. 2025 Jul;6(7):1263-1282.

doi: 10.1038/s43018-025-01003-3. Epub 2025 Jun 30.

Self-amplifying NRF2-EZH2 epigenetic loop converts KRAS-initiated progenitors to invasive pancreatic cancer

Laura Antonucci¹, Na Li², Angeles Duran³, Isidoro Cobo^{4

5

6}, Chiara Nicoletti⁷, Kosuke Watari¹, Shuvro Prokash Nandi^{4

8}, Feng Zhu⁹, Yongmei Zhao¹⁰, Irene Riahi¹¹, Motoyuki Tsuda¹², Vidhi M Shah^{12

13}, Terry Morgan¹⁴, Trent Waugh¹³, Luca Caputo⁷, Yuan Liu¹, Alexandra Rundberg Nilsson¹, Hongxu Xian¹, Jelena Todoric^{1

15}, Li Gu¹, Elsa Sanchez-Lopez¹⁶, Guido Eibl¹⁷, Emily A Vucic¹⁸, Michal Krawczyk¹⁹, Qianlan Xu¹⁹, Andrew M Lowy²⁰, Georgia Hatzivassiliou²¹, Merone Roose-Girma²¹, Dorota Skowronska-Krawczyk¹⁹, David A Scott²², Dafna Bar-Sagi¹⁸, Pablo Tamayo²³, Ying Wu¹⁰, Rosalie C Sears^{12

13

24}, Christopher K Glass⁴, Ludmil B Alexandrov^{4

8

25}, Pier Lorenzo Puri⁷, David W Dawson¹¹, Yinling Hu⁹, Maria T Diaz-Meco³, Jorge Moscat³, Michael Karin²⁶

Affiliations

¹ Laboratory of Gene Regulation and Signal Transduction, Departments of Pharmacology and Pathology, University of California, San Diego School of Medicine, La Jolla, CA, USA.
² Department of Radiation Oncology, Shandong Cancer Hospital and Institute, Shandong First Medical University, Jinan, China.
³ Department of Pathology and Laboratory Medicine and Sandra and Edward Meyer Cancer Center, Weill Cornell Medicine, New York, NY, USA.
⁴ Department of Cellular & Molecular Medicine, University of California, San Diego School of Medicine, La Jolla, CA, USA.
⁵ Division of Clinical Immunology & Rheumatology, Department of Medicine, Heersink School of Medicine, University of Alabama at Birmingham, Birmingham, AL, USA.
⁶ Comprehensive Arthritis, Musculoskeletal, Bone and Autoimmunity Center, University of Alabama at Birmingham, Birmingham, AL, USA.
⁷ Development, Aging and Regeneration Program, Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA, USA.
⁸ Moores Cancer Center, University of California, San Diego School of Medicine, La Jolla, CA, USA.
⁹ Cancer Innovation Laboratory, Center for Cancer Research, National Cancer Institute, Frederick National Laboratory for Cancer Research, Frederick, MD, USA.
¹⁰ CCR Sequencing Facility Bioinformatics Group, Bioinformatics and Computational Science Directorate, Frederick National Laboratory for Cancer Research, Frederick, MD, USA.
¹¹ Departments of Pathology and Laboratory Medicine, Jonsson Comprehensive Cancer Center, David Geffen School of Medicine at University of California, Los Angeles, Los Angeles, CA, USA.
¹² Department of Molecular and Medical Genetics, School of Medicine Oregon Health and Science University, Portland, OR, USA.
¹³ Brenden-Colson Center for Pancreatic Care, School of Medicine Oregon Health and Science University, Portland, OR, USA.
¹⁴ Department of Pathology, School of Medicine Oregon Health and Science University, Portland, OR, USA.
¹⁵ Department of Laboratory Medicine, Medical University of Vienna, Vienna, Austria.
¹⁶ Department of Orthopedic Surgery, University of California, San Diego School of Medicine, La Jolla, CA, USA.
¹⁷ Department of Surgery, University of California, Los Angeles, Los Angeles, CA, USA.
¹⁸ Department of Biochemistry and Molecular Pharmacology, New York University School of Medicine, New York, NY, USA.
¹⁹ Department of Physiology and Biophysics, Center for Translational Vision Research, School of Medicine, University of California, Irvine, Irvine, CA, USA.
²⁰ Department of Surgery, Division of Surgical Oncology, Moores Cancer Center, University of California, San Diego School of Medicine, La Jolla, CA, USA.
²¹ Genentech, Inc., San Francisco, CA, USA.
²² Cancer Metabolism Core, Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA, USA.
²³ Center for Novel Therapeutics and Division of Medical Genetics, Department of Medicine, Moores Cancer Center, University of California, San Diego School of Medicine, La Jolla, CA, USA.
²⁴ Knight Cancer Institute, Oregon Health and Science University, Portland, OR, USA.
²⁵ Department of Bioengineering, University of California, San Diego, La Jolla, CA, USA.
²⁶ Laboratory of Gene Regulation and Signal Transduction, Departments of Pharmacology and Pathology, University of California, San Diego School of Medicine, La Jolla, CA, USA. karinoffice@health.ucsd.edu.

PMID: 40588523
PMCID: PMC12331375
DOI: 10.1038/s43018-025-01003-3

Self-amplifying NRF2-EZH2 epigenetic loop converts KRAS-initiated progenitors to invasive pancreatic cancer

Laura Antonucci et al. Nat Cancer. 2025 Jul.

. 2025 Jul;6(7):1263-1282.

doi: 10.1038/s43018-025-01003-3. Epub 2025 Jun 30.

Authors

Affiliations

¹ Laboratory of Gene Regulation and Signal Transduction, Departments of Pharmacology and Pathology, University of California, San Diego School of Medicine, La Jolla, CA, USA.
² Department of Radiation Oncology, Shandong Cancer Hospital and Institute, Shandong First Medical University, Jinan, China.
³ Department of Pathology and Laboratory Medicine and Sandra and Edward Meyer Cancer Center, Weill Cornell Medicine, New York, NY, USA.
⁴ Department of Cellular & Molecular Medicine, University of California, San Diego School of Medicine, La Jolla, CA, USA.
⁵ Division of Clinical Immunology & Rheumatology, Department of Medicine, Heersink School of Medicine, University of Alabama at Birmingham, Birmingham, AL, USA.
⁶ Comprehensive Arthritis, Musculoskeletal, Bone and Autoimmunity Center, University of Alabama at Birmingham, Birmingham, AL, USA.
⁷ Development, Aging and Regeneration Program, Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA, USA.
⁸ Moores Cancer Center, University of California, San Diego School of Medicine, La Jolla, CA, USA.
⁹ Cancer Innovation Laboratory, Center for Cancer Research, National Cancer Institute, Frederick National Laboratory for Cancer Research, Frederick, MD, USA.
¹⁰ CCR Sequencing Facility Bioinformatics Group, Bioinformatics and Computational Science Directorate, Frederick National Laboratory for Cancer Research, Frederick, MD, USA.
¹¹ Departments of Pathology and Laboratory Medicine, Jonsson Comprehensive Cancer Center, David Geffen School of Medicine at University of California, Los Angeles, Los Angeles, CA, USA.
¹² Department of Molecular and Medical Genetics, School of Medicine Oregon Health and Science University, Portland, OR, USA.
¹³ Brenden-Colson Center for Pancreatic Care, School of Medicine Oregon Health and Science University, Portland, OR, USA.
¹⁴ Department of Pathology, School of Medicine Oregon Health and Science University, Portland, OR, USA.
¹⁵ Department of Laboratory Medicine, Medical University of Vienna, Vienna, Austria.
¹⁶ Department of Orthopedic Surgery, University of California, San Diego School of Medicine, La Jolla, CA, USA.
¹⁷ Department of Surgery, University of California, Los Angeles, Los Angeles, CA, USA.
¹⁸ Department of Biochemistry and Molecular Pharmacology, New York University School of Medicine, New York, NY, USA.
¹⁹ Department of Physiology and Biophysics, Center for Translational Vision Research, School of Medicine, University of California, Irvine, Irvine, CA, USA.
²⁰ Department of Surgery, Division of Surgical Oncology, Moores Cancer Center, University of California, San Diego School of Medicine, La Jolla, CA, USA.
²¹ Genentech, Inc., San Francisco, CA, USA.
²² Cancer Metabolism Core, Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA, USA.
²³ Center for Novel Therapeutics and Division of Medical Genetics, Department of Medicine, Moores Cancer Center, University of California, San Diego School of Medicine, La Jolla, CA, USA.
²⁴ Knight Cancer Institute, Oregon Health and Science University, Portland, OR, USA.
²⁵ Department of Bioengineering, University of California, San Diego, La Jolla, CA, USA.
²⁶ Laboratory of Gene Regulation and Signal Transduction, Departments of Pharmacology and Pathology, University of California, San Diego School of Medicine, La Jolla, CA, USA. karinoffice@health.ucsd.edu.

PMID: 40588523
PMCID: PMC12331375
DOI: 10.1038/s43018-025-01003-3

Abstract

Pancreatic ductal adenocarcinoma (PDAC) emerges from mutant KRAS-harboring but dormant low-grade pancreatic intraepithelial neoplasia (PanIN). To examine the role of oxidative stress, a putative PDAC risk factor, we established an organoid-based transformation system. Although the prototypic oxidant H₂O₂ induced organoid transformation, its effect was nonmutational and was mediated by the oxidant-responsive transcription factor NRF2, which induced the histone methyltransferase EZH2. Congruently, nonoxidizing NRF2 activators triggered organoid malignant conversion through NRF2 and EZH2, establishing a hitherto unknown epigenetic mechanism underlying PanIN-to-PDAC progression. While NRF2 induced EZH2 gene transcription in mouse and human PDAC, EZH2, a general repressor, coactivated transcription of NRF2-encoding NFE2L2 and interacted with other transcription factors to induce genes that sustain PDAC metabolic demands. The self-amplifying NRF2-EZH2 epigenetic loop also accounted for inflammation-driven PanIN-to-PDAC progression in vivo and was upregulated in established human PDAC, whose malignancy was maintained by NRF2 binding to the EZH2 promoter.

PubMed Disclaimer

Conflict of interest statement

Competing interests: G.H. and M.G. are employees of Genentech/Roche and hold company shares. M. Karin is a founder and scientific advisory board member of Elgia Therapeutics and has received research support from Merck, Jenssen and Gossamer Bio. L.B.A is a compensated consultant and has equity interest in io9 and Genome Insight and his spouse is an employee of Biotheranostics. L.B.A. is also an inventor of US patent 10,776,718 for source identification by non-negative matrix factorization. L.B.A. declares US provisional applications 63/289,601, 63/269,033, 63/483,237, 63/366,392 and 63/367,846. L.B.A. and S.P.N. also declare US provisional applications 63/412,835 and 63/492,348. R.C.S. is a consultant for Larkspur Biosciences and member of the scientific advisory board of RAPPTA Therapeutics and has received sponsored research support from Cardiff Oncology, AstraZeneca Partner of Choice grant award. The remaining authors declare no competing interests.

Figures

**Extended Data Fig. 1. Enduring EZH2, MAPK signaling and NOX1 in CAE-stimulated PanIN to PDAC progression.**
a, CAE treatment and mouse analysis scheme. b, EZH2 and NRF2 IB densitometry from Fig. 1a. Arbitrary units (A.U.) relative to ERK1/2. Mean±SEM (n=6 for all the experimental groups, 9 for WT CTRL; independent runs for biological replicates). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value= 0.0043, 0.0097, 0.0361, 0.0001, 0.0006, 0.0248, 0.0279, 0.0230, 0.0064, 0.0194; ns-not significant). c, NQO1 and EZH2 IHC quantification from Fig.1b. Kruskal-Wallis, Dunn’s multiple comparisons test (adjusted P value= 0.0009, 0.0132, 0.0039, 0.0067, 0.0024, 0.0004, 0.0003, 0.0012, 0.0013, <0.0001; ns-not significant). Mean±SEM (from left to right, n/fields=20, 10, 14, 24, 9, 21, 13, 10, 13, 13, 7, 19). d, IB of total and phosphorylated ERK1/2 in 8-wo WT and *Kras*^G12D/PEC pancreata 3 and 7 days after CAE injection. HSP70: loading control (top); Densitometric quantitation of phosphorylated to total ERK1/2 normalized to HSP70 (bottom). Arbitrary units (A.U.). Mean±SEM (n=4 for all experimental groups, 6 for WT and *Kras*^G12D/PEC CTRL, independent runs for biological replicates). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value= 0.0082, 0.0117, 0.0006, 0.0012, <0.0001; ns-not significant). e, Differentially expressed NOX1 complex genes in pancreata of indicated genotypes and treatments as in d (n/pancreata=3 for *Kras* CTRL 3d, 4 for WT CAE 7d, 2 for all the other experimental groups). Vertical rows-different specimens. Horizontal rows-individual genes, colored according to log-transformed transcript intensity in z-scored units. TPM (transcripts per million). Green: replicates with low expression (z-score = −2); red: replicates with high expression (z-score = +2). f, Staining of NOX1 and 4-HNE in pancreata treated as in e (f, top). Scale bars, 50μm. Quantification of positively stained areas (f, bottom). Mean±SEM (from left to right, n=14, 6, 11, 18, 10, 6, 31, 23, 21, 26, 23, 48, fields). Kruskal-Wallis test, Dunn’s multiple comparisons test (adjusted P value= 0.0179, 0.0013, 0.0021, 0.0001, 0.0266, 0.0003, 0.0087, <0.0001; ns-not significant). g, IB analysis of 4-HNE-modified proteins in above samples. HSP90: loading control. h, Schematic of CAE and ML385 (NRF2i) treatment of WT and *Kras*^G12D/PEC mice. i, Quantification of EZH2 and NQO1 positive areas in *Kras*^G12D/PEC pancreata from Fig. 2c. Mean±SEM (n/fields=10, 12 for CAE EZH2 IHC, 14 for CAE+ML385). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value <0.0001, ns-not significant). j, IHC staining for the indicated antigens (left), and quantification of positively stained areas (right) in 8-mo WT and *Kras*^G12D/PEC pancreata treated as indicated 4- to 6-weeks post CAE. Scale bars, 50μm. Mean±SEM (from top to bottom, for each experimental group indicated left to right, n/fields=12, 11, 5, 3 CTRL; 6, 7, 3, 9 CAE; 10, 9, 3 CAE+ML385). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value = 0.001, 0.0087, <0.0001, ns-not significant). The immunoblot data in the figure are representative of 2–3 independent runs.

**Extended Data Fig. 2. EZH2 is upregulated in *Kras*^G12D/PEC/Ikkα^ΔPEC PDAC.**
a, H&E and EZH2 staining of pancreata from 3- and 6-mo mice of the indicated genotypes. Scale bars, 50μm. **b-c,** Contingency tables (b) and corresponding graphical presentation (c), of strong, weak/moderate, or negative EZH2 staining frequency among ADM (top) and PanIN (bottom) of the indicated genotypes. Between 76–85 ADMs and between 47–75 PanINs were analyzed in 16-wo mice. Significance was determined by contingency Chi-square test for trend. P<0.0001. **d-e**, EZH2 and NRF2 proteins (d) and mRNAs (e) in 12-wo pancreata. NQO1: NRF2 target; HSP70: loading control. Mean±SEM (n=3 mice/group). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value =0.0120, 0.0161, 0.0007, 0.0006, <0.0001, ns-not significant). The IBs are representative of 3 biological replicates and 2 replicate blots. **f-g,** *Ezh2* mRNA amounts in acinar (UEA⁺) and ductal (DBA⁺) (f), or EPCAM⁺ (g) cells from 12-wo pancreata of the indicated genotypes. Mean±SEM (n=3 mice). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value <0.0001, ns-not significant). h, *Nfe2le, Nqo1* and *Ezh2* mRNA amounts in the indicated 8-wo pancreata. Mean±SD (n = 3 mice). Multiple paired t-tests, Holm-Šídák method (P value= 0.000018, 0.000198, <0.000001). **i-j,** EZH2 and NQO1 proteins (i) and mRNAs (j) in the indicated 18-mo pancreata; HSP70: loading control. Mean±SD (n= 4 mice/group). Unpaired, two-tailed t-tests (adjusted P value = 0.0004, 0.000468, 0.000329, 0.005333, 0.000904, 0.000501, 0.000194, <0.0001).

**Extended Data Fig. 3. EZH2 and NRF2 are upregulated during human PanIN to PDAC progression.**
a, Indicative serial sections from 2 PDAC TMAs stained for EZH2 and NRF2. Patients #57 (UCLA) and ST-00008601 (OHSU) have low NRF2 (score 1) and EZH2 (histoscore 60) expression and patients #108 (UCLA) and ST-00004242 (OHSU) have high NRF2 (score 2 and 3) and EZH2 (histoscore 300 and 240) expression (images captured with a 20x objective, Scale bars, 100μm). b, Contingency tables for high and low NRF2 and EZH2 expression group membership (PDAC UCLA TMA, top; PDAC and PanIN OHSU TMA, center, bottom). c, EZH2 histoscores of normal ducts, PanIN1 or PanIN2/3 lesions from resection specimens in the indicated TMAs (PDAC UCLA TMA, left; PDAC and PanIN OHSU TMA, right) and bar graphs showing NRF2 and EZH2 expression in PanIN1 and PanIN2/3 (OHSU TMA). Mean±SD (left: n=34 normal duct; n=22 PanIN1; n=16 PanIN2/3; right: n=11 normal duct; n=15 PanIN1; n=16 PanIN2/3). Kruskal-Wallis, Dunn’s multiple comparisons test (adjusted P value =0.029, PanIN1 vs Normal Duct; 0.036 PanIN2/3 vs Normal Duct; 0.0003, 0.0506 ns-not significant). d, *EZH2* mRNA in preneoplastic human epithelial cell profiles representing different stages of multistep pancreatic tumorigenesis (GSE19650). Normal ducts, intraductal papillary-mucinous adenoma (IPMA), intraductal papillary-mucinous neoplasm (IPMN), and intraductal papillary-mucinous carcinoma (IPMC) (left). Kruskal-Wallis P value=0.0006. *EZH2* mRNA in PDAC tumor tissue (n=70) and adjacent non-malignant tissue (Adj. Normal, n=25) in publicly available cohorts (GSE15471, GSE16515) (right). Student’s t-test P<0.0001. e, Computational analysis of human PDAC TCGA data plotting relative *EZH2* and *NQO1* mRNA amounts at the top and bottom quartiles of EZH2-expressing PDACs. Pearson correlation: 0.3914 (P=0.0001); Spearman correlation: 0.5274 (P<0.0001). f, Heat map showing status of selected mutations in TCGA PAAD (PDAC) samples sorted according to the amplitude of the NRF2_EZH2 loop transcriptional signature projected using single-sample GSEA. See Methods, Fig.7, Extended Data Fig.8 and Supplementary Tables 1,2.

**Extended Data Fig. 4. NRF2 stabilization accelerates KRAS-initiated PanIN to PDAC progression.**
a, Schematic of key signaling pathways that control NRF2 expression and activation. b, Schematic of the conditional *Nfe2l2*^E79Q allele and *Nrf2*^Act-PEC generation by CRE-mediated excision-recombination. c, Representative PCR genotyping of the indicated alleles. d, NQO1 IB from 16-wo pancreata of indicated genotypes. WT: *Pdx1-Cre*, littermate control; ERK1/2: loading control. e, Pancreas/bw ratio in the indicated 12- and 16-wo mice. Mean±SD (n/mice= 20 WT; 6 *Kras*^G12D/PEC; 7 *Kras*^G12D/PEC;*Nrf2*^Act-PEC; 18 *Nrf2*^Act-PEC). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value=0.0448, 0.006, 0.005, <0.0001, ns-not significant). f, Pancreatic areas occupied by ADM and PanIN lesions per high power field (HPF) in the indicated mice. Mean±SD (n/mice= 8 *Kras*^G12D/PEC; 5 *Kras*^G12D/PEC;*Nrf2*^Act-PEC 2–3mo; 7Kras^G12D/PEC;*Nrf2*^Act-PEC 4–5mo). Multiple unpaired t-tests, Holm-Šídák method (P<0.000001). g, Contingency tables of strong, weak/moderate, and negative EZH2 staining frequency among ADMs (top) and PanINs (bottom) of the indicated genotypes (See Fig. 3b, and Extended Data Fig. 2b). h, mRNA amounts of NRF2 targets including *Mdm2,* acinar-identity genes, and ductal and progenitor cell markers in 16-wo pancreata of indicated genotypes. WT: littermate control; Mean± SD or SEM (*Epcam*) (n/mice for WT, *Kras*^G12D/PEC, and *Kras*^G12D/PEC;*Nrf2*^Act-PEC =7, 4, 6 *Nqo1*, 4, 4, 3 *Gstm5*; 6, 5, 4 *Mdm2*; 6, 6, 5 *Amy2a1;* 4, 6, 6 Nr5a2; 5, 5, 6 *Bhlha15*; 7 *Krt19*; 4, 5, 5 *Pdx1*; 6, 5, 8 *Epcam;* 8, 4, 5 *Aldh1*; 5, 7, 6 *Acta2*). One-way ANOVA, Dunnett’s multiple comparisons test or Kruskal Wallis, Dunn’s multiple comparisons test (*Acta2*). Adjusted P value=0.0002, 0.0013, 0.0011, 0.0015, 0.0001, 0.0016, 0.0166, 0.0177, 0.0073, 0.0039, 0.0073, 0.0055, 0,0023, 0.0152, <0.0001, ns-not significant. i, Alcian blue and Sirius red staining (left), and corresponding quantifications (right) of 16-wo pancreata of indicated genotypes, detecting high acidic mucin in PanIN and desmoplasia, respectively. Scale bars: 50μm. Mean±SD (n/fields=10, 20, top; 9, 14, bottom). Unpaired, two-tailed t-tests (P=0.0036, 0.0008). j, Heat maps of differentially expressed NOX1 complex and RAS-MAPK signaling component genes in 8-wo pancreata of the indicated genotypes.

**Extended Data Fig. 5. NRF2 activation converts non-malignant *Kras*^G12D-expressing cells to malignant progenitors and upregulates EZH2, independently of EED.**
a, Experimental scheme depicting conversion of non-malignant *Kras*^G12D/PEC organoids to PDAC progenitors. b, Pancreas/bw ratio 4 wk after orthotopic transplantation of *Kras*^G12D/PEC organoids treated as indicated. Mean±SD (n=5 mice); Two-tailed Mann-Whitney test; P=0.0317. c, H&E, EZH2, Ki67, phospho-ERK1/2, and KRAS^G12D staining (left), and corresponding quantifications (right), of pancreata from (b). Mean±SD (From top to bottom, n/fields=11, 8, 7, 16 *Kras*^G12D/PEC; 13, 11, 8, 13 *Kras*^G12D/PEC+ H₂O₂); two-tailed Mann-Whitney test (Ki67, and P-ERK1/2); unpaired, two-tailed t-tests (EZH2, and KRAS^G12D); P=0.0003, <0.0001. Scale bars, 50μm. d, H&E and Ki67 IHC of sectioned low-passage *Kras*^G12D/PEC organoids treated as indicated (top), and Ki67 quantification (bottom). Mean±SEM (n/fields=13 *Kras*^G12D/PEC ;18 *Kras*^G12D/PEC+H₂O₂); Two-tailed, Mann-Whitney test: P=0.0027. e, EZH2, NQO1, RAS^G12D and αSMA IB of CTRL and H₂O₂ treated *Kras*^G12D/PEC organoids. Mouse fibroblasts served as αSMA positive control. HSP70, tubulin: loading controls. f, Brightfield microscopy (left), and quantification of *Kras*^G12D/PEC organoids (right) untreated or treated for 7 days with H₂O₂ or SFN −/+ 10nM trametinib (Tram) for 72 h before harvesting. Mean±SEM (for each experimental group left to right, n/fields =41, 22, 35, 24, 44, 28, 23, 21, 56, 33, 68, 33); Kruskal Wallis, Dunn’s multiple comparisons test (vs *Kras*^G12D/PEC), two-tailed Mann-Whitney test (Tram vs *Kras*^G12D/PEC H₂O₂ or SFN); adjusted P value= 0.0467, 0.0005, 0.0004, 0.0009, <0.0001, ns-not significant. Scale bars, 100μm. g, mRNA amounts in above organoids. Mean±SEM (n/biological replicates =5, 7, 8, 9 *Ezh2* mRNA7, 8 for *Nfe2l2* mRNA). Kruskal Wallis, Dunn’s multiple comparisons test (vs *Kras*^G12D/PEC), and Two-tailed, Mann-Whitney test (vs *Kras*^G12D/PEC +H₂O₂ or +SFN); adjusted P value=0.0111, 0.0222, 0.0059, 0.0013, 0.0009, 0.0002, <0.0001, ns-not significant. h, Area (top), and mRNA amounts (bottom) of *Kras*^G12D/PEC organoids treated as indicated or transfected with NRF2^Act vector. Mean±SEM (n/organoids replicates=108 +DMSO, 83 +SFN, 18 +NRF2^Act), Kruskal Wallis, Dunn’s multiple comparisons test (top); Mean±SD (n/biological replicates =6 for all experimental groups, and 5, *Nqo1*/+SFN), Two-way ANOVA, Dunnett’s multiple comparisons test (bottom); adjusted P value=0.0090, 0.0030, 0.0179, 0.0017, 0.0004, <0.0001. i, Area (left) and MTT assay (right) of *Kras*^G12D/PEC organoids treated as indicated. MTT signal %: Mean ± SD (n/biological replicates= 4 *Kras*^G12D/PEC, and *Kras*^G12D/PEC+H₂O₂; 5 WT and *Kras*^G12D/PEC+SFN), One-way ANOVA, Dunnett’s multiple comparisons test. Organoid area: Mean ± SEM (n/organoids= 69 *Kras*^G12D/PEC+SFN; 158 *Kras*^G12D/PEC +H2O2; 117 *Kras*^G12D/PEC), Kruskal Wallis, Dunn’s multiple comparisons test. P=0.0371, < 0.0001, ns-not significant. **j-k,** Brightfield microscopy of parental, EZH2^KO and NRF2^KO *Kras*^G12D/PEC organoids treated as indicated (j), and their number and area (k). Mean±SD (each experimental group, from left to right n/fields=34, 28, 20, 17, 19, 12, k, left), Mean±SD (n/organoids= 32, 166, 57, 85, 43, 83, k, right), Kruskal Wallis, Dunn’s multiple comparisons test; adjusted P value=0.0003, 0.0026, 0.0017, 0.0164, <0.0001, ns-not significant. Scale bars, 100μm. l, Brightfield microscopy (top), area (bottom, left), and number (bottom, right) of parental, and EED^KO *Kras*^G12D/PEC organoids treated as indicated. Mean ± SD (for each experimental group, left to right n/organoids=103, 121, 73, 100, bottom left; n/fields= 23, 30, 27, 36, bottom right). Organoid number: One-way ANOVA, Dunnett’s multiple comparisons test; Organoid Area: Kruskal Wallis, Dunn’s multiple comparisons test (adjusted P value= 0.0001, 0.0003, 0.0443, <0.0001, ns-not significant). Scale bars, 100μm. m, EED, acetyl- and 3me-H3K27 (top) and mRNA (bottom) amounts in CTRL and EED^KO *Kras*^G12D/PEC organoids treated as indicated. H3, tubulin: loading controls. Mean±SD (n=8 biological replicates). Kruskal Wallis, Dunn’s multiple comparisons test (adjusted P value=0.0005, 0.0084, 0.0001, 0.0136, <0.0001, ns-not significant). The immunoblot data in the figure are representative of 2–3 independent runs.

**Extended Data Fig. 6. NOX1 is dispensable for organoid transformation although it promotes CAE-induced *in-vivo* transformation.**
a, qRT-PCR analysis of indicated mRNAs in *Kras*^G12D/PEC organoids treated for 7 days with SFN or DMSO. Mean ± SD (n/biological replicates= 3 for *Epcam* DMSO, 4 for *Krt19, Sox9, Epcam* SFN, *Cd8* DMSO and *Adgre1*, 5 for *Lif* SFN, 6 for all the other groups); 2-way ANOVA, Šídák’s multiple comparisons test (*Krt19, Sox9*), two-tailed, Mann-Whitney test *(Cd8, Cxcl12, Lif, Saa3)* and unpaired, two-tailed t-tests (adjusted P value= 0.01000, 0.0002, 0.0091, 0.0126, 0.0173, ns-not significant). **b-c,** Brightfield microscopy of *Kras*^G12D/PEC organoids treated for 7 d with DMSO, SFN or KI696 (b). Scale bars, 100μm. Number, area (c, top), and mRNA of the above organoids (c, bottom). Mean±SD (from left to right n/field=36, 56, 62, c, top-left; n/organoids=43, 20, 61 c, top-right; n/biological replicates=4, 6, 7 for DMSO and 8 for all the other groups, c, bottom). Top: Kruskal Wallis, Dunn’s multiple comparisons test; Bottom: One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value=0.0041, 0.0312, 0.0328, 0.0031, 0.0005, 0.0062, <0.0001). d, Schematic of NAC and CAE treatment of *Kras*^G12D/PEC mice. e, H&E staining (top) and pancreas to body weight ratio (bottom) of *Kras*^G12D/PEC pancreata 2 wk after CAE −/+NAC treatment. Mean±SD (n=4 mice); Unpaired, two-tailed t-tests; P=0.0161. Scale bars, 50μm. f, mRNA amounts in above pancreata. Mean±SEM (n/pancreata=5 for *Ezh2* CAE; 6 for *Ezh2* and *Nqo1* CAE+NAC; 8 for *Nfe2l2* and *Duox1* CAE+NAC, and 7 for all the other groups). Mann-Whitney two-tailed test; *Ezh2,* and *Duox1:* Unpaired, two-tailed t-test (adjusted P value=0.0003, 0.0047, 0.0006, 0.0065, 0.0530, <0.0001). g, Brightfield microscopy of CTRL and NOX1^KD *Kras*^G12D/PEC organoids treated as indicated. Scale bars, 100 μm. h, NOX1 IB of CTRL and NOX1^KD *Kras*^G12D/PEC 2D cultures. HSP70- loading control. i, Numbers (left), and mRNA amounts (right) of *Kras*^G12D/PEC organoids treated as indicated. Mean±SD (n/biological replicates= 4, 5 for DMSO and NOX1^KD+H₂O₂ *Ezh2*, 6 NOX1^KD+H₂O₂ *Nfe2l2,* 8 for NOX1^KD+SFN *Nfe2l2*, and 4 for all the other experimental groups for mRNA; n/organoids for each experimental group left to right = 26, 34, 39, 22, 27, 29); One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value=0.0069, 0.0218, 0.0073, 0.0012, 0.0029, 0.0031, 0.0004, 0.0033, 0.0005, <0.0001; ns-not significant). j, Schematic of NOX1^KD *Kras*^G12D/PEC organoid transplantation and CAE treatment. k, Relative 8-OHdG content of *Kras*^G12D/PEC organoids treated as indicated. Mean ± SD (n=4 biological replicates); One-way ANOVA, Dunnett’s multiple comparisons test, ns-not significant. l, Schematic representation of the deep exome sequencing setup used to detect mutations in *Kras*^G12D/PEC organoids. Created in BioRender. Nandi, S. (2025) https://BioRender.com/h33k361. m, Total number of coding mutations that are unique (red) or shared (blue) between any two samples. n, Stacked bar plot showing total number of mutations in each replicate. Red dot indicates individual sample. The bounds of the boxplots represent the interquartile range divided by the median, and Tukey-style whiskers extend to a maximum of 1.5 × interquartile range beyond the box. The P-value from Wilcoxon rank-sum test between NT (n=4 biological replicates) and treated groups (each group n=4) are indicated. o, Oncoplot following variant effect prediction of all coding missense mutations between the analyzed samples. Gene name and corresponding amino acid changes are indicated to the left. The number of samples without any given missense mutation is indicated to the right. The IBs are representative of 3 *biological replicates and 2 replicate* blots.

**Extended Data Fig. 7. NRF2 recruitment to the *EZH2* promoter establishes a malignancy supportive autoregulatory loop.**
a, Alignment of the WT and the ARE^KO *EZH2* promoter sequences. b, Densitometric quantitation of EZH2, NRF2, NQO1, 3me-H3K27, H3 and HSP70 normalized to ERK1/2, from the IBs in Figure 4a. Arbitrary units (A.U.) are relative to CTRL. Mean±SD (n=4 biological replicates). Two-way ANOVA, Šídák’s multiple comparisons test, P=0.0319 NQO1, 0.0069 EZH2, 0.00393 3meH3K27, <0.0001, ns-not significant. **c-d,** mRNA amounts of ARE-containing (*EZH2, NFE2L2, NQO1*) and ARE-independent (*ACTB* and *GAPDH*) genes (c), and EZH2 and NQO1 IBs (d, top) and their quantitation (d, bottom) in human PDAC 1444 cells with ARE-deleted (ARE^KO) *EZH2* promoter. HSP70: loading control. A.U. normalized to HSP70 relative to CTRL. Mean±SD (n=3 biological replicates). Multiple paired t-tests, Holm-Šídák method (adjusted P value=0.016198, 0.022657, 0.022657, 0.003023, 0.050181, ns-not significant). e, Targeted ChIP probing NRF2 recruitment to the *Ezh2* (left) and *Nqo1* (right) promoters in the indicated PEC preparations. Mean ±SD (n=4 biological replicates). WT: littermate controls. Multiple paired t-tests, Holm-Šídák’s method (adjusted P value=0.021402, 0.005929, 0.016271, 0.018177, 0.003321, ns-not significant). f, MTT assays of CTRL and ARE^KO MIA PaCa-2 cells cultured for the indicated times (days). Mean±SD (n=6 biological replicates). Two-way ANOVA, Šídák’s multiple comparisons test (adjusted P value <0.0001, ns-not significant). g, S.C. tumors (left), and their volumes (right) formed by CTRL and ARE^KO PDX1444 cells. Mean ± SD (n=4 tumors). Paired, two-tailed t-tests P=0.0492. **h-i,** H&E and IHC staining (h) and corresponding quantification of positive area (i) of above tumors. Scale bars, 50μm. Mean ± SD (n =4 AREKO; 7 CTRL/EZH2; 6 CTRL/Ki67). Unpaired, two-tailed t-tests, P=0.0348, 0.0056. j, *EZH2* and *NFE2L2* mRNAs in CTRL and ARE^KO PDX1444 s.c. tumors. Mean±SD (n=4 tumors). Multiple unpaired t-tests, Holm-Šídák method (adjusted P value=0.017463). k, EZH2 and 3me-H3K27 in control and EZH2^KO mouse (right) and human (left) PDAC cells. H3, actin, and ERK1/2: loading controls. l, IB analysis of CTRL and EED^KO UN-KC6141 cells. ERK1/2: loading control. m, S.C. tumors (left), and their relative volume and weight (right) formed by parental and EZH2^KO UN-KC6141 cells. Mean±SEM (n=10 CTRL, 9 EZH2KO/weight, 8 EZH2KO/volume). Two-tailed Mann-Whitney test (adjusted P value <0.0001). n, EZH2 in above tumors. H3: loading control. o, S.C. tumors (left), and their relative volume (right) formed by CTRL and EED^KO UN-KC6141 cells. Mean±SEM (n=8 CTRL, 6 EED^KO tumors). Unpaired, two-tailed t-tests, ns-not significant. p, MTT assays of CTRL, EZH2^KO, and EED^KO UN-KC6141 cells. Mean±SD (n=7 biological replicates). Two-way ANOVA, Dunnett’s multiple comparisons test, P<0.0001, ns-not significant. q, Sphere formation by CTRL and EZH2^KO mouse UN-KC6141, and human MIA PaCa-2 cells. Mean±SD (n=15, left; n=5 CTRL, 6 EZH2^KO HMF right). Two-tailed Mann-Whitney test (adjusted P value = 0.0043, <0.0001). r, mRNA amounts in CTRL and EZH2^KO human and mouse PDAC cells. *Nfe2l2* and *Nqo1*: Mean±SD (n=7 and 6 CTRL, and 9 EZH2^KO biological replicates), Mann-Whitney test. *NFE2L2*, and *NQO1*: Mean±SD (n=6 biological replicates), unpaired two-tailed t-tests (adjusted P value =0.0002, <0.0001). s, mRNA amounts in CTRL, EZH2^KO and EED^KO UN-KC6141 cells. Mean±SD (n=4 biological replicates). Two-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value =0.0299 <0.0001, ns-not significant). IBs are representative of 3–4 *biological replicates*.

**Extended Data Fig. 8. The NRF2-EZH2 autoregulatory loop controls malignancy-supportive genes.**
a, Principal component analysis of transcriptomic relationships between 8-wo WT and *Kras*^G12D/PEC pancreata (n/pancreata=3 for *Kras* CTRL 3d, 4 for WT CAE 7d, 2 for all the other experimental groups) 3 and 7 days after CAE administration. b, Principal component analysis of transcriptomic relationships between pancreata (n=3 mice) from the indicated 8-wo mice. c, Dataset enrichment graphs of RNA-seq data from 8-wo *Kras*^G12D/PEC/Nrf2^Act-PEC vs *Nrf2*^Act-PEC (top); *Nrf2*^Act-PEC vs WT (center); *Kras*^G12D/PEC vs WT (bottom) pancreata. The latter comparison highlights the absence of major deviations from the WT transcriptome in young, healthy *Kras*^G12D/PEC pancreata. d, Dataset enrichment graphs of RNA-seq data from 8-wo *Kras*^G12D/PEC pancreata treated −/+ CAE for 3 days (n/pancreata=3 for *Kras* CTRL 3d; 2 *Kras* CAE 3d). e, Volcano plot representation of differentially expressed genes (DEG) in CTRL vs EZH2^KO UN-KC6141 cells. Red and blue points mark significantly elevated or diminished DEG, respectively (n=3 biological replicates). f, mRNA amounts in s.c. tumors formed by CTRL and EZH2^KO UN-KC6141 cells. Mean±SEM(n=4 tumors *Amy2a1, Bhlha15,* CTRL *Ptf1a*, EZH2^KO *Krt19; 5* tumors *Nr5a2,* EZH2^KO *Ptf1a*; 6 tumors *Cdkn2a*); Two-tailed unpaired t-tests, two-tailed Mann-Whitney test (*Cdkn2a* mRNA); P=0.0103 *Amy2a1*, 0.0214 *Ptf1a*, 0.0274 *Bhlha15*; 0.0014 *Krt19,* 0.0022 *Cdkn2a;* 0.0009 *Nr5a2*. g, Genome browser representation of ATAC-seq (top), 3meH3K27 ChIP-seq (middle), and H3K27ac ChIP-seq (bottom) tracks covering the indicated promoters in CTRL and EZH2^KO UN-KC6141 cells. h, Targeted ChIP of repression (3meH3K27) and activation (acH3K27 and acH3K14) marks (represented as ratio of modified to unmodified H3, top panels), p300, POLII (middle panels) and EZH2 recruitment to the *NR5A2/Nr5a2* promoter in CTRL and EZH2^KO human (top, left) and mouse (bottom) cells. Mean±SD (n/biological replicates= 4 EZH2, p300 ChIP; 6 H3 ChIP; 3 POLII ChIP). Top-left: Two-way ANOVA, Šídák’s multiple comparisons test; 3meH3K27/H3 enrichment: two-tailed unpaired t-tests. Multiple unpaired, t-tests, Holm-Šídák method and Welch correlation (*Nr5a2* promoter). Adjusted P value = 0.0008, 0.004090, 0.000006, 0.005101, 0.017070, 0.040053, 0.025986, <0.0001, ns- not significant. The primer sets used to assess EZH2 recruitment are depicted at the bottom. i, Targeted ChIP of EZH2 recruitment and 3me-H3K27 enrichment at the *Nr5a2* promoter in primary PEC from 12-wo mice of indicated genotypes. Upper panel: Mean±SD (n=4 biological replicates). Bottom panel: Mean±SEM (n=3 biological replicates). Two-way ANOVA, Šídák’s (top) and Holm-Šídák’s (bottom) multiple comparisons tests (adjusted P value =0.006, 0.0262, 0.0214, <0.0001. ns- not significant). WT: littermate controls. j, Correlation between *EZH2* mRNA amounts and the depicted mRNAs in human PDAC TCGA data (n=147). Significantly correlated genes with FDR-correlated Spearman P values <0.05 or −LogBH p value >1.3 are indicated by the bars on the right. k, Frequency of DBA⁺, and CD133⁺ cells in CTRL and EZH2^KO UN-KC6141 cells. Mean±SEM (n=3 biological replicates). Unpaired, two-tailed t-tests, P<0.0001. Gating strategy are included in Supplementary Fig.1.

**Extended Data Fig. 9. PRC2 catalytic activity is not required for EZH2-mediated NRF2 upregulation and tumor growth.**
a, Schematic of co-IP experiments probing the residual amounts of EED-free solo-EZH2. b, Serial IP-IB (bottom) and total level (IN;Input, top) of EZH2 and EED in CTRL, EZH2^KO and EED^KO KC6141 cells. First IP: anti-EED or IgG. The supernatant after 1^st IP was subjected to 2^nd IP with anti-EZH2. ERK1/2: loading control. c, Co-IP of STAT3 and RELA/p65 (right), PPARγ and E2F1 (left), and POLII and EED (bottom) with EZH2 in parental and EZH2^KO KC6141 cells. d, Schematic depiction of WT and catalytically dead EZH2^ΔSET proteins. e, IB analysis of EZH2^KO KC6141 cells transduced with Myc-tagged Empty, EZH2^WT or EZH2^ΔSET vectors (left) and *Ezh2* and *Nfe2l2* mRNAs in same cells (right). Mean±SD (n=6 *Ezh2,* 8 *Nfe2l2,* biological replicates). Two-way ANOVA, Dunnett’s multiple comparisons test, and Kruskal Wallis, Dunn’s multiple comparisons test (*Nfe2l2* mRNA); adjusted P value = 0.0012, 0.0015, <0.0001, ns- not significant. f, H&E and IHC staining of s.c. tumors formed by EZH2^KO KC6141 cells reconstituted as above, and corresponding quantification of positively stained areas. Scale bars, 50μm. Mean±SD (Corresponding n/fields for EZH2^KO+Empty, EZH2^KO+ EZH2^WT, EZH2^KO+EZH2^ΔSET n= 4,5,7 EZH2 IHC; 14,15,22 Ki67 IHC; 6,9,11 NQO1 IHC), One-way ANOVA, Dunnett’s multiple comparisons test, and Kruskal Wallis, Dunn’s multiple comparisons test (Ki67 quantification); adjusted P value = 0.0016,0.0026, 0.0015, 0.0024, <0.0001, ns- not significant. g, Principal component analysis of transcriptomic relationships between CTRL, EZH2^KO, and EED^KO UN-KC6141 cells (n=2 biological replicates). h, Volcano plot of DEG between EZH2^KO and EED^KO KC6141 cells. Red and blue points mark DEG with significantly elevated or diminished expression, respectively (n=2 biological replicates). i, Venn diagram showing comparative analysis of up and down regulated genes in EZH2^KO vs CTRL, and EED^KO vs CTRL KC cells (n=2 each). The IBs are representative of *2 replicate*s.

**Extended Data Fig. 10. EZH2 activates lipogenic genes in mouse and human PDAC.**
a, Genome browser representation of ATAC-seq (top), and H3K27ac ChIP-seq (bottom) tracks covering the indicated genomic regions in CTRL and EZH2^KO KC6141 cells. b, mRNAs related to lipid and cholesterol metabolism in ductal (DBA⁺) or EPCAM⁺ cells isolated from 12-wo pancreata of the indicated genotypes. Mean±SEM (n=3, and n=4, *Srebp2, Acaca, Fasn, Hmgcr,* biological replicates in DBA⁺ cells). Two-way ANOVA, Šídák’s multiple comparisons test, adjusted P value = 0.000110, 0.000005, 0.000044, 0.000515, 0.000085, 0.038788, 0.027582, 0.000724, 0.014170, 0.038788, 0.000873, 0.038788. c, Computational analysis of TCGA data comparing expression of indicated mRNAs between the top and bottom quartiles of human PDACs based on *EZH2* mRNA amounts. d, Mass isotopomer distribution (partial) for palmitate in KC6141 cells incubated for 48 hours with ¹³C-glucose and 500nM Fasnall for 24 hours. Mean±SEM (n=3 biological replicates). Multiple unpaired, t-tests, Holm-Šídák method, adjusted P value = 0.022641, 0.009106, 0.042924, 0.001377, 0.017274, 0.037656, ns-not significant. e, 3me-H3K27 IB of MIA PaCa-2 cells treated for 96 hours with 1μM EPZ-6438. H3 and actin: loading controls. The IB is representative of 2 *biological replicates*. **f-g,** MTT assays of KC6141 (f) and MIA PaCa-2 (g) cells treated with Fasnall. Mean±SD (n/biological replicates=4, and 8 KC6141 0.0 μM FASNALL, n=5 Mia PaCa-2 0 FASNALL, 8 for 1 μM, and 7 for all the other groups). One-way ANOVA, Holm-Šídák’s multiple comparisons test, adjusted P value <0.0001. h, MTT assays of CTRL and EZH2^KO UN-KC6141 (left) and MIA PaCa-2 (right) cells treated for 72 hours as indicated. Mean±SD (n/biological replicates=7 CTRL KC6141, 5 EZH2^KO and Mia PaCa-2 CTRL FASNALL, 4 Mia PaCa-2 CTRL DMSO). Kruskal Wallis, Dunn’s multiple comparisons test; adjusted P value =0.0001, 0.0005, 0.0179, 0.0020, ns- not significant.

**Fig. 1:. Enduring NRF2 and EZH2 induction in CAE-challenged *Kras*^G12D/PEC pancreata.**
a, 3me-H3K27 IB, and NRF2 and EZH2 proteins (top) and mRNAs (bottom) in 8-wo CAE-treated WT and *Kras*^G12D/PEC pancreata. ERK1/2, HSP70, H3: loading controls. mRNA amounts are expressed relative to WT CTRL d3. Mean±SEM (n= 4, 8 WT CAE d7, 6 *Kras*^G12D/PEC CTRL d3 mice). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value=0.0291, 0.0008, 0.0028 vs WT CTRL d3; 0.0104, 0.0007, 0.0002, vs *Kras*^G12D/PEC CTRL d3; ns-not significant). b, H&E and IHC staining of above pancreata. c, H&E staining of 8-wo WT pancreata at d3 after last CAE injection. d, NRF2 inhibition (ML385) and haploinsufficiency (*Nrf2*^+/−) attenuate CAE induction of NQO1, NRF2 and EZH2 proteins (top) and mRNAs (top-right, bottom) in WT mice. Tubulin, H3: loading controls. Mean±SEM (*Ezh2* WT CAE, CTRL, ML385, CAE+ML385 n=3; *Nfe2l2* CAE n=5; all the other groups n=4 mice). Bottom panels, One-way ANOVA, Dunnett’s multiple comparisons test; top-right, Two-tailed t-tests (adjusted P value= 0.0406, 0.003, 0.0036, 0.0006, 0.0002, < 0.0001, ns-not significant). Scale bars, 50 μm. IBs are representative of 2–3 independent runs.

**Fig. 2:. NRF2 activation accelerates malignant progression.**
**a-c**, H&E (a), *Ezh2* and *Nfe2l2* mRNAs (b) and IHC (c) in 8-mo WT and *Kras*^G12D/PEC pancreata treated with ML385 4–6 wks post CAE injections. *Ezh2* mean±SEM (WT CTRL n=4 mice; WT CAE n=5 mice; WT CAE+ML385 n=6 mice; *Kras* CTRL, *Kras* CAE, *Kras* CAE+ML385 n=4 mice); *Nfe2l2* mean±SEM (WT CTRL n=4 mice; WT CAE n=4 mice; WT CAE+ML385, Kras CTRL, *Kras* CAE, *Kras* CAE+ML385 n=3 mice). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value=0.0002, <0.0001, ns-not significant). d, H&E stained 8- and 16-wo pancreata of indicated genotypes. e, Kaplan-Meier survival analysis (n=15 *Kras*^G12D/PEC, 21 *Kras*^G12D/PEC/Nrf2^Act-PEC mice). Log-rank (MantelCox) test, p<0.0001. Scale bars in 2a and 2d, 50 μm.

**Fig. 3:. NRF2 induces EZH2 expression during PanIN to PDAC progression.**
a, IHC (left) and quantification (right) of 16-wo *Kras*^G12D/PEC and *Kras*^G12D/PEC/Nrf2^Act-PEC pancreata. Mean±SD (From top to bottom panels, n=17, 14, 21, 28, 15, 16, 8 fields for *Kras*^G12D/PEC; n= 24, 19, 26, 30, 21, 4, 5, 9, 10 fields for *Kras*^G12D/PEC/Nrf2^Act-PEC). Unpaired, two-tailed t-tests (P value=0.0001, 0.0186, <0.0001). Scale bars, 50 μm. b, Graphical presentation of strong, weak/moderate, or negative EZH2 staining frequency of ADMs (top, n=76, 82 fields) and PanINs (bottom, 47, 69 fields) in 16-wo mice. Contingency Chi-square test for trend. p<0.0001 (see Extended Data Fig. 2b and 4g). **c,d,** EZH2, NRF2 and NQO1 (c) and *Ezh2* and *Nfe2l2* mRNAs (d) in 8-wo pancreata of indicated genotypes. H3, tubulin: loading controls. Mean±SEM (n=3 mice for *Nfe2l2;* n=4, 8, 6, 7 mice for *Ezh2,* left to right). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value <0.0001, ns-not significant). Scale bars: 50 μm. IBs are representative of 3 independent runs.

**Fig. 4:. The NRF2-EZH2 axis converts non-malignant *Kras*^G12D organoids to malignant progenitors.**
a, EZH2 and 3me-H3K27 IB of CTRL and EZH2^KO *Kras*^G12D/PEC 2D cultures. ERK1/2, HSP70: loading controls. **b-c,** Brightfield microscopy (b), number (c, left), area (c, center) and mRNA quantification (c, right) of low-passage CTRL and EZH2^KO *Kras*^G12D/PEC organoids treated −/+ H₂O₂ (150 mM, for 7 days). Scale bars, 100 μm. Left panel: Mean±SD (n=14, 13, 7, 6 visual fields for each group, left to right), One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value =0.0027, <0.0001). Center panel: Mean±SEM (n= 20, 156, 9, 22 organoids, for each group, left to right), Kruskal Wallis test (adjusted P value =0.0117, 0.0069, 0.0166). Right panel: Mean±SD (n=3 biological replicates); Two-way ANOVA, Tukey’s multiple comparisons test (adjusted P value =0.0002, 0.0182, <0.0001, ns-not significant). d, H&E and IHC staining of pancreata transplanted with CTRL and EZH2^KO organoids treated as in (b). After 4 wk, EZH2^KO organoids generated one small tumor on which the images were focused (d, bottom). Scale bars, 100 μm (H&E) and 50 μm (IHC). e, Pancreas/bw ratio, and EZH2 and Ki67 positive areas in the IHC from (d). Mean±SD (n= 5, 6 mice, left panel; *Kras*^G12D/PEC+H₂O₂ n= 22, 14, *Kras*^G12D/PEc; EZH2^KO +H₂O₂ n= 24, 31 fields, center and right panel). Two-tailed, Mann-Whitney test (P value =0.0173, <0.0001). **f,g,** EZH2, NRF2, and NQO1 IB (f) and mRNA amounts (g) in above pancreata. HSP70: loading control; Mean±SD (n=5; *Kras*^G12D/PEC+H₂O₂ *Ezh2* n=6 mice). Multiple unpaired, two-tailed t-tests, and Mann–Whitney test (*Ezh2* mRNA) (P value =0.0043, 0.0077, 0.0110, 0.0011).

**Fig. 5:. NRF2 activation but not NOX1 is essential for *Kras*^G12D organoids transformation.**
a, Brightfield microscopy (left) and number (right) of *Kras*^G12D/PEC organoids treated (7 d) with sulforaphane (SFN) or DMSO (top) or formed by *Kras*^G12D/PEC PEC transfected with empty or NRF2^E79Q (NRF2^Act) vectors (bottom). Scale bars, 100 μm. Mean±SD (n=6 biological replicates); Multiple unpaired, two-tailed t-tests (P value =0.0026). b, H&E and IHC of pancreata 4 wk after transplantation of CTRL or NRF2^KO organoids treated as indicated. Scale bars, 50 μm. c, Pancreas/bw ratio, and EZH2 and NRF2 positive areas (top) and mRNA amounts (bottom) in above pancreata. Mean±SD (from left to right, n=11,17, 7 fields for EZH2 IHC; n= 13, 16, 7 fields for NRF2 IHC; Pancreas/BW, and bottom panels n=4 mice). One-way ANOVA, Dunnett’s multiple comparisons test (Top left, center, adjusted P value =0.0061, 0.0365, <0.0001; bottom adjusted P value = 0.0110, 0.0022, 0.0012, 0.0002, 0.02, 0.0055, 0.0027). Top-right panel: Kruskal Wallis test (adjusted P value =0.0003, 0.0107). d, H&E and IHC of pancreata transplanted with CTRL or NOX1^KD *Kras*^G12D/PEC organoids, 4 wk post CAE. Scale bars, 50 μm. e, Pancreas/bw ratio, and EZH2 and Ki67 IHC positive areas (top), and mRNA amounts (bottom) in pancreata from (d). Mean±SD (for each experimental group, left to right, n=13, 23, 11 fields for EZH2 IHC; 15, 23, 12, fields for NRF2 IHC; top-left panel n=4, 5, 6 mice); Mean±SEM (for each experimental group, left to right, bottom mRNA n=4, 9, 6 pancreata for *Ezh2* mRNA; n=7, 8, 6, pancreata for *Nfe2l2* mRNA; n=4, 4, 6 pancreata for *Nqo1, Mox1, Duox2*). One-way ANOVA, Dunnett’s multiple comparisons test; Top-right and bottom-left panels: Kruskal Wallis test (adjusted P value = 0.0014, 0.0018, 0.0017, 0.023, 0.0238, 0.0012, 0.0046, 0.0014, 0.0002, 0.0001, 0.0057, <0.0001, ns-not significant). IB are representative of 3 biological replicates and 2 independent runs.

**Fig. 6:. The malignancy sustaining NRF2-EZH2 loop is maintained by a single ARE in the *EZH2* promoter.**
a, Top, *EZH2* and *NFE2L2* mRNAs, and EZH2, 3me-H3K27, NRF2 and NQO1 proteins in ARE^KO MIA PaCa-2 cells with ARE-disrupted *EZH2* promoter. Bottom, schematic description and primer locations. Green arrows: primers encompassing the *EZH2* promoter region; red arrows: primers encompassing the *EZH2* 3’ flanking region (negative control). H3, HSP70, ERK1/2: loading controls. Mean±SD (n=3, biological replicates). Multiple unpaired two-sample t-tests, Holm-Šídák method (adjusted P value =0.028081, 0.013635). b, Targeted ChIP probing NRF2, small MAF and POLII recruitment to the *EZH2* promoter and 3’ flanking region and *NQO1* promoter (positive control) in CTRL and ARE^KO MIA PaCa-2 cells. Mean±SD (n=4 biological replicates). Two-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value = 0.0288, 0.0454, <0.0001; ns-not significant). **c,d,** Subcutaneous tumors formed by CTRL and ARE^KO MIA PaCa-2 cells (c) and their relative volume and weight (d). Mean±SEM (CTRL n=5 tumors; ARE^KO n=8 tumors). Two-tailed Mann-Whitney test (P value = 0.0085). **e,f** H&E and IHC of above tumors (e) and quantification of EZH2 and Ki67 positive cells, and NQO1, NRF2, and 3meH3K27 positive areas (f). Scale bars, 50 μm. Mean±SD (n=7, 22,13, 24 fields for CTRL; n= 6, 4, 5, 20 fields for ARE^KO). Unpaired, two-tailed t-tests (top), and two-tailed Mann-Whitney test (center/bottom left) (P value =0.0001, 0.0231, 0.0185, <0.0001). g, *EZH2* and *NFE2L2* mRNAs in above tumors. Mean±SD (n=4 ARE^KO, 5 CTRL tumors). Two-way ANOVA, Šídák’s multiple comparisons test (P value =0.0405, 0.0036). h, EZH2 recruitment to the *NFE2L2* promoter in CTRL and EZH2^KO cells. Mean±SEM (n=5 IgG, 6 anti-EZH2 technical replicates). Two-way ANOVA, Šídák’s multiple comparisons test (P value =0.0002). i, Genome browser ATAC-seq (top) and H3K27ac ChIP-seq (bottom) tracks covering the *Nfe2l2* promoter (top) and enhancer (bottom) in CTRL and EZH2^KO KC6141 cells. j, The *Nfe2l2* promoter and enhancer regions with AP-1, NRF2, E2F1 and NF-κB binding sites marked in green, yellow, blue, and purple, respectively. k, TF recruitment to the *Nfe2l2* promoter in CTRL and EZH2^KO KC6141 cells. Mean±SD (n=3 left panel, 4 right panel, biological replicates). Two-way ANOVA, Šídák’s multiple comparisons test (adjusted P value = 0.0105, <0.0001, ns-not significant). IBs are representative of 3 biological replicates and 2 independent runs.

**Fig. 7:. The NRF2-EZH2 loop cooperates with KRAS to activate cancer supportive genes and suppress acinar identity genes.**
a, Differentially expressed genes (DEG) in 8-wo WT and *Kras*^G12D/PEC pancreata (n/pancreata=3 for *Kras* CTRL 3d, 4 for WT CAE 7d, 2 for all the other experimental groups) 3 and 7 d after CAE or CTRL injections. Vertical rows-different specimens. Horizontal rows-individual genes, colored according to log-transformed transcript intensity in z-scored units. Days=d, TPM=transcripts per million. Green: replicates with low expression (z-score = −2); red: replicates with high expression (z-score = +2). **b,c,** RNA-seq dataset enrichment graphs of 8-wo *Kras*^G12D/PEC/Nrf2^Act-PEC vs *Kras*^G12D/PEC pancreata (n=3 pancreata) (b), and 8-wo 7 d CAE vs CTRL, *Kras*^G12D/PEC pancreata (n=2, pancreata) (c). d, Venn diagram showing integrative analysis of hallmark DEG sets from (a-c). e, RNA-seq dataset enrichment graphs of CTRL vs EZH2^KO KC cells (n=3 biological replicates). f, Venn diagram showing comparative analysis of genes with lower H3K27ac signal and downregulated genes in EZH2^KO vs CTRL KC cells (n=3, biological replicates) (left) and topmost significant terms corresponding to genes with reduced H3K27ac signal (right).

**Fig. 8:. EZH2 activates PDAC anabolic genes.**
a, Volume (left) of s.c. tumors (right) formed by EZH2^KO KC6141 cells transduced with Empty, EZH2^WT or EZH2^ΔSET vectors. Mean±SEM (EZH2^KO +Empty, n=8 tumors, EZH2^KO+ EZH2^WT or EZH2^ΔSET, n=4 tumors). Kruskal Wallis, Dunn’s multiple comparisons test (adjusted P value = 0.0191, 0.0066, ns - not significant). b, mRNA amounts in above tumors. Mean±SEM (n/tumors=3 center, 4 left and right panels). One-way ANOVA, Dunnett’s multiple comparisons test (adjusted P value = 0.0021, 0.0195, 0.0014, 0.0013, 0.0236, 0.0133, 0.0018, 0.0003, 0.0009,0.0013, 0.0141, 0.0128, 0.0019, 0.0320, <0.0001, ns-not significant). c, Venn diagram showing integrative analysis of hallmark DEG sets from CTRL vs EED^KO, and CTRL vs EZH2^KO KC cells. d, Plot showing metabolic genes that are downregulated in EZH2^KO vs CTRL KC cells (Log₂FC) identified by RNA-seq (left), and corresponding H3K27ac ChIP-seq (right). e, mRNA amounts in KC cells (top), and SFN or DMSO treated *Kras*^G12D/PEC organoids (bottom). Mean±SD (n/biological replicates =3 CTRL, 6 EZH2^KO top panel; each experimental group from left to right, n/biological replicates = 9, 10, 8, 6, 6 *Fasn* mRNA; n=8, 9, 6, 7, 6 for *Hmgcr*; n=7, 6, 4, 5, 5 for *Cpt1a*; n=8, 9, 9, 7, 9 for *Srebp2*). Multiple unpaired, two-sample t-tests, Holm-Šídák method (top); One-way ANOVA, Dunnett’s multiple comparisons test (bottom), and Kruskal Wallis, Dunn’s multiple comparisons test (*Cpt1a* mRNA bottom) (adjusted P value = 0.0002, 0.0001, 0.0008, 0.0213, 0.0182, 0.0054, 0.0336, 0.0456, 0.0045, 0.0046, 0.0010, 0.0007, <0.0001, ns-not significant). f, Fatty acid (FA) (C18:1a=oleic, C18:1b=cis-vaccenic) amounts in *Kras*^G12D/PEC organoids treated for 4 days −/+ H₂O₂ (top), and parental and EZH2^KO KC cells (bottom). Mean±SD (n=4 biological replicates, top), Mean±SEM (bottom) (n=3, biological replicates). Two-way ANOVA, Šídák’s multiple comparisons test (top), and Multiple unpaired, t-tests, Holm-Šídák method (bottom) (adjusted P value = 0.0001, 0.0049, 0.0184, 0.0055, 0.0317, 0.0089, 0.001539, 0.002420, 0.027804, 0.007272, 0.008731, 0.0273552, 0.008748, 0.020056, ns-not significant). g, Brightfield microscopy (left) of *Kras*^G12D/PEC organoids treated for 7 d with DMSO or SFN −/+ 2.5 μM EPZ6438 or 500 nM Fasnall, and organoid number (right). Scale bars, 100 μm. Mean±SD (from left to right each experimental group n/biological replicates = 18, 17, 6, 15), Kruskal Wallis, Dunn’s multiple comparisons test (adjusted P value =0.0279, 0.0002, 0.0067, ns-not significant).

See this image and copyright information in PMC

References

1. Hayes JD, Dinkova-Kostova AT & Tew KD Oxidative Stress in Cancer. Cancer Cell 38, 167–197 (2020). - PMC - PubMed
1. Principe DR et al. The Current Treatment Paradigm for Pancreatic Ductal Adenocarcinoma and Barriers to Therapeutic Efficacy. Front Oncol 11, 688377 (2021). - PMC - PubMed
1. Xu Y, Liu J, Nipper M & Wang P Ductal vs. acinar? Recent insights into identifying cell lineage of pancreatic ductal adenocarcinoma. Ann Pancreat Cancer 2 (2019). - PMC - PubMed
1. Kopp JL et al. Identification of Sox9-dependent acinar-to-ductal reprogramming as the principal mechanism for initiation of pancreatic ductal adenocarcinoma. Cancer Cell 22, 737–750 (2012). - PMC - PubMed
1. Hingorani SR et al. Preinvasive and invasive ductal pancreatic cancer and its early detection in the mouse. Cancer Cell 4, 437–450 (2003). - PubMed

MeSH terms

Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions

Substances

Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions

Grants and funding

LinkOut - more resources

Full Text Sources
- Nature Publishing Group
- PubMed Central
Medical
- MedlinePlus Health Information
Molecular Biology Databases
- Mouse Genome Informatics (MGI)
Miscellaneous
- NCI CPTAC Assay Portal

Save citation to file

Email citation

Add to Collections

Add to My Bibliography

Your saved search

Create a file for external citation management software

Your RSS Feed

Self-amplifying NRF2-EZH2 epigenetic loop converts KRAS-initiated progenitors to invasive pancreatic cancer

Affiliations

Self-amplifying NRF2-EZH2 epigenetic loop converts KRAS-initiated progenitors to invasive pancreatic cancer

Authors

Affiliations

Abstract

Conflict of interest statement

Figures

References

MeSH terms

Substances

Grants and funding

LinkOut - more resources

Full Text Sources

Medical

Molecular Biology Databases

Miscellaneous