Inhibition of S-Adenosylmethionine-Dependent ... · knockdown suggests novel mechanisms contribute...

Oncogenes and Tumor Suppressors

Inhibition of S-Adenosylmethionine-DependentMethyltransferase Attenuates TGFb1-InducedEMT andMetastasis in Pancreatic Cancer: PutativeRoles of miR-663a and miR-4787-5pHardik R. Mody2,3, Sau Wai Hung3, Mohammad AlSaggar3, Jazmine Griffin3, andRajgopal Govindarajan1,2,3

Abstract

The identification of epigenetic reversal agents for use incombination chemotherapies to treat human pancreatic ductaladenocarcinomas (PDAC) remains an unmet clinical need. Phar-macologic inhibitors of Enhancer of Zeste Homolog 2 (EZH2) areemerging as potential histone methylation reversal agents forthe treatment of various solid tumors and leukemia; however,the surprisingly small set of mRNA targets identified with EZH2knockdown suggests novel mechanisms contribute to theirantitumorigenic effects. Here, 3-deazaneplanocin-A (DZNep),an inhibitor of S-adenosyl-L-homocysteine hydrolase andEZH2 histone lysine-N-methyltransferase, significantly repro-grams noncoding microRNA (miRNA) expression and dam-pens TGFb1-induced epithelial-to-mesenchymal (EMT) signalsin pancreatic cancer. In particular, miR-663a and miR-4787-5pwere identified as PDAC-downregulated miRNAs that werereactivated by DZNep to directly target TGFb1 for RNA inter-

ference. Lentiviral overexpression of miR-663a andmiR-4787-5preducedTGFb1 synthesis and secretion inPDACcells andpartiallyphenocopied DZNep's EMT-resisting effects, whereas lockednucleic acid (LNA) antagomiRNAs counteracted them. DZNep,miR-663a, and miR-4787-5p reduced tumor burden in vivo andmetastases in an orthotopic mouse pancreatic tumor model.Taken together, these findings suggest the epigenetic reprogram-ming of miRNAs by synthetic histonemethylation reversal agentsas a viable approach to attenuate TGFb1-induced EMT features inhumanPDACanduncover putativemiRNA targets involved in theprocess.

Implications: The findings support the potential for synthetichistone methylation reversal agents to be included in futureepigenetic–chemotherapeutic combination therapies for pan-creatic cancer. Mol Cancer Res; 14(11); 1124–35. �2016 AACR.

IntroductionThe tendency of pancreatic tumors to metastasize to lymph

nodes, the abdominal cavity, and the liver at a very early stagecompromises the effectiveness of chemotherapeutic agents(1, 2). Hence, challenges to prevent tumor spread and drugresistance continue to exist for pancreatic cancer treatment.Evidence shows that epithelial-to-mesenchymal (EMT), distinctcellular events characterized by loss of epithelial phenotypesand gain of mesenchymal phenotypes, contributes to the earlydissemination of primary tumor cells and metastatic spread(2–4). Further, the process of EMT itself has been shown topromote stem-like characteristics of tumor cells and increase

drug resistance (2–4). Therefore, therapeutic strategies directedagainst EMT would likely prove useful in controlling metastasisand drug resistance in PDAC patients.

Transforming growth factor beta 1 (TGFb1) is one of the criticalcytokines known to drive EMT in pancreatic cancer (5, 6). It bindsand activates TGFb Receptor 2 (TGFBR2), propagating signalingthrough phosphorylation of Smad proteins and associating withtranscription factors to regulate gene expression (7). Although theprimary roles of TGFb1 are to inhibit epithelial cell proliferationand orchestrate mesenchymal development (8), loss of growthcontrol responses and activation of EMT become salient inadvanced tumors (9, 10). These outcomes are due to the loss ofnormal responsiveness to the ligand as a result of eithermutationsin TGFb receptors (e.g., TGFBR2) and/or intracellular Smadproteins (e.g., Smad4; refs. 11, 12).Other nongeneticmechanismshave also shown to contribute to aberrant TGFb signaling(13, 14). Evidence also presents that the aforementioned events,largely in a stochastic manner, allows transcriptional reprogram-ming of a network of EMT-related genes and alterations in thenucleosome state to promote a sustainedmesenchymal signature(9, 15).

Recent studies have uncovered novel epigenetic mechanismssuch asmethylation of lysine in histones to preferentially amplifypro-tumorigenic signaling of cancer cells to EMT (16–19). Inparticular, a striking correlation between TGFb1 and EZH2, ahistone methyl transferase enzyme in the Polycomb Repressive

1The Comprehensive Cancer Center, The Ohio State University,Columbus, Ohio. 2Division of Pharmaceutics and PharmaceuticalChemistry, The Ohio State University, Columbus, Ohio. 3Departmentof Pharmaceutical and Biomedical Sciences, The University ofGeorgia, Athens, Georgia.

Note: Supplementary data for this article are available at Molecular CancerResearch Online (http://mcr.aacrjournals.org/).

Corresponding Author: Rajgopal Govindarajan, The Ohio State University,Room R542, 500 West 12th Avenue, Columbus, OH 43210. Phone: 614-247-8269; Fax: 614-292-2588; E-mail: [email protected]

doi: 10.1158/1541-7786.MCR-16-0083

�2016 American Association for Cancer Research.

MolecularCancerResearch

Mol Cancer Res; 14(11) November 20161124

on January 14, 2021. © 2016 American Association for Cancer Research. mcr.aacrjournals.org Downloaded from

Published OnlineFirst September 13, 2016; DOI: 10.1158/1541-7786.MCR-16-0083

http://mcr.aacrjournals.org/

Complex 2 (PRC2) involved in silencing of genes via H3K27Me,was recognized to regulate EMT (20, 21). Other members of thePRC2 including EED and JARID2 have also been reported to driveTGFb-induced EMT, suggesting a possible widespread effect ofhistone methylation in the steps leading to EMT (22, 23). Unlikethe advancements in the biological understanding of EZH2 onTGFb-induced EMT, the cancer pharmacology of EZH2 inhibitorsis lagging behind. Despite new and more specific EZH2 inhibitorcandidates currently being evaluated in clinical trials for lympho-ma and multiple myeloma (24), their utilities in metastaticcontrol of solid tumors remains obscure. We have previouslyreported DZNep, a carbocyclic analogue of adenosine thatdepletes cellular levels of EZH2 and erases the H3K27Me mark,to curtail growth, proliferation, and anticancer nucleoside ana-logue resistance of pancreatic cancer cells (25). Here, we reportDZNep's role in dampening TGFb1-induced EMT features inPDAC. We present evidence for DZNep-induced miRNA repro-gramming as a mechanism for EMT inhibition in PDAC.

Materials and MethodsCell culture, MTT cytotoxicity assays, clonogenic assays,Western blotting, real-time PCR analysis, and MiRNAoverexpression

These procedures were performed as described previously(25, 26). All cell lines were received from ATCC and were usedwithin 25 passages, not exceeding a period of 2 to 3 months ofrevival. The ATCC uses morphological, cytogenetic and DNA pro-file analyses for characterization of cell lines. Human pancreaticductal epithelial (HPDE) cells were received fromDr. Ming Tsao ofthe Ontario Cancer Institute. The L3.6pl cell line was received fromDr. Isiah D. Fidler at The University of Texas MD Anderson CancerCenter. BothHPDE and L3.6pl cell lines were handled as other celllines and were genotyped by DNA fingerprinting (PowerPlex 16;Promega) as per the manufacturer's instructions. The growthconditions of cell lines were performed as described previously(26). Panc 10.05 was grown in RPMI-1640Mediumwith 15% FBSand 10 U/mL human recombinant insulin while CFPAC-1 wasgrown in Iscove's Modified Dulbecco's Medium with 10% FBS.

Tumor RNATotal RNA from normal pancreas and PDAC were procured

from Asterand (Supplementary Table S1).For materials, scratch wound assay, Transwell invasion assay,

ELISA, MicroRNA microarray, luciferase in vitro reporter assay,knockdown of miRNA, RT2 profiler PCR array, orthotopic pan-creatic tumor xenograft model (27), and statistical analysis, seeSupplemental Materials and Methods.

ResultsDZNep resists TGFb1-induced EMT in pancreatic cancer cells

To investigate DZNep effects on TGFb1-induced EMT, wetested TGFb1-induced changes in morphology and growth oftwo moderately–poorly differentiated PDAC cell lines, viz.,MIA PaCa-2 and PANC-1. Recombinant-derived human TGFb1(10 ng/mL; 72 hours) induced distinct EMT-like, morpholog-ical changes in both MIA PaCa-2 and PANC-1 (Fig. 1A) but notin normal HPDE (data not shown). More spindle-shaped cellswith elongated cellular processes and diminished cell-to-cellcontacts (Fig. 1A) as well as reduced expression of epithelialmarkers (E-cadherin and cytokeratin 8/18) and increased

expression of mesenchymal markers (N-cadherin and vimen-tin) were noted with TGFb1 treatment (Fig. 1B). TGFb1-induced EMT changes were independent of changes in cellproliferation in MIA PaCa-2 with only a slight growth reductionin PANC-1 (12.23% � 0.35%; P < 0.05; Supplementary Figs. S1and S2). On the contrary, TGFb1 significantly reduced cellproliferation in normal HPDE (30.25% � 1.99%; P < 0.005;Supplementary Fig. S1). These data confirmed the presence ofTGFb1-mediated EMT-like features in MIA PaCa-2 and PANC-1and were therefore used for further studies.

In subconfluent conditions, MIA PaCa-2 appears more plas-tic with the presence of both epithelial- and mesenchymal-looking cells while PANC-1 appears predominantly epithelial(Fig. 1A). Interestingly, DZNep treatment visibly increased theepithelial morphologic characteristics of both cell lines thatwere more readily identifiable in MIA PaCa-2 than PANC-1(Fig. 1A). Biochemically, DZNep increased expression of E-cad-herin and cytokeratin 8/18 and decreased expression of vimen-tin and N-cadherin in PANC-1 as judged by Western blottinganalysis (Fig. 1B). Similar results were also obtained in MIAPaCa-2 (except that N-cadherin was not expressed; Fig. 1B),L3.6pl, PANC10.05, CFPAC-1, and SW1990 (SupplementaryFig. S3), suggesting the effects are widespread in PDAC. DZNeptreatment also resisted migratory and invasive characteristics ofMIA PaCa-2 and PANC-1 (Fig. 1C–F) and even retained itsgrowth inhibitory effects in the presence of TGFb1 (Supple-mentary Fig. S2). Combined treatment of DZNep and TGFb1opposed the morphological and biochemical EMT changes inMIA PaCa-2 and PANC-1 (Fig. 1A and B; Supplementary Fig.S4). Further, the presence of DZNep significantly inhibitedTGFb1-induced increases in cell migration and invasion in MIAPaCa-2 and PANC-1 (Fig. 1C–F). DZNep in the presence ofTGFb1 increased gemcitabine cytotoxicity in PDAC cell lines(Fig. 1G and H) but not in HPDE (Fig. 1G and I). Together,these data identified DZNep's antagonistic effects on TGFb1-induced EMT features in PDAC cell lines.

DZNep inhibits endogenous TGFb1 protein expression andsecretion in a dose- and time-dependent manner

The increase in epithelial characteristics of inherently mesen-chymal MIA PaCa-2 (Fig. 1A; top labeled DZNep) was suggestiveof the inhibition of endogenous TGFb1 by DZNep. To test thispossibility, we assayed the endogenous TGFb1 inMIAPaCa-2 andPANC-1. Because cellular TGFb1 exists as either a latent TGFb1complex (290 kDa) or a biologically active monomeric or homo-dimeric form (13 kDa and 26 kDa, respectively; ref. 28), we firstexamined the identity of TGFb1 in MIA PaCa-2 and PANC-1.Western blotting analysis displayed endogenous TGFb1 in bothcell lines to migrate predominantly as a 26-kDa homodimericTGFb1, although a band corresponding to the monomeric form(13 kDa) was faintly detectable in some cases (Fig. 2A). Anonspecific�20-kDa protein was also detectable. Under the sameconditions, recombinant hTGFb1 also migrated as both mono-meric and dimeric forms (Fig. 2A), so we focused on the better-detectable 26-kDa homodimeric bands for subsequent experi-ments. Treatment with DZNep distinctly reduced the 26-kDaTGFb1 immunoreactivity in both MIA PaCa-2 and PANC-1 ina dose- and time-dependent fashion (Fig. 2A). Significant reduc-tions in TGFb1 were evident from 48 hours of 10 mmol/L DZNeptreatment with almost complete loss of immunoreactivityobserved at�96 to 120 hours inMIA PaCa-2 (Fig. 2A). Reduction

DZNep Resists EMT of Pancreatic Cancer via miRNAs

www.aacrjournals.org Mol Cancer Res; 14(11) November 2016 1125




in TGFb1 with DZNep treatment was also evident in several otherPDAC cell lines (Fig. 2B). Intriguingly, none of these changes inTGFb1 expression appeared to be transcriptionally regulated asreal-timePCRanalysis of total RNA extracted fromDZNep-treatedcells showed a lack of change in TGFb1 transcripts (Fig. 2C andSupplementary Fig. S5).

Because TGFb1 plays a vital role in modulating tumormicroenvironment via autocrine/paracrine mechanisms (29),we examined whether DZNep can inhibit secretory levels ofTGFb1. Using a quantitative sandwich ELISA, we assayed theculture media of DZNep-treated PDAC cells for changes inextracellular TGFb1 (Fig. 2D). Consistent with a reduction inTGFb1 protein, DZNep (10 mmol/L) decreased TGFb1 secretioninto the extracellular media in a time-dependent manner(�40%–50% reduction in 5–6 days; Fig. 2D). In addition, wetested whether the reductions in TGFb1 protein and secretion

can affect its downstream signaling by investigating the phos-phorylation status of Smad2 (a critical mediator of the TGFbsignaling pathway) in DZNep-treated PANC-1 (Fig. 2E). West-ern blotting analysis revealed DZNep to significantly reducep-Smad2 without altering total Smad2 (Fig. 2E). Consistently,DZNep significantly inhibited the expression of several knownTGFb1-induced genes (Fig. 2F). Collectively, these data iden-tified that DZNep can significantly inhibit TGFb1 synthesis andsecretion as well as TGFb signaling in PDAC.

DZNep-induced epigenetic reprogramming of microRNAs inpancreatic cancer cells

The mechanism of DZNep inhibition of TGFb1 remainedunclear, but it was unlikely that TGFb1 was regulated at thetranscriptional level. The lack of a major change in the slope ofthe line representing TGFb1 secretory levels after DZNep

0-2-4-60

50

100

150

% C

ellu

lar v

iabi

lity HPDE

DZNep-TGF-GemTGF-Gem

Control

TGFβ1DZNep

DZNep+T

GFβ10

50

100

150

200

250

MIA PaCa-2PANC-1

% C

ells

inva

ded

n.s.

*

***

****

***

Control

TGFβ1DZNep

DZNep+T

GFβ10

50

100

150

MIA PaCa-2PANC-1

% W

ound

clo

sure

***

***

***

***

***

***

B

C D

E

MIA

PaC

a-2

PAN

C-1

0 h48 h

TGFβ1Control TGFβ1 + DZNepDZNep

0 h48 h

F

MIA

PaC

a-2

PAN

C-1

TGFβ1Control TGFβ1 + DZNepDZNep

48 h48 h

DZNep

MIA PaCa-2 PANC-1

- - + +- + - +

- - + +- - +TGFβ1 +

E-cadherin

N-cadherin

β-Actin

Keratin 8/18

Vimentin

G

H

I

0-2-4-60

50

100

150

Log concentration of gemcitabine (mmol/L)



% C

ellu

lar v

iabi

lity

HPDEAsPc-1L3.6plPANC-1MIA PaCa-2Capan-1HPAF II

0-2-4-60

50

100

150

% C

ellu

lar v

iabi

lity

AsPc-1L3.6plPANC-1MIA PaCa-2Capan-1HPAF II

TGFβ1Control DZNep TGFβ1 + DZNepA

MIA PaC

a-2PAN

C-1

Figure 1.

DZNep inhibits EMT and chemoresistance in pancreatic cancer.A,DZNep resisted TGFb1-inducedmorphologic EMT features. Phase-contrast images of live cells aftertreatment (72 hours).Originalmagnification,�10.B,DZNep resisted TGFb1-induced changes in epithelial andmesenchymalmarkers.Whole-cell lysates (50mg) fromcells treated with TGFb1, DZNep, or both for 72 hours subjected to Western blotting for EMT markers. b-Actin, the internal loading control, is shown with arepresentative blot. C, representative images of cell monolayers subjected to a scratch wound-assay shows DZNep inhibited cell migration. Originalmagnification, �4. D, quantification of wound closure measurements. E, a representation of cells invaded into a Matrigel-coated Transwell insert after crystalviolet staining. F, invaded cells were counted and plotted. G and H, DZNep resisted TGFb1-induced gemcitabine chemoresistance in pancreatic cancer cell lines.3 � 103 cells seeded in a 96-well plate were treated with TGFb1 (24 hours) in the presence (H; dotted lines) or absence (G and H; solid lines) of DZNep (24 hours)followed by an MTT cytotoxicity analysis with gemcitabine. I, DZNep did not increase cytotoxicity in TGFb1-treated HPDE. For all experiments, cells weretreated with DZNep at 10 mmol/L and TGFb1 at 10 ng/mL. Points, mean of triplicates; bars, SD. n ¼ 3.

Mody et al.

Mol Cancer Res; 14(11) November 2016 Molecular Cancer Research1126




treatment (comparedwith untreated control; Fig. 2D) also arguedagainst probable alterations in the posttranslational stability ofTGFb1. Hence, we postulated that DZNep could reprogrammiRNAs in cancer cells to repress TGFb1 protein by RNA inter-ference. To test this hypothesis, we screened >1,900 humanmiRNAs (referenced in the Sanger miRBase Release 18.0.) usingmicroarrays to identify possible global alterations in MIA PaCa-2treated with DZNep as opposed to untreatedMIA PaCa-2 (Fig. 3Aand B). Further, the miRNA profile of MIA PaCa-2 was comparedwith that of HPDE to understandmiRNAs differentially expressed(Fig. 3C and Supplementary Fig. S6). The heat map generatedfrom hierarchical cluster analysis showed a significant reductionin miRNAs in cancer cells with 52.3% of miRNAs (91 out of 174detectable miRNAs) downregulated in MIA PaCa-2 (Fig. 3C–Eand Supplementary Fig. S6; Supplementary Table S3). Similarly,we identified 68.8% of miRNAs (161 out of 234 detectablemiRNAs) downregulated in 10 different patient-derived PDACtissues compared with two normal pancreatic tissues (data not

shown). Thirty-five miRNAs were found commonly downregu-lated in both MIA PaCa-2 and PDAC tissues (SupplementaryTable S4). Interestingly, DZNep increased the expression of >50miRNAs (22 miRNAs at 8 hours; 34 miRNAs at 72 hours; Fig. 3Band D; Supplementary Table S5) and even partially reversed thedownregulation pattern of miRNAs seen in MIA PaCa-2 (Fig. 3Eand F; Supplementary Table S6).

We next examined whether any of the DZNep-upregulatedmiRNAs can target TGFb1 for RNA interference. First, weutilized publicly available algorithms to predict bindingof DZNep-upregulated miRNAs to the 30UTR of TGFb1. Theinitial analysis identified several DZNep-upregulated miRNAsto possibly regulate known players in the TGFb signalingpathway including TGFb1 (Supplementary Table S8). MiR-663a and miR-4787-5p in particular were predicted to bindto the 30UTR of TGFb1 (Supplementary Table S8). Interestingly,those miRNAs were significantly downregulated by 5- to10-fold in MIA PaCa-2 (Fig. 3F). Further comparison of

B

D

C F

A

MIA

PaCa-2

PAN

C-1

p-Smad2

t-Smad2

DZNep (mmol/L) 0 0.1 1 10 20 50 0 TGFβ1 + + + + + + --

0 0.1 1 10 20 500.0

0.5

1.0

1.5

2.0

DZNep (mmol/L)

Rel

ativ

e TG

Fβ1

expr

essi

on

Rel

ativ

e TG

Fβ1

expr

essi

onTranscriptProtein MIA PaCa-2

0 0.1 1 10 20 500.0

0.5

1.0

1.5

2.0

DZNep (mmol/L)

TranscriptProtein PANC-1

E

rTG

Fβ

0 0.1 1 10 20 50 0DZNep

(mmol/L)

TGF-β1dimer

TGF-β1monomer

TGF-β1dimer

TGF-β1monomer

β-Actin

β-Actin

13 kDa

26 kDa

26 kDa

13 kDarT

GFβ24 h 48 h 96 h 120 h

_ _ _ _ + + + +DZNep (10 mmol/L)

TGFβ1dimer

13 kDa

26 kDa

TGFβ1monomer

TGFβ1dimer

TGFβ1monomer

β-Actin

β-Actin

26 kDa

13 kDa

72 h

* *

* *

TGFβ1

β-Actin

_ + _ + _ +AsPC-1 L3.6pl PANC-1DZNep

(10 mmol/L)

MIA PaCa-2 Capan-1 HPAF-II _ + _ + _ +

72

96

120

144

0

50

100

150

200

250

Time (h)

ControlDZNep (10 μmol/L) MIA PaCa-2

72 96 120

144

0

50

100

150

200

250

Time (h)

% T

GFb

1 Se

cret

ion

% T

GFb

1 Se

cret

ion

ControlDZNep (10 mmol/L) PANC-1

COL1A1

KANK4

SERPINE1

TGFBIGDF5

PDGFBBMP7

0

10

20

30

4050

100

Rela

tive

expr

essi

on TGFb1 DZNepTGFb1 + DZNep

Figure 2.

DZNep inhibits TGFb1 synthesis and secretion in pancreatic cancer. A, DZNep decreased TGFb1 protein in a time- and dose-dependent manner. Whole-cell lysates(50 mg) of DZNep-treated cells subjected to Western blotting for TGFb1 expression. A rhTGFb1 used as a positive control and b-actin as a loading control. Theposition of a nonspecific band is indicated by an asterisk. B, DZNep decreased TGFb1 protein in PDAC cell lines. C, DZNep-mediated decrease in TGFb1 wasindependent of transcriptional changes. Densitometric values of TGFb1 protein (A, right), and transcript levels were compared. GUSB used as an internalcontrol for qRT-PCR. D, DZNep reduced TGFb1 secretion into the culture supernatants. The extent of decrease in secretion was measured by ELISA. E, DZNepattenuated TGFb1 signaling in a dose-dependent fashion. Whole-cell lysates of PANC-1 treated with TGFb1 (10 ng/mL; 60 minutes) and DZNep (24 hourspretreatment) subjected to Western blotting for total and p-Smad2. F, DZNep decreased transcript levels of TGFb-responsive genes as determined byqRT-PCR. Points, mean of triplicate; bars, SD. n ¼ 3.






DZNep-upregulated miRNAs with those downregulated in MIAPaCa-2 showed that DZNep restored expression of 17 miRNAswith miR-663a and miR-4787-5p expressions restored at boththe 8 hours and 72 hours of time points (Fig. 3E and F). Real-time PCR analysis with individual primers/probes further val-idated downregulation of miR-663a and miR-4787-5p in sev-eral PDAC tissues (Fig. 4A) and cell lines (Fig. 4B) as well asrestoration of expression with DZNep in PDAC cell lines (Fig.4C and D). DZNep-induced expression of miR-663a and miR-4787-5p was dose- and time-dependent with maximal changesobserved within 8 to 24 hours of 0.1–1 mmol/L DZNep treat-ment in MIA PaCa-2 and PANC-1 (Fig. 4E and F). Induction ofmiR-663a and miR-4787-5p was specific to DZNep and not toany other clinically used nucleoside analogues (SupplementaryFig. S7). Importantly, EZH2 shRNA induced miR-663a andmiR-4787-5p levels and decreased TGFb1 expression in MIAPaCa-2 cells, suggesting EZH2 silencing can phenocopyDZNep's actions (Supplementary Fig. S8).

DZNep-induced miR-663a and miR-4787-5p directly targetTGFb1 for RNA interference

On subsequent DNA analysis, we identified five binding sitesfor the miR-663a seed sequence and a single probable bindingsite for miR-4787-5p within a 60-bp region of the beginning ofthe TGFb1 30UTR (Fig. 5A). To experimentally evaluate whethermiR-663a or -4787-5p can target TGFb1 for RNA interference,we examined the changes in the expression of an engineereddual luciferase construct in which the 30UTR of TGFb1 wascloned downstream of a luciferase cDNA. Interestingly, DZNeptreatment of PANC-1 itself significantly destabilized the lucif-erase activity of the engineered construct in both a dose- andtime-dependent fashion (Fig. 5B), providing preliminary evi-dence for the DZNep-reprogrammed miRNAs to alter bindinginteractions with TGFb1 30UTR. To test if miR-663a ormiR-4787-5p can directly bind to the TGFb1 30UTR and directthe transcripts for RNA interference, we repeated the luciferaseassays on MIA PaCa-2 and PANC-1 stably expressing miR-663a

205-

5p31

-5p

429

31-3

p70

8-5p 50

312

5b-5

p14

1-3p

200a

-3p

200c

-3p

376c

200b

-3p

155-

5p44

36b-

3p34

2-3p

9-3p

100-

5p44

84 203

1268

b18

1a-5

p36

5445

3062

5-5p

4513

181c

-5p

139-

5p12

07-5

p14

9-3p

181b

-5p

4324

4787

-5p

30c-

1-3p

4508

4492

132-

3p19

7214

6912

4619

15-3

p14

6a-5

p30

2d-3

p42

4-5p

3196

5001

-5p

4532

625-

3p36

76-5

p18

1a-2

-3p

663a 21

092

163

848

00-3

p76

245

1617

-3p

194-

5p28

6136

6546

87-3

p44

6639

6031

7819

3a-5

p12

73g-

3p19

2-5p

4497

22-5

p22

-3p

26a-

5p44

8844

23-3

p57

4-5p

3940

-5p

3149

3195

3656

27b-

3p14

6b-5

p14

8a-3

p64

4b-3

p44

4319

7336

07-5

p15

877-

1-3p

185-

5p13

0a-3

p45

5-3p

423-

5p11

8036

0910

6a-5

p-2500,000-2250,000-2000,000

-15,000-10,000

-5,000

-50-40-30-20-10

0

miR

NA

Lev

els

(fol

d de

crea

ses)

miRNA changes in MIA PaCa-2 vs. HPDE

miRNAs Downregulated (≥1.5-fold)

A DB

-2 0 +2

C

F

E

-2 0 +2

1268

b42

5742

9144

4836

76-3

p45

2187

7-5p

663a

4734

3195

30c-

1-3p 92

119

4-5p

4739 76

512

4653

2-5p

4646

-5p

361-

5p44

4447

87-5

p31

7845

3458

4-5p

1280

4690

-5p

4442

4459

30d-

5p14

9-3p

191-

5p36

0923

a-3p

103a

-3p

140-

3p46

4444

637g

-5p

135b

-5p

345-

5p36

76-5

p29

a-5p

149-

5p39

40-5

p29

c-3p

4472

5001

-5p

4454

3197

-6-4-202468

10121416

miR

NA

Lev

els

(fold

cha

nges

) miRNA changes by DZNep (8 h)

miRNAs upregulated (≥1.5-fold)

4448

202-

5p12

4631

8244

8566

3a62

5-5p

424-

5p65

4-5p

10a-

5p10

b-5p

3676

-3p

4530

10a-

3p50

2-3p

4787

-5p

760

584-

5p32

0e 137

1280

1973

3178

320d

320c

320a

4284

4497

500a

-3p

1273

g-3p

3196

4492

378d

19a-

3p17

-5p

106a

-5p

378i

378f

378e

532-

5p37

8g19

5-5p

4463

30e-

5p21

-5p

4444

362-

5p 765

193b

-3p

4442

4488

30b-

5p30

c-5p 15

242

8930

1a-3

p15

b-5p

4466

7e-5

p39

40-5

p45

5-3p

140-

3p18

1a-5

p47

07-5

p15

b-3p

4455

146b

-5p

149-

5p42

6714

6a-5

p18

1b-5

p30

65-5

p44

7231

9736

1-5p

4436

b-3p

194-

5p42

9136

05-5

p44

23-3

p92

1-500,000-400,000

-15,000-7,500

-6-4-202468

10m

iRN

A L

evel

s (fo

ld c

hang

es)

miRNA changes by DZNep (72 h)

miRNAs upregulated (≥1.5-fold)

663a

1246

4787

-5p

3178

1268

b31

9530

c-1-

3p 921

194-

5p62

5-5p

424-

5p45

3019

7344

9712

73g-

3p31

9644

92

-28-24-20-16-12

-8-4048

12

miR

NA L

evel

s (fo

ld c

hang

es)

miRNA downregulated in MIA PaCa-2 vs. upregulated by DZNep

DZNep (8 h)DZNep (72 h)

PanCan

MIA PaCa-2 vs. HPDE

52.3 %

5.7%

42 %

DZNep (8 h) vs. Control(MIA PaCa-2)

24.5 %

44.9 %30.6 %

DZNep (72 h) vs. Control

.

(MIA PaCa-2)

22.2 %

35.8 %

42 %

% Unchanged miRNAs% Downregulated miRNAs% Upregulated miRNAs

Figure 3.

DZNep reprograms PDAC-downregulated miRNAs in pancreatic cancer. A, clustering of miRNAs differentially expressed in DZNep (1 mmol/L)-treated MIA PaCa-2versus untreated. Heat maps with statistically significant (P < 0.1) changes in miRNA expression shown by a 2-color system (light green, lowest expression;dark red, highest expression). Each row represents expression of a singlemiRNA,while each column represents a single sample. n¼ 2.B, fold changes inmiRNAs after8 hours or 72 hours of DZNep treatment in MIA PaCa-2. Purple bars indicate miRNAs upregulated by DZNep at both time points. C, miRNAs downregulatedin MIA PaCa-2 compared against HPDE. D, Pie charts showing expression patterns of miRNAs. A fold change of � or � 1.5 considered as upregulated ordownregulated respectively. E,Venn diagram comparingmiRNAs downregulated inMIA PaCa-2 against HPDE (blue; n¼ 2) with those upregulated by DZNep inMIAPaCa-2 at 8 hours (orange) or 72 hours (green). F, fold changes of miRNAs downregulated in MIA PaCa-2 (Pan Can) but upregulated after DZNep treatment.

Mody et al.





or miR-4787-5p (Fig. 5C). Stable expression was achieved usinglentiviral gene constructs wherein�3- to 5-fold and�10- to 12-fold increases in miR-663a and miR-4787-5p expression,respectively, was achieved (Supplementary Fig. S9). The resultsobtained from these luciferase assays indicated that bothmiR-663a and miR-4787-5p can directly bind to the 30UTR ofTGFb1 and destabilize luciferase expression in MIA PaCa-2 andPANC-1 (Fig. 5C). Western blotting analysis using cell lysatesand ELISA on culture supernatant showed a reduction in TGFb1protein synthesis (Fig. 5D) and secretion (Fig. 5E) in cellsoverexpressing miR-663a and miR-4787-5p. Again, thesechanges were independent of TGFb1 transcriptional variations(Fig. 5F). The reductions in luciferase activities noted withmiRNAs were effectively counteracted with their respectiveLNA-derived antagomirs (Fig. 5G), suggesting the observedeffects were specific to miR-663a and miR-4787-5p. Addition-ally, LNA-derived antagomiRNAs significantly antagonized therespective miRNA-induced reductions in TGFb1 protein (Fig.5H and I). These data identified the direct binding of miR-663aand miR-4787-5p to the 30UTR of TGFb1 and supported RNAinterference as a mechanism of DZNep inhibition of the TGFbsignaling pathway in PDAC.

miR-663a and miR-4787-5p contribute to DZNep-mediatedEMT control in pancreatic cancer

We next investigated whether miR-663a and miR-4787-5p cansuppress TGFb1-induced EMT in PDAC as observed with DZNep(Fig. 1). Overexpression of either miR-663a or miR-4787-5p

significantly induced E-cadherin and decreased vimentin (Fig.6A) similar toDZNep (Fig. 1B). BothmiR-663a andmiR-4787-5palso increased cytokeratin 8/18 and decreased N-cadherin inPANC-1 (Fig. 6A). Both miR-663a– and miR-4787-5p–overex-pressing cells exhibited reduced cell migration (Fig. 6B and D)and invasion (Fig. 6C and E and Supplementary Fig. S10), similarto that observed with DZNep treatment. Morphologically,miR-663a and miR-4787 overexpression visibly increased theepithelial appearance of MIA PaCa-2 as observed with DZNeptreatment; however, the changes between control and miR-over-expressing PANC-1 were indistinguishable (data not shown).Both miR-663a and miR-4787-5p significantly diminished thep-Smad2 level without altering total Smad2 in PANC-1 (Fig. 6F).As seen with DZNep, miR-663a and miR-4787-5p also signifi-cantly reduced expression of the TGFb-responsive genes exam-ined in Fig. 2F. However, to further extend the understanding ofthe miR-induced effects on the entire TGFb axis, we profiledtranscript changes of 84 pathway-related genes (SupplementaryTable S2). PCR array data analyses of Ct values identified anegative TGFb pathway scores of �0.4 for miR-663a (P < 0.01)and�0.2 formiR-4787-5p (P¼n.s.), respectively, confirming therepressive roles of miRNAs on the TGFb signaling pathway (Fig.6G and H). As expected, TGFb activated the pathway and gen-erated a positive pathway score ofþ0.865 (P < 0.001; Fig. 6G andH). In addition to attenuation of TGFb signaling, both miRNAsmoderately reduced growth and cell proliferation rates in MIAPaCa-2 and PANC-1 cells as observed from MTT and colonyformation assays (Fig. 6I and J).

00.0

01 0.01 0.1 1 10

0

2

4

6

Rel

ativ

e ex

pres

sion

miR-663amiR-4787-5p

DZNep (mmol/L)

8 hMIA PaCa-2

Control

8 h 72h

0

1

2

3

4

Rel

ativ

e m

iR-4

787-

5p e

xpre

ssio

n

HPDEAsPC-1L3.6plPANC-1MIA PaCa-2Capan-1HPAF II

*

*

***

*****

DZNep (1 mmol/L)

miR-4787-5p

*****

***

Control

8 h 72 h

0

2

4

6

Rel

ativ

e m

iR-6

63a

expr

essi

on

HPDEAsPC-1L3.6plPANC-1MIA PaCa-2Capan-1HPAF II

**

******

**

*

**

DZNep (1 mmol/L)

miR-663a

**

Normal

(n = 3)

1171

675F

2578

70A1

2578

72A1

2578

73A1

2578

75A1

2578

76A1

1143

862F

1143

867F

1100

174F

Avg Tumor

0.0

0.5

1.0

1.5

Rel

ativ

e ex

pres

sion

miR-663amiR-4787-5p

Moderately differentiated

Moderately-poorly differentiated

****

A C

D

E

FB

HPDE

AsPC-1

L3.6pl

PANC-1

MIAPaC

a-2

Capan

-1

HPAF-II0.0

0.5

1.0

1.5

Rel

ativ

e ex

pres

sion

miR-663amiR-4787-5p

*******

***

* ***

***

***

******

0 8 24 48 720

2

4

6

Rela

tive

expr

essi

on

miR-663amiR-4787-5p

Time (h)

MIA PaCa-2DZNep (1 mmol/L)

Figure 4.

DZNep restores miR-663a and miR-4787-5p expression in pancreatic cancer. A and B, MiR-663a and miR-4787-5p are downregulated in patient-isolatedPDAC tissues (A) and cell lines (B) as determined by qRT-PCR. GUSB and U6B were used as an internal control for assaying miR-663a and miR-4787-5p. C and D,DZNep increased miR-663a (C) and miR-4787-5p (D) expression in PDAC cell lines. E and F, DZNep increased miR-663a and miR-4787-5p expressions in adose- and time-dependent manner. Bars, SD. n ¼ 3.






DZNep, miR-663a, and miR-4787-5p suppress the growth andmetastasis of pancreatic cancer xenografts

Finally, we investigated whether the effects of DZNep andmiRNAs on EMT resistance could be recapitulated in vivo. Athymicnude mice were orthotopically implanted with PANC-1 stablytransduced with control, miR-663a, or miR-4787-5p lentiviralconstructs. Mice injected with control PANC-1 were also treatedwith DZNep (3 mg/kg, i.p., twice a week) until sacrificed. At theend of 10 weeks, tumor growth was found to be significantlysmaller in DZNep-treated mice as well as in mice transplantedwith miR-663a or miR-4787-5p cells as compared with untreatedmice injected with control cells (Fig. 7A and B). Consistently, 10-week tumor weights (Fig. 7C) and volumes (Fig. 7D) weresignificantly less for the DZNep-treated group as compared withthe control group. Mice transplanted miR-663a– or miR-4787-5p–overexpressing cells also showed reduced tumor weights and

volumes against control groups (Fig. 7A–D). Furthermore, aspredicted from in vitro data (Fig. 1 and Fig. 6), the total numberof metastatic lesions in the secondary organs (liver, spleen, lungs,and kidneys) was significantly less in DZNep-treated (mean ¼0.83� 0.8), miR-663a–transduced (mean¼ 1.5� 1.4), andmiR-4787-5p–transduced (mean¼0.83� 1.6)mice as comparedwithcontrol groups (mean ¼ 4.83 � 2.9; Fig. 7E–G). In fact, 67% ofmice in the DZNep-treated or miRNA-transduced group showedno metastatic lesions in the liver (main site of pancreatic cancermetastasis), while 83.33% of mice in the control group demon-strated visible liver metastases. DZNep-treated or miRNA-trans-duced mice also exhibited reduced metastatic foci in the spleen,lungs, and kidneys as compared with control groups (Fig. 7F andG).Overall, DZNep andmiRNAs in tumors were well tolerated bythe mice as evident by no significant changes in hepatic enzyme(sGPT and sGOT) levels or bodyweight (Fig. 7H). Finally, DZNep

Control

miR-66

3a

miR+LNA-66

3a

miR-47

87-5p

miR+L

NA-4787

-5p0.0

0.5

1.0

1.5

Rel

ativ

e TG

Fβ1

trans

crip

tsR

elat

ive

TGFβ

1 tra

nscr

ipts

PANC-1

60

80

100

120

% L

ucife

rase

act

ivity

miR-663aLNA-663amiR-4787-5pLNA-4787-5p

- + + - - - - + - - - - - + + - - - - +

PANC-1

***

***

***

72

96

120

144

0

50

100

150

200

250

Time (h)

ControlmiR-663a MIA PaCa-2miR-4787-5p

Control

miR-66

3a

miR-47

87-5p

60

80

100

120

% L

ucife

rase

act

ivity

**

*

PANC-1MIA PaCa-2

***

***

8 h 16 h

24 h

60

80

100

120

% L

ucife

rase

act

ivity

Control DZNep (10 μmol/L)DZNep (20 μmol/L)

PANC-1

n.s.n.s.

n.s.

****

**

A

miR-663a(13−19)

GGTCCCGCCCCGCCCCGCCCCGCC…CCCCACCCCGCCCCGCCCCC…

miR-663a(8−14) miR-663a(18−24) miR-663a(43−49)

miR-663a(48−54)

miR-4787-5p(51−57)

831

AGGCGGGGCGCCGCGGGACCGCmiR-663a -GCGGGGGUGGCGGCGGCAUCCCmiR-4787-5p -

D

CB

FE

HG miR-663a _ + + + + _

LNA-663a 0 1 5 25 _ _

β-Actin

PANC-1

miR-4787-5p _ + + + + _

LNA-4787-5p 0 1 5 25 _ _

TGFβ1

TGFβ1

β-Actin

PANC-1

(nmol/L)

(nmol/L)

TGF-β1

β-Actin

MIA

PaCa-2

PAN

C-1

TGF-β1

β-Actin

I

Control

miR-66

3a

miR-47

87-5p

0.0

0.5

1.0

1.5 MIA PaCa-2PANC-1

72

96

120

144

0

50

100

150

200

250

Time (h)%

TG

Fβ1

Secr

etio

n

% T

GFβ

1 Se

cret

ion

ControlmiR-663a PANC-1miR-4787-5p

Figure 5.

miR-663a and miR-4787-5p directly target TGFb1 for RNA interference. A, schematic representation of predicted miR-663a and miR-4787-5p binding site(s) in theTGFb1 30-UTR. B, DZNep reduced luciferase activity in cells transfected with the TGFb1 30-UTR luciferase reporter construct. C, MiR-663a or miR-4787-5p directlybinds to TGFb1 30UTR. Cells stably overexpressing miRNAs were transfected with the TGFb1 30-UTR luciferase reporter and percentage of activities wasrecorded.D,MiR-663a andmiR-4787-5p decreased TGFb1 protein.Whole cell lysates (50mg) of cells stably overexpressingmiRNAs subjected toWestern blotting todetermine TGFb1 protein levelswith b-actin as a loading control. E,miRNAoverexpression decreased TGFb1 secretion into the culture supernatants. F,miR-663a andmiR-4787-5p did not affect TGFb1 transcripts. Total RNA from cells stably overexpressing miRNAs subjected to qRT-PCR to determine TGFb1 transcriptlevels with GUSB as an internal control. G, LNA-663a or LNA-4787-5p (50 nmol/L; 48 hours) antagonized respective miRNA-induced decreases in luciferaseactivity of cells. H and I, LNA-663a or LNA-4787-5p antagonized respective miRNA-induced decreases in TGFb1 protein levels (H) but had no effecton TGFb1 transcripts (at 50 nmol/L; 48 hours; I). Points, mean of triplicate; bars, SD. n ¼ 3.

Mody et al.





also induced regression of tumors in mice (Fig. 7I and Supple-mentary Fig. S11). These results support that DZNep andmiR-663a and miR-4787-5p can suppress the metastatic capacityof orthotopically implanted pancreatic tumor cells.

DiscussionIncreasing evidence is recognizing the critical role of EZH2-

induced chromatin alterations and transcriptional reprogram-ming of genes in EMT (20). For instance, SOX4 was identifiedas a master regulator of EMT by directly controlling EZH2 expres-sion in normal and cancerous breast epithelial cells (21). EZH2was shown to support cancer cell invasion and metastasis via theTGFb1 axis in various solid tumors and serves as a predictiveindicator of treatment outcomes in PDAC and ovarian cancerpatients (30, 31). Despite recognizing EZH2 regulation of EMT,studies on the development of pharmacological inhibitors of

EZH2 to mitigate EMT progression are limited. While DZNepinhibition of EZH2 has been shown to increase the expression ofthe epithelialmarker E-cadherin in certain cancer types (e.g., renalcell carcinoma; ref. 32), its role in modulating overall EMTcharacteristics in solid tumors remains elusive. The current studyuncovers the role of DZNep in dampening TGFb1-induced EMTcharacteristics in PDAC. To our knowledge, this is the first studythat demonstrates a direct effect of a histone methyltransferaseinhibitor in decreasing TGFb1 protein and secretory levels incell lines representing various subtypes of PDAC. Pharmacolog-ical treatment with DZNep distinctly induced epithelial charac-teristics of PDAC cells, counteracted TGFb1-induced mesenchy-mal characteristics, and inhibited migratory and invasive signalsthat drive tumor progression. The mechanism of DZNep inhibi-tion of TGFb1 was identified as epigenetic reprogramming ofmiRNAs, which included a subset of miRNAs downregulated inPDAC. Significant contributions came from two epigenetically

24 48 72 960

200

400

600

800

Time (h)

% C

ellu

lar v

iabi

lity

ControlmiR-663amiR-4787-5p

PANC-1

******

*** ***

****

24 48 72 960

500

1,000

1,500

Time (h)

% C

ellu

lar v

iabi

lity

ControlmiR-663amiR-4787-5p

MIA PaCa-2

***

***

***

***

Control

miR-66

3a

miR-47

87-5p

0

50

100

150

% C

ells

inva

ded

MIA PaCa-2PANC-1

********

*

Control

miR-66

3a

miR-47

87-5p

0

20

40

60

% W

ound

clo

sure

MIA PaCa-2PANC-1

******

******MIA PaCa-2 PANC-1

E-cadherin

N-cadherin

β-Actin

Keratin 8/18

Vimentin

Control

48 h48 h

0 h0 h

miR-663a miR-4787-5pB

A

C

D

E

G

HI

TGFβ1 _ +

Control

_ + miR-663a

_ + miR-4787-5p

p-Smad2

t-Smad2

TGFb1

miR-4787-5p

miR-663a

-1.0 -0.8 -0.6 -0.4 -0.2 0.0TGFb Pathway activity score

0.2 0.4 0.6 0.8 1.0

PAN

C-1

F

ControlTG

Fβ1

Empty vectorm

iR-663am

iR-4787-5p

Control Control

miR-663a miR-4787-5p

J

PAN

C-1

Control miR-663a miR-4787-5p

MIA

PaC

a-2

PAN

C-1

MIA

PaC

a-2

PAN

C-1

48 h48 h

Figure 6.

miR-663a and miR-4787-5p resist EMT features and slow cellular growth in pancreatic cancer cells. A,miR-663a and miR-4787-5p increased epithelial markers anddecreased mesenchymal markers. Whole-cell lysates (50 mg) of cells overexpressing miRNAs were subjected toWestern blotting for EMT markers with b-actin as aloading control. B and D, miR-663a and miR-4787-5p decreased cell migration. Representative images of cells subjected to a scratch wound assay (B) andquantification of wound closure measurements (D). Original magnification, �4. C and E, miR-663a and miR-4787-5p decreased cell invasion. F, miR-663a andmiR-4787-5p reduced p-Smad2 levels. Whole-cell lysates of TGFb1 (10 ng/mL; 60 minutes)-treated cells subjected to Western blotting to detect total andp-Smad2. G and H, miRNAs decreased expression of TGFb1-responsive genes and attenuated the TGFb1 signaling pathway. Total RNA from cells subjected to aRT2-PCR profiler array. Heat maps showing transcript level changes of TGFb-responsive genes (H) and overall TGFb pathway activity scores depicted (G).I and J,miR-663a andmiR-4787-5p reduced cellular proliferation as determined byMTT (I) and colony formation (J) assays. Points, mean of triplicate; bars, SD. n¼ 3.






reprogrammed miRNAs (miR-663a and miR-4787-5p) thatdirectly targeted TGFb1 for RNA interference and reduced TGFb1synthesis and secretory levels. These findings reveal an alternativemechanism of TGFb1 regulation of EMT in PDAC and betterexplain the antimetastatic effects of synthetic histone methyltransferase inhibitors that have so far only been known to upre-gulate a small number of coding genes.

Earlier, we reported DZNep to induce delayed cytotoxic andapoptotic effects inmoderately–poorly differentiatedMIAPaCa-2and PANC-1 (25). While we demonstrate here that DZNep canalso dampen EMT in these cell types, we also observed some cell-line–specific differences with respect to TGFb1 activation of EMTcharacteristics. It is likely that the inherent genetic variations in theTGFb1 signaling components of these cell types contribute tovariations (33, 34). While PANC-1 has intact TGFBR2, it ismutated in MIA PaCa-2 (33). Consistently, a modest growthinhibitory effect of TGFb1 was evident in PANC-1 but not in MIA

PaCa-2. Next, we noted exogenous TGFb1 to activate the endog-enous synthesis in PANC-1 but not in MIA PaCa-2 (Supplemen-tary Fig. S12) and the autostimulatory effects in PANC-1 wereattenuated by DZNep. Further, TGFb1 phosphorylated Smad3 inPANC-1 but not significantly in MIA PaCa-2 (Supplementary Fig.S12B). Moreover, the endogenous levels of TGFb1 were signifi-cantly higher in PANC-1 than in MIA PaCa-2 (SupplementaryFig. S12B). Given the genetic aberrations in the TGFb1 signalingpathway in MIA PaCa-2, the mechanism by which TGFb1 triggersmesenchymal characteristics in MIA PaCa-2 remains perplexing;however, it becomes increasingly clear that noncanonical TGFbpathways, including the epigenetic pathways, could play prom-inent roles in deciding the EMT signature of PDAC cells. Furtherintriguing was that DZNep resisted TGFb1-induced EMT charac-teristics at approximately equal magnitudes in both these celltypes. These observations suggest that epigenetic reprogrammingof miRNAs could likely play a dominant role in inhibiting TGFb1

Control

DZNep

miR-66

3a

miR-47

87-5p

0

1,000

2,000

3,000

Tum

or v

olum

e (m

m3 )

*** *** ***

Control

DZNep

miR-66

3a

miR-47

87-5p

0

1

2

3

4

5

Tum

or w

eigh

t (g)

*** *** ***

Control DZNep miR-663a miR-4787-5pA

C

B

E

F G

H

Control DZNep miR-663a miR-4787-5p

Control

DZN

epm

iR-

663am

iR-

4787-5p

D

Liver

Spleen

Lung

Kidney0

50

100

150

Met

asta

sis

inci

denc

e (%

)

ControlDZNepmiR-663amiR-4787-5p

5 8 12 15 20 22 27 30 34 46 50 54 57 62 66

20

22

24

26

28

30

Days post implantation

Bod

y w

eigh

t (g) Control

DZNepmiR-663amiR-4787-5p

Control

DZNep

miR-66

3a

miR-47

87-5p

0

2

4

6

8

10

No o

f met

asta

tic le

sion

s pe

r mou

se

** ***

I

28 30 32 34 36 38 400

1,000

2,000

3,000

Days post implantation

Ave

rage

tum

or v

olum

e (m

m3 )

for a

gro

up

ControlDZNep

***

Figure 7.

DZNep and miRNAs suppress pancreatic cancer growth and metastasis in vivo. A and B, DZNep and miRNAs reduced primary tumor burden in an orthotopicpancreatic xenograft model. Representative photographs of tumor xenografts (A) and excised tumor specimens (B) after 10 weeks of cell injections.C and D, average tumor weights (C) and volumes (D) in DZNep-treated and miRNA-expressed conditions. Each point represents data obtained from one mousewithin a group. E, representative images of tumor metastases to the liver with arrows pointing to metastatic foci. Tumor characteristics were confirmed byhistopathology. F and G, DZNep, miR-663a, and miR-4787-5p reduced overall metastasis in secondary organs. F, the total number of metastatic nodules permouse was added and plotted as a data point. The bar indicates the mean number of lesions per mouse within a group. G, metastatic incidences per organ.H, average body weights of mice with each point representing average mice weight (n ¼ 6) for a group. I, DZNep induced regression of tumors in mice (n ¼ 3for each group). Arrows indicate days of DZNep treatment. Bars, SD.

Mody et al.





in PDAC cells. In addition to p-Smad2, DZNep suppressedp-Smad3 induced by TGFb1 in PANC-1 and decreased endoge-nous p-Smad3 levels inMIAPaCa-2 and PANC-1 (SupplementaryFig. S12B). DZNep also suppressed many of the TGFb1-inducedcellular EMT phenotypes, including their migratory and invasivebehaviors in both cell types. Even in the presence of TGFb1,DZNep consistently increased nucleoside analog cytotoxicity inPDAC cell lines regardless of whether TGFb1 promoted furtherchemoresistance or not. Although themechanism of howDZNepinhibits exogenous TGFb1-induced EMT characteristics in pan-creatic cancer cells warrants further investigation, overall thesefindings support the overriding potential of synthetic epigeneticreversal agents in inhibiting TGFb-mediated EMT characteristicsin PDAC.

The identification of profound reprogramming of miRNAs byDZNep uncovered several putative mechanisms involved in cyto-static growth control and EMT reversal in PDAC. For instance,DZNep induced the expressions of �50 miRNAs by �1.5–10folds. Of these miRNAs, several candidates have been reportedearlier to confer antitumorigenic roles in various solid tumors(35–41). While they provide supportive evidence for DZNep'sgrowth control effects in PDAC, comparison of DZNep-inducedmiRNAs with downregulated miRNA datasets in PDAC furtherrevealed which miRNAs are playing a role in the TGFb1-inducedEMT. Particularly, we focused onmiR-663a andmiR-4787-5p fortheir consistent downregulation in PDAC tissues and cell linesand upregulation by DZNep at various time points. Interestingly,miR-4787-5p is a novel miRNA with no studies available in theliterature. MiR-663a has been previously reported as a primate-specific miRNA silenced in other solid tumors by DNA methyl-ation (42–44). This study provides primal evidence for silencingof miR-663/4787 in human PDAC due to histone methylationand for the use of DZNep as a pharmacologic activator ofmiR-663/4787 expression for EMT reversal.

Intriguingly, the majority of DZNep-induced miRNAs, partic-ularly miR-663a and miR-4787-5p (>90%), were high in GCcontent (Supplementary Table S7). The mechanism of inductionof GC-rich miRNAs by DZNep is unclear at present, but it isinteresting to note that the 30UTR region of the TGFb1 mRNA,where most miRNAs bind and function, also has GC-rich regions(45). Further, miR-663a and miR-4787-5p reduced TGFb1 pro-tein levels without affecting TGFb1 mRNA by directly binding tothe TGFb1 30UTR and causing RNA interference. Even in thepresence of Actinomycin-D (transcription inhibitor), exogenousmiR-663a or miR-4787-5p had no effect on TGFb1 mRNA (datanot shown), suggesting the lack of changes in TGFb1 mRNA arenot due to high turnover of the TGFb1 mRNA. These findings aresupportive of a physiologic role for miR-663a and miR-4787-5pin regulating TGFb1 function. Indeed, as this article is completed,two independent groups showed miR-663 involvement in regu-lating TGFb signaling in glioblastoma and thyroid carcinoma(46–48). Surprisingly, miR-663a and miR-4787-5p also attenu-ated TGFb1-induced p-Smad2, which suggests their role in reg-ulating other signaling components in addition to TGFb1. Bio-informatics analyses failed to predict the involvement of anycanonical TGFb signaling players as targets for miR-663a ormiR-4787-5p, and hence the mechanism of miR-663a andmiR-4787-5p mediated suppression of p-Smad2 requires furtherinvestigation.

We suspect that several other DZNep-reprogrammedmiRNAs are also capable of directly targeting the TGFb1 30UTR

to RNA interference because DZNep induced a much strongerTGFb1 RNA interference effect than either miR-663a or miR-4787-5p in PANC-1 cells. Recent findings in the laboratoryshowed DZNep to reduce the expression of TGFb receptors andcould provide a plausible explanation for DZNep's antagonisticeffects on exogenous TGFb. While this is further explored in thelaboratory, we also note that DZNep could target other keyplayers (in addition to TGFb) in the TGFb signaling pathway(Supplementary Table S8). Together, these scenarios providesupport for why DZNep imparted stronger growth and metas-tasis inhibition effects on PANC-1 derived tumors in vivo thanthat observed with either miRNAs. Further, it is likely that thedecreased metastatic load in DZNep/miR arms, at least in part,could be due to decreased size of primary tumors. However, wedid not always find a correlation between tumor size andmetastatic lesions (Fig. 7B–G). For instance, miR-4787-5ptumors were relatively larger as compared with DZNep andmiR-663a; however, it was as efficient, if not better, as them insuppressing metastasis. Nonetheless, at this point we do nothave further evidence to explain the apparent discrepancies.Experimentally, we found no evidence for increased caspase-3cleavage in tumors from mice belonging to DZNep and miRgroups (Supplementary Fig. S13), whereas our in vitro data fromthis study (Fig. 6I and J) and our previous study on culturedpancreatic cancer cells showed DZNep to reduce cellular pro-liferation (25). These results indicate that the reduced tumorsizes noted with DZNep and miRNAs could be more likelyattributed to reduced cell proliferation and not increased apo-ptosis. While the current study confirms the role of miR-663aand miR-4787-5p in growth control and EMT resistance inPDAC, the functions of numerous other PDAC-downregulatedand DZNep-induced miRNAs remain to be interrogated.

The development of DZNep as a potential therapeutic com-pound may present some limitations. For example, DZNepexerted somenonspecific actions such as increasing the expressionof certain potential oncomiRNAs (e.g., miR-10a and miR-10b)and decreasing the expression of certain potential tumor suppres-sor miRNAs (i.e., miR-181b). Further, DZNep's antitumorigeniceffects occurred only at lower micromolar concentrations,although this may not be a significant concern because mostFDA-approved nucleoside analogues exhibit an effective thera-peutic plasma concentration at this range. DZNep has also beenreported to cause an unintended nephrotoxicity in rodents overlong-term usage, with some studies even suggesting DZNep canhinder normal developmental functions.While this could be verywell attributed to DZNep's global histone methylation alteringeffects, an attempt was made in this study to refine the applica-bility of histone methylation reversal agents by investigating itsmolecular targets and identifying candidates responsible for itsbeneficial actions. As the causes for cancer initiation and progres-sion are multifactorial, it also becomes imperative to globallyreprogram the cancer genome in a directed manner to desirablyimpart growth and EMT control. In this regard, both microRNAsand DZNep-like compounds offer a tremendous potential tofuture research in this direction. Identification of more favorableDZNep analogues exploiting desirably reprogrammedmiRNAs ordirect delivery of reprogrammed miRNAs into tumors offers twopotential avenues for therapeutic development. The recent pre-clinical pharmacokinetic studies optimizing intravenous admin-istration of DZNep for utilities in advanced-stage solid tumors isalso very encouraging (49).






It is well known that EMT facilitates migration and invasion oftumor cells that leads tometastasis and development of chemore-sistance. EMT is triggered by perturbations of the autocrine/paracrine TGFb signaling toward late-stage cancer. Hence, thispathway remains an attractive target of intervention to controlEMT. Plasma TGFb1 levels are also elevated in advanced pancre-atic carcinoma patients who are presented with increased risk ofpoor prognosis (50). TGFb1 levels are also significantly higher inpatients not responding to chemotherapy as compared with theresponding group (50). These clinical correlations suggest thatTGFb ligands and/or their receptors could be important targets incurtailing PDAC progression. Indeed, humanized monoclonalantibodies targeting TGFb ligands and receptors, antisense oligo-nucleotides targeting TGFb ligands, and small-molecule inhibi-tors of TGFb receptors have been developed and tested (5). Whilea few remain in phase I and II clinical trial evaluations, unfortu-nately, many of them have already been terminated due to lowefficacies or high toxicities (5). So far, LY2157299, a small-molecule inhibitor of TGFBR1, is the most advanced TGFb sig-naling inhibitor currently under clinical development (5, 51). Theapproach used in this study provides a new direction for inves-tigating small-molecule inhibitors of TGFb signaling in PDAC viamicroRNA reprogramming of cancer cells.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: H.R. Mody, S.W. Hung, R. GovindarajanDevelopment of methodology: H.R. Mody, S.W. Hung, M. AlSaggar,R. GovindarajanAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): H.R. Mody, M. AlSaggar, J. Griffin, R. GovindarajanAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): H.R. Mody, R. GovindarajanWriting, review, and/or revision of the manuscript: H.R. Mody, S.W. Hung,R. GovindarajanAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): M. AlSaggar, R. GovindarajanStudy supervision: R. Govindarajan

Grant SupportThis work was supported by the award NIH1R01CA188464

(R. Govindarajan).The costs of publication of this articlewere defrayed inpart by the payment of

page charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received March 11, 2016; revised July 29, 2016; accepted August 27, 2016;published OnlineFirst September 13, 2016.

References1. Ryan DP, Hong TS, Bardeesy N. Pancreatic adenocarcinoma. N Engl J Med

2014;371:1039–49.2. Satoh K, Hamada S, Shimosegawa T. Involvement of epithelial to mesen-

chymal transition in the development of pancreatic ductal adenocarcino-ma. J Gastroenterol 2015;50:140–6.

3. SmithBN,BhowmickNA.Role of EMT inmetastasis and therapy resistance.J Clin Med 2016;5;pii: E17. doi: 10.3390/jcm5020017.

4. Javle MM, Gibbs JF, Iwata KK, Pak Y, Rutledge P, Yu J, et al. Epithelial-mesenchymal transition (EMT) and activated extracellular signal-regulatedkinase (p-Erk) in surgically resected pancreatic cancer. Ann Surg Oncol2007;14:3527–33.

5. Neuzillet C, de Gramont A, Tijeras-Raballand A, deMestier L, Cros J, FaivreS, et al. Perspectives of TGF-beta inhibition inpancreatic andhepatocellularcarcinomas. Oncotarget 2014;5:78–94.

6. Truty MJ, Urrutia R. Basics of TGF-beta and pancreatic cancer. Pancreatol-ogy 2007;7:423–35.

7. Shi Y, Massague J. Mechanisms of TGF-beta signaling from cell membraneto the nucleus. Cell 2003;113:685–700.

8. Massague J. TGFbeta signalling in context. Nat Rev Mol Cell Biol 2012;13:616–30.

9. HeldinCH, VanlandewijckM,Moustakas A. Regulation of EMTby TGFbetain cancer. FEBS Lett 2012;586:1959–70.

10. Katsuno Y, Lamouille S, Derynck R. TGF-beta signaling and epithelial–mesenchymal transition in cancer progression. Curr Opin Oncol2013;25:76–84.

11. de Caestecker MP, Piek E, Roberts AB. Role of transforming growth factor-beta signaling in cancer. J Natl Cancer Inst 2000;92:1388–402.

12. Kubiczkova L, Sedlarikova L, Hajek R, Sevcikova S. TGF-beta – an excellentservant but a bad master. J Transl Med 2012;10:183.

13. Chowdhury S, Ammanamanchi S, Howell GM. Epigenetic targeting oftransforming growth factor beta receptor II and implications for cancertherapy. Mol Cell Pharmacol 2009;1:57–70.

14. Guo L, Zhang Y, Zhang L, Huang F, Li J, Wang S. MicroRNAs, TGF-betasignaling, and the inflammatorymicroenvironment in cancer. Tumour Biol2016;37:115–25.

15. Kiesslich T, Pichler M, Neureiter D. Epigenetic control of epithelial–mesenchymal transition in human cancer. Mol Clin Oncol 2013;1:3–11.

16. DanquahM, Singh S, Behrman SW, Mahato RI. Role of miRNA and cancerstem cells in chemoresistance and pancreatic cancer treatment. ExpertOpinDrug Deliv 2012;9:1443–7.

17. McCleary-Wheeler AL, Lomberk GA, Weiss FU, Schneider G, Fabbri M,Poshusta TL, et al. Insights into the epigenetic mechanisms controllingpancreatic carcinogenesis. Cancer Lett 2013;328:212–21.

18. Liu S, Ye D, Guo W, Yu W, He Y, Hu J, et al. G9a is essential for EMT-mediatedmetastasis andmaintenance of cancer stem cell-like characters inhead and neck squamous cell carcinoma. Oncotarget 2015;6:6887–901.

19. Zhang L, Deng L, Chen F, Yao Y, Wu B, Wei L, et al. Inhibition of histoneH3K79 methylation selectively inhibits proliferation, self-renewal andmetastatic potential of breast cancer. Oncotarget 2014;5:10665–77.

20. Rao RA, Dhele N, Cheemadan S, Ketkar A, Jayandharan GR, Palakodeti D,et al. Ezh2 mediated H3K27me3 activity facilitates somatic transitionduring human pluripotent reprogramming. Sci Rep 2015;5:8229.

21. Tiwari N, Tiwari VK, Waldmeier L, Balwierz PJ, Arnold P, Pachkov M, et al.Sox4 is a master regulator of epithelial–mesenchymal transition by con-trolling Ezh2 expression and epigenetic reprogramming. Cancer Cell2013;23:768–83.

22. Oktyabri D, Tange S, Terashima M, Ishimura A, Suzuki T. EED regulatesepithelial–mesenchymal transition of cancer cells induced by TGF-beta.Biochem Biophys Res Commun 2014;453:124–30.

23. Tange S, Oktyabri D, Terashima M, Ishimura A, Suzuki T. JARID2 isinvolved in transforming growth factor-beta-induced epithelial–mesen-chymal transition of lung and colon cancer cell lines. PLoS One 2014;9:e115684.

24. Kondo Y. Targeting histone methyltransferase EZH2 as cancer treatment.J Biochem 2014;156:249–57.

25. Hung SW, Mody H, Marrache S, Bhutia YD, Davis F, Cho JH, et al.Pharmacological reversal of histone methylation presensitizes pancreaticcancer cells to nucleoside drugs: in vitro optimization and novel nano-particle delivery studies. PLoS One 2013;8:e71196.

26. Bhutia YD, Hung SW, Krentz M, Patel D, Lovin D, Manoharan R, et al.Differential processing of let-7a precursors influences RRM2 expressionand chemosensitivity in pancreatic cancer: role of LIN-28 and SET onco-protein. PLoS One 2013;8:e53436.

27. Kim MP, Evans DB, Wang H, Abbruzzese JL, Fleming JB, Gallick GE.Generation of orthotopic and heterotopic human pancreatic cancer xeno-grafts in immunodeficient mice. Nat Protoc 2009;4:1670–80.

28. Hinck AP, Archer SJ, Qian SW, Roberts AB, SpornMB,Weatherbee JA, et al.Transforming growth factor beta 1: three-dimensional structure in solutionand comparisonwith theX-ray structure of transforming growth factor beta2. Biochemistry 1996;35:8517–34.

Mody et al.





29. Subramanian G, Schwarz RE, Higgins L, McEnroe G, Chakravarty S,Dugar S, et al. Targeting endogenous transforming growth factorbeta receptor signaling in SMAD4-deficient human pancreatic car-cinoma cells inhibits their invasive phenotype1. Cancer Res 2004;64:5200–11.

30. Rao ZY, Cai MY, Yang GF, He LR, Mai SJ, Hua WF, et al. EZH2 supportsovarian carcinoma cell invasion and/or metastasis via regulation of TGF-beta1 and is a predictor of outcome in ovarian carcinoma patients.Carcinogenesis 2010;31:1576–83.

31. Hasegawa S, Nagano H, Konno M, Eguchi H, Tomokuni A, Tomimaru Y,et al. A crucial epithelial to mesenchymal transition regulator, Sox4/Ezh2axis is closely related to the clinical outcome in pancreatic cancer patients.Int J Oncol 2016;48:145–52.

32. Liu L, Xu Z, Zhong L, Wang H, Jiang S, Long Q, et al. Enhancer of zestehomolog 2 (EZH2) promotes tumour cell migration and invasion viaepigenetic repression of E-cadherin in renal cell carcinoma. BJU Int2016;117:351–62.

33. Deer EL, Gonzalez-Hernandez J, Coursen JD, Shea JE, Ngatia J, Scaife CL,et al. Phenotype and genotype of pancreatic cancer cell lines. Pancreas2010;39:425–35.

34. Jazag A, Ijichi H, Kanai F, Imamura T, Guleng B, Ohta M, et al. Smad4silencing in pancreatic cancer cell lines using stable RNA interference andgene expression profiles induced by transforming growth factor-beta.Oncogene 2005;24:662–71.

35. Leung YK, Chan QK, Ng CF, Ma FM, Tse HM, To KF, et al. Hsa-miRNA-765as a key mediator for inhibiting growth, migration and invasion infulvestrant-treated prostate cancer. PLoS One 2014;9:e98037.

36. Liang L, Li X, Zhang X, Lv Z, He G, Zhao W, et al. MicroRNA-137, anHMGA1 target, suppresses colorectal cancer cell invasion andmetastasis inmice by directly targeting FMNL2. Gastroenterology 2013;144:624–35e4.

37. Ma F, Song H, Guo B, Zhang Y, Zheng Y, Lin C, et al. MiR-361-5p inhibitscolorectal and gastric cancer growth and metastasis by targeting staphy-lococcal nuclease domain containing-1. Oncotarget 2015;6:17404–16.

38. Sun V, Zhou WB, Nosrati M, Majid S, Thummala S, de Semir D, et al.Antitumor activity of miR-1280 in melanoma by regulation of Src. MolTher 2015;23:71–8.

39. Zhang Q,Wei T, Shim K,Wright K, Xu K, Palka-Hamblin HL, et al. Atypicalrole of sprouty in colorectal cancer: sprouty repression inhibits epithelial-mesenchymal transition. Oncogene 2016;35:3151–62.

40. Zhang Y, He X, Liu Y, Ye Y, Zhang H, He P, et al. microRNA-320a inhibitstumor invasion by targeting neuropilin 1 and is associated with livermetastasis in colorectal cancer. Oncol Rep 2012;27:685–94.

41. Zhang Y, Li T, Guo P, Kang J, Wei Q, Jia X, et al. MiR-424-5p reversedepithelial-mesenchymal transition of anchorage-independent HCC cellsby directly targeting ICAT and suppressed HCC progression. Sci Rep2014;4:6248.

42. Yang Y, Wang LL, Wang HX, Guo ZK, Gao XF, Cen J, et al. The epigenet-ically-regulated miR-663 targets H-ras in K-562 cells. FEBS J 2013;280:5109–17.

43. Shi Y, ChenC, Zhang X, LiuQ, Xu JL, ZhangHR, et al. Primate-specificmiR-663 functions as a tumor suppressor by targeting PIK3CD and predicts theprognosis of human glioblastoma. Clin Cancer Res 2014;20:1803–13.

44. Zang W, Wang Y, Wang T, Du Y, Chen X, Li M, et al. miR-663 attenuatestumor growth and invasiveness by targeting eEF1A2 in pancreatic cancer.Mol Cancer 2015;14:37.

45. Scotto L, Assoian RK. AGC-rich domainwith bifunctional effects onmRNAand protein levels: implications for control of transforming growth factorbeta 1 expression. Mol Cell Biol 1993;13:3588–97.

46. HuW, Xu S, Yao B, HongM,Wu X, Pei H, et al. MiR-663 inhibits radiation-inducedbystander effects by targeting TGFB1 in a feedbackmode. RNABiol2014;11:1189–98.

47. LiQ, ChengQ, Chen Z, Peng R, ChenR,Ma Z, et al.MicroRNA-663 inhibitsthe proliferation,migration and invasionof glioblastoma cells via targetingTGF-beta1. Oncol Rep 2016;35:1125–34.

48. Wang Z, Zhang H, Zhang P, Dong W, He L. MicroRNA-663 suppresses cellinvasion and migration by targeting transforming growth factor beta 1 inpapillary thyroid carcinoma. Tumour Biol 2016;37:7633–44.

49. Sun F, Lee L, Zhang Z, Wang X, Yu Q, Duan X, et al. Preclinical pharma-cokinetic studies of 3-deazaneplanocin A, a potent epigenetic anticanceragent, and its human pharmacokinetic prediction using GastroPlus. Eur JPharm Sci 2015;77:290–302.

50. Vizio B, Novarino A, Giacobino A, Cristiano C, Prati A, Ciuffreda L, et al.Potential plasticity of T regulatory cells in pancreatic carcinoma in relationto disease progression and outcome. Exp Ther Med 2012;4:70–8.

51. Herbertz S, Sawyer JS, Stauber AJ, Gueorguieva I, Driscoll KE, Estrem ST,et al. Clinical development of galunisertib (LY2157299 monohydrate), asmall molecule inhibitor of transforming growth factor-beta signalingpathway. Drug Des Dev Ther 2015;9:4479–99.






2016;14:1124-1135. Published OnlineFirst September 13, 2016.Mol Cancer Res Hardik R. Mody, Sau Wai Hung, Mohammad AlSaggar, et al. Cancer: Putative Roles of miR-663a and miR-4787-5p

1-Induced EMT and Metastasis in PancreaticβAttenuates TGFInhibition of S-Adenosylmethionine-Dependent Methyltransferase

Updated version

10.1158/1541-7786.MCR-16-0083doi:

Access the most recent version of this article at:

Material

Supplementary

http://mcr.aacrjournals.org/content/suppl/2016/10/06/1541-7786.MCR-16-0083.DC1

Access the most recent supplemental material at:

Cited articles

http://mcr.aacrjournals.org/content/14/11/1124.full#ref-list-1

This article cites 51 articles, 3 of which you can access for free at:

Citing articles

http://mcr.aacrjournals.org/content/14/11/1124.full#related-urls

This article has been cited by 2 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mcr.aacrjournals.org/content/14/11/1124To request permission to re-use all or part of this article, use this link



http://mcr.aacrjournals.org/lookup/doi/10.1158/1541-7786.MCR-16-0083

http://mcr.aacrjournals.org/content/suppl/2016/10/06/1541-7786.MCR-16-0083.DC1

http://mcr.aacrjournals.org/content/14/11/1124.full#ref-list-1

http://mcr.aacrjournals.org/content/14/11/1124.full#related-urls

http://mcr.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://mcr.aacrjournals.org/content/14/11/1124


Inhibition of S-Adenosylmethionine-Dependent ... · knockdown suggests novel mechanisms contribute...

Documents

Transcript of Inhibition of S-Adenosylmethionine-Dependent ... · knockdown suggests novel mechanisms contribute...