Direct in vivo CRISPR gene editing in the mouse pancreas

To functionally test putative PDAC cancer genes in vivo, we employed a multiplexed CRISPR/Cas9 genome editing approach to generate knock-out clones directly in the pancreatic epithelium of tumor-prone mice. We used conditional Lox-Stop-Lox-(LSL)-KrasG12D and LSL-Cas9-GFP mice crossed to the pancreas-specific PDX1-Cre driver line (termed KC mice) and injected an adeno-associated virus that expresses a sgRNA and the H2B-RFP fluorescent marker (AAV-sgRNA-RFP) (Fig.1a). Cre-mediated excision of Lox-Stop-Lox cassettes resulted in expression of oncogenic KrasG12D, Cas9 and GFP and formation of hundreds of cytokeratin19 positive (CK19) pancreatic intraepithelial neoplasia (PanIN) precursor lesions, which can be lineage-traced by virtue of red fluorescence (Supplementary Fig.1a, b). To validate the efficiency of CRISPR/Cas9-mediated mutagenesis, we injected sgRNAs targeting GFP, which revealed a knock-out efficacy of 786% (Supplementary Fig.1c).

A Experimental design of the in vivo PDAC CRISPR screen, showing gene selection from long-tail mutations, pancreatic injection of AAV libraries and tumor sequencing. B Tumor-free survival of Pdx1-Cre;LSL-KrasG12D;LSL-Cas9-GFP mice transduced with a sgRNA library targeting putative pancreatic cancer genes (n=23) or a control sgRNA library (n=13) C Representative whole-mount, H&E and immunofluorescent images of an H2B-RFP+ pancreatic PDAC-library tumor: Scale bar 2mm. H&E image: scale bar 250m. Representative immunofluorescence image shows H2B-RFP and CK19 expression. Scale bar 50m. Similar results were observed in all collected tumor. D Representative pie charts showing tumor suppressor genes with enriched sgRNAs in tumor DNA obtained from three different pancreatic tumors and a control-transduced pancreas with multifocal PanINs. E Bar graph showing putative tumor suppressor genes with enriched sgRNAs in tumor DNA obtained from the PDAC mouse model (sgRNA enriched per tumors are indicated by color).

KC mice exhibited rapid growth of pre-invasive PanINs precursor lesions but showed very slow progression to invasive PDAC with a median latency of 14 months (Fig.1b). Additional genetic alterations such as loss of transformation-related protein 53 (Trp53), p16Ink4a, Lkb1 or inactivation of TGF- signaling was previously shown to cooperate with KrasG12D and induces rapid PDAC development within 3-5 month16,17,18,19,20. To test whether our direct in vivo CRISPR approach can reveal genetic interactions, we recapitulated cooperation between oncogenic KrasG12D and loss of p53 (Trp53). Indeed, Cas9-mediated ablation of Trp53 in KC mice triggered rapid PDAC formation with a median latency of 14 weeks, while littermates transduced with scrambled control sgRNAs remained cancer-free for over 1 year (Supplementary Fig.1d). This is in line with previous efforts using CRISPR/Cas9 gene editing in KRasG12D mice21,22 and demonstrates that this approach can be used to test for genetic cooperation between PDAC genes.

In pancreatic cancer, 125 genes show recurrent somatic mutations6,7. To assess these genes in vivo, we established a sgRNA library targeting the corresponding mouse orthologs (4 sgRNAs/gene; 500 sgRNAs) and a library of 420 non-targeting control sgRNAs (Supplementary Data1). Of note, we did not include sgRNAs targeting well-established PDAC driver genes such as Trp53 or Smad416,17,18,19,20.

Next, we optimized the parameters for an in vivo CRISPR screen. Using a mixture of AAV-GFP and AAV-RFP, we determined the viral titer that transduces the pancreatic epithelium at clonal density (MOI<1). Higher viral titers were associated with double infections, whereas a 15% overall transduction level minimized double infections while generating necessary clones sufficient to screen (Supplementary Fig.1e). Using multicolor Rosa26-Lox-Stop-Lox(R26-LSL)-Confetti Cre-reporter mice, we next determined the viral titer required to generate thousands of discrete clones within the pancreatic epithelium (Supplementary Fig.1f). Thus, at a transduction level of 15% and a pool of 500 sgRNAs, each sgRNA would be introduced into at least 50 CK19+ epithelial cells within a single pancreas.

To uncover long-tail genes that cooperate with oncogenic KRasG12D and accelerate PDAC development, we injected the experimental and the control AAV-sgRNA libraries into the pancreas of 23 and 13 KC mice, respectively. Next-generation sequencing confirmed efficient AAV transduction of all sgRNAs (Supplementary Fig.2a). Importantly, KC mice transduced with the long-tail PDAC sgRNA library developed pancreatic cancer significantly faster than littermates transduced with the control sgRNA library (31 versus 59 weeks; p<0.0001) (Fig.1b and c). In addition, 13/23 (56%) KC mice transduced with the long-tail PDAC sgRNA library developed liver and/or lung metastasis, while only 1/13 (~8%) littermate mice transduced with the control sgRNA library developed metastasis (Supplementary Fig.2b-e), indicating the existence of strong tumor suppressors within the long-tail of PDAC associated genes.

To identify these PDAC driver genes, we examined the sgRNA representation in 151 tumors. 78% of tumors showed strong enrichment for single sgRNAs, indicating a clonal origin. In contrast, the pancreas of control-transduced mice with multifocal PanINs showed enrichment of several non-template control sgRNAs (Fig.1d). We prioritized genes that were targeted by 2 sgRNAs and knocked out in multiple tumors and/or metastatic foci, resulting in 8 candidate tumor suppressor genes (Fig.1e, Supplementary Data2). These candidates included well-known PDAC tumor suppressor genes, such as Cdkn2a23, Rnf4324, Fbxw725 or NF226, as well as genes with poorly understood function, such as Usp15 and Scaf1.

Pancreatitis is a risk factor for the development of PDAC in humans and cooperates with oncogenic KRas mutations to induce PDAC formation in mice23,27. Therefore, we repeated our screen and treated mice with chronic, low doses of cerulein to induce mild pancreatitis. As expected, cerulein treatment significantly accelerated PDAC development in KC mice transduced with the PDAC sgRNA library (17 versus 32 weeks median survival, p<0.0001), and a trend towards faster PDAC development in KC mice transduced with the control library (Supplementary Fig.2f). In line with the previous screen, Cdkn2a was the top-scoring gene followed by Rnf43 and the newly identified genes, Usp15 and Scaf1 (Supplementary Fig.2g), further supporting their function as strong suppressors of pancreatic cancer in KC mice.

The multi-domain deubiquitinase USP15 regulates diverse processes, such as the p53 tumor suppressor pathway28, MAPK signaling29, Wnt/beta-catenin signaling30, TGF- signaling31,32,33, NfKb signaling32,34,35 and chromosome integrity36,37, either through regulated de-ubiquitination of direct substrates such as MDM2, APC, SMADs or TGF- receptors, or de-ubiquitination-independent functions such as through protein-protein interactions38.

To validate the tumor suppressive function of Usp15, we first injected KC mice individually with one library or one newly designed sgRNA. All transduced mice developed highly proliferative pancreatic tumors with much shorter latencies compared to mice transduced with the non-targeting control sgRNAs (sgCrtl) (Fig.2a). In fact, age-matched control KC mice only exhibited PanINs at the time when USP15 knockout mice exhibit aggressive PDACs (Fig.2b). All tested tumors exhibited efficient CRISRP/Cas9-mediated mutagenesis of Usp15 (Supplementary Fig.3a and b).

A Tumor-free survival of Pdx1-Cre;LSL-KrasG12D;LSL-Cas9-GFP mice injected with CRISPR AAV-sgRNAs targeting the indicated gene or non-targeting control sgRNA (sgCtrl, n=6). Two independent sgRNAs were used (sgUsp15_1 n=9, sgUsp15_2 n=9). Log-Rank test (Mantel-Cox). B Representative H&E images showing multifocal PanINs in sgCtrl transduced pancreas and PADC tumors in sgUsp15 transduced pancreas. Scale bar 100m. C Tumor-free survival of Pdx1-Cre;LSL-KrasG12D mice with the indicated Usp15 genotype where + indicates the wildtype allele and indicates a conditionally deleted allele. KrasG12D +/- Usp15 +/+ (n=12); KrasG12D +/- Usp15 +/- (n=11); KrasG12D +/- Usp15 -/- (n=13). Log-Rank test (Mantel-Cox). D Representative H&E images of mice with the indicated genotype showing multifocal PanINs and PADC tumors. Scale bar 100m. E Cell proliferation curves of KC cells transduced with the indicated sgRNA obtained using the IncuCyte live-cell imaging. Cells were grown for five days and data are expressed as cell confluence percentage (%; meanSD, n=3 independent experiments, Two-way ANOVA (sgUsp15_1 p=3.43e-8: sgUsp15_2 p=4.89e-9), Dunnetts multiple comparison. F Cell proliferation curves of KC cells expressing ubiquitin variants inhibiting Usp15 (Ubv15.1a and Ubv15.1/d) or wildtype ubiquitin (Ubwt) as control. (%; meanSD, n=3 independent experiments, Two-way ANOVA (Ubv15.1a p=2.74e-6: Ubv15.1/d p=2.06e-7), Dunnetts multiple comparison. G Tumor-free survival of NSG (NOD Scid Gamma) mice after orthotopic injection sgCtrl (n=5) or sgUsp15_1/2 (n=5; n=5) KC cells. Two independent sgRNAs were used. Log-Rank test (Mantel-Cox). H Dose-response curves for KPC sgCtrl or sgUsp15 cells treated with the indicated concentration of Olaparib (meanSD, n=3 independent experiments). Two-way ANOVA (sgUsp15_1 p=0.0349: sgUsp15_2 p=0.0431), Dunnetts multiple comparison.

To further confirm the tumor suppressive role and rule out any confounding effect of Cas9 endonuclease expression, we generated conditional Usp15fl/fl; KRasG12D; Pdx1-Cre. This conventional knock-out approach recapitulated our CRISPR/Cas9 findings (Fig.2c and d), validating our in vivo CRISPR approach. Interestingly, Usp15fl/+ heterozygous mice also manifested significantly shorter disease-free survival (Fig.2c). To assess whether tumor development was due to Usp15 loss of heterozygosity, we used fluorescence-activated cell sorting (FACS) to isolate tumor cells from Usp15 homozygous, heterozygous and wild-type KRasG12D tumors. Western Blot analysis revealed Usp15 expression in Usp15 heterozygous tumor cells, albeit at a reduced level compared to control tumors (Supplementary Fig.3c), indicating Usp15 functions as a haploinsufficient tumor suppressor.

Next, we established primary PDAC cell lines from KC mice as well as KC mice with concomitant expression of the hotspot p53R270H mutant (KPC) and used CRISPR/Cas9 to knock-out Usp15 (Supplementary Fig.3d). Loss of Usp15 significantly increased proliferation of these KC cells (Fig.2e), while it did not affect KPC cells (Supplementary Fig.3e), presumably, because those cells are at the maximal proliferation rate. Similar results were obtained using ubiquitin variants (UbVs) that bind and block the catalytic domain of Usp1538, indicating that this tumor suppressive function is de-ubiquitination dependent (Fig.2f). Upon orthotopic injection, Usp15 knock-out KC cells also formed allograft tumor faster than non-targeting control cells (Fig.2g). Together, these data show that Usp15 regulates tumor cell proliferation in a cell-autonomous manner and loss of Usp15 increases a cells ability to form allograft tumors.

Consistent with a previous report36, we also found that loss of Usp15 sensitizes pancreatic cancer cells to Poly-(ADP-ribose) polymerase inhibition (PARPi) by Olaparib. This increased drug sensitivity was stronger in KPC cells than KC cells and was also seen in response to Gemcitabine, one of the most commonly used chemotherapies to treat pancreatic cancer (Fig.2h and Supplementary Fig3f, Fig.4a and b). KC cells were overall more sensitive to Gemcitabine likely due to the intact p53 response (Supplementary Fig3g). Importantly, loss of USP15 also sensitized allograft tumors in vivo towards Olaparib treatment (Supplementary Fig.4c). In addition, we found that Olaparib and Gemcitabine treatment significantly increases expression of Usp15 in KC and KPC cells (Supplementary Fig.4d). In line with its haploinsufficient tumorigenic effect, heterozygous loss of Usp15 also significantly increased proliferation and sensitized to Olaparib treatment, but not as pronounced as complete Usp15 loss (Supplementary Fig.4e and f). As such, Usp15 appears to function as a double-edged sword in pancreatic cancer, where the loss of Usp15 enhances tumor progression in the initial stages of tumorigenesis but sensitizes to certain treatment regimens in the later stages.

Given the wide range of USP15 substrates and USP15-regulated pathways with well-known functions in cancer, we set out to elucidate USP15s exact role in PDAC suppression. First, we transcriptionally profiled primary KC cells transduced with sgRNAs targeting Usp15 or non-template controls sgRNAs. Inactivation of Usp15 resulted in dramatic changes in gene expression compared to scrambled control KrasG12D tumor cells (794 differentially expressed genes (DEG), false discovery rate (FDR, Benjamini-Hochberg)<0.05 and absolute log2 fold-change > 1, Fig.3a and Supplementary Data3). Gene set enrichment analyses (GSEA) revealed significantly upregulated gene sets associated with xenobiotic detoxification, glutathione metabolism, anabolic processes, and oxidative phosphorylation (Fig.3b and Supplementary Data3). These findings are in line with USP15s known role in negatively regulating NRF239 (encoded by the NFE2L2 gene), the master regulator of glutathione metabolism and the redox balance of a cell. In addition, NRF2 expression is induced by oncogenic KRAS and known to stimulate proliferation and suppress senescence of PDAC cells40. Indeed, Usp15 knock-out cells exhibited significantly increased levels of Nrf2 (Supplementary Fig.5a).

A Volcano Blot showing differential expressed genes between Usp15-knockout compared to sgCtrl control KC cells. Wald test and Benjamini-Hochberg (BH)-adjusted P-value. Two independent sgRNAs, two biological duplicates. B Bar graph showing gene set enrichment analysis (GSEA) of Usp15-knockout compared to sgCtrl control KC cells. GSEA nominal p-values. Two independent sgRNAs, two biological duplicates. C GSEA plots and Heatmaps of log2 counts per million for selected differentially expressed pathways and genes in sgUsp15 versus sgCtrl control KC cells. GSEA nominal p-values. Two independent sgRNAs, two biological duplicates. D Expression levels of genes related to TNF signaling evaluated by RT-qPCR. Results were normalized with Gapdh and are expressed in fold change compared to Ctrl (meanSEM, n=3 independent experiments). Cells were incubated with 10ng/mL TNF-for 30min. Two-sided T-test, Rel-B p=0.043; TRAF-1 p=0.037/p=0.034; NFKB1 p=0.036; Rel-B p=0.042; TRAF-1 p=0.039; CXCL2 p=0.028; CXCL3 p=0.047/p=0.043; NFKB1 p=0.038/p=0.040; NFKB2 p=0.039/p=0.043.

GSEA also revealed depleted genes sets associated with inflammatory responses, TNF, TGF and p53 signaling (Fig.3b-d and Supplementary Fig.5b), all pathways with well-known tumor suppressive function in PDAC development17,41. Quantitative RT-PCR confirmed reduced expression of TNF and TGF responsive genes at baseline as well as TNF/TGF-stimulated conditions (Fig.3d and Supplementary Fig.5c). In addition, loss of USP15 reduced TNFinduced cell death and TGF-induced migration (Supplementary Fig.5d and e). Together, these data indicate that Usp15 functions as a strong haploinsufficient PDAC tumor suppressor potentially by regulating tumor suppressive cytokine signaling pathway.

Our second new hit, SCAF1 (SR-Related CTD Associated Factor 1), is a member of the human SR (Ser/Arg-rich) superfamily of pre-mRNA splicing factors. It interacts with the CTD domain of the RNA polymerase II (RNAPII) and is thought to be involved in pre-mRNA splicing42. Its close homologs SCAF4 and SCAF8 were recently shown to be essential for correct polyA site selection and RNAPII transcriptional termination in human cells43. SCAF1 was also one of the top-scoring hits in a screen for genes that can restore homologous recombination in BRCA1-deficient cells and thus conferred resistance to PARP inhibition44. However, the molecular function of SCAF1 remains completely elusive.

First, we validated the tumor suppressive function of Scaf1 by injecting KC mice individually with one library or one newly designed sgRNA. All transduced mice developed highly proliferative pancreatic cancer with much shorter latencies compared to mice transduced with the non-targeting control sgRNAs (Fig.4a, b). Of note, both Scaf1 sgRNAs induced high CRISPR/Cas9-mediated mutagenesis and resulted in significantly reduced Scaf1 mRNA expression (Supplementary Fig.6a, b). Similar to Usp15 knockout cells, we also found that primary Scaf1 knockout KC cells exhibited increased proliferation in culture and formed tumors faster when injected orthotopically into mice compared to scrambled control KC cells (Fig.4c and d). Scaf1 knockout cells also exhibited significantly increased sensitivity to Olaparib in vitro and in vivo (Fig.4e and Supplementary Fig.6c and d), again phenocopying Usp15 knockout cells.

A Tumor-free survival of Pdx1-Cre;LSL-KrasG12D;LSL-Cas9-GFP mice injected with CRISPR AAV targeting the indicated gene or non-targeting control sgRNA (sgCtrl n=6). Two independent sgRNAs were used (sgScaf1_1 n=8, sgScaf1_2 n=7). Log-Rank test (Mantel-Cox). B Representative H&E images showing multifocal PanINs in sgCtrl-transduced pancreas and PADC tumors in sgScaf1-transduced pancreas. Scale bar 100m. C Cell proliferation curves of KC sgCtrl and sgScaf1 cells were obtained using the IncuCyte live-cell imaging and data are expressed as cell confluence percentage (%; meanSD, n=3 independent experiments). Two-way ANOVA (sgScaf1_1 p=0.0039: sgScaf1_2 p=0.00042), Dunnetts multiple comparison. D Tumor-free survival of NSG mice orthotopically injected with sgCtrl (n=5) or sgScaf1_1/2 (n=5; n=5) KC cells. Two independent sgRNAs were used. Log-Rank test (Mantel-Cox). E Dose-response curves for KPC sgCtrl or sgScaf1 cells treated with the indicated concentration of Olaparib (meanSD, n=3 independent experiments). Two-way ANOVA (sgScaf1_1 p=0.0084: sgScaf1_2 p=0.0028), Dunnetts multiple comparisons. F Representative Western Blot of Usp15 in KC cells transduced with the indicated sgRNAs and treated as indicated. This experiment was repeated independently two times with similar results. G Cell growth curves of KC sgCtrl and sgScaf1cells expressing the listed isoform of USP15 or an empty vector (EV). Data are expressed as cell confluence percentage (%; meanSD, n=3 independent experiments); two-way ANOVA (sgScaf1+EV p=0.0342), Dunnetts multiple comparison. Dose-response curves for KC sgCtrl and sgScaf1 cells expressing the listed isoforms of USP15 or EV and treated with Olaparib. (%; meanSD, n=3 independent experiments; two-way ANOVA (sgScaf1+EV p=0.00129), Sidak multiple comparison H Tumor-free survival of NSG mice orthotopically injected with sgCtrl (n=5) or sgScaf1 KC cells expressing the listed isoforms of USP15 (n=5: n=5) or EV (n=5; n=5). Log-Rank test (Mantel-Cox).

Interestingly, we found a connection between Scaf1 and Usp15. Scaf1 knockout cells exhibited reduced expression of full-length Usp15 (molecular weight of ~125kDa) and showed expression of a 25kDa short Usp15 isoform under homeostatic as well as Olaparib and gemcitabine treatment (Fig.4f and Supplementary Fig.6e and f). SCAF1 KO tumors also exhibited lower levels of full-length USP15 (Supplementary Data Fig.6g). To examine a potential function of this truncated isoform, we cloned and transduced the long and the short isoforms into primary Usp15 knock-out KC cells (Supplementary Fig.6h). While full-length Usp15 was able to supress the hyperproliferative phenotype of Usp15 knock-out cells, the short isoform failed to suppress the cell proliferation (Supplementary Fig.7a). Similarly, re-expressing the full-length but not the short Usp15 isoform reversed the sensitivity of Usp15 knock-out KC cells to Olaparib and gemcitabine (Supplementary Fig.7b). In addition, overexpression of the full-length or the short Usp15 isoform did not alter proliferation of wildtype KC cells (Supplementary Fig.7c), indicating that the short isoform does not exhibit dominant negative functions. However, expression of the long but not the short Usp15 isoform or a catalytically-dead USP15 isoform suppressed the hyperproliferation and Olaparib sensitivity as well as the increased in vivo tumorigenesis of Scaf1 knock-out cells (Fig.4g and h and Supplementary Fig.7c-f). Together these data indicate that the short isoform has no tumor suppressive functions or alters the response to PARP inhibition and that Scaf1s tumor suppressive function is at least in part routed by regulating the expression of full-length Usp15.

To further elucidate the effects of Scaf1, we transcriptionally profiled Scaf1 knockout KC cells. Inactivation of Scaf1 resulted in 625 differentially expressed genes (DEG) (false discovery rate (FDR)<0.05 and absolute log2 fold-change > 1, Fig.5a and Supplementary Data3) compared to scrambled control KrasG12D tumor cells. GSEA revealed significantly upregulated gene sets associated with nucleotide metabolism, glutathione metabolism, microtubule polymerization, and oxidative phosphorylation as well as downregulation gene sets associated with TNF signaling, one-carbon metabolism, xenobiotic catabolic processes, mTorc1/mTOR signaling, hypoxia and p53 signaling (Fig.5b and Supplementary Fig.7g). In addition, we found a trend towards downregulated TGF signaling (Fig.5b and Supplementary Data3), reminiscent of the pathways altered in Usp15 knock-out cells.

A Volcano Blot showing differential expressed genes between Scaf1-knockout compared to control KC cells. Wald test and Benjamini-Hochberg (BH)-adjusted P-value. Two independent sgRNAs, two biological duplicates. B Bar graph showing gene set enrichment analysis of Scaf1-knockout compared to control KC cells. GSEA nominal p-values. Two independent sgRNAs, two biological duplicates. C Bar graph showing gene set enrichment analysis of Usp15-knockout and Scaf1-knockout compared to sgCtrl control KC cells treated with Olaparib (1M). GSEA nominal p-values. Two independent sgRNAs, two biological duplicates. D Expression levels of genes related to HH signaling evaluated by RT-qPCR in the indicated KC cell lines. Results were normalized with Gapdh and are expressed in fold change to CTRL (meanSEM, n=3 independent experiments). Cells were treated with 100nM Smoothened Agonist (SAG) and 1M Olaparib. Two-sided T-test, for sgUSP15: NRP2 p=0.021/p=0.018; PTCH1 p=0.037/p=0.033; GLI1 p=0.033; NRP2 p=0.042/p=0.037; for sgScaf1: PTCH1 p=0.046; PTCH2 p=0.040/p=0.042/p=0.033; GLI1 p=0.037/p=0.031; NRP2 p=0.042.

Lastly, we set out to elucidate how Usp15 and Scaf1 regulate the response of pancreatic cancer cells to PARP inhibition. Interestingly, transcriptional profiling and GSEA following Olaparib treatment revealed that both Usp15 and Scaf1 knock-out cells, exhibited downregulation of hedgehog signaling, TGF signaling and axon guidance by netrin as well as upregulation of glycolysis as the top dysregulated pathways compared to Olaparib-treated control KC cells (Fig.5c and Supplementary Data4). Together, this indicates a common mechanism leading to increased sensitivity to PARP inhibition shared between Usp15 and Scaf1 knock-out cells. Indeed, quantitative RT-PCR confirmed reduced expression of hedgehog target genes at baseline as well as upon sonic hedgehog stimulation (Fig.5d). Thus, Scaf1 and Usp15 knockout cells share several alterations such as upregulated TNF signaling and downregulated TGF, hedgehog and p53 signaling but also several distinct pathways.

To extend our findings from mouse to human cancers, we analyzed 295 PDAC samples from The Cancer Genome Atlas45,46,47. Mutations and homozygous deletion of USP15 and Scaf1 are rare as expected for long-tail mutation and were found in only 2.4% and 1.4% of PDAC samples, respectively. However, an additional 25% and 13% of PDAC cases showed shallow deletions of USP15 and SCAF1, respectively, indicative of heterozygous loss of these genes (Fig.6a). Focal USP15 and SCAF1 copy-number losses have been identified in independent large-scale genome studies48,49. In addition, allelic copy number loss also coincided with reduced expression of USP15 and SCAF1 and patients with deep or shallow USP15 or SCAF1 deletions showed a significant trend towards a shorter overall survival (Fig.6b and Supplementary Fig.8a). Given our genetic and biochemical data linking SCAF1 and USP15, we next considered patients with deep or shallow USP15 or SCAF1 deletions as a group (=37% of patients) and found a significantly shorter overall survival (Supplementary Fig.8b). This raises the possibility that USP15 and potentially also SCAF1 function in a haploinsufficient manner, which is in line with the increased tumorigenesis found in the Usp15fl/+; KRasG12D; Pdx1-Cre mice.

A Oncoprint of the indicated genes in PDAC samples (n=293, TCGA). B Kaplan-Meier survival analyses of PDAC patients with deep or shallow USP15 or SCAF1 deletion. (n=293, TCGA) Log-Rank test (Mantel-Cox). C Tumor-free survival of NSG mice orthotopically injected with sgCtrl (n=5), sgUsp15_1/2 (n=5; n=5) or sgSCAF1_1/2 (n=5; n=5) PANC-1 cells. Two independent sgRNAs were used. Log-Rank test (Mantel-Cox) D Dose-response curves for sgCtrl, sgUsp15 or sgSCAF1 PANC-1 cells treated with Olaparib. (%; meanSD, n=3 independent experiments) two-way ANOVA (sgUsp15_1 p=0.0242; sgUsp15_2 p=0.0387; sgScaf1_1 p=0.0281; sgScaf1_2 p=0.0371), Dunnetts multiple comparison. E sgUSP15 and sgOR2W5 PDO Competition Assay. sgUSP15 and control sgOR2W5 patient-derived organoids were disassociated into single cells and mixed in a 20:80% ratio. Organoid cultures were passaged, and a sample was collected every ~7 days. Percentage of DNA indels is tracked overtime by sanger-sequencing and TIDE analysis.

Next, we assessed the expression of USP15 in 4 human pancreatic cancer cell lines. While, PANC1 and HPAFII exhibited expression of the small as well as the long USP15 isoform, MiaPACA2 and BXPC3 cells only exhibited low-level expression of the long USP15 isoform, indicating that USP15 is also downregulated in some human pancreatic cancer cell lines (Supplementary Fig.8c).

To functionally test USP15 and SCAF1, we genetically ablated these genes in human PANC1 cells (Supplementary Fig.8d and e). Importantly, genetic ablation of SCAF1 resulted in increased expression of the short USP15 isoform, indicating that this mechanism is conserved from mouse to human cells (Supplementary Fig.8f). Similarly, to our autochthonous mouse experiments, we also found that loss of USP15 or SCAF1 in PANC1 cells resulted in accelerated tumorigenesis and increased sensitivity to Olaparib and Gemcitabine (Fig.6c, d and Supplementary Fig.8g). We also observed increased NRF2 protein levels in USP15 knockout PANC1 cells, which showed further elevated upon inhibition of TXNRD1/2 and antioxidant imbalance by auranofin treatment50 (Supplementary Fig.8h), akin to our findings in mouse KC cells. USP15 knockout PANC1 cells also exhibited increased sensitivity to auranofin treatment (Supplementary Fig.8i).

Lastly, we genetically ablated USP15 in patient-derived organoids (PDOs) from 3 different pancreatic cancer patients using Cas9 ribonucleotide particles. We set up competitive growth assays to assess the relative fitness of USP15 knockout PDOs compared to OR2W5 knockout PDOs. Of note, the OR2W5 olfactory receptor is not expressed in pancreatic PDOs and thus serves as control. We mixed the USP15 knockout and the OR2W5 knockout PDOs at a 1:4 ratio and followed their relative growth by quantifying the percent of USP15 and OR2W5 mutations over time using Sanger sequencing. Within ~10 passages, we observed that the PDO cultures were almost completely taken over by USP15 knockout cells (Fig.6e). Together, these data demonstrate the tumor suppressive function of USP15 and SCAF1 in pancreatic cancer by modulating several important signaling pathways and that loss of USP15 and SCAF1 sensitizes to Gemcitabine and Olaparib.

Read more from the original source:
In vivo CRISPR screens reveal SCAF1 and USP15 as drivers of pancreatic cancer - Nature.com

Comments are closed.