Previous studies in our laboratory identified that 3-deazaneplanocin A (DZNep), a carbocyclic adenosine analog and histone methyl transferase inhibitor, suppresses TGFβ-induced epithelial-to-mesenchymal (EMT) characteristics. In addition, DZNep epigenetically reprograms miRNAs to regulate endogenous TGFβ1 levels via miR-663/4787-mediated RNA interference (Mol Cancer Res. 2016 Sep 13. pii: molcanres.0083.2016) (1). Although DZNep also attenuates exogenous TGFβ-induced EMT response, the mechanism of this inhibition was unclear. Here, DZNep induced miR-202-5p to target both TGFβ receptors, TGFBR1 and TGFBR2, for RNA interference and thereby contributes to the suppression of exogenous TGFβ-induced EMT in pancreatic cancer cells. Lentiviral overexpression of miR-202 significantly reduced the protein levels of both TGFβ receptors and suppressed TGFβ signaling and EMT phenotypic characteristics of cultured parenchymal pancreatic cancer cells. Consistently, transfection of anti-miRNAs against miR-202-5p resulted in increased TGFBR1 and TGFBR2 protein expressions and induced EMT characteristics in these cells. In stellate pancreatic cells, miR-202 overexpression slowed growth as well as reduced stromal extracellular membrane matrix protein expression. In orthotopic pancreatic cancer mouse models, both immunodeficient and immunocompetent, miR-202 reduced tumor burden and metastasis. Together, these findings demonstrate an alternative mechanism of DZNep in suppressing TGFβ signaling at the receptor level and uncover the EMT-suppressing role of miR-202 in pancreatic cancer.
Implications: These findings support the possibility of combining small molecule–based (e.g., DZNep analogs) or large molecule–based (e.g., miRNAs) epigenetic modifiers with conventional nucleoside analogs (e.g., gemcitabine, capecitabine) to improve the antimetastatic potential of current pancreatic cancer therapy. Mol Cancer Res; 15(8); 1029–39. ©2017 AACR.
One of the key reasons for the aggressiveness of pancreatic tumors is attributed to a tumor cell remodeling process called epithelial-to-mesenchymal transition (EMT; refs. 2–4). Mechanistically, EMT is characterized by the loss of cell–cell contacts and cell polarity and gain of mesenchymal phenotypes, including increased motility, invasiveness, and chemoresistance (2–4). TGFβ1 is a multifunctional cytokine secreted into the tumor microenvironment that primarily promotes the EMT process (5, 6). Although TGFβ1 is a growth-inhibitory cytokine to normal cells, parenchymal tumor cells exposed to TGFβ do not exhibit growth inhibition, but instead favor increased tumor cell dissemination, colonization, and clonal expansion with acquisition of drug refractiveness (5–9). In addition, TGFβ induces significant stromal reorganization to provide a structural framework necessary for supporting tumor malignancy (10, 11). Thus, a therapeutic strategy to curtail the TGFβ-triggered EMT process is expected to improve pancreatic cancer therapy, although several challenges exist for translational advancements in this direction.
TGFβ signaling is initiated by the binding of the TGFβ ligand to its receptor (TGFBR2; ref. 12). The receptor–ligand complex recruits TGFBR1 to phosphorylate downstream signaling players including receptor-associated Smad2 and Smad3 (R-Smads; ref. 12). The R-Smads further complex with a coSmad, Smad4, to translocate to the nucleus, which leads to the activation of critical transcription factors governing EMT. These include the Snail family of zinc-finger transcription factors (Snail1, Snail2, and Snail3), Smad-interacting proteins (ZEB2), and basic helix-loop-helix factors (Twist1 and Twist2; ref. 13). In addition to the canonical players involved in the Smad-dependent pathway, TGFβ also works in concert with several noncanonical players, including those in the PI3K/AKT, MEK/ERK, and β-catenin/WNT pathways, to coregulate the EMT process (13). Several mutations, frameshift deletions, and monoallelic losses of genes involved in the TGFβ signaling pathway are reported to contribute to aberrant functional consequences of TGFβ signaling in cancers (14).
In addition to genetic alterations, epigenetic alterations, including DNA and histone methylations, are now increasingly being recognized to play critical roles in the interconversions of epithelial–mesenchymal plasticity. In particular, EZH2 histone methyl transferase (a part of the polycomb repressive complex 2) that mediates trimethylation of lysine 27 in histone H3 was recently reported to epigenetically silence genes essential in maintaining an epithelial phenotype (15, 16). In addition, overexpression of EZH2 protein or mutations in the EZH2 gene is also linked to the acquisition of EMT characteristics and cancer progression in tumor settings (17). These findings support the notion that targeting of EZH2 could be of potential therapeutic value for treating EMT-active tumors such as pancreatic tumors. As a result, synthetic EZH2 inhibitors are currently being evaluated in preclinical and clinical trials for various solid tumors (18), although their applicability in pancreatic cancer treatment remains less understood.
An earlier study in our laboratory demonstrated 3-deazaneplanocin A (DZNep), a carbocyclic analog of adenosine that depletes cellular levels of EZH2 to presensitize pancreatic cancer cells to nucleoside analog drugs (19). Recently, we reported DZNep to inhibit TGFβ-induced EMT characteristics of pancreatic cancer cells (1). The latter study also revealed DZNep to induce distinct epigenetic reprogramming of miRNAs, a class of small noncoding RNAs that downmodulate target gene expression, to attenuate endogenous synthesis of TGFβ1 in parenchymal pancreatic cancer cells (1). Particularly, two DZNep-reprogrammed and pancreatic ductal adenocarcinoma (PDAC)-downregulated miRNAs, miR-663a and miR-4787-5p, were identified to directly target TGFβ1 for RNA interference (1). As a result, DZNep inhibited not only the synthesis of TGFβ1 but also its secretion into the extracellular environment and its associated autocrine and paracrine responses. Intriguingly, in addition to attenuation of endogenous TGFβ, we observed DZNep to inhibit exogenous TGFβ-stimulated EMT response in pancreatic cancer cells (1). However, the mechanism(s) regarding this process was not known and suggested the presence of other modes of action.
In this study, we addressed whether any of the DZNep-induced miRNAs were capable of attenuating exogenous TGFβ1-stimulated EMT responses in pancreatic cancer cells. We report the DZNep-induced miR-202-5p to target both TGFB receptors, TGFBR1 and TGFBR2, for RNA interference to attenuate the exogenous TGFβ1-induced EMT response at the receptor level. MiR-202 also inhibited cell migration and invasion as well as metastasis in orthotopic pancreatic cancer mouse models. Together, these data uncovered a novel mechanism of DZNep action in suppressing the EMT phenotype and further support the pharmacotherapeutic evaluation of miR-202 for the intervention of pancreatic cancer progression.
Materials and Methods
Materials and reagents, cell culture, Western blotting, real-time PCR analysis, miRNA overexpression, knockdown of miRNAs, scratch wound assay, MTT growth assays, and RT2 profiler PCR array
These procedures were carried out as described previously (1, 20). A breast cancer cell line, MCF-7, was received from the American Type Culture Collection (ATCC) and grown in DMEM supplemented with 10% FBS. The murine PDAC cell line, PAN 02, was procured from National Cancer Institute DCTDC Tumor Repository and was grown in RPMI 1640 media supplemented with 10% FBS and 2 mmol/L l-Glutamine. The details of KPC cell line established from a genetically engineered mouse model of pancreatic cancer, PDX-1-CRE, LSL-KRasG12D, LSL-Trp53−/−, are described earlier (21, 22), and it was grown in DMEM (without sodium pyruvate) supplemented with 10% FBS. Primary human pancreatic stellate cells (PSC; #3830) were purchased from ScienCell Research Laboratories.
Growth conditions of PSCs
PSCs were grown in stellate cell medium (ScienCell Research Laboratories, #5301) supplemented with FBS, Stellate Cell Growth Supplement, and penicillin/streptomycin solution (provided with the media) in poly-lysine–coated flasks and were used in the study within 10 passages. Procedures, including miRNA-202 overexpression, in PSCs and subsequent analysis, including Western blotting, MTT growth assays, and real-time PCR analysis, were carried out in a similar fashion as for other pancreatic cancer cell lines. Collection of PSCs-conditioned media was carried out as per procedure described elsewhere (23).
The rabbit polyclonal TGFBR1 (sc-398), TGFBR2 (sc-400), collagen 1alpha1 (sc-8784-R), and fibronectin (sc-9068) antibodies, the mouse monoclonal Smad3 (sc-101154) antibody, and the goat polyclonal collagen 3alpha1 (sc-8781) antibody were purchased from Santa Cruz Biotechnology, Inc. The rabbit polyclonal p-Smad2 (#3101), rabbit monoclonal Smad2 (#3122), and vimentin (#5741) antibodies and the mouse monoclonal cytokeratin 8/18 (#4546) antibody were purchased from Cell Signaling Technology. The mouse E-cadherin antibody was a kind gift from Dr. Parmender Mehta (University of Nebraska Medical Center, Nebraska). The mouse monoclonal N-cadherin antibody was purchased from BD Biosciences. The mouse monoclonal α-smooth muscle action (SMA) (A2547) and β-actin (A1978) antibody were purchased from Sigma-Aldrich.
Primers and constructs
The Taqman primer probes for EZH2 (Hs01016789_m1), TGFBR1 (Hs00610320_m1), TGFBR2 (Hs00234253), CTGF (Hs01026927_g1), GUSB (Hs00939627_m1), miR-202-5p (002362), pri-miR-663a (Hs03304850_pri), miR-4787-5p (464332_mat), and RNU6B (001093) were purchased from Applied Biosystems. Pre–miR-202 lentiviral constructs (HmiR0166-MR03) for overexpressing miR-202-5p and the miRNA scrambled control clone (CmiR0001-MR03) were purchased from GeneCopoeia, Inc. The GIPZ lentiviral EZH2 shRNAs (RHS4531-EG2146) were purchased from Dharmacon GE. For knockdown studies, miR-202-5p inhibitors (Assay ID: MH12682) were purchased from Applied Biosystems.
Luciferase in vitro reporter assay
The luciferase reporter assay was carried out as described elsewhere (20) with some modifications. Overall, 5 × 103 cells (empty vector or miR-202-5p overexpressing) were seeded in a 96-well plate. Within 24 hours, cells were transfected with 0.1 μg of TGFBR1 (ID: HmiT064498-MT01, GeneCopoeia) or TGFBR2 (ID: HmiT018053-MT01, GeneCopoeia) 3′ untranslated region (3′UTR) dual-luciferase vector and treated with DZNep (10 μmol/L), anti–miR-202-5p (50 nmol/L), or solvent. Within 24 to 48 hours, cells were lysed and subjected to the Dual-Luciferase assay (Promega) as per the manufacturer's instructions.
Orthotopic pancreatic tumor models
Animal experiments were performed as per the protocol approved by the Institutional Animal Care and Use Committee at the University of Georgia and The Ohio State University. Orthotopic pancreatic cancer xenograft mice models were generated as described elsewhere (1). Briefly, cells grown to 90% confluency in culture dishes were trypsinized, centrifuged, and resuspended at a concentration of 1 × 105 cells/100 μL in sterile HBSS containing 1% v/v serum-free Matrigel. Eight-week-old female athymic nude mice (homozygous; 002019; Jackson Laboratories) were injected with cells orthotopically in the tail region of the pancreas. The cell suspension was allowed to solidify, and mice were recovered and monitored for changes in their weight twice a week. Eight weeks after cell injection, tumors were excised and assessed for weight and size for end-point assays. Tumor volume was measured with a digital caliper and calculated by the formula: tumor volume = π/6 (L × S × S), where L is the longest diameter and S is the shortest diameter of the tumors. In addition, the liver was monitored for metastatic tumor lesions, and tissue sections at the normal-tumor interface were collected for histologic examination of metastatic foci. For establishing syngeneic orthotopic model of pancreatic cancer, 1 × 106 KPC cells were injected in 8- to 10-week-old female C57BL/6 mice as described above. Three weeks after cell injection, tumors were excised and assessed for weight. For another orthotopic model of pancreatic cancer in immunocompetent mice, 0.5 × 106 PAN 02 cells were injected in 8- to 10-week-old female C57BL/6 mice as described above. Nine weeks after cell injections, tumors were weighed, tumor progression was monitored, and metastasis and hemorrhagic ascites were assessed.
Data analysis was conducted using GraphPad Prism 6 (Graphpad Software, Inc.). The statistical significance of the data was determined using the Student t test between two groups or using one-way (one variable) or two-way (two independent variables) ANOVA with multiple comparisons test for multiple experimental groups. All P values <0.05 were considered significant unless otherwise stated. P values <0.05, <0.01, and <0.005 are indicated with *, **, and ***, respectively, unless otherwise stated.
DZNep inhibits TGFβ receptor expressions and induces miR-202-5p expression in pancreatic cancer cells
DZNep's resistance to exogenous TGFβ1-induced EMT characteristics could be due to the functional inhibition of TGFβ receptors or intracellular Smad proteins. To test this possibility, we assayed for the changes in the expression levels of TGFβ receptor I (TGFBR1), TGFβ receptor II (TGFBR2), Smad2, and Smad3 in control- and DZNep-treated cells. As shown by Western blotting analysis in Fig. 1A, DZNep distinctly reduced the expression of both TGFBR1 and TGFBR2 in PANC-1 without altering the expression of Smads (Smad3 in Fig. 1A; Smad2 shown in previous study; ref. 1). Similar results were also obtained in another pancreatic cancer cell line (BxPC-3) and a breast cancer cell line (MCF-7; both of which are previously shown to undergo EMT on TGFβ treatment) as well as in other pancreatic cancer cell lines (Capan-1, L3.6pl, and MIA PaCa-2) and murine PDAC cell lines (KPC and PAN 02; Supplementary Fig. S1A). DZNep inhibited TGFBR1 and TGFBR2 protein expressions in a dose-dependent manner within a concentration range of 1 to 50 μmol/L (72-hour treatment; Fig. 1B). These results identify DZNep's inhibition of TGFBR1 and TGFBR2 protein expression and possibly also DZNep's action at the receptor level to inhibit exogenous TGFβ-stimulated EMT response.
Because our previous studies identified DZNep to significantly reprogram miRNA expression in human PDAC (1), we postulated that one or more of the DZNep-induced miRNA(s) could be involved in the regulation of TGFβ receptor expressions. Bioinformatic analysis predicted DZNep-induced miRNAs, miR-4646-5p and miR-361-5p, to directly target TGFBR1, whereas miR-1273g-3p, miR-19a-3p, and miR-3182 were predicted to directly target TGFBR2 for RNA interference (1). Furthermore, DZNep-induced miR-320a/c/d and miR-202-5p were predicted to target both TGFBR1 and TGFBR2. As DZNep maximally induced miR-202-5p (∼7-folds; microarray data) as compared with miR-320 members (∼2-folds; microarray data) and because two independent algorithms (miRDB and microRNA.org) predicted TGFBR1 and TGFBR2 as direct targets for miR-202-5p (in contrast to only TargetScan for miR-320 members), we further directed our research on miR-202-5p. Real-time PCR analysis of DZNep-treated (10 μmol/L; 36 hours) cells validated that DZNep induced the expression levels of miR-202-5p by at least 4- to 5-fold in pancreatic cancer cells in which it is otherwise generally poorly expressed (Fig. 1C). Similarly, DZNep induced miR-202-5p levels in BxPC-3 and MCF-7 (TGFβ–EMT-responsive cell lines) as well as in KPC cells (Supplementary Fig. S1B). Furthermore, DZNep induced miR-202-5p levels in PANC-1 in a dose- and time-dependent fashion with maximal induction at 10 μmol/L for 36 to 48 hours of treatment (Fig. 1D and E). In addition, EZH2 shRNAs (Fig. 1F) increased miR-202-5p levels, suggesting that DZNep's actions are at least partially mediated via EZH2 inhibition (Fig. 1G).
miR-202-5p directly targets TGFBR1 and TGFBR2 for RNA interference
Sequence analysis showed two complementary binding regions to the miR-202-5p seed sequence within the 3′UTRs of TGFBR2 and a single binding site within the 3′UTR of TGFBR1 (Fig. 2A). To experimentally test whether miR-202-5p binds to the 3′UTRs of TGFBR1 and/or TGFBR2 for targeting to RNA interference, we transiently transfected PANC-1 cells stably expressing miR-202-5p (Fig. 2B–D) with a dual-luciferase reporter construct in which the TGFBR1 3′UTR or TGFBR2 3′UTR were cloned downstream of a firefly luciferase cDNA (see Materials and Methods). PANC-1 cells were chosen for this analysis because they endogenously expressed abundant levels of both TGFBR1 and TGFBR2 and harbored no receptor mutations or loss of SMAD expression. In addition, our earlier studies identified DZNep to significantly reprogram miRNA expression in PANC-1 (1). As shown in Fig. 2E, both miR-202-5p overexpression and DZNep treatment significantly reduced luciferase activities demonstrating a functional role of miR-202-5p on DZNep inhibition of TGFBR1 and TGFBR2 expressions. Supportively, the reductions in luciferase activities by DZNep and miR-202-5p were effectively counteracted by a miR-202-5p antagomiric construct (50 nmol/L for 48-hour treatment; Fig. 2E). Western blotting analysis using cell lysates of miR-202-5p–overexpressing PANC-1 showed a reduction of both TGFBR1 and TGFBR2 protein expressions, similar to that observed with DZNep (Fig. 2F). Consistently, depletion of endogenous miR-202-5p in PANC-1 cells by transfection of anti–miR-202-5p constructs (50 nmol/L for 48-hour treatment) increased the protein expression of both receptors when compared with that in control vector-transfected cells (Fig. 2F). These results were independently verified in MIA PaCa-2, wherein miR-202-5p expression was the most abundant strand (Supplementary Figs. S2 and S3). However, the changes were observed only for TGFBR1 protein levels and not TGFBR2 (due to lower expression of TGFBR2) with miR-202-5p alterations in MIA PaCa-2. Finally, we investigated if there was any correlation between miR-202-5p and its targets, TGFBR1 and TGFBR2, in pancreatic cancer. For this, we compared the expression levels of miR-202-5p with that of TGFBR1 and TGFBR2 (Supplementary Fig. S2) in a panel of pancreatic cancer cells to determine the Pearson's correlation coefficient. As shown in Fig. 2G, miR-202-5p expression levels negatively correlated with both TGFBR1 and TGBR2 with Pearson r = −0.529 and −0.470, respectively. Overall, these results support that DZNep-induced miR-202-5p directly targets TGFBR1 and TGFBR2 for RNA interference in pancreatic cancer.
miR-202 resists TGFβ1-induced EMT in pancreatic cancer
Because both TGFB receptors are indispensable for TGFβ1-mediated actions (e.g., EMT) and miR-202-5p expression inhibited TGFB receptor expressions, we directly tested whether miR-202-5p can counteract TGFβ1-mediated EMT in pancreatic cancer (as seen with DZNep; ref. 1). To do so, we treated PANC-1 transfected with empty vector or stably overexpressing miR-202-5p with TGFβ1 and analyzed for possible changes in cellular morphology, EMT markers, and cell migration. As expected, recombinant-derived human TGFβ1 (10 ng/mL; 48 hours) induced distinct EMT-like morphologic changes as judged from scattering of cells, more spindle-shaped cells, elongated cellular processes, and reduced cell-to-cell contacts in empty vector–transfected cells (Fig. 3A). However, such morphologic changes were significantly inhibited in miR-202-5p–overexpressing cells (Fig. 3A) similar to that observed with DZNep, although the magnitude of inhibition of miR-202 was lesser than DZNep (1). In addition, cells stably overexpressing miR-202-5p appeared more epithelial-like (Fig. 3A) as compared with control cells, suggesting that miR-202-5p is resisting endogenous TGFβ1-induced changes as well. This is expected when miR-202 targets nascent TGFBR1 and TGFBR2 transcripts for RNA interference. Western blotting analysis of control-transfected cells treated with TGFβ1 (10 ng/mL; 48 hours) showed reduced expression of epithelial markers (e.g., E-cadherin) and increased expression of mesenchymal markers (e.g., N-cadherin and vimentin; Fig. 3B, left). However, such changes were absent or reduced, especially for changes in mesenchymal markers N-cadherin and vimentin, in miR-202-5p–overexpressing cells treated with TGFβ1 (Fig. 3B, left). Consistent with morphologic observations of miR-202-5p–overexpressing cells appearing more epithelial-like (Fig. 3A, left, bottom), these cells showed increased expression of the epithelial markers E-cadherin and cytokeratin 18 and reduced expression of the mesenchymal markers N-cadherin and vimentin (Fig. 3B). In addition, such changes were reduced or absent when miR-202-5p–overexpressing cells were transfected with anti–miR-202-5p (50 nmol/L; 48 hours; Fig. 3B, right), confirming miR-202-5p to play a role in EMT resistance.
Subsequent to biochemical characterization of EMT markers, we examined the role of miR-202-5p in modulating a key phenotypic EMT characteristic, i.e., cell migration. MiR-202-5p reduced cell migration in a wound-healing assay both in the presence and absence of TGFβ1 treatment (Fig. 3C and D). Further, to test whether miR-202 can significantly influence TGFβ-induced activation of TGF signaling in pancreatic cancer cells, we treated control or miR-202–overexpressing PANC-1 with TGFβ1 and observed for changes in the phosphorylation of Smad2. Western blotting analysis of control-transfected cells treated with TGFβ1 (10 ng/mL; 15 to 60 minutes) showed increased phosphorylated levels of Smad2 with no significant changes in total Smad2 levels, whereas in miR-202-5p–overexpressing cells, such increments in p-Smad2 levels were either absent or reduced (Fig. 3E). More importantly, anti–miR-202-5p treatment (50 nmol/L; 36 hours) rescued DZNep's effects on p-Smad2 levels in TGFβ1-treated PANC-1 (Fig. 3F), suggesting that DZNep's effects are at least partially mediated by miR-202-5p in pancreatic cancer. Since anti–miR-202-5p was observed to significantly antagonize both miR-202-5p and DZNep's actions, we assessed the effect of anti–miR-202-5p on miR-663a and miR-4787-5p (DZNep-induced miRNAs previously shown to target TGFβ1 ligand) in PANC-1 (1). As shown in Supplementary Fig. S4, anti–miR-202-5p was found to be specific to its target (miR-202) and did not alter the levels of TGFβ1-targeting miR-663a and miR-4787-5p in PANC-1. To extend the understanding of the effect of miR-202-5p on the players involved in the TGFβ signaling axis, we profiled several genes involved in TGFβ signaling using a nested PCR array (1). Analysis of Ct values of all 84 genes profiled in the RT-PCR array identified a strong positive pathway score of +0.865 (P < 0.001) in PANC-1 treated with TGFβ, indicating the activation of the TGFβ pathway as shown in our previous studies (1). In contrast, a negative TGFβ pathway score of (∼–0.6; P < 0.005) was determined in PANC-1 with miR-202-5p overexpression, thereby confirming the repressive effect of miR-202-5p on the TGFβ signaling axis (Fig. 3G–H). The extent of inhibition was much higher than that observed with either miR-663a (–0.4; P < 0.01) or miR-4787 (–0.2; miR-4787-5p; P = n.s.) which act on the TGFβ1 ligand, suggesting a stronger influence of DZNep on TGFβ signaling axis through acting on the receptors (1). In addition, MTT assays showed that miR-202-5p reduced the cellular proliferation rate in pancreatic cancer cells (Fig. 3I).
MiR-202-5p overexpression in PSCs slows growth, suppresses stromal-associated ECM proteins, and reduces viability of pancreatic cancer cells grown in PSC-conditioned media.
PSCs are the main cell types that form the fibrotic stroma in PDAC (24). TGFβ signaling plays an important role in the activation of PSCs that leads to an elevated production and accumulation of extracellular matrix (ECM) proteins and thereby fibrotic stromal deposition (11, 24). As such, the activated PSCs act as supporting cells for PDAC progression and metastasis. Because miR-202-5p suppressed TGFβ signaling in pancreatic cancer cells, we next wanted to investigate if the pathophysiologic role of miR-202-5p was similar in PSCs. We overexpressed miR-202-5p in PSCs (Fig. 4A) using a lentiviral gene transfer method (20) and evaluated the effects of miR-202-5p on various phenotypes in PSCs. To begin, we evaluated the effects of miR-202-5p on the growth of PSCs especially because we found miR-202-5p to suppress growth of pancreatic cancer cells (Fig. 3I). As shown by the MTT cell proliferation assay, miR-202-5p significantly slowed the growth of PSCs as compared with control cells (Fig. 4B). We next wanted to investigate whether miR-202-5p suppressed TGFβ signaling in PSCs and thereby activated PSC-mediated production of ECM proteins. Hence, we checked the levels of three ECM proteins (Collagen 1alpha1, collagen 3alpha1, and fibronectin), a PSC activation marker (α-SMA), and a downstream mediator of TGFβ1-induced fibrotic effect (CTGF, connective tissue growth factor). Indeed, overexpression of miR-202-5p reduced ECM protein levels in PSCs as compared with control cells (Fig. 4C, left). In addition, TGFβ1 induced levels of α-SMA and fibronectin, indicating that the activated state of PSCs was effectively counteracted with miR-202-5p overexpression in PSCs (Fig. 4C, right). This was also evident from the reduction in the TGFβ1-induced expression of CTGF by miR-202-5p in PSCs (Fig. 4D). Finally, to check the paracrine effects of miR-202-5p–overexpressing PSCs on the growth of the pancreatic cancer cells, we grew PANC-1 cells in conditioned media of miR-202-5p–overexpressing or control activated PSCs and measured cell viability. As shown in Fig. 4E, PANC-1 cells in conditioned media of miR-202-5p–overexpressing and activated PSCs showed a reduced growth as compared with control. Taken together, these results show that miR-202-5p suppresses the TGFβ signaling in PSCs and thereby the autocrine and paracrine effects of PSCs and pancreatic cancer cells.
Exogenous overexpression of miR-202 suppresses the growth and metastasis of pancreatic cancer xenografts in immunocompromised and immunocompetent mouse models
To study whether miR-202-5p can direct in vivo suppression of the TGFβ signaling–induced increase in metastasis, we investigated whether the in vitro effects of miR-202-5p could be recapitulated in athymic nude mice orthotopically implanted with PANC-1 cells stably overexpressing miR-202-5p or empty vector–transfected cells (Fig. 5A). As shown in Fig. 5B and C, mice transplanted with miR-202-5p–overexpressing cells at the end of 8 weeks show reduced tumor growth as compared with mice transplanted with empty vector–transfected cells. Consistently, average tumor weights (n = 3; Fig. 5D) and average tumor volumes (n = 3; Fig. 5E) were significantly reduced in mice with miR-202-5p–overexpressing xenografts as compared with control. Moreover, gross pathologic examination of the liver, one of the main sites for pancreatic tumor metastasis, showed the average number of metastatic lesions per mouse to be reduced in the miR-202-5p group of mice as compared with control group (Fig. 5F and G). Furthermore, 100% of mice (3 of 3) in the control group showed liver metastases compared with a reduction to approximately 33% (1 of 3) in the miR-202-5p group (Fig. 5H). Overall, there were no significant changes in mice body weights, indicating that miR-202-5p was well tolerated by the mice (data not shown). These results support that miR-202-5p can suppress metastasis in vivo in an orthotopic model of pancreatic tumor xenografts.
In order to evaluate if miR-202-5p exerted similar effects in a mouse model with an active immune system, we established a syngeneic mouse model of pancreatic cancer by orthotopically implanting KPC cells (from genetically engineered mouse model of pancreatic cancer) in the pancreas of C57BL/6 mice (Fig. 5I). Consistent with the human pancreatic xenograft model, miR-202-5p overexpression reduced tumor burden in the KPC syngeneic model as well (Fig. 5J and K). However, the KPC syngeneic mice developed extensive peritoneal carcinomatosis and dissemination along with intense hemorrhagic ascites and gastric obstruction (data not shown). Hence, these mice were sacrificed as early as on day 21 after cell implantation, at which none had developed liver metastasis. As a result, we were not able to evaluate the effect of miR-202-5p on liver metastasis in syngeneic model of pancreatic cancer. Hence, we carried out a similar study in another immunocompetent mouse model of pancreatic cancer which was developed in a similar fashion using PAN 02 cells (from carcinogen-induced mouse model of pancreatic cancer; Fig. 5L). Indeed, miR-202-5p not only reduced primary tumor burden (Fig. 5M and N) but also reduced the aggressiveness of PAN 02 pancreatic tumors as evidenced by the absence of hemorrhagic ascites as well as liver metastasis (Fig. 5O–Q). Overall, these results show that miR-202-5p can reduce tumor aggressiveness in mouse models with an active immune system as well.
Previous studies in our laboratory showed DZNep to suppress TGFβ1-induced EMT characteristics in PDAC (1, 19). We identified epigenetic reprogramming of miRNAs, especially miR-663a and miR-4787-5p, that directly targeted TGFβ1 for RNA interference, as a mechanism of DZNep inhibition of endogenous TGFβ1. However, the posttranscriptional mechanism of DZNep-mediated inhibition of endogenous TGFβ1 did not provide explanations for DZNep resisting exogenous TGFβ1-induced EMT effects. The current study contributes to the better understanding of the pharmacologic mechanisms of DZNep by revealing the miR-202-5p effect on TGFβ receptors. To our knowledge, this is the first study that demonstrates the direct effects of a histone methylation inhibitor (DZNep) in reducing both TGFBR1 and TGFBR2 protein expressions across a panel of PDAC cells. The mechanism of DZNep inhibition of TGFBR1 and TGFBR2 was attributed, at least partially, to miR-202-5p, which directly targeted both TGFBR1 and TGFBR2 3′UTRs for RNA interference. MiR-202 overexpression reduced the protein levels of both TGFBR1 and TGFBR2, and, consequently, resisted TGFβ1-induced EMT-like characteristics in PDAC. Thus, these findings uncover a novel mechanism of DZNep-mediated inhibition of TGFβ1-induced EMT in PDAC.
We suspected DZNep to affect TGFβ1 signaling by targeting either the TGFβ1 receptors or players downstream of TGFβ1 in the pathway. By screening endogenous expression profiles in basal states, we observed expressional differences in the endogenous levels of TGFB receptors (TGFBR1 and TGFBR2) and Smads (Smad2, Smad3, and Smad 4; data not shown) in cell lines representing various subtypes of PDAC. While this partially explains the inherent differences in TGFβ signaling noted between various PDAC cell lines, DZNep's effect was consistent in reducing the protein levels of TGFBR1 and TGFBR2 across multiple PDAC cell lines. However, DZNep effects were independent of Smad protein alterations. The strong reduction in TGFβ receptors expression and lack of alteration of Smad proteins are likely to be beneficial from a therapeutic perspective considering the tumor-suppressor effects of certain Smads in pancreatic cancer.
Next, DZNep's ability to induce several miRNAs that could target TGFB receptors also suggested a posttranscriptional mechanism of TGFBR1 and TGFBR2 inhibition. It is likely that several DZNep-induced miRNAs could target TGFBR1 and/or TGFBR2 and possibly mediate DZNep-induced EMT effects in PDAC. Particularly, we focused on miR-202-5p in this study since it was predicted to bind to the 3′UTRs of both TGFBR1 and TGFBR2 for RNA interference, and it was one of the miRNAs maximally induced by DZNep in PANC-1 (∼4–5-fold). Although miR-202-3p has been previously reported to be a tumor-suppressor miRNA for various types of cancer (e.g., hepatocellular carcinoma, osteosarcoma, and colorectal cancer; refs. 25–27), its role in any other types of cancer including pancreatic cancer is not investigated. Previous studies identified both miR-202-3p and miR-202-5p as evolutionarily conserved gonadal miRNAs regulated by a sex-determining transcription factor, SOX-9, with a potential role in gonadal differentiation (28). However, the current study for the first time presents miR-202-5p as potential growth-suppressing miRNA candidate that targets both TGFBR1 and TGFBR2 in PDAC. In addition, miR-202-5p suppresses the TGFβ signaling axis, reduces the expression of several genes associated with TGFβ signaling, and counteracts TGFβ1-induced EMT-like characteristics of PDAC. In this study, we overexpressed miR-202 with a lentiviral construct of pre–miR-202 that resulted in overexpression of both miR-202-5p and miR-202-3p strands. Interestingly, in addition to miR-202-5p, bioinformatics analysis also predicted miR-202-3p to target TGFBR1 3′UTR. Although we confirmed that the observed phenotypic effects of miR-202 were specific to miR-202-5p (using antagomiRs specific for miR-202-5p and not -3p), it is not unlikely that contributions may have also occurred from miR-202-3p. In addition to miR-202-5p, DZNep induced the expression of other miRNAs, including miR-4646-5p, miR-361-5p, miR-3182, miR-320a, miR-320c, miR-320d, miR-1273g-3p, and miR-19a-3p (1), which were also predicted to target TGFBR1 and/or TGFBR2. However, understanding the possible functional contributions of these miRNAs in mediating DZNep-induced inhibitions of TGFBR1 and TGFBR2, and thereby resistance to TGFβ-induced EMT, warrants further studies.
Although we have earlier demonstrated DZNep to reprogram miR-663a and miR-4787-5p to inhibit TGFβ1 (1), in this study, we found DZNep to induce miR-202-5p to inhibit TGFBR1 and TGFBR2 for RNA interference in PDAC (Fig. 6). Because silencing EZH2 with shRNAs moderately increased the levels of these miRNAs in pancreatic cancer cells, it is likely that the DZNep's action of miRNAs induction is partially via EZH2 inhibition. Taken together, these findings explain the role and mechanistic aspects of synthetic histone methylation reversal agents in inhibiting TGFβ signaling and TGFβ-induced EMT in PDAC. Indeed, as we complete our studies, another independent group very recently showed DZNep's effect in inhibiting TGFβ-induced activation of renal fibroblasts and deposition of ECM protein in an obstructed kidney by preventing Smad7 (an inhibitor of TGFβ signaling) degradation and reducing TGFBR1 expression (29). Hence, it is also likely that DZNep can also suppress TGFβ-induced activation of PSCs and fibrosis associated with stromal accumulation and deposition, both of which along with EMT are also known to contribute to the development of chemotherapeutic resistance of pancreatic cancer. Overall, these findings are consistent with our previous work which demonstrated DZNep to induce chemosensitizing properties in pancreatic cancer (19). Nonetheless, DZNep is a nonspecific, global histone methylation inhibitor reported to cause several unintended toxicities (e.g., nephrotoxicity) in mice that halted the clinical development of the compound (30). Hence, by understanding the mechanistic aspects of DZNep, further efforts were made in this study to identify the molecular targets of DZNep responsible for its beneficial therapeutic actions. On this basis, these findings so far further support the utilities of histone demethylation agents for combined epigenetic–chemotherapeutic combination therapies in pancreatic cancer. Next logical steps include combination of miR-663a/miR-4787-5p and/or miR-202-5p with specific chemotherapeutic agents to improve the overall effectiveness of pancreatic cancer therapy. In fact, it seems single nanoparticle formulations combining a cocktail of miRNAs (here miR-663a/miR-4787-5p targeting TGFβ1 and miR-202-5p targeting TGFBR1 and TGFBR2) seems a plausible strategy to combat the aberrant TGFβ signaling in pancreatic cancers.
In summary, our studies identify a nucleoside-derived compound, DZNep, and specific DZNep-induced miRNAs to attenuate TGFβ signaling by acting at the ligand and/or receptor levels. These findings support the possibility of combining small molecule–based (e.g., DZNep analogs) or large molecule–based (e.g., miRNAs) epigenetic reversal agents with conventional nucleoside analogs (e.g., gemcitabine, capecitabine) to improve the antimetastatic potential of current pancreatic cancer therapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: H.R. Mody, S.W. Hung, R.K. Pathak, R. Govindarajan
Development of methodology: H.R. Mody, S.W. Hung, R. Govindarajan
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.R. Mody, R.K. Pathak, J. Griffin, Z. Cruz-Monserrate
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.R. Mody, R.K. Pathak, R. Govindarajan
Writing, review, and/or revision of the manuscript: H.R. Mody, S.W. Hung, R. Govindarajan
Study supervision: R. Govindarajan
This work was supported by the award NIH 1R01CA188464 (to R. Govindarajan).
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.