Abstract
Objective Yap1 oncogene is essential for KRAS-induced pancreatic ductal adenocarcinoma (PDAC) initiation. We recently demonstrated that YAP1 is capable of bypassing KRAS-dependency. However, genetic alterations of YAP1 pathway are rare in human PDAC and its role in tumor maintenance remains unclear. This study investigates YAP1 function in transcriptionally distinct subsets of PDAC and explores the molecular mechanisms for YAP1 activation.
Design Conditional Yap1 allele was crossed into LSL-KrasG12D/+; Trp53R172H/+; Mist1-CreERT2+ mice to study the requirement of Yap1 for PDAC development in adult mice. YAP1 function in advanced PDAC was analyzed in human tissues, PDAC cell lines, patient-derived xenograft samples and mouse PDAC cells, using in vitro and in vivo assays and gene expression analyses. WNT5A expression and its effect on YAP1 activation as well as tumorigenic activity was studied in aforementioned human and mouse systems using genetic and pharmacological approaches. The bypass of KRAS-dependency by WNT5A was evaluated in KrasG12D-driven mouse PDAC cells.
Results Yap1 deletion blocked KrasG12D/Trp53R172H/+ induced PDAC development in mice. YAP1 activation correlated with squamous subtype of human PDAC and poor prognosis. YAP1 activation enhanced malignant phenotype and YAP1 depletion suppressed the tumorigenic activity specifically in squamous subtype tumors. WNT5A was significantly overexpressed in squamous subtype tumors and positively correlated with YAP1 activity. Conversely, WNT5A depletion resulted in YAP1 inactivation and inhibition of tumorigenic activity. WNT5A expression enabled cell survival and tumor growth in the absence of oncogenic Kras.
Conclusions YAP1 oncogene is a major driver for squamous subtype PDAC whose activation is mediated by WNT5A overexpression.
Significance of this study What is already known on this subject?
Pancreatic ductal adenocarcinoma (PDAC) is recently classified into distinct molecular subtypes with the squamous subtype correlated with worst prognosis.
The molecular subtypes of PDAC are based on gene expression and are not consistently associated with any genetic alterations, such as KRAS mutations.
YAP1 oncogene is required for oncogenic KRAS-induced PDAC development.
YAP1 amplification in genetically engineered mouse model of PDAC enables the bypass of KRAS-dependency.
What are the new findings?
YAP1 is highly activated in squamous subtype PDAC and is required for tumorigenic activity.
YAP1 activation in squamous subtype PDAC is mediated by WNT5A overexpression.
WNT5A overexpression in PDAC tumor cells bypasses KRAS-dependency.
How might it impact on clinical practice in the foreseeable future?
YAP1 activation signature may serve to stratify PDAC patients for YAP1 targeted therapy.
The WNT5A-YAP1 axis can serve as therapeutic target for squamous subtype PDAC with WNT5A overexpression and YAP1 activation.
Tumors with YAP1 activation may quickly develop resistance to KRAS targeted therapy.
Introduction
Pancreatic ductal adenocarcinoma, expected to become the second most common cause of cancer-related mortality in the US by 20301, is a highly heterogeneous disease with complex genetic and molecular diversity. While majority of PDACs share near-ubiquitous oncogenic mutations of KRAS and the frequent inactivation of TP53, SMAD4 and CDKN2A tumor suppressors, most additional somatic mutations occur at low individual prevalence, suggesting diverse mechanisms underlying PDAC progression2. Although others and we have shown that KRAS oncogene is essential for tumor maintenance3,4, effective targeting of KRAS pathway is yet to be achieved clinically. Moreover, recent studies have revealed a subpopulation of PDAC cells may bypass KRAS-dependency in advanced tumors5–7, pointing to a critical need to identify context-specific vulnerabilities to improve patient outcome.
Recent large-scale molecular analyses have confirmed that human PDAC can be classified into several subtypes based on transcriptome profiles8–11. Although the subtype nomenclature varies between individual reports, the molecular signature of each subtype is largely overlapping across these studies. Since it has been suggested that the expression signatures of some of the subtypes, such as ADEX/exocrine or immunogenic subtype, are likely derived from non-neoplastic cells9,11, the molecular signatures of tumor cells largely fall into two categories, namely the squamous/quasimesenchymal/basal-like and the progenitor/classical subtypes. Although the expression subtypes are not consistently associated with any somatic mutation or other genetically altered pathways11, the squamous subtype reproducibly exhibits worse prognosis compared to other subtypes8,10,11. This suggest the biological phenotype of these tumor subgroups are driven by subtype-specific molecular mechanisms other than genetic alterations and thus, identifying the oncogenic pathways that drive the squamous subtype tumors will reveal subtype-specific vulnerabilities to treat these highly malignant tumors.
Yes-associated protein 1 (YAP1) is a transcriptional coactivator and plays critical roles in controlling tissue growth during organ development12. Its activity is kept in check by the upstream Hippo pathway, composed of the MST1/2-LATS1/2 kinases cascade, which phosphorylates YAP1 at multiple serine residues and sequesters YAP1 in cytoplasm for degradation13. Recent studies showed that YAP1 activation is commonly observed in tumor cells and in vivo studies using genetically engineered mouse (GEM) models have shown that hyper-activation of YAP1 results in acinar-to-ductal metaplasia, a trans-differentiation process tightly correlated with PDAC initiation14,15. Conversely, pancreas-specific Yap1 deletion in GEM models abolished PDAC development driven by oncogenic Kras, suggesting YAP1 is essential for tumor initiation16,17. Recent studies also indicate that YAP1 activation is associated with more aggressive behavior of PDAC such as epithelial-to-mesenchymal transition (EMT) and liver metastasis 18,19. In addition, we recently showed that Yap1 amplification is sufficient to bypass KRAS dependency in an inducible PDAC GEM model5, implicating YAP1 may substitute oncogenic KRAS to sustain tumor growth. However, the requirement for YAP1 in tumor maintenance in advanced human PDAC is yet to be thoroughly verified. Though, YAP1 is activated in human PDAC16,20, the mechanisms that lead to YAP1 activation in PDAC cells remain largely elusive. Here we show that YAP1 is not only required for PDAC initiation in adult animals, but also remains activated in a subset of advanced human PDACs with squamous subtype feature and is required for their tumorigenic activity. We further identify WNT5A overexpression as one of the mechanisms leading to YAP1 activation in the deadliest form of PDAC.
Material and Methods
Transgenic Mice
For generation of tamoxifen-inducible PDAC GEM model, Mist1CreERT2/+21 mice were used for conditional activation of mutant KrasG12D and mutant Trp53R172H in the mature pancreas. For Yap1 deletion, these mice were further crossed with Yap1fl/fl mice22. For most efficient recombination, tamoxifen was administered (i.p.) to 6-week old mice in corn oil once daily for 5 days. The recombination efficiency was tested using PCR primers designed specifically to detect wildtype and recombinant alleles of Kras, Trp53 and Yap1 in pancreatic tissue. All manipulations were approved under MD Anderson Cancer Center (MDA) Institutional Animal Care and Use Committee (IACUC) under protocol number 00001549.
Xenograft Studies
All xenograft studies were carried out in NCr nude mice (Taconic) and were approved by the MD Anderson IACUC under protocol number 00001549. Details for the xenograft studies are described in Supplementary Methods.
Cell Culture and Establishment of Primary PDAC lines
Human pancreatic cell lines SNU410, HPAC, HPAFII, PL45, PaTu8988S and PaTu8988T were obtained from ATCC. Pa04C was established from resected patient tumors and maintained as low passage (<10)23 and cultured according to recommended protocols. Establishment and maintenance of primary mouse PDAC lines was performed as described previously 4,5. Mouse PDAC cell line, PD3077, was gift from Dr. Ben Stanger, University of Pennsylvania Perelman School of Medicine. The human patient PDX cell lines were maintained in RPMI-1640 medium containing 10% FBS (Clontech). KRAS mutation status, molecular subtypes and tumor grade information are listed in Supplementary Table 1.
Reagents
Doxycycline (RPI), PE Annexin V Apoptosis Detection Kit I (BD Biosciences), BOX5 (EMD Millipore).
Immunostaining and Western Blot Analysis
Immunohistochemical (IHC) analysis was performed as described earlier24. Details for immunofluorescence staining, western blot and primary antibody information are described in Supplementary Methods.
Lentivirus Mediated shRNA Knockdown
All lentiviral shRNA clones targeting YAP1, WNT5A and non-targeting shRNA control were obtained from Sigma Aldrich in the pLKO vector. The clone IDs for the shRNA are listed in Supplementary Materials.
Crispr-Cas9 Mediated Gene Knockout (KO)
sgRNAs targeting mouse Wnt5a or Yap1 were cloned into pSpCas9(BB)-2A-Puro (Addgene, æ62988) and transfected into target cells. After 2 μg/ml puromycin selection for 1 week, single cell clones are isolated and analyzed by T7E1 assay and Western blot. Sequences for Wnt5a and Yap1 sgRNA are listed in Supplementary Materials.
TMA staining and analysis
Immunohistochemical staining for YAP1 was performed on 5-μm unstained sections from the tissue microarray blocks, which included 92 (MD Anderson Cancer Center (MDA)) or 83 (Johns Hopkins University School of Medicine (JHU)) PDAC samples from patients who underwent upfront surgery. The immunohistochemical staining for YAP1 was reviewed by a pathologist (H.W.). The expression of YAP1 was classifies as YAP1-low and YAP1-high using the median score for total YAP1 expression (nuclear plus cytoplasmic expression) as a cutoff.
Statistical Analysis
Tumor volume and tumor free survivals were analyzed using GraphPad Prism. To assess distributional differences of variance across different test groups, the Mann-Whitney test was used. Other comparisons were performed using the unpaired Student t-test. For all experiments with error bars, standard deviation (SD) was calculated to indicate the variation with each experiments and data, and values represent mean ± SD.
Results
YAP1 is required for PDAC development in adult mice
To evaluate the role of YAP1 in human PDAC, tissue microarray (TMA) analysis in a cohort of 92 human PDAC showed that 47% PDACs exhibit high YAP1 protein expression in tumor epithelium compared to surrounding stromal tissue. The median overall survival (OS) for YAP1-low group was 38.3 months compared to 25.3 months for YAP1-high group (p=0.02) (Fig.1A&B). Such association between elevated YAP1 protein and poor survival was further validated in an independent cohort of 83 PDAC patients (p=0.0475) (Fig.1C), implicating YAP1 may promote adverse biological outcomes in PDAC. The in vivo function of YAP1 during PDAC development was further characterized using genetically engineered mouse (GEM) models. Previous studies have shown that conditional Yap1 deletion at early embryonic stage blocked mutant KRAS-driven initiation of acinar to ductal metaplasia (ADM), premalignant lesions and subsequent tumor formation in GEM models16,17. To more faithfully recapitulate the PDAC initiation of human patients, we investigated the requirement of YAP1 for PDAC when the tumor initiation is started in adult pancreas. Tamoxifen-induced acinar-specific activation of Cre recombinase in adult pancreas of the Mist1-CreERT2; LSL-KrasG12D/+; LSL-Trp53R172H/+ (MKP) model leads to rapid PDAC development accompanied by induction of nuclear YAP1 expression in tumor cells (Fig. 1D, left, YAP-WT). In contrast, acinar-specific deletion of Yap1 in the MKPY (YAP-KO) model completely blocked tumor development, resulting in relatively normal pancreas (Fig.1D-F). While all MKP mice succumbed to PDAC with a median survival of 103 days (n=29), Yap1-null MKPY mice remained entirely free of any overt disease and pathological lesion (n=29) (Fig.1G). This could be attributed to the effect of YAP1 on cell proliferation and survival, as evident by loss of both Ki67 and survivin (BIRC5) staining in YAP-KO pancreas (Fig.1H & S1A). The protective effect of Yap1 deletion on KrasG12D and Trp53R172H/+ ‐induced PDAC development was further confirmed with an independent tamoxifen-inducible Elastase-CreERT2; LSL-KrasG12D/+; LSL-Trp53R172H/+ model (Fig.S1B).
YAP1 is activated in squamous subtype of human PDAC
Gain-of-function studies showed that ectopic expression of mutant YAP1S127A, a constitutively active YAP1 resistant to cytoplasmic retention and degradation 13, drastically enhanced the anchorage-independent growth, migration and invasion capacity of human PDAC cell line PaTu8988S cells and Pa04C cells, an early passage patient-derived cell line (Fig.2A-C, S2A), suggesting active YAP1 may promote the tumorigenicity and metastatic spread of PDAC cells. Indeed, YAP1S127A expression diminished necrotic regions within primary tumor core and enhanced the distal metastasis of Pa04C cells in an orthotopic xenograft model (Fig.2D&E, S2B-C). This was accompanied by induction of canonical YAP1 target genes, such as CYR61, CTGF and AXL in both cultured cells and xenograft tumors (Fig.S2D&E), suggesting YAP1 activation contributing to this aggressive phenotype. Gene expression microarray and subsequent Gene Set Enrichment Analysis (GSEA) in these cells confirmed an established YAP1 signature (Fig.S2F), enrichment in pathways associated with tumor development and metastasis, and the underlying cellular processes responsible, such as cell proliferation, cell cycle progression, migration, motility and EMT (Supplementary Tables 2–3). By QPCR, we also confirmed upregulation of several genes reported to be involved in EMT in Pa04C-YAPS127A tumors (Fig.S2E). Interestingly, cellular processes like cell cycle progression and signaling pathways significantly activated in Pa04C-YAP1S127A such as MYC, IL6-STAT3, TGF, RhoA, E2F, TNF etc., along with Hippo signaling pathway were part of all four gene signatures (GP2-5) associated with squamous subtype of PDAC8 (Supplementary Table 4–5), implicating the role of YAP1 activation in this most aggressive subtype.
To address this, we analyzed the expression profiles of human PDAC from the recent ICGC collection8 and found that tumors of squamous subtype exhibit elevated expression of genes that are known to be associated with YAP1 activation25 (Fig.2F&G, Supplementary Table 6). Moreover, the expression of YAP1 activation signature is significantly correlated with that of the squamous subtype signature in both TCGA and ICGC datasets (Fig.2H), underscoring the tight relationship between YAP1 activation and squamous subtype tumors. YAP1 pathway activation is significantly correlated with poor survival in PDAC patients (Fig.2I). Squamous subtype has also been shown to be associated with poor prognosis8,10. Thus, to exclude the possibility that correlation between YAP1 activation and shortened patient survival is merely a reflection of enrichment of YAP1 signature in squamous subtype, we further analyzed the prognostic value of YAP1 signature in PDAC subtypes other than the squamous subtype. Indeed, the median patient survival significantly declines with the increase of YAP1 activation signature in those non-squamous subtype tumors (Fig.2J), indicating that YAP1 pathway activation is an independent prognosis factor in PDAC patients.
To further exclude the possibility that YAP1 activation signature in squamous subtype PDAC is largely derived from tumor stroma, we analyzed the transcriptome data of human PDAC cell lines from CCLE dataset and that of a collection of 47 PDAC PDX models, after the expression reads from murine host were omitted. We failed to identify cell lines or PDX models that are enriched with ADEX or immunogenic signatures (Fig.S2G&H, and data not shown), consistent with the notion that the molecular signatures of these subtypes are likely derived from non-tumor cells9. Not surprisingly, the molecular signatures of human PDAC cell lines or PDXs mostly clustered under either progenitor or squamous subtype (Fig. S2G&H). In accordance with the analysis of ICGC data, YAP1 activation signature was consistently elevated in squamous subtype (Fig.2K&L). Interestingly, multiplexed gene expression analysis using the NanoString platform showed that, in contrast to the parental PaTu8988S cells that belonged to progenitor subtype, the molecular signature of PaTu8988S cells expressing YAP1S127A was highly similar to that of PaTu8988T cells, which were derived from the same patient as PaTu8988S cells26 but of squamous subtype (Fig.S2I, Supplementary Table 7). The induction of squamous subtype by YAP1 activation is further verified in vivo where the Pa04C-YAPS127A tumors exhibited induction of squamous subtype gene expression and suppression of progenitor subtype gene expression (Fig.S2J). Together, our data indicates that YAP1 activation is capable of driving the molecular signature of squamous subtype.
YAP1 is essential for the maintenance of squamous subtype PDACs
To further investigate the requirement of YAP1 in squamous subtype PDAC, we conducted loss-of-function studies with shRNA in a panel of human PDAC cell lines and early-passage primary cell lines derived from human PDX tumors. Knockdown of YAP1 strongly suppressed the colony formation capacity of PDAC cell lines (PaTu8988T, SNU410 and PL45) and PDX lines (PATC148 and PATC153) belonging to squamous subtype with strong YAP1 activation signature (Fig.3A-D, S2G-H). In contrast, progenitor subtype cell lines with low YAP1 signature, including PaTu8988S, HPAF-II, HPAC PATC102 and PATC108 cells, were less sensitive to YAP1 depletion (Fig.3A-D, S2G-H). Moreover, inducible knockdown of YAP1 in established tumors induced regression in PaTu8988T tumors while the growth of PaTu8988S tumors were not affected (Fig.S3A-B), indicating the role of YAP1 for tumor maintenance in squamous subtype PDAC.
Others and we have previously showed that YAP1 activation enables the bypass of oncogene addiction in multiple cancer types, including PDAC5,27–29. Indeed, compared to PaTu8988S cells, PaTu8988T cells were more resistant to KRAS knockdown with shRNA (Fig.3E-F&S3C). Expression of YAP1S127A partially rescued the growth of PaTu8988S cells upon KRAS depletion (Fig.3E-F&S3C), indicating that YAP1 activation could enable bypass of KRAS-dependence in PDAC cells. Accordingly, pathway analysis of human PDAC expression profiles in the ICGC dataset indicated that gene signatures induced upon KRAS knockdown 30 or suppressed by oncogenic KRAS expression 31 were significantly upregulated in squamous subtype tumors (Fig.S3D), implying relatively low KRAS activity in these tumors. By using an inducible KrasG12D-driven PDAC GEM model, we have recently obtained a collection of spontaneous relapse tumors following KrasG12D extinction in advanced PDAC (iKras‐ tumors)5. These tumors neither expressed oncogenic Kras nor exhibited strong activation of KRAS surrogates, and are thus deemed as KRAS-independent. Interestingly, molecular signature of the iKras‐ tumors (Supplementary Table 8), which is composed of genes that highly expressed in iKras‐ tumors compared to iKras+ tumors 5, was also significantly enriched in squamous subtype of human PDACs (Fig.3G&S3E), further supporting the notion that squamous subtype tumors are relatively KRAS-independent. Consistent with the YAP1 activation in human squamous subtype PDAC, YAP1 signature is significantly enriched in the mouse iKras‐ tumors (Fig.3H). While YAP1 knockdown exhibit minimal effect on progenitor subtype Kras-driven mouse tumor lines (AK192 and 19636) 4,5, YAP1 depletion significantly suppressed the proliferation and colony formation capability of squamous subtype cells (PD3077) 32 and iKras‐ tumor cells (Fig.3I-J, S3F), underscoring the YAP1 dependency in squamous subtype tumors. Interestingly, reinduction of KRASG12D expression with doxycycline in iKras‐ cells fails to sustain colony formation following YAP1 knockdown (Fig.S3G-I), suggesting YAP1 activation may render oncogenic KRAS obsolete in PDAC cells.
WNT5A overexpression contributes to YAP1 activation in PDAC
In an effort to identify mechanism for YAP1 activation, genomic analysis of primary human tumors from ICGC or TCGA dataset didn’t reveal frequent copy number changes or mutations of YAP1 locus (Fig.S4A). While Yap1 is amplified in a subset of mouse iKras‐ tumors5, most iKras‐ tumors harbor no genomic alteration of Yap1 despite enrichment of YAP1 signature (Fig. 3H&S4B), indicating non-genetic mechanisms cause YAP1 activation in PDAC. In addition, no obvious difference in YAP1 transcription level was observed between squamous and progenitor subtype PDAC cell lines (Fig. S4C). YAP1 activation occurs by its nuclear translocation 13, which indeed, was evident in YAP1-dependent human PDAC cells, including PaTu8988T, SNU410 and PL45 cells, and iKras‐ mouse tumors or the derived cell lines without Yap1 amplification (iKras‐ Yap1Amp-) (Fig.4A&B, S4D). Conversely, YAP1 was localized in both cytoplasm and nuclei in PDAC cells with weak YAP1 signature, such as PaTu8988S, HPAF-II and HPAC cells, as well as progenitor subtype KrasG12D-driven mouse tumors and iKras‐ tumors with Yap1 amplification (iKras‐ Yap1Amp+) (Fig.4A&B). Accordingly, YAP1 activated human and mouse PDAC cells exhibit reduced phosphorylation at S127, the target site for directly upstream LATS1/2 kinases, along with corresponding trend of decrease in LATS1/2 phosphorylation, an indicator of LATS1/2 activity (Fig.S4E-H). Surprisingly, no obvious difference was observed for MST1/2 phosphorylation or their total protein level (Fig.S4E-H), suggesting YAP1 activation in PDAC cells may through direct modulation of LATS1/2 activity, independent of MST1/2.
To identify the mechanisms leading to YAP1 activation, we compared the differentially regulated pathways in iKras‐ Yap1Amp- vs iKras‐ Yap1Amp+ cells with GSEA. Interestingly, one of the pathways significantly elevated in iKras‐ Yap1Amp- cells was non-canonical WNT pathway33 (Fig.4C, Supplementary Table 9), which has been recently shown to suppresses Hippo signaling and activate YAP1 in adipocytes34. Importantly, a moderate but significant correlation between non-canonical WNT signature and YAP1 activation signature was observed in both TCGA and ICGC dataset (Fig.4D&S5A), suggesting control of YAP1 activation by non-canonical WNT pathway in PDAC. A survey of WNT ligands identified Wnt5a, the prototypic non-canonical WNT ligand35, to be exclusively overexpressed in the iKras‐ Yap1Amp- tumor cells at both mRNA and protein levels (Fig.4E&S5B). Moreover, WNT5A expression was also significantly elevated in squamous subtype compared to progenitor subtype tumors in the TCGA, while the difference in ICGC dataset does not reach statistical significance (Fig.4F&S5C). Deletion of Wnt5a with CRISPR in two independent iKras‐ Yap1Amp- tumor cell lines lead to the induction of YAP1 phosphorylation at S127, increase in cytoplasmic retention of YAP1 protein as well as downregulation of YAP1 downstream target genes (Fig.4G-I). Additionally, ectopic expression of Wnt5a in KrasG12D-driven tumor cells leads to decrease in YAP1 phosphorylation (Fig.4J), further supporting the notion that WNT5A expression drives YAP1 activation in mouse PDAC cells. We further validated these findings in human PDAC cell lines where WNT5A expression was found to be highly elevated in the YAP1-dependent PaTu8988T cells, in contrast to PaTu8988S cells, derived from same patient but with low YAP1 activation (Fig.4E). Ectopic expression of WNT5A in PaTu8988S cells reduced YAP1 phosphorylation and enhanced YAP1 nuclear localization (Fig.4J&K). On the other hand, depletion of WNT5A in PaTu8988T cells with shRNA resulted in elevated YAP1 phosphorylation (Fig.4L). Taken together, our data indicates that WNT5A overexpression can lead to YAP1 activation in PDAC cells.
WNT5A overexpression enables tumor maintenance and bypass of KRAS-dependence
At functional level, Wnt5a deletion in iKras‐ Yap1Amp- tumor cells with CRISPR significantly inhibited colony formation (Fig.5A&B). In agreement with the genetic ablation, treatment with WNT5A antagonist, BOX-5, specifically induced YAP1 phosphorylation and abolished the colony formation ability of iKras‐ Yap1Amp- tumor cells, but not the Yap1Amp+ cells (Fig.5C-E). Moreover, Wnt5a deletion also significantly inhibited xenograft tumor growth in vivo, which was rescued by reconstituted WNT5A expression (Fig.5F&G). Importantly, in contrast to the predominant nuclear staining of YAP1 in the parental iKras‐ Yap1Amp- cells, Wnt5a knockout tumors exhibited significant amount of cytoplasmic YAP1 whereas WNT5A reconstitution restored YAP1 nuclear accumulation without affecting total YAP1 expression level (Fig.5H-I, S5D), implicating the diminished tumor growth to decreased YAP1 activity. Consistent with the role of YAP1 in driving squamous subtype, expression of squamous subtype genes (Cav1 and Pappa) was suppressed in Wnt5a KO tumors whereas the expression of progenitor subtype genes (Tff1 and Muc13) was induced, which was partially reversed upon WNT5A reconstitution (Fig.S5E). In addition, expression of constitutive active YAP1S127A largely rescued the inhibitory effect of Wnt5a deletion on tumor growth (Fig.5G), thus indicating WNT5A overexpression in mouse PDAC cells promotes tumor growth by activating YAP1. In agreement with this notion, depletion of WNT5A in human PDAC cell line, PaTu8988T cells, significantly inhibited their colony formation ability (Fig.5J).
Since YAP1 activation can maintain tumor growth upon genetic extinction of KRAS oncogene in PDAC5,29, we next investigated if WNT5A overexpression can also serve to bypass KRAS-dependency. Indeed, ectopic expression of WNT5A in KrasG12D-driven iKras tumor cells and KRAS-dependent PaTu8988S cells partially restored the colony formation upon KRAS depletion (Fig.6A, S6A-B). In addition, forced WNT5A expression in KrasG12D-driven iKras tumor spheres was able to maintain cell viability upon extinction of KrasG12D by doxycycline withdrawal whereas most control tumor cells expressing GFP underwent apoptosis (Fig.6B-D). Importantly, the survival effect of WNT5A upon KrasG12D extinction was largely abolished upon YAP1 knockdown (Fig.6B-D), indicating that YAP1 is required for WNT5A-induced bypass of KRAS-dependence. Indeed, similar to the effect of YAP1S127A, ectopic WNT5A expression in iKras tumor cells showed KRAS-independent tumor growth when injected orthotopically into nude mice, whereas GFP-expressing iKras tumor cells failed to maintain tumor growth in the absence of doxycycline (Fig. 6E-G). Importantly, WNT5A-induced KRAS-independent tumor growth is abolished in Yap1 deleted cells (Fig. S6C-D), underlying the requirement of YAP1 for WNT5A-mediated bypass of KRAS-dependence. Notably, all WNT5A-driven tumors showed lower MAPK activity and strong nuclear YAP1 accumulation as compared to KrasG12D-driven tumors (Fig.6H). Together, these results indicate that WNT5A overexpression can activate YAP1 and substitute for oncogenic Kras-driven tumor maintenance.
WNT5A-YAP1 axis functions in primary human PDAC
Since WNT5A is not frequently overexpressed in established human PDAC cell lines (Fig.S2G), we further validated the WNT5A-YAP1 axis in our collection of PDAC PDXs. As shown in Fig.7A-C, WNT5A expression is elevated in squamous subtype tumors and is significantly correlated with YAP1 signature in both PDXs and TCGA dataset, although the correlation is relatively moderate. Accordingly, WNT5A was highly expressed while YAP1 phosphorylation was relatively low in squamous subtype PATC148 and PATC153 cells, while two cell lines derived from progenitor subtype tumors, PATC102 and PATC108, exhibited elevated YAP1 phosphorylation along with absence of WNT5A expression (Fig.7D). This is in accordance with the elevated expression of YAP1 target gene CYR61 and the dependence on YAP1 oncogene in squamous subtype PDXs (Fig.3C-D, 7D).
Consistent with our findings in known human PDAC cell lines and primary mouse tumor lines, shRNA-mediated depletion of WNT5A in PDX-derived PATC148 (KrasG12D) and PATC153 (Kras WT) cells caused increase in YAP1 phosphorylation, suppression of the colony formation ability and diminished tumor growth in vivo (Fig.7E-H), supporting the role of WNT5A for the tumorigenic activity. In contrast, WNT5A shRNA has minimal effect on the in vivo tumor growth of PATC108 cells. Consistent with the role of WNT5A in bypass of KRAS-dependence, knockdown of KRAS elicited less inhibition on the growth of high WNT5A expressing PATC148 (KRASG12D) cells compared to low WNT5A expressing PATC102 (KRASG12D) and PATC108 (KRASG12D) cells, with KRAS WT PATC153 cells being resistant to KRAS knockdown (Fig.7I&J). Together, our data indicates that WNT5A overexpression in squamous subtype PDACs contributes to YAP1 activation and tumor growth.
Discussion
YAP1 is activated in many cancer types and has been shown to be essential for tumor initiation and progression, including pancreatic cancer36. In this study, we found that PDAC development driven by KrasG12D and Trp53R172H in adult mice was marked by YAP1 activation, evident by its strong nuclear expression in tumor epithelium (Fig.1D). Subsequently, concurrent deletion of Yap1 in adult pancreas completely blocked PDAC development (Fig.1E-G). Immunohistochemical staining on these tumor tissues showed marked decrease in cell proliferation index, as measured by Ki-67 staining (Fig.S1A) and high Survivin (BIRC5) expression (Fig.1H). Since, Survivin expression overlapped with YAP1 expression in Yap1-WT PDAC but completely lost in Yap1-KO pancreas, it is likely regulated directly by YAP1 at transcription level, as shown in esophageal squamous cell carcinoma37. Survivin is a known anti-apoptotic protein, which is expressed only in tumor cells38 and mostly during G2-mitotic phases of cell cycle39. The mostly nuclear expression of Survivin that is observed in YAP-WT tumor sections (Fig.1H) supports its predominant role in regulating cell cycle. Accordingly, higher Survivin expression was also in agreement with predominant gene signatures associated with cell cycle progression in Pa04C-YAP1S127A cells (Supplementary Table 3) and higher percentage of Pa04C-YAP1S127A cells in G2/M phase by cell cycle analysis (data not shown).
We provided evidence that YAP1 is highly activated in squamous subtype PDACs and required for their tumorigenic function. Moreover, we show that enforced YAP1 activation was able to convert the gene expression signature of PDAC cells from progenitor to squamous subtype, further indicating that YAP1 may function as a major driver for squamous subtype tumors and thus could be a potential therapeutic target in this highly malignant subgroup. The squamous subtype PDAC was characterized by high expression of mesenchyme-associated genes with PDAC cell lines enriched within squamous subtype exhibiting features of EMT8,10,40. Moreover, the KRAS-independent mouse PDAC tumors, which were YAP1-dependent and reminiscent of squamous subtype tumors, were also highly enriched with EMT signature (data not shown), thus implicating YAP1 in promoting aggressive behavior of squamous subtype PDAC through EMT. In fact, YAP1 has been shown to be a potent inducer of EMT in tumor cells29. Recently, ZEB1, a key EMT transcription factor, was shown to be critical for PDAC metastasis41, and required for the transcription of YAP1 downstream targets by direct bind with YAP142. Further studies are warranted to verify whether EMT is indeed required for YAP1-driven aggressive behavior, such as metastasis in PDAC.
Despite the emerging role of YAP1 as a major oncogene in multiple cancer types, genetic alterations of YAP1 gene or its upstream Hippo pathway are relatively uncommon43,44. YAP1 amplification or mutations in NF2, an upstream negative regulator of YAP1 activity, has been reported in around 1% of human PDAC8,45. Therefore, the activation of YAP1 in advanced human PDAC is likely due to non-genetic factors regulating inhibitory upstream Hippo kinases. Our data showing constitutive nuclear localization of YAP1 protein in YAP1-dependent tumor cells, indicates that the suppression of Hippo signaling is the major mechanism for YAP1 activation in PDAC. While investigating upstream regulator of YAP1 activation, we found evidence that WNT5A overexpression leads to YAP1 activation and bypass of KRAS-dependency in KRAS-independent mouse PDAC cells and a subset of human squamous subtype PDACs. WNT5A is a prototypic non-canonical WNT ligand35 and has been implicated in the pathogenesis of PDAC46,47. It has been recently shown that WNT5A-mediated non-canonical WNT pathway suppresses Hippo signaling and activates YAP1 through G protein-dependent activation of Rho GTPases34. WNT5A can also engage multiple additional downstream signaling pathways, including SRC and PKC, which have been shown to activate YAP1 through direct phosphorylation or indirectly through the regulation of Rho GTPases and LATS activity48–53. Interestingly, it was recently reported that non-canonical WNT and FZD8 mediated calcium signaling counteracts the tumorigenic activity of oncogenic KRAS54. On the other hand, FZD1 was shown to be important for WNT5A-mediated YAP1 activation34. It’s possible that the engagement of specific receptors by WNT5A determines its signaling and biological output in PDAC. It remains to be further validated if any particular one or all of these mechanisms are responsible for WNT5A-mediated YAP1 activation in PDAC. In addition, it’s likely additional non-canonical WNT ligands are also involved in YAP1 activation in PDAC. Among them, WNT7B expression is also elevated in squamous subtype PDAC and is correlated with YAP1 activation signature (Data not shown). It remains to be determined whether the additional non-canonical WNT ligands also contribute to YAP1 activation in PDAC.
While our data indicates WNT5A overexpression in tumor cells functions in cell-autonomous manner to activate YAP1 oncoprotein, tumor cells may also get active WNT5A signaling through paracrine mechanisms. Interestingly, WNT5A has been shown to be highly expressed in PDAC stroma fibroblast55,56, and our preliminary data implicates that stroma WNT5A level is significantly correlated with tumor cell YAP1 level in human PDAC (data not shown). Therefore, stromal WNT5A could possibly contributes to YAP1 activation in tumor cells, given that the exuberant desmoplastic stroma is a defining characteristic of PDAC2. In this scenario, the tumor-stroma interaction will thus play an instrumental role in orchestrating the heterogeneous YAP1 activation in bulk tumor which may, in turn, define the molecular heterogeneity and diverse biological phenotypes of PDAC.
Taken together, with agents targeting the Hippo-YAP pathway under development43, our study showing the critical role of WNT5A-mediated YAP1 activation in a subset of pancreatic tumors of squamous subtype provides viable therapeutic targets for this most malignant form of human PDAC.
Author Contributions
B.T., W.Y., S.F-B., S.C., Q.W., L.Y., X.Z. and S.G. performed data collection and interpretation. J.Y., provided statistical and bioinformatics analysis. C.Z. and S.B. contributed to animal breeding. Q.C., C.B., Y.K., H.Z., H.W., J.F. and M.K. provided clinical specimen and data analyses on PDX samples. J.Z., H.W. and A.M. performed pathology analyses. G.D., D.P. and A.M. provided crucial feedback on manuscript. W.Y., S.G. and H.Y. drafted the manuscript. W.Y., S.G. and H.Y. did conception, design and supervision of the study.
Competing Interests
The authors declare no competing interests.
Supplementary Materials and Method
Immunostaining and Western Blot Analysis
For immunofluorescence staining, mouse and human cells were fixed with 4% paraformaldehyde-PBS for 15 min. Following Triton-X100 permeabilization and blocking, cells were incubated with primary antibodies overnight at 4°C following with Alexa 594-conjugated secondary antibodies at 4°C for 1 hour (Thermo Fisher Scientific, 1:1000). Samples were mounted using VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories) and immunofluorescence was detected using Olympus confocal microscopy. For western blot analysis, cells were lysed on ice using RIPA buffer supplemented with protease and phosphatase inhibitors (Sigma).
Primary Antibodies for Immunostaining and Western Blot Analysis: Yap (æ14074, Cell Signaling), pYAP (æ4911, Cell Signaling), Lats1 (æ3477, Cell Signaling), pLats1(æ8654, Cell Signaling), Mst1 (æ3682, Cell Signaling), Mst2 (æ3952,Cell Signaling), p-MST1/2 (æ3681, Cell Signaling), Merlin(æ12888, Cell Signaling), Phospho-Merlin (æ13281,Cell Signaling), Wnt5a (æ2530, Cell Signaling), Ki-67 (VP-K451, Vector Laboratories), Cyr61 (sc-13100, Santa Cruz Biotechnology), CTGF (sc-14939, Santa Cruz Biotechnology), AXL (8661, Cell Signaling), pErk (4376, Cell Signaling), pMEK (4376, Cell Signaling), Ck-19 (16858-1-AP, Proteintech), Actin (A2228, Sigma Aldrich), Vinculin (V4139, Sigma Aldrich), Kras (sc-30, Santa Cruz Biotechnology).
Ectopic expression of YAP1 and WNT5A in mouse and human cells
To generate YAP1S127A-expressing stable Pa04C cells, Pa04C cells were transfected with a linearized pcDNA3.1 plasmid with or without YAP1 cDNA containing S127A substitution. Two days post-transfection using Lipofectamine1000, cultures were selected in G418 (Sigma) and single clones were picked and expanded for further analysis. Overexpression of YAPS127A or WNT5A in human or mouse cells other than Pa04C were acheieved with lentivral infection. Briefly, lentivirus infection was performed by transfecting 293T cells with either GFP control, YAP1S127A, or WNT5A cloned in pHAGE lentivirus vector {EF1α promoter-GW-IRES-eGFP (GW: Gateway modified)}. The virus was concentrated using ultracentrifuge and added to target cells in a 6-well plate containing 10ug/ml of polybrene (Millipore). 48 hours after infection GFP positive cells were selected by flow sorting.
Lentivirus Mediated shRNA Knockdown
The clone IDs for shRNA are as follows: sh_mouse Yap1-1 (TRCN0000238436), sh mouse Yap1-2 (TRCN0000095864), sh_huYap1-1 (TRCN0000107265), sh_huYap1-2 (TRCN0000107266), sh_huWnt5a-1 (TRCN0000062717), sh_huWnt5a-2 (TRCN0000288987), sh_hu Kras-1 (TRCN0000033260), sh_hu Kras-2 (TRCN0000033262).
Crispr-Cas9 Mediated Gene Knockout
Sequences for Wnt5a sgRNA are as follows:
sgRNA-1 F: CTTGAGAAAGTCCTGCCAGT; R: ACTGGCAGGACTTTCTCAAG.
sgRNA-2 F: GAAACTCTGCCACTTGTATC; R: GATACAAGTGGCAGAGTTTC.
sgRNA-3 F: TATACTTCTGACATCTGAAC; R: GTTCAGATGTCAGAAGTATA.
sgRNA-4 F: ACAGCCTCTCTGCAGCCAAC, R: GTTGGCTGCAGAGAGGCTGT.
Sequences for Yap1 sgRNA are as follows:
sgRNA-1 F: ACCAGGTCGTGCACGTCCGC; R: GCGGACGTGCACGACCTGGT.
sgRNA-2 F: CCCCGCGGACGTGCACGACC; R: GGTCGTGCACGTCCGCGGGG.
Xenograft Studies
For orthotopic xenografts, 5 × 105 cells were injected pancreatically into NCr nude mice (Taconic) and tumor growth was monitored with bioluminescent imaging as described1. For Sub-Q xenografts, 1 × 106 cells (mouse tumor cells) or 3 × 106 cells (human PDAC or PDX cells) were injected subcutaneously into the lower flank of NCr nude mice. Tumor volumes were measured every 7 days starting from Day 7 post injection and calculated using the formula, volume = length × width2/2.
Immunoprecipitation Assay
Immunoprecipitation of YAP1 complexes was performed by lysing cells with 1% NP40 lysis buffer containing phosphatase and protease inhibitor cocktails on ice for 45 minutes. 1 mg of lysate was incubated overnight at 4°C with primary antibodies followed with protein A/G Plus-agarose (sc-2003, Santa Cruz) for 3 hours at 4°C. Immunoprecipitates were washed three times with lysis buffer, then re-suspended with 2X sample buffer boiled for 5 min and detected by western blot analyses.
Quantitative Real-time Polymerase Chain Reaction Analysis
RNA from cell lines and pancreas tissues was isolated using RNeasy Mini Kit (Qiagen) and first-strand cDNA was synthesized from 2 μg total RNA using random primers and Omniscript® Reverse Transkriptase Kit (Qiagen). Actin was used as housekeeping gene. Real-time polymerase chain reaction experiments were performed in triplicates and are displayed ± SD.
Clonogenic assay
500-2000 cells were seeded into each well of 6-well plate in triplicates and incubated to allow colony formation for 10-20 days. The colonies were stained with 0.2% crystal violet in 80% methanol for 30 minutes at room temperature and de-stained upon which they were scanned and colonies counted using Image J (http://rsb.info.nih.gov/ij).
Gene Expression and PDAC-Subtype Analysis
mRNA expression profiling on Illumina microarrays were performed according to the manufacture’s protocol. Raw data was processed using Genome Studio (GSGX Version 1.6.0) and analysis was done using group quantile normalization with background subtraction. The software package LIMMA (Linear Models for Microarray Data) was applied to detect significantly differentially expressed probes using Benjamini-Hochberg adjusted p-values. For GSEA analysis, gene sets collection from MSigDB 3.0 and Kyoto Encyclopedia of Genes and Genomes (KEGG) were included in the analysis.
For molecular subtype analysis, we combined subtype specific genes from Collisson et al and Bailey et al studies to construct a “NanoString signature” comprised of 32, 23, and 17 subtype specific genes for squamous, progenitor, and ADEX subtypes, respectively (see supplementary table 7). To call PDAC subtypes in human PDAC cell lines or PDXs, we used the following algorithm. First, subtype signature scores were calculated by summing up Z scores from subtype specific genes which were ceiled at 2.5 and bottomed at −0.5. Kmeans two separation were then done to call high and low groups within each subtype. SQ/Pro subtypes are then called as squamous gene high/progenitor gene low and vice versa. ADEX subtype is called as ADEX gene high and SQ/Pro gene low. The remaining tumors are undefined. The nanoString call results are further confirmed by heatmap visualiation using Collission genes (data not shown)2. The clusters were viewed using Java TreeView (version 1.1.6r4) 3.
Supplementary Figure Legends
Acknowledgements
We thank the laboratory of Dr. Ben Stanger for sharing mouse PDAC cell lines, PD3077. We would like to thank the Institute for Applied Cancer Science (IACS), the Flow Cytometry and Cellular Imaging Core, the Department of Veterinary Medicine at The University of Texas MD Anderson Cancer Center (Cancer Center Support Grant, CA016672). We thank Dr. Ronald DePinho, Dr. Alan Wang, Dr. Mien-Chie Hung, Dr. Guocan Wang, Dr. Baoli Hu, Dr. Xin Zhou and Dr. Jihye Paik for helpful discussions and critical reviews. The research was supported by the Pancreatic Cancer Action Network-AACR Career Development Award to H.Y.; the Pancreatic Cancer Action Network-AACR Pathway to Leadership Award to W.Y.; Seed Grant from Hirshberg foundation for pancreatic cancer research to H.Y. and W.Y., P01 Grant (P01CA117969 12, NIH) to H.W., J.B.F., M.K., G.F.D., A.M. and H.Y.