Patient samples and RNA-seq analysis
Table of Contents
RNA-seq data from 395 patients with PDA was obtained from the University Health Network (Toronto), Sunnybrook Health Sciences Centre (Toronto), Kingston General Hospital (Kingston), McGill University (Montreal), Mayo Clinic (Rochester), Massachusetts General Hospital (Boston) and Sheba Medical Center (Tel Aviv) and has been described previously1,12,13,14. The samples were provided with informed patient consent, and approval for the study was obtained from the Institutional Review or Research Ethics Board of each site. Genetic alterations and clinical data of patients with PDA are provided in Supplementary Table 1.
RNA sequencing was performed as described previously13. In brief, RNA was isolated using PicoPure RNA Isolation Kit, treated with RNase-free DNase, and quantified using Qubit dsRNA High Sensitivity kit. The RNA quality was determined using both RNA Screen Tape Assay and the 2200 TapeStation Nucleic Acid System. RNA libraries were prepared using TruSeq RNA Access. Reads were aligned to the human reference genome (hg38) and transcriptome (Ensembl v84) using STAR v.2.5.3a41.
Principal component analysis
PCA was applied to the RNA-seq data of 395 PDA patient samples to inspect the alternative splicing and gene expression changes with the prcomp function in R. The samples were manually filtered to include only samples that were clustered according to their clinical annotation to reduce background noise. 63% of the samples (136 primary tumour samples and 113 metastatic samples) were utilized for further analysis.
Alternative splicing and gene expression analysis of patient samples
Alternative splicing analysis was performed with PSI-Sigma (version 1.9c)15 on the filtered samples (249 samples). Ensembl gene annotation (version 87) was used as the reference transcriptome. ΔPSI and P value based on exon coordinates were used to identify significant splicing changes between the two clusters. Significant events were identified with imposed cut-offs |ΔPSI| > 10% and nominal P value < 0.05. P values were calculated using PSI-Sigma bioinformatic analysis. Supplementary Table 2. Based on the gene names, the pathway enrichment analysis was conducted using the Reactome database42. Gene sets were limited to between 5 and 500 genes, and pathways were filtered for a statistical threshold of P value < 0.05 using over-representation analysis (hypergeometric distribution) test. Reactome analysis is provided in Supplementary Table 4. For gene expression analysis, reads were mapped to the human genome (hg38) using STAR (version 2.5.3a)41. Ensembl gene annotation (version 87) was used as the reference transcriptome. DESeq2 (version 3.16)43 was used to estimate the significance of differential expression between the sample groups. Overall, gene expression changes were considered to be significant if they passed the false-discovery rate (FDR) threshold of <5%.
Sequence motif enrichment analysis
The XSTREME package44 was used to identify motifs in the 5′ splice site sequences of the cassette exons with significant splicing changes. The 5′ splice site sequence is defined as the 500 bp sequence downstream of the alternative exon (+1 to +500 positions). RBFOX2 motif (GCWUG) was manually added to the motif collection of RNA-binding proteins in the XSTREME database (Ray2013 Homo sapiens). The sequence motif enrichment analysis was performed for the following comparisons: (1) primary tumour samples versus metastatic tumour samples, (2) events identified in the RHO pathways, and (3) the reciprocal splicing changes between RBFOX2 knockout in BxPC3 cells and RBFOX2 overexpression in X50 cells. By default, XSTREME reports 6- to 15-mer motifs whose E-value ≤ 0.05. The program uses Fisher’s exact test or the binomial test to determine the significance of each motif found. Sequence motif enrichment analysis is provided in Supplementary Table 3.
PDA PDX generation in nude mice
Xenograft models were established from tumour samples at Sheba Medical Center, as described previously21,45. X50 PDX cell line was generated from pleural effusion of a patient with stage IV pancreatic adenocarcinoma, X139 PDX cell line was derived from a liver metastasis of patient with stage IV pancreatic adenocarcinoma, and X252 PDX cell line was generated from the primary pancreatic tumour of a patient with stage II pancreatic adenocarcinoma. All samples were obtained with approval of the patients at the Sheba Medical Center. In brief, core needle biopsies and pleural effusion from PDAC tumours were collected and implanted subcutaneously into NOD-SCID mice. Xenografts were propagated and serially passaged into new recipient mice. Tumour chunks were bio-banked in 90% serum + 10% DMSO for future experiments and cryopreserved in liquid nitrogen for DNA, RNA and protein extraction. PDX-derived cells were generated by tissue dissociation and cultured in RPMI-1640 supplemented with 1% l-glutamine and 10% fetal bovine serum (FBS) (Biological Industries). PDA PDX generation in nude mice was performed in accordance with the guidelines of Sheba Medical Center Institutional Animal Care and Use Committee (IACUC) (5539/13).
Cell lines and tissue culture
The BxPC3, HEK293T, Phoenix-AMPHO and HEK293 cells lines were originally obtained from the American Type Culture Collection (ATCC). BxPC3 cell line was grown in RPMI-1640 supplemented with 10% FBS, 1% l-glutamine, and 1% penicillin-streptomycin. HEK293T, Phoenix-AMPHO and HEK293 cell lines were grown in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin.
For immunoblotting, cells were lysed in Laemmli buffer for 5 min at 95 °C, lysates were separated on 10% or 12% SDS–PAGE gels and transferred to PVDF membranes (Invitrogen). The list of antibodies used in this study and the corresponding dilutions are provided in Supplementary Table 12.
mRNA stability assay
For mRNA half-life measurements, BxPC3 and X50 cells were treated with 10 μg ml−1 actinomycin D for 0, 2, 4, 6 and 8 h. RBFOX2 and 18S rRNA mRNA levels were measured by RT–qPCR. A list of RT–qPCR primers used in this study is provided in Supplementary Table 11.
Protein stability assay
For RBFOX2 protein stability assay, BxPC3 and X50 cells were treated with 10 μg ml−1 cycloheximide for 0, 2, 4, 6 and 8 h. RBFOX2 protein levels were detected by Immunoblot analysis. The list of antibodies used in this study are provided in Supplementary Table 12.
CRISPR–Cas9-directed mutations for knockout and splicing modulation
For CRISPR–Cas9 directed knockout, sgRNAs were designed using CHOPCHOP (version 3), a web tool for selecting target sites for CRISPR–Cas946. BxPC3 and X252 cells were transduced with LentiCRISPR v2 vector (Addgene #52961) containing sgRNAs targeting RBFOX2. sgRNA targeting the exon–intron junction (RBFOX2 EIJ sgRNA) was designed manually and confirmed by the CHOPCHOP tool.
For CRISPR–Cas9-directed splicing modulation, sgRNAs were designed using the CHOPCHOP platform to target the 3′ and 5′ end splice sites of the target exon in order to induce skipping of MPRIP, MYL6 and CLSTN1. MPRIP exon 23 inclusion sgRNA was designed to target the RBFOX2 motif downstream of exon 24, which was designed manually and confirmed by the CHOPCHOP tool. Modulation of all three targets simultaneously was achieved using dual sgRNAs, targeting the 3′ splice site target exons of MYL6 and MPRIP, which were cloned into lentiCRISPR v2-Blast (Addgene #52961). This plasmid was then transduced into X50 metastatic cells with CLSTN1 3′ splice site sgRNA47. A list of sgRNAs used in this study is provided in Supplementary Table 10.
Lentiviruses were produced by co-transfection of HEK293T cells with psPax2 (Addgene #12260), pMD2.G (Addgene #12259) and either Flag SFFV2 ΔRRM RBFOX2-puro (Twist Bioscience) or specific sgRNAs LentiCRISPRV2, using FuGENE-HD (Promega E2312) and OptiMEM (Gibco 51985-026). One day after transfection, the medium was replaced, and 48 h after transfection, viruses were collected and filtered through a 0.45 µm membrane. BxPc3, X50, and X252 cells were infected with the viruses with the addition of polybrene (10 μg ml−1, Sigma 107689). Selection with puromycin (2 μg ml−1, Sigma P8833) was initiated 2 days after infection. Immunoblot analysis was performed to confirm overexpression using antibodies against either Flag or RBFOX2. Validation of the CRISPR-induced skipping event was performed by RT–PCR using specific primers for targeted splicing events.
Retroviruses were produced by co-transfection of Phoenix-AMPHOcells with pCL-Eco (Addgene #12371), pMD2.G (Addgene #12259) and Flag pWZL-RBFOX2-hygromycin (received from A. Krainer) using FuGENE-HD (Promega E2312) and OptiMEM (Gibco 51985-026). One day after transfection, the medium was replaced, and 48 h after transfection, viruses were collected and filtered through a 0.45 µm membrane. X50 and X139 cells were infected with the viruses with the addition of polybrene (10 μg ml−1) (Sigma 107689). Selection with hygromycin (50 μg ml−1, Merck 400052) was initiated 2 days after infection. Immunoblotting was performed to confirm overexpression using antibodies against either Flag or RBFOX2.
Six-well plates were coated with a bottom layer containing 2 ml agar mixture (culture media, 20% FBS, 1% agar). After the bottom layer solidified, 1 ml of top agar mixture (culture media, 20% FBS, 0.3% agar) containing the cells (2 × 104 cells per well of BxPC3 or X50 cells in triplicate) was added. After this layer had solidified, 2 ml of media (culture media, 10% FBS) was added. Plates were incubated for 10–21 days, colonies from ten different fields were counted, and the average number of colonies per well was calculated.
For proliferation quantification, 2,000 cells were seeded per well in 96-well plates in 4 replicates, fixed, and stained with methylene blue. Cell density was determined every 24 h (up to 96 h) by absorbance of the methylene blue dye at 655 nm, measured on a plate reader (Bio-Rad) according to the manufacturer’s instructions. For MBQ167 treatment, cells were treated with either 0.05 µM MBQ167 or DMSO for 24 h after seeding.
Quantitative wound healing assay
We used an automated Incucyte wound maker tool to create precise, uniform cell-free zones in cell monolayers enabling real-time, automated measurement of cell migration with a tool to measure the density of the wound region relative to the density of the cell region. Migration was assessed by performing scratch wound healing assays using a real-time cell imaging system (IncuCyte Live cell). One-hundred thousand cells per well were plated onto an IncuCyte 96-Well ImageLock Plate (IncuCyte 4379) (n = 3 independent experiments, each with 8 replicates). Twenty-four hours later, the confluent monolayer of cells was washed in PBS and scratched with the IncuCyte: 96-pin wound-making tool (IncuCyte WoundMaker). Starvation media (basic culture media supplemented with 1% FCS) was applied. Plates were transferred to the IncuCyte live-cell imaging system, and the 96-well wound assay protocol was run on the software. X50, BxPC3, X252, and X139 cells were imaged at 1-h intervals for 24 h to monitor cell migration. For MBQ167 treatment, cells were treated with either 0.05 µM MBQ167 or DMSO for 24 hours. Analysis was performed using the Cell Migration Analysis software (IncuCyte 4400).
Trypan blue exclusion survival assay
BxPC3 cells (1 × 106 cells) were seeded on 6 wells plate in triplicates. The following day cells were treated with either 0.05 µM MBQ167 or DMSO as a vehicle. After 24 h, cells were collected, including the floating dead cell fraction, and resuspended in HBSS. The percentage of dead cells was determined on a Bio-Rad cell counter using 0.4% trypan blue.
RNA-seq analysis of RBFOX2-manipulated cells
mRNA was isolated using oligo-dT purification from three independent biological replicates from the following cell conditions: BxPC3 CRISPR control, BxPC3 with two different RBFOX2 sgRNAs, X50 empty vector and X50 with RBFOX2 overexpression.
RNA-seq experiments were performed using NextSeq 2000 system (Illumina). Libraries were prepared using TruSeq RNA Sample preparation kit. More than 100 × 106 reads of 150 bp from each side (paired-end) per sample were generated. Reads were mapped to the human genome (hg38) using STAR (version 2.5.3a) with two-pass mode41. Alternative splicing analysis was performed using PSI-Sigma (version 1.9c)15. Ensembl gene annotation (version 87) was used as the reference transcriptome.
To identify significant reciprocal splicing changes, we compared the overlap between two comparisons: (1) RBFOX2 OE (versus pWZL(−)) and RBFOX2 sgRNA-1 (versus CRISPR control), 217 significant events, and (2) RBFOX2 OE (versus pWZL(−)) and RBFOX2 sgRNA-2 (versus CRISPR control), 222 significant events. We compared ΔPSI and P values based on exon coordinates. An exon must fit three criteria to be considered as ‘oppositely spliced’: (1) the exon has |ΔPSI| > 10% and nominal P value < 0.05 in either KO or OE condition, (2) ΔPSI direction is the opposite between KO and OE conditions, and (3) |ΔPSI| > 5% in both KO and OE conditions. P values were calculated using PSI-Sigma bioinformatic analysis. A list of differentially spliced events in RBFOX2 manipulated cells is provided in Supplementary Table 6. Pathway enrichment analysis was conducted by using Reactome database42 based on the gene names of opposite splicing changes between KO and OE conditions. Gene sets were limited to between 5 and 500 genes, and pathways were filtered for a statistical threshold of P < 0.05 using over-representation analysis (hypergeometric distribution) test. Reactome analysis is provided in Supplementary Table 4.
Total RNA was isolated using TRI Reagent (Sigma T9424). For cDNA synthesis, 1 µg of total RNA was reverse transcribed to cDNA with iScript cDNA Synthesis kit (Bio-Rad 1708891). RT–PCR was conducted on 1 µl of cDNA using PCRBIO HS Taq Mix Red kit (BIOSYSTEMS PB10.23.02) to confirm splicing and splicing modulation by CRISPR–Cas9. PCR conditions were as described in the manufacturer’s protocol. PCR products were separated either on a 2% agarose gel or using the LabChip GX microfluidics platform. The list of primers used in this study are provided in Supplementary Table 11.
Total RNA was extracted with TRI Reagent (Sigma), and 1 μg of total RNA was reverse transcribed using iScript cDNA Synthesis kit (Bio-Rad 1708891). qPCR was performed on the cDNA using SYBR green (Applied Biosystems) and the Step one plus real-time PCR system (Applied Biosystems). Normalization was performed using 18S rRNA primers. Primers are listed in Supplementary Table 11.
BxPC3 cells (1 × 104 cells) were seeded in μ-Slide 8 well-ibiTreat, tissue culture-treated (IBIDI 80826) and incubated overnight. Cells were fixed with 2 % paraformaldehyde in PBS for 20 min, then permeabilized with 0.1% Triton X-100 for 20 min and blocked with BSA and 1% FBS for 10 min. Cells were stained with Paxillin antibody for 2 h (1:200 BD Transduction Laboratories 612405), then incubated with a secondary antibody for 1 h (1:100 Goat anti-Mouse Alexa Fluor 488 A-11029). Antifade Mounting Medium with DAPI (Vector laboratory H-1200) was used to stain the nuclei and mount the samples. Images were acquired using a spinning disk confocal microscope (Nikon) with 60× and 100× objectives. The fields of view were randomly chosen. Image analysis was performed using NIS Elements version 4.13 imaging software.
Metastatic and tumour formation in vivo experiments were performed in accordance with the guidelines of IACUC at the Hebrew University (MD-15-14634-5). The study is in compliance with all the relevant ethical regulations. NOD-SCID mice (Jackson Laboratories, 0001303) were ordered at 6 weeks of age. The mice were housed under standard laboratory conditions in specific-pathogen-free cages in an animal room at constant temperature (19–23 °C) and regulated humidity under a 12 h:12 h light-dark cycle and received standard laboratory chow and water ad libitum. All mice entered the experiments at 8–12 weeks of age. Both male and female mice were used for the experiments.
In vivo metastasis model
X50 mCherry or GFP-labelled or BxPC3 GFP-labelled cells (1 × 106 cells) in PBS were injected intravenously into NOD-SCID mice. One month after injection, the lungs were removed and analysed for metastases. For azathioprine and MBQ167 treatments, mice were treated with either 10 mg kg−1 of azathioprine or 3 mg kg−1 of MBQ167 starting 1 week after intravenous injection of cells. Azathioprine and MBQ167 were injected intraperitoneally every 3 days for a total of 15 doses. The mice were closely monitored on a daily basis for any signs of disease. Mice were killed at 30 days post-injection (endpoint), or earlier if they failed to thrive, experienced a weight loss of greater than 10% of their total body weight, or showed any signs of infection, as per our IACUC. These limits were not exceeded in any of our experiments. Images were obtained using a fluorescent binocular microscope. Quantification of lung metastases was performed using NIS Elements version 4.13 imaging software.
Tumour formation in mice
BxPC3 and X50 cells (1 × 106 cells) in PBS were injected subcutaneously into NOD-SCID mice. For azathioprine and MBQ167 treatments, mice were treated either with 10 mg kg−1 of azathioprine or 3 mg kg−1 of MBQ167 starting 1 week after intravenous injection of cells. Azathioprine and MBQ167 were injected intraperitoneally every 3 days for a total of 15 doses. The mice were monitored for tumour volume and weight. All mice were sacrificed approximately 6 weeks after injection and before their tumour volume reached the maximum allowed limit of 2,000 mm3 as permitted by our IACUC guideline. No experiments exceeded this limit.
Tissues were fixed in 4% formaldehyde for 16 h. Tissues were embedded in paraffin, and 5 µm sections were cut and mounted on slides. Slides were rehydrated through a series of xylene and ethanol washes, and antigen retrieval was performed in a 10 mM citrate buffer in a pressure cooker. Tissues were blocked in 2.5% normal horse serum blocking solution (Vector Laboratories S-2012) and subjected to staining with anti-GFP antibody (abcam ab6673). ImmPRESS HRP anti-goat IgG polymer (Vector Laboratories MP-7401) was used as a secondary antibody. ImmPACT DAB peroxidase substrate (Vector Laboratories SK-4105) was used as a substrate. Hematoxylin (Vector Laboratories H-3404) was used as a counterstain. Slides were imaged using Aperio Digital Pathology Slide Scanners (Leica Biosystems).
Serine/threonine kinase predictions
The kinase predictions were based on experimental biochemical data of their substrate motifs. Synthetic peptide libraries, containing 198 peptide mixtures, that explored amino acid preference up to 5 residues N-terminal and C-terminal to the phosphorylated Ser/Thr to determine the optimal substrate sequence specificity for recombinant Ser/Thr kinases were utilized. In total, 303 kinases were profiled. Their motifs were quantified into position-specific scoring matrices (PSSMs) and then applied computationally to score phosphorylation sites based on their surrounding amino acid sequences. These PSSMs were ranked against each site to identify the most favourable kinases48. Serine/threonine kinome analysis is provided in Supplementary Table 8.
Structural illustrations were generated with PYMOL, using predicted models of MPRIP on alphaFold: AF-Q6WCQ1-F1-model_v3_1 (Exon 23 included) and AF-B9EGI2-F1-model_v3 (Exon 23 excluded)49.
HEK293 cells were transfected with 24 µg plasmid DNA per 100 mm plate of either WPRE(−) empty plasmid or Flag-MPRIP exon 23-skipped isoform, or Flag-MPRIP exon 23-included isoform plasmids. Cells were lysed 48 h later in CHAPS buffer. Protein concentrations were normalized, and the samples were brought up to the same volume in CHAPS lysis buffer. A total of 50 μl of total lysate was set aside for immunoblotting. The remaining lysate was incubated overnight with 1 μg of anti-Flag M2 affinity gel (Sigma, A2220). The resin was washed 4 times with wash buffer, incubated with 50 µl of 2× Laemmli buffer, heated at 95 °C for 5 min, and separated by SDS–PAGE.
Mass spectrometry analysis
The immunoprecipitation beads were resuspended in 100 µl of 8 M urea, 10 mM DTT, 25 mM Tris-HCl pH 8.0 and incubated for 30 min, followed by addition of iodoacetamide to a concentration of 55 mm and incubated for 30 min in the dark. The urea was diluted by the addition of 7 volumes of 25 mM Tris-HCl pH 8.0. Trypsin (0.4 µg) was added (Promega), and the beads were incubated overnight at 37 °C with gentle agitation. The released peptides were desalted by loading the whole bead supernatant on C18 stage tips50.
Mass spectrometry analysis was performed using a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific) coupled online to a nanoflow UHPLC instrument (Ultimate 3000 Dionex, Thermo Fisher Scientific). Eluted peptides were separated over a 120-min gradient run at a flow rate of 0.15–0.3 μl min−1 on a reverse phase 25-cm-long C18 column (75 μm internal diameter, 2 μm, 100 Å, Thermo PepMap RSLC, from Thermo Fisher Scientific). The survey scans (380–2,000 m/z, target value 3 × 106 charges, maximum ion injection times 50 ms) were acquired and followed by higher energy collisional dissociation-based fragmentation (normalized collision energy 25). A resolution of 70,000 was used for survey scans, and up to 15 dynamically chosen most abundant precursor ions were fragmented (isolation window 1.6 m/z). The MS/MS scans were acquired at a resolution of 17,500 (target value 105 charges, maximum ion injection times 120 ms).
Mass spectrometry data analysis
Mass spectra data were processed using the MaxQuant computational platform, version 2.0.3.051. Peak lists were searched against the Homo sapiens Uniprot FASTA sequence database UP000005640 appended with MPRIP isoforms. The search included cysteine carbamidomethylation as a fixed modification and oxidation of methionine as variable modifications. Match between runs option was selected. Peptides with a minimum of seven amino acid lengths were considered, and the required FDR was set to 1% at the peptide and protein levels. Relative protein quantification in MaxQuant was performed using the label-free quantification (LFQ) algorithm51. MaxLFQ allows accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction52. Statistical analysis was performed using the Perseus statistical package, Perseus computational platform for comprehensive analysis of proteomics data53. Only those proteins for which at least two valid LFQ values were obtained in at least one sample group were accepted for statistical analysis. After application of this filter, a random value was substituted for proteins for which LFQ could not be determined (‘imputation’ function of Perseus) using default parameters. Volcano plot and t-tests were obtained using the following parameters in the volcano plot Perseus function: Randomization: 250, FDR: 0.05, S0: 0.1. Mass spectrometry analysis is provided in Supplementary Table 9.
Tables and graphs for statistical analysis were created using GraphPad Prism 9 (GraphPad Software). P values < 0.05 were considered significant. Statistical significance between two groups was determined by one- or two-tailed Student’s t-test, and for experiments with more than two groups was determined by one- or two-way ANOVA, details of statistical analyses are indicated in figure legends. All the data in the graphs are shown as mean ± s.d. unless stated otherwise. All experiments were performed a minimum of three independent times with similar results, each containing at least three technical replicates (wells) for each condition.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.