Mouse iKras 4292 cell lines were obtained from Dr. Marina Pasca di Magliano [University of Michigan](Collins et al., 2012). Cells were maintained in RPMI 1640 media supplemented with 10% (v/v) FBS and 10 μg/ml doxycycline hyclate (doxy, Sigma Aldrich, Cat# D9891) for regular culture. For Kras^G12D induction experiments, cells were deprived of doxy at least 48 h for beginning experiments (0 h, baseline). Then, fresh media containing doxy 10 μg/ml was added, and cells maintained under culture during multiple time intervals (12, 24 and 48 h). Human hTERT-E6/E7-HPNE (ATCC Cat# CRL-4037, RRID: CVCL_C468) and hTERT-E6/E7-HPNE-KRAS^G12D (ATCC Cat# CRL-4039, RRID: CVCL_C470) pancreatic cell lines were obtained from ATCC and maintained in appropriate media according to recommendations. All cells were cultured at 37°C in a humidified incubator with 5% CO2 and were used to a maximum of 30 passages. All cell lines tested negative for Mycoplasma with last testing performed in July 2020.

Mouse pancreas (∼100 mg) was cut in small pieces and homogenized in protein extraction buffer [10 mM Tris–HCl, 1 mM EDTA, 1% (v/v) Triton-X100, 1% (w/v) Sodium Deoxycholate, 0.1% (w/v) SDS, 140 mM NaCl and 1 mM PMSF] with freshly added protease and phosphatase inhibitors (Thermo Fisher Scientific). Homogenized tissue was moved to a 2 mL microcentrifuge tube containing 1 stainless steel bead (7 mm diameter) and further dissociated using TissueLyser LT (2 min at 50 Hz). The lysate was then sonicated twice (amplitude 5 for 10 s) and supernatant was cleared by centrifugation (10,000 rpm, 10 min, 4°C). HPNE and iKras proteins were collected from cells by lysis in protein extraction buffer supplemented with phosphatase and protease inhibitors. Equal amounts of protein from pancreas or cell lysates were electrophoresed on 12% SDS-PAGE gels and transferred to nitrocellulose membranes (GE Healthcare). Membranes were blocked in either 5% (w/v) milk or 3% (w/v) BSA for 1 h, and primary antibody incubations were performed overnight at 4°C with rocking. Supplementary Table 1 contains primary antibodies used. Anti-mouse or anti-rabbit HRP-conjugated secondary antibodies (1:5000, Millipore) were incubated on the membranes for 1 h at room temperature with agitation, followed by detection with chemiluminescence (Thermo Fisher Scientific). Quantification of bands in n = 3 experiments was completed using ImageJ and statistical significance determined by Student’s t-tests in GraphPad Prism 7.

iKras 4292 cells were stimulated with doxy at the indicated intervals, fixed using 4% Formaldehyde and washed three times with 1× PBS. Cells were permeabilized using 0.2% Triton X-100 for 10 min, PBS washed and blocked with 1% (w/v) BSA in PBS-Tween 20 solution. Slides were then incubated with EHMT2, EHMT1 and WIZ primary antibodies at 4°C overnight and labeled using Alexa Fluor-conjugated secondary antibodies. For proximity ligation assays (PLA) assays, the Duolink PLA Starter Kit (Sigma Aldrich, Cat# DUO92102) was used according to manufacturer’s instructions. Coverslips were mounted using Prolong Gold antifade mounting media (with DAPI, Life technologies, Cat# P36930). Antibodies used are listed in Supplementary Table 1. Images were acquired using a Zeiss LSM510 confocal microscope with a 63× magnification objective.

Immunoprecipitation (IP) of EHMT2 was performed by conjugating the EHMT2 antibody (Thermo Fisher Scientific Cat# PA5-34971, RRID: AB_2552320) or control IgG antibody to Protein A/G Magnetic Beads (Thermo Fisher Scientific) through disuccinimidyl suberate crosslinking. iKras 4292 cells were grown and then stimulated with doxy for 24 h. Cells were lysed with IP Lysis/Wash Buffer (Thermo Fisher Scientific-Pierce), and lysates were incubated overnight with the antibody conjugates at 4°C. Immunoprecipitated complexes were washed, eluted, and run on a 4–15% Criterion^TM Tris–HCl Protein Gel (Bio-Rad). The gel was subsequently visualized with BioSafe^TM Coomassie Stain (Bio-Rad). Each gel lane was de-stained, dehydrated, dried, and subjected to trypsin digestion. Subsequently, liquid chromatography (LC)-ESI-MS/MS analysis was performed on a Thermo Scientific LTQ Orbitrap mass spectrometer at the Mayo Clinic Proteomics Core.

Animals were housed in standard mouse housing with controlled temperature, humidity and light cycles and provided with ad libitum standard rodent chow and water. Mice were euthanized using CO₂ following institutional guidelines. Tissues were collected and preserved in formaldehyde for 24 h and then moved to 70% (v/v) ethanol for histological analysis or freshly processed for cellular studies. B6.129S4-Kras^{TM 4Tyj}/J (LSL-Kras^G12D, IMSR Cat# JAX:008179, RRID: IMSR_JAX:008179) (Jackson et al., 2001), B6.FVB-Tg(Pdx1-Cre)6Tuv/J (Pdx1-Cre, IMSR Cat# JAX:014647, RRID: IMSR_JAX:014647) (Hingorani et al., 2003), Ptf1a^{TM 1(cre)Hnak}/RschJ (P48^Cre/+, IMSR Cat# JAX:023329, RRID: IMSR_JAX:023329) (Nakhai et al., 2007), and B6.Cg-Tg(CAG-cre/Esr1^∗)5Amc/J (CAGGCre-ER^TM, IMSR Cat# JAX:004682, RRID: IMSR_JAX:004682) (Hayashi and McMahon, 2002) were purchased from Jackson Laboratories. EHMT2 flox/flox (EHMT2^fl/fl) animals were a generous gift received from Dr. Oltz (Tachibana et al., 2007). All animals have been described previously and were maintained on a C57Bl/6 background. EHMT2^fl/fl mice were crossed with animals containing LSL-Kras^G12D, Pdx1-Cre, P48^Cre/+ or tamoxifen-inducible (CAGGCre-ER^TM) Cre drivers. Animals were weaned after 3 weeks and 0.4–0.8 cm length of tail was taken for genotyping. DNA was extracted from the tail pieces using Qiagen’s DNeasy^® Blood and Tissue kit (Cat# 69506). PCR was performed as previously described (Tachibana et al., 2007) with products run out an 2.5% agarose gel. Animals that were used in aging studies with no additional treatments were sacrificed at 4, 6, or 8 weeks.

To induce chronic pancreatitis, caerulein was administered via IP injection to 4-week-old mice once a day, five times a week, for a total of 4 weeks at a dose of 50 μg/kg. Animals were given 7 days with no treatment before sacrifice to allow recovery. Animal and pancreas weights were recorded at the time of sacrifice and tissue was taken for RNA, protein and histological analysis.

Immunohistochemistry was performed on mouse pancreas tissue as described previously (Mathison et al., 2013). Primary antibodies (Supplementary Table 1) were incubated overnight at 4°C. Slides were developed with Nova Red (Vector Laboratories, Cat# SK-4800) and counterstained with Mayer hematoxylin. Five random fields (10× objective) per slide, containing at least 1,000 cells per field, were imaged, and counted.

CAGGCre-ER^TM and CAGGCre-ER^TM;EHMT2^fl/fl mice were euthanized and pancreas immediately injected with 2 mL of chilled collagenase P (1.33 mg/ml), dissected and cut into small pieces for digestion (20 min at 37°C with gentle shaking). Acinar cells were collected by centrifugation and resuspended in 3D culture base medium (RPMI 1640 medium supplemented with 1% FBS, 0.1 mg/ml soybean trypsin inhibitor, sodium pyruvate 1× and antibiotics). 96-well plates were pre-coated with Matrigel (Corning, Cat# 356231, 30 μL) for 1 h at 37°C. Cell suspensions were mixed 1:1 with Matrigel and plated (100 μL/well). The cell/Matrigel mix was allowed to solidify for 1 h at 37°C before addition of 175 μL 3D culture media. For the first 48 h, DMSO or 4-OHT (10 mmol/L) was added, followed by EGF (20 ng/mL) or TGFα (50 ng/mL) for an additional 5–6 days (Martinelli et al., 2016). Pictures from ductal structures were taken after 8 days in culture using a Canon EOS Rebel Xsi camera mounted on a Nikon Eclipse TS100 microscope at 10× magnification. Duct size was measured using ImageJ software, and plotted values represent fold-change of treated groups over respective controls.

For RNA extraction from tissue, a section of pancreas was taken at time of euthanasia, minced in RNAlater and snap frozen with liquid nitrogen. RNA extraction was completed with ToTALLY RNA Kit (Ambion, Cat# AM1910) according to manufacturer’s protocols. Briefly, frozen samples were lysed with tissue Denaturation Solution and disrupted in the TissueLyser LT (Qiagen, 2 min at 50 Hz) with a single, 7 mm stainless steel bead. RNA isolation continued with Phenol:Chloroform:IAA and Acid-Phenol:Chloroform extractions to the aqueous phase and a final isopropanol precipitation. Precipitated RNA was cleaned up according to Qiagen’s protocol on the miRNeasy column (Qiagen, Cat# 217004), with an on-column DNA digestion and RNaseOUT (Invitrogen, Cat# 10777019) added to final elution to reduce degradation during storage at −80°C. Isolation of RNA from cells was completed according to Qiagen’s protocol for the RNeasy kit, including an on-column DNase digestion.

RNA was quantified by Qubit (Invitrogen) and quality assessed with a Fragment Analyzer (Agilent), with the highest quality (typically RINs > 6, DV200 > 80%) utilized for library preparations. Pancreas RNA was prepared and sequenced in the Mayo Clinic Medical Genomics Core using the Illumina TruSeq RNA v2 library preparation kit and the Illumina High Seq-2000 with 101 bp paired end reads. Reads were aligned to the mouse reference transcriptome Gencode vM23 (GRCm38.p6) with at least 24 million mapped read pairs acquired per sample. Sequencing reads were processed through the GSPMC workflow including MapRseq3 (Kalari et al., 2014) and differential expression calculated by EdgeR (McCarthy et al., 2012). Differentially expressed genes (DEGs) were filtered based on a false discovery rate (FDR) < 0.1 and an absolute FC ≥ | 1.5| called between at least one condition and the reference control sample. For RNA-seq on caerulein-treated animals, DEGs were filtered with FC ≥ | 2| and FDR < 0.05. Pathway analysis of DEGs was completed with RITAN (R package-rapid integration of annotation, network, and molecular database) (Zimmermann et al., 2019), which queries different annotation resources to analyze enrichment for an input set of genes. RITAN performs a hypergeometric test and generates FDR adjusted p-values, called q-values, for assessing pathway significance, which queries different annotation resources to analyze enrichment for an input set of genes. The Molecular Signatures Database (MSigDB) hallmark gene set collection (Liberzon et al., 2015) was used as the annotation resource. Pathways with q-values < 0.05 were considered enriched. Gene network and upstream regulatory analyses were performed using Ingenuity^® Pathway Analysis (IPA^®; Qiagen). To quantitatively determine the percentage of infiltrating immune cell types, we utilized a “digital cytometry” method called the quanTIseq deconvolution algorithm, which estimates the absolute proportions of relevant infiltrating immune cell types from bulk RNA-seq profiles (Finotello et al., 2019). The cell type fraction scores provided by this method allow intra-sample and inter-sample comparisons of 10 immune cell type fractions. We applied the quanTIseq method through an R package called Immunedeconv (Sturm et al., 2019) on DEGs from the Pdx1-Cre;EHMT2^fl/fl, Pdx1-Cre;LSL-Kras^G12D, and Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl dataset (1039 DEGs), as well as the caerulein-treated animals dataset (4654 DEGs).

Reverse transcriptase was performed using RT² first strand kit (Qiagen) following manufacturer’s protocol. Real-time PCR was performed in a volume of 20 ul using SYBR Green Master Mix and CFX96 Real Time System (Bio-Rad). Primer sequences are shown in Supplementary Table 2. RT² Profiler PCR Array (SA Biosciences, Qiagen) was used to examine the expression patterns of 84 genes involved in cell cycle. The array was performed following manufacturer instructions. ddCT was obtained using Qiagen software and all groups were normalized to Pdx1-Cre mice, with an absolute FC ≥ | 1.5| for the generation of a heatmap using R-studio.

Senescence-associated β-galactosidase activity was assessed in fresh pancreas tissue using the Cell Signaling β-Galactosidase Staining kit, according to manufacturer’s protocol (Cell Signaling Technologies, Cat# 9860S). Briefly, under terminal anesthesia, pancreas from LSL-Kras^G12D;EHMT2^+/+ and LSL-Kras^G12D;EHMT2^fl/fl with both Pdx1-Cre and P48^Cre/+ drivers were collected, washed in PBS (1×) and placed in a well of a 12-well dish containing 1 mL of fixative solution for 15 min. Pancreas were then rinsed twice with PBS (1×) and stained with β-galactosidase solution at 37°C overnight in a dry incubator. Next morning, pancreata were washed with PBS (1×), and images acquired.

Animal care and all protocols were reviewed and approved by the Institutional Animal Care and Use Committees of Mayo Clinic Rochester (IACUC protocols A00002240-16 and A24815) and the Medical College of Wisconsin (AUA00005963).

RNA-seq datasets that support the findings of this study have been deposited to the public database repository Gene Expression Omnibus (GEO) under accession code GSE169525.

By regulating H3K9me2 levels, EHMT2 plays a role in human pancreatic cancer (Yuan et al., 2012, 2013; Casciello et al., 2015; Pan et al., 2016). Initial studies in P48^Cre/+ Kras^G12D mice indicated that EHMT2 deficiency prolonged survival by reducing PanIN growth, which was accompanied by a decreased number of phosphorylated Erk-positive and Dclk1-positive cells, both populations that contribute to PDAC initiation (Kato et al., 2020). To extend these studies, here, we use several distinct models; two purely genetic ones to express Kras^G12D for studying initiation combined with EHMT2 deletion (Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl and P48^Cre/+;LSL-Kras^G12D;EHMT2^fl/fl), an inducible Cre to investigate EHMT2 inactivation ex vivo (CAGGCre-ER^TM;EHMT2^fl/fl), as well as the model of pancreatitis-associated promotion (Guerra et al., 2007). Prior to crossing with Kras^G12D, both Pdx1-Cre;EHMT2^fl/fl and P48^Cre/+;EHMT2^fl/fl reproduced at Mendelian ratios and thrived similar to controls. Pancreata from 10-day and 4 to 6-week-old animals demonstrated normal structure and histology (Figure 1A and Supplementary Figure 1A). Body weights of EHMT2 knockout mice were not different from controls for both Pdx1-Cre and P48^Cre/+ animals (Figure 1B and Supplementary Figure 1B). The pancreatic cell nuclei from control animals were positive for H3K9me2, while from EHMT2^fl/fl mice displayed global reduction in H3K9me2 (8.0 ± 1.4% for Pdx1-Cre vs. 4.0 ± 1.0% for Pdx1-Cre;EHMT2^fl/fl, p = 0.08; Figures 1C,D). Reverse transcriptase quantitative PCR (RT-qPCR) showed that Pdx1-Cre;EHMT2^fl/fl mice display reduced EHMT2 levels in whole pancreas (1.94 ± 0.31 FC) (Figure 1E). EHMT2 inactivation in the P48^Cre/+ model also showed reduced H3K9me2 by IHC (9.02 ± 1.26% for P48^Cre/+ vs. 1.85 ± 0.28% for P48^Cre/+;EHMT2^fl/fl, p < 0.05; Supplementary Figures 1C,D) with a reduction in EHMT2 transcript by RT-qPCR (3.6 ± 0.55 FC, Supplementary Figure 1E). Cre-mediated EHMT2 exon excision in the pancreas of Pdx1-Cre;EHMT2^fl/fl and P48^Cre/+;EHMT2^fl/fl mice was detected by PCR (Supplementary Figure 1F). Notably, for the Pdx1-Cre or P48^Cre/+-driven models, organ-specific EHMT2 inactivation occurs at embryonic day 8.5 or 9.5, respectively (Herreros-Villanueva et al., 2012). Therefore, the H3K9me2 pathway is not required for pancreas exocrine development independent of the Cre model used for EHMT2 inactivation. Subsequently, we crossed Pdx1-Cre;EHMT2^fl/fl and P48^Cre/+;EHMT2^fl/fl animals to LSL-Kras^G12D mice. IHC staining demonstrated that LSL-Kras^G12D;EHMT2^fl/fl mice displayed global reduction of H3K9me2 (17.1 ± 2.7% for Pdx-Cre;LSL-Kras^G12D vs. 3.5 ± 1.3% for Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl, p < 0.05; Figures 1C,D). We also detected a 2.27 ± 0.25 FC reduction of EHMT2 transcript in whole pancreas from Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl mice compared to Pdx-Cre;LSL-Kras^G12D animals (Figure 1E). Similar results were obtained using P48^Cre/+ with reductions in the H3K9me2 mark (17.5 ± 1.3% for P48^Cre/+;LSL-Kras^G12D vs. 4.6 ± 1.0% for P48^Cre/+;LSL-Kras^G12D;EHMT2^fl/fl, p < 0.01; Supplementary Figures 1C,D) and EHMT2 transcript (4.8 ± 0.25 FC for P48^Cre/+;LSL-Kras^G12D;EHMT2^fl/fl vs. P48^Cre/+;LSL-Kras^G12D mice, p < 0.001; Supplementary Figure 1E). Western blots on pancreas protein lysates from Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl animals showed reduction in EHMT2 and H3K9me2 levels when compared to Pdx-Cre;LSL-Kras^G12D (3.5 ± 0.2 FC for EHMT2, p < 0.05 and 2.46 ± 0.34 FC for H3K9me2, p < 0.05, respectively, in Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl vs. Pdx1-Cre;LSL-Kras^G12D mice; Figure 1F). Noteworthy, while Kras^G12D expression results in larger pancreas size (Hingorani et al., 2003), knockout of EHMT2 counteracts this effect (pancreas-to-body weight ratios: 1.45 ± 0.04% for Pdx1-Cre;LSL-Kras^G12D vs. 1.17 ± 0.03% for Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl mice, p < 0.001; Figure 2A). Similar results were found in P48^Cre/+;LSL-Kras^G12D mice (1.43 ± 0.06% from P48^Cre/+;LSL-Kras^G12D vs. 1.1 ± 0.02% from P48^Cre/+;LSL-Kras^G12D;EHMT2^fl/fl mice, p < 0.001; Supplementary Figure 2A). Hence, while genetic inactivation of EHMT2 is tolerated by the organ during development, its presence appears to be critical for supporting Kras^G12D-induced pancreatic neoplastic growth (Figure 2B and Supplementary Figure 2B). Next, blinded histological examination of tissues was performed by two separate pathologists (V.A. and B.P.). We found that Kras^G12D-expressing EHMT2 knockout mice with both Cre models had limited acinar tubular complexes and atypical flat lesions with a reduction in ADM formation (Figures 2C,D and Supplementary Figures 2C,D), which is the earliest morphological hallmark of PDAC initiation (Esposito et al., 2014; Murtaugh and Keefe, 2015). In support of this observation, western blot analyses of lysates from Kras^G12D-expressing pancreata showed reduced cytokeratin 19 (Krt19) levels in Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl animals (2.12 ± 0.28 FC on Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl vs. Pdx-Cre;LSL-Kras^G12D mice, Figure 2E). Figure 2F demonstrates that Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl mice indeed had reduced PanIN when compared to Pdx1-Cre;LSL-Kras^G12D mice, (0.14 ± 0.07% vs. 0.39 ± 0.05%, respectively; p < 0.05). Similar findings were obtained with P48^Cre/+;LSL-Kras^G12D and P48^Cre/+;LSL-Kras^G12D;EHMT2^fl/fl animals (0.56 ± 0.05% vs. 0.23 ± 0.08%, respectively; p < 0.05, Supplementary Figure 2E). Our studies, stringently obtained with the two most often used Cre-drivers for KRAS-mediated initiation, suggest that EHMT2 inactivation antagonizes PanIN formation, as early as its ADM precursor stage.

To mechanistically complement our in vivo data, we next investigated whether EHMT2 is involved in the ADM process, using a genetically engineered ex vivo approach with acinar explants from a tamoxifen-inducible model of EHMT2 knockout. This model carries a CAGGCre-ER^TM transgene, which has the chicken β-actin promoter/enhancer coupled with the cytomegalovirus immediate-early enhancer to drive high levels of Cre expression, allowing this enzyme to translocate into the nucleus (Hayashi and McMahon, 2002) upon tamoxifen treatment. Acinar cells from CAGGCre-ER^TM and CAGGCre-ER^TM;EHMT2^fl/fl animals were isolated and grown in 3D matrigel in the absence (−) or presence (+) of 4-Hydroxytamoxifen (4-OHT) to inactivate EHMT2. Western blot showed that EHMT2 protein levels were reduced after 4-OHT treatment (Figure 3A). Subsequently, acinar cell explants were cultured in presence of EGF or TGFα for 5–6 days to activate the KRAS-MAPK pathway, which induces ex vivo ADM formation. Phase microscopy of 3D cultures showed that EGF and TGF-α induced duct-like structures from acini isolated from CAGGCre-ER^TM (−/+4-OHT) controls, demonstrating ADM formation (Figure 3B). No significant difference was observed in the size of duct-like structures formed from acini of CAGGCre-ER^TM alone animals either in the absence or presence of 4-OHT upon EGF treatment (17.51 ± 0.6 FC with DMSO vs. 19.07 ± 0.6 FC with 4-OHT) or after TGFα treatment (14.3 ± 0.7 FC with DMSO vs. 15.05 ± 1.4 FC with 4-OHT; Figure 3B). CAGGCre-ER^TM;EHMT2^fl/fl without 4-OHT treatment also demonstrated ADM formation after EGF and TGFα treatment (Figure 3C). However, upon EHMT2 inactivation (+4-OHT), the quantification of duct-like structures revealed significantly reduced ADM formation with EGF (19.08 ± 0.7 FC with DMSO vs. 8.36 ± 0.6 FC with 4-OHT; p < 0.001) or TGFα (17.58 ± 0.5 FC with DMSO vs. 8.19 ± 0.75 FC with 4-OHT; p < 0.001; Figure 3C). Thus, congruent with the antagonism to Kras^G12D-mediated ADM and PanIN initiation found in LSL-Kras^G12D;EHMT2^fl/fl mice, loss of EHMT2 impairs the phenotypic conversion of acinar cells to more duct-like structures in 3D ex vivo cultures stimulated with growth factors that activate the EGF-KRAS pathway, further supporting the rigor of our in vivo observations.

We also investigated whether Kras^G12D behaves as an upstream regulator of EHMT2 activity. For this purpose, we utilized a GEMM-derived PDAC cell model in which oncogenic Kras^G12D expression is induced by doxycycline (Collins et al., 2012) (iKras 4292) to evaluate downstream effects on the EHMT2-H3K9me2 epigenetic pathway. We induced Kras^G12D expression and collected lysates at various timepoints to examine levels of proteins from this pathway (Figure 4A). Kras^G12D protein levels were detected at 12 h, reaching maximum after 48 h (10.57 ± 1.6-fold change (FC) vs. 0 h; p < 0.01; Supplementary Figure 3A). We found that Kras^G12D increased protein levels for both, H3K9 methyltransferases EHMT2 and EHMT1, peaking at 48 h (4.6 ± 1.2 FC for EHMT2 at 48 h vs. 0 h, p < 0.05, and 4.18 ± 1.7 FC for EHMT1 at 48 h compared to 0 h, p < 0.01; Supplementary Figure 3A). EHMT2 and EHMT1 form an enzymatically active heterotrimer with WIZ, which stabilizes the complex on chromatin (Simon et al., 2015). We also observed increased levels of WIZ (3.2 ± 1.3 FC at 48 h vs. 0 h; p < 0.05; Supplementary Figure 3A). Levels of the H3K9me2 mark increased upon Kras^G12D induction as well (2.50 ± 0.4 FC at 48 h vs. 0 h; p < 0.001; Supplementary Figure 3A). In addition, using immunofluorescence-based confocal microscopy, we detected increased protein levels for EHMT2, EHMT1, WIZ, and H3K9me2 (Supplementary Figures 3B–E). Indeed, the mean fluorescence intensity (MFI) for EHMT2 at baseline (0 h) was 2.81e4 ± 2.41e3 compared with 6.97e4 ± 3.21e3 at 48 h (p < 0.001; Supplementary Figure 3C). For EHMT1, MFI values were 2.40e4 ± 2.21e3 at 0 h vs. 6.42e4 ± 3.91e3 at 48 h (p < 0.001; Supplementary Figure 3C). For WIZ, the MFI increased from 3.59e4 ± 3.25e3 at 0 h to 8.60e4 ± 4.96e3 at 48 h (p < 0.001; Supplementary Figure 3D). H3K9me2 MFI values were 1.38e4 ± 3.51e3 for 0 h vs. 1.20e5 ± 1.01e4 at 48 h (p < 0.001; Supplementary Figure 3E). Lastly, using a set of genetically engineered human pancreatic duct-derived cell models specifically designed to study KRAS^G12D, hTERT-HPNE E6/E7 (HPNE), and hTERT-HPNE E6/E7 KRAS^G12D (HPNE-KRAS^G12D), we observed a similar increase in the EHMT2-EHMT1-WIZ complex along with di-methylation of its H3K9 substrate. Congruent with the data described above, HPNE-KRAS^G12D cells displayed higher protein levels of EHMT2, EHMT1, WIZ, and H3K9me2 compared to its HPNE counterpart (Figure 4B and Supplementary Figure 3F). Thus, activation of oncogenic KRAS, the earliest genetic event in PDAC, results in increased levels of the EHMT2 complex and its catalytic product, the H3K9me2 mark.

Next, since stability and methyltransferase activity depends on formation of the complex (Ueda et al., 2006), we immunopurified EHMT2 from iKras 4292 cells and performed mass spectrometry, which revealed increased hetero-trimerization of EHMT2, EHMT1, and WIZ after 24 h of Kras^G12D expression (Supplementary Figure 3G). We also used PLA to detect, in situ, the interaction between endogenous EHMT2 and EHMT1 or WIZ (Figures 4C,D and Supplementary Figures 4A,B). PLA signals from EHMT2 + EHMT1 complexes increased after Kras^G12D induction, with MFI of 6.29e3 ± 3.98e2 at baseline and 1.75e4 ± 1.12e3 after 24 h (p < 0.001, Figure 4C). This effect was also accompanied by increased PLA signals for EHMT2 + WIZ complexes upon Kras^G12D expression (MFI of 7.89e3 ± 6.47e2 at baseline vs. 1.33e4 ± 1.16e3 at 24 h, p < 0.001, Figure 4D). The induction of EHMT2, EHMT1, and WIZ protein levels, as well as the formation of their complex detected by mass spectrometry and PLA are important since we did not detect changes in their mRNA levels either in the absence or presence of Kras^G12D (Supplementary Figure 4C). This result was recapitulated in the human KRAS^G12D cell models, as we found no statistical difference in EHMT2, EHMT1 and WIZ transcript levels from HPNE cells compared to HPNE-KRAS^G12D cells (Supplementary Figure 4D). These comparable transcript levels in the absence and presence of oncogene activation suggest that KRAS^G12D-effects on the EHMT2 complex are possibly not via transcriptional regulation of these genes but rather potentially due to protein stability. Given this result, we also performed staining for the H3K9me2 mark, by immunohistochemistry (IHC) in wild-type and Kras^G12D expressing mice to translate these effects to the in vivo situation. Pancreata were harvested from both, Pdx1-Cre;LSL-Kras^G12D and P48^Cre/+;LSL-Kras^G12D mouse models. Upon quantification of positively stained nuclei compared to total nuclei, we found that Pdx1-Cre;LSL-Kras^G12D animals had higher levels of H3K9me2 compared to control Pdx1-Cre mice (17.1 ± 2.7% vs. 8.0 ± 1.4%, p < 0.05; Figure 4E). Similar results were obtained when Kras^G12D activation was driven by Cre expression from the Ptf1a-P48 promoter via knock-in (17.5 ± 1.3% in P48^Cre/+;LSL-Kras^G12D mice compared to 9.0 ± 2.1% in P48^Cre/+ controls, p < 0.01; Figure 4E). Thus, based on these collective data, we conclude that EHMT2, EHMT1 and WIZ proteins form a stable complex upon Kras^G12D activation, which leads to enhanced deposition of the H3K9me2 mark, supporting a role for this epigenetic pathway downstream of the most commonly mutated oncogene in PDAC.

Oncogenic KRAS mounts transcriptional responses that program the growth-promoting phenotype that characterizes pancreatic cancer initiation (Waters and Der, 2018). This observation, along with the fact that EHMT2 is a well-known regulator of transcriptional responses (Shankar et al., 2013), led us to carry out RNA-seq from whole pancreas to evaluate gene expression networks affected by EHMT2 inactivation. We performed these experiments comparatively in four murine lines, namely Pdx1-Cre, Pdx1-Cre;EHMT2^fl/fl, Pdx1-Cre;LSL-Kras^G12D, and Pdx-Cre;LSL-Kras^G12D;EHMT2^fl/fl. We found a total of 767 DEGs (with 257 upregulated and 510 downregulated) in Pdx1-Cre;EHMT2^fl/fl, 790 DEGs (691 up, 99 down) in Pdx1-Cre;LSL-Kras^G12D and 878 DEGs (305 up, 573 down) in Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl, normalized to Pdx1-Cre mice, with an absolute FC ≥ | 1.5| and FDR < 0.1 (Figure 5A and Supplementary Table 3). The largest subset of regulated genes across all three experimental conditions was comprised of 530 DEGs. The second largest subset of 188 DEGs were differentially expressed only in EHMT2 inactivated samples, irrespective of Kras^G12D (Figure 5A). On the other hand, Kras^G12D carrying mice showed 114 DEGs irrespective of their EHMT2 status (Figure 5A). Principal component analysis (PCA) using all 1039 DEGs led to separation of samples by cluster and experimental conditions (Figure 5B). Next, we performed calculations of distances among cluster centroids for experimental samples and controls (marked by empty circles in Figure 5B), as a similarity measure among genetic conditions. The distance between the EHMT2 inactivated groups was 21.7 while the distance between Pdx1-Cre;EHMT2^fl/fl and Pdx1-Cre;LSL-Kras^G12D was 46.8 and between the two Kras^G12D activated groups was 48.7. Thus, EHMT2 inactivation in the Pdx1-Cre;LSL-Kras^G12D EHMT2^fl/fl animals clustered much closer to Pdx1-Cre;EHMT2^fl/fl and significantly separated from Kras^G12D mice with EHMT2 intact, indicating that EHMT2 inactivation is characterized by a transcriptional profile that is functionally antagonistic to oncogenic KRAS. Hierarchical clustering of normalized DEG across all three conditions (1039 genes) revealed four major expression patterns (Figure 5C). When compared to Pdx1-Cre controls and Kras^G12D animals, the 245 genes defining cluster 1 (green) were upregulated in EHMT2^fl/fl and Kras^G12D;EHMT2^fl/fl mice, while 647 genes characterized cluster 2 (yellow) with genes downregulated upon EHMT2 inactivation, either alone or with Kras^G12D. Cluster 3 (pink) displayed 65 genes upregulated in Kras^G12D animals, and Cluster 4 (purple) was defined by 82 genes downregulated in Kras^G12D animals, independent of EHMT2 status. Notably, ADM and PanIN-related mRNA markers, such as Gkn1, Gkn2, and Muc5a, were downregulated in Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl animals (Supplementary Figure 5A), supporting our histological observations. Together, these studies revealed that mice with EHMT2 inactivated in their pancreas epithelial cells displayed a gene expression pattern antagonistic to the Kras^G12D-regulated gene expression program.

We used different tools for data mining the biological and mechanistic significance of the gene expression networks identified by RNA-seq. Hierarchical clustering identified four main patterns of gene expression. Cluster 1 genes represented those upregulated upon EHMT2 inactivation that remained elevated in EHMT2^fl/fl crossed to Kras^G12D mice. Notably, cluster 1 included a gene network corresponding to the P21 pathway (q-value = 1.81 × 10^–5; Figure 5D and Supplementary Table 4), which causes growth arrest. Cluster 2 genes, downregulated by EHMT2 inactivation, involved gene networks that participate in KRAS signaling (q-value = 1.5 × 10^–10) and immunoregulatory/inflammation-related pathways (q-value = 2.3 × 10^–14), (e.g., type I interferon response, IL6-Jak-Stat3 signaling, and IL2-Stat5 signaling; Supplementary Table 4). This is important since inflammatory responses are required for Kras^G12D-mediated initiation (di Magliano and Logsdon, 2013). No biological pathway was found to be significant in cluster 3, while cluster 4 contained genes downregulated in Kras^G12D animals, which participate in EMT (q-value = 2.3 × 10^–2). Additional gene network enrichment and visualization analyses using semantic-based algorithms (Krämer et al., 2014) showed that EHMT2 inactivation results in antagonism of transcriptional networks corresponding to inflammatory responses as part of gene networks that participate in KRAS signaling (Figure 5E). Upstream Regulatory Analysis linked the upregulation of Cdkn1a/p21, Chek2, Bax, Jun, and Ccng1, among other genes, to a P53-like transcriptional network activated in EHMT2^fl/fl mice, and this molecular phenotype was transferred to the Kras^G12D;EHMT2^fl/fl mice (Figure 5F). Therefore, EHMT2 inactivation antagonizes oncogenic Kras^G12D at the molecular level, at least in part, by upregulating well-known pathways for cell cycle inhibitory checkpoints, though also unexpectedly, downregulating inflammatory responses.

RNA-seq counts from pancreas are dominated by the high expression of multiple key digestive enzyme genes (Hoang et al., 2016), occasionally masking genes expressed at lower levels that can be causal of phenotypes. For this reason, we also utilized more sensitive, pathway-specific RT-qPCR-based gene expression profiling focused on cell cycle-related transcripts to monitor the status of the entire gene expression network that regulates cell cycle. Results from EHMT2^fl/fl, Kras^G12D, or Kras^G12D;EHMT2^fl/fl mice were normalized to those from Pdx1-Cre mice. Indeed, this approach revealed that Kras^G12D;EHMT2^fl/fl and EHMT2^fl/fl animals clustered closer to each other than with Kras^G12D, sharing a similar expression profile for cell cycle-related genes (Supplementary Figure 5B). Genes forming Cluster 1 were involved in cell cycle arrest and senescence after replication stress (Cdkn1a/p21 and Chek2) (Aliouat-Denis et al., 2005; Cook, 2009), remaining upregulated in both EHMT2^fl/fl and Kras^G12D;EHMT2^fl/fl mice (orange, Supplementary Figure 5B). Genes within cluster 2 contained networks that also function in replication stress responses (Atr, Wee1, Brca1, and Rad9a), which often also result in senescence induction (Gralewska et al., 2020), as well as genes, including Skp2 and Cdc6, involved in origin licensing during DNA replication (Cook, 2009). These genes ranged in levels from unchanged to slightly increased by Kras^G12D but underwent further upregulation upon EHMT2 inactivation, whether in the absence or presence of Kras^G12D (purple, Supplementary Figure 5B). Cluster 3 genes were characterized by the presence of the replication stress kinase Chek1 and several cyclins (Ccna2, Ccnb1, Ccnd1, and Ccnd2). These genes were downregulated by Kras^G12D alone but markedly upregulated in both EHMT2^fl/fl and Kras^G12D;EHMT2^fl/fl mice (green, Supplementary Figure 5B). Thus, combined these data suggest that EHMT2 inactivation alone increases the levels of cell cycle regulatory molecules known to function as checkpoints to arrest cell growth in response to replication stress, congruent with emerging data suggesting that this protein localizes and works at the replication fork (Estève et al., 2006; Dungrawala et al., 2015; Ferry et al., 2017) and pharmacological data that has implicated a role of this protein in senescence without much mechanistic insight (Yuan et al., 2012). Notably, these expression networks were inherited when EHMT2^fl/fl mice were crossed to the Kras^G12D background. While cluster 4 was primarily formed by Kras^G12D-downregulated transcripts, its pattern displayed a mixed profile of downregulated, unchanged, and minimally upregulated transcripts in the EHMT2^fl/fl and Kras^G12D;EHMT2^fl/fl mice (red, Supplementary Figure 5B). EHMT2^fl/fl and Kras^G12D;EHMT2^fl/fl mice blunted the Kras^G12D-mediated downregulation of transcripts in this cluster, which encompassed DNA damage and apoptosis-related genes. Thus, these data further support the significant role that EHMT2 inactivation plays in the regulation of gene expression networks that are primarily involved in signaling for replication stress, cell cycle arrest and senescence.

Based on the discovery that EHMT2 inactivation upregulates the Cdkn1a/p21-Chek2 pathway, which when participating in prolonged cell cycle arrest, leads to senescence (Aliouat-Denis et al., 2005; d’Adda di Fagagna, 2008), we hypothesized that loss of EHMT2 may prevent Kras^G12D-mediated cell growth, at least in part, through this mechanism. Thus, we performed fresh tissue senescence-associated β-galactosidase staining on Pdx1-Cre;LSL-Kras^G12D and Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl pancreas tissue. Tissue from Kras^G12D;EHMT2^fl/fl mice displayed marked blue staining indicating increased senescence compared to Kras^G12D mice with EHMT2 intact (Figure 6A). Similar results were obtained with P48^Cre/+;Kras^G12D;EHMT2^fl/fl (Supplementary Figure 6A). Concomitantly, using western blot using lysates from Pdx1-Cre;LSL-Kras^G12D and Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl mice (n = 6/group), we found that EHMT2 inactivation, which upregulated Cdkn1a/p21 mRNA, also increased the levels of its encoded cell cycle regulator protein (2.19 ± 0.38 FC for Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl vs. Pdx1-Cre;LSL-Kras^G12D; p < 0.05; Figure 6B and Supplementary Figure 6B). IHC performed on tissue from these animals further supported the increase of P21 protein in the nuclei of pancreas cells with EHMT2 inactivated (0.45 ± 0.05% from Pdx1-Cre;LSL-Kras^G12D vs. 12.03 ± 2.00% from Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl animals, p < 0.05; and 0.26 ± 0.05% from P48^Cre/+;LSL-Kras^G12D vs. 4.35 ± 0.06% from P48^Cre/+;LSL-Kras^G12D;EHMT2^fl/fl mice, p < 0.001; Figure 6C and Supplementary Figure 6C). Thus, EHMT2 inactivation exerts its inhibitory effects on Kras^G12D, at least in part, by one cellular mechanism, namely senescence. At the molecular level, we reveal that genetic inactivation of EHMT2 generates a transcriptional profile, which is antagonistic to oncogenesis and transferred to the cross with Kras^G12D. This profile is represented by linked nodes that converge on the upregulation of P21-mediated pathways, a finding that is validated by evidence gathered at the transcriptional level, protein level, and by in situ immunostaining.

Pancreatic Intraepithelial Neoplasia formation and progression requires infiltration of immune cells, which is known to be initiated by Kras activity (Guerra et al., 2007; di Magliano and Logsdon, 2013). In fact, an inflammatory response supports cells to overcome the oncogene-induced senescence (OIS) barrier (Guerra et al., 2011). Considering that our transcriptional signatures show an antagonistic effect of EHMT2 inactivation on the expression of Kras-mediated inflammatory genes (Figures 5D,E), we performed histopathological examination of 8-week-old Pdx1-Cre pancreata to evaluate the inflammatory cell infiltration. On the Pdx1-Cre background, we found that 100% of Kras^G12D animals showed the presence of polymorphonuclear (PMN) leukocytes or histiocytes compared to only 33.3% of Kras^G12D;EHMT2^fl/fl animals (Figure 7A). When evaluating P48^Cre/+ animals, again 100% of Kras^G12D animals had infiltration of these inflammatory cells as opposed to only 81.8% of Kras^G12D;EHMT2^fl/fl animals (Figure 7A). Subsequently, the extent of PMN or histiocyte luminal infiltration was scored as absent, rare, focal or substantial. For Pdx1-Cre mice, all Kras^G12D animals (100%) presented with rare infiltration, while these cells were absent in the majority of Kras^G12D;EHMT2^fl/fl animals (66.6%) with few mice demonstrating rare (16.7%) or focal (16.7%) infiltration (Figure 7B). In a similar manner, all P48^Cre/+;Kras^G12D mice showed either focal (60%) or substantial (40%) infiltration. However, in P48^Cre/+;Kras^G12D;EHMT2^fl/fl animals, infiltration of these cells was mostly absent (18.2%) or rare (72.7%) with a minor portion of focal (9.1%). Additionally, we examined the degree of inflammation at foci of acinar tubular complexes and atypical flat lesions, which was categorized as absent, mild, moderate or severe. All of the Pdx1-Cre;Kras^G12D animals showed moderate (40%) to severe (60%) inflammation, compared to 66.6% of the Kras^G12D;EHMT2^fl/fl animals with only mild inflammation and few with moderate (16.7%) to severe (16.7%) inflammation (Figure 7C). Likewise, P48^Cre/+;Kras^G12D animals presented with moderate (20%) to severe (80%) inflammation, whereas those carrying inactivating EHMT2^fl/fl alleles most commonly demonstrated mild inflammation (72.7%) with only 18.2% scored as moderate (Figure 7C). For both Cre backgrounds, Kras^G12D animals predominantly had a mixed cell type infiltrate (60% for Pdx1-Cre and 100% for P48^Cre/+), while Kras^G12D;EHMT2^fl/fl animals had a lymphoplasmacytic predominant cell type (50% for Pdx1-Cre and 45% for P48^Cre/+), rather than neutrophilic or mixed (Figure 7D). Utilizing a deconvolution approach for bulk RNA-seq (Finotello et al., 2019), we estimated the proportions of different types of immune cell populations. This approach revealed that Pdx1-Cre;Kras^G12D animals had a markedly higher amount of DEG markers for T cells, neutrophils, B cells, and M1 macrophages (28.4, 28.3, 20.6, and 6.4%, respectively) when compared to Pdx1-Cre;EHMT2^fl/fl (0.20, 6.98, 0.26, and 0.40%, respectively) and Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl (0.06, 4.96, 0.08, and 0.27%, respectively; Figure 7E). Together, these data corroborate our gene network enrichment findings, demonstrating that EHMT2 inactivation promotes an anti-inflammatory phenotype that antagonizes Kras^G12D, leading to reduced inflammatory cell infiltration seen by both, histology and deconvolution of RNA-seq data.

Prolonged or chronic inflammatory conditions promote the initiation and development of neoplastic and fibrotic events, serving as a major risk factor leading to PDAC (Momi et al., 2012). Pancreatitis induced by caerulein, a decapeptide that stimulate Gq-coupled growth regulatory receptors (e.g., CCK), accelerates the effects of oncogenic Kras (Murtaugh and Keefe, 2015). Pancreatitis-accelerated Kras-induced neoplastic growth in mice experimentally models the inflammation-associated progression, in which the microenvironment aids the growth-promoting process (Perez-Mancera et al., 2012). We injected 4-week-old Pdx1-Cre;LSL-Kras^G12D and Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl mice with caerulein for 4 weeks to induce repeated chronic pancreatitis. Macroscopic examination of the pancreas in animals sacrificed after 4 weeks of caerulein treatment demonstrated that EHMT2 deletion antagonizes the known effect of Kras^G12D to increase pancreas-to-body weight ratios (Figure 8A; 3.27 ± 0.09% for Pdx1-Cre;LSL-Kras^G12D vs. 1.43 ± 0.06% for Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl; p < 0.001; n = 5 and 13, respectively). Histopathology showed extensive and dysplastic PanIN lesions with caerulein treatment in Kras^G12D animals, while pancreatic tissues from Kras^G12D mice with EHMT2 inactivation displayed more limited dysplasia (Figure 8B). Measurements of pancreas area with mucin-rich PanIN lesions detected by Alcian Blue staining substantiated that Kras^G12D;EHMT2^fl/fl mice had reduced area occupied by PanIN lesions, even when challenged by the inflammation-stimulating conditions (Figure 8C; 1.28 ± 0.13% Kras^G12D;EHMT2^fl/fl + caerulein vs. 2.64 ± 0.23% Kras^G12D + caerulein; p < 0.001). We conclude that EHMT2 deletion antagonizes Kras^G12D-mediated cell growth even after enhanced stimulation by pancreatitis.

We performed RNA-seq in both caerulein-treated Kras^G12D, and Kras^G12D;EHMT2^fl/fl animals and compared them with untreated Kras^G12D mice as controls (Figure 8D; fold change ≥ |2| and FDR < 0.05). We identified 3881 caerulein-induced DEGs in Kras^G12D animals (2024 upregulated and 1857 downregulated). The upregulated subset included not only 6 distinct AP-1 transcription factors, which participate in growth, but also 19 different collagen genes and 11 chemokine ligands (Supplementary Table 5), reflecting the functional expansion of the tumor microenvironment. Among downregulated networks, we found 10 different subunits of mitochondrial respiratory chain complexes (Supplementary Table 5). Caerulein-treated Kras^G12D;EHMT2^fl/fl mice displayed 3612 DEGs (1949 upregulated and 1663 downregulated). Of these, 2839 DEGs were shared with caerulein-treated Kras^G12D mice, but 773 DEGs were unique to EHMT2 inactivation (Figure 8D). Thus, overall, chronic caerulein treatment of Kras^G12D;EHMT2^fl/fl mice induces significant changes of gene expression. Consequently, we used PCA to measure genome-wide level differences in the transcriptional landscape of these animals by comparing the position of centroids in 3D (Supplementary Figure 7A). The centroid separation in 3D of the Kras^G12D and caerulein-treated Kras^G12D mice was 114, while for Kras^G12D and caerulein-treated Kras^G12D;EHMT2^fl/fl, it was 105. The distance between the two caerulein treated conditions was 60, indicative of a dominant effect of caerulein administration on gene expression. Hierarchical clustering of RPKM for these DEGs (4654 genes), depicted as a heatmap in Figure 8E, identified four main expression patterns or clusters, which were annotated for pathway enrichment analyses (Figure 8F). Cluster 1 included metabolic gene networks that were downregulated by caerulein in Kras^G12D mice regardless of EHMT2 status (Figure 8F and Supplementary Table 6). This observation is relevant in light of the emergent relationship between metabolism and cell growth regulation by oncogenic KRAS (Kimmelman, 2015). Another important downregulated gene network, ontologically known as KRAS signaling “down,” contained genes typically downregulated by KRAS signaling (q-value = 6.1 × 10^–10), which is congruent with caerulein signaling via the Gq pathway (Goldsmith and Dhanasekaran, 2007). EMT gene networks were upregulated by caerulein independent of the EHMT2 status, forming cluster 4 (q-value = 1.6 × 10^–22; Figures 7E,F and Supplementary Table 6). Cluster 2 was comprised of genes with expression upregulated by caerulein in Kras^G12D;EHMT2^fl/fl mice in comparison with caerulein-treated Kras^G12D mice with EHMT2 intact (Figure 8E). Networks in cluster 2 were prominently represented by immunoregulatory genes (Figure 8F and Supplementary Table 6), such as H2-Oa (human gene HLA-DOA), H2-Ob (human gene HLA-DOB), H2-DMb2 (human gene HLA-DMB), Cd2, Cd96, Cd28, Cd8a, Ccl5, Lck, and Zap70 (q-value = 1.5 × 10^–14), congruent with lymphocyte infiltration of T-cell origin. Finally, cluster 3 contained genes downregulated in Kras^G12D animals carrying EHMT2 inactivation (Figure 8E), primarily representing EMT processes (q-value = 7.7 × 10^–12), TNF-α signaling via NFκB (q-value = 2.3 × 10^–12), as well as genes upregulated in response to KRAS signaling (q-value = 1.4 × 10^–08) among others (Figure 8F and Supplementary Table 6). Overall, these data suggest that during the process of pancreatitis-enhanced carcinogenesis by Kras^G12D, EHMT2 influences immunomodulatory processes, while reducing EMT, TNF-α signaling via NFκB, and KRAS signaling, which have well-documented effects on growth promotion (Thiery et al., 2009; Taniguchi and Karin, 2018; Waters and Der, 2018). Functional inferences were further gathered by building gene expression networks represented with color-coded nodes, corresponding to fold changes in gene expression (Supplementary Figure 7B). These networks better illustrate that both, Kras and TNF-α-NFκB pathways were upregulated by pancreatitis to a significantly lesser extent in mice carrying EHMT2 inactivation. This effect is particularly evident for a portion of the KRAS-associated gene network, composed of Itga2, Etv4, and Wnt7a, and the TNF-α-NFκB pathway, containing Areg, Fos, and Cxcl5 (human gene CXCL6) (Supplementary Figure 7B). Notably, the growth inhibitory networks that included Cdkn1a/p21 and Chek2, which were upregulated in animals with EHMT2 inactivation under basal conditions (Figure 5F and Supplementary Figure 5), were no longer a hallmark of the caerulein-treated Kras^G12D;EHMT2^fl/fl transcriptome. Using the same RNA-seq deconvolution approach shown in Figure 7E from pancreas tissue, we found that the treatment of Kras^G12D mice with caerulein increased regulatory T cells (Tregs), NK cells, myeloid dendritic cells, monocytes and M2 macrophages (Figure 8G). Surprisingly, caerulein-treated Kras^G12D animals with inactivation of EHMT2 not only had the presence of the immune cell types found in the caerulein-treated Kras^G12D mice, but in addition, increased the infiltration of CD8+ and non-regulatory CD4+ T cells, B cells and neutrophils (Figure 8G). This result demonstrated that inactivation of this histone methyl transferase, in epithelial cells, leads to a change in the Kras^G12D immune landscape in response to pancreatitis by significantly enriching the infiltrate with both cytotoxic and non-regulatory T cells and B cells, which are known to work in concert to mount more efficient antitumor responses (Wörmann et al., 2014). Immunohistochemical analyses, using CD3 as a pan-T cell marker, supported the increased immune infiltration in the pancreas of caerulein-treated Kras^G12D;EHMT2^fl/fl mice (18.76 ± 2.88% for caerulein-treated Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl vs. 8.96 ± 1.66% for caerulein-treated Pdx1-Cre;LSL-Kras^G12D; p < 0.05; n = 4) that was not detected in their corresponding untreated cohort (0.57 ± 0.14% for Pdx1-Cre;LSL-Kras^G12D;EHMT2^fl/fl vs. 1.08 ± 0.28% for Pdx1-Cre;LSL-Kras^G12D; n = 4; Supplementary Figure 7C). In summary, our data demonstrate that genetic inactivation of EHMT2 interferes with caerulein-induced promotion of Kras^G12D-induced effects at the gross and histopathological levels. At a molecular level, EHMT2 inactivation affects gene expression networks involved in growth and immunoregulatory processes. This genotype-phenotype integration of transcriptomic data suggests that targeting EHMT2 for inactivation ameliorates cell growth- and inflammation-associated Kras^G12D functions, most likely by a combined effect not only on the targeted pancreatic epithelial cells but also in its contributions to cell populations in the microenvironment, such as those from the immune system.