Mutant SF3B1 promotes malignancy in PDAC

The splicing factor SF3B1 is recurrently mutated in various tumors, including pancreatic ductal adenocarcinoma (PDAC). The impact of the hotspot mutation SF3B1K700E on the PDAC pathogenesis, however, remains elusive. Here, we demonstrate that Sf3b1K700E alone is insufficient to induce malignant transformation of the murine pancreas, but that it increases aggressiveness of PDAC if it co-occurs with mutated KRAS and p53. We further show that Sf3b1K700E already plays a role during early stages of pancreatic tumor progression and reduces the expression of TGF-β1-responsive epithelial–mesenchymal transition (EMT) genes. Moreover, we found that SF3B1K700E confers resistance to TGF-β1-induced cell death in pancreatic organoids and cell lines, partly mediated through aberrant splicing of Map3k7. Overall, our findings demonstrate that SF3B1K700E acts as an oncogenic driver in PDAC, and suggest that it promotes the progression of early stage tumors by impeding the cellular response to tumor suppressive effects of TGF-β.

Despite compelling evidence on the oncogenic role of mutated SF3B1 in hematologic malignancies, its contribution to the formation or progression of solid tumors is less understood.Since splicing deregulation has been reported as a hallmark of PDAC, with SF3B1 being recurrently mutated (Bailey et al., 2016), we aimed at elucidating the impact of the frequently occurring SF3B1 K700E mutation to the pathogenesis of this tumor type.We demonstrate that Sf3b1 K700E increases malignancy in a mouse model for PDAC and decreases the sensitivity to TGF-β-responsive EMT genes.Experiments in pancreatic organoids and cell lines further provide evidence that SF3B1 K700E protects from TGF-β-induced cell death and that TGF-β-resistance is partly mediated through missplicing of Map3k7.Together, our work suggests that SF3B1 K700E exerts its oncogenic role in PDAC by dampening the tumor-suppressive effect of TGF-β.

Results
SF3B1 K700E is a tumor driver during early stages of PDAC formation RNA processing has been previously identified as a hallmark of pancreatic cancer in the published Pancreatic Cancer Australian (PACA-AU) cohort (Bailey et al., 2016).Validating these findings, we found that genes encoding for the splice factors RBM10, SF3B1, and U2AF1 are also frequently mutated in the Pancreatic Cancer Canadian (PACA-CA) cohort (Figure 1-figure supplement 1A).In accordance with its described function as a tumor suppressor (Hernández et al., 2016), 56% of the mutations found in RBM10 lead to a truncated protein.Conversely, in SF3B1 and U2AF1 the majority of mutations were missense mutations that occurred at hotspot sites, indicating a neomorphic function of the mutated proteins.Like in other cancer types, also in PDAC the most frequently found mutation in SF3B1 led to a lysine (K) to glutamic acid (E) change at position 700 (SF3B1 K700E ; Figure 1-figure supplement 1B).Therefore, we experimentally tested if the SF3B1 K700E mutation contributes to PDAC malignancy by generating a mouse model where the Sf3b1 K700E mutation is specifically activated in the pancreas using Ptf1a-Cre (Figure 1A, Figure 1-figure supplement 1C).First, we tested if Sf3b1 K700E/+ alone is sufficient to induce PDAC formation.However, heterozygous activation of the Sf3b1 K700E allele did not have any effect on survival of mice or weight of the pancreas after 300 days (Figure 1B and  C).Furthermore, assessing the pancreata histologically for two prognostic markers for PDAC, Cytokeratine-19 (CK19) and Mucin 5AC (MUC5AC), revealed no difference to wild type control mice (CK19 was restricted to pancreatic ducts and MUC5AC was not expressed; Figure 1-figure supplement 1D, E).Next, we assessed if the K700E mutation increases aggressiveness of PDAC, and crossed the Sf3b1 K700E allele into the Ptf1a-Cre; Kras G12D/+ ; Trp53 fl/fl (KPC) mouse model, which is known to induce PDAC within 2-3 months (Bardeesy et al., 2006a;Hingorani et al., 2005;Marino et al., 2000).Importantly, KPC-Sf3b1 K700E/+ animals displayed a significantly shorter survival (mean survival 57 days vs. 64 days), and an increased tumor size compared to KPC mice at the age of 9 weeks (Figure 1D-F).However, since at this timepoint similar amounts of tissue fibrosis and CK19 positive cells were observed in Sf3b1 K700E/+ mutant vs. Sf3b1 +/+ KPC tumors (Figure 1-figure supplement 1F-I), we next assessed if SF3B1 K700E already promotes malignancy at early stages of tumorigenesis.Indeed, when we analyzed pancreata of 5-week-old mice, already 67% of KPC-Sf3b1 K700E/+ mice vs. 25% of KPC mice Further supporting that SF3B1 K700E already plays a role in early stages of carcinogenesis, Sf3b1 K700E/+ also had an effect on tumor malignancy when introduced into Kras G12D/+ (KC) mice, which is a model for pre-cancerous pancreatic neoplasms with sporadic PDAC formation after a prolonged latency period (Hingorani et al., 2005).At 43 weeks of age, the area of neoplastic pancreas tissue was increased, and in 64% of KC-Sf3b1 K700E/+ mice vs. 10% of KC mice the lesions had already progressed to PDAC (Figure 1L; Figure 1-figure supplement 1L).In addition, the time of disease-free survival was significantly shortened in KC-Sf3b1 K700E/+ mice vs. KC mice (Figure 1K).Finally, in a model for advanced cancer, where KPC and KPC-Sf3b1 K700E cells were harvested from fully developed PDAC tumors and expanded in vitro before being orthotopically transplanted, we did not observe differences in tumor growth (Figure 1-figure supplement 1M, N).Together, our data demonstrate that the SF3B1 K700E contributes to PDAC malignancy by accelerating the formation of precursor lesions.

SF3B1 K700E reduces expression of EMT genes in pancreatic tumors
In order to elucidate the functional impact of SF3B1 K700E on the transcriptome, we isolated cancer cells of mouse tumors by fluorescence-activated cell sorting (FACS) of Epithelial cell adhesion molecule (EpCAM) positive cells and performed RNA-sequencing (RNA-seq) (Supplementary file 1).High purity of isolated tumor cells was confirmed by the absence of sequencing reads for Trp53 exons 2-10, which are excised via Cre-recombination specifically in tumor cells (Figure 2-figure supplement 1A), and by the presence of the Sf3b1 K700E mutation in 38% of the transcripts (Figure 1D).Principal component analysis separated the sequenced replicates (3 KPC and 4 KPC-Sf3b1 K700E/+ tumors) according to the genotype, indicating a major impact of the K700E mutation on the transcriptome (Figure 2figure supplement 1B).We next performed gene set enrichment analysis (GSEA), which revealed IFN-α-response as the most significantly enriched pathway in KPC-Sf3b1 K700E/+ tumor cells (Figure 2-figure supplement 1C).This result is in line with a previous study, which found that aberrant splicing caused by SF3B1 inhibition or oncogenic SF3B1 mutations induces an IFN-α-response through retinoic acid-inducible gene I (RIG-I) mediated recognition of cytosolic aberrant RNA-species (Chang et al., 2021).Interestingly, epithelial-mesenchymal transition (EMT) emerged as the most significantly attenuated gene set in KPC-Sf3b1 K700E/+ cells (Figure 2A and B,.We first confirmed downregulation of the most significantly depleted gene of the EMT gene set, the glycoprotein Tenascin-C (Tnc), by qPCR on additional KPC-Sf3b1 K700E/+ tumor samples (Figure 2C) and by histology in KPC-Sf3b1 K700E/+ PDAC sections (Figure 2D and E).Next, we assessed if the reduction of EMT genes was induced cell-autonomously by the Sf3b1 K700E/+ mutation, or if it was an indirect consequence of the altered micro-environment in Sf3b1 K700E KPC tumors.We therefore compared the expression of the 15 most significantly depleted EMT genes (Figure 2B) in vitro in Sf3b1 K700E vs. Sf3b1 WT KPC pancreatic organoids.Importantly, 71% of the analysed genes were significantly reduced in KPC-Sf3b1 K700E/+ vs. KPC organoids, with none of the genes showing a trend towards elevated expression (Figure 2F).To rule out that differences in EMT gene expression could be a consequence of differences in the tumor stage between Sf3b1 K700E versus Sf3b1 WT KPC tumors, we next established non-cancerous pancreatic organoids from LSL-Kras G12D/+ ; Trp53 fl/fl ; Sf3b1 flK700E/+ or Sf3b1 +/+ and LSL-Kras G12D/+ ; Trp53 fl/ p-value.(D) Representative photographs of WT, KPC, and KPC-Sf3b1 K700E/+ pancreata of mice 9 weeks old mice.(E) Survival of KPC and KPC-Sf3b1 K700E/+ mice, p-value was determined by Log-rank (Mantel-Cox) testing.(F) Pancreatic weight of KPC and KPC-Sf3b1 K700E/+ mice at 9 weeks of age.Two-tailed unpaired t-test was used to compute the indicated p-value.(G) Representative micrograph images of H&E, MUC5AC, β-amylase and CK19 staining of KPC and KPC-Sf3b1 K700E/+ pancreata of 5-week-old mice.Scale bar is 50 μM (H&E) or 100 μM (IF).(H) Quantification of β-amylase (H), MUC5AC (I) and CK19 (J) staining shown in (G).Per specimen, the average value of 5 random microscopy fields is displayed, a two-tailed unpaired t-test was used to compute the indicated p-value.Error bar represents standard error of the mean (SEM).(K) Survival of Kras G12D/+ and Kras G12D/+ ; Sf3b1 K700E/+ mice.p-Value was determined by Log-rank (Mantel-Cox)

SF3B1 K700E confers resistance to TGF-β1-induced apoptosis in organoids and cell lines
The two major EMT-promoting cytokines are TNF-α and TGF-β (Bulle and Lim, 2020).We therefore determined if the EMT genes that are most significantly downregulated by Sf3b1 K700E in KPC tumours are induced by TNF-α or TGF-β.Importantly, we found that 80% of the analysed genes were strongly induced by TGF-β in tumour derived KPC cells (Figure 2-figure supplement 1E), whereas TNF-α significantly upregulated only 20% of these genes (Figure 2-figure supplement 1F).Furthermore, we observed that in vitro induced Sf3b1 K700E KPC cells fail or only partially induce 6 of the 9 assessed EMT genes when stimulated with TGF-β (Figure 2-figure supplement 1G).In line with these results, in vitro induced Sf3b1 K700E KPC organoids exhibit a sixfold decrease in matrigel invasion compared to KPC control organoids when stimulated with TGF-β (Figure 2G and H).Moreover, human PANC-1 cells expressing SF3B1 K700E display a reduced migratory rate compared to control PANC-1 cells expressing wildtype SF3B1 in a wound healing assay upon TGF-β stimulation (Figure 2I and J).
In pancreatic lesions TGF-β induces EMT, followed by apoptosis of the affected cells in a process termed lethal EMT (David et al., 2016).This prompted us to speculate that SF3B1 K700E could drive PDAC progression by reducing sensitivity of epithelial cells to TGF-β-mediated apoptosis.Performing immunofluorescence staining for cleaved caspase 3 in KPC tumors, we first confirmed that the majority of apoptotic cells reside in the lumen of PanINs (Figure 3-figure supplement 1A, B), and that these cells are negative for the epithelial marker E-cadherin and positive for the mesenchymal marker Fibronectin-1 (Figure 3-figure supplement 1C; Hruban et al., 2006).We then analysed Sf3b1 K700E tumors by immunofluorescence staining.In line with our hypothesis, we observed a reduction in luminally extruded cells and a reduction in cleaved caspase 3-positive cells (CC3, Figure 3A and B, Figure 3-figure supplement 1D).To further analyse the impact of Sf3b1 K700E on the tumor suppressive effect of TGF-β, we exposed Sf3b1 WT and Sf3b1 K700E KPC tumor organoids to TGF-β1.We again observed reduced cleaved caspase 3 and 7 activity in Sf3b1 mutant organoids, and a greatly increased survival rate (72% vs 17% surviving organoids, Figure 3C-E).Since results from our mouse models indicate that the Sf3b1 K700E mutation already plays a role during early stages of carcinogenesis (Figure 1G-L, Figure 1-figure supplement 1G), and the tumor-suppressing effect of TGF-β is most prominent on pre-cancerous epithelial cells (Massagué, 2008), we additionally established organoid lines with-and without Sf3b1 K700E from non-cancerous mouse pancreata.While Sf3b1 K700E led to slightly reduced proliferation without supplementation of TGF-β1 (Figure 3-figure supplement indicated p-value.(D) Representative Immunofluorescence staining of CK19 (red) and TNC (green) in murine PDAC samples, counterstained with DAPI (blue).Scale bar is 50 μm.(E) Quantification of TNC staining in KPC (n=6) and KPC-Sf3b1 K700E/+ (n=6) tumors.The averaged area of TNC staining in three randomly chosen fields per tumor specimen was compared by a two-tailed unpaired t-test.(F) The expression of the EMT genes displayed in (B) was assessed by RT-qPCR in tumor cells (KPC, n=11 and KPC-Sf3b1 K700E/+ n=11) after one passage of ex-vivo culture.For analysis, Ct-values of the indicated genes were normalized to Actb and a two-tailed unpaired t-test was used to compute the indicated p-values.(G) Representative micrographs of KPC (n=3) and KPC-Sf3b1 K700E/+ (n=3) cancer organoid lines treated with TGF-β1 (10 ng/ml) for 48 hr.Matrigel was detached prior to staining with crystal violet, allowing quantification of cells migrated through the matrigel matrix and attached at the cell culture plate.Scale bar is 1 mm (panel above) or 100 μm (panel below).(H) Quantification of micrographs shown in (G).The fraction of attached organoids was calculated by dividing the number of attached organoids by the number of total organoids contained in the matrigel dome.The experiment was performed independently three times for every cell line, the average of all replicates is shown.Error bar represents SEM.Two-tailed unpaired t-test was used to compute the indicated p-value.(I) Representative micrographs of wound healing assay of PANC-1 WT and PANC-1 SF3B1 K700E/+ cells pre-treated with TGF-β1 (10 ng/ml) for 24 hours at the indicated time points after performing the scratch.Scale bar is 100 μm.(J) Quantification of wound healing assay of PANC-1 WT and PANC-1 SF3B1 K700E/+ cells, displaying the ratio of wound-closure of cells treated with TGF-β1 or without (Ctrl).The experiment was independently performed three times.Error bar represents standard deviation (SD).A two-tailed unpaired t-test was used to compute the indicated p-values.
To assess if reduced sensitivity to TGF-β is also observed in human pancreas cells containing the SF3B1 K700E mutation, we stably overexpressed either wildtype or mutated SF3B1 in the human pancreatic duct cell line H6c7.This cell line is derived from healthy pancreatic tissue, which unlike all tested human PDAC-derived cell lines (BxPC-3, Mia PaCa-2, PANC-1, and PSN-1) is still partially responsive to the growth-suppressive effect of TGF-β signalling (Figure 3J, Figure 3-figure supplement 1I).In line with our results from murine PDAC, overexpression of mutant SF3B1 K700E resulted in an increased viability upon TGF-β1 exposure compared to overexpression of wildtype SF3B1 (Figure 3J), while no effect of SF3B1 K700E on proliferation was observed in absence of TGF-β1 treatment in human pancreatic duct cells and PDAC cell lines (Figure 3-figure supplement 1J).

SF3B1 K700E reduces TGF-β sensitivity through Map3k7 missplicing in organoids and cell lines
To identify how SF3B1 K700E could mediate the observed TGF-β resistance in pancreatic cells, we next assessed the impact of the mutation on RNA-splicing.By analysing RNA-seq data from sorted KPC vs. KPC-Sf3b1 K700E/+ tumor cells we predominantly identified alternative 3' splice events (Figure 4A, Supplementary file 2), with cryptic 3' ss showing an upstream adenosine enrichment and a less pronounced polypyrimidine tract most frequently located 8-14 bases upstream of the canonical 3' ss (Figure 4B and C).These findings are in accordance with previous splice-analyses performed in various murine SF3B1 mutant tissues (Liu et al., 2021;Mupo et al., 2017;Obeng et al., 2016;Yin et al., 2019).Next, we sought to determine which of the identified SF3B1 K700E -dependent alternative splice events are conserved between mice and humans.Due to limited publicly available RNA-seq datasets in human PDAC, we analysed a pan-cancer dataset containing samples of 32 different cancer types.In agreement with previous studies, we found that also in human cancers the SF3B1 K700E hotspot mutation leads to a predominant use of cryptic 3' ss (Figure 4-figure supplement 1A, B, Supplementary file 2; Alsafadi et al., 2016;DeBoever et al., 2015;Kesarwani et al., 2017;Tang et al., 2020;Wang et al., 2016).Importantly, we further identified 11 genes that contained an alternative 3' splice-event linked to SF3B1 K700E in human tumors as well as in KPC mice (Figure 4D).Of those genes, MAP3K7 (formerly known as TGF-β activated kinase 1/TAK1) in particular raised our attention.It is a well-described effector of cytokine-signalling that mediates non-canonical TGF-β signalling (Kim and Choi, 2012), and it was recently shown to induce EMT and apoptosis in TGF-β-stimulated human mammary cells (Tripathi et al., 2019).Using targeted RNA-seq, we discovered that one third of Map3k7 transcripts were misspliced in pancreata of Sf3b1 K700E/+ and KPC-Sf3b1 K700E/+ mice (Figure 4E and F, Figure 4-figure supplement 1C).Confirming inter-species conservation of this alternative splice-event, MAP3K7 was also misspliced in H6c7 cells and human PDAC cell lines overexpressing SF3B1 K700E (Figure 4F, Figure 4-figure supplement 1D).Notably, when we further assessed p-value.(F) Representative microscopy images of WT and Sf3b1 K700E/+ organoids exposed to 10 ng/ml TGF-β1 for 48 hr.Scale bar is 500 μM.(G) Quantification of CC3/7 in WT and Sf3b1 K700E/+ organoids exposed to 10 ng/ml TGF-β1 for 48 hr.The experiment was repeated independently four times, error bar represents SEM.Two-tailed unpaired t-test was used to compute the indicated p-values.(H) Quantification of viable organoids of the indicated genotype exposed to 10 ng/ml TGF-β1 for 48 hr, normalized to organoid numbers of untreated control samples.Each data point shows a different organoid line.For each organoid line, the experiment was independently performed three times, the average of replicates and SEM is plotted.Two-tailed unpaired t-test was used to compute the indicated p-values.(I) Organoid count of organoids cultured in medium containing 1 ng/ml TGF-β1 for up to 120 days.One organoid line per genotype was used, the experiment was repeated three times independently.(J) Viability of the human pancreatic duct cell line H6c7 overexpressing wildtype or mutated SF3B1 after 72 hr of exposure to 10 ng/ml TGF-β1 assessed by crystal violet staining.The optical density of TGF-β1-treated cells was normalized to untreated control cells.The experiment was independently performed three times, error bar represents SD.A two-tailed unpaired t-test was used to compute the indicated p-value.
The online version of this article includes the following figure supplement(s) for figure 3:    SF3B1 K700E dependent alternative splicing of Ppp2r5a, which was reported to impair apoptosis via post-translational modification of BCL2 in leukaemia (Liu et al., 2020a), we did not observe significant alternative 3'ss usage or mRNA expression of Ppp2r5a in KPC-Sf3b1 K700E tumor cells (Figure 4-figure supplement 1E, F).This observation indicates tissue-specificity of Ppp2r5a missplicing.
Since Sf3b1 K700E dependent missplicing in MAP3K7 was shown to result in reduced RNA and protein levels of MAP3K7 in leukaemia (North et al., 2022), we hypothesized that the reduced responsiveness to TGF-β signalling in SF3B1 K700E mutant pancreas cells is caused by lower MAP3K7 levels.Indeed, RT-qPCR and western blotting confirmed a reduction in MAP3K7 levels in vitro and in vivo in Sf3b1 K700E mutant pancreatic cells (Figure 5A-D).Next, we tested whether this reduction could explain the observed resistance to TGF-β in Sf3b1-mutant PDAC, and assessed the expression of EMT genes in TGF-β-treated KPC cells with a stable knock-down of MAP3K7 (Figure 5E, Figure 5figure supplement 1A).Indeed, a decrease of Map3k7 mRNA levels to 35% (SD ± 10%) led to a reduced expression in 7 out of 10 EMT genes (Figure 5-figure supplement 1B).Moreover, knocking down Map3k7 in pre-cancerous pancreas organoids led to increased viability upon TGF-β1-treatment (Figure 5F, Figure 5-figure supplement 1C), and chemical inhibition of p38, one of the major effectors of MAP3K7, partially protected organoids against TGF-β1-induced cell death (Figure 5G).Further supporting our hypothesis that Sf3b1 K700E mediates resistance to TGF-β1 via MAP3K7, overexpression of the full-length isoform of MAP3K7 in TGF-β1-treated Sf3b1 K700E mutant organoids significantly decreased their viability (Figure 5H1).
To further assess conservation of this mechanism between mice and humans, we next treated human pancreatic H6c7 cells as well as human pancreatic organoids before TGF-β1 exposure with Takinib (TAKi), a chemical inhibitor for MAP3 K7.In line with our results in murine cells, also in human pancreatic cells inhibition of MAP3 K7 led to a significant increase in survival (Figure 5J and K).Moreover, conditioning human PANC-1 PDAC cells with TAKi prior to TGF-β1 exposure resulted in a significant decrease in migratory capacity in wound healing assays (Figure 5L and M).Taken together, our results suggest that Sf3b1 K700E mediates resistance of pancreatic epithelial cells to TGF-β1 via MAP3 K7, providing a potential mechanism for its role of in PDAC progression.

Discussion
While the roles of the most frequently mutated tumor driver genes in PDAC are well understood, oncogenes occurring at lower rates are less well studied (International Cancer Genome Consortium et al., 2010).One of these genes is SF3B1, for which hotspot mutations occur in various tumor types, including PDAC.While the molecular function of oncogenic SF3B1 on RNA-splicing is well described, how deregulation of misspliced genes contribute to malignancy in different cancer entities is only partially understood.Previous studies have shown that in chronic lymphocytic leukaemia (CLL) SF3B1 K700E leads to missplicing of PPP2R5A, which in turn stabilizes c-Myc and thereby promotes aggressiveness of tumor cells (Liu et al., 2020a;Yang et al., 2021).Furthermore, in breast cancer SF3B1 K700E causes a tumor-promoting effect through missplicing in MAP3 K7 and downstream activation of NF-κB-signalling (Liu et al., 2021).
In our study we analysed the oncogenic function of SF3B1 K700E in the context of PDAC.Using a conditional mouse model, we provide the first experimental evidence that Sf3b1 K700E indeed promotes PDAC progression.While the Sf3b1 K700E mutation alone did not induce malignant transformation in the mouse pancreas, co-occurrence with KRAS and p53 mutations increased the severity of PDAC.Effects of SF3B1 K700E were already observed during early stages of tumor progression and resulted in reduced expression of TGF-β-responsive EMT genes.Our experiments further revealed that SF3B1 K700E reduces TGF-β-induced cell-death in pancreatic duct cell lines and organoids, providing a potential mechanism Dashed line indicates cropping of gel image.(F) Representative gel image of RT-PCR amplicon of Map3k7 cDNA isolated from WT and Sf3b1 K700E/+ pancreata and from the indicated human cell lines.The upper band of the gel image represents the non-canonical transcript variant, the lower band represents the canonical isoform.The length of the amplicon derived from murine pancreas cells is 114 bp (canonical) and 137 bp (non-canonical) and from human cell lines 94 bp (canonical) and 114 bp (non-canonical).
The online version of this article includes the following figure supplement(s) for figure 4:  for the oncogenicity of SF3B1 K700E in PDAC.In line with this hypothesis, pancreatic epithelial cells have previously been found to undergo lethal EMT when stimulated with TGF-β (David et al., 2016), and acquiring TGF-β resistance is considered to be essential in early stages of PDAC tumorigenesis (Hezel et al., 2012).Interestingly, akin to Sf3b1 K700E , deletion of the vital TGF-β signalling component Smad4 only triggered malignant transformation of the pancreas in combination with oncogenic Kras mutations (Bardeesy et al., 2006b).Splicing analysis of RNA-seq datasets from both murine and human cancers identified several genes that are misspliced in the SF3B1 K700E background.Among the identified genes was Map3 k7, which is known to mediate non-canonical TGF-β signaling (Kim and Choi, 2012) and to induce EMT and apoptosis in TGF-β stimulated human mammary cells (Tripathi et al., 2019).In this study, we show that reducing Map3 k7 levels by shRNA mediated knockdown in pancreatic cell lines and organoids also impairs TGF-β-mediated EMT and cell death, although not to the same extend as Sf3b1 K700E .Thus, while aberrant splicing of other genes is likely also contributing to the observed resistance to TGF-β, our experiments indicate that TGF-β resistance in SF3B1 K700E mutant pancreatic ducts is at least partly mediated via MAP3 K7.Importantly, this hypothesis is also supported by previous studies, which demonstrated that SF3B1 mutations induce 3' missplicing of MAP3 K7 in various tumor entities (Bondu et al., 2019;Li et al., 2021;Lieu et al., 2022;Liu et al., 2020b;Wang et al., 2016;Zhang et al., 2019), that aberrant splicing by SF3B1 K700E reduces MAP3 K7 protein levels (North et al., 2022), and that MAP3 K7 mediates TGF-β-induced EMT and apoptosis in mammary epithelial cells (Tripathi et al., 2019).Notably, depending on the cellular context, MAP3 K7 has been shown to exert proapoptotic stimuli through activation of MAPK p38 and Jun N-terminal kinase (JNK; Mihaly et al., 2014), or anti-apoptotic stimuli through activation of the NF-κB pathway (Mukhopadhyay and Lee, 2020).Thus, while our results show that SF3B1 K700E mediated missplicing and inactivation of MAP3 K7 protects pancreatic duct cells from TGF-β induced cell death, in other tissues or in combination with other mutations, the reduction of MAP3 K7 levels by SF3B1 K700E could trigger different phenotypic responses.
One limitation of our study is the lack of in vivo evidence supporting our hypothesis that SF3B1 K700E promotes PDAC progression through MAP3 K7 missplicing.We attempted to study the effect of altered MAP3 K7 levels in an orthotopic PDAC transplantation model.However, we found that the model was not suitable for adressing this question, since introducing SF3B1 K700E in KPC cells had no significant impact on tumor growth after transplantation.This outcome aligned with our expectations, given that KPC cells were isolated from fully established PDAC, where SF3B1 K700E does not play the same role as in early-stage PDAC.Additional studies using autochthonous PDAC models, where Map3 k7 levels are modulated in early stage KPC or KPC-Sf3b1 K700E/+ tumors, would be required to foster in vivo evidence for the functional role of Map3 k7 in SF3B1 K700E mutant PDAC.
organoids.(E) RT-qPCR analysis of Map3k7 in cells transduced with a shMap3k7 compared to a control shRNA.Error bar represents SD, a two-tailed unpaired t-test was used to compute the indicated p-values.(F) Quantification of viable WT and Sf3b1 K700E/+ murine pancreatic duct organoids or WT transduced with control shRNA (shCtrl) or shRNA targeting Map3k7 (shMap3k7).The organoids were exposed to 10 ng/ml TGF-β1 for 24 hr prior to analysis.The experiment was independently performed three times.Error bar represents SD, a two-tailed unpaired t-test was used to compute the indicated p-value.(G) Viability of wildtype (WT) organoids cultured in medium containing 10 ng/ml TGF-β1, supplemented with chemical inhibitors targeting the indicated effectors of TGF-β1-signalling.Two independent experiments are summarized.Further details are provided in the methods section.(H) RT-qPCR analysis of Map3k7 in cells transduced by lentivirus with an overexpression construct of Map3k7, compared to overexpression of GFP.Error bar represents SD, a two-tailed unpaired t-test was used to compute the indicated p-values.(I) Quantification of viable murine pancreatic duct organoids with stable overexpression of Map3k7, exposed to 10 ng/ml TGF-β1 for 96 hr.Data represents one organoid line per condition, the experiment was independently performed three times.Error bar represents SD, a two-tailed unpaired t-test was used to compute the indicated p-value.(J-K) Viability of human pancreatic duct H6c7 cells (J) or human pancreatic duct organoids (K) exposed to 10 ng/ml TGF-β1 with addition of the MAP3K7 inhibitor Takinib (TAKi, 5 μM) or DMSO.The viability was assessed after 48 hr of TGF-β1 treatment and normalized to cells grown in absence of TGF-β1.The experiment was independently performed four (J) or three (K) times.(L) Representative micrographs of wound healing assay of PANC-1 cells pretreated with TGF-β1 (10 ng/ml) for 24 hr, with or without addition of 5 μM TAKi, at the indicated time points after performing the scratch.Scale bar is 50 μM.(M) Quantification of wound healing assay shown in (L).The experiment was independently performed three times.Error bar represents SD, a twotailed unpaired t-test was used to compute the indicated p-values.
The online version of this article includes the following figure supplement(s) for figure 5:  Taken together, this study provides a first demonstration that oncogenic SF3B1 K700E promotes tumor progression in vivo in a mouse model for PDAC.Based on data from pancreatic organoids and cell lines, we further suggest that SF3B1 K700E promotes PDAC progression by reducing tumor-suppressive TGF-β signalling through missplicing of MAP3 K7.

Animal models
Sf3b1 K700E/+ mice were a gift from E. A. Obeng (Dana-Farber Cancer Institute, Boston, USA) and B. L. Ebert (Brigham and Women's Hospital, Harvard Medical School, Boston, USA).LSL-Kras G12D/+ , LSL-Trp53 R172H/+ and Ptf1a-Cre mice were purchased from the Jackson Laboratory (Bar Harbor, Maine, USA).All Sf3b1 K700E/+ and KPC-Sf3b1 K700E/+ mice were bred in a C57BL/6 J background, Sf3b1 K700E/+ ; Kras G12D/+ mice were a C57BL/6J-BALB/c strain.Female and male mice were used for all experiments.Animals displaying dwarfism were excluded from analysis.The minimum of animals needed for the study was estimated by Fisher-Yates analysis.Due to the observed variance of the mouse model, more mice than initially estimated were used for the study.For the orthotopic transplantation models, KPC cells were derived from the autochthonous model described above by digestion of fully grown tumors and plating the digest in cell culture dishes.After multiple passages, KPC cells were grown to 80% confluency, and 10 5 cells were injected into the pancreas of BALB/cJ mice in matrigel, diluted in a 1:1 ratio with DMEM.2Two weeks upon transplantation, mice were sacrificed.Mice were held in a specific-pathogen-free (SPF) animal facility at the ETH Phenomics Center EPIC (ETH Zurich, Switzerland).All animal experiments were conducted in accordance with the Swiss Federal Veterinary Office (BVET) guidelines (license no.ZH055/17).

Immunohistochemistry
Murine tissue specimens were dissected and fixed in 10% neutral buffered formalin for 48-72 hr.Thereafter, formalin was replaced with 70% ethanol before paraffin-embedding and sectioning at a thickness of 4 μm.Hematoxylin and eosin stainings were performed according to the manufacturer's instructions.Anti-Cleaved Caspase 3 (Asp175) antibody (Cell Signaling Technology), anti-Amylase (100-4147) antibody (ThermoFisher), anti-Cytokeratin 19 (ab15463) antibody (ThermoFisher), anti-E-Cadherin (Clone 36) antibody (BD Biosciences), anti-FN-1 antibody (Chemicon), anti-MUC5AC (45M1) antibody (Novus Biologicals) and anti-TNC (Clone 578) antibody was used according to the manufacturer's recommendations.TNC and amylase staining was quantified by calculating the average of Raw Int Density of 3-5 randomly chosen microscopy fields per specimen using ImageJ.CK19 and MUC5AC staining was quantified by counting stained cells of 3-5 randomly chosen microscopy fields per specimen.ADM structures were quantified by counting ADM structures of 5 microscopy fields per specimen of tissue not yet (or only minorly) progressed to PanINs.Luminal necrotic cells were defined as shed cells residing within the lumen of PanIN-lesions.

Organoid growth assay
Growth of organoids was assessed with CellTiter-Glo 3D (Promega).For absolute quantification of ATP levels, standard curves with defined concentrations of ATP were used for every measurement according to the manufacturer's instructions.As approximation of proliferation rate, the ratio of ATP concentrations at the indicated time points was calculated.

Crystal violet assay
To measure proliferation of cell lines, 5000 cells were seeded per 96 wells.At the indicated time points, cells were stained with 0.5% (w/v) crystal violet (Sigma-Aldrich) dissolved in an aqueous solution with 20% Methanol (v/v).After washing, plates were allowed to air-dry and the crystal violet was dissolved in 10% acetic acid.Optical density was measured at 595 nM in an Infinite 200 plate reader (Tecan).

Organoid viability assay
For short-term treatment, organoids were seeded as fragments in 10 μl of Matrigel and allowed to form spheres for 24 hr in regular organoid medium.Organoids were thereafter exposed to the indicated concentration of TGF-β1 (Thermo Fisher).To assess viability of the organoids, intact organoids were counted and compared to untreated organoids after 48 hr of TGF-β1-exposure.This method of quantification was validated by correlating counts of intact organoids with ATP levels as described above (data not shown).For long-term treatment, organoids were seeded as fragments in 40 μl of Matrigel.After allowing to form spheres for 24 hr, the indicated concentration of TGF-β1 was added.After 4 days of TGF-β1-treatment, matrigel-drops were imaged and the number of intact organoids was counted.Then, organoids were reseeded as fragments in normal organoid medium and TGF-β1 was added after 24 hr.Every 4th passage, organoids were split in a 1:1 ratio in the 1 ng/ml TGF-β1 condition.
Commercial cell lines were seeded and TGF-β1 was added at the indicated concentrations 12 hr after plating.

Organoid invasion assay
Organoids were seeded at equal density in 40 μl of matrigel in 24-well plates.Twenty-four hr after seeding, organoid growth medium was supplemented with 10 ng/ml TGF-β1 (Thermo Fisher).Nienty-six hr after seeding, matrigel domes were detached by rinsing and the migrated organoids (i.e. cells attached to the cell culture dish) were stained by crystal violet.The fraction of attached organoids was calculated by dividing the number of attached organoids by the number of attached organoids plus the number of non-attached organoids (i.e.organoids residing in the matrigel dome).
Cleaved-caspase 3/7 assay Organoids were seeded 10 μl of Matrigel and allowed to form spheres for 24 hr in regular organoid medium.Organoids were thereafter exposed to 10 ng/ml TGF-β1 (Thermo Fisher) overnight.Cleavage of Caspase 3 and 7 was quantified by using Caspase-Glo 3/7 Assay System (Promega) according to the manufacturer's instructions.
Wound healing assay 8x10 5 human PANC-1 cells, overexpressing SF3B1-WT or SF3B1 K700E , were seeded in six-well plates.Ten ng/ml  or DMSO [1:1000] in the respective experiments) was added overnight prior scratch formation with a 10 μl pipette tip.After scratch formation, fresh medium including TGF-β1 (and TAKi or DMSO) was added.0 hr, 4 hr, and 7 hr after scratch formation, images were taken and the width of the scratch was analysed, performing four measures at standardized positions for each image using ImageJ / Fiji (NIH).To calculate the effect of TGF-β on gap closure in Figure 2J, distance of migration (gap width 0 hr after scratch formation -gap width 4 hr after scratch formation) of cells cultured with TGF-β was divided by distance of migration of cells cultured with control medium.
shRNA-mediated Map3k7 knockdown shRNA targeting murine Map3 k7 was purchased from Sigma-Aldrich (TRCN0000022563).A pLKO.1puro Non-Target shRNA was used as control.Lentivirus was produced by PEI-based transfection of HEK293T cells.Briefly, HEK293T cells were seeded at 70% confluency in 6-well plates, and the following plasmids were transfected: PAX2 plasmid (1100 ng), VSV-G plasmid (400 ng), cargo plasmid (1500 ng).Medium was changed 12 hr after transfection and the virus-containing supernatant collected after 36 hr.Organoids were dissociated into single cells by Tryp-LE treatment for 5 min at 37°C and consecutive mechanical disruption.After centrifugation, 10% Lentivirus-containing supernatant in organoid medium (v/v) was added to the cell suspension.After a 4-6 hr incubation at 37°C, cells were seeded in matrigel as described above.Organoids were selected in 2 ng/ml Puromycin after the first passage for at least 5 days.

Overexpression of Map3k7
Murine Map3 k7 (full-length isoform) was amplified from cDNA of murine WT duct organoids and cloned into a Lenti-backbone (addgene #73582).Production of lentivirus and transduction of organoids was performed as described above.Organoids stably overexpressing GFP (addgene #17488) were used as experimental control.

Library preparation
The quantity and quality of the isolated RNA was determined with a Qubit (1.0) Fluorometer (Life Technologies, California, USA) and a Tapestion (Agilent, Waldbronn, Germany).The SMARTer Stranded Total RNA-Seq Kit -Pico Input Mammalian (Clontech Laboratories, Inc, A Takara Bio Company,California, USA) was used in the succeeding steps.Briefly, total RNA samples (0.25-10 ng) were reversetranscribed using random priming into double-stranded cDNA in the presence of a template switch oligo (TSO).When the reverse transcriptase reaches the 5' end of the RNA fragment, the enzyme's terminal transferase activity adds non-templated nucleotides to the 3' end of the cDNA.The TSO pairs with the added non-templated nucleotide, enabling the reverse transcriptase to continue replicating to the end of the oligonucleotide.This results in a cDNA fragment that contains sequences derived from the random priming oligo and the TSO.PCR amplification using primers binding to these sequences can now be performed.The PCR adds full-length Illumina adapters, including the index for multiplexing.Ribosomal cDNA is cleaved by ZapR in the presence of the mammalian-specific R-Probes.Remaining fragments are enriched with a second round of PCR amplification using primers designed to match Illumina adapters.The quality and quantity of the enriched libraries were validated using Qubit (1.0) Fluorometer and the Tapestation (Agilent, Waldbronn, Germany).The product is a smear with an average fragment size of approximately 360 bp.The libraries were normalized to 10 nM in Tris-Cl 10 mM, pH8.5 with 0.1% Tween 20.

Cluster generation and sequencing
The TruSeq SR Cluster Kit HS4000 or TruSeq PE Cluster Kit HS4000 (Illumina, Inc, California, USA) was used for cluster generation using 8 pM of pooled normalized libraries on the cBOT.Sequencing was performed on the Illumina HiSeq 4000 paired end at 2X126 bp or single end 126 bp using the TruSeq SBS Kit v4-HS (Illumina, Inc, California, USA).

RNAseq data analysis
Adapters have been trimmed with trimmomatic (v0.35).Pairs for which both reads passed the trimming have been mapped to the murine genome using STAR (v2.7.0a) and indexed BAM files obtained with samtools (v1.9).Reads were counted with featureCounts from subread package (v1.5.0).The read counts have been processed in a statistical analysis using edgeR (v3.24.3), obtaining a list of genes ranked for differential expression by p-value and Benjamini-Hochberg adjusted p-value as the estimate of the false discovery rate.All data is summarized in Supplementary file 1.

Gene set enrichment analysis
Gene set enrichment analysis (v.4.1.0,Broad Institute, MIT) was used to determine enriched gene sets in KPC or KPC-Sf3b1 K700E/+ tumor cells.Standard parameters of the software were used to perform the analysis.Molecular Signatures Database v7.4,Hallmark Gene Sets (H) was used to query enriched gene sets.The input gene expression matrix contained read-count information (count per million) of 21,633 genes.

Alternative splicing analysis
We ran a 2-pass alignment of the fastq files using STAR v2.7 (Dobin et al., 2013) using the GRCm38.p6 genome as reference.The gene annotation used was GENCODE v.m25.For gene expression quantification, we used a custom script, available at GitHub (copy archived at Kahles, 2021); commit hash d074114f1d0a9f518c9cd039f68de0cdf8d583ff.SplAdder v.2.2 (Kahles et al., 2016) was run to build splicing graphs and determine splice events.Differential splicing events were determined by calculating a log(psi +x) transformation of the percent spliced in (calculated as ratio of reads supporting the splice event over the number of reads supporting the alternate event).Splice events that did not show any variability over the samples were removed and missing values were mean imputed.After standardization, a two-sided t-test was used to calculate p-values of splice events differences between KPC and KPC-Sf3b1 K700E mice.All data is summarized in Supplementary file 2.

Motif analysis
Consensus 3' ss motif in proximity of the canonical and the cryptic 3' ss in sorted KPC-Sf3b1 K700E/+ tumor cells was assessed by query 30-40 bases spanning the respective 3'ss of the 7 main splice events for a motif using weblogo-sequence creator (https://weblogo.berkeley.edu/logo.cgi).

RT-PCR and quantitative RT-PCR (qPCR)
RNA-extraction was performed with QIAGEN RNeasy Mini Kit, and cDNA was generated with GoScript Reverse Transcriptase kit (Promega) according to the manufacturers' instructions.RT-PCR was performed with GoTaq G2 Green Master Mix (Promega) and gene-specific primers.Amplicons were fractionated on 2% TBE gel (Life Technologies) supplemented with 0.01% GelRed (Biotium).For qPCR, 2 μL of 1:10-diluted cDNA was added to 8 μl of 5 x HOT FIREPol Evagreen qPCR Supermix (SolisBiodyne).Raw gel images can be found in Source data 1.RT-qPCR was performed with a Light-Cycler480 II (Roche).Relative gene expression was determined with the comparative CT method.Genes with a median CT value of more than 33 cycles and a difference of less than 3.3 cycles to the template control (H 2 O) were defined as not detectable.Sequences of all primers used in this study are listed in Supplementary file 3.

NGS-based isoform quantification of Map3k7
Primers generating an amplicon including the exon 4 and exon 5 junction of Map3k7 cDNA were used.Briefly, a gene-specific amplicon was generated in a 20 μL reaction for 35 cycles with GoTaq G2 Green Master Mix (Promega).The PCR product was purified using the NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel).Thereafter, the isolated product was amplified for eight cycles using primers with sequencing adapters.After column-based isolation of the amplicon and quantification of DNA-yield using a Qubit 3.0 fluorometer and the dsDNA HS assay kit 392 (Thermo Fisher), paired-end sequencing was performed on an Illumina Miseq.The sequencing data was subsequently analyzed with CRISPResso2 (Clement et al., 2019).

Western blotting
Cells were lysed in RIPA buffer, supplemented with Protease Inhibitor (Cell Signaling Technologies) and PhosStop (Sigma-Aldrich) and centrifuged for 10 min at 21,000 g.The protein concentration of the supernatant was determined using Pierce BCA assay (ThermoFisher) and a standard curve of albumin.Then, samples were heated for 5' at 95°C in Lämmli buffer and protein lysates were resolved on polyacrylamide Mini-PROTEAN TGX gels (Bio-Rad) and transferred onto nitrocellulose membrane by wet-transfer.The following antibodies were used for immunoblotting: Recombinant anti-GADPH (EPR16891, Abcam) and rabbit monoclonal anti-MAP3K7 (anti-TAK1, D94D7, Cell Signaling Technology) IRDye-conjugated secondary antibodies (donkey anti-goat: LI-COR cat.no.926-32214; antirabbit: LI-COR cat.no.926-68073) were used for signal detection by an Odyssey Imager (LI-COR) imaging system.Raw gel images can be found in Source data 1.

Simmler
Figure 1.Sf3b1 K700E increases aggressiveness of murine PDAC.(A) Schematic overview of the Sf3b1 K700E/+ knockin model.Arrows indicate direction of the translatable code of the gene.An A to G conversion at base position 2098, induced by Cre-recombination, results in a lysine to glutamic acid mutation at amino acid position 700 of the protein.(B) Survival of WT and Sf3b1 K700E/+ mice followed over 300 days (n=8).(C) Pancreas weight of WT and Sf3b1 K700E/+ mice at 300 days of age (3 males and 3 females for each genotype).Two-tailed unpaired t-test was used to compute the indicated Figure 1 continued on next page testing.(L) Percentage of mice at 43 weeks of the indicated genotypes showing PanINs (blue) or PanINs and PDAC formation (red).p-Value indicates significance of the difference in PDAC formation, computed by Chi-square test.The online version of this article includes the following figure supplement(s) for figure 1:

Figure 4
Figure 4 continued on next page
Figure 5 continued