Kras^G12D induces EGFR-MYC cross signaling in murine primary pancreatic ductal epithelial cells

Abstract

Epidermal growth factor receptor (EGFR) signaling has a critical role in oncogenic Kras-driven pancreatic carcinogenesis. However, the downstream targets of this signaling network are largely unknown. We developed a novel model system utilizing murine primary pancreatic ductal epithelial cells (PDECs), genetically engineered to allow time-specific expression of oncogenic Kras^G12D from the endogenous promoter. We show that primary PDECs are susceptible to Kras^G12D-driven transformation and form pancreatic ductal adenocarcinomas (PDAC) in vivo after Cdkn2a inactivation. In addition, we demonstrate that activation of Kras^G12D induces an EGFR signaling loop to drive proliferation. Interestingly, pharmacological inhibition of EGFR fails to decrease Kras^G12D-activated ERK or PI3K signaling. Instead our data provide novel evidence that EGFR signaling is needed to activate the oncogenic and pro-proliferative transcription factor c-MYC. EGFR and c-MYC have been shown to be essential for pancreatic carcinogenesis. Importantly, our data link both pathways and thereby, explain the crucial role of EGFR for Kras^G12D-driven carcinogenesis in the pancreas.

Introduction

Although novel chemotherapeutic regimens increased the overall survival of patients with advanced pancreatic ductal adenocarcinoma (PDAC), its prognosis remains dismal. The epidermal growth factor receptor (EGFR) signaling pathway plays an outstanding role in the carcinogenesis of the disease ^{1, 2}. Despite being active in only a subgroup of patients, the EGFR inhibitor erlotinib is currently the only known targeted therapeutic for PDAC. EGFR belongs to a receptor tyrosine kinase (RTK) family including ErbB2, ErbB3, and ErbB4³. Seven ligands, epidermal growth factor (EGF), transforming growth factor α (TGFα), betacellulin (BTC), heparin-binding EGF-like growth factor (HB-EGF), amphiregulin (ARG), epiregulin (EPR), and epigen (EGN) can induce receptor dimerization and consecutive activation ³. Effectors acting downstream of EGFR in the Kras^G12D-induced circuit that drive tumor development in the pancreas remain incompletely understood.

In this study we generated a novel mouse model of PDAC, in which expression of the Kras^G12D allele can be induced by a tamoxifen-inducible Cre recombinase in primary pancreatic ductal epithelial cells (PDECs). We provide evidence that Kras^G12D-driven proliferation of PDECs depends on an EGFR signaling loop engaging the oncogenic transcription factor c-MYC (MYC afterwards).

Results and Discussion

Mutations of the Kras oncogene are one of the earliest genetic events and have been shown to drive carcinogenesis in the pancreas ⁴. To activate the expression of one allele of oncogenic Kras^G12D from the endogenous gene promoter in PDECs, we isolated PDECs from R26^CreERT2;LSL-Kras^G12D/+ mice (Fig. 1A). These cells show presence of ductal markers and the absence of acinar or endocrine markers (Fig. S1A). PDECs express genes associated with a progenitor state (Fig. S1A). Activation of the Cre recombinase in these cells by 4-hydroxytamoxifen (4-OHT) induced efficient recombination of the Kras locus (Fig. 1B) and more than 90% of the PDECs are recombined after 8 days of 4-OHT treatment (Fig. S1B–D). Expression of oncogenic Kras^G12D induced GTP-bound Ras to an extent observed in murine Kras^G12D-driven PDAC cell lines (Fig. 1C). In addition, ERK becomes phosphorylated indicating activated canonical Kras signaling (Fig. 1D and 1E).

Figure 1. Activation of canonical Kras signaling in PDECs.

A) Genetic strategy to activate Kras^G12D-expression in PDECs (R26^CreERT2;LSL-Kras^G12D/+). Isolation of PDECs and mouse lines are described in detail in the supplementary material and methods section. All animal studies were conducted in compliance with European guidelines for the care and use of laboratory animals and were approved by the Institutional Animal Care and Use Committees (IACUC) of Technische Universität München, Regierung von Oberbayern, and the University of Pennsylvania. The R26^CreERT2 mouse line was described in ⁴⁸ and LSL-Kras^G12D line in ⁴⁹. B) Genotyping PCR of the indicated PDECs treated with 4-hydroxy-tamoxifen (4-OHT) (200 nM) (Sigma-Aldrich, München, Germany) over time. WT: wild type allele; LSL: Lox-Stop-Lox allele; STOP del: recombined LSL-allele. Primer sequences are depicted in the supplementary material and methods section. C) Ras pull-down assay (Raf-RBD Protein GST beads (Cytoskeleton, Denver, CO, USA)) from vehicle or 4-OHT (200 nM) treated PDECs. The murine Kras^G12D-driven PDAC cell line PPT-6037 was used as a positive control. Western blot of pan-Ras expression (clone 10, #05-516, Merck-Millipore, Darmstadt, Germany) (β-actin (Sigma-Aldrich): loading control) Irrelevant lanes were excised and the merger originated from the same gel. D) Western blot of phospho-ERK (Thr202/Tyr204) (#4370, Cell Signaling Technology, Danvers, MA, USA) and pan-ERK (#4696, Cell Signaling Technology) from vehicle or 4-OHT (200 nM) treated PDECs over the indicated time points (α-tubulin (Sigma-Aldrich): loading control). E) Quantification of ERK phosphorylation. PDECs from R26^CreERT2;LSL-Kras^G12D/+ mice were treated with 4-OHT (200 nM) over time. pan-ERK and phospho-ERK were determined in western blots and quantified using the Odyssey Infrared Imaging System (Li-Cor Biosciences, Bad Homburg, Germany), assuring measurements in the linear range. Shown is the relative ERK phosphorylation of four independent experiments using four individual PDEC lines.

One road to PDAC originates in the pancreatic acinar cells likely via acinar-to-ductal metaplasia (ADM) and pancreatic intraepithelial neoplasia (PanIN) ⁵. Although the contribution of ductal cells to the carcinogenesis in the pancreas is still a matter of debate ⁶, available data suggest that ductal cells seem relatively refractory to Kras^G12D-driven transformation ⁷. Therefore, we investigated whether PDECs can form PDAC in vivo. We orthotopically transplanted ex vivo tamoxifen-treated PDECs from R26^CreERT2;LSL-Kras^G12D/+ mice into the pancreas of immunodeficient mice. However, none of the transplanted mice (n=3) developed PDAC in the investigated time period of 51 days. Furthermore, we detected no pre-malignant lesions in the pancreas of these mice (Fig. S2A). In contrast, it has been reported that transplantation of PDECs, engineered to express Kras^G12D, into C57Bl/6 mice, leads to the formation of ductal structures resembling early PanIN lesions ⁸. Considering low efficacy of Kras^G12D-dependent tumor initiation, the number of orthotopically transplanted PDEC cells (1×10⁶ versus 0.15×10⁶ cells) might account for this discrepancy. Indeed, after increasing the number of transplanted PDECs to 7.5×10⁵ cells, formation of PanIN-like structures (lineage label [YFP] and keratin 19 [K19] positive) was detected (Fig. S2B). Besides activating mutations in the Kras gene, the tumor suppressor Cdkn2a is frequently lost in pre-neoplastic lesions. To model the human disease, we isolated PDECs from R26^CreERT2;LSL-Kras^G12D/+;Cdkn2a^lox/lox mice (Fig. S2C). Tamoxifen treatment of these cells induced rapid loss of p16^Ink4a expression (Fig. S2D) and canonical Kras^G12D signaling is activated (Fig. S2E). Orthotopic transplantation of Cdkn2a-deficient PDECs resulted in the development of invasive, proliferative, K19 positive, and metastatic PDAC (Fig. S2F–S2H). Thus, our model system suggests that PDECs contain a cellular population that is susceptible to Kras^G12D-induced transformation.

Expression of Kras^G12D in PDECs induces proliferation (Fig. 2A and 2B) accompanied by induction of cell cycle genes, like cyclin D1 or cyclin A (Fig. 2C). This is in agreement with observations that Kras^G12D prevents premature senescence of PDECs ⁹ and induces a proliferative response ^9–12.

Figure 2. Kras^G12D-driven proliferation in PDECs depends on an EGFR-loop.

A) White light microscopic images of PDECs from R26^CreERT2;LSL-Kras^G12D/+ mice treated for five days with 200 nM 4-OHT or left as vehicle treated controls. Scale bars: 100 μm. B) PDECs from R26^CreERT2;LSL-Kras^G12D/+ mice were treated with 200 nM 4-OHT or were left as vehicle treated controls. After 7 days, 50.000 PDECs were seeded and cell number was determined for two additional days (two biological replicates performed as triplicates). C) PDECs from R26^CreERT2;LSL-Kras^G12D/+ mice were treated with 200 nM 4-OHT over time and cyclin D1 ((HD11), sc-246, Santa Cruz Biotechnology, Dallas, Tx, USA) and cyclin A ((H-432), sc-751, Santa Cruz Biotechnology) expression was measured in western blots. Different lysates were blotted to different membranes and loading was controlled by β-actin. D) Enrichment plots of EGFR signatures and corresponding heatmaps (top 20 EGFR controlled genes induced by Kras^G12D) from microarrays of vehicle (control: ctrl.) or 4-OHT (200 nM, 3 days) treated PDECs. NES: normalized enrichment score; FDR: false discovery rate. EMBL-EBI ArrayExpress Accession number: E-MTAB-2592. See supplementary material and methods for a description of the microarray analysis and gene set enrichment analysis. E) Relative amphiregulin (Areg) and epiregulin (Ereg) mRNA expression in 4-OHT (200 nM) treated PDECs was determined by qPCR using cyclophilin A mRNA expression as reference. Primers are depicted in supplementary material and methods. One way ANOVA *p-value < 0.05. F) and G) PDECs were treated for 6 days with 4-OHT (500 nM) and erlotinib or gefitinib (10 μM each; LC Laboratories, Woburn, MA, USA) were added for the last 24 hours as indicated. F) phospho-EGFR (#2234, Cell Signaling Technology) and pan-EGFR ((1005), sc-03, Santa Cruz Biotechnology) western blot (α-tubulin: loading controls). G) cyclin D1 and phospho-Rb (#8516, Cell Signaling Technology) western blot. Different lysates were blotted to different membranes and loading was controlled by β-actin.

In order to identify pathways driving Kras^G12D induced proliferation, we used gene set enrichment analysis of mRNA expression profiles (GSEA). Several of the gene sets significantly enriched in Kras^G12D expressing cells are linked to signatures controlled by the EGFR family (Fig. 2D and Supplemental Table 1). Accordingly, Kras^G12D induced expression of the EGFR ligands amphiregulin and epiregulin (Fig. 2D and 2E), arguing for autocrine stimulation. Consistently, in murine PanIN organoids derived from ductal cells of Pdx1-Cre;LSL-Kras^G12D/+ mice ¹³, Kras^G12D induced expression of EGFR ligands (Fig. S3). Along with upregulation of EGFR ligands, increased receptor auto-phosphorylation was observed (Fig. 2F). To test whether this EGFR phosphorylation is critical for mutant Kras-regulated proliferation we utilized the EGFR inhibitors erlotinib and gefitinib. Of note, in PDAC models, gefitinib has been demonstrated to be more specific for EGFR than erlotinib ¹⁴. Both inhibitors diminished the Kras^G12D-induced EGFR phosphorylation (Fig. 2F) and decreased expression of cell cycle regulators, like cyclin D1 (Fig. 2G). A link between the EGFR loop and cyclin D1 was recently described in Kras^G12D-driven cancer formation in the pancreas in vivo ¹. Importantly, the Kras^G12D-mediated inactivation of the Rb-dependent restriction point in the G1-phase of the cell cycle is controlled by EGFR (Fig. 2G).

EGFR signaling networks engage ERK-, PI3K- and STAT3-controlled pathways. Although expression of oncogenic Kras^G12D induced phosphorylation of ERK, AKT and its substrate GSK3β, both EGFR inhibitors did not distinctly change activation of these pathways (Fig. 3A and 3B). STAT3 phosphorylation was neither induced nor modulated by EGFR inhibitors (Fig. 3C). Involvement of EGFR in Ras induced transformation is context dependent ^{1, 2, 15} and the context appears to direct the signaling hubs engaged. Indeed, the hypomorphic waved-2 (wa2) EGFR receptor variant reduced AKT activation but not ERK phosphorylation in primary keratinocytes of K5-SOS-F mice ¹⁵. In the pancreas, the EGFR-controlled signaling hubs are incompletely understood ¹⁶. Despite expression of the Kras^G12V oncogene, most acinar cells express low EGFR levels and nuclear phospho-AKT or phospho-STAT3 staining is absent ². Inflammatory stimuli increase EGFR expression and induce AKT and STAT3 phosphorylation in acinar cells in the context of oncogenic Kras^{G12V 2}, arguing that AKT and STAT3 are part of the EGFR signaling network in this inflammatory context. Furthermore, EGFR signaling increased the signaling output of the canonical Kras pathway to induce acinar-to-ductal metaplasia (ADM) in Kras^G12D expressing acinar cells ¹. In established Kras-driven PDAC models, the EGFR is inconsistently linked to active ERK, AKT or STAT3 ^{1, 2}. Together, these observations clearly demonstrate that the EGFR signaling network is modulated by cell-autonomous (e.g. tumor suppressor status) and non-autonomous (e.g. inflammatory environment) conditions.

Figure 3. MYC is a downstream effector of EGFR.

PDECs were treated for 6 days with 4-OHT (500 nM) and gefitinib or erlotinib (10 μM each) were added for the last 24 hours as indicated. Western blot of A) phospho- and pan-ERK (α-tubulin: loading controls). B) phospho-AKT (#9271 and #9275, Cell Signaling Technology) and -GSK3β (#9323, Cell Signaling Technology) as well as pan-AKT ((C67E7), #4691, Cell Signaling Technology). Different lysates were blotted to different membranes and loading was controlled by β-actin. C) phospho-STAT3 ((D3A7), #9145, Cell Signaling Technology) and pan-STAT3 ((C-20):sc-482, Santa Cruz Biotechnology) western blot. D) Transcription factor gene signatures (TFT-MSigDB) significantly downregulated in EGFR inhibitor (EGFRi) treated PDECs. PDECs were treated as described in A). E) GSEA enrichment plots of MYC signatures and corresponding heatmaps (top 40 MYC controlled genes inhibited by the EGFR inhibitors) from microarrays of PDECs treated as described in A). NES: normalized enrichment score; FDR: false discovery rate. EMBL-EBI ArrayExpress Accession number: E-MTAB-2592. See supplementary material and methods for a description of the microarray analysis and gene set enrichment analysis. F) PDECs were treated as described in A). Relative Ccna2, E2F1, eIf4e, Hspe1, Skp2, Ncl, and Odc1 mRNA expression levels were determined by qPCR using beta-actin mRNA expression as reference. Primers are depicted in supplementary material and methods. One way ANOVA *p-value < 0.05. G) PDECs from 2 months old Pdx1-Flp;FSF-Kras^G12D/+;FSF-R26^CAG-CreERT2;R26^mT/mG;MYC^lox/lox mice were isolated. In these cells expression of Kras^G12D is induced in vivo and expression of floxed genes can be manipulated by the treatment of cells with 4-OHT. The cells were treated with 4-OHT (500 nM) for 24 hours. The green fluorescent protein (GFP)-expressing cells were FACS (fluorescence-activated cell sorting) sorted as recently described ⁵⁰. Relative Myc, Ccna2, E2f1, Hspe1, Ncl, and Odc1 mRNA expression levels were determined by qPCR using beta-actin mRNA expression as reference and compared to untreated cells, in which expression was set to 1. The Pdx1-Flp, FSF-Kras^G12D, and the FSF-R26^CAG-Cre-ERT2 mouse lines were described recently in ²⁵. The R26^mT/mG mouse line is described in ⁵¹ and the MYC^lox line in ⁵².

The observation that ERK and AKT remain phosphorylated, despite an inactivation of EGFR auto-phosphorylation and cell cycle progression, suggests that different hubs sense the signal. To identify EGFR-engaged pathways, we again performed transcriptome profiling of EGFR inhibitor-treated PDECs. Enriched gene sets in Kras^G12D expressing PDECs with an active EGFR loop were linked to the cell cycle, DNA replication and repair as well as anabolic pathways (Supplemental Table 2). Since we intended to identify an integration of the EGFR loop with the cell cycle machinery, we focused on transcription factors (TFT-MSigDB). We detected that the majority of EGFR-controlled gene sets were linked to the pro-proliferative E2F transcription factor family, corroborating the link of the EGFR loop to the cell cycle (Fig. 3D and Supplemental Table 3). Additionally, we observed six signatures associated with transcription factors of the MYC family (Fig. 3D, 3E, and Supplemental Table 3). Performing a GSEA with curated gene sets (C2) of the MSigDB revealed 18 significant MYC gene sets linked to EGFR (Supplemental Table 4). MYC adopts a prominent role in Ras-driven cancers ^17–21. Since MYC is strongly linked to the cell cycle of exocrine progenitors in the pancreas ^{22, 23} and to the E2F pathway in cell-based PDAC models ²⁴, we investigated the role of MYC. First, we quantified mRNA levels of MYC target genes upon treatment with EGFR inhibitors. Both inhibitors reduced the expression of the MYC target genes Ccna2, E2f1, eIf4e, Hspe1, Skp2, Ncl, and Odc1 (Fig. 3F), indicating robust cross signaling between EGFR and MYC. Second, to demonstrate the regulation of these genes by MYC in the context of murine PDECs, we used a novel dual-recombinase system, allowing the time-specific manipulation of genes ²⁵. We isolated murine PDECs at PanIN stages (2 months old mice) from Pdx1-Flp;FSF-Kras^G12D/+;FSF-R26^CAG-CreERT2;R26^mT/mG;Myc^lox/lox mice. We treated these PDECs for 24 hours with tamoxifen and sorted GFP expressing cells by FACS to investigate expression of MYC and its target genes. Here, Myc mRNA expression is reduced to 20% compared to untreated controls, accompanied by a reduced mRNA expression of all investigated target genes (Fig. 3G).

Oncogenic activity of MYC is regulated by phosphorylation of threonine 58 and serine 62 residues at the N-terminal MYC homology box I ^{26, 27}. Indeed, mutant Kras^G12D activation induces N-terminal phosphorylation of MYC (Fig. 4A). Increased MYC phosphorylation was detected over time (Fig. 4A and 4B). Furthermore, Kras^G12D induces MYC protein expression (Fig. 4B), which is accompanied by a slight induction of Myc mRNA (Fig. 4C). Both EGFR inhibitors prevent Kras^G12D-induced MYC protein expression (Fig. 4B) and significantly reduce Myc mRNA expression (Fig 4C). MYC regulation is controlled at multiple levels in PDAC ²¹. How the EGFR loop is connected to MYC expression is currently unknown and awaits further investigations. Nevertheless, our results show that Kras^G12D induces MYC expression, phosphorylation, and MYC target gene expression in an EGFR-dependent manner, suggesting that the Kras^G12D-induced EGFR network is engaging MYC as an important effector.

Figure 4. MYC expression is regulated by the autocrine EGFR-loop.

A) Indicated PDECs were treated with 4-OHT (500 nM) over time. Western blot for phospho-MYC (#9401, Cell Signaling Technology) (β-actin: loading control). B) PDECs were treated with 4-OHT (500 nM) as indicated. Gefitinib or erlotinib (10 μM each) were added for the last 24 hours of incubation. Westernblot for phospho-MYC and MYC (#9402, Cell Signaling Technology). Different lysates were blotted to different membranes and loading was controlled by β-actin or α-tubulin as indicated. To detect phosphorylated MYC, PDECs were lysed by directly boiling in protein loading buffer. C) Indicated PDECs were treated with 4-OHT (500 nM) for 6 days. Gefitinib or erlotinib (10 μM each) were added for the last 24 hours of incubation. Relative Myc mRNA expression levels were determined by qPCR using beta-actin mRNA expression as reference. One way ANOVA *p-value < 0.05. D) Genetic strategy to activate Kras^G12D-expression in Hnf1β-positive PDECs (Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom). The Hnf1b-Cre^ER mouse line was described in ³⁰ and the R26^Tom reporter mouse line in ⁵³. E) PDECs from Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom mice were treated with 4-OHT (1 μM) for 15 days. Afterwards gefitinib or erlotinib (10 μM each) were added for additional 24 hours or the cells were left as vehicle treated controls. Western blot of phospho-EGFR, pan-EGFR, phospho-ERK and pan-ERK, MYC, and phospho-Rb. Different lysates were blotted to different membranes and loading was controlled by β-actin or α-tubulin as indicated. F) PDECs from Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom mice were treated with 4-OHT (1 μM) for 15 days. Afterwards gefitinib or erlotinib (10 μM each) were added for additional 24 hours. Relative Myc mRNA expression levels were determined by qPCR using beta-actin mRNA expression as reference. G) PDECs from Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom mice were treated with 4-OHT (1 μM) for 15 days. Afterwards, 2.000 cells per well were seeded in a 96 well plate in quadruplicates (n=4). After 24 hours the cells were treated with erlotinib (10 μM), gefitinib (10 μM), or 10058-F4 (80 μM) or were left as vehicle treated controls. To determine relative growth, fluorescence (excitation: 560 nm, emission 590 nm) was measured daily over 8 days with a BMG FLUOstar OPTIMA Microplate Reader (BMG Labtech, Ortenberg, Germany). One way ANOVA *p-value < 0.05.

Acinar cells cultured in suspension dedifferentiate and activate a ductal gene expression program ^{5, 28}. Therefore, the possibility exists that our PDEC preparations contain dedifferentiated acinar cells. However, in our hands, we were never able to serial passage and subculture acinar cells. This is in agreement with observations that ex vivo acinar cells lack the capacity to proliferate due to a p53-dependent growth arrest ²⁹. Since we subculture and serial passage the PDEC lines (only passage 3 to 8 were used for all experiments), a contamination with acinar cells is unlikely. To further address this point, lineage tracing technology using the ductal marker Hnf1b-Cre^ER mouse line was used ³⁰ (Fig. 4D). First, we activated Cre recombinase in a Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom mouse by the i.p. application of tamoxifen in vivo ³¹. Two weeks after the last tamoxifen administration we isolated PDECs. 87% of these cells express the reporter gene tdTomato (Fig. S4A), arguing for the ductal origin of the prepared cells. To further corroborate the EGFR-MYC loop in the Hnf1β-lineage, we isolated PDEC lines from untreated Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom mice. We adapted the tamoxifen treatment regime since 0.5 μM 4-OHT insufficiently recombined the Kras locus in Hnf1b-Cre^ER PDEC lines (Fig. S4B). Even upon an increased 4-OHT dose, only 54% of the cells expressed tdTomato after 7 days (Fig. S4C). However, the fraction of tdTomato expressing cells was increased to 91% after 15 days (Fig. S4D). Then the Kras locus is recombined (Fig. S4E) and the canonical Ras-pathway is activated (Fig. 4E). Additionally to ERK phosphorylation, the EGFR becomes phosphorylated, MYC protein and mRNA expression is induced and the G1-phase restriction point becomes inactivated upon the expression of Kras^G12D (Fig. 4E and Fig. 4F). Both EGFR inhibitors prevent EGFR and Rb phosphorylation as well as MYC expression whereas ERK phosphorylation is not influenced (Fig. 4E and Fig. 4F). Together, the lack of proliferation of acinar cells ex vivo and our lineage tracing experiments argue that the described pathway and biology acts in the ductal lineage.

To further demonstrate the impact of the EGFR-dependent loop towards the proliferative capacity of Kras^G12D expressing Hnf1b-Cre^ER;LSL-Kras^G12D/+;R26^Tom PDECs, we measured growth of erlotinib- and gefitinib-treated cells. Both EGFR inhibitors reduce proliferation with a similar potency (Fig. 4G). To compare the effects of the EGFR blockade with the effects of a direct MYC inhibitor, we treated the cells with 10058-F4. This MYC inhibitor prevents the dimerization of MYC with MAX ³². Similar to EGFR inhibitors, 10058-F4 reduces proliferation of Kras^G12D expressing PDECs, supporting an important function of MYC downstream of the EGFR.

Overall, PDECs are a valid model to gain mechanistic insights into Kras-driven processes in a specific pancreatic context ^{6, 8–12, 33}. In addition, PDECs are a tool for genetic screening experiments ³⁴. A multipotent subpopulation of adult pancreatic ductal cells capable of reprogramming towards the endocrine lineage was recently described, arguing for a stem cell population in the ductal compartment ³⁵. Consistently, our observations demonstrate that PDECs express progenitor markers and are susceptible to Kras^G12D-dependent transformation. In agreement, a non-islet Pdx1-positive PDEC subpopulation in the adult pancreas with a stem-like phenotype was described, harboring tumorigenic and metastatic capacity upon the expression of Kras^{G12D 36, 37}. In line with our data, MYC activation by Kras^G12D was observed in this model and MYC was linked to the evolution of pancreatic cancer cells with stem cell-like features and metastatic potential ^{36, 37}. In addition, we analyzed recently published microarray datasets generated from duct and duct-like cells of Pdx1-Cre;LSL-Kras^G12D/+ mice ³⁸ at pre-malignant disease stages. Indeed, we observed Kras^G12D-induced EGFR- and MYC-signatures using gene set enrichment analysis (Fig. S5), suggesting that the molecular changes occurring in vivo are recapitulated by our in vitro model. Furthermore, our data show that an EGFR-loop contributes to Kras^G12D-driven proliferation of PDECs, well in line with recent data from Kras-dependent mouse models ^{1, 2}.

In addition to this EGFR loop, a Kras^G12D-activated autocrine loop engaging the insulin like growth factor 1 receptor (IGF1R) in Cdnk2a/Trp53-double deficient PDECs has recently been described ¹¹. Whether EGFR- and IGF1R-dependent loops act in parallel to modulate signaling thresholds and whether the usage of such circuits is determined by tumor suppressive programs, awaits further analysis. However, in vivo findings indicate that the need of EGFR signaling to develop Kras^G12D-driven PDACs is bypassed in a p53-deficient background ^{1, 2}, arguing that tumor suppressors determine the need of such loops.

In contrast to the requirement of EGFR for oncogenic Kras-induced pancreatic carcinogenesis, molecular hubs of the EGFR network are incompletely defined. Although the cooperation of Ras and MYC oncogenes to transform cells has been described in the last century ³⁹ and many underlying molecular processes are known ⁴⁰, our data link for the first time two essential components of Kras-driven transformation in a pancreatic pre-neoplasia equivalent model. Like the EGFR loop, MYC essentially contributes to the carcinogenesis in the pancreas ^{19–21, 41, 42}. Especially, in embryonic stem cell-based genetically engineered mouse models the distinct effect of MYC silencing on disease progression and tumor formation was recently demonstrated ¹⁹. MYC is highly expressed in multipotent pancreatic progenitor cells ⁴³ and can autonomously drive tumor initiation and progression in the pancreas ^{21, 44–46}. The EGFR ligand TGFα dramatically accelerates MYC-driven carcinogenesis in the pancreas in vivo ⁴⁷, which is consistent with our observation of an EGFR-MYC cross signaling. Therefore, connecting EGFR to MYC underscores the importance of the EGFR network in the pancreatic carcinogenesis.