The Tumor Suppressor BAP1 Regulates the Hippo Pathway in Pancreatic Ductal Adenocarcinoma

Somatic mutations inactivate BAP1 in the majority of metastatic uveal melanomas (1) and in approximately 60% of malignant pleural mesotheliomas (2, 3). Some breast, lung, and renal cell cancers also harbor somatic BAP1 mutations (4–7). Germline BAP1 mutations cause a tumor predisposition syndrome characterized by benign melanocytic tumors (8) and mesothelioma, uveal melanoma, and other cancers (9, 10). The spectrum of tumors caused by BAP1 mutations continues to expand as more families with germline mutations are identified. BAP1 loss during mouse development is embryonic lethal, whereas BAP1 deficiency in the adult mouse causes a severe myeloproliferative disease reminiscent of chronic myelomonocytic leukemia (11). BAP1 is expressed ubiquitously (12) and is found in chromatin-associated complexes containing HCFC1, OGT, ASXL1/2, FOXK1/2, and KDM1b (11, 13–15). The Drosophila BAP1 ortholog Calypso influences histone dynamics by deubiquitinating histone 2A at lysine 118 (H2A-K118ub; ref. 14).

BAP1 has been linked to a multitude of biological processes, including DNA damage (16), cell cycle (17), gluconeogenesis (18), and metabolic homeostasis (19). However, the mechanism by which BAP1 suppresses tumorigenesis, the signaling pathways that are deregulated in BAP1-deficient tumors, and the critical BAP1 substrate(s) in various cancers remain largely unexplored.

In this study, we demonstrate a role for BAP1 as a tumor suppressor in pancreatic cancer. BAP1 loss cooperated with oncogenic Kras to form pancreatic ductal adenocarcinoma in mice. The Hippo pathway was deregulated in these BAP1-deficient tumors because negative regulator of the pathway, LATS2, was ubiquitinated and degraded. Hippo pathway deregulation leads to hyperactivation of the key downstream oncoproteins YAP and TAZ. Therefore, we identify YAP/TAZ repression as a potential therapeutic strategy for certain BAP1-deficient tumors.

Bap1^fl/fl mice crossed to the inducible general deleter C567BL/6 NTac- Gt(ROSA)26Sor^{tm9(Cre/ESR1)Arte} were described previously (11). CreERT2⁺ mice aged 6–8 weeks were injected intraperitoneally with 60 mg/kg tamoxifen dissolved in sunflower oil daily for 5 days. To generate the pancreatic-specific deletion, the Bap1^fl/fl mice were crossed to Pdx1.cre mice (20). The Genentech Institutional Animal Care and Use Committee approved all protocols.

Bap1 genotyping primers 5′ CCA TCA GTG ACT ACT GGG GAG CAA C, 5′ ACA GAT GGC TGG GCA CAT CTG, and 5′ GAA CCC TCC GTT GCA TAG TGT TG amplified 234 bp wild-type (WT), 350 bp floxed, and 503 bp knockout (KO) DNA fragments. Cre primers 5′- GCT AAA CAT GCT TCA TCG TCG GTC and 5′-CCA GAC CAG GCC AGG TAT CTC TG amplified a 582 bp DNA fragment.

Recipient animals, Bap1^fl/fl;Rosa26.creERT2⁺ or Bap1^wt/wt;Rosa26.creERT2⁺, received 2 doses of 525 Rads from a ¹³⁷Cs source separated by a 4-hour interval. Donor bone marrow cells from Bap1^wt/wt animals were injected into the tail vein. Reconstituted mice were given water containing 0.11 mg/mL polymyxin B and 1.1 mg/mL neomycin for 2 weeks and then switched to regular water.

To generate pancreatic ductal adenocarcinoma (PDAC) BAP1 KO tumor–derived cell lines, the tumor tissue was minced with a pair of scalpels in a 10-cm Petri dish containing 20 mL RPMI + 10% FBS. Tissue fragments with the medium were transferred to a T75 tissue culture flask and placed into a 37°C, 5% CO₂ incubator. After 1 or 2 days to allow for cellular attachment, the medium was removed and replaced with fresh growth medium. Contaminating fibroblasts were removed by differential trypsinization. Once monolayer cultures were established and passaged a few times, the growing cultures were passaged by trypsinization at appropriate intervals and split ratios. BAP1 WT reconstituted PDAC and H226 cells were generated by transducing cells with lentiviral BAP1 construct for 72 hours, followed by selection with puromycin.

ME202, 92-1, OMM1.3, UPMM-2, and UPMM-3 cells have been described previously (21). MP38 and MP46 cells were kindly provided by Sergio Roman-Roman (Institut Curie, Paris, France; ref. 22). All other cell lines were from the ATCC. Cell lines were maintained in RPMI1640 or DMEM/F12 (Gibco) supplemented with 10% FBS (Sigma), 50 U/mL penicillin, 50 U/mL streptomycin, and 2 mmol/L l-glutamine (Gibco). Doxycycline was purchased from BD and cycloheximide was from Calbiochem.

Cell lines used were checked at the Genentech core lab by the following methods.

Short tandem repeat (STR) profiles were determined for each line using the Promega PowerPlex 16 System. This is performed once and compared with external STR profiles of cell lines (when available) to determine cell line ancestry.

SNP profiles are performed each time new stocks are expanded for cryopreservation. Cell line identity is verified by high-throughput SNP profiling using Fluidigm multiplexed assays. SNPs were selected on the basis of minor allele frequency and presence on commercial genotyping platforms. SNP profiles are compared with SNP calls from available internal and external data (when available) to determine or confirm ancestry.

All stocks were tested for Mycoplasma prior to and after cells are cryopreserved. Two methods are used to avoid false positive/negative results: Lonza Mycoalert and Stratagene Mycosensor. Cell growth rates and morphology are also monitored for any batch-to-batch changes.

For coimmunoprecipitation experiments, HEK293T cells were transfected with Flag- tagged BAP1 or LATS2 constructs for 72 hours followed by cell lysis using RIPA buffer and immunoprecipitation with FLAG-M2 agarose beads (Sigma) for overnight at 4°C. After washing with RIPA buffer, coimmunoprecipitated endogenous proteins, such as LATS1, 2, MOB1, and BAP1 were then detected by immunoblotting.

For immunoblotting, cells were harvested in RIPA lysis buffer containing protease and phosphatase inhibitors cocktail (Roche). Lysates were then analyzed for immunoblotting. Antibodies directed against BAP1, LATS1, LATS2, YAP/TAZ, pYAP, MOB1, MST1, MST2, NF2, GAPDH, α-tubulin, and β-actin were purchased from Cell Signaling Technology. MAX antibody was from Santa Cruz Biotechnology.

Cells grown on 10-cm dishes were harvested after washing with cold PBS. Cell pellets were then incubated with 400 μL of Buffer A [10 mmol/L HEPES pH 7.9, 10 mmol/L KCl, 1 mmol/L EDTA, 0.1 mmol/L EGTA, 0.2% NP-40, 10% glycerol] on ice for 20 minutes. Cytoplasmic fraction was collected by taking the supernatant from centrifugation for 30 seconds at 1,400 × g. For the preparation of nuclear fraction, the pellet was then incubated with 200 μL of RIPA buffer for 30 minutes at 4°C. After centrifugation for 12 minutes at 16,000 × g, nuclear fraction was collected by retaining the supernatant.

All siRNA oligos and a nonspecific nontargeting control were purchased from Dharmacon. siRNA oligos were transfected into cells for 48 hours using the Lipofectamine RNAiMAX reagent (Invitrogen).

For RNA preparation, tumors were dissociated and lysed for RNA isolation using RNeasy kit (Qiagen). For quantitative PCR, cDNA was prepared by reverse transcription and PCR was performed using TaqMan probes for YAP, TAZ, CTGF, CYR61, LATS1, LATS2, and HPRT (Life Science). Relative expression to HPRT of target genes was assessed. The siRNA and probe information used in this study are listed in Supplementary Table S1.

For the in cell deubiquitination assay, Myc-Flag-human LATS2 and HA-ubiquitin (ub) plasmids were transfected into HEK293T cells at a ratio of 1:1.5 with Fugene 6. After 48 hours, cells were lysed in NP-40 buffer [1% NP-40, 120 mmol/L NaCl, 50 mmol/L Tris, pH 7.4, 1 mmol/L EDTA, pH 7.4, 20 mmol/L cysteine protease inhibitor N-ethylmaleimide (NEM), protease and phosphatase inhibitors (Roche)]. Proteins were denatured with 1% SDS and then diluted to 0.05% SDS prior to anti-HA immunoprecipitation.

For the in vitro deubiquitination assay, ubiquitinated LATS1 and 2 were isolated by transiently cotransfecting HEK293T cells with MYC-LATS1/MYC-LATS2 and V5-ubiquitin as described above. BAP1 protein was purified from BAP1-overexpressing HEK293T cells by immnoprecipitation using anti-BAP1 antibody. In vitro reaction was performed by incubating poly-ub-LATS with BAP1 in 2× reaction buffer (100 μmol/L Tris pH 8.0, 40 mmol/L NaCl, 200 μg/mL BSA, 8 μmol/L EDTA, pH 8.0, 4 mmol/L DTT) at 25°C for indicated times. Reaction was terminated by adding SDS sample buffer followed by 5-minute boiling at 95°C. The result was assessed by immunoblot analysis.

Histopathologic analyses were performed on routinely processed formalin-fixed paraffin-embedded tissue sections. Antibodies used for IHC included carboxypeptidase A1 (R&D Systems goat polyclonal, 0.05 μg/mL), cytokeratin 19 (University of Iowa rat monoclonal TROMA III, 10 μg/mL), HES1 (MBL rat monoclonal NM1, 1 μg/mL), cleaved notch 1 (Val1744; Cell Signaling Technology, rabbit monoclonal D3B8, 10 μg/mL), Ki-67 (NeoMarkers rabbit monoclonal SP6, 1:200 dilution), YAP1 (Cell Signaling Technology, rabbit polyclonal, 0.036 μg/mL), CD3 (Thermo Fisher Scientific, rabbit monoclonal SP7, 1:200 dilution), TAZ (Sigma rabbit polyclonal, 0.25 μg/mL), BAP1 (Genentech, rabbit monoclonal 2G12, 1 μg/mL), CD68 (Serotec rat monoclonal FA-11, 5 μg/mL), insulin (Dako guinea pig polyclonal, 5.5 μg/mL), glucagon (Cell Signaling Technology, rabbit polyclonal, 0.0417 μg/mL), and cleaved caspase 3 (Asp175; Cell Signaling Technology, rabbit monoclonal 9661L, 0.05 μg/mL). Cleaved caspase-3, Ki-67, and carboxypeptidase A1 IHC was performed on the Ventana Discovery XT platform with CC1 standard antigen retrieval (Ventana) and OmniMap detection (Ventana). Cleaved notch IHC was also performed on the Ventana Discovery XT platform with OmniMap detection, but with CC1 extended retrieval. Cytokeratin 19, CD3, CD68, and glucagon IHC was performed on a Dako autostainer with Target (Dako) antigen retrieval and ABC-Peroxidase Elite (Vector Laboratories) detection. Insulin IHC was performed on the Dako autostainer with ABC- Peroxidase Elite detection, but with no antigen retrieval. TAZ and BAP1 IHC was also performed on a Dako autostainer with Target retrieval, but detection was performed with PowerVision polymer-HRP (Leica Biosystems). HES-1 and YAP IHC were performed on the Dako autostainer with TSA- HRP detection. EDTA pH8 (LabVision) heat-induced epitope retrieval was used for YAP and Target retrieval was used for HES-1. 2,3-diaminobenzidine (DAB) was used as the chromogen for all studies and hematoxylin was used as a counterstain. Isotype control antibodies were used as negative control and positive control tissues and/or cell lines were used to validate each antibody. For quantitative image analysis, slides were digitally scanned using a Hamamatsu slide scanner and immunolabeling was quantified using MatLab (MathWorks) software package.

BAP1 loss in adult mice causes lethal myeloid dysplasia within 4–6 weeks (11). To investigate the role of BAP1 in nonhematopoietic cells, we used tamoxifen to induce BAP1 loss in Bap1^fl/fl Rosa26^CreERT2/+ chimeric mice that had their hematopoietic system reconstituted with wild-type bone marrow (hereafter referred to as BAP1 BMC-KO). Bap1^+/+ Rosa26^CreERT2/+ mice reconstituted with wild-type bone marrow (hereafter referred to as BAP1 BMC-WT) served as negative controls (Fig. 1A; ref. 19). Zero out of 28 BAP1 BMC-KO mice analyzed between 3 and 10 months after tamoxifen treatment had CMML-like disease. After 3 months, however, all BAP1 BMC-KO mice (n = 11) exhibited pancreatic acinar atrophy (loss of acinar tissue) with relatively increased duct profiles (possibly due to parenchymal collapse or acinar ductal metaplasia), which was not associated with acute inflammatory infiltrates (Fig. 1B) and multifocal hepatocellular death (individual shrunken and hypereosinophilic hepatocytes) with regenerative changes (variation in hepatocellular size, mitotic activity) that disrupted the hepatic architecture (Supplementary Fig. S1A). In addition, 2 of the 11 BAP1 BMC-KO mice had tumors. One mouse had an intestinal sarcoma (Supplementary Fig. S1B) and the other had a pulmonary adenoma (Supplementary Fig. S1C). BAP1 BMC-WT controls (n = 9) examined at 3 months after tamoxifen treatment had no significant histologic lesions or tumors.

Another cohort of mice was analyzed at 5–7 months after tamoxifen treatment. All BAP1 BMC-KO mice (n = 12) had the aforementioned pancreatic and liver lesions, plus 5 out of 12 had evidence of bile duct hyperplasia (Fig. 1C) and 1 out of 12 had a hemangioma- like lesion (Supplementary Fig. S1D). In contrast, the only finding in the BAP1 BMC-WT control group (n = 13) was one mouse with minimal bile duct hyperplasia. Interestingly, at 10 months after tamoxifen treatment, 2 of 5 BAP1 BMC-KO mice had multiple tumors. One mouse had both a colonic adenoma and an ovarian granulosa cell tumor, while the second mouse had a hair follicle tumor, cutaneous hemangiosarcoma, and pulmonary adenoma (Fig. 1D). The latter mouse also had a small lobular proliferative salivary gland lesion with epithelial piling that was associated with chronic inflammation (Supplementary Fig. S1E), but it was difficult to distinguish between adenoma versus dysplastic regenerative hyperplasia. A vascular lesion characterized by dilated vascular channels replaced the ovary of a third mouse, which was consistent with angiectasis or hemangioma (Supplementary Fig. S1F). In the BAP1 BMC-WT control group, 1 of 5 mice had pancreatic acinar atrophy and a hepatocellular adenoma. The liver from this mouse was confirmed to express BAP1 protein by IHC, suggesting it is less likely to represent a genotyping error (Supplementary Fig. S1G). In summary, tumors were observed at varying times in 5 of 28 BAP1 BMC-KO mice compared with 1 of 27 BAP1 BMC-WT controls. The tumor incidence in the BAP1 BMC-KO mice was lower than expected, but this could reflect variable deletion of the Bap1 gene from different tissues. For example, Bap1 mRNA is eliminated quite efficiently from liver, but is only moderately reduced in lung, heart, and kidney (19).

Given that all BAP1 BMC-KO mice developed exocrine pancreatic atrophy within 4 weeks of tamoxifen treatment (Fig. 1B), we explored the consequences of BAP1 deficiency in pancreas in more detail. Initial hypereosinophilia of acinar cells was followed by progressive lobular acinar cell loss with an associated relative increase in duct profiles (Fig. 1B). BAP1- deficient pancreas contained increased numbers of cleaved caspase-3–expressing apoptotic cells (Supplementary Fig. S2A), suggesting that BAP1 is required for acinar cell survival. Increased cell proliferation as noted by Ki67 immunolabeling was also observed in BAP1-deficient pancreas, which was interpreted to represent a compensatory response to the increased cell death (Supplementary Fig. S2B). Despite the dramatic changes in the exocrine pancreas, BAP1 loss did not impact the number or distribution of insulin-positive beta cells or glucagon-positive alpha cells within pancreatic islets (Fig. 1E). Keratin 19 and carboxypeptidase A1 also were expressed normally in ductal cells and in acinar cells, respectively (Fig. 1E). Bap1^fl/fl mice bearing a Pdx1.cre transgene for pancreas-specific gene deletion also exhibited acinar atrophy by 6–8 weeks of age, indicating an essential role for BAP1 in pancreas development and/or homeostasis. Keratin 19 and carboxypeptidase A1 expression was unaltered in ducts and remaining acinar cells, respectively, and there was increased expression of the proliferation marker Ki-67 (Fig. 1F).

BAP1 mutations in patients are mostly found in the background of a sensitizing oncogenic mutation. To evaluate the role of BAP1 as a tumor suppressor in the pancreas, we compared Bap1^+/+ Kras^lsl/+ Pdx1.cre and Bap1^fl/fl Kras^lsl/+ Pdx1.cre mice that carried a conditional allele of mutant Kras G12D (Fig. 2A). At 8–12 weeks of age, 13 of 13 Bap1^+/+ Kras^lsl/+ Pdx1.cre mice had multifocal, grades 1–3 pancreatic intraepithelial neoplasia (PanIN) lesions (Fig. 2B and C), consistent with previous reports (20). In contrast, 6 of 8 (75%) age-matched Bap1^fl/fl Kras^lsl/+ Pdx1.cre mice had histologic lesions that represented a continuum from ductal hyperplasia to PanIN to PDACs in the same mouse. While some ducts were increased in size with dilated lumina, but maintained an organized single epithelial layer, other areas, sometimes of the same duct, were characterized by irregular epithelial piling, loss of luminal architecture, and irregular extension into the stroma suggestive of PDAC (Supplementary Fig. S2C). Mixed inflammatory infiltrates were present in the tumor-associated stroma (Supplementary Fig. S2D). Although there was some variability in the size and relative distribution of hyperplasia to PanIN to PDAC in individual lesions, tumors in the BAP1-deficient mice resulted in significant effacement of the normal pancreatic architecture (Fig. 2B; Supplementary Fig. S2C). One Bap1^fl/fl Kras^lsl/+ Pdx1.cre mouse was histologically normal and may represent a genotyping error. Keratin 19 expression was restricted to normal and neoplastic ductal tissues in all genotypes examined, confirming the classical ductal phenotype and origin of the PanIN and PDAC lesions. Similar to BAP1 BMC-KO and Bap1^fl/fl Pdx1.cre mice, Bap1^+/+ Kras^lsl/+ Pdx1.cre PanIN lesions, and Bap1^fl/fl Kras^lsl/+ Pdx1.cre tumors contained many cells expressing Ki-67 (Fig. 2D). BAP1 deficiency also increased the incidence of squamous papillomas in Kras^lsl/+ Pdx1.cre mice (Supplementary Fig. S3A and S3B). All of our comparison was done across aged-matched animals. These tumors consisted of papillary masses lined by hyperplastic stratified squamous epithelium, often with hyperkeratosis. Some papillomas had associated sebaceous gland hyperplasia. Collectively, these data are consistent with BAP1 functioning as a tumor suppressor.

Given that BAP1 loss potentiated Kras-driven PDAC in mice, we evaluated 50 human PDAC samples for evidence of BAP1 loss by IHC (Supplementary Fig. S4A). Normal human pancreas contained nuclear BAP1 labeling in acinar cells, islets, ducts, and some stromal cells. Three out of 50 PDAC samples had weak or inconsistent labeling throughout the sample, including the stroma, which could indicate fixation issues; therefore, these samples were excluded from further analyses. Thirty-eight of the remaining 47 (81%) PDAC samples had moderate to strong BAP1 labeling in nearly all tumor cells. However, in 8 tumor samples (17%), there was weaker labeling in subsets of tumor cells often with intermingled weak and moderately labeled cells. This heterogeneity in labeling suggests that there is decreased BAP1 expression in at least a subset of the tumor cell population. Interestingly, 1 of the 47 (2%) tumors had complete loss of BAP1 expression in the tumor cells, but maintained strong labeling in the adjacent stroma. While the significance of heterogeneous BAP1 expression in PDAC needs to be evaluated further, it is noteworthy that decreased BAP1 expression in patients with gastric and colorectal cancer is associated with poor prognosis (23, 24). Copy number and mRNA expression analysis of patients with pancreatic adenocarcinoma from The Cancer Genome Atlas (n = 149) revealed recurrent shallow deletion of BAP1 (28%, n = 42; Supplementary Fig. S4B and S4C; refs. 25, 26). Furthermore, samples with shallow copy number deletion of BAP1 were significantly associated with lower BAP1 mRNA expression (Supplementary Fig. S4B and S4C), suggesting partial loss of BAP1 gene may lead to decrease BAP1 abundance. Tumors with low BAP1 expression defined as those with homozygous deletions and/or BAP1 mRNA expression less than 1.5 SDs below the mean were associated with poorer survival (Supplementary Fig. S4D). Taken together, this suggests that shallow deletion of BAP1 may contribute to decreased BAP1 expression and subsequently to the pathogenesis of a subset of human PDACs. Therefore, BAP1 loss may also contribute to human PDAC.

Notch signaling is implicated in the progression of PanIN lesions and malignant transformation of the pancreas (27, 28). Therefore, we determined whether pancreatic BAP1 loss enhanced Notch signaling by IHC for the Notch intracellular domain (NICD) and the Notch target gene HES1. Although KrasG12D-expressing pancreas expressed more HES1 than wild-type pancreas, BAP1 deficiency did not impact expression of HES1 or nuclear NICD (Supplementary Fig. S5A). Therefore, BAP1 loss does not appear to activate Notch signaling in the pancreas. Furthermore, we evaluated a KrasG12D/Bap1 KO cell line with and without BAP1 reconstitution to determine whether BAP1 expression influenced TGFβ signaling. By Western blotting, pSMAD2/3 was similar in cells with and without BAP1 reconstitution, suggesting that these transcriptional changes are not due to changes in TGFβ signaling (Supplementary Fig. S5B).

Next, we determined whether BAP1 loss altered Hippo signaling because transcriptional coactivators YAP and TAZ, which are key effectors of the Hippo pathway, are increased significantly in human PDAC (29, 30). YAP is essential for progression to PDAC in the context of oncogenic KrasG12D and p53 deficiency in the mouse (30). In addition, pancreatic loss of the tumor suppressor genes Mst1 and Mst2 in the Hippo pathway causes exocrine pancreatic atrophy (31) and therefore bears some resemblance to pancreatic Bap1 deletion. Intriguingly, BAP1-deficient PDACs exhibited more intense YAP1 and TAZ expression than the KrasG12D PanIN lesions (Fig. 3A). Immunoblotting confirmed that the Bap1^fl/fl Kras^lsl/+ Pdx1.cre pancreas expressed more YAP and TAZ than the Bap1^+/+ Kras^lsl/+ Pdx1.cre pancreas (Fig. 3B). Increased protein expression appeared to be due to either transcriptional or posttranslational events since Yap mRNA level was not changed while Taz expression was increased (Fig. 3C). Expression of two Hippo target genes Ctgf and Cyr61 was also upregulated in the BAP1-deficient PDACs (Fig. 3C), reflecting activation of Hippo signaling. We speculated that deregulation of the Hippo pathway might render tumor cells derived from BAP1-null PDACs sensitive to knockdown of the key pathway effectors, YAP and TAZ, and their cognate transcription factors TEAD1-4. Indeed, we observed decreased cell viability after knockdown of YAP/TAZ and TEAD1-4, suggesting that the cells are dependent on YAP, TAZ, and TEADs for their survival (Fig. 3D). Next, we addressed whether reconstitution of BAP1 alters the dependency of these cell lines on these factors. BAP1 reconstitution partially, but significantly rescued the cell death phenotype associated with knockdown of YAP/TAZ or TEAD1-4 (Fig. 3D). Collectively, our data suggest that BAP1 loss leads to activation of Hippo signaling that drives cell growth in PDAC.

Hippo pathway activity is controlled by a negative feedback loop, whereby YAP and TAZ (which are repressed by the Hippo pathway) induce expression of upstream pathway members including LATS1/2 (32, 33). We investigated whether BAP1 loss affects levels of any of the components in the Hippo kinase cascade. LATS1, LATS2 and MOB1 within the core kinase complex were all markedly reduced in the BAP1-deficient KrasG12D PDACs as compared with the KrasG12D PanIN lesions (Fig. 4A). However, other upstream regulators of the Hippo pathway MST1, MST2, and NF2 did not exhibit differential expression. The residual BAP1 expression in the BAP1 KO samples is likely due to incomplete deletion and/or infiltrating wild-type stroma in the pancreatic tissue. Interestingly, while Lats1 expression was not changed, Lats2 mRNA was significantly increased despite its protein level being decreased in BAP1-deficient PDACs, which is suspected to be due to hyperactivation of YAP/TAZ as a previous study has shown that YAP/TAZ selectively induce Lats2 expression (Fig. 4B; ref. 34). Next, we assessed whether LATS1/2 might be substrates for BAP1-mediated deubiquitination and stabilization. To address this, we evaluated the stability of LATS1 and LATS2 in a BAP1-deficient tumor cell line (KT4) derived from Kras^G12D/BAP1^fl/fl mice before and after reconstitution with BAP1. LATS2 was stabilized at the protein level by ectopic expression of BAP1 (Fig. 4C and D), whereas Lats2 transcripts were not increased (Fig. 4E). In contrast, LATS1 abundance was unaffected by reconstitution of BAP1 expression. Surprisingly, LATS1 expression was not affected by 24 hours of cycloheximide treatment regardless of BAP1 expression (Fig. 4C). It has been reported that half-life of LATS1 is longer than LATS2 under certain conditions, and this further suggested that BAP1 could differentially stabilize LATS2 to regulate Hippo signaling (35, 36). Overall, our data suggest that BAP1 directly regulates LATS2 abundance in tumor cells, whereas the effect seen on LATS1 in vivo could be linked to other BAP1-independent mechanisms that might amplify the signaling to accelerate tumorigenesis during PDAC progression. Because LATS1 influences YAP and TAZ localization and activity but not the level of expression, the rapid increase in YAP and TAZ expression in our tumors could be a result of their rapid development instead of a direct consequence of BAP1 loss.

Given that BAP1 increased the amount of LATS2 in the PDAC tumor cell line, we next examined whether this required the catalytic activity of BAP1. BAP1-deficient cells contained more LATS2 after reconstitution with wild-type BAP1, but not after reconstitution with the catalytically inactive BAP1-mutant C91A (Fig. 5A). Interestingly, subcellular fractionation suggested that BAP1 deubiquitinating activity largely altered LATS2 abundance (Fig. 5B and D). Coimmunoprecipitation (IP) experiments in 293T cells detected an interaction between endogenous LATS1, LATS2, MOB1, and ectopic flag-tagged wild-type BAP1, but this was not seen using BAP1 C91A (Fig. 5E). Endogenous BAP1 in 293T cells also interacted with flag-tagged LATS2 (Fig. 5F). It is worth noting that BAP1 is a nuclear DUB although our ectopic, reconstituted BAP1 localizes both in the nucleus and cytoplasm (37). Thus, we examined where endogenous BAP1/LATS interaction occurs using a relevant pancreatic cancer cell line. In PaTu8988T cells, which express relatively high levels of BAP1 and LATS, BAP1 expression as well as BAP1/LATS interaction was observed only in the nuclear fraction while LATS1/2 was expressed in both fractions (Fig. 5G). Together with our observation that BAP1 reconstitution specifically affects LATS2 expression, these data suggest that endogenous BAP1 regulates LATS2 in the nucleus. Interestingly, BAP1 interacts with both LATS1 and 2, although BAP1 only affects LATS2 protein abundance (Fig. 5E–G). To investigate the differential effect of LATS1 and LATS2 on the BAP1- LATS1/LATS2 complex formation and function, we performed nuclear BAP1 coimmunoprecipitation under either LATS1 or 2 depleted conditions in 293T cells. Interestingly, LATS2 depletion significantly reduced YAP association to the complex while loss of LATS1 resulted in a milder reduction of YAP association (Fig. 5H).

This led us to examine the kinase activity of LATS on YAP in basal or serum-starved conditions. Consistently, doxycycline-induced BAP1 increased pYAP expression via increased LATS2, but not by LATS1 (Fig. 5I). Conversely, knockdown of LATS decreased YAP phosphorylation in a LATS2-dependent manner. Given the known role of serum starvation in YAP phosphorylation, we further demonstrated that upon BAP1 reconstitution, LATS2-dependent YAP phosphorylation was increased by serum starvation (Fig. 5J). These data suggest that nuclear BAP1/LATS2 regulation of YAP phosphorylation has functional relevance in vitro. To assess whether BAP1 is a direct DUB for LATS1 and LATS2, we performed an in-cell deubiquitylation assay using HA-tagged ubiquitin with IP followed by Western blot analysis to detect ubiquitylated LATS1/2 and WT BAP1 or catalytic dead BAP1 C91A. WT BAP1, but not BAP1 C91A, reduced ubiquitylation of cotransfected LATS2 and to a somewhat lesser extent LATS1 (Fig. 5K). Furthermore, the in vitro deubiquitylation assay with purified ubiquitylated LATS1 or -2 and purified BAP1 demonstrated that BAP1 directly deubiquitylates LATS2 while ubiquitylated LATS1 was not as significantly impacted (Fig. 5L). Notably, the ubiquitylation efficiency on LATS2 was much greater than on LATS1 in vitro. This might also explain the difference between the stability of endogenous LATS1 and LATS2 in the cycloheximide chase experiment in Fig. 4C. These data are consistent with LATS2 being a BAP1 substrate, although this might be context dependent as well because LATS1 and LATS2 can also compensate for each other.

BAP1 is mutated in a significant fraction of uveal melanomas and malignant mesotheliomas (1, 9). In addition, BAP1 and NF2 mutations in mesothelioma are often mutually exclusive (38). Therefore, we investigated whether BAP1 regulation of LATS2 abundance is also observed in mesothelioma cells. BAP1-deficient NCI-H226 mesothelioma cells expressed negligible LATS2 protein, but after reconstitution with WT BAP1, LATS2 was readily detected (Fig. 6A). We speculated that deregulation of the Hippo pathway in the NCI-H226 cells might render them sensitive to knockdown of the key pathway effectors, YAP and TAZ, and their cognate transcription factors TEAD1-4. Indeed, viability of parental NCI-H226 cells was reduced upon knockdown either of YAP, TAZ, or TEAD1-4 (Fig. 6B and C). We also examined whether reconstitution of BAP1 affects the dependency on these factors. BAP1 reconstitution completely rescued cells from single YAP or TAZ knockdown and partially, but significantly, rescued cells from combinatorial knockdown of YAP/TAZ or TEAD1-4 (Fig. 6B and C). Consistently, these data suggest that BAP1 plays a role in regulating Hippo signaling through LATS2 stabilization in mesothelioma.

Given the links of BAP1 mutation in uveal melanoma, we next investigated whether BAP1 loss correlated with increase in expression of YAP and TAZ in uveal melanoma cell lines. Considerable variability in YAP and TAZ levels was observed when immunoblotting uveal melanoma cell lines that either expressed BAP1 or had undetectable BAP1 (Supplementary Fig. S6), although some BAP1-mutant cell lines had higher expression of YAP and TAZ and also lower expression of LATS2.

In this study, we show that BAP1 is essential for normal pancreatic acinar tissue architecture (Fig. 1), and that its loss cooperates with KrasG12D to cause PDAC in mice (Fig. 2). At the molecular level, BAP1-deficient pancreatic tumors appear to coopt the Hippo pathway, resulting in increased expression of the transcriptional coactivators YAP and TAZ (Figs. 3 and 4). We present data that is consistent with BAP1 deubiquitylating LATS2, resulting in stabilization of the LATS kinase complex that targets YAP and TAZ for proteasomal degradation (Figs. 5 and 6). Even though we observe significant deregulation of the Hippo pathway in BAP1-deficient pancreatic tumors, the canonical role of BAP1 in deubiquitination of H2A may also play a role in the pathology. Future studies are needed to address differential roles of BAP1 as a DUB in pancreatic cancers.

YAP and TAZ are upregulated in a large fraction of human PDAC tumors (29). In addition, YAP activation was identified as a key resistance mechanism by which pancreatic, colon, and lung cancers bypass addiction to oncogenic Kras (39, 40). Although YAP is dispensable for initiation of pancreatic tumors and acinar ductal metaplasia in mice, it is critical for tumor progression to invasive PDAC and tumor maintenance (30). However, the detailed underlying mechanism remains unknown. There are multiple paths to deregulated Hippo signaling in cancer, including amplification of the effectors YAP and TAZ or loss of upstream tumor suppressors. Our study suggests that loss of the tumor suppressor BAP1 is another mechanism by which the Hippo pathway may be deregulated in cancer.

We demonstrate that BAP1 KO pancreatic tumors maintain high levels of YAP and TAZ, which probably contribute to tumor cell proliferation and survival. While ubiquitination of YAP and TAZ (41, 42) and nuclear ubiquitination of LATS (43) has been implicated in regulation of the Hippo pathway, the role of deubiquitination in the context of cancer has been largely unexplored.

The development and progression of PDAC is strongly influenced by the microenvironment, fibroblasts, and infiltrating immune cells (44, 45). It is noteworthy that BAP1-deficient PDACs in vivo exhibited a marked decrease in LATS1 and LATS2, whereas only the stability of LATS2 was altered in tumor cell lines in vitro. This suggests a possible tumor cell intrinsic role of BAP1 in regulating LATS2.

Two previous studies demonstrated infrequent BAP1 loss in PDAC ranging from 0.33 to 2.4% (46, 47). Similarly, we found BAP1 loss in 2% of the human PDAC samples evaluated (Supplementary Fig. S4A and S4B). When we investigated copy-number alterations, we identified a correlation between the presence of shallow deletions and decreased BAP1 expression in PDAC (Supplementary Fig. S4B and S4C). In additon, we observed weak BAP1 immunolabeling in 17% of the PDAC samples evaluated, which is comparable with another study that identified weak BAP1 expression in 13.5% of PDACs (Supplementary Fig. S4A; ref. 47).

Previously, USP9X has been shown to regulate Hippo signaling pathway as a broad DUB for angiomotin, Kibra, WW45, and LATS2 (48–50). Under physiologic conditions, BAP1 is mainly localized in the nucleus (37). Here, we highlight the context-dependent roles of DUBs and their spatial regulation. We propose that BAP1 directly regulates nuclear LATS2 abundance in tumor cells. In addition, BAP1-deficient pancreatic tumor cells might overcome the intrinsic control of YAP/TAZ activity by disabling the LATS-dependent negative feedback loop (34).

In conclusion, our results identify a tumor suppressor function for BAP1 in pancreatic cancer and link the molecular mechanism to suppression of the Hippo pathway. We propose that BAP1 deficiency deregulates the negative feedback loop in the Hippo pathway by deubiquitinating LATS2. Future studies will address whether BAP1 deficiency coopts the Hippo pathway in additional cancers and promotes tumor development leading to poor prognosis. The Hippo pathway has emerged as a promising target in oncology and several small-molecule inhibitors are currently at various stages of development. Targeting the Hippo pathway could provide therapeutic opportunities for certain BAP1-deficient tumors like uveal melanoma, malignant mesothelioma, and pancreatic tumors.

All Genentech authors are shareholders in Roche. M. Carbone has pending patent applications on BAP1 and provides consultation for mesothelioma diagnosis. No potential conflicts of interest were disclosed by the other author.

Conception and design: H.-J. Lee, T. Pham, J.D. Webster, A. Dey

Development of methodology: H.-J. Lee, T. Pham, R. Noubade, J.D. Webster

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.-J. Lee, T. Pham, M.T. Chang, D. Barnes, A.G. Cai, R. Noubade, X. Chen, C. Tran, T. Hagenbeek, X. Wu, W. Lee, J.D. Webster, A. Dey

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-J. Lee, M.T. Chang, A.G. Cai, C. Tran, J. Eastham-Anderson, W. Lee, B.C. Bastian, M. Carbone, J.D. Webster, A. Dey

Writing, review, and/or revision of the manuscript: H.-J. Lee, T. Pham, A.G. Cai, M. Carbone, J.D. Webster, A. Dey

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Pham, A.G. Cai, K. Totpal, J. Tao

Study supervision: H.-J. Lee, W. Lee, A. Dey

We thank Vishva Dixit and Kim Newton for discussions and critical reading of the manuscript. Thanks to members of the Dixit and Dey laboratories for advice and discussions and core laboratories for technical assistance. This research was also supported by an R01 grant CA142873 from the NCI and a Stein Innovation Award from Research to Prevent Blindness and the Gerson and Barbara Baker Distinguished Professorship (to B.C. Bastian), and a Young Investigator Award from the Melanoma Research Alliance (to X. Chen). This work was also supported in part by NCI R01 CA198138 and by the University of Hawai'i Foundation, which received unrestricted donations to support mesothelioma research from Honeywell International Inc. (to M. Carbone).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.