Article Text

Download PDFPDF

Original article
Loss of Setd2 promotes Kras-induced acinar-to-ductal metaplasia and epithelia–mesenchymal transition during pancreatic carcinogenesis
  1. Ningning Niu1,
  2. Ping Lu1,
  3. Yanlin Yang1,
  4. Ruizhe He2,
  5. Li Zhang1,
  6. Juanjuan Shi1,
  7. Jinghua Wu1,
  8. Minwei Yang2,
  9. Zhi-Gang Zhang3,
  10. Li-Wei Wang4,
  11. Wei-Qiang Gao1,5,
  12. Aida Habtezion6,
  13. Gary Guishan Xiao7,
  14. Yongwei Sun2,
  15. Li Li1,5,
  16. Jing Xue1
  1. 1State Key Laboratory of Oncogenes and Related Genes, Renji-Med X Clinical Stem Cell Research Center, Shanghai Cancer Institute, Shanghai Jiao Tong University School of Medicine Affiliated Renji Hospital, Shanghai, China
  2. 2Department of Biliary-Pancreatic Surgery, Shanghai Jiao Tong University School of Medicine Affiliated Renji Hospital, Shanghai, China
  3. 3State Key Laboratory of Oncogenes and Related Genes, Shanghai Cancer Institute, Shanghai Jiao Tong University School of Medicine, Shanghai, China
  4. 4Department of Oncology, Shanghai Jiao Tong University School of Medicine Affiliated Renji Hospital, Shanghai, China
  5. 5School of Biomedical Engineering and Med-X Research Institute, Shanghai Jiao Tong University, Shanghai, China
  6. 6Division of Gastroenterology and Hepatology, Stanford University School of Medicine, Stanford, California, USA
  7. 7School of pharmaceutical Science and Technology, Dalian University of Technology, Dalian, China
  1. Correspondence to Dr Yongwei Sun, Department of Biliary-Pancreatic Surgery, Renji Hospital, School of Medicine, Shanghai Jiao Tong University, 160 Pujian Rd, Shanghai 200127, China; syw0616{at}126.com, Dr Li Li, State Key Laboratory of Oncogenes and Related Genes, Renji-Med X Clinical Stem Cell Research Center, Ren Ji Hospital, School of Biomedical Engineering & Med-X Research Institute, Shanghai Jiao Tong University, 1954 Huashan Rd, Shanghai, China; lil{at}sjtu.edu.cn and Jing Xue, State Key Laboratory of Oncogenes and Related Genes, Stem Cell Research Center, Ren Ji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai 200127, China; xuejing0904{at}126.com

Abstract

Objective SETD2, the sole histone H3K36 trimethyltransferase, is frequently mutated or deleted in human cancer, including pancreatic ductal adenocarcinoma (PDAC). However, whether SETD2/H3K36me3 alteration results in PDAC remains largely unknown.

Design TCGA(PAAD) public database and PDAC tissue array with SETD2/H3K36me3 staining were used to investigate the clinical relevance of SETD2 in PDAC. Furthermore, to define the role of SETD2 in the carcinogenesis of PDAC, we crossed conditional Setd2 knockout mice (PdxcreSetd2flox/flox) together with KrasG12D mice. Moreover, to examine the role of SETD2 after ductal metaplasia, Crisp/cas9 was used to deplete Setd2 in PDAC cells. RNA-seq and H3K36me3 ChIP-seq were performed to uncover the mechanism.

Results SETD2 mutant/low expression was correlated with poor prognosis in patients with PDAC. Next, we found that Setd2 acted as a putative tumour suppressor in Kras-driven pancreatic carcinogenesis. Mechanistically, Setd2 loss in acinar cells facilitated Kras-induced acinar-to-ductal reprogramming, mainly through epigenetic dysregulation of Fbxw7. Moreover, Setd2 ablation in pancreatic cancer cells enhanced epithelia–mesenchymal transition (EMT) through impaired epigenetic regulation of Ctnna1. In addition, Setd2 deficiency led to sustained Akt activation via inherent extracellular matrix (ECM) production, which would favour their metastasis.

Conclusion Together, our findings highlight the function of SETD2 during pancreatic carcinogenesis, which would advance our understanding of epigenetic dysregulation in PDAC. Moreover, it may also pave the way for development of targeted, patients-tailored therapies for PDAC patients with SETD2 deficiency.

  • pancreatic cancer
  • cancer genetics
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known on this subject?

  • SETD2/H3K36me3 with tumour suppressive function is implicated in several tumours.

  • SETD2, the sole histone H3K36 trimethyltransferase, is frequently mutated or deleted in human pancreatic cancer.

What are the new findings?

  • SETD2 deficiency in human pancreatic ductal adenocarcinoma (PDAC) is linked to poor prognosis.

  • Loss of Setd2 facilitates oncogenic Kras-induced acinar-to-ductal reprogramming.

  • Setd2 is a barrier to Kras driven epithelia–mesenchymal transition (EMT) in PDAC cells.

How might it impact on clinical practice in the foreseeable future?

  • Our findings pave the way for development of targeted, patients-tailored therapies for PDAC patients with SETD2 deficiency.

Introduction

Pancreatic ductal adenocarcinoma (PDAC) is a lethal malignancy with a poor prognosis, largely due to late diagnosis and its intrinsic high metastatic feature.1 Consequently, understanding the molecular mechanism of tumour initiation and metastasis is imperative for the development of novel therapeutic strategies for PDAC.1

Genome sequencing of pancreatic cancer revealed four major driving mutations, including KRAS, TP53, CDKN2A and SMAD4,2 of which about 90% of patients with PDAC harboured KRAS mutations.3 The significance of KRAS mutations for PDAC initiation and maintenance has been well demonstrated in mouse models.4 Mouse models expressing mutant Kras in exocrine compartment of pancreas (acini) usually develop distinct precursor lesion known as pancreatic intraepithelial neoplasia (PanIN) and then progress to metastatic lesion in the context of additional genetic and epigenetic mutations (e.g., TP53 and KDM6A) as well as environmental insults (eg, inflammation).4–6

Based on current knowledge, epigenetic regulator SETD2 is the sole enzyme responsible for trimethylation of lysine 36 on histone 3 (H3K36me3), a histone mark related to actively transcribed regions.7 SETD2 is frequently mutated or deleted in leukaemia and various solid tumours including renal cell carcinoma,8 glioma,7 as well as colorectal and breast cancer.9 SETD2/H3K36me3 with tumour suppressive function is implicated in many aspects, including DNA repair, transcriptional initiation and elongation and alternative mRNA splicing.9–11 Moreover, evidence from human pancreatic cancer genome sequencing has linked SETD2 to pancreatic oncogenesis; however, its causal role in pancreatic cancer remains unknown.

Accumulating evidence demonstrates that PanIN lesion is mainly derived from pancreatic acinar cells by undergoing acinar-to-ductal metaplasia.12–14 Following Kras activation or tissue injury (e.g., pancreatitis), acinar cells would transdifferentiate into a ductal-like morphology. In conditions of tissue injury, these ductal-like cells rapidly redifferentiate back into acinar compartments for tissue regeneration. Oncogenic Kras initiates ADM but hijacks the healing progression by blocking redifferentiation and then promotes PanIN formation.15 Developmental signalling pathways are activated and tightly controlled during acinar-to-ductal reprograming during tissue regeneration. Additionally, defined transcriptional factors (e.g., Ptf1a, Mist1, Nr5a2 and Pdx1) maintained acinar cell integrity, and impaired mutant Kras-induced ADM has been described before.16–18 Thus, dysregulation of genes related to developmental pathways or acinar homeostasis maintenance would promote Kras-induced ADM and PanIN formation. Metastasis to distant organs is a major cause of death in patients with PDAC. Like most solid tumours, epithelia–mesenchymal transition (EMT), characterised as loss of epithelia identity (E-cadherin expression) is the first step for pancreatic cancer cells dissemination from local site. Moreover, a recent study pointed out cell autonomous expression of ECM genes in pancreatic cancer cells might contribute to their spread to distal organs.19

To explore the role of Setd2 in PDAC, we generated mice harbouring pancreatic-specific Setd2 depletion together with oncogenic Kras mutation in this study. According to mice phenotype and human PDAC database, we defined Setd2 as a putative tumour suppressor gene in PDAC. Mechanistically, we delineated the function of Setd2 both in acinar cells homeostasis and EMT process, indicating its dual function in PDAC initiation and metastasis. Thus, understanding the molecular mechanism of Setd2 in PDAC may pave the way for the development of targeted, patients-tailored therapies for PDAC patients with SETD2 mutations.

Materials and methods

Mice

Setd2f/f mice were generated by Shanghai Biomodel Organism Co. using conventional homologous recombination in embryonic stem (ES) cells as previously reported.9 The Pdxcre mice were purchased from The Jackson Laboratory. Setd2f/f mice were mated with Pdxcre mice to generate Pdxcre;Setd2f/f (SC) mice in C57BL/6 background. Furthermore, LSL-KrasG12D;Setd2f/f mice were interbred with SC mice to generate Pdxcre;LSL-KrasG12D;Setd2f/f (KSC) and Pdxcre;LSL-KrasG12D (KC) mice housing under the same condition were used as control mice. Mice were harvested at indicated time for pancreas histology investigation. The mice that lose weight more than 20% within 1 week will be euthanised and counted as death.

Human subjects

The PDAC tissue microarray from Renji hospital (Shanghai, China) was approved with Local Ethics Committee approval and patient consent. Briefly, clinical parameters of patients with pancreatic cancer were collected, including age, sex, stage, pathological diagnosis, differentiation status, TNM status and survival. None of the patients underwent preoperative chemotherapy or radiation therapy prior to surgery.

Statistical analysis

Survival curve was estimated by Kaplan-Meier methods. Two-way analysis of variance (ANOVA) or one-way ANOVA together with Bonferroni’s post hoc test was used for multiple groups analysis. Unpaired Student’s t-test was used to determine statistical significance in the rest experiments, and p value less than 0.05 was considered significant: * for p<0.05, ** for p<0.01, *** for p<0.001. Values are expressed as mean±SEM or mean±SD (Prism 6; GraphPad Software). Unless indicated, results are from at least two or three independent experiments.

Other detailed methods can be found in the online supplementary materials and methods.

Results

SETD2 deficiency predicts aggressive phenotype and poor prognosis of PDAC

Genetic analyses of multiple tumours have uncovered the existence of SETD2 mutations, including pancreatic cancer (online supplementary figure S1A). PDAC is a major form of pancreatic cancer. There were approximately 87% of cases containing KRAS mutations in four public PDAC databases, including TCGA (PAAD), UTSW,20 ICGC21 and QCMG3 (cases=775). In addition to KRAS, other three driven mutations, TP53, CDKN2A and SMAD4, occupied from 27% to 58% (figure 1A). There were about 3% of patients that carried SETD2 mutations in a total of 775 patients with PDAC, and nearly all SETD2 mutations were accompanied with aberrant KRAS (figure 1A).

Figure 1

Mutations or low expression of SETD2 predicts aggressive phenotype and poor prognosis of PDAC. (A) Oncomap showing the distribution of human SETD2, KRAS, TP53, CDKN2A and SMAD4 mutation in pancreatic cancer patients from TCGA (PAAD), UTSW, ICGC and QCMG database (775 cases in total). (B–D) Gene set enrichment analysis (GSEA) of gene expression between hKC (harbouring KRAS multination alone, 33 cases) and hKSC (harbouring both KRAS and SETD2 mutations, 3 cases) from TCGA (PAAD) database. The significantly enriched hallmark pathways were listed. (C and D) DNA repair (C) and epithelial–mesenchymal transition (D) enrichment plots in patients with hKSC compared with patients with hKC are depicted. (E and F) Representative IHC staining indicates low and high expression of SETD2 (E) or H3K36me3 (F) in pancreatic tumour. Scale bar=100 µm. (G and H) Overall survival (OS) of patients with PDAC (from Renji cohort 1) based on SETD2 or H3K36me3 expression, respectively. (I) Statistic analysis of positive correlation between SETD2 and H3K36me3 expression, p<0.0001 (χ2 test). IHC, i mmunohistochemistry; NES,  normalised enrichment score; PDAC, pancreatic ductal adenocarcinoma.

To explore the possible function of SETD2 in PDAC, we compared the transcriptome differences between KRASmutTP53WTSETD2WTpatients (hKC, 33 cases) and KRASmutTP53WTSETD2mutpatients (hKSC, 3 cases) in TCGA (PAAD) cohort (figure 1B). We performed gene set enrichment analysis (GSEA) to gain a global view of the distinguished transcriptome profiles between patients with hKC and hKSC. Several pathways related to aggressive phenotype were enriched in hKSC, including oxidative phosphorylation, interferon alpha and gamma response, MYC targets, DNA repair and EMT (figure 1B–D). In addition, we performed alternatively analysis based on SETD2 mRNA level. In line with previous result, genes related to DNA repair and EMT were significantly upregulated in PDAC with lower SETD2 expression (online supplementary figure S1B–D).

To further assess the clinical relevance of SETD2 in PDAC, we performed immunohistochemistry (IHC) staining with validated antibodies against SETD2 and H3K36me3 on PDAC tissue array from two independent cohorts. Our analysis showed that low expression of SETD2 and H3K36me3 were both associated with poor prognosis of patients with PDAC (figure 1E–H; online supplementary figure S1D,E). More importantly, the expression of SETD2 and H3K36me3 is positively correlated in our tissue array cohorts (figure 1I). Taken together, these findings suggest a critical role of SETD2/H3K36me3 in pancreatic carcinogenesis.

Setd2 loss facilitates Kras-induced and inflammation-induced ductal metaplasia

To study the impact of Setd2 loss in pancreatic cancer, we crossed previously described Setd2f/f 22 with Pdxcre transgenic strains to ablate Setd2 in pancreas (Pdxcre;Setd2f/f, referred to as SC mice) (online supplementary figure S2A). Tissue-specific depletion of Setd2 was confirmed by qPCR (online supplementary figure S2B). Immunoblotting analyses revealed the reduced expression of Setd2 and H3K36me3 in both total pancreas and isolated acinar cells from SC mice compared with littermate Setd2f/f mice (online supplementary figure S2C,D). We observed that pancreata weight of SC mice increased slightly compared with age-matched Setd2f/f mice (figure 2A). In comparison with littermate Setd2f/f mice, we noticed the enlarged size of acinar cells and higher percentage of Ki-67+ acinar cells in SC mice (online supplementary figure S2E,F). These results indicate that loss of Setd2 affects homeostasis of pancreatic acini. To further determine the Setd2 role in acinar homeostasis, we applied caerulein-induced pancreatitis recovery model on Setd2f/f and SC mice. In comparison with Setd2f/f mice, we observed much more ADM-like lesions and less amylase+ signals in pancreata of SC mice on day 3 after last injection of caerulein (figure 2A–C). Normally, metaplastic lesions resolved on day 7 in wildtype mice, while pancreata from SC mice still maintained visible ADM-like lesions. All results suggest that Setd2 loss enhances acinar-to-ductal reprogramming on inflammation.

Figure 2

Setd2 loss facilitates inflammation-induced and Kras-induced ductal metaplasia. (A) Setd2f/f and Pdxcre; Setd2f/f (SC) mice were administrated with caerulein (100 µg/kg, 8 hourly injection/day) for consecutive 2 days. Pancreas tissues were collected at indicated days for H&E staining; (B) Quantification of ADM area (%) in pancreas tissue from indicated mice at day 3 (D3). (C) The IHC staining of amylase and H3K36me3 on indicated pancreas tissue. (D) Relative pancreas weight of Setd2f/f, Pdxcre;Setd2f/f(SC), Pdxcre;LSL-KrasG12D (KC) and Pdxcre;LSL-KrasG12D; Setd2f/f (KSC) mice. (E) Pancreatic tissues from indicated mice for staining of H&E, CK19, Ki67, Sirus red and H3K36me3. Scale bars=100 µm. (F) Quantification of the affected pancreatic area from indicated mice at 3 months or 12 months of age. (G) Ratio of early lesion (ADM and early PanIN) and late lesion (late PanIN and PDAC) in affected pancreatic areas from each group as shown in (F); (H) Representative images and statistic of acinar-to-ductal metaplasia of acinar explants on transforming growth factor α  (TGFα) stimulation from indicated mice. IHC, immunohistochemistry; PanIN, pancreatic intraepithelial neoplasia; PDAC, pancreatic ductal adenocarcinoma.

To investigate whether Setd2 contributed to Kras-induced transformation, we generated Pdxcre;LSL-KrasG12D;Setd2f/f mice (referred to as KSC mice) (online supplementary figure S2A). In line with literatures,23Pdxcre;LSL-KrasG12D mice (referred to as KC mice) develop infrequent ADM lesions at the age of 3 months and gradually progress to PDAC around 12 months of age. In sharp contrast to KC mice, additional depletion of Setd2 led to abnormal pancreatic morphology, accompanied with accelerated acini loss and ductal lesion formation at 3 months of age (figure 2D,E). Next, we compared histological and molecular features of ductal lesions between KC and KSC mice. As shown, KSC mice in 3 months of age displayed larger affected pancreatic area with much higher grades of PanIN/PDAC lesions, compared with the age-matched KC mice (figure 2F,G). Hence, ablation of Setd2 significantly accelerated KrasG12D-induced ADM/PanIN formation. In line with in vivo data, pancreatic acinar explants derived from SC mice dedifferentiated more rapidly into ductal-like structures than explants from Setd2f/f mice in response to transforming growth factor α (TGFα) (figure 2H). Collectively, Setd2 loss synergises oncogenic Kras to facilitate ADM transition and progression towards PanIN and PDAC.

Loss of Setd2 alters gene sets related to acinar cell homeostasis

To understand the molecular consequence of Setd2 loss in acinar cells, we obtained gene expression profile of pancreata from SC and littermate Setd2f/f mice (6 weeks old) by RNA sequencing (RNA-Seq). We found distinct transcriptome difference between SC and Setd2f/f (figure 3A). There were 282 upregulated and 207 downregulated genes on Setd2 loss in pancreas (FC>1.5, p<0.05), in which several acinar-specific genes were downregulated, such as Pnlip and Amy2a1 (figure 3A and online supplementary table S1). Of note, top 10 downregulated genes were associated with mature acinar cell state (such as Sostdc1, Vtn, Fgf21 and Cxcl12),24 which were further validated by qPCR assay in acinar cells from SC and littermate Setd2f/f mice (figure 3A,B and online supplementary table S2). In addition, GSEA analysis on differentially expressed genes revealed that proliferation-related gene sets were enriched in transcriptome of Setd2-deficient pancreas, which is consistent with previous observation (figure 3C). More interestingly, MYC targets signature were enriched in Setd2-deficient pancreas (figure 3D). Previous study has proven that repression of Myc is required for maturation of preacinar cells.25 Moreover, Myc overexpression led to induction of dedifferentiated state of acinar, which would contribute to initiation of PDAC.25 The expression of Myc was further validated with immunoblotting and qPCR assays in primary acinar cells from Setd2f/f and SC mice. As indicated in figure 3E,F, we found that Myc protein level was increased, while its mRNA level was downregulated in the absence of Setd2, suggesting that Myc might be indirectly regulated by Setd2 at protein level (figure 3E,F). Together, these findings shed light on possible mechanism of acinar cells dysregulation in the absence of Setd2.

Figure 3

Loss of Setd2 alters gene sets related to acinar cell homeostasis. (A) Heat map illustrates gene expression of pancreas tissue from Setd2f/f and Pdxcre;Setd2f/f (SC) mice. (B) Expression of indicated genes were validated in acinar cells isolated from Setd2f/f and SC mice, respectively. Results are presented as the mean±SEM (n=4). (C) Gene set enrichment analysis (GSEA) of gene expression between Setd2f/f and SC are depicted. (D) MYC signature enrichment plots in SC compared with Setd2f/f. (E) Gene expression of Myc in acinar cells from Setd2f/f and SC mice. Results are presented as the mean±SD, n=4 per condition. (F) Western blot analysis of Setd2, Myc, β-catenin and H3K36me3 protein level in acinar cells isolated from Setd2f/f and SC mice, respectively.

Setd2 regulates E3-ligase Fbxw7 via H3K36me3

To study the epigenetic alteration dependent on Setd2 and causatively link them with changes in gene expression, we next performed chromatin immunoprecipitation sequence (ChIP-Seq) to map the distribution of H3K36me3 in pancreatic acinar cells. Consistently with previous reports,22 H3K36me3 were widely distributed in whole genomic regions, including gene bodies, promoters, 3′ UTR and intergenic zones (figure 4A,B). Collectively, the majority of H3K36me3 marks were found at the gene bodies of target genes (figure 4A,B). The pattern of peak distribution determines that Setd2 and H3K36me3 regulate gene transcription mainly through other regulation zones but not promoter regions. There were 7199 H3K36me3 peaks (on 3833 genes) were identified, among which there were 74 differentially expressed (including downregulated and upregulated) genes in Setd2-deficient acinar cells (figure 4C and online supplementary table S3-4). Moreover, gene ontology (GO) analysis revealed that these overlapping 74 genes were closely related to DNA metabolism, cell–cell signalling, cell cycle and proliferation and developmental processes (figure 4D and online supplementary table S4). As shown in figure 3, depletion of Setd2 increased Myc expression at protein level, suggesting that Setd2/H3K36me3 might positively regulate an E3 ligase-coding gene related to Myc protein expression. As we expected, Fbxw7, a well-defined E3 ligase of Myc, was listed among these 74 genes.26 Downregulated Fbxw7 level was validated in acinar cells from SC mice, comparing with cells from Setd2f/f mice (figure 4D). Using ChIP-qPCR, we further confirmed the existence of H3K36me3 and Setd2 marks at the introns of Fbxw7 gene, and the signals decreased along with Setd2 loss (figure 4E,F). Furthermore, both immunoblotting and IHC assays demonstrated the reduction of Fbxw7 protein level, as accompanied with increased Myc expression in Setd2-deficient acinar cells (figure 4G,H). Surprisingly, other targets of Fbxw7, such as mTOR, β-catenin and cyclin E1, had no obviously change, suggesting Myc as a main target of Fbxw7 in pancreatic acinar cells. Moreover, we found strong correlation between SETD2 and FBXW7 mRNA levels based on human GTEx-Pancreas database (figure 4I). Fbxw7, a frequently mutated gene in pancreatic cancer, has shown to accelerate KrasG12D-induced pancreatic tumourigenesis.27 Next, we sought to investigate whether loss of Fbxw7 would affect homeostasis of acinar cells. To this end, we found that knockdown of Fbxw7 in acinar cells could significantly elevated Myc expression and accelerated the ADM formation (figure 4J,K).

Figure 4

Setd2 regulates E3-ligase Fbxw7 via H3K36me3. (A) Normalised read density of H3K36me3 ChIP-seq signals of Setd2f/f acinar cells from 5 kb upstream of the TSS to 5 kb downstream of the TES. (B) Analysis of the occupancy of H3K36me3 ChIP-seq peaks in gene bodies and intergenic regions. (C) Venn diagram illustration of H3K36me3 peaks of Setd2f/f acinar cells and the overlap with total differential expression genes determined by RNA sequencing (Pdxcre; Setd2f/fvs Setd2f/f mice). GO enrichment analysis of 74 overlapping genes. (D) The expression of Fbxw7 was validated in acinar cells isolated from Setd2f/f and Pdxcre;Setd2f/f mice by qPCR assay. Results are presented as the mean±SD (n=4). (E and F) ChIP-qPCR analyses of H3K36me3 and Setd2 occupancy to the locus of Fbxw7 gene, IgG was used as the control. (G) Western blot analysis of Setd2, Fbxw7, mTOR, β-catenin, c-Myc, Mcl1, Cyclin E1 and H3K36me3 protein level in acinar cells isolated from indicated mice. (H) Immunohistochemistry assay of H3K36me3 and Fbxw7 in pancreas of 3-month-old Setd2f/f and Pdxcre;Setd2f/f mice. (I) The Pearson product-moment pair-wise gene correlation analysis between SETD2 and FBXW7 with GTEx-Pancreas expression database. The log-scale axis was used for visualisation. (J) Western blot analysis of Fbxw7 and c-Myc level in primary acinar cells transfected with siCon, siFbxw7-1 and siFbxw7-3, respectively. (K) Downregulation of Fbxw7 accelerated TGFα-induced in vitro ADM formation in primary acinar cells. Representative images of acinar-to-ductal metaplasia of acinar explants on TGFα stimulation from indicated group. Scale bars=500 µm. ChIP-seq, chromatin immunoprecipitation sequence; GO, gene ontology; TGFα, transforming growth factor α.

Loss of Setd2 enhances EMT and metastasis

Low expression of Setd2 was associated with poor prognosis of patients with PDAC (figure 1E and F). It is well known that the metastasis of pancreatic cancer cells to distant organs (including liver and lung) is often the leading cause of death in patients with PDAC.28 Consistently, KSC mice died as early as 5 weeks old, and almost 40% of the mice died (or 20% weight loss within 1 week) around 3 months of age. In addition, significant liver and lung metastases were observed in some KSC mice at 13–15 weeks of age (the ratio is 47.4% and 42.1%, respectively) (figure 5A,B). EMT, where cells lose epithelia identity to disseminate into a vascular or lymphatic system, is the first step for metastasis.29 In line with previous GSEA analyses (figure 1D and online supplementary figure S1D), we found a prominent decreased expression of E-cadherin and β-catenin on pancreatic malignant lesions from KSC mice, in comparison with KC mice (figure 5C).

Figure 5

Loss of Setd2 enhances EMT and metastasis. (A) Kaplan-Meier survival curve between KC (n=20) and KSC (n=32) mice. (B) Representative H&E staining of lung and liver tissues from KSC mouse at 13–15 weeks age. (C) Immunohistochemistry assays of E-cadherin and β-catenin in the pancreas from indicated mice. Scale bars=50 µm. (D) The expression of indicated proteins in Setd2WT, Setd2KO1 and Setd2KO2 cells derived from KPC1199 with immunoblotting assay. (E) Representative images of Setd2WT, Setd2KO1 and Setd2KO2 cells (KPC1199 derived). Scale bars=50 µm. (F) The incorporation of EdU in Setd2WT, Setd2KO1 and Setd2KO2 cells for 2 hour. (G) Wound healing assay of Setd2WT, Setd2KO1 and Setd2KO2 cells for 20 hours. Representative images were taken with phase contrast microscopy. Scale bars=200 µm. (H) Transwell-based invasion assay with Setd2WT, Setd2KO1 and Setd2KO2 cells for 24 hours. Relative invasion ability was quantified with OD595. Scale bars=50 µm. (I) In vivo lung metastasis assay with Setd2WT and Setd2KO cells. Representative images of lung metastases and quantitation of metastasis nodules. Arrowheads denote the visible metastasis nodules. Scale bars=1 cm. (I) Representative H&E staining of lung from indicated groups. EMT, epithelia–mesenchymal transition.

To investigate whether the Setd2 loss affected EMT process, we further depleted Setd2 with CRISPR/Cas9 in KPC1199, a murine pancreatic cancer cell line derived from KPC mice (Pdxcre;LSL-KrasG12D;LSL-TP53R172H).30 We generated two Setd2-KO cell lines (Setd2KO1 and Setd2KO2) via targeting exons encoding SET domain and post-SET domain respectively, and each one harbours a frameshift mutation in Setd2 allele (online supplementary figure S3A), resulting in Setd2 and H3K36me3 loss at protein level (figure 5D). In line with what we found on KSC mice, both Setd2-KO cell lines (Setd2KO1 and Setd2KO2) showed drastically impaired expression of epithelial markers (E-cadherin and β-catenin) and elevated expression of mesenchymal marker (N-cadherin and S100A4) (figure 5D). Consistent with the immunoblotting result, PDAC cells were transformed into mesenchymal-like morphology on depletion of Setd2 (figure 5E and online supplementary figure S3B). Furthermore, we found that Setd2 loss would promote the cell migration and invasion capabilities by wound healing and transwell-based invasion assays, in spite of lower proliferative capability in vitro (figure 5F-H). To further confirm the role of Setd2 in tumour metastasis, we grafted PDAC cells into tail vein or splenic vein as lung or liver metastasis mouse model. Setd2 deficiency led to obviously increase in tumour burden in the lungs and liver respectively (figure 5I,J; online supplementary figure S3C and D). Hence, we reason that Setd2 deficiency enhances EMT of PDAC cells and thus facilitates their distal metastasis.

Sustained AKT activation and α-catenin loss mediate EMT and migration in Setd2-null PDAC cells

To gain mechanistic insight into how Setd2 loss promotes EMT process, the gene expression profile of Setd2WT and Setd2KO cells was examined by RNA-seq. We found 3993 differentially expressed genes on Setd2 loss (fold change >2; p value <0.05), of which 2333 genes were decreased and 1660 genes were increased (figure 6A; online supplementary table S5). In line with pervious observation, several pathways related to EMT and migration, including PI3K-AKT signalling,31 focal adhesion,32 cytokine–cytokine receptor interaction,33 ECM–receptor interaction34 and TGF-β signalling,35 were upregulated on Setd2 deficiency (figure 6B). Consistently, we found that phospho-Akt level was dominantly activated in Setd2-null PDAC cells, in comparison with Setd2WT cells (figure 6C). When we took a closer look at these upregulated genes in PI3K-AKT pathway, we found that more than half of the genes were secreted cytokines, extracellular matrix or plasma membrane receptors, all of which were involved in ECM–receptor or cytokine–receptor interaction (online supplementary figure S4A). Upregulated genes were further validated with qPCR assay (online supplementary figure S4B). Strikingly, inhibition of PI3K-AKT signalling impaired cell migration ability while upregulating Cdh1 expression and decreasing ECM-related gene expression (online supplementary figure S4C–F). The above findings uncover that sustained activation of AKT signalling, at least partially mediates the enhanced migration ability of Setd2-null PDAC cells.

Figure 6

α-Catenin loss mediates EMT and migration in Setd2-null PDAC cells. (A) Heat map of RNA-Seq data to compare the gene expression of Setd2WT and Setd2KO derived from KPC1199 cells. (B) KEGG pathway enrichment in RNA-Seq data (Setd2KOvs Setd2WT). (C) Western blot analysis of E-cadherin, β-catenin, p-Akt, Akt, p-Erk1/2, Erk1/2, p-p38, p38, p-Smad2/3, Smad2/3, Smad4 and H3K36me3 levels in Setd2WT, Setd2KO1 and Setd2KO2 derived from KPC1199 cells. (D) Normalised read density of H3K36me3 ChIP-seq signals of Setd2WT and Setd2KO derived from KPC1199 cells from 5 kb upstream of the TSS to 5 kb downstream of the TES. (E) Analysis of the occupancy of H3K36me3 ChIP-seq peaks in gene bodies and intergenic regions of indicated cells. (F) Venn diagram illustration of H3K36me3 peaks only occupied in Setd2WT derived from KPC1199 cells, and the overlap with differential expression genes determined by RNA sequencing. KEGG pathway enrichment of the 222 overlapping genes. (G) Expression of Ctnna1 was validated in Setd2WT, Setd2KO1 and Setd2KO2 cells derived from KPC1199 by qPCR. Results are presented as the mean±SD (n=4). (H) Western blot analysis of E-cadherin, α-catenin, β-catenin, H3K36me3, Gapdh and total histone H3 levels in indicated cells. (I and J) ChIP-qPCR analysis of H3K36me3 and Setd2 occupancy to the locus of Ctnna1 gene, IgG was used as the control. (K) immunofluorescence assay of E-cadherin, α-catenin and β-catenin in Ctnna1-depleted KPC1199 cells. Scale bars=20 µm. (L) Transwell-based migration assay were performed in KPC1199 cells transfected with siCon, siCtnna1-1 and siCtnna1-3 for 16 hours. Relative migration ability was quantified with OD595. Scale bars=50 µm. (M) The Pearson product-moment pair-wise gene correlation analysis between SETD2 and CTNNA1 in TCGA (PAAD), GSE15471 and GSE16515 databases. The log-scale axis was used for visualisation. ChIP-seq, chromatin immunoprecipitation sequence; EMT, epithelia–mesenchymal transition; PDAC, pancreatic ductal adenocarcinoma.

Figure 7

Loss of Setd2 promotes Kras -induced pancreatic carcinogenesis. ADM, acinar to ductal metaplasis; EMT, epithelial–mesenchymal transition.

To identify direct targets of Setd2/H3K36me3 in PDAC cells, we next performed chromatin ChIP-Seq to map the distribution of H3K36me3 both in Setd2WT and Setd2KO cells. As shown in figure 6D,E, the distribution pattern of H3K36me3 in whole genomic regions was similar to that in acinar cells. The holistic H3K36me3 occupancy were brought down in Setd2-deficient cells (figure 6D,E). We acquired 2745 H3K36me3 peaks (on 1805 genes) only located in Setd2WT cells (q-value<0.1), suggesting that these genes might be directly regulated by Setd2 (figure 6F). Through integrating RNA-Seq and ChIP-Seq data, there were 222 candidate targets of Setd2/H3K36me3 (figure 6F and online supplementary table S6). The mRNA level of Cdh1 and EMT-related transcriptional factors (Twist1 and Zeb2) were altered on Setd2 loss; however, none of them were included (online supplementary figure S5A-C and table S6). These 222 genes were functionally associated with Focal adhesion, ECM–receptor interaction and adhesion junction (figure 6F). Among them, Ctnna1 caught our attention. In current view, α-catenin (Ctnna1) is a component of cadhesome, bridging E-cadherin/β-catenin with actin cytoskeleton to maintain integrity of adherent junction.36 Downregulation of Ctnna1 was validated both in mRNA and protein level on Setd2 loss in PDAC cells (figure 6G,H). Using ChIP-qPCR, we further confirmed that H3K36me3 and Setd2 signals could be enriched at the gene bodies of Ctnna1, and the signals decreased along with Setd2 loss (figure 6J). Similar results were obtained in PDC0034, a human PDAC cell lines from patient-derived xenograft (online supplementary figure S6A–C). Consistent with previous reports,37 we demonstrated that knockdown of Ctnna1 in PDAC cells could disrupt E-cadherin expression (figure 6K and online supplementary figure S7A–C) and facilitate cells migration (figure 6L and online supplementary figure S7D). More importantly, we found strong and positive corelationship between mRNA levels of CTNNA1 and SETD2 in several PDAC databases (figure 6M).38 Thus, loss of α-catenin on Setd2 deficiency contributes to EMT and migration of PDAC cells.

Discussion

Epigenetic dysregulation has been implicated in many cancers.39 Of note, histone methyltransferases (HMTs) are frequently mutated/deleted in a spectrum of human tumours; however, whether the HMTs alterations are the driving force in tumourigenesis remains largely unknown. Evidence from human genomes sequencing has linked SETD2 to PDAC,40 but its causal role in pancreatic cancer has not been reported yet. In this study, we found that Setd2 deletion affected pancreatic acini homeostasis and facilitated initiation and metastasis on oncogenic Kras mutation in PDAC (figure 7). To our knowledge, this is the first animal model that interprets the function of Setd2 in PDAC. Moreover, our in-depth studies provide insights into how Setd2 loss co-opts with oncogenic Kras in the development of pancreatic cancer.

Kras activation in acinar cells would trigger acinar-to-ductal reprogramming and PanIN formation.15 41 42 Hence, exploring the mechanism of acini differentiation/maturation would foster a better understanding of PanIN initiation. For example, the genes related to acini integrity/maturation (e.g., Ptf1a, Mist1 and Nr5a2), had been reported as a barrier to Kras-induced ADM.17 43 44 Acinar cells differentiation/maturation determines organ size, and acinar homeostasis must be tightly controlled by gene network.45 In our study, we first observed that Setd2 loss affected pancreas size and acini homeostasis during development and facilitated ADM only under oncogenic Kras activation.25 46 To further understand the role of Setd2 after ductal metaplasia, we deleted Setd2 through CRISPR/Cas9 in mouse PDAC cell, harbouring Kras and Tp53 mutations. Strikingly, loss of Setd2 in PDAC enhanced AKT activation via autocrine loop of ECM/cytokine signals, at least partially contributed to enhanced EMT. In addition, recent study observed pancreatic circulating tumour cells have cell-autonomous expression of ECM genes, which might contribute to their spread to distal organs.19 47 Of note, EMT is a common feature for most aggressive tumours.48 Thus, enhanced EMT on Setd2 loss might be implicated in other type of cancers, especially in those with Kras mutation.

Recent studies have shown that some genes (e.g., PDX1 and KLF4) have a context-dependent protumourigenic or tumour suppressive function in the development of PDAC, which raises a concern about their targeting therapy.18 49 Hence, it is very important to comprehensively evaluate gene function throughout pancreatic carcinogenesis. In our study, Setd2 loss has both impacted on acinar-to-ductal reprogramming and epithelia–mesenchymal transition, suggesting tumour suppressive role of Setd2 at different stages of pancreatic carcinogenesis. Given that we did not observe any pancreatic precancerous lesion in Setd2 deficient mice, we hypothesised that Setd2 loss had to cooperate with other driver mutations (e.g., KrasG12D). Very recently, Qin group reported that Setd2 loss could counteract with Apcmin/+ to promote colorectal carcinogenesis via augmenting Wnt/β-catenin signalling.9 Setd2 loss did not affect Wnt/ β-catenin signalling in the context of KrasG12D, suggesting distinct role of Setd2 under different genetic background. Besides active regulation of gene transcription, Setd2/H3K36me3 has been shown to regulate DNA repair and alternative mRNA splicing. Moreover, non-histone targets of Setd2 have been found, including α-TUBULIN and STAT1, which implicates a vital role in mitosis and virus infection respectively.50 51 Thus, other mechanisms might be involved to mediate the function of Setd2 in the development of pancreatic cancer.

In summary, our findings highlight the dual suppressive role of SETD2/H3K36me3 in acinar-to-ductal reprograming and EMT during pancreatic carcinogenesis, in the context of oncogenic Kras activation, via regulation of distinct targets in different cell types. Thus, our studies on the molecular mechanism of SETD2 will advance our understanding of epigenetic dysregulation at different stages of PDAC development, as well as other human cancers with SETD2 deficiency upon KRAS mutation.

Acknowledgments

We give our special thanks to Jing Geng, who is responsible for Setd2f/f mice breeding at the early stage of this project. We thanks to Prof Yujie Tang for giving us valuable suggestion.

References

View Abstract

Footnotes

  • NN and PL contributed equally.

  • Contributors JX, NN and PL designed experiment and interpreted data; NN and PL performed most of the experiments; YY, RH, JS, JW, MY and LZ assisted in some experiments; L-WW, W-QG and ZZ provided the key materials; ZZ, GGX and AH assisted in some discussion; JX and NN wrote the manuscript; JX, LL and YS provided the overall guidance.

  • Funding This work was supported by Program for Professor of Special Appointment (Eastern Scholar) at Shanghai Institutions of Higher Learning No TP2015007 (JX); Shanghai Municipal Education Commission-Gaofeng Clinical Medicine Grant Support No 20161312 (JX); The Shanghai Youth Talent Support Program (JX); Shanghai Rising-Star Program No 19QA1408300 (NN); The Science and Technology Commission of Shanghai Municipality No 18140902700 and No 19140905500 (LL); State Key Laboratory of Oncogenes and Related Genes No KF01801 (LL); Innovation Research Plan supported by Shanghai Municipal Education Commission No ZXGF082101 (LL); National Natural Science Foundation of China No 81702938 and No 81770628 (JX), No 81772938 (LL), No 81802307 (PL), No 81874175 (YS) and No 81802317 (MY).

  • Competing interests None declared.

  • Ethics approval All protocols for animal use and euthanasia were reviewed and approved by RenJi Hospital Institutional Animal Care and Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Correction notice This article has been corrected since it published Online First. The author name Li-Wei Wang has been corrected.

  • Patient consent for publication Not required.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.