.1.0.2014061664 was used with an false discovery rate (FDR) cut-off of 0.05 and the parameters ‘--nomodel’, ‘--format BAMPE’, ‘--gsize 2652783500’, ‘--keep-dup all’ and optionally the ‘--broad’ flag for the histone marks H3K4me1 and H3K27me3. The data processing procedure was implemented as a fully containerised pipeline using the Common Workflow Language V.1.065 and is publicly available.66

UTX peaks were associated with the closest TSS using the mouse gene model annotation information from R/Bioconductor packages TxDb.Mmusculus.UCSC.mm10.knownGene (V.3.10.0) and BSgenome.Mmusculus.UCSC.mm10 (V.1.4.0). Association with gene model features was performed using package ChIPseeker (V.1.24.0).67 Coverage at genomic features was summarised using bwtool suite68 and visualised as profile plots and heatmaps using custom R code. For the global analysis of histone marks at TSSs, the read coverage at a 2 kb radius around TSSs obtained from Gencode Mouse Release 23 was summarised using the Genomation package (V.1.12.0).69 Empirical cumulative distribution functions for coverage at TSSs being close (<3 kb) or not close to Kdm6a peaks were estimated and visualised using the stats_ecdf functionality of ggplot2.70

Human patient samples

Seventy-six HCCs and corresponding surrounding non-tumour liver tissues were used for the study. Liver tissues were collected at the Universities of Greifswald (Greifswald, Germany) and Regensburg (Regensburg, Germany).

Pancreatic cancer samples were collected from patients undergoing surgery at the University Hospital Mainz (Mainz, Germany). Samples from 84 patients were included in the study from which both high-grade PanINs and PDAC samples were available.

Human HCC tissue microarray

The HCC tissue microarray used in this study contained 720 representative tissue cores (diameter: 1 mm) distributed on seven slides. In total, 40 histologically normal livers, 174 cirrhosis, 14 dysplastic nodules and 476 HCCs were spotted (87× G1, 311× G2, 76× G3, 2× G4 HCCs). For the evaluation of individual immunohistochemical stains, the KDM6A intensity were evaluated. Staining intensity was scored from 0 to 3; 0=unstained, 1=weakly, 2=moderately and 3=strongly positive.