Epigenetic reprogramming in pancreatic premalignancy and clinical implications

Pancreatic cancer (PC) is the most lethal human cancer, with less than 10% 5-year survival. Pancreatic premalignancy is a genetic and epigenomic disease and is linked to PC initiation. Pancreatic premalignant lesions include pancreatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasm (IPMN), and mucinous cystic neoplasm (MCN), with pancreatic acinar-to-ductal metaplasia (ADM) as the major source of pancreatic premalignant lesions. Emerging evidence reveals that an epigenetic dysregulation is an early event in pancreatic tumorigenesis. The molecular mechanisms of epigenetic inheritance include chromatin remodeling; modifications in histone, DNA, and RNA; non-coding RNA expression; and alternative splicing of RNA. Changes in those epigenetic modifications contribute to the most notable alterations in chromatin structure and promoter accessibility, thus leading to the silence of tumor suppressor genes and/or activation of oncogenes. The expression profiles of various epigenetic molecules provide a promising opportunity for biomarker development for early diagnosis of PC and novel targeted treatment strategies. However, how the alterations in epigenetic regulatory machinery regulate epigenetic reprogramming in pancreatic premalignant lesions and the different stages of their initiation needs further investigation. This review will summarize the current knowledge of epigenetic reprogramming in pancreatic premalignant initiation and progression, and its clinical applications as detection and diagnostic biomarkers and therapeutic targets in PC.


Introduction
Pancreatic cancer (PC) is one of the most lethal diseases, with an incidence nearly equal to the mortality; e.g., there are 495,773 new cases and 466,003 deaths worldwide in 2020 (1). The 5-year survival rate of PC patients is less than 10%, the lowest among common cancers (2). The poor patient survival is mainly due to the fact that PC is asymptomatic in the early stages and difficult for early diagnosis. Although improvements in diagnosis, surgical procedure, and adjuvant therapy have contributed to a sustained decrease in the overall mortality rate, the death rate in PC has increased over the past decades (3). The development of minimally invasive biomarker assays for early diagnosis and effective clinical management of PC is urgently needed to reduce high morbidity and mortality.
The PC develops through the accumulation of a series of genomic and epigenomic alterations and progresses into invasive carcinoma. The process can take up to 15-20 years to develop from the occurrence of the first initiating mutational event. Understanding and discovering early PC and its precursor lesions are fundamentally important to effectively control this deadly disease and improve the outcome of patients. It is generally believed that pancreatic acinar-toductal metaplasia (ADM) is the major source of pancreatic premalignant lesions (4). ADM is a reversible process during injury or acute pancreatitis and becomes irreversible with chronic pancreatitis or Kras mutations (5). Pancreatic premalignant lesions include pancreatic intraepithelial neoplasia (PanIN), intraductal papillary mucinous neoplasm (IPMN), and mucinous cystic neoplasm (MCN), with PanIN being the most frequent (4,6). PC is well investigated at the genetic level, while the molecular basis of its precursor, particularly epigenetics, remains to be explored.
Epigenetics studies the inheritance of phenotypes that occur without a corresponding change in DNA sequence. The molecular mechanisms of epigenetic inheritance include DNA methylation, RNA methylation, non-coding RNA, alternative splicing, histone modifications, chromatin remodeling, and phosphorylation (7)(8)(9)(10)(11). Epigenetic changes contribute to the silencing of tumor suppressor genes and the activation of oncogenes. Aberrant epigenetic alterations link to many genes with important roles in PC initiation and progression (12,13). Epigenetic dysregulation is an early event in pancreatic tumorigenesis. The cancer-associated epigenetic changes occur hours after pancreatic injury at the beginning of ADM and before the appearance of widespread PanIN lesions and involve an "acinar-to-neoplasia" chromatin remodeling that leads to the early malfunction of genes that initiate PC (14). PC initiation results from a complex interaction of genetic and epigenetic damage that triggers changes in cell identity and tissue state and leads to a neoplastic cell fate, i.e., from normal lineage-specifying to cancer-defining (14,15).
In this review, we first present an overview of the basic molecular mechanisms of epigenetic reprogramming in pancreatic premalignancy and describe the potential use of epigenetic reprogramming for early PC detection and targeted intervention.

Expression of DNA methyltransferases
DNMT1 is the most abundant and primarily a DNA maintenance methyltransferase and is responsible for copying methylation patterns after DNA replication (20). DNMT1, DNMT3A, and DNMT3B mRNA levels significantly increase from normal ducts to PanINs to PC (74). Similar to the mRNA results, DNMT1 protein expression increases from normal ducts with or without inflammation to precancerous lesions (PanINs or IPMNs) to PC (64). DNMT1 protein overexpression is involved in multistage tumorigenesis of the pancreas. PC patients with high levels of DNMT1 protein expression have a poorer outcome than those with low levels of DNMT1 expression (64,75). Higher DNMT1 expression correlates with advanced stages of the disease (64). DNMT1 is oncogenic in promoting malignant phenotype and stemness in PC. These effects are achieved via promoter hypermethylation of tumor suppressor genes, e.g., cyclin-dependent kinase inhibitors, tumor suppressor microRNAs, and suppressors of epithelial-mesenchymal transition (EMT) (21).
DNMT2 methylates specifically cytosine 38 in the anticodon loop. DNMT2 is also involved in DNA recombination, DNA damage recognition, and mutation repair. The levels of DNMT3A, DNMT3B, and DNMT3L are usually elevated in cancer tissues and cells, in part due to the hypermethylation of promoter CpG-rich regions of tumor suppressor genes (22). DNMT3A and DNMT3B function as de novo methyltransferases (20). Patients having PC with higher levels of DNMT3A and/or DNMT3B expression have a poorer prognosis than those with a lower-level expression (74). In addition, the expressions of DNMT1 and DNMT3a in PC tissues are all higher than those in adjacent normal tissues (23, 76). However, the potential changes in their expression and function and underlying mechanisms in pancreatic premalignancy remain to be defined.
O 6 -Methylguanine DNA methyltransferase (MGMT) is another important DNMT and DNA-repair protein that removes mutagen from the O 6 position of guanine. O 6 -Methylguanine often mispairs with thymine during replication and results in G-to-A point mutations if the adduct is not removed (18). MGMT methylation is associated with mutations of KRAS, TP53, and SMAD4 in PC (77). There is approximately 25% loss of MGMT in PC (78) and pancreatic endocrine tumors (79).

DNA methylation of tumor genes
DNA methylation of CpG islands in the promoter regions of tumor-related genes is associated with the silencing of those genes (12). Aberrant DNA methylation is associated with chromosomal instability, while deamination of 5-methylcytosine to thymine increases gene mutagenicity (80). Aberrant methylation of the ppENK gene increases progressively from PanIN-1A and PanIN-3 to PC. Aberrant methylation of p16 gene is also increased from PanIN-1A to PanIN-3. The methylation of p16 and ppENK genes increases with PanIN grades, suggesting that methylation of these genes is an intermediate or late event during pancreatic carcinogenesis (8). Moreover, the majority of IPMNs exhibit hypermethylation of CpG islands. Hypermethylation of p16 and ppENK occurs significantly higher in high-grade IPMNs than in low-grade IPMNs. The methylation of multiple CpG islands is one of the important pathways linked to the development and progression of IPMNs (16).
The aberrant methylation is identified in the majority of the earliest PanIN-1A lesions. NPTX2 increases in methylation from

m 6 A writers
METTL3 mRNA and protein are significantly higher in PC than in adjacent tissues. The increased expression of METTL3 is associated with advanced stages and decreased survival (85). Moreover, the m 6 A RNA methylation of IGF2BP2, IGF2BP3, METTL3, FTO, YTHDC1, and YTHDF1 is increased, while that of ALKBH5 METTL16, RBM15, and ZC3H13 is downregulated (86). Overexpression of METTL3 causes m 6 A modification and an excessive maturation of miR-25-3p. The upregulation of miR-25-3p inhibits PHLPP2 and activates AKT-p70S6K oncogenic signaling, forming a malignant METTL3-miR-25-3p-PHLPP2-AKT axis and promoting PC development and progression (13).
METTL14 is an enzyme that modulates m 6 A methylation. METTL14 overexpression significantly promotes pancreatic tumorigenesis by directly targeting the mRNA of p53 effector related to PMP-22 (25), MPKa, ERK1/2, and mTOR signaling pathways (88), cytidine deaminase, an enzyme inactivating gemcitabine (89), and Cdc2-like kinases 1/SRSF5/cyclin L2 pathways (24). WTAP plays an important role in transcriptional and post-transcriptional regulation. WTAP is an independent prognostic indicator for PC. The WTAP promotes malignancy and suppresses gemcitabine chemo-sensitivity in PC (26). The pseudogene WTAPP1 RNA significantly increases in PC and is associated with poor prognosis in patients. Decreasing WTAPP1 RNA significantly suppresses the growth and metastasis of PC cells (90). However, their involvement in pancreatic premalignancy is unclear.

m 6 A erasers
Fat mass and obesity-associated (FTO) protein is a critical demethylase, or m 6 A eraser, in regulating mRNA stability by removing m 6 A residues in mRNA. The role of FTO in cancer cell homeostasis remains unclear. FTO is required for PC cell proliferation, and its knockdown causes a compromised DNA synthesis and PC cell proliferation but an increase in apoptosis. FTO interacts with MYC and bHLH transcription factors and enhances their stability by decreasing m 6 A levels. Reduced FTO expression results in elevated levels of m 6 A RNA modification in PC. Moreover, FTO demethylates the m 6 A modification of Praja ring finger ubiquitin ligase 2 (PJA2), reduces its mRNA degradation, suppresses Wnt signaling, and ultimately inhibits the malignant phenotype of PC cells (27). This novel mechanism for the regulation of PC malignancy by FTO could be a therapeutic target of PC (91).
Alkylation repair homolog protein 5 (ALKBH5) is an m 6 A demethylase. Its overexpression globally downregulates RNA m 6 A levels. ALKBH5 is generally downregulated in PC cells (92). P53mediated activation of ALKBH5 transcription regulates the m 6 A modifications in a feedback loop (93). The overview and clinical implications of epigenetic reprogramming in pancreatic premalignancy and PC progression. Epigenetic reprogramming contributes to ADM, and its progression to PanIN and then PC. The initiation and progression of pancreatic premalignancy occur after epigenetic reprogramming such as DNA methylation, RNA methylation, microRNAs, lncRNAs, circRNAs, alternative splicing, histone modification, and phosphorylation. Those epigenetic changes potentially constitute biomarkers for early detection and diagnosis and targeted treatment in PC (data from ClinicalTrials.gov). PC, pancreatic cancer; ADM, acinar-to-ductal metaplasia; PanIN, pancreatic intraepithelial neoplasia.
ALKBH5 substrates include those mRNAs encoding mitochondrial iron importers SLC25A37 and SLC25A28 and ubiquitin ligase FBXL5. Downregulation of FBXL5 in tumor samples correlates with shorter patient survival. ALKBH5 overexpression incurs FBXL5-mediated degradation of and a significant reduction in iron regulatory protein IRP2 and the modulator of EMT, SNAI1. ALKBH5 overexpression leads to a significant reduction in intracellular iron levels as well as cell migration and invasion, which could be rescued by FBXL5 knockdown (28). Wnt inhibitory factor 1 (WIF-1) is also ALKBH5 target gene. Increased ALKBH5 decreases WIF-1 RNA methylation and transactivates WIF-1, thus inhibiting Wnt signaling and PC (94). ALKBH5 activates PER1 by m 6 A demethylation in an m 6 A-YTHDF2dependent manner. Upregulated PER1 reactivates ATM-CHK2-P53/ CDC25C signaling and inhibits PC cell growth (93).

m 6 A readers
The m 6 A modifications are recognized by m 6 A-binding proteins, such as YTHDF1/2/3, IGF2BP1/2, and HNRNPA2/B1. These proteins are regarded as "m 6 A readers" (84). Abnormal modifications in the m 6 A levels are closely related to tumorigenesis. Specifically, IGF2BP2 is a member of the IGF2BP protein family. IGF2BP2 has the capacity of binding to many transcripts. The expression of IGF2BP2 is upregulation with a poor prognosis in PC. IGF2BP2 protein is gradually upregulated from normal pancreas, PanIN to PC in the Kras G12D mouse model. IGF2BP2 promotes aerobic glycolysis and cellular proliferation via directly binding to and stabilizing Glut1 mRNA (29, 96). The hnRNP A2/B1 is upregulated in the PanIN and pancreatic cancer tissues (97). Moreover, hnRNP A2/B1 increases in PC tissues and cell lines (98).

MicroRNAs
MicroRNAs are small non-coding RNAs of 20-22 nucleotides. Growing evidence suggests that miRNAs interfere with cellular The overview of the biogenesis, functions, and mechanisms of epigenetic reprogramming in pancreatic premalignancy. At the chromatin level, histone methylation, acetylation and deacetylation, and histone ubiquitination occur. DNA and RNA can be modified by methylation. Alternative splicing leads to different isoforms of mRNA and proteins. MiRNAs can regulate different key downstream gene targets that mediate cellular signaling pathways. The lncRNAs are cell type-specific and cell state-specific. The lncRNAs can act as miRNA sponges, regulate mRNA stability, and interact with proteins. The circRNAs are transcribed by the RNA polymerase and can be processed through alternative splicing of the pre-mRNA. The circRNAs can act as miRNA sponges, interact with proteins, and translate into proteins. functions in pancreatic premalignancy and malignancy (41). The direct targeting of miRNAs provides a novel approach to the development of targeted intervention of early PC.
As a critical component of the miRNA processing machinery, Dicer is an enzyme containing RNase III. Dicer essentially maintains acinar cell identity. Acinar cells without Dicer exhibit increased plasticity, as evidenced by loss of polarity and initiation of ADM and EMT. Moreover, homozygous Dicer deletion accelerates the formation of ADM but not PanIN, whereas heterozygous Dicer deletion accelerates PanIN initiation, suggesting complex roles for Dicer in the regulation of normal and neoplastic pancreatic epithelial identity (99). Let-7b and miR-495 repress HNF6 and express in developing acini. Let-7b and miR-495 expressions are downregulated in Dicer-deficient acini and pancreatitis-induced ADM. Moreover, suppressing let-7b and miR-495 causes similar effects in Dicer-deficient acini and metaplastic cells, i.e., induction of HNF6 and other hepatic genes and suppression of acinar differentiation. This gene network of Let-7b, miR-495, and their targets establish and maintain pancreatic acinar cell differentiation (32).
MiR217HG encodes miR-216a, miR-216b, and miR-217, which are enriched in pancreatic acinar cells. In germline knockout mice of miR-216a, miR-216b, or miR-217, the acini from each of the three miRNA-knockout mice produce many more ducts than those from control mice. There is an increased expression in ductal genes and reduced expression in acinar cells after acinar transdifferentiation. In caerulein-induced acute pancreatitis, miR-216a-and miR-216b-KO mice have more pancreatic duct glands and delayed recovery (33).
MiR-802 is highly expressed in pancreatic acinar cells and silenced during the early stages of pancreatic injury or cellular transformation. Knockout of miR-802 promotes ADM formation in Ptf1a +/Cre ;Kras +/LSL-G12D mice. Downregulation of miR-802 activates Sox9 and enhances the expression levels of ductal identity genes but attenuates the expression of acinar identity genes. MiR-802 also inhibits PC development by suppressing oncogenic Kras-induced ADM (34).

MicroRNAs in PanIN
MicroRNA deregulation is a very early event in the progression of PC. MiRNA-21, miRNA-221, miRNA-222, and let-7a are upregulated in PanIN lesions, with peak expression occurring in PanIN-2 and PanIN-3 (100, 101). MiR-148a and miR-217 expression levels are downregulated in PanIN-2 and PanIN-3 and PC, whereas the level of miR-196 is significantly elevated in PC and PanIN-2 and PanIN-3 as compared with normal pancreatic parenchyma. Moreover, miR-10b is highly expressed in PanIN-2 and PanIN-3. These markers could be used as biomarkers to distinguish PC and its precursors from benign lesions (35). MiRNA-196 is most specifically overexpressed miRNA in PanIN-3, which has potential use as a marker for PanIN-3 lesions (102).
The expression of miRNA-148a is significantly downregulated in PanIN-1B, PanIN-2, and PanIN-3 lesions as compared with normal duct and PanIN-1A. The methylation level in the coding region of miRNA-148a is significantly higher in PC samples than in chronic pancreatitis samples. MiRNA-148a downregulation occurs at an early stage of PC initiation and progression (103). MiRNA-148a targets the mRNA of DNMT3B. The downregulation of miRNA-148a occurs early in premalignancy lesions, leading to DNMT3B upregulation and the inactivation of tumor suppressor genes (36). MiR-155 is significantly overexpressed in both PanIN-2 and PanIN-3, while miR-21 is highly expressed only in PanIN-3 (104). MiR-196b only is expressed in PanIN-3 tissue (102).
MiR-483-3p is overexpressed in PanIN and PC. Moreover, miR-483-3p expression levels correlate with increases in the PanIN grade. Circulating miR-483-3p levels are significantly increased in the serum and serum exosomes of PC patients. Interestingly, serum levels of miR-483-3p can distinguish patients with early-stage PC (≤2 cm) from healthy controls. MiR-483-3p expression negatively correlates with Smad4 expression in PC and PanIN tissues (37).

MicroRNAs in IPMN
The expression of miRNA-101 is significantly lower and EZH2 is dramatically higher in malignant IPMN than in benign IPMN. The loss of miRNA-101 could be a trigger for the tumorigenesis of IPMN by upregulation of EZH2 (38). Downregulation of miRNA-126 is detected in PC as compared to low malignant pancreatic benign cystic tumors. MiRNA-126 directly targets KRAS transcript through its binding site within 3′-UTR (39).

MicroRNAs in MCN
The expression of miR-17-3p, miR-21, and miR-221 is significantly higher in mucinous than in non-mucinous cysts (108). MiR-224-5p expression is significantly higher while PTEN expression is lower in pancreatic mucinous cystadenocarcinoma cells than in normal tissues. Overexpression of miR-224-5p promotes cell proliferation and invasion. PTEN is the direct target gene of miR-224-5p and negatively correlates with each other (41).

Long non-coding RNAs
LncRNAs are RNA molecules and usually consist of more than 200 nucleotides without protein-coding potential (109).

LncRNAs in pancreatic premalignancy
Neat1 is a large intergenic non-coding RNA (lincRNA) and is p53-regulated with a key role in suppressing transformation and tumorigenesis. Neat1 deficiency promotes the development of ADM, PanIN, and PC in Kras G12D mice (110). LncRNA Ppp3ca and lincRNA1611 expressions are significantly high in PC cells and tissues. Ppp3ca is low in normal pancreatic tissues and PanIN-I and PanIN-II tissues, while it is upregulated in PanIN-III and PC tissues (111).

LncRNAs in PC
N 6 -methyladenosine of LINC00857 is highly upregulated and enhances its RNA stability (43). The LINC00857 competes with endogenous RNA for sponging miR-150-5p, then upregulates its target E2F3, promotes proliferation, and inhibits apoptosis in PC cells. DUXAP8 is upregulated whereas miR-448 is downregulated in PC tissue and cells. DUXAP8 directly targets miR-448, which directly binds to WTAP. DUXAP8 sponges miR-448 to modulate the malignancy of PC cells (44).
LncRNA MALAT1 is highly expressed in PC tissues, while the large tumor suppressor 1 (LATS1) expression is downregulated and YAP1 is upregulated. MALAT1 influences proliferation, migration, and invasion in PC by regulating Hippo-YAP signaling (45). MALAT1 also interacts with RNA binding protein HuR, dramatically enhances the regulation of TIA-1, and has further effects on inhibiting autophagy (46). CERS6 antisense RNA 1 (CERS6-AS1) is highly increased in PC. CERS6-AS1 competitively binds to miR-15a-5p as a molecular sponge in PC, increasing the expression of fibroblast growth factor receptor 1 (FGFR1), which also is a direct target of miR-15a-5p (47).
LncRNA PART1 expression is significantly upregulated in PC and correlates with tumor size, clinical stage, and poor overall survival. PART1 serves as a molecular sponge of miR-122. PART1 knockdown significantly reduces cell proliferation and invasion of PC (50). LncRNA ELFN1-AS1 is highly upregulated in PC. The knockdown of ELFN1-AS1 significantly increases cancer cell apoptosis and growth arrest (119).
The expression level of LINC00460 is increased in PC. Moreover, miR-320b directly targets LINC00460, whose knockdown leads to a downregulation in ARF1 expression. LINC00460 modulates the miR-320b/ARF1 axis, which inhibits tumor cell proliferation, migration, and invasion (51). LINC00460 directly targets and attenuates the tumor suppressor miR-491-5p and accelerates PC progression (120).

Circular RNAs
Alternative splicing of both exons and introns of the pre-mRNA can generate circRNAs. The circRNAs usually act as miRNA sponges because of their multiple miRNA response elements in the sequence. Certain circRNAs can also bind to proteins with RNA-binding sites. They can also encode for and translate into various proteins.
There are several circRNAs that are known to be important in PC development and progression. CircEYA3 is highly elevated in PC and associated with a poorer prognosis. CircEYA3 works as an endogenous miR-1294 sponge to upregulate c-Myc expression and promotes PC progression (52). CircNEIL3 is increased in PC and promotes PC progression. CircNEIL3 regulates ADAR1 expression by sponging miR-432-5p and induces the editing of glioma-associated oncogene 1 mRNA, thus affecting the cell cycle and promoting EMT in PC (121). The expression of circ_0075829 is also upregulated and regulates the LAMTOR3/p-ERK signaling pathway via sponging miR-1287-5p in PC. Knockdown of circ_0075829 significantly suppresses the malignancy of PC cells (53). The expression of CircCCT3 is significantly increased in PC tissues and cell lines. Patients with high expression of CircCCT3 have significantly poorer survival than those with low CircCCT3 expression. Moreover, CircCCT3 facilitates VEGFA and VEGFR2 expression as a sponge for miR-613 and promotes the malignancy of PC (54).

Alternative splicing in pancreatic premalignancy
Alternative splicing has an important role in tumorigenesis, and the splicing-derived isoforms are potentially powerful diagnostic, therapeutic, and prognostic factors. Alternative splicing produces multiple actin regulators, hMENA protein isoforms with hMENA 11a and hMENADv6 (7). hMENA is absent in normal pancreas and lowgrade PanINs and very weak in PanIN-3, while high hMENA protein levels are present in PC. However, the hMENA11a isoform is only strongly expressed in 26% of PC cases. In the absence of hMENA 11a , increased expression of hMENA and hMENADv6 is key to SMAD2mediated TGF-b1 signaling and EMT. The absence of hMENA 11a significantly correlates with poor prognosis. The hMENA splicing and associative pathways are promising targets for the prognostic biomarkers and treatment in PC (55).
Tissue factor (TF) is a membrane glycoprotein and an enzymatic cofactor of the serine protease FVII/FVIIa and triggers the blood coagulation cascade. The murine alternatively spliced tissue factor (masTF) lacks exon 5 and has a distinct 93-amino-acid C-terminus. The intensity of masTF expression increases from early PanINs to PC phenotype, suggesting that masTF could promote PC growth (56).
The alternative splicing of FGFR-2 produces two variants: IIIb and IIIc. The expression of FGFR-2 IIIb correlates with a poor prognosis and may promote migration and invasion in PC (57). FGFR-2 IIIc is highly expressed in PC tissues and cell lines, which correlates with increased proliferation, migration, and invasion. High FGFR-2 IIIc levels in PC contribute to cancer stemness and could be a novel target for PC therapy (58).

Histone modification in pancreatic premalignancy
Epigenetic regulation can effectively change gene expression through histone post-translational modification and chromatin remodeling (122) (Table 1; Figures 1, 2). The factors that regulate epigenetics undergo certain expression changes in ADM, leading to the silence of genes specific to acinar cells and upregulation of the genes specific to ductal cells.

Histone methylation
NSD1 and SETD2 genes encode two histone H3K36 methyltransferases and are altered in approximately 10% of PC cases. NSD1 protein expression gradually increases in PanIN, IPMN, and MCN, and plays an important role in PC. Also, NSD1 expression is significantly higher in metastatic PC than in normal pancreas and primary cancer. The expression of SETD2 significantly decreases in metastatic PC and PanIN as compared to that in the normal pancreas (59). Moreover, the expression of SETD2 is downregulated in human PC and mouse models. The decreased expression of SETD2 mainly silences Fbxw7 through epigenetics to relieve its inhibitory effect on c-Myc (123).
The multiple endocrine neoplasia syndrome type 1 (MEN1) is a component of the macromolecular SET domain histone methyltransferase complex, which mediates the methylation of histone 3 on lysine 4 (60). The expression of Men1 is significantly reduced in PC (124). The absence of Men1 in the mouse pancreas impairs caerulein-induced pancreas regeneration and accelerates Kras-induced tumor formation (61). The tumor-promoting effect of Men1 deletion is related to the overexpression of the transcription factor Jund. Men1 inhibits the transcription of Jund, and Men1 deletion leads to Jund upregulation (125).
NFATc2 binds to the p15 INK4b promoter site, induces the H3K9 trimethylation and heterochromatin formation, allows docking of heterochromatin protein HP1g to the complex, and promotes PC cell growth (62). NFATc2 expression increases in PanIN lesions. Enhancer of zeste 2 homolog 2 (EZH2) is a histone methyltransferase, silences Nfatc1 epigenetically by histone methylation, and promotes acinar cell redifferentiation. However, NFATC1 is required for EZH2-mediated transcriptional activation of KRAS signaling in PC cells (126).

Histone acetylation
Histone acetylation is a chromatin modification and plays important roles in gene regulation, DNA replication, and DNA damage response. Pancreatic acinar cells with Kras mutations acquire histone H4 acetylation prior to the phenotype of premalignant lesions (127). Acetyl-CoA is upregulated in Krasmutant acinar cells, and the mevalonate pathway supports ADM formation. Specific knockout of the acetyl-CoA-producing enzyme ATP-citrate lyase in the pancreas suppresses ADM and carcinogenesis. The acetyl-CoA regulates histone acetylation and facilitates cell plasticity and proliferation (63). In acute or chronic pancreatitis, inhibition of histone deacetylase (HDAC) can decrease the formation of ADM and inflammatory cell infiltration (128), suggesting that it is feasible to inhibit ADM development by modulations of epigenetics. Also, PC patients having a higher expression of HDAC1 have lower survival than those having a lower expression. High expression of HDAC1 correlates with advanced stages of and reflects the malignancy of PC (64).

Histone ubiquitination
The expression levels of ubiquitination significantly increase in the nucleus and cytoplasm in premalignant atypical acinar cells and cells of malignant adenocarcinoma in situ (129). Ring1b catalyzes histone modification H2AK119ub and silences acinar cell transcription factors, which is a key mechanism for triggering acinar cell dedifferentiation and PC development (66). Ring1b expression increases in PC (130), and knockout of acinar-specific Ring1b attenuates the formation of ADM induced by Kras (65).
B cell-specific Moloney murine leukemia virus insertion site 1 (BMI1) belongs to the polycomb group of proteins, including polycomb repressive complex 1, which consists of BMI1, RING1, and RING2 and acts as an E3-ubiquitin ligase. Upregulation of BMI1 in early PanINs may be a protective response to Kras-driven oxidative stress. BMI1 is highly expressed in human and murine PC (131, 132). BMI1-knockout mice have an increased expression of the Ink4a/Arf locus and a decreased b-cell mass and impaired glucose tolerance. The depletion of Bmi1 greatly decreases the levels of H2AK119Ub at the Chk2 locus. In pancreatic carcinogenesis, the role of Bmi1 is independent of the Ink4a/Arf expression, because its downregulation abrogates PanIN formation even in Ink4a-null animals (67). In addition, high expression of H2AK119Ub1 is significantly associated with poor prognosis and metastasis (66).

Chromatin remodeling
Chromatin remodeling affects gene transcription by changing the binding of transcription factors to chromosomes. BRG1 is the ATPase subunit of the SWI/SNF chromatin remodeling complex and is often inactivated in PC (68). In the presence of Kras oncogenic mutations, duct-specific BRG1 deletion leads to the formation of IPMN (69, 133). BRG1 binds to the Sox9 promoter region, regulates its expression, and recruits Pdx-1 through chromatin conformation changes (134). ARID1A, another component of the SWI/SNF complex, has also been frequently mutated in PC (70). ARID1A deletion changes the cancer precursor spectrum from PanIN to IPMN and accelerates the onset of invasive cancer (71).

Phosphorylation in pancreatic premalignancy
S6 ribosomal protein (S6) phosphorylation is involved in PC cell proliferation. As compared to that in normal cells, the expression of S6 phosphorylation is significantly upregulated in IPMN and is associated with glucose uptake increase in malignant cells of IPMN (9). Therefore, its expression level can be detected in pancreatic juice as a biomarker of malignancy of IPMN.
pSmad3C gradually downregulates from low-grade dysplasia and high-grade dysplasia to malignant IPMN, whereas pSmad3L upregulates. There are inverse relationships between the expression of pSmad3C and that of c-Myc, and between pSmad3C and Ki-67. There are positive relationships between the expression of pSmad3L and that of c-Myc, and between pSmad3L and Ki-67 (135).
As a member of the ezrin and moesin protein family, Ezrin acts as a cross-linker between the cell membrane and actin cytoskeleton. Ezrin is activated by threonine and tyrosine phosphorylation (72). The expression of phosphorylated ezrin (p-ezrin, tyr354) is higher significantly in IPMN than in invasive carcinoma. Moreover, p-ezrin (tyr353) is highly expressed in all grades of PanINs. The p-ezrin (tyr354) expression is associated with invasion in IPMNs, while p-ezrin (tyr353) expression plays an important role in early PanIN development (73).

Epigenetic biomarkers for early PC detection and diagnosis
The discovery of specific biomarkers is essential for the detection of PC premalignancy and diagnosis of PC at its early stages (136). Currently, only the serum carbohydrate antigen 19-9 (CA19-9) is a Food and Drug Administration (FDA)-approved marker for clinical treatment efficacy in PC (137). However, CA9-9 has limitations, including ineffectiveness, low sensitivity, and low specificity. Other biomarkers such as carcinoembryonic antigen (CEA) and cancer antigen 125 (CA125) are ineffective as early PC biomarkers (138). Evidently, epigenetic biomarkers have many potential advantages including stability in serum and non-invasive, economical, and convenient detection (Table 2; Figures 1, 3).

DNA methylation biomarkers
DNA methylation changes during the development of cancer. The methylation usually occurs in the regulatory region of the gene and leads to the inactivation of the gene. Small amounts of DNA are released into the blood and can be detected in a blood sample. The DNA methylation changes can be tumor-specific and potentially usable tumor biomarkers. The alteration in hypermethylation of DNA from plasma is a diagnostic biomarker for PC (NCT02079363). Hypermethylation of cysteine dioxygenase 1 (CDO1) promoter is specific to IPMN and may increase with IPMN tumor progression, suggesting that CDO1 is a potential diagnostic marker of pancreatic cystic diseases (139).

MiRNAs and circRNAs as biomarkers 8.2.1 Plasma-associated miRNAs
Plasma miRNA profiling could be a sensitive and specific bloodbased biomarker assay for PC in the clinic. MiR-155, miR-196a, miR-  (140). The combination of CA19-9, miR-16, and miR-196a in plasma is more effective for distinguishing PC from non-PC (normal or chronic pancreatitis), especially in early tumor screening (141). MiRNA-483-3p and miRNA-21 from blood plasma could be PC biomarkers. MiR-483-3p expression is significantly increased in PC patients than in IPMN patients (142).

Tissue-associated miRNAs
MiRNA-155 and miRNA-21 are also significantly overexpressed in IPMN lesions (40). Locked nucleic acid in situ hybridization has verified that the expression of miRNA-155 is much higher in IPMNs than in normal ducts and that the expression of miRNA-21 is much higher in IPMNs than in normal ducts. The specificity of miRNA-155 upregulation as a biomarker of IPMNs is observed in 60% of IPMN cases in pancreatic juice samples compared to 0% of healthy controls (143).
MiR-196b is overexpressed selectively in PanIN-3 and PC lesions as compared with lower-grade PanINs and normal pancreatic ducts. These samples are isolated by laser capture microdissection. The expression of miR-196b is restricted in PanIN-3 lesions and could be an early diagnostic marker (102). The expression of hsa-miR-96 and hsa-miR-217 is significantly downregulated with increasing grades of PanINs and IPMNs as compared with normal pancreatic ductal epithelium in formalin-fixed paraffin-embedded specimens. They may serve as potential early biomarkers (144).

8.2.2.1
Ongoing clinical trials (selected data from ClinicalTrials.gov) There are clinical trials on epigenetic biomarkers for the early detection of PC (Table 2). For example, NCT04765410 analyzes miRNA profile (84 miRNAs) using qRT-PCR array in every participant to identify the correlation between microRNA expression in PC tissue and tumor aggressive behavior, treatment response, and patient survival. Samples are obtained from the participants through fine-needle aspiration during endoscopic ultrasonography. NCT03432624 tries to improve the efficacy of PC diagnosis by combined detection of microRNA, conventional tumor markers, and imaging. NCT04406831 determines whether the detection of certain miRNAs present in the blood of PC patients is important to the early diagnosis of the disease and whether miRNA detection in PC patients helps predict treatment response and produces prognostic information.
In the NCT02531607 trial, the patients are assigned to control, non-carcinoma, carcinoma non-pancreatic, and pancreatic ductal adenocarcinoma. Comparative efficacy of biomarkers, brush cytology, CA19-9, and biliary and blood biomarkers including lipidomics, miRNAs, and proteomics is determined for the diagnosis of pancreatic malignancy.
In the NCT04584996 trial, paired samples of PC tissue and associated normal pancreatic tissue will be collected after pancreatic surgery. The expression levels of circRNAs will be determined, and significantly dysregulated candidate circRNAs will be selected. Candidate circRNA expression in blood plasma samples as a clinically diagnostic biomarker in PC and superior sensitivity than serum CA19-9 will be evaluated. Moreover, the expression of candidate circRNAs in patient tissue, blood, bile, and biopsy samples will be explored as biomarkers for PC diagnosis. Finally, bioinformatic data will be reviewed to determine the abilities of miRNAs to bind candidate circRNAs.

Therapeutic implications of epigenetic regulators
Unlike genetic changes, epigenetic alterations can be easily reversed by pharmacological manipulations. Numerous studies have focused on understanding the key epigenetic regulators that regulate these epigenetic processes and developing molecule inhibitors targeting these regulators. Increasing clinical trials are conducted on epigenetic alterations as valuable targets to develop cancer therapies (Table 3; Figure 1).
Many clinical trials on epigenetic regulators are being carried out for treating PC according to the data from ClinicalTrials.gov. The DNMT inhibitors 5-azacytidine and decitabine can trap DNMT proteins during the S-phase and degrade the trapped DNMTs and hypomethylated CpG islands of tumor suppressor genes (145,146). Azacitidine and decitabine are used for hematological cancers in clinics.
There are clinical trials on the epigenetic treatment of PC (Table 3). For example, NCT04257448 aims to determine the safety and tolerability of HDAC inhibition by romidepsin and DNMT inhibition by azacitidine or both agents, in combination with gemcitabine/nab-paclitaxel in patients with advanced PC. Similarly, a phase II trial (NCT01845805) is ongoing to improve outcomes in patients with resected PC and high recurrence risk using oral azacitidine therapy. This trial is for patients who have resected PC and have completed adjuvant therapy or who are unable to receive adjuvant therapy due to an increased CA19-9 or advanced disease. Moreover, the phase I trial of 5-azacitidine plus gemcitabine (NCT01167816) is for patients having advanced PC. The primary objective is to define the maximum tolerated dose (MTD) of gemcitabine and azacitidine in patients with unresectable and previously untreated PC and to determine the effect of azacitidine therapy on DNA methylation in peripheral blood cells.
In the NCT02847000 trial, patients with PC who have stopped responding to one or more chemotherapy drugs are asked to take part in this study on decitabine therapy that decreases DNMT1 protein level. The study hopes to find out whether decitabine will have an effect on PC. Decitabine is also being given with another drug, tetrahydrouridine (THU), to improve the exposure of the PC cells to decitabine. Phase II trials (NCT04705818) are conducted in parallel to assess the efficacy of durvalumab when prescribed with tazemetostat, separately, in advanced PC. Because preclinical data in a PC mouse model demonstrates an improvement in survival with the combined use of a hypomethylating agent and immune therapy, NCT03264404 phase II study is designed to determine the effectiveness of combining immune therapy, pembrolizumab, with a hypomethylating agent, azacitidine, for locally advanced or metastatic PC. Another phase I study (NCT01478685) is to use CC-486 alone or in combination with ABI-007 or carboplatin in patients with relapsed or refractory solid tumors.
The bromodomain and extra-terminal domain (BET) binds to acetylated histone lysine residues within chromatin and interacts with gene promoters/enhancers (147). After binding, BET proteins recruit transcription complexes, thereby promoting gene transcription. INCB054329 and INCB057643 are new oral BET inhibitors that are tested in phase I and II clinical trials for patients with advanced cancer (NCT02431260 and NCT02711137) (148).
Vorinostat (Zolinza) is an HDAC inhibitor. In this first study (NCT00948688), vorinostat will be given along with radiation therapy (RT) and 5-FU to determine whether vorinostat combined with radiation and 5-FU may improve PC treatment. The phase I and II trials (NCT00831493) determine the MTD of vorinostat plus radiation therapy in patients with locally advanced PC. Another phase I trial (NCT00983268) studies the side effects and optimal dose of vorinostat when used in combination with capecitabine and radiation therapy for patients with non-metastatic PC. In the NCT02349867 phase I study, concurrent chemoradiation uses a regimen of sorafenib and vorinostat with gemcitabine and radiation followed by chemotherapy in patients with PC to find the recommended phase II dose of the concurrent chemoradiation combination. The U.S. FDA has approved vorinostat (Merck & Co., Inc.) to treat cutaneous manifestations in patients with progressive, persistent, or recurrent cutaneous T-cell lymphoma disease or following two systemic therapies (149).
Phase II (NCT05053971) will determine the MTD of entinostat and BET bromodomain inhibitor ZEN-3694 (ZEN003694) in combination in patients having advanced and refractory PC, and the overall response rate of entinostat and ZEN003694 in advanced/ progressive PC. Similarly, the phase II clinical trial (NCT03250273) of combined use of entinostat and nivolumab is for patients having metastatic or previously treated unresectable PC.
Panobinostat (LH589) produces diverse effects on endothelial cells and suppresses tumor angiogenesis. Bortezomib (Velcade) triggers cell death in PC cells. A phase II study (NCT01056601) of panobinostat in combination with bortezomib focuses on Patients with PC progressing on gemcitabine monotherapy or gemcitabine in combination. Moreover, a phase II trial (NCT00004861) compares the effectiveness of gemcitabine with or without tacedinaline (CI-994) in treating patients with advanced PC.
Finally, Atu027 is a liposomal formulation of siRNA targeting protein kinase N3 and can suppress cancer growth. In vitro Atu027 suppresses PKN3 function in endothelial cells and impairs their tube formation (150). Disease stabilization has been achieved in 41% of patients at end of treatment (151). In the NCT01808638 clinical trial, combined use of Atu027 and gemcitabine appeared to be safe and well tolerated. In patients with advanced PC, twice-weekly Atu027 significantly improved outcomes (152).

Conclusion and future perspectives
Effective management of PC urgently needs early detection through the discovery of highly sensitive and specific biomarkers for pancreatic premalignancy and novel interventional strategies with precise therapeutic targets. PanIN, MCN, and IPMN are the main pancreatic premalignant lesions. PanINs are the most frequent and too small to detect clinically. Identification and validation of precursor biomarkers are fundamentally important. Epigenetics are important regulatory mechanisms in pancreatic premalignant, including cellular proliferation and apoptosis, making them appealing candidates for diagnostic, therapeutic, and prognostic biomarkers. The expression profiles of various epigenetics molecules offer an exciting and promising opportunity for biomarker development for early diagnosis of PC. In this review, we describe several epigenetic biomarkers with the potential for early PC detection and diagnosis, including methylation of CDO1, miR-21, miR-25, miR-155, and miR-483-3p. However, there is no sufficient evidence to support the clinical use of these candidate biomarkers because of their mediocre performance and lack of thorough validation. Nonetheless, the performance of some investigational biomarkers could be improved through combination with existing tumor biomarkers (e.g., serum CA19-9) and radiological features (137,138).
DNA methylation is a major epigenetic program regulating gene transcription. Aberrant hypermethylation of CpG islands in the promoters/enhancers of tumor suppressor genes and hypomethylation of oncogenes are key to early tumorigenesis (153,154). m 6 A is the most abundant reversible methylation in mammalian mRNA and plays crucial roles in tumorigenesis. Histone post-translational modification and chromatin remodeling are the most complicated types of epigenetic programs and are crucial in PC development and progression (122).
Likewise, miRNAs in pancreatic premalignancy regulate cell proliferation, growth, metastasis, invasion, and apoptosis (41,43). The direct targeting of miRNAs provides a novel insight into the development of diagnostic biomarkers and targeted therapy of PC. Currently, only the microRNA-25 and miRNA profiles have been in clinical trials, while not a single miRNA biomarker is applied in the clinic for the detection and diagnosis of PC or premalignant lesions. Investigation into the regulation of miRNAs within pancreatic premalignant lesions using an effective and credible approach remains a major challenge. Both lncRNAs and circRNAs are emerging and promising molecules that need further systematical studies for their full potential as biomarkers in pancreatic premalignant lesions.
Preclinical studies and clinical trials have shown that epigenetic alterations are potentially reversible through the use of DNMT, EZH2, BET, and HDAC inhibitors. However, it remains to determine whether PC patients benefit from receiving epigenetic inhibitors with improved survival, and none of the available epigenetic regulators that entered clinical trials has passed phase II. Nevertheless, the potential of targeting epigenetic regulators should be systematically investigated, particularly in combination with chemotherapy, radiotherapy, immunotherapy, and targeted therapy to increase therapeutic efficacy and avoid toxicity.
With improving knowledge and understanding of the molecular basis on the initiation and progression of pancreatic premalignancy and with potential identification and validation of novel diagnostic biomarkers and therapeutic targets, it becomes possible not only for early detection and diagnosis of PC but also for screening of patients with a high risk for developing a malignant transformation and clinical management of these patients before the development of high-grade dysplasia or invasive PC.

Author contributions
WZ, TJ, and KX reviewed the literature for the article and wrote, reviewed, and/or edited the manuscript before its submission. All authors contributed to the article and approved the submitted version.