YAP1 and TAZ Control Pancreatic Cancer Initiation in Mice by Direct Up-regulation of JAK–STAT3 Signaling

Abstract

Background & Aims

Pancreatitis is the most important risk factor for pancreatic ductal adenocarcinoma (PDAC). Pancreatitis predisposes to PDAC because it induces a process of acinar cell reprogramming known as acinar-to-ductal metaplasia (ADM)—a precursor of pancreatic intraepithelial neoplasia lesions that can progress to PDAC. Mutations in KRAS are found at the earliest stages of pancreatic tumorigenesis, and it appears to be a gatekeeper to cancer progression. We investigated how mutations in KRAS cooperate with pancreatitis to promote pancreatic cancer progression in mice.

Methods

We generated mice carrying conditional alleles of Yap1 and Taz and disrupted Yap1 and Taz using a Cre-lox recombination strategy in adult mouse pancreatic acinar cells (Yap1fl/fl;Tazfl/fl;Ela1-CreERT2). We crossed these mice with LSL-KrasG12D mice, which express a constitutively active form of KRAS after Cre recombination. Pancreatic tumor initiation and progression were analyzed after chemically induced pancreatitis. We analyzed pancreatic tissues from patients with pancreatitis or PDAC by immunohistochemistry.

Results

Oncogenic activation of KRAS in normal, untransformed acinar cells in the pancreatic tissues of mice resulted in increased levels of pancreatitis-induced ADM. Expression of the constitutive active form of KRAS in this system led to activation of the transcriptional regulators YAP1 and TAZ; their function was required for pancreatitis-induced ADM in mice. The JAK–STAT3 pathway was a downstream effector of KRAS signaling via YAP1 and TAZ. YAP1 and TAZ directly mediated transcriptional activation of several genes in the JAK–STAT3 signaling pathway; this could be a mechanism by which acinar cells that express activated KRAS become susceptible to inflammation.

Conclusions

We identified a mechanism by which oncogenic KRAS facilitates ADM and thereby generates the cells that initiate neoplastic progression. This process involves activation of YAP1 and TAZ in acinar cells, which up-regulate JAK–STAT3 signaling to promote development of PDAC in mice.

See Covering the Cover synopsis on page 380; see editorial on page 393.

The most common and lethal tumor of the pancreas, pancreatic ductal adenocarcinoma (PDAC), exhibits the histologic morphology and marker expression of pancreatic duct cells. However, many of these tumors originate from the other major exocrine cell type, acinar cells, through a reprogramming process known as acinar-to-ductal metaplasia (ADM).¹ ADM is characterized by a change in marker expression: acinar cells positive for amylase also begin to express the ductal cell marker CK19. These poorly differentiated cells express the pancreatic progenitor cell markers Pdx1 and Nestin^{2, 3} and are important precursors of malignancy.⁴ Pancreatic intraepithelial neoplasia (PanIN) lesions have a well-established histologic progression toward PDAC,⁵ and can arise from ADM lesions.¹ ADM is the earliest pre-neoplastic lesion that predisposes to PDAC, making acinar-to-ductal reprogramming a crucial step in pancreatic cancer initiation.

KRas mutations are found in >90% of all PDAC cases⁶ and they occur early, in low-grade pre-neoplastic lesions.^{7, 8} Although the importance of KRas mutation in the development of ADM, PanIN, and pancreatic tumors is well established,⁹ the mechanisms by which oncogenic Ras leads to PDAC are not fully understood.

Pancreatitis is a well-known risk factor for PDAC development in humans. Patients with hereditary pancreatitis show a 50-fold increase in pancreatic cancer incidence.¹⁰ In mouse models, both acute and chronic inflammation of the pancreas accelerate pancreatic cancer progression.^{1, 11} Pancreatitis can be induced experimentally by injection of caerulein, which induces acinar cell death and inflammation.¹² In caerulein-induced pancreatitis, inflammation induces acinar cells to reprogram to form ADM lesions,¹¹ which may be transient, or in the presence of a KRas^G12D mutation, persistent.¹³ Transforming growth factor–α (TGFα) administration or overexpression also cooperates with KRas^G12D to induce ADM in vitro.^{3, 14} Pancreatic inflammation facilitates tumorigenesis by inducing ADM.

Expression of oncogenic KRas^G12D at endogenous levels in acinar cells triggers progression to pancreatic adenocarcinoma very inefficiently. Stronger transgenic expression of oncogenic Ras,¹⁵ or additional oncogenic stimuli, such as mutant p53¹⁶ or TGFα overexpression,¹⁷ are required. Many of these additional stimuli directly or indirectly increase Ras activity, suggesting that a positive feedback loop amplifying oncogenic Ras signaling is required for PDAC progression.¹⁸

The reliance of established PDAC tumors on KRas^G12D expression can be alleviated by up-regulation of the transcriptional regulator YAP1.¹⁹ YAP1 has also been implicated in progression from PanIN to PDAC,²⁰ but its contribution at the reprogramming stage is unclear.

Here we show that KRas mutation sensitizes acinar cells to reprogramming by activating YAP1 and TAZ signaling, which in turn up-regulates components of the JAK–STAT3 pathway. This increases sensitivity to inflammatory stimuli, which induces widespread ADM among KRas-mutant acinar cells. Importantly, inhibiting KRas^G12D- and pancreatitis-induced reprogramming by inactivating YAP1/TAZ also prevents progression to PanIN, highlighting the importance of this mechanism for pancreatic cancer initiation.

Materials and Methods

Mouse Lines

For the generation of conditional Yap1 and Taz/Wwtr1 knockout mice, embryonic stem (ES) cell clones carrying the respective targeted alleles were obtained from the Knockout Mouse Project (KOMP) Repository (Yap1<tm1a(KOMP)Mbp>; ES clone E08) and the European Conditional Mouse Mutagenesis (EUCOMM) Program (Wwtr1<tm1a(EUCOMM)Wtsi>; ES clone A08). Targeted ES cell clones (agouti C57BL/6 parental cell line JM8A3.N1) were injected into blastocysts of C57BL/6 background to generate chimeric mice, which were then crossed with Flpo-expressing mice (Tg(CAG-Flpo)1Afst; C57BL/6 background)²¹ to remove the neo-selection cassette and obtain mice carrying the conditional floxed alleles Yap1<tm1c(KOMP)Mbp> and Wwtr1<tm1c(EUCOMM)Wtsi> isogenic on C57BL/6. The Flpo-allele was bred out and mice were maintained on a C57BL/6 background. For genotyping and validation of knockout alleles, see Supplementary Figure 4 and Supplementary Materials and Methods.

The Stat3^fl/fl, Pdx1-Cre, R26-LSL-YFP, LSL-KRas^G12D, and Ela1-Cre^ERT2 mouse lines have been described (see Supplementary Materials and Methods).

Acute Pancreatitis Induction

Acute pancreatitis in C57BL/6 mice was induced by 8 intraperitoneal injections of caerulein (Sigma-Aldrich, St Louis, MO) dissolved in phosphate-buffered saline (PBS) given at 1-hour intervals on 2 consecutive days at a dose of 50 μg/kg body weight per injection (high dose). Other strains received a similar treatment schedule except with 6 injections at a dose of 40 μg/kg body weight per injection (low dose). Control animals received injections of PBS only.

Gene Set Enrichment Analysis

Gene set enrichment analysis was performed on the published transcription profile of C57BL/6 Jackson mice treated with caerulein or PBS²² and on the transcription profile of mouse liver organoids overexpressing YAP1,²³ with the software developed by the Broad Institute of the Massachusetts Institute of Technology and Harvard University. The settings are listed in Supplementary Materials and Methods.

Primary Acinar Cell Culture

Mice were sacrificed by cervical dislocation, the pancreas was dissected out and transferred to ice-cold Hank’s balanced salt solution (HBSS) supplemented with penicillin/streptomycin (Sigma-Aldrich). Pancreata were cut into small pieces and digested with 2 mg/mL Collagenase P (Roche Diagnostics, Indianapolis, IN) in HBSS for 15 minutes at 37°C. Cells were washed 3 times with HBSS supplemented with 5% fetal bovine serum and then filtered through 500-μm and 105-μm nylon meshes (Spectrum Laboratories, Rancho Dominguez, CA). The cell suspension was layered on top of 30% fetal bovine serum in HBSS, centrifuged at 1000 rpm for 2 minutes and the cell pellet was resuspended in acinar cell culture medium ([Waymouth’s medium; Life Technologies, Carlsbad, CA] supplemented with 1% fetal bovine serum, penicillin/streptomycin [Sigma-Aldrich], 1 μg/mL dexamethasone [Sigma-Aldrich] and 100 μg/mL soybean trypsin inhibitor [Sigma-Aldrich]). The cell suspension was then infected with adenoviruses either obtained from the University of Iowa Carver College of Medicine (Ad5-GFP and Ad5-Cre-GFP), at a concentration of 1.25 × 10⁷ plaque-forming units/mL, or generated by cloning (Ad-GFP, Ad-GFP-YAP1-5SA, and Ad-GFP-TAZ-S89A), at a concentration of 5 × 10⁵ plaque-forming units/mL. Acinar cells were incubated with adenoviruses for 1 hour at 37°C. Six-well tissue culture plates were coated with an 800 μL layer of collagen solution (4 mg/mL rat tail collagen; BD Biosciences, San Jose, CA), 10% 10× Waymouth’s medium (Sigma-Aldrich) and 0.02 mol/L NaOH; Sigma-Aldrich). The cell suspension was mixed 1:1 with collagen solution and plated onto the collagen layer. The acinar cell/collagen mix was allowed to solidify for 1 hour at 37°C before adding medium. Medium was changed on day 1 and day 3 after the isolation. Quantifications were done on day 5 after isolation. To harvest the cells, the collagen matrix was digested with 1 mg/mL collagenase P (Roche Diagnostics) diluted in HBSS for 30 minutes at 37°C.

Production of Recombinant Adenoviruses

The vectors harboring the coding sequences of the phosphorylation mutants of YAP1 (5SA) and TAZ (S89A), kindly provided by S. Piccolo, were used to subclone the YAP1 and TAZ complementary DNAs in the pAd-Track-cytomegalovirus (CMV) vector (Addgene, Cambridge, MA; #16405). This results in the expression of YAP1/TAZ under the control of the CMV promoter and pAd-Track-CMV empty vector was used as control. The pAd-Track-CMV contains a GFP complementary DNA under control of a second CMV promoter. Recombinant adenoviruses expressing GFP, YAP1-5SA in combination with GFP or TAZ-S89A in combination with GFP were generated following the protocol as described.²⁴ Briefly, human HEK293A cells were co-transfected with pAd-Easy-1 (Addgene #16400) together with pAd-Track-CMV, pAd-Track-CMV-YAP1-5SA, or pAd-Track-CMV-TAZ-S89A. Adenoviruses were harvested 14 days after transfection. To generate higher viral titer, fresh HEK293A were infected with the adenoviruses and grown for 7 days before harvesting and this process was repeated 3 times. Adenoviruses were purified using the Adeno-X Maxi Purification Kit (TaKaRa Clontech, Mountain View, CA) according to the manufacturer’s protocol. Virus titers were determined by infecting HEK293T cells with a dilution series of the adenoviruses. After infection, cells were overlaid with agarose and plaques were visualized by MTT (Sigma-Aldrich) and quantified 7 days later. This protocol generated virus titer ranging from 1 to 6 × 10⁸ plaque-forming units/mL.

Immunoblot Analysis and Ras-GTP Pulldown Assay

Protein lysates were obtained from dissected pancreata, primary acinar cells or PANC1 cells and homogenized in Cell Lysis Buffer (Cell Signaling, Danvers, MA) containing proteasome inhibitor (Sigma-Aldrich). Immunoblots were performed as described.²⁵ Polyvinylidene difluoride membranes (GE Healthcare, Little Chalfont, Buckinghamshire, United Kingdom) were incubated with antibodies against YAP1 (Cell Signaling; #4912 and Santa Cruz Biotechnology, Santa Cruz, CA; sc-15407); p-YAP1 (Ser127 in humans and Ser112 in mice; Cell Signaling; #4911); TAZ (Sigma-Aldrich; HPA007415); CK19 (DHSB; TROMA-III); STAT3 (Cell Signaling; #4904); p-STAT3 (Tyr705; Cell Signaling; #9131); LATS1 (Cell Signaling; #3477); p-LATS1 (Ser909; Cell Signaling; #9157); LIFR (Santa Cruz Biotechnology; sc-659); PDX1 (R&D Systems, Minneapolis, MN; MAB2419); SOX9 (Millipore, Billerica, MA; AB5535); p-Erk1/2 (Thr202/Tyr204; Cell Signaling; #9106); Erk1/2 (Cell Signaling, #9102); glyceraldehyde-3-phosphate dehydrogenase (Abcam, Cambridge, MA; ab9485); and ACTIN (Abcam; ab49900). Active Ras levels were determined by Ras-GTP pull-down assay (ThermoFisher Scientific, Logan, UT) using the Raf Ras-binding-domain fused to GST (Raf-RBD-GST) according to the manufacturer’s instructions. Signal intensities of bands were quantified using ImageJ software.

Statistics

Statistical analysis was performed using GraphPad Prism 6.0a software (GraphPad, La Jolla, CA) and 2-tailed Student t tests were used to generate P values. Normalized enrichment scores and nominal P values for the gene set enrichment analysis were generated using the software developed by the Broad Institute of the Massachusetts Institute of Technology and Harvard University. P < .05 was considered significant.

Study Approval

Mouse experiments were carried out with the approval of the London Research Institute’s Ethical Review Committee according to the UK Animals (Scientific Procedures) Act 1986.

Results

YAP1 and TAZ Are Up-Regulated in Pancreatitis

ADM caused by pancreatitis is an initiating step in pancreatic tumor development.¹ To find molecular regulators of this process, we performed gene set enrichment analysis²⁶ on the published transcriptional profile of mouse pancreata treated with caerulein²² to identify pathways up-regulated in pancreatitis (Supplementary Figure 1A). Many of these pathways, such as Wnt, Notch, and JAK–STAT3, have previously been implicated in either acinar cell regeneration or in inducing ADM upon injury.^{13, 27, 28, 29} Intriguingly, the highest gene set enrichment scores were for YAP1 target genes (Figure 1A and Supplementary Figure 1A).

Figure 1.

YAP1 and TAZ are up-regulated in pancreatitis. (A) Gene set enrichment analysis of transcription data from caerulein-treated pancreata²² identified enrichment of the conserved YAP1 signatures reported by Yimlamai et al²³ and Cordenonsi et al.⁴⁷ Normalized enrichment score (NES) and nominal (NOM) P values are shown. (B) Scheme showing model of caerulein-induced acute pancreatitis in C57BL/6 mice. (C) Quantitative reverse transcription polymerase chain reaction of YAP1/TAZ target genes in caerulein-treated vs PBS-treated pancreata. n = 3 mice each group; means ± SEM are shown; ^∗∗∗P < .005, Student t test. (D) Immunoblots of pancreas lysates from caerulein- and PBS-treated animals. p-Lats1, phosphorylated Lats1 (Ser909); p-Yap1, phosphorylated Yap1 (Ser112). (E) Immunohistochemical stains of YAP1 and TAZ in pancreata of caerulein- and PBS-treated mice. Scale bars = 50 μm. n = 5–7 mice analyzed. (F) Triple immunofluorescence of pancreata from caerulein- and PBS-treated animals showing the duct marker CK19, the acinar marker amylase and either YAP1 or TAZ. Arrowheads indicate ADM cells. Scale bars = 20 μm. n = 5–7 mice analyzed. DAPI, 4′,6-diamidino-2-phenylindole dihydrochloride.

To investigate a possible role for YAP1/TAZ signaling in pancreatitis-induced ADM, we treated C57BL/6 mice with caerulein to induce pancreatitis (Figure 1B). In untreated mice, YAP1 and TAZ proteins showed nuclear localization in duct cells of the pancreas, with low or absent expression in acinar cells and islet cells^{20, 30} (Supplementary Figure 1B). Consistent with the gene set enrichment analysis, the YAP1/TAZ target genes Ctgf, Cyr61, Ankrd1, and Amotl2 were overexpressed in caerulein-treated animals compared with control PBS-treated mice (Figure 1C). Caerulein treatment reduced activatory phosphorylation of the YAP1/TAZ inhibitory kinase Lats1 (Figure 1D and Supplementary Figure 2A and B). In line with this, we observed increased YAP1 and TAZ protein levels (Figure 1D and Supplementary Figure 2A and B), accompanied by strong nuclear localization, indicating more active YAP1/TAZ in caerulein-treated animals (Figure 1E). Levels of the duct cell marker CK19 were also increased, consistent with acinar-to-ductal reprogramming (Figure 1D and Supplementary Figure 2A and B). To determine the identity of the YAP1/TAZ-responsive cells, we performed immunofluorescence co-staining with pancreatic cell type specific markers. Cells coexpressing CK19 and the acinar cell marker amylase showed high levels of nuclear YAP1 and TAZ, with some stromal cells also positive (Figure 1F). Thus, YAP1/TAZ activity is elevated in the injured, inflamed pancreas, particularly in the subset of cells undergoing ADM.

YAP and TAZ Are Up-Regulated in Acinar-to-Ductal Metaplasia Lesions Induced by KRas^G12D

We next asked whether YAP1/TAZ up-regulation is restricted to caerulein-induced ADM or also occurs during ADM induced by oncogenic Ras. The pancreatic tumor mouse model LSL-KRas^G12D;Pdx1-Cre develops PDAC with long latency and displays the whole spectrum of pre-neoplastic lesions³¹ (Figure 2A). We found a strong signal for both YAP1 and TAZ in ADM lesions of these mice, and an up-regulation of YAP1 and TAZ in pancreatic protein lysates (Figure 2A and B and Supplementary Figure 3A and B). Analysis of YAP1 and TAZ immunostaining on human pancreatic tissue microarrays showed strong up-regulation of YAP1 in human ADM and PanIN pre-neoplastic lesions compared with normal acini (Figure 2C–E). TAZ was strongly up-regulated in ADM and to a lesser extent in PanIN lesions in mouse models and human tissues (Figure 2A and C–E). Surprisingly, we found a significant down-regulation of YAP1 and TAZ protein levels in human PDAC samples compared with ADM lesions (Figure 2C–E), suggesting a functional role of YAP1 and TAZ particularly during tumor initiation.

Figure 2.

YAP1 and TAZ are up-regulated in ADM lesions induced by KRas^G12D. (A) Immunohistochemical analysis of YAP1 and TAZ in normal (LSL-KRas^G12D) pancreas and pancreas from 6-month-old LSL-KRas^G12D;Pdx1-Cre mice. Scale bars = 50 μm. Dotted lines highlight mouse PanIN lesions. n = 4–5 mice analyzed. (B) Immunoblot analysis of pancreas lysates from control LSL-KRas^G12D and LSL-KRas^G12D;Pdx1-Cre mice. p-Lats1, phosphorylated Lats1 (Ser909); p-Yap1, phosphorylated Yap1 (Ser112). (C) Immunohistochemical analysis of YAP1 and TAZ on human pancreatic tissue microarrays. Scale bars = 50 μm. (D, E) Quantification of immunohistochemical staining intensity, ranging from absent (0) to highest (6) for YAP1 (D) and TAZ (E) on human pancreatic tissue microarrays. n = number of tissue cores. Means ± SD are shown. ^∗∗P < .01; ^∗∗∗P < .005, Student t test.

YAP1 and TAZ Are Necessary and Sufficient for Acinar-to-Ductal Metaplasia Induction

To determine the functional requirement for YAP1/TAZ signaling in acinar-to-ductal reprogramming, we generated conditional floxed alleles of both Yap1 and Taz (Supplementary Figure 4A–F). YAP1 and TAZ were efficiently deleted in Yap1^fl/fl;Taz^fl/fl primary acinar cells (Supplementary Figure 5A and B). We combined the floxed Yap1^fl/fl and Taz^fl/fl alleles with the Cre-inducible oncogenic LSL-KRas^G12D allele, shown to induce ADM in cultured acinar cells,¹⁴ and assayed ADM in vitro (Figure 3A). As expected, Cre-mediated activation of KRas^G12D transformed acinar cell clusters into tubular duct-like structures (Figure 3B). Deletion of Yap1 significantly reduced the ability of KRas^G12D to induce ADM, but deletion of Taz did not. However, in double-mutant Yap1/Taz acinar cells, KRas^G12D failed to induce ADM (Figure 3B and Supplementary Figure 5C and D).

Figure 3.

YAP1 and TAZ are necessary and sufficient for ADM induction. (A) Scheme showing in vitro assay for KRas^G12D-induced ADM. (B) Brightfield images of pancreatic acinar cell clusters on day 5 after isolation from LSL-KRas^G12D and LSL-KRas^G12D;Yap1^fl/fl;Taz^fl/fl mice. Cells were infected with adenoviruses encoding either GFP (Ad-GFP) or GFP plus Cre (Ad-Cre) on day 0. White arrows: tubular ductal structures. Scale bars = 50 μm. Quantification of tubular ductal structures of the indicated genotypes with n = 3–4 mice per group; means ± SEM are shown; ^∗∗∗P < .005, n.s., not significant, Student t test. (C) Experimental design of caerulein-induced acute pancreatitis in the acinar-specific and tamoxifen-inducible Ela1-Cre^ERT2 mouse model combined with the R26-LSL-YFP lineage tracer. (D) Triple immunofluorescence of pancreata from R26-LSL-YFP;Ela1-Cre^ERT2 and R26-LSL-YFP;Yap1^fl/fl;Taz^fl/fl;Ela1-Cre^ERT2 mice treated with caerulein as indicated in (C). Scale bars = 20 μm. Dotted white lines indicate ADM lesions. Quantification of GFP-positive ADM cells (CK19/amylase double-positive) as a percentage of total GFP-positive cells in caerulein-treated mice of the indicated genotypes. n = 3 mice per group; means ± SEM are shown; ^∗∗P < .01, Student t test. (E) Scheme showing in vitro assay for YAP1 and TAZ-induced ADM. (F) Brightfield images (upper panels) and GFP signal (lower panels) of acinar cells after 5 days in culture, infected on day 0 with Ad-GFP, Ad-YAP1-5SA-GFP, or Ad-TAZ-S89A-GFP. Scale bars = 25 μm. Quantification of GFP-positive tubular ductal structures at day 5, in n = 3 experiments; means ± SEM are shown; ^∗∗∗P < .005, Student t test. DAPI, 4′,6-diamidino-2-phenylindole dihydrochloride.

To address the requirement for YAP1 in vivo, we combined Yap1^fl/fl with the Pdx1-Cre mouse line. Caerulein-induced ADM lesions co-expressing CK19 and amylase were readily observed in Yap1^fl/fl;Pdx1-Cre mice. However, these lesions were positive for YAP1 nuclear staining (Supplementary Figure 6A), suggesting that ADM occurred in cells that had escaped Yap1 deletion. Consistent with this, polymerase chain reaction and immunoblot analysis showed that the recombination efficiency and consequent depletion of YAP1 was incomplete in Yap1^fl/fl;Pdx1-Cre mice (Supplementary Figure 6B and C). As an alternative model, we used a tamoxifen-inducible Cre line under the control of the acinar-specific Ela1 promoter (Ela1-Cre^ERT2) in combination with a GFP lineage tracer (R26-LSL-YFP) to allow identification of recombined cells (Figure 3C). In this model, GFP-positive acinar cells that had lost both Yap1 and Taz were significantly impaired in caerulein-induced ADM formation (Figure 3D), supporting a requirement for YAP1 and TAZ in ADM formation in vivo. Importantly, co-deletion of Yap1 and Taz did not affect acinar cell viability (Supplementary Figure 7A).

We next examined the effects of ectopic YAP1/TAZ activation in acinar cells by generating adenoviral vectors encoding constitutively active phosphorylation mutants of YAP1 (Ad-GFP-YAP1-5SA) and TAZ (Ad-GFP-TAZ-S89A) (Supplementary Figure 7B–D). These mutants are predominantly nuclear and able to induce target gene expression^{32, 33}: as expected, adenoviral delivery of YAP1-5SA, as well as TAZ-S89A, resulted in up-regulation of their target genes Ctgf and Cyr61 in primary acinar cells (Supplementary Figure 7E). Strikingly, acinar explants infected with either Ad-GFP-YAP1-5SA or Ad-GFP-TAZ-S89A, but not Ad-GFP, underwent conversion to a duct cell morphology (Figure 3E and F). These data suggest that activation of the transcriptional regulators YAP1 and TAZ in acinar cells is sufficient to induce ADM.

YAP1/TAZ Activation Increases Expression of JAK–STAT3 Pathway Components

To investigate the molecular mechanisms downstream of YAP1/TAZ, we analyzed the published transcription profile of cells overexpressing YAP1.²³ Gene set enrichment analysis of BIOCARTA-annotated pathways in this dataset³⁴ identified the interleukin (IL)6–JAK–STAT3 signaling cascade as strongly regulated by YAP1 overexpression (Supplementary Figure 8A). Importantly, JAK–STAT3 signaling was also strongly up-regulated in mice with pancreatitis (Supplementary Figure 1A), and STAT3 is known to be important for oncogenic KRas-induced pancreatic tumor development.^{28, 29, 35} To validate the involvement of this pathway downstream of YAP1/TAZ, we analyzed the expression of JAK–STAT3 pathway components in the human PDAC cell line PANC1, which shows constitutively active YAP1/TAZ (Supplementary Figure 8B and C). The receptors LIFR, GP130, IL13RA1, and the transcription factor STAT3 were down-regulated upon YAP1 and TAZ knockdown in PANC1 cells (Figure 4A). Accordingly, YAP1/TAZ knockdown reduced STAT3 and LIFR protein levels (Figure 4A and Supplementary Figure 8D). Overexpression of constitutively active YAP1-5SA or TAZ-S89A led to increased protein levels of STAT3 and LIFR in primary mouse acinar cells (Figure 4B and Supplementary Figure 9A and B).

Figure 4.

YAP1/TAZ activation controls JAK-STAT3 signaling via transcription of STAT3. (A) Quantitative reverse transcription polymerase chain reaction (RT-PCR) of the indicated genes and immunoblots of lysates from PANC1 cells expressing short-hairpin RNAs (shRNAs) against YAP1 and TAZ (shYAP/TAZ-#1 and -#2) or non-targeted shRNA control (shNT). n = 3; means ± SEM shown. (B) Immunoblots of primary acinar cell lysates 2 days after isolation and infection with Ad-GFP, Ad-YAP1-5SA-GFP, or Ad-TAZ-S89A-GFP. (C) Quantitative chromatin immunoprecipitation PCR of the LIFR, GP130, and STAT3 loci (promoter and intron regions) using either IgG or an antibody against TEAD4. ZFP37 intron = negative control region; CTGF promoter = positive control region. n = 3 experiments; means ± SEM shown. (D) Scheme showing luciferase expression construct with human STAT3 promoter region. Base pair (bp) numbers indicate position relative to STAT3 transcription start site. Red box indicates GT-IIc TEAD binding motif; genomic sequence of WT GT-IIc shown below. Mutant STAT3 promoter construct with a 6-bp deletion of the GT-IIc site (mut GT-IIc) was used as control. Luciferase expression analysis using wild-type and mutant STAT3 promoter regions in AR42J acinar cells 1 day after transfection with empty vector, YAP1-5SA, or TAZ-S89A. An artificial TEAD luciferase reporter consisting of 8 GT-IIc motifs was used as positive control. n = 3–5 experiments; means ± SEM shown. (E) Immunohistochemistry showing phosphorylated STAT3 (Tyr705) in pancreata of caerulein- and PBS-treated C57BL/6 mice. Scale bars = 50 μm. n = 3 mice analyzed. (F) Immunoblots of pancreas lysates from PBS- and caerulein-treated animals on day 2 and day 7 after treatment. p-Stat3, phosphorylated Stat3 (Tyr705). Quantitative RT-PCR of JAK–STAT3 pathway genes in caerulein-treated vs PBS-treated pancreata 2 days after treatment. n = 3–4 mice per group; means ± SEM shown. ^∗P < .05; ^∗∗P < .01; ^∗∗∗P < .005, n.s., not significant, Student t test.

We next tested whether the JAK–STAT3 pathway genes are direct transcriptional targets of the YAP1/TAZ-dependent TEAD transcription factors.³⁶ The promoter regions of STAT3, LIFR, and GP130 contain TEAD4 binding sites, as analyzed by genome-wide ENCODE chromatin immunoprecipitation experiments³⁷ (Supplementary Figure 9C). We confirmed that TEAD4 binding was significantly enriched at the promoters of STAT3, LIFR, and GP130 by chromatin immunoprecipitation (Figure 4C). In the STAT3 promoter region, one TEAD consensus binding site GT-IIc³⁸ was found 312 bp to 306 bp upstream of the STAT3 transcription start site (Figure 4D). We used the entire STAT3 promoter region including this site in a luciferase reporter construct transfected into the rat pancreatic acinar cell line AR42J. Overexpression of either YAP1-5SA or TAZ-S89A in this cell line significantly increased the luciferase activity using the STAT3 promoter construct compared with empty vector control (Figure 4D). Mutation of the TEAD binding site abolished the YAP1- and TAZ-induced luciferase activity (Figure 4D). Deletion of Stat3 did not change the expression or protein levels of YAP1 and TAZ in primary acinar cells (Supplementary Figure 10A–C). Our results suggest that, through TEAD, YAP1 and TAZ regulate the gene expression of the cytokine receptors LIFR and GP130 and the transcription factor STAT3.

JAK–STAT3 Pathway Activation in Acinar-to-Ductal Metaplasia Lesions

To examine whether JAK–STAT3 signaling is involved in ADM, we analyzed STAT3 activation in response to caerulein-induced pancreatitis. Caerulein treatment led to widespread active, phosphorylated STAT3 (pTyr705-STAT3) in the injured pancreas (Figure 4E), in accordance with previous reports.²⁹ STAT3 phosphorylation and up-regulation of STAT3 and LIFR proteins were also detected in caerulein-treated pancreas lysates (Figure 4F and Supplementary Figure 10D and E), and Stat3, Lifr, Gp130, and Jak1 were up-regulated in injured pancreata at the transcriptional level (Figure 4F), in agreement with our bioinformatic analysis (Supplementary Figure 1A). Co-staining with CK19 and amylase revealed increased staining intensity of STAT3 and LIFR in KRas^G12D-induced ADM cells in the Pdx1-Cre model (Supplementary Figure 11A and B).

YAP1/TAZ-Dependent Extreme Acinar-to-Ductal Metaplasia Sensitization by Oncogenic KRas^G12D

The up-regulation of the JAK–STAT pathway by YAP1/TAZ predicted an increased susceptibility of KRas^G12D-mutant acinar cells to inflammation, independently of oncogenic transformation. To test this, we challenged control (R26-LSL-YFP;Ela1-Cre^ERT2) and R26-LSL-YFP;LSL-KRas^G12D;Ela1-Cre^ERT2 mice with caerulein to induce pancreatitis (Figure 5A). As expected, control mice treated with low doses of caerulein showed acinar cell loss, inflammation, and a limited number of ADM lesions. However, R26-LSL-YFP;LSL-KRas^G12D;Ela1-Cre^ERT2 mice displayed a dramatic increase in ADM lesions double-positive for amylase and CK19 after caerulein treatment, although the recruitment of inflammatory F4/80-positive macrophages was similar to caerulein-treated control mice (Figure 5B, C, and Supplementary Figure 12A). Lineage tracing with the Cre-inducible R26-LSL-YFP allele confirmed the acinar cell origin of the ADM lesions (Supplementary Figure 12B). Thus, oncogenic Ras mutations dramatically increase the susceptibility of acinar cells to undergo ADM, possibly by increasing acinar cell sensitivity to inflammatory cytokines through up-regulation of Stat3, Lifr, and Gp130.

Figure 5.

YAP1 and TAZ are required for KRas^G12D-induced ADM in vivo in response to pancreatitis. (A) Scheme showing experimental design of caerulein-induced acute pancreatitis in cooperation with acinar-specific KRas^G12D mutation. (B) H&E stain and amylase and CK19 antibody stains of pancreata from mice of the indicated genotypes. Scale bars: H&E (top row) = 2 mm; (2nd row) = 200 μm; (3rd row) = 50 μm; amylase = 200 μm; CK19 (5th row) = 200 μm; (6th row) = 50 μm; double immunofluorescence = 50 μm. (C) Quantification of CK19⁺amylase⁺ ADM pancreatic area in pancreatic sections of the genotypes shown in (B). n = 3 mice per genotype; means ± SEM are shown; ^∗∗∗P < .005, Student t test. (D) Immunoblots of pancreas lysates from mice of the indicated genotypes treated as shown in (A). DAPI, 4′,6-diamidino-2-phenylindole dihydrochloride.

We next tested the significance of YAP1/TAZ and STAT3 in Ras-induced ADM sensitization directly. Strikingly, co-deletion of Yap1 and Taz completely blocked the enhanced formation of these pancreatitis-induced ADM lesions by KRas^G12D (Figure 5B and C). Pancreata of R26-LSL-YFP;LSL-KRas^G12D;Yap1^fl/fl;Taz^fl/fl;Ela1-Cre^ERT2 mice showed normal morphology with some signs of caerulein-induced acinar cell loss and inflammation similar to the extent observed in control mice (Figure 5B and Supplementary Figure 12A). Additionally, in agreement with previous results,²⁹ loss of STAT3 significantly reduced caerulein-induced CK19-positive ductal structures in the mutant KRas^G12D background (Figure 5B and C). Ras-induced ADM formation resulted in increased expression of the pancreatic progenitor marker Pdx1 and the duct marker Sox9, which was dependent on YAP1/TAZ (Figure 5D and Supplementary Figure 13A). These results suggest that expression of KRas^G12D in normal, untransformed acinar cells results in extreme sensitization to pancreatitis-induced ADM, which is mediated by YAP1/TAZ and STAT3.

JAK–STAT3 Pathway Acts Downstream of YAP1/TAZ in Response to Oncogenic Ras and Inflammatory Stimuli

We next addressed the function of YAP1/TAZ and JAK-STAT3 signaling as effectors of oncogenic Ras. KRas^G12D was activated in adult acinar cells using tamoxifen-inducible LSL-KRas^G12D;Ela1-Cre^ERT2 mice, and gene expression was analyzed 7 days later, when pancreas histology was unaltered and no tumor initiation had occurred. KRas^G12D activation in normal untransformed acinar cells resulted in increased expression of the YAP1/TAZ target genes Ctgf and Cyr61, accompanied by transcriptional induction of Stat3, Lifr, and Gp130 (Figure 6A). Importantly, activation of STAT3 pathway genes by oncogenic KRas^G12D in this context occurred without caerulein treatment. After deletion of Yap1^fl/fl;Taz^fl/fl, KRas^G12D was unable to induce YAP1/TAZ target genes or Stat3, Lifr, and Gp130 expression (Figure 6A). KRas^G12D activation in primary acinar cells in vitro increased YAP1 and TAZ protein levels, STAT3 protein levels and STAT3 phosphorylation (Figure 6B and Supplementary Figure 13B), but deletion of Yap1^fl/fl;Taz^fl/fl modestly reduced STAT3 protein levels and greatly impaired the KRas^G12D-induced phosphorylation of STAT3 in acinar cells in vitro (Figure 6B and Supplementary Figure 13B). In agreement with these results, caerulein treatment of tamoxifen-induced LSL-KRas^G12D;Ela1-Cre^ERT2 mice led to a dramatic increase in STAT3 phosphorylation at tyrosine 705, which was completely abolished in the absence of YAP1 and TAZ (Figure 6C, Figure 5D, and Supplementary Figure 13A). YAP1 and TAZ deletion also diminished the caerulein-induced phosphorylation of STAT3 in pancreatic acinar cells wild type for KRas (Supplementary Figure 14A and B). These data indicate that, even in the absence of an inflammatory stimulus or tumorigenesis, oncogenic Ras mutation leads to YAP1/TAZ and JAK–STAT3 pathway activation, and YAP1 and TAZ act upstream of the JAK-STAT3 signaling pathway in oncogenic KRas-induced ADM.

Figure 6.

KRas^G12D- and caerulein-mediated activation of STAT3 depends on YAP1 and TAZ. (A) Quantitative reverse transcription polymerase chain reaction of the YAP1/TAZ target genes Ctgf, Cyr61, and Amotl2 and the STAT3 pathway genes Stat3, Lifr, and Gp130 in pancreas lysates from Ela1-Cre^ERT2, LSL-KRas^G12D;Ela1-Cre^ERT2 and LSL-KRas^G12D;Yap1^fl/fl;Taz^fl/fl;Ela1-Cre^ERT2 mice 7 days after the last tamoxifen injection. n = 3–5 mice per genotype; means ± SEM are shown; ^∗P < .05; ^∗∗P < .01; n.s., not significant, Student t test. (B) Immunoblot analysis of primary acinar cells from wild-type (WT), LSL-KRas^G12D and LSL-KRas^G12D;Yap1^fl/fl;Taz^fl/fl mice, 2 days after isolation and infection with Ad-Cre adenovirus. pStat3, phosphorylated Stat3 (Tyr705). (C) Immunohistochemistry with antibodies against phosphorylated STAT3 (Tyr705) and STAT3 in pancreata of the indicated genotypes. Mice were treated with tamoxifen and caerulein as indicated in Figure 5A. Scale bars = 50 μm. n = 3–5 mice analyzed.

Deletion of YAP1 and TAZ Reduces Inflammation-Induced Ras Activation

Inflammatory insults lead to hyperstimulation of oncogenic Ras activity, and interfering with the inflammatory signaling cascade interrupts this feedback activation.¹⁸ We next tested whether YAP1/TAZ affects oncogenic Ras activity. Caerulein treatment increased Ras activity in mice expressing KRas^G12D in acinar cells (Supplementary Figure 15A and B), as shown previously.¹⁸ Deletion of Yap1 and Taz in the KRas^G12D background reduced Ras activation after caerulein treatment (Supplementary Figure 15A and B). Overexpression of constitutively active YAP1 in primary acinar cells increased Ras activity (Supplementary Figure 15C). Hyperactivating oncogenic Ras by TGFα resulted in a more efficient differentiation of acinar to duct cells in vitro³⁹ (Supplementary Figure 16A and B). However, YAP1 and TAZ were required for KRas^G12D-induced ADM, even in the presence of TGFα (Supplementary Figure 16A and B).

YAP1 and TAZ Are Required for KRas^G12D-Induced Pancreatic Intraepithelial Neoplasia Formation

To determine whether the block to ADM in Yap1^fl/fl;Taz^fl/fl;LSL-KRas^G12D;Ela1-Cre^ERT2 mice affects the development of pancreatic cancer, we assessed PanIN lesions 3 months after the transient induction of pancreatitis by caerulein (Figure 7A). At this stage, LSL-KRas^G12D;Ela1-Cre^ERT2 mice displayed CK19/amylase double-positive ADM lesions and CK19- and AB/PAS-positive PanIN lesions (Figure 7B–D). Yap1 and Taz co-deletion resulted in a striking decrease in ADM lesions and these mice were free of PanIN lesions, similar to control mice (Figure 7B–D). These results show that acinar cell-specific YAP1/TAZ signaling is essential for oncogenic KRas^G12D-induced PanIN formation in the context of pancreatitis, and that STAT3 signaling appears to be a critical downstream factor in this process.

Figure 7.

YAP1 and TAZ are required for KRas^G12D-induced PanIN. (A) Scheme showing experimental design of caerulein-induced PanIN formation in the KRas^G12D background. (B) H&E, amylase, and CK19 antibody, and Alcian Blue/periodic acid–Schiff (AB/PAS) stains of pancreata from mice of the indicated genotypes. Scale bars: H&E = 200 μm; CK19 = 200 μm; double immunofluorescence = 50 μm; AB/PAS (4th row) = 200 μm; AB/PAS (5th row) = 50 μm. (C) Quantification of CK19⁺amylase⁺ ADM lesions in control, LSL-KRas^G12D;Ela1-Cre^ERT2 and LSL-KRas^G12D;Yap1^fl/fl;Taz^fl/fl;Ela1-Cre^ERT2 mice. n = 3–4 mice per genotype; means ± SEM are shown; ^∗P < .05, Student t test. (D) Quantification of AB/PAS-positive PanIN lesions in control, LSL-KRas^G12D;Ela1-Cre^ERT2, and LSL-KRas^G12D;Yap1^fl/fl;Taz^fl/fl;Ela1-Cre^ERT2 mice. n = 3–4 mice per group; means ± SEM are shown; ^∗P < .05, Student t test. (E) Scheme showing the roles of YAP1 and TAZ in pancreatic cancer initiation by inducing ADM in response to oncogenic KRas^G12D and inflammation. See text for details.

Discussion

Acinar-to-ductal reprogramming is now established as an important precursor to PanIN lesions and progression to pancreatic cancer, but the molecular mechanisms that drive this process have remained unclear. Our results confirm that KRas^G12D mutation synergizes with inflammatory insults to induce widespread and persistent ADM lesions, and we elucidate how this synergy occurs (Figure 7E). We find that the transcriptional regulators YAP1 and TAZ are required for KRas^G12D to promote acinar-to-ductal reprogramming, both in vitro and in vivo. In the absence of YAP1 and TAZ, KRas^G12D is unable to induce PanIN lesions, suggesting that acinar-to-ductal reprogramming via YAP1/TAZ is an obligate step in pancreatic cancer initiation from acinar cells.

Mechanistically, active YAP1 and TAZ interact with the transcription factor TEAD, which binds and up-regulates the expression of several genes in the JAK–STAT3 pathway. This pathway controls the response to inflammatory cytokines like IL6, which is secreted during caerulein-induced pancreatitis and also present at high levels in pancreatic tumors, promoting PanIN progression in KRas^G12D mice.^{35, 40} A recent study found that the IL6 co-receptor GP130, up-regulated by YAP1/TAZ in our study, also activates YAP1 independently of STAT3.⁴¹ Additionally, LIFR has been shown to act upstream of YAP1 signaling.⁴² This would create a positive feedback loop amplifying YAP1 activity.

Our data in mouse models expressing KRas^G12D are consistent with findings in cultured cells showing that Ras signaling both stabilizes and activates YAP1.^{43, 44} Recent studies have implicated YAP1 as an important driver of pancreatic cancer progression at later stages in collaboration with oncogenic Ras, and as a mechanism of escape from Ras-RAF-MEK-targeted therapies.^{19, 20, 45} However, data from patient samples has shown that the incidence of KRas codon 12 mutations is already >90%, even in the earliest pre-neoplastic lesions.⁸ Here we show that YAP1 acts downstream of oncogenic Ras at this early reprogramming stage. There is likely redundancy between YAP1 and TAZ because co-deletion of Yap1 and Taz is more efficient in blocking KRas^G12D- induced ADM in the in vitro assay compared with either single mutant. YAP1 seems able to compensate for TAZ deletion but not vice versa, however, overexpression of constitutively active YAP1 or TAZ alone is sufficient to induce ADM.

We noted that inactivation of YAP1 and TAZ reduced ADM formation more efficiently than Stat3 deletion (Figure 5). It has been shown that the Hippo pathway also interacts with other signaling pathways that have implications in ADM, such as Wnt and Notch pathways.^{23, 46} YAP1/TAZ-dependent regulation of these pathways might also contribute to ADM formation.

The presence of the KRas^G12D mutation in ADM cells in an inflammatory microenvironment is known to be sufficient for progression to PanIN lesions and pancreatic cancer.¹ Additionally, inflammation in the pancreas creates a feedback loop leading to hyperactivation of oncogenic KRas^G12D (Figure 7E).¹⁸ The KRas^G12D mutation and subsequent hyperactivation promotes pancreatic cancer at 2 levels: it creates a susceptible “cell of origin” by inducing acinar-to-ductal reprogramming; and it also acts as the transforming driver that enables progression toward malignancy. This model explains why KRas mutations in pancreatic cancer are so common, and why they frequently appear very early, as a gatekeeper mutation, in pancreatic acinar cells.