GNAS^R201C Induces Pancreatic Cystic Neoplasms in Mice That Express Activated KRAS by Inhibiting YAP1 Signaling

Abstract

Background & Aims:

Mutations at hot spots in GNAS, which encodes stimulatory G-protein, alpha subunits, are detected in ~60% of intraductal papillary mucinous neoplasms (IPMNs) of the pancreas. We generated mice with KRAS-induced IPMNs that also express a constitutively active form of GNAS in pancreas and studied tumor development.

Methods:

We generated p48-Cre; LSL-KrasG12D; Rosa26R-LSL-rtTA-TetO-GnasR201C mice (Kras;Gnas mice); pancreatic tissues of these mice express activated KRAS and also express a mutant form of GNAS (GNAS^R201C) upon doxycycline administration. Mice that were not given doxycycline were used as controls and survival times were compared by Kaplan-Meier analysis. Pancreasta were collected at different timepoints after doxycycline administration and analyzed by histology. Pancreatic ductal adenocarcinomas (PDACs) were isolated from mice and used to generate cell lines, which were analyzed by reverse transcription PCR, immunoblotting, immunohistochemistry, and colony formation and invasion assays. Full-length and mutant forms of YAP were expressed in PDAC cells. IPMN specimens were obtained from 13 patients with IPMN undergoing surgery and analyzed by immunohistochemistry.

Results:

All Kras;Gnas mice developed pancreatic cystic lesions that resemble human IPMNs; the grade of epithelial dysplasia increased with time. None of the control mice developed cystic lesions. Approximately one-third of Kras;Gnas mice developed PDACs, at a median of 30 weeks after doxycycline administration, whereas 33% of control mice developed PDACs. Expression of GNAS^R201C did not accelerate the development of PDACs, compared with control mice. However, the neoplasms observed in Kras;Gnas mice were more differentiated, and expressed more genes associated with ductal phenotypes, than in control mice. PDACs isolated from Kras;Gnas mice had activation of the Hippo pathway; in cells from these tumors, phosphorylated YAP1 was sequestered in the cytoplasm—this was also observed in human IPMNs with GNAS mutations. Sequestration of YAP1 was not observed in PDAC cells from control mice.

Conclusions:

In mice that express activated KRAS in the pancreas, we found expression of GNAS^R201C to cause development of more differentiated tumors, with gene expression pattern associated with the ductal phenotype. Expression of mutant GNAS caused phosphorylated YAP1 to be sequestered in the cytoplasm, altering tumor progression.

INTRODUCTION

Pancreatic ductal adenocarcinoma (PDAC) is the 3^rd most common cause of cancer-related mortality in the United States ¹. The multistep progression of PDAC can occur on the backdrop of non-cystic, microscopic precursor lesions known as pancreatic intraepithelial neoplasias (PanINs), or alternatively, through a cystic pathway that portends a distinct natural history ². Overall, pancreatic cysts are identified in greater than 2 % of the general patient population, rising to 10% amongst individuals above 70 years of age ³. Many of the cystic lesions are essentially benign, and harbor no risk of progression to PDAC. In contrast, mucin-secreting cysts of the pancreas, specifically intraductal papillary mucinous neoplasms (IPMNs), are bona fide precursor lesions of PDAC ³. Patients with non-invasive IPMNs have a 5-year survival rate of 90-100 % upon surgical resection, which is reduced to about 50% for those with an invasive component ⁴ Nonetheless, in contrast to PDAC that arises through the non-cystic (PanIN) pathway to invasive neoplasia, and which is nearly always fatal, cancers arising in the context of IPMNs demonstrate a more indolent course. Further, even within the universe of IPMNs, the vast majority does not progress to PDAC, although the biological basis for this favorable outcome has, thus far, remained elusive.

Progress in translational research on pancreatic cystic lesions has been hampered by lack of animal models that recapitulate the biology and genetic of the cognate human disease. In 2011, we performed a comprehensive next generation sequencing (NGS) assessment of IPMNs, and identified two recurrent driver alterations, oncogenic mutations of KRAS and GNAS ⁵. Specifically, activating mutations of KRAS were observed in 80% of cases, while activating mutations of GNAS, which encodes for the alpha subunit of a stimulatory G-protein (Gαs protein), were present in 66% of IPMNs. GNAS mutations were restricted to codon 201, a previously described “hotspot” in epithelial neoplasms. Overall, 96 % of IPMNs harbored either KRAS or GNAS mutations ^{5, 6}, underscoring the importance of these alternations in IPMN pathogenesis. Further, our prior studies have confirmed that GNAS mutations are observed early in IPMN pathogenesis, including in precursor lesions of IPMNs, the so-called “incipient IPMNs”, establishing these mutations as an early driver of mucinous neoplasms of the pancreas ⁷. However, the precise functional role played by mutant Gαs protein in cooperation with oncogenic Ras during IPMN progression has been incomplete due to lack of an autochthonous model that recapitulates the cognate human disease process.

Under homeostatic conditions, Gαs transiently activates adenylate cyclase and increases cyclic adenosine monophosphate (cAMP) production as a second messenger. Although Gαs possesses intrinsic enzymatic activity that hydrolyzes bound GTP to GDP, thereby returning them to active form, GNAS R201 mutations markedly reduce efficiency of GTP hydrolysis that render Gαs constitutively active. This results in sustained elevation of the intracellular cAMP level ⁸, which subsequently activates protein kinase A (PKA) and downstream signaling cascades ⁹. Activating mutations of GNAS were first described in the context of endocrine neoplasms like pituitary adenomas and juvenile granulosa cell tumors ^{8, 10, 11}, and subsequently, in epithelial tumors, such as in a minor subset of colorectal cancers ¹². The recent generation of a genetically engineered mouse (GEM) model of conditional (doxycycline inducible) GNAS^R201C activation by Gutkind and colleagues ¹³ has now allowed us to examine the role of constitutively active Gαs in IPMN pathogenesis. While our GEM model faithfully recapitulates the multistep progression of human IPMNs, we also elucidate the molecular mechanisms through which Gαs elaborates a cystic differentiation with the context of mutant Ras signaling. Specifically, we demonstrate that Gαs does not behave as a traditional cooperating oncogene, but rather as a modulator of differentiation of Ras-induced pancreatic neoplasia (“onco-modulator”) through attenuation of YAP1 signaling within epithelial cells. Our studies likely clarify the natural history of IPMNs and the low rate of progression to invasive neoplasia on the background of cystic precursors.

MATERIALS AND METHODS

Genetically engineered mice

All animal studies were carried out according to the MD Anderson Institutional Care and Use of Animals Committee (IACUC)-approved protocols. To model the co-expression of mutant KRAS and GNAS in the murine pancreatic epithelium, we used a previously described doxycycline (doxy)-inducible GNAS^R201C model ¹³. In this model, tissue-specific expression of GNAS^R201C is regulated by crossing the mice to an appropriate Cre-driver (in this case, p48-Cre). This leads to removal of a lox-STOP-lox (LSL) cassette upstream of a reverse tetracycline transactivator (rtTA) and expression of rtTA in the pancreatic epithelium. In the presence of doxy, rtTA binds to the Tet operon (TetO) upstream of Gnas^R201c leading to transgene expression in the pancreas ^{13, 14} We generated compound heterozygous p48-Cre; LSL-KrasG12D; Rosa26R-LSL-rtTA-TetO-GnasR201C mice (Kras;Gnas mice) (Figure 1A). Doxycycline diet was fed at a dose of 0.0060 %. The time course for supplementation is outlined in the Figure 1B. The confirmation of recombination of Kras^G12D and transcription of Gnas^R201C are shown in the pancreatic lysates in Supplementary Figure 1A and B.

Figure 1. Concomitant expression of GNAS^R201C and KRAS^G12D in the mouse adult pancreas induces the development of cystic neoplasms resembling human IPMN.

(A) Schema for conditional activation of pancreas-specific GNAS^R201C expression using doxycycline (doxy) on a background of constitutive KRAS^G12D expression.

(B) Kras;Gnas mice were maintained on a doxy diet from the age of 8 weeks to induce GNAS^R201C. Littermate Kras;Gnas mice without doxy served as isogenic controls.

(C) Gross pathology of the pancreas in Kras;Gnas mice maintained without doxycycline (left) or with doxycycline (right), showing multiple cystic lesions throughout the pancreas of latter mice (time of harvest = 28 weeks).

(D) Representative pictures of pancreatic precursor lesions arising in mice with or without doxy induction, including murine pancreatic intraepithelial neoplasia (mPanIN) in control mice pancreata and intraductal papillary lesions in Kras;Gnas mice fed doxy diet until the indicated time points (scale bar, 50μm).

(E) Quantification of papillary structures in pancreata with or without GNAS^R201C expression at the indicated time points.

(F) Quantification of pancreatic lesions in Kras;Gnas mice pancreata with or without GNAS^R201C expression at the indicated time point.

(G) Quantification of the disease stages and histology in Kras;Gnas mice without doxy (n=18) and with doxy administration (n=17).

(H) Kaplan-Meier survival analysis for Kras;Gnas mice without doxy (n=18) and with doxy administration (n=17).

Histological evaluation of pancreas

To histologically evaluate the multistep progression of pre-invasive neoplasms, including mPanINs and murine IPMNs (mIPMNs), cystic papillary lesions (mIPMNs) in our mouse model were defined as epithelium-lined cystic structures ≥500μm, and composed of papillary epithelial proliferations with fibrovascular cores and varying degrees of cellular atypia, which is a sine qua non of human IPMNs ¹⁵. Pre-invasive ductal lesions which do not meet the criteria of mIPMN were defined as mPanINs, in other words, <500μm in greatest diameter with no discernible fibrovascular core. The grade of ductal lesions was evaluated in representative pancreatic sections of cohorts of Kras;Gnas mice, with or without doxycycline at the age of 18, 28, and more than 40 weeks, respectively (Figure 1B). Only the highest-grade lesion per pancreatic lobule was evaluated, using published criteria for low-grade (LG) or high-grade (HG) mPanIN and mIPMN ^15-17 An average of 171 total pancreatic lobules were counted from 3 independent animals for each group. The data were expressed as a percentage of the total counted pancreatic lobules. Cystic papillary structures ≥500μm in the representative pancreatic sections which met the definition were counted to assess the neoplastic nature of Kras;Gnas mice. In terms of invasive carcinomas, they were graded based on the most poorly differentiated component when multiple histopathlogical grades were observed.

Establishment of primary pancreatic ductal adenocarcinoma cell lines with doxycycline inducible mutant GNAS^R201C

Freshly isolated murine PDAC tissues were minced by sterile scalpel and washed with PBS containing 1% penicillin and streptomycin and re-suspended in RPMI with 10 % fetal bovine serum (FBS) and 1% penicillin and streptomycin, and seeded in 60 mm collagen coated dishes. Five days after seeding of tumor pieces, residual tissue was removed and attached colonies comprised of murine PDAC cells were isolated using 150μl cloning cylinders (Sigma-Aldrich, St Louis, MO) and expanded. For the purpose of this study, the two cell lines used in all of the experiments are designated as “LGKC-1” and “LGKC-2”, respectively. Both cell lines were used at less than passage 10 to minimize passage-dependent artifacts ¹⁸. To confirm these cell lines were epithelial in nature and derived from pancreatic epithelium with Kras^G12D recombination, genotyping PCR for Kras^G12D, qPCR for Gnas, and western blotting to detect epithelial markers Cytokeratin 19 and E-cadherin expression were performed (Supplementary Figure 2).

Allograft model for assessing tumor growth rates and differentiation

For in vivo growth and differentiation assays, murine PDAC cells (with or without additional genetic modifications of mYap1 as described in the text) were suspended in 100μl of 50% Matrigel/PBS mix and injected subcutaneously into the lower flanks of age-matched female NUDE mice (n=4 per group). Twenty four hours after injection, the mice were randomized to be fed normal diet or doxycycline diet (0.6%). Tumor volumes were measured every 3-4 days for 17 days and tumor weight was measured at the end of the experiment. Histological assessment was conducted for grade of tumor differentiation, and additional immunohistochemistry and western blot analyses on the explanted tumors were conducted, as described below.

Immunofluorescence and Immunohistochemistry

For immunofluorescence, LGKC cells were seeded on the collagen coated glass bottom dishes (Thermo fisher scientific, Grand Island, NY) and incubated 48 hours with RPMI media containing 10% FBS. Cells were incubated for 24 hours in the media with either PBS or 1μg/ml doxycycline to induce mutant GNAS and fixed with 100% cold methanol. The cells were blocked in 1xPBS, 1% BSA and 0.3% Triton X-100 for 1 hour followed by incubation with primary antibodies listed in Supplementary Table 1 at 4°C overnight. After washing, cells were incubated in conjugated secondary antibody. The proportion of cells per field showing YAP1 localization in the nucleus or cytoplasm was quantified. In 3 independent experiments, 4 fields with 30 to 50 cells were counted per condition per experiment. Methodologies for immunofluorescence and immunohistochemistry of paraffin-embedded samples were described in the Supplementary Materials and Methods.

Protein extraction and Western blotting

Protein extraction and western blotting were performed as described previously ²⁰. Additional details on methodology and assessment are provided in the Supplementary Materials and Methods.

Statistical analysis

All analyses were performed in triplicate or greater and the means obtained were used for independent t-tests. Statistical analyses were carried out using the Prism 6 statistical analysis program (GraphPad Software, La Jolla, CA). Asterisks denote statistical significance (non-significant or NS, P>0.05; *P<0.05; **P<0.01; and***P<0.001). All data are reported as mean ± standard error of the mean (s. e. m.).

RESULTS

Expression of GNAS^R201C in adult p48-Cre; LSL-KrasG12D; Rosa26R-LSL-rtTA-TetO-GnasR201C (Kras;Gnas) mice leads to cystic precursor lesions resembling human IPMN.

To determine how Gnas mutations alter the natural history of Kras-driven murine pancreatic neoplasia, we fed doxycycline (doxy) diet to adult Kras;Gnas mice starting at the age of 8 weeks to induce GNAS^R201C on a constitutive mutant KRAS^G12D expressing background (Figure 1A, B). In the absence of doxy, i. e. with mutant KRAS^G12D expression alone, the pancreata mainly demonstrated mouse pancreatic intraepithelial neoplasias (mPanINs), which are non-cystic precursor lesions of PDAC (Figure 1C, left). However, upon co-expression of mutant GNAS^R201C, 100 % of the mice developed cystic lesions beginning as early as 10 weeks of doxy treatment (Figure 1C, right). The neoplastic cells expressed apomucins MUCIN1 and MUCIN5AC but not MUCIN2, which resembles the phenotype pancreatbiliary subtype of human IPMNs (Supplemental Figure 3). There were no variations of mucin expression pattern across the histologic grades in Kras;Gnas mouse pancreata with GNAS^R201C expression. However, significant reduction of MUC5AC expression in mPanIN-HG lesions in those with KRAS^G12D expression alone, likely due to increased nuclear to cytoplasmic ratio.

To further evaluate the nature of pancreatic precursor lesions with or without mutant GNAS^R201C, the number of papillary structures with fibrovascular cores, which are a diagnostic sine qua non of IPMNs, was counted in representative pancreatic sections from doxy administered Kras;Gnas mice at the ages of 18 and 28 weeks, respectively. For each comparison, littermate Kras;Gnas mice with no doxy (i. e. with mutant KRAS^G12D expression alone) were used as control. The number of papillary structures was significantly increased in the pancreata of mice co-expressing mutant KRAS^G12D and GNAS^R201C, in comparison to those with mutant KRAS^G12D at both time points. In addition, more papillary structures were observed in the pancreata of the mice at the later time point, suggesting co-expression of mutant GNAS^R201C progressively enhances the IPMN-specific phenotype (Figure 1D, E).

The distribution for grades of all exocrine pancreatic precursor lesions including acinar-ductal metaplasia (ADM), mPanIN, and mouse IPMN (mIPMN) were also assessed to determine the effect of mutant GNAS^R201C on multistep progression of KRAS^G12D-driven neoplasia (Figure 1F). The proportion of pancreatic parenchyma affected by ADM or low-grade mPanIN or mIPMN was not significantly different, irrespective of the co-expression of mutant GNAS^R201C at the indicated time points of 18 and 28 weeks (Figure 1F). However, in terms of high-grade lesions, the percentage of pancreatic lobules with high-grade (HG) lesions were significantly higher in the pancreata with mutant Kras^G12D and Gnas^R201C co-expression than those in the pancreata with mutant Kras^G12D alone (1.9 v. s. 13.1 %, p=0.006, and 5.8 v. s. 23.6 %, p= 0.016 in 18 and 28 weeks old mice, respectively, Figure 1F). To assess the specific effect of GNAS^R201C on high-grade lesions, these lesions were stratified into mPanINs or mIPMNs in each cohort. There were significantly larger number of mIPMN-HG in Kras;Gnas mice receiving doxy diet (Supplementary Figure 4). These results indicate that GNAS^R201C co-expression accelerates formation of precursor lesions in the setting of mutant KRAS^G12D.

In spontaneously aged mice to 40 weeks and beyond, irrespective of the administration of doxy, virtually the entire pancreatic parenchyma was eventually replaced by ductal neoplastic lesions intermingled fibrotic areas due, in part to the constitutive expression of the mutant KRAS^G12D allele. Overall, 10 of 17 (59%) doxycycline-administered Kras;Gnas mice harbored mIPMN-HG, while only 6 of 18 (33%) no doxy control KRAS^G12D-expressing mice harbored mPanIN-HG (Figure 1G), which is consistent with the findings observed at earlier time points (Figure 1F).

Despite a greater prevalence of high-grade precursor lesions, however, there was no significant difference in the frequency of invasive carcinoma upon co-expression of GNAS^R201C. While the median survival of doxy administered Kras;Gnas mice were significantly shorter than control mice without doxy (Figure 1H). Of note, however, 8 of 10 doxy administered Kras;Gnas mice without invasive carcinoma were euthanized due to abdominal distention caused by large cyst formation within the pancreas, suggesting that measuring median survival alone was a confounding factor for assessing natural history of these mice.

In a cohort of 35 Kras;Gnas mice that was monitored for progression to invasive carcinoma (17 with doxy, and 18 without doxy administration), five mice in the doxy arm (29%) and six mice in the control arm (33%) developed invasive neoplasms at a median interval of 43 weeks. All of the invasive carcinomas developing in the doxy-induced Kras;Gnas mice were well to moderately differentiated ductal adenocarcinomas (Figure 2A, Supplementary Table 2); no colloid carcinomas, which are mucin-rich invasive cancers arising in the context of intestinal type human IPMNs, were observed. On the other hand, control (no doxy) mice predominantly develop poorly differentiated or undifferentiated adenocarcinomas (5 of 6 cases). Of the mice that developed cancers in each cohort, Kras;Gnas mice with doxy administration exhibited predominantly locally invasive PDAC without distant organ metastasis (4 of 5 cases), while control mice exhibited somewhat higher metastatic rate (3 of 6 cases). Immunohistochemical assessment of phospho-Map kinase (pMAPK) expression in PDAC arising in Kras;Gnas mice with or without doxy showed distinctly higher expression of pMAPK upon GNAS^R201C co-expression (Figure 2B), consistent with the additive effect of cAMP-signaling on activation of the MAP kinase pathway. To investigate mutational evolution in our model, immunohistochmical staining of P16, SMAD4, and TP53 was performed using sections containing invasive carcinomas and the status of these genes based on the nuclear expression in comparison to that of pre-invasive lesions ²¹. We found at least one of 3 genes was lost in 10 of 11 invasive carcinomas in Kras;Gnas mice with and without doxy administration, suggesting inactivation of tumor suppresser genes was required for progression to invasive carcinoma (Supplementary Table 3 and Supplementary Figure 4).

Figure 2. Generation of isogenic pancreatic cancer cell lines with inducible cyclic AMP (cAMP) activation from invasive carcinomas arising in Kras;Gnas mice.

(A) Representative gross appearance and histology of invasive carcinomas arising in the pancreas of Kras;Gnas mice maintained on either control diet or doxycycline diet (scale bar, 50μm).

(B) Immunostaining for phospho-mitogen-activated protein kinase (pMAPK) in invasive carcinomas arising in pancreata of Kras;Gnas mice fed either control diet or doxycycline diet (scale bar, 25μm).

(C) Morphology of two independent isogenic cell lines – LGKC-1 and LGKC-2 – generated from invasive carcinomas arising in Kras;Gnas mice. Mutant Kras^G12D allele is constitutively expressed in the lines, while GNAS^R201C expression can be induced by addition of doxycycline (scale bar, 100μm).

(D) LGKC-1 and LGKC-2 cells show significant upregulation of cAMP levels upon doxycycline induced Gnas^R201C activation at 12 and 24 hours post-induction.

(E) Western blot analysis confirms activation of phospho-protein kinase A (pPKA) substrates in LGKC-1 and LGKC-2 cells upon doxycycline induced Gnas^R201c activation.

(F) Western blot analysis demonstrates induced GNAS^R201C activates MAPK pathway in LGKC-1 cell and inactivates AKT pathway in LGKC-2 cell.

Co-expression of Gnas^R201C on Kras^G12D background paradoxically attenuates the oncogenic phenotype of murine PDAC cells

For further mechanistic analysis, we established PDAC cell lines from primary tumors that developed in the pancreata of 2 independent Kras; Gnas mice at the age of 8 and 10 months, respectively, which we named Linker; Gnas; Kras; Cre-1 (LGKC-1) and LGKC-2 cell lines (Figure 2C). Both of these cell lines represent an isogenic system, which constitutively express Kras^G12D, but Gnas^R201C is only co-expressed upon addition of doxy in vitro. Therefore, these lines can serve as a conduit for elucidating how aberrant cAMP alters signaling pathways on the backdrop of constitutively active Ras. Indeed, in vitro assays confirmed that both LGKC-1 and LGKC-2 lines significantly upregulated cAMP levels upon doxy administration (Figure 2D), as well as phospho-Protein Kinase A (pPKA) substrates (Figure 2E). Interestingly, significant modulations of Kras canonical pathway were observed by inducing GNAS^R201C in these cell lines. As well as autochthonous model, GNAS^R201C activated MAPK pathway in LGKC-1 cell, while phosphorylation of AKT was significantly reduced in LGKC-2 cell (Figure 2F), consistent with a previous study ²².

We then assessed anchorage-independent growth and invasiveness of LGKC-1 and LGKC-2 lines with or without doxy administration. Rather unexpectedly, the number of colony in soft agar was significantly decreased in both LGKC-1 and LGKC-2 lines by the induced expression of GNAS^R201C, with a more profound reduction of colony formation observed in the LGKC-2 line (Figure 3A). On the same lines, invasive capacity of both lines was also attenuated in an in vitro Matrigel assays, upon GNAS^R201C induction (Figure 3B). Mirroring the paradoxical in vitro data, subcutaneous allografts established from both LGKC-1 and LGKC-2 lines in doxy-fed athymic mice demonstrated reduced tumor weights compared to control allografted mice (Figure 3C); albeit, in contrast to LGKC-2 allografts, in LGKC-1 cells, the reduction in mean tumor weights did not reach statistical significance. Overall, however, the compendium of in vitro and in vivo data underscored the surprising finding that co-expression of GNAS^R201C on a KRAS^G12D background attenuated the transformed phenotype for established PDAC cells. The most remarkable difference was however in the histology of isogenic allografted tumors, where expression of Gnas^R201C was with doxy resulted in striking epithelial differentiation, which was not apparent in matched allografts established without doxy (Figure 3D). This effect of mutant GNAS^R201C on PDAC differentiation seems to be consistent with the findings in autochthonous Kras;Gnas mice. The epithelial differentiation induced by GNAS^R201C was more prominent in LGKC-1 line, compared to LGKC-2, but this was not surprising given that the isogenic LGKC-2 allografts without doxy were essentially undifferentiated.

Figure 3. Expression of GNAS^R201C in LGKC cells induces a striking differentiation phenotype in vivo.

(A) Quantification of colony formation in LGKC-1 and LGKC-2 cells treated with or without doxycycline for 2 weeks (* = P <0.05).

(B) Quantification of invasion assays in LGKC-1 and LGKC-2 cell treated with or without doxycycline for 48 h (** = P <0.01).

(C) Subcutaneous allografts derived from LGKC-1 (left) and LGKC-2 cells (right) in athymic mice fed either normal diet or doxycycline diet (n = 8 mice for each cell line and doxycycline treatment) (* = P <0.05).

(D) Representative histology of LGKC-1 (top panels) and LGKC-2 allografts (bottom panels) in athymic mice fed either normal diet or doxycycline diet (scale bar, 50μm).

Co-expression of Gnas^R201C on Kras^G12D background activates upstream Hippo kinase cascade and attenuates YAP1 signaling

We probed for the potential downstream pathways that might account for this striking induction of differentiation in mutant KRAS^G12D tumors upon co-expression of GNAS^R201C. Microarray analysis of the isogenic lines showed a prominent YAP1 repression signature in LGKC lines upon doxycycline induction (Supplementary Figure 5A), suggesting that cAMP signaling might be suppressing this critical transcriptional co-activator implicated in PDAC pathogenesis. Indeed, in both LGKC-1 and LGKC-2 cells, activation of GNAS^R201C signaling was associated with significant reduction in credentialed YAP1 target genes (mCtgf, mCyr61, mThbs1, mItgb5), albeit the effects were more striking in LGKC-2 cells (Supplementary Figure 5B), supporting the significant anchorage independent growth inhibition observed in these cell lines (Figure 3A). As a transcriptional co-activator, activated YAP1 is localized in the nucleus, while cytoplasmic sequestration is indicative of phosphorylation and inactivation by the upstream Hippo kinase cascade. Indeed, upon GNAS^R201C activation, YAP1 was substantially sequestered in the cytoplasm in both LGKC-1 and LGKC-2 cells (Figures 3A-B), consistent with inactivation. Assessment of nuclear and cytoplasmic extracts also confirmed evidence for cytoplasmic sequestration upon Gnas^R201C activation (Figure 4C), with no changes in the levels of the main YAP1 binding partner and transcription factor Tead2. In LGKC-1 and LGKC-2 cells, doxycycline induction led to time-dependent phosphorylation of the immediate upstream kinase LATS1, which in turn, phosphorylates and inactivates YAP1 at two distinct serine residues (S127 and S397) (Figure 4D). The impact of doxycycline induction on YAP1 activation status in both LGKC lines was pheno-copied by the cAMP elevator forskolin (Figure 4E), which further reiterated that the effects were through the canonical signal transduction pathway downstream of Gαs. Orthogonal validation for a role of mutant GNAS^R201C activation in regulating YAP1 activation through cytoplasmic sequestration was borne out by immunohistochemical staining for YAP1 expression in Kras;Gnas mice, with or without doxycycline (Figure 5A). In contrast to the strong nuclear localization of YAP1 protein in pancreatic precursor lesions (ADM, mPanINs) of all histological grades in mice without doxycycline administration (Top), cytoplasmic localization was the predominant pattern seen in Kras;Gnas mice receiving doxycycline (Bottom). This suggests that the impact of GNAS^R201C activation on the Hippo pathway is an early event, occurring at the step of pre-invasive neoplasia.

Figure 4. Induction of mutant Gnas^R201C leads to YAP1 cytoplasmic sequestration through the activation of Hippo pathway.

(A) Representative photomicrographs of non-confluent LGKC-1 (left) and LGKC-2 cells (right) untreated or treated with doxycycline for 24 hours and stained for YAP1 (green) and nuclei (DAPI, blue) (scale bar; 50μm).

(B) The proportion of cells per field showing YAP1 localization in the nucleus or cytoplasm. n=3 independent experiments.

(C) Nuclear and cytoplasmic extracts from LGKC-1 and LGKC-2 cell lines demonstrates reduced nuclear YAP1 localization upon doxycycline induction. Of note, c-Myc protein levels (a YAP1 target gene) are also reduced in both cell lines upon doxycycline induction.

(D) Western blot analysis LGKC-1 and LGKC-2 cell lines demonstrates activation of the upstream Hippo kinase cascade (specifically phosphorylation of LATS1) upon doxycycline induction, which is accompanied by inactivating phosphorylation of YAP1 at residues S127 and S397.

(E) The cAMP elevator, forskolin, phenocopies the effects of Gαs induction on LATS1 and YAP1 S127/S397 phosphorylation.

Figure 5: Cystic neoplasia in the murine pancreas progresses along a YAP1 independent pathway and can be bypassed using a phosphorylation resistant YAP1 mutant allele.

(A) Immunohistochemical assessment of YAP1 nuclear localization shows reduced nuclear YAP1 localization in acinar-ductal metaplasia, low-grade and high-grade pancreatic precursor lesions arising in Kras;Gnas mice fed doxycycline versus those on control diet (scale bar, 50μm).

(B) In LGKC-1 allografts established with cells expressing the Hippo phosphorylation resistant YAP^S127A allele, expression of doxycycline-induced Gαs fails to induce the differentiation phenotype (right panels) observed with wild type YAP1 expression (left panels) (scale bar, 100μm).

As the most profound impact of GNAS^R201C-dependent YAP1 inactivation appears to be on the differentiation status of tumors, we assessed whether sustained YAP1 activity would “rescue” this phenotype. In paired allograft experiments, where we stably expressed either wild type YAP1, or the phosphorylation resistant YAP1^S127A mutant allele in LGKC-1 cells, the differentiation phenotype observed with doxycycline induction was completely abrogated when YAP1^S127A was expressed in LGKC-1 allografts (Figure 5B). To further explore the mechanistic basis of this striking epithelial differentiation phenotype, we examined the dynamics of α-E-catenin localization in LGKC cells, derivative allografts and in the autochthonous models, with and without induction of GNAS^R201C expression. In the epidermis, membranous localization of α-E-catenin has been implicated as a mediator of epidermal differentiation, wherein it also plays a role as a tumor suppressor through sequestering YAP1 in the cytoplasm ²³. Under sub-confluent conditions, addition of doxycycline to LGKC-1 cells was associated with strong membranous expression of α-E-catenin that coincided with cytoplasmic YAP1 sequestration (Figure 6A). Moreover, in subcutaneous allografts generated from LGKC-1 cells, the previously described epithelial differentiation phenotype was also similarly associated with membranous α-E-catenin localization and nuclear YAP1 exclusion (Supplementary Figure 6A); this differential compartmentalization was also phenocopied in the autochthonous model. Specifically, while mPanIN lesions in Kras;Gnas mice with or without doxycycline induction expressed strong α-E-catenin membrane localization, this pattern was lost in the setting of mutant KRAS^G12D induced invasive cancers, while retained in the differentiated PDAC that arose in the setting of concomitant GNAS^R201C expression (Figure 6B).

Figure 6. Cytoplasmic sequestration of YAP1 induced by mutant GNAS enhances membranous expression of α-E-catenin.

(A) Representative photomicrographs of immunofluorescence assay for YAP1 and α-E-catenin localization in LGKC-1 cells treated with (bottom panels) or without doxycycline (top panels) for 24 h. Green = YAP1; Red = α-E-catenin; DAPI = nucleus, blue (scale bar, 50μm)

(B) Representative photomicrographs of immunofluorescence assay for YAP1 and α-E-catenin localization in precursor lesions and invasive cancers arising in autochthonous Kras;Gnas mice with (bottom two panels) or without (top two panels) doxycycline treatment. While membranous α-E-catenin expression is retained in pancreatic precursor lesions irrespective of Gαs expression (i.e. irrespective of doxycycline treatment), only the differentiated invasive cancers arising in the doxycycline-fed, mutant GNAS expressing Kras;Gnas mice show retained membranous α-E-catenin localization. In contrast, minimal membranous α-E-catenin localization is seen in the invasive cancers arising in the absence of mutant GNAS expression (scale bar, 50μm).

YAP1 localization in human IPMNs in relationship to mutant GNAS status

In light of the impact of GNAS mutation on YAP1 localization in murine tissues, we performed immunohistochemical staining for YAP1 in human IPMNs with known GNAS mutations (either GNAS^R201C or GNAS^R201H). Seven of 8 (88%) GNAS wild type IPMNs exhibited strong nuclear staining for YAP1 (Supplementary Figure 6B, left); in contrast, 6 of 14 (44%) IPMNs with mutant GNAS had minimal YAP1 expression in IPMN lesional tissue, despite robust internal controls (ducts and centroacinar cells that expressed strong nuclear YAP1 on the same section) (data not shown). Of the remaining 8 of 14 GNAS mutant IPMNs, two (14%) had expression restricted to the cytoplasm only (see representative image in Supplementary Figure 6B, right), three (21%) had a higher score of YAP1 expression in the cytoplasm compared to the nucleus (consistent with cytoplasmic sequestration), and three (21%) had similar levels of expression in both compartments (Supplementary Table 3). While we did not observe an obvious effect of GNAS mutations on the grade of IPMN differentiation, it is noted that the 22 lesions we queried with known GNAS status were non-invasive IPMNs, and there were no PDAC arising from these lesions.

DISCUSSION

Our study demonstrated that targeted expression of a mutant GNAS^R201C allele in the adult mouse pancreas, concurrently with an activating KRAS^G12D mutation, induces the development of pancreatic cystic neoplasms, including their eventual multistep progression to PDAC, which mirrors the cognate IPMN-PDAC progression model in humans. In addition, we demonstrate that canonical Gαs-cAMP-PKA signaling activates the inhibitory Hippo kinase cascade and sequesters the transcriptional co-activator protein YAP1 in the cytoplasm, which has a profound impact on the differentiation status of invasive carcinomas that arise in the context of mutant GNAS^R201C. Although a previous study has reported a mouse model displaying IPMN-like pancreatic cystic lesions by the co-expression of mutant Kras^G12D and Gnas^R201C alleles during embryonic development, this model did not recapitulate the entire spectrum of IPMN progression because of the early death of mice (median, 5 weeks) ²⁴. Thus, our genetically engineered mouse (GEM) model of IPMNs provides the first bona fide cross-species platform where we can potentially generate longitudinal data points for biomarker and imaging studies, with the intent of translational research in early diagnosis and interception of PDAC

Pancreas-specific expression of mutant TP53, or inactivation of SMAD4 or CDKN2A, demonstrate robust oncogenic cooperation with mutant Kras^G12D, leading to accelerated formation of mPanINs, and invasive carcinoma with higher frequency (ranging from 38 to 96%) and shorter latency (median 14 to 25 weeks) ^25-27, respectively, in comparison to mice with pancreas-specific Kras^G12D mutation alone ²⁸. On the contrary, in the Kras;Gnas mice, we did not observe the stated acceleration, with a median time to invasive carcinoma of 46 weeks (38 weeks post doxycycline) comparable to control mice, suggesting that GNAS^R201C is not functioning as a traditional cooperating oncogene, but rather, as an “onco-modulator” that alters the phenotype of pre-invasive lesions arising in the context of mutant KRAS^G12D and leads to invasive carcinomas with a predominantly well differentiated morphology. Expression profiling studies identified repression of the transcriptional co-activator YAP1 as a signature of GNAS^R201C co-expression on the backdrop of mutant KRAS^G12D, through activation of the upstream inhibitory Hippo kinase cascade. Many of the downregulated transcripts (mCtgf, mCyr6, mMyc, mThbs1, mItgb5) have been previously implicated not only in cell proliferation, but also anchorage independent growth. Thus, the impact on colony formation is likely to be a combination of more than single deregulated transcript but rather a combination of multiple influences due to the profound impact of YAP1 on the transcriptional program on neoplastic cells. Notably, YAP1 is inactivated through phosphorylation-dependent cytoplasmic sequestration, which we observed not only in murine autochthonous or allograft models upon induction of Gαs, but also in patient-derived IPMNs specimens with known GNAS mutations. We were able to demonstrate the primacy of YAP1 sequestration in the Gαs-induced differentiation phenomenon, and the role of Hippo-induced phosphorylation as the mediator of this effect, through rescue experiments performed with a phosphorylation resistant mutant allele that completely reversed the differentiation effect in vivo.

YAP1 has been implicated as a critical transcriptional co-activator in both development and neoplasia, including that of the pancreas. We, and others, have previously demonstrated the transforming capabilities of YAP1 as an oncogene in the pancreatic epithelium ^{29, 30}; specifically, YAP1 has been demonstrated to exhibit both cancer cell intrinsic growth properties (through induction of JAK-STAT3 and MYC signaling, amongst others) ^{31, 32}, as well as a profound impact on the tumor microenvironment by creating a permissive immune milieu ³³. Conversely, genetic ablation of YAP1 in the murine pancreas on a mutant Kras or Kras; p53 background blocks progression to PDAC, although the earliest precursor lesions (acinar ductal metaplasia) are not impeded ³⁴. Notably, despite the abrogation of YAP1 signaling in our model, we do encounter eventual progression in a subset of mice to invasive cancers, suggesting that constitutive cAMP activation likely has additional downstream effects that sustain the neoplastic phenotype and bypass YAP1 inhibition. We are currently in the process of analyzing the secretomes of Kras;Gnas versus Kras isogenic cell lines (LGKC-1 and LGKC-2) to identify potential pro-metastatic factors that might be enhanced or attenuated in the setting of mutant GNAS expression. Another key difference from the prior study pertains to embryonic ablation of YAP1 ³⁴, which likely has more profound impact on the pool of pancreatic progenitors amenable to neoplastic transformation ³⁵, than Gαs-induced inhibition of YAP1 in the adult pancreas. Irrespective, we believe that the inhibition of YAP1 signaling plays a profound role in attenuating the unchecked progression of IPMNs to PDAC, and could explain the favorable biology of these cystic neoplasms. This recalcitrance to invasive neoplasia might also explain why GNAS mutations are uncommon in PanIN lesions, which are the more common precursor to “conventional” adenocarcinoma and harbor a distinct natural history compared to IPMN-induced cancers.

While phosphorylation of YAP1 through the Hippo kinase cascade results in cytoplasmic sequestration, there is a strong selection pressure during carcinogenesis to unleash this important oncogenic signal through phosphatase-dependent de-phosphorylation and nuclear translocation. Prior studies in epidermal keratinocytes and breast epithelium have shown the importance of the cell-cell adhesion protein α-E-catenin in “fencing in” phosphorylated YAP1 ^{23, 36-38}, and shielding the phosphorylated protein from phosphatases like PP2A. Further, downregulation of α-E-catenin in cutaneous squamous cell carcinoma models was shown to result in poorly differentiated cancers with nuclear YAP1 translocation ^{23, 36}. In Kras;Gnas mice, we find that Gαs-induced YAP1 sequestration in the cytoplasm is accompanied by robust expression of membranous α-E-catenin, likely functioning as a cytoplasmic “restraint”, while nuclear translocation of YAP1 (seen with mutant KRAS^G12D expression alone) is accompanied by loss of membrane expression of α-E-catenin. While some studies have suggested that α-E-catenin itself regulates the activity of YAP1 through a non-canonical Hippo-independent pathway ³⁹, we do not find definitive evidence to this effect in our model, since Gαs induction phosphorylates the canonical Hippo kinase LATS1, and at least in the precursor lesions arising in the setting of KRAS^G12D alone, we observe nuclear YAP1 expression despite robust membranous α-E-catenin localization (the latter, however, is lost in the invasive carcinomas).

Several studies with well-characterized human IPMN cohorts demonstrated the pattern of GNAS mutations in the different histologic subtypes of IPMNs. GNAS mutations typically occur in IPMNs, mostly of intestinal subtype, ranging from 83 to 100%. GNAS mutations in pancreatobiliary type of human IPMNs were less frequent in comparison to either intestinal or gastric subtypes, ranging from 33 to 71% ^{5, 6, 40, 41}. Despite of the higher frequency of GNAS R201 mutations in intestinal type of human IPMNs, Kras;Gnas mice with GNAS^R201C expression displayed pancreatobiliary type of mIPMN based on the apomucin expression pattern, as well as other IPMN mouse models ^{24, 26}. Considering that GNAS R201 mutations increase intracelluer cAMP level and forskolin, a cAMP elevator, positively regulates Mucin 2 gene expression in human colon cancer cell ⁴², GNAS R201 mutations might be direct regulator of MUC2 and associated with the intestinal subtype in human IPMN. The human-mouse discrepancy in terms of apomucin expression indicates the presence of Muc2 gene silencing mechanisms in murine IPMN models, such as methylation of the MUC2 gene promoter observed in human pancreatic cancer cells ⁴³.

The study findings regarding YAP1 inactivation induced by mutant GNAS in the presence of mutant KRAS indicate the possible utility of GNAS status to stratify PDAC patients for targeted therapy or immunotherapy. In KRAS mutant pancreatic cancer cell with high YAP1 expression, YAP1 inhibition significantly enhanced the antitumor efficacy of a pan-RAF inhibitor ⁴⁴, which suggests KRAS;GNAS mutant PDAC is more sensitive to therapies targeting of the major downstream pathway effector of KRAS. In addition, YAP inhibition blocks multiple secreted factors including CTGF and CYR61 and decreases the formation of desmoplasia, recruitment of immunosuppressive myeloid-derived suppressor cell and tumor-associated macrophage ³⁵. Thus, better response to immunotherapy, such as T- cell check point inhibitors, in GNAS mutated PDAC patients might be expected. We believe the generated GEM model is ideal for testing of these hypotheses.

In summary, we describe a mouse model of IPMNs that recapitulates the genetic landscape of the most common drivers and the multistep progression observed in the cognate human cystic neoplasms. We identify the potential mechanistic basis for the relatively indolent biology of the cystic pathway to invasive neoplasia in the pancreas, and the role of constitutively active Gαs as an onco-modulator of epithelial differentiation in the context of mutant KRAS. The Kras;Gnas model (and derivative reagents, such as isogenic cell lines) will be useful reagents for the community to pursue cross-species translational research endeavors in the context of early detection and cancer interception.