Epigenetic Silencing of 15-Hydroxyprostaglandin Dehydrogenase by Histone Methyltransferase EHMT2/G9a in Cholangiocarcinoma

This article is featured in Highlights of This Issue, p. 355

Cholangiocarcinoma (CCA) is highly aggressive malignancy of the biliary tract with poor prognosis. The incidence and mortality rate of CCA are rising worldwide; currently, there is no effective chemoprevention or systemic therapy (1, 2). Being considered as a prototype of cancers associated with inflammation, CCA typically forms in the setting of cholestasis and chronic inflammation that cause reparative injury and biliary epithelial cell proliferation (1–3). Accordingly, primary sclerosing cholangitis with chronic inflammation of the biliary tree is the most common predisposing condition for CCA in the Western world (1, 2).

Prostaglandin E₂ (PGE₂) is a potent lipid mediator that elicits a wide range of biological effects associated with inflammation and cancer (3). Studies from our group (3–10) and others (11–15) have shown that PGE₂ synthesized by cylooxygenase-2 (COX-2) plays an important role in CCA progression. Besides PGE₂ biosynthesis, the status of PGE₂ signaling is modulated by 15-hydroxyprostaglandin dehydrogenase (15-PGDH) which catalyzes the oxidation of the 15(S)-hydroxyl group of PGE₂ thus eliminating the biological activity of PGE₂ (16, 17).

The significance of 15-PGDH in carcinogenesis is attested by the fact that 15-PGDH mRNA and protein levels are significantly decreased in multiple human cancers (18–21), including CCA (5, 9). These clinical observations support an important anticancer effect of 15-PGDH. In our previous studies (5, 9), we have shown that 15-PGDH inhibits CCA cell growth in vitro and in xenograft models and that the expression of 15-PGDH in CCA cells is regulated by miRNA-21 and miRNA-26a/b at posttranscriptional level. However, the role of 15-PGDH in CCA development has not been previously investigated and it remains unknown whether and how the expression of the 15-PGDH gene is controlled through transcriptional mechanisms.

Recent studies from our group and others suggest a potentially important role of histone modification in the carcinogenesis and progression of CCA (22–25). Given that epigenetic modification is a key mechanism for the silence of antitumor genes and that CCA lacks clinically significant mutations in the 15-PGDH gene, we consider a high priority to explore whether epigenetic mechanisms may be implicated for regulation of 15-PGDH gene expression in CCA.

In the current study, we provide novel evidence that G9a (also known as Euchromatic histone-lysine N-methyltransferase 2, EHMT2), a H3K9 monomethyltransferase and dimethyltransferase, is recruited specifically to the promoter of the 15-PGDH gene by the Myc/Max/E-box complex which consequently represses 15-PGDH expression through methylation of H3K9 and prevention of pSTAT4 binding to 15-PGDH promoter element. Our data indicate that these mechanisms are impotently implicated in the regulation of CCA cell growth and invasion, in vitro and in vivo. Moreover, our experimental findings provide the first evidence that 15-PGDH prevents CCA development which further support 15-PGDH as a key antitumor molecule in cholangiocarcinogenesis.

Human CCA cell lines CCLP1 and SG231 were cultured in DMEM (Invitrogen) containing 10% FBS (Sigma-Aldrich) and antibiotics (100 U/mL penicillin and 100 μg/mL streptomycin) in a humidified 5% CO₂ incubator at 37°C. Cells within 10 passages were used for experiments. All cell lines were latest authenticated by short tandem repeat DNA profiling analysis at ATCC in Oct 2021 and each excluded as having Mycoplasma contamination using the LookOut Mycoplasma PCR Detection Kit (Sigma-Aldrich) regularly. For establishment of CCA cells with stable depletion of G9a, cells were transfected with pSMP-G9a short hairpin RNA (shRNA) or the control vector (pSMP-Luc shRNA). After 48 hours of transfection, the cells were cultured in DMEM containing 1 μg/mL Puromycin (Calbiochem). The selection medium was replaced every 3 days for the next 4 weeks. Subsequently, distinct colonies of surviving cells were transferred onto 6-well plates and the cultures were maintained under the same selection medium. The efficiency of G9a depletion was verified by Western blotting analysis using anti-G9a antibody.

Eight-week-old mice (C57BL/6J or FVB, purchased from The Jackson Laboratory) were anesthetized by isoflurane inhalation through vaporizer to approximately 3% to 5% for induction and approximately 1% to 3% for maintenance with oxygen flow to approximately 0.5–1.0 L/minute. The anesthetized mouse was placed into the restraining tube. Tail vein injections were performed by using a 27-gauge needle (BD). Briefly, 10 μg pT3-EF1a-NICD, 5 μg pT3-EF1a-myr-Akt-HA, 1.25 μg pCMV(CAT)T7-SB100, and 15 μg of pT3-EF1a-15PGDH or vector control plasmids were diluted in 10% body weight volume (mL) of sterile PBS and injected into a lateral tail vein of mouse within 5 to 7 seconds. The injected mice were closely monitored and sacrificed at indicated timepoints. The liver and body weights were recorded, and the livers were harvested for further analyses. All the mouse experimental procedures in this study were approved by the Institutional Animal Care and Use Committee of Tulane University (New Orleans, LA).

Five-week-old male Non-Obese Diabetic CB17-Prkdc/severe combined immunodeficiency (NOD.CB17-Prkdc^SCID/J) mice were purchased from The Jackson Laboratory. G9a knockdown or control CCLP1 cell suspensions were mixed with BD Matrigel high concentration matrix (BD Biosciences) at 1:1 ratio, and then cell suspension mixtures (1 × 10⁶ cells/10 μL) were injected directly into the livers of mice using a Hamilton syringe (model 75 N SYR; Hamilton Co.). The mice were observed closely for tumor formation. At sacrifice, the livers were dissected surgically and the tumor volume was calculated by the following formula: Tumor volume = Width² × Length/2.

Chromatin immunoprecipitation (ChIP) assay was performed using a SimpleChIP Enzymatic Chromatin IP kit (Cell Signaling Technology) according to the manufacturer's instructions. Each immunoprecipitation was performed using 1 μg of following antibodies: anti-H3K9me2 (#4658), anti-Myc (#9402), anti-Max (#4739), anti-G9a (#3306; from Cell Signaling Technology), anti-TFIID (sc-421), anti-c-Ets1 (sc-55581), anti-pSTAT4 (sc-28296), anti-Foxp3 (sc-53876), anti-HNF1A (sc-393925), anti-HNF3a (sc-101058; from Santa Cruz Biotechnology). Isotype normal IgG was used as a negative control. ChIP primers were designed to amplify the different sites in 15-PGDH promoter regions by real-time PCR. The primers are as follows: P_set1 F1, 5′- GAG CAA GGA ACC TCT GTC CC -3′; R1, 5′- TGT CAT CAT CAA CAG GCG CT -3′; P_set2 F2, 5′- TGC TTA GCG GCT TAC CAA CA -3′; R2, 5′- GGG GAA ATG GGA GTT GAG CA -3′; P_set3 F3, 5′- CCT GGT GCT CAC CTG AGT ATT -3′; R3, 5′- ATG GCA ACA TGC TGG GAG AA -3′; P_set4 F4, 5′- CTG GAC AGT GGC AGT GGA AA -3′; R4, 5′- AGC AAG GAC TGA GGT CTA GAG AA -3′. Bio-Rad SsoAdvanced Universal SYBR Green Supermix was used to amplify the target DNA fragments on a Bio-Rad C1000 Thermal Cycler.

The growth of cells was measured by WST-1 assay. CCA cells (2 × 10³ per well) with or without G9a depletion were seeded in 96-well plates at least in triplicate and treated with vehicle control or 15-PGDH inhibitors (sw033291 or dmPGE₂). At the indicated timepoint, the culture medium was removed, and cells were incubated with 100 μL of serum-free medium containing 10 μL of WST-1 reagents for 2 hours at 37°C. The absorbance of each sample was measured at 450 nm by using an automatic enzyme-linked immunosorbent assay plate reader (VersaMax Microplate Reader; Molecular Devices).

A total of 5 × 10³ CCA cells with indicated modification were plated in 10-cm dishes and then the cells were cultured for 14 days to allow colony formation. The colonies were fixed with 100% methanol and then stained with 0.1% crystal violet solution. For sw033291 or dmPGE₂ treatment, the indicated cells (5 × 10³) were plated in 10-cm dishes and treated with the indicated doses of sw033291, dmPGE₂ or vehicle. After 14 days of cell culture, colony formation was determined by crystal violet staining.

Cells or tissues were lysed in a RIPA buffer (50 mmol/L Tris-HCl pH 8.0, 150 mmol/L NaCl, 5 mmol/L EDTA, and 1% NP-40 with protease inhibitor) supplied with phosphatase and protease inhibitors cocktails (Roche). After being denatured, samples were separated on SDS-PAGE gel and then transferred onto the nitrocellulose membrane (Bio-Rad). The milk-blocked blots were incubated with different primary antibodies at 4°C overnight. After 3× washing with PBS-T, blots were incubated with IRDye 800CW or 680LT conjugated secondary antibodies (LI-COR Biosciences) and scanned using LI-COR Odyssey Imaging system. The following primary antibodies were used: anti-15PGDH (#160615) from Cayman Chemical, anti-Myc (#9402), anti-pAkt (#9271), anit-G9a (#3306), anti-H3K9me2 (#4658), anti-STAT4 (#2653) from Cell Signaling Technology, anti-GAPDH (#G8795), anti-Flag (#F3165), anti-β-actin (#A5316) from Sigma-Aldrich, and anti-pSTAT4 (sc-28296) from Santa Cruz Biotechnology.

Thymidine Incorporation Assays were performed using Click-iT EdU Alexa Fluor 594 Imaging Kit from Thermo Fisher Scientific according to the manufacture manual. Briefly, CCA cells with or without G9a depletion were incubated with sw033291 (5 μmol/L), dmPGE₂ (5 μmol/L) or vehicle control for 24 hours. After that, cells were incubated with 5 μmol/L EdU (5-ethynyl-2′-deoxyuridine) in serum-free DMEM for 24 hours. Then cells were fixed by 3.7% formaldehyde in PBS, followed by detection of incorporated EdU with Alexa 594-labeled azide according to the manual. Nuclei were counterstained by Hoechst 33342.

Microarray datasets GSE76297 for CCA and nontumor liver samples gene expression generated by Affymetrix Human Transcriptome Array 2.0 array platform were identified and downloaded from GEO (Gene Expression Omnibus of NCBI). The raw data were processed using oligo and related R packages with Robust Multi-array Average approach for background normalization as per the package instruction. Heatmap was generated by using R package ComplexHeatmap. For transcription factor (TF) binding sites identification, 15-PGDH gene upstream sequence (−2,000 to +1) was put into algorithm PROMO as a query sequence to search for potential binding sites of putative TFs (26).

Results are presented as mean ± SD from a minimum of three replicates. Difference between groups was evaluated by SPSS 19.0 statistical software (IBM) with one-way analysis of variance, repeated measure analysis of generalized linear model or Wilcoxon signed-rank test. A P value < 0.05 was considered as statistically significant.

Studies from our lab have shown that 15-PGDH inhibits the growth of CCA cells in vitro and in xenograft models (5, 9), although the impact of 15-PGDH on CCA development is not known prior to the current study. We sought to further determine the role of 15-PGDH in CCA development by employing a mouse model of CCA induction via hydrodynamic tail vein injection of sleeping beauty (SB) transposase-based plasmids expressing Notch1 intracellular domain (NICD) and constitutively active Akt as described previously (27). To this end, we constructed a transposase-based 15-PGDH plasmid and co-injected it hydrodynamically with NICD and Akt plasmids into wild-type mice. As shown in Fig. 1A, while hydrodynamic tail vein injection of NICD/Akt/SB vectors rapidly induced the development of CCA, we observed that mice injected with 15-PGDH in combination with NICD/Akt/SB plasmids (15-PGDH group) developed fewer and smaller tumor nodules when compared with the control mice (injected with control vector in combination with NICD/Akt/SB). The mice in the 15-PGDH group showed significantly lower liver-to-body weight ratios and better survivals when compared with the mice in the control group (Fig. 1B and C). IHC staining for the cell proliferation marker Ki67 showed lower percentages of Ki67-positive cells in 15-PGDH–overexpressed tumors than in control tumors (Fig. 1D). Immunoblotting analysis confirmed successful expression of the injected plasmids in the liver tumor tissues (Fig. 1E). These findings demonstrate that 15-PGDH overexpression effectively prevents CCA development, and the results establish an important tumor inhibiting role of 15-PGDH in cholangiocarcinogenesis.

Epigenetic gene silencing is one of the key mechanisms underlying dysregulation of tumor-suppressor genes in cancers (28, 29). For the regulation of 15-PGDH expression, previous studies have reported a potential role of 15-PGDH promoter hypermethylation in gastric and colon cancer cells (30, 31). In our system, we treated human CCA cells (CCLP1) with the DNA methyltransferase inhibitor 5-Aza-2′-deoxycytidine (5-Aza-CdR) and observed that the level of 15-PGDH expression in CCLP1 cells was not altered by (5-Aza-CdR) treatment (Supplementary Fig. S1A). In parallel, we analyzed DNA-methylation and corresponding RNA-sequencing data of patients with CCA from TCGA-CHOL dataset (including nine nontumor tissues and 36 CCA tissues). Our analyses indicate that 38.9% (14/36) patients have consistent hypermethylation of CpG island located around the transcription starting site of 15-PGDH gene (denoted by Composite Element cg00906130, cg02822257, cg07372795 and cg11073923 in Infinium HumanMethylation450 BeadChip). However, the hypermethylation status of the indicated CpG island have little impacts on 15-PGDH mRNA level in CCA tissues, as indicated by the negligible correlations between the methylation levels (denoted by beta-scores) of these four elements and 15-PGDH mRNA levels in CCA patients’ tissues (Supplementary Fig. S1B). These findings suggest that DNA methylation does not appear to represent a predominant mechanism for regulation of 15-PGDH gene expression in CCA.

Our data presented in the current study provide novel evidence for an important role of the histone methyltransferase (HMT) G9a in regulation of 15-PGDH gene expression in CCA. We analyzed the expression levels of all 53 known human HMTs (32) in CCA and nontumor tissues of dataset GSE76297, which comprises gene expression data from 91 intrahepatic CCA and 92 adjacent nontumor tissues (including 90 tumor-nontumor tissue pairs; ref. 33). We observed that 22 HMTs were significantly increased (>1.2 folds) in CCA tissues compared with nontumor tissues (as shown in Fig. 2A; Supplementary Table S1), which include G9a. Our analyses indicate that G9a is the most negatively correlated HMT with 15-PGDH in patients’ tissues (Fig. 2B) after assessing the correlation between the mRNA levels of 15-PGDH and these 22 increased HMTs. The negative correlation of these two mRNAs was further verified by using additional dataset GSE107943, which comprises gene expression data from 30 CCA and 27 adjacent nontumor tissues (including 27 tumor-nontumor tissue pairs; ref. 34; Fig. 2B).

Consistent with the above-described results in human data, we detected a noticeable reduction of 15-PGDH and a significant increase of G9a in mouse CCA induced by hydrodynamic tail vein injection of NICD/Akt/SB vectors (Fig. 2C and D). In our initial study describing the role of G9a in CCA (25), we showed that treatment with the G9a specific inhibitor, UNC0642, significantly inhibited CCA growth and reduced tumor burden in NICD/Akt-injected mouse model (25). Using the CCA tissues derived from those mice for further immunoblotting, we observed that inhibition of G9a by UNC0642 increased the expression of 15-PGDH in CCA tissues (Fig. 2E).

We next assessed the effect of G9a on 15-PGDH expression in cultured human CCA cells (CCLP1 and SG231) transfected with G9a shRNA plasmid or treated with the G9a inhibitor UNC0642. Successful inhibition of G9a HMT activity by G9a shRNA or UNC0642 was reflected by decline of H3K9me2 levels (Fig. 2F and G). Western blotting analyses indicate that G9a depletion by shRNA or G9a inhibition by UNC0642 increased the expression of 15-PGDH in CCA cells (Fig. 2F and G). These findings demonstrate an important role of G9a for suppression of 15-PGDH expression in CCA cells.

Next, we sought to investigate whether G9a depletion would impact tumorigenic potential of CCLP1 cells in vivo. For this purpose, 1 × 10⁶ CCLP1 cells with or without G9a knockdown were mixed with matrigel and inoculated orthotopically into the livers of SCID mice through liver injection. We observed that, by 7 weeks post-injection, the recipient mice receiving G9a depleted CCLP1 cells developed much smaller tumor nodules when compared with the mice receiving control CCLP1 cells (Fig. 3A and B). Given that PGE₂ plays an important role in tumor invasion and metastasis (35), we further examined the intrahepatic metastasis in the tumor-adjacent liver tissues. We observed that the mice receiving G9a-depleted CCLP1 cells exhibited significantly less metastatic foci when compared with the group receiving control CCLP1 cells (Fig. 3C). Western blotting assays confirmed reduced G9a and elevated 15-PGDH protein levels in tumors formed by G9a-depleted CCLP1 cells (Fig. 3D).

To further assess the functional implication of 15-PGDH in G9a-mediated CCA cell growth, we evaluated whether 15-PGDH inhibition could rescue the inhibition of CCA cell growth induced by G9a-depletion. For this purpose, we utilized CCA cell lines with stable G9a knockdown (as shown above in Fig. 2F, successful reduction of G9a protein by shRNA in the stable cell lines were verified by immunoblotting analysis). We observed that G9a depletion by shRNA significantly decreased the proliferation of CCA cells when compared with their corresponding control cells (Fig. 4A and B). G9a knockdown also reduced the clonogenicity of CCA cells (Fig. 4C). Importantly, treatment with two separate 15-PGDH inhibitors, including 16,16-dimethyl PGE₂ (an agonist of prostaglandin E receptor and a competitive inhibitor of 15-PGDH) or sw033291 (a noncompetitive inhibitor of 15-PGDH), rescued the proliferation and clonogenic capability of G9a-depleted CCA cells (Fig. 4A–C). These results, along with the data that G9a depletion increased 15-PGDH levels in mouse CCA (Fig. 3D), demonstrate the role of 15-PGDH in G9a-mediated CCA progression.

Given that G9a catalyzes the monomethylation and dimethylation of H3K9 with high specificity (36), we further performed ChIP assay by using anti-H3K9me2 antibody and primer pairs specific for 15-PGDH promoter. We observed that 15-PGDH promoter fragments were greatly enriched by anti-H3K9me2, and this enrichment was repressed significantly by UNC0642 treatment (Fig. 5A and B). Thus, the H3K9 residues in 15-PGDH promoter region are heavily dimethylated, which support the role of G9a in the silencing of 15-PGDH gene expression. Our findings suggest that G9a inhibits 15-PGDH expression in CCA through modifying the methylation status of H3K9 in the promoter of 15-PGDH gene.

Next, we sought to further investigate how G9a was specifically recruited to the promoter region of the 15-PGDH gene. Recently, Dingar and colleagues had explored Myc/Max interactome systematically by using BioID (proximity‐dependent biotin identification) mass spectrometry in multiple cell types; it came to our attention that G9a is included as a high-confidence interactor with Myc/Max heterodimer (37, 38). Through bioinformatics analyses, we have found that there are four putative Myc/Max-bound E-box elements (CAGCTG×3 and CACCTG×1) in human 15-PGDH promoter region (−2,000 to +1; ref. 39), as indicated in Fig. 5A. On the basis of these observations, we postulate that Myc/Max heterodimer could mediate the recruiting of G9a to 15-PGDH promoter. To validate this hypothesis, we performed ChIP experiments with anti-Myc, anti-Max, or anti-G9a antibodies in CCA cells. Our data showed that the 15-PGDH promoter fragments were significantly enriched by all of the above three antibodies (Fig. 5C and D). These observations indicate the association of these proteins with the 15-PGDH promoter. Next, we carried out co-immunoprecipitation assays to determine the association of Myc/Max with G9a in CCA cells. As shown in Fig. 5E, these three proteins form complex in CCLP1 cells. Taken together, our findings suggest that Myc/Max complex recruits G9a to the promoter region of the 15-PGDH gene which leads to methylation of the H3K9 residues and silencing of 15-PGDH gene expression.

Our next goal was to identify TFs that can activate 15-PGDH transcription and can be inhibited by G9a in CCA cells. We performed bioinformatics analyses using PROMO algorithm with strictest screening criteria (26); this approach led to the identification of six TFs (TFIID, c-Ets-1, STAT4, Foxp3, HNF1A, and HNF3a) which have putative binding sites located at the upstream (−2,000 to +1) of the human 15-PGDH gene (Fig. 6A). We then carried out ChIP assay with a set of four primer pairs to evaluate the effects of G9a depletion on the interactions between the predicted TFs and the 15-PGDH gene promoter. While the ChIP results revealed the binding of TFIID, c-Ets-1, and pSTAT4 to the 15-PGDH promoter, we observed that only pSTAT4 binding was significantly boosted by G9a knockdown in CCLP1 cells (with consistent phosphorylation status of STAT4), which was further verified in SG231 cells (Fig. 6B–D; Supplementary Fig. S2A). We observed that STAT4 protein levels were significantly decreased in human CCA tissues when compared with the nontumor tissues (Supplementary Fig. S2B). On the basis of these findings, we evaluated whether forced overexpression of STAT4 would enhance 15-PGDH expression in CCA cells. Indeed, we observed a considerable rise of 15-PGDH protein level in CCLP1 cells transfected with STAT4-expressing plasmid (Fig. 6E). To further determine whether there was a correlation between STAT4 and 15-PGDH in CCA tissues, we performed coexpression analyses between these two genes in CCA patients’ tissues from dataset GSE76297; this analysis revealed positive correlation between STAT4 and 15-PGDH in human CCA tissues (Fig. 6F). These observations, in combination with the data that G9a inhibition decreases H3K9 methylation level (Fig. 2E–G) and reduces H3K9me2 association with 15-PGDH promoter (Fig. 5B) suggest that G9a-mediated H3K9 methylation prevents the binding of pSTAT4 to its target element in the 15-PGDH promoter.

Taken together, our findings disclose a novel mechanism for regulation of 15-PGDH gene expression in CCA cells: G9a is recruited to 15-PGDH promoter by Myc/Max/E-box complex and the recruited G9a catalyzes H3K9 methylation which prevents the binding of TFs (i.e., pSTAT4) leading to transcriptional repression of 15-PGDH gene expression (outlined in Fig. 7). Our data support an important role of this novel regulatory mechanism in cholangiocarcinogenesis.

PGE₂ synthesis involves cyclooxygenase-1 and -2 (COX-1 and -2). While COX-1 is constitutively active in most tissues to maintain physiologic level of PGE₂, COX-2 is highly inducible in various tissues by pro-inflammatory and mitogenic stimuli, including cytokines and growth factors (3, 40). Compelling evidence has shown that COX-2–derived PGE₂ plays an important role in promoting cell survival, stimulating cell proliferation and motility, inducing angiogenesis, and suppressing immune surveillance (3, 4, 6, 7, 11–13, 15, 17, 41, 42). Studies from several laboratories have shown that PGE₂ synthesized by COX-2 plays an important role in CCA progression (4, 6, 7, 11, 12, 14, 15, 41). PGE₂ is known to activate G protein–coupled receptors and then initiates diverse intracellular events in CCA, such as protein kinase activation and proinflammatory cytokines production, which consequently lead to increased cell growth and tumor invasion (3). In addition to the function-based ligand-receptor interaction, COX2/PGE₂ signaling also interacts with other key pathways in CCA, such as iNOS (43), EGFR (8, 14), and Wnt/β-catenin pathway (4).

In addition to biosynthesis, PGE₂ level is tightly controlled by its metabolic degradation. The NAD⁺-dependent 15-PGDH catalyzes the oxidation of the 15(S)-hydroxyl group of PGE₂ which leads to inactivation of PGE₂ (16, 17). Previous studies from our group (5, 9) have shown that (i) 15-PGDH mRNA and protein levels are decreased in CCA; (ii) 15-PGDH inhibits CCA cell growth in vitro and in xenograft models, and (iii) the expression of 15-PGDH in CCA cells is regulated by miRNA-21 and miRNA-26a at posttranscriptional level. However, the role of 15-PGDH in CCA development has not been previously investigated and it remains unknown whether and how the expression of the 15-PGDH gene is controlled through transcriptional mechanisms.

In the current study, by employing a novel mouse CCA model induced by hydrodynamic tail vein injection of NICD/Akt plasmids, we observed that overexpression of 15-PGDH efficiently prevent CCA development. These experimental findings support 15-PGDH as an important antitumor molecule in cholangiocarcinogenesis. These results, along the existing data on 15-PGDH in CCA, may have important clinical relevance, given that 15-PGDH is much more specific than selective COX-2 inhibitors with respect to PGE₂ signaling inhibition. Different from 15-PGDH, whose main metabolic substrates are limited to prostaglandins and eicosanoids (16), selective COX-2 inhibitors may have severe off-target effects by interacting with a broad spectrum of molecules such as cGMP phosphodiesterase, PPARs, NFκB, Akt, and 15-lipoxygenase-1 (3). Our findings suggest that induction of 15-PGDH may represent an effective substitute approach for COX-2 inhibitors treatment in chemoprevention and adjuvant anticancer therapy.

Epigenetic modifications are characteristic of all cancers, from apparently normal precursor tissue to advanced metastatic disease, and these epigenetic modifications confer growth advantage upon tumor cells through disrupting the functions of diverse genes (28, 29). In the current study, our findings reveal an important epigenetic mechanism for regulation of 15-PGDH expression in CCA cells. Specifically, our results indicate a key role of the HMT EHMT2/G9a for inhibition of 15-PGDH gene expression in CCA. The latter assertion is supported by the steady increase of 15-PGDH expression in G9a-depleted CCA cells in culture and in mice; and by the findings from bioinformatics analyses indicating that G9a and 15-PGDH mRNA levels are inversely correlated in human CCA tissues.

In addition to epigenetic modifications, genetic mutation is the other key mechanism that silences tumor suppressors in cancer. However, we were unable to identify any mutation in 15-PGDH exons after performing data mining analysis across more than 450 patients with bile tract cancer from various publications based on different high-throughput platforms (44, 45). Instead, our findings in the current study all point toward the notion that G9a-mediated histone methylation is a key mechanism for 15-PGDH repression in CCA.

The chromatins of many tumor suppressor genes in cancer are marked by both DNA methylation and histone H3K9 methylation. Although the exact relationship between these epigenetic modifications is not fully elucidated, it is becoming clear that histone methylation occurs during the initial phases of gene silencing in cancer, and DNA methylation might then spread over the gene promoter in the process of cell division to maintain associated histone modifications and gene silence (28, 46, 47). Compelling evidences in various experimental systems with deficient DNA methyltransferase function have already shown that G9a alone can induce gene silence through methylating H3K9 (46–48). For example, H3K9 methylation associated with silencing of p16^INK4 gene can occur prior to and independently of DNA methylation in colon cancer (47); G9a can block Oct-3/4 gene reactivation through methylating H3K9 independent of its Dnmt3a/b recruiting function in embryonic cells (48). These facts are in agreement with our observation that G9a depletion/inhibition effectively alters 15-PGDH expression, while inhibition of DNA methyltransferase by 5-Aza-CdR exhibits little effect on 15-PGDH level in CCA cells. All of these phenomena fit well with the view that G9a is a master structural regulator with an important role in early development and carcinogenesis (46, 48). In this context, our experimental findings are noteworthy as they provide further evidence in support of G9a as a promising epigenetic oncotarget for developing inhibitors as anticancer therapeutic drugs (49). In this context, is worth mentioning that we have demonstrated in our previous study that the G9a inhibitor UNC0642 effectively prevents the development of CCA induced by hydrodynamic tail vein injection of NICD and Akt plasmids in mice (25).

Another novel aspect of this study is the identification of a functional connection between STAT4 and 15-PGDH in CCA. According to the Human Protein Atlas database, increased STAT4 is considered as a favorable prognostic maker in most types of human cancer (50). While previous studies have focused on the activation of STAT4 signaling by cytokines in immunocytes, the detailed functional role of STAT4 in cancer cells has not been fully considered. Our results presented in this study reveal a possible mechanism underlying tumor-inhibiting role of STAT4 in cancer cells, which is to up-regulate the transcription of its target gene 15-PGDH. These facts support a tumor-inhibitory role of STAT4 in CCA.

In addition, we would like to reiterate the inhibiting effect of c-Myc on 15-PGDH in CCA. In our previous study, we have demonstrated that c-Myc is involved in 15-PGDH repression through regulating miR-26a/b and their host genes (9). Here, we describe a separate novel mechanism underlying the repressing effect of c-Myc on 15-PGDH expression. Specifically, our data demonstrate that Myc/Max, when binding to E-box elements located in the promoter of 15-PGDH can recruit G9a to form a DNA-protein complex. Then, this complex drives the transcription repression of 15-PGDH through methylating adjacent H3K9 residues.

In summary, our findings presented in the current study disclose a novel mechanism of epigenetic regulation of 15-PGDH by G9a in CCA. Therefore, disrupting the c-Myc–elicited, G9a-mediated transcriptional repression axis and/or inducing the reexpression of 15-PGDH may represent a potentially effective strategy for the chemoprevention and adjuvant therapy of CCA.

J. Zhang reports grants from Department of Defense during the conduct of the study. No disclosures were reported by the other authors.

J. Zhang: Conceptualization, data curation, funding acquisition, investigation, writing–original draft. W. Chen: Data curation, formal analysis, investigation. W. Ma: Resources, investigation. K. Song: Investigation. S. Lee: Resources. C. Han: Resources, project administration. T. Wu: Conceptualization, resources, data curation, supervision, funding acquisition, project administration, writing–review and editing.

This work was supported by the NIH grants CA102325, CA219541, CA226281, and Department of Defense grant CA180361.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.