Hypomethylating therapy in an aggressive stroma-rich model of pancreatic carcinoma

Therapeutics, Targets, and Chemical Biology

Hypomethylating Therapy in an Aggressive Stroma-RichModel of Pancreatic Carcinoma

Reena Shakya1, Tamas Gonda2, Michael Quante2, Martha Salas1, Samuel Kim1, Jenna Brooks1,Steffen Hirsch1, Justine Davies1, Angelica Cullo1, Kenneth Olive2,3, Timothy C. Wang2,Matthias Szabolcs3, Benjamin Tycko1,3, and Thomas Ludwig1,3

AbstractPancreatic ductal adenocarcinoma (PDAC) is a lethal malignancy that resists current treatments. To test

epigenetic therapy against this cancer, we used the DNA demethylating drug 5-aza-20-deoxycytidine (DAC) in anaggressive mouse model of stromal rich PDAC (KPC-Brca1 mice). In untreated tumors, we found globallydecreased 5-methyl-cytosine (5-mC) in malignant epithelial cells and in cancer-associated myofibroblasts (CAF),along with increased amounts of 5-hydroxymethyl-cytosine (5-HmC) in CAFs, in progression from pancreaticintraepithelial neoplasia to PDAC. DAC further reduced DNA methylation and slowed PDAC progression,markedly extending survival in an early-treatment protocol and significantly though transiently inhibiting tumorgrowth when initiated later, without adverse side effects. Escaping tumors contained areas of sarcomatoidtransformation with disappearance of CAFs. Mixing-allografting experiments and proliferation indices showedthat DAC efficacy was due to inhibition of both the malignant epithelial cells and the CAFs. Expression profilingand immunohistochemistry highlighted DAC induction of STAT1 in the tumors, and DAC plus IFN-g producedan additive antiproliferative effect on PDAC cells. DAC induced strong expression of the testis antigen deletedin azoospermia-like (DAZL) in CAFs. These data show that DAC is effective against PDAC in vivo and provide arationale for future studies combining hypomethylating agents with cytokines and immunotherapy. Cancer Res;73(2); 885–96. �2012 AACR.

IntroductionPancreatic ductal adenocarcinoma (PDAC) is a devastat-

ing cancer for which current chemotherapy offers onlymodest improvements in survival. Surgical resection of localtumors can prolong survival, but more than 70% of patientspresent with advanced disease, lowering the predicted over-all survival to just 4- to 5 months. As a result, PDAC is thefourth leading cause of cancer-related mortality in Europeand North America. PDACs are among the most highlydesmoplastic tumors known: typically, neoplastic epithelialcells comprise only a small fraction of the tumor mass,suggesting that the stromal cells play a significant role in

the biology of these tumors. In line with this idea, experi-mental data in various tumor systems have consistentlyshown that cancer-associated myofibroblasts (CAF) andother stroma cells actively promote tumor growth andprogression (1–8). Indeed, multiple lines of evidence arestarting to suggest that targeting CAFs may be an effectiveapproach to treat cancer (9, 10). For example, one of us (K.Olive) recently used an inhibitor of the hedgehog pathway totarget the stromal cells of pancreatic tumors in geneticallyengineered mice, resulting in substantial albeit transientresponses in most tumors when combined with the cyto-toxic nucleoside analog gemcitabine (9). More recently, anenzyme that degrades hyaluronic acid, a key component ofthe extracellular matrix, was used to treat PDAC-bearingmice, resulting in depletion of CAFs and a significant sur-vival benefit when combined with gemcitabine (11, 12).

Drugs that target global DNA methylation are anothernew and promising approach against solid tumors. Hypo-methylating agents such as 5-aza-20-deoxycytidine (decita-bine; DAC) are already used in low-dose regimens to prolongsurvival in patients with myeloid leukemias and myelodys-plastic syndrome (AML/MDS; refs. 13, 14), and are now beingstudied in similar low-dose protocols for their effects againstsolid tumors (15). In principle, hypomethylating agentscould exert antitumor effects not only on the neoplasticcell population, but also by killing or growth-arrestingcancer-associated stromal cells. In this regard, we previouslyreported consistent findings of reduced global DNA

Authors' Affiliations: 1Institute for Cancer Genetics; 2Division of Digestiveand Liver Diseases, Department of Medicine; and 3Department of Pathologyand Cell Biology, Columbia University Medical Center, New York, New York

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

R. Shakya and T. Gonda are co-first authors of this article.

Current address for R. Shakya and T. Ludwig: Department ofMolecular andCellular Biochemistry, Ohio State University Wexner Medical Center,Columbus, OH 43210.

Corresponding Author: Benjamin Tycko, Columbia University, 1130 St.Nicholas Avenue, New York, NY 10032. Phone: 212-851-5280; Fax: 212-851-5284; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-12-1880

�2012 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 885

methylation and focally increased gene-specific methylationin CAFs from gastric carcinomas (16). These observationssuggest a therapeutic opportunity to concurrently targetmalignant epithelial cells and their supportive CAFs usinghypomethylating drugs, to achieve a net antiproliferativeeffect from activation of growth suppressor genes andinduction of a genome-wide hypomethylation crisis in thetumor cells, which already have hypomethylated genomes atbaseline (10). Here, we test these ideas by using low-dosesingle-agent DAC in a rapid onset stroma-rich mouse modelof PDAC.

Materials and MethodsGenetically modified mice

All mice used in this study were housed in an Associationfor Assessment and Accreditation of Laboratory Animal Care(AAALAC)-accredited facility and all procedures relating to thecare and use of animals were conducted at Columbia University(NewYork) in accordancewith theNIHguidelines. TheBrca1flex2

and p53flex7 mice have been previously described (17, 18). Themouse strains p53LSL-R270H (strain number 01XM3), KrasLSL-G12D

(strain number 01XJ6), and Pdx1-cre (strain number 01XL5)were obtained from the National Cancer Institute (NCI) Freder-ick Mouse Repository. All mice generated in this study weremaintained on a mixed 129/B6 genetic background.

DAC administration to miceIntraperitoneal injection of DAC (Sigma-Aldrich) was con-

ducted once weekly according to the treatment schedulesoutlined. A total of 5 mg/mL dilutions were made in PBS freshevery day of treatment. Hamilton syringes were used to injectthemice with DAC [100 mL; dose of DAC 1 mg/g of body weight,similar to prior studies in other mouse models of cancer (19,20)] after weighing them on each day of treatment. An equalvolume of PBS was injected in the control animals. Mice weresacrificed using isoflurane inhalation and cervical disclocation.At necropsy the entire pancreas was removed and weighed.Tissue was fixed in 4% formaldehyde or snap-frozen for furtheranalysis.

Immunohistochemistry and immunofluorescenceFor detecting nuclear 5-methyl-cytosine (5-mC), we used a

mouse monoclonal antibody (Ab-1, Calbiochem), with a pro-tocol that has been previously validated by our group (16). Anidentical protocol was followed for staining or costaining witha polyclonal anti-5-hydroxymethyl-cytosine (5-HmC) antibody(Cat # 39769, ActiveMotif). Standard IHC and immunofluores-cence staining protocols were used for detecting alpha-smoothmuscle actin (ASMA; ab5694, Abcam), p53 [Novocastra, rabbitpolyclonal antibody (CM5), Cat# NCL-p53-CM5p], STAT1(ab31639, Abcam), STAT2 (ab53132, Abcam), Ki67 (M7248,Dako), and deleted in azoospermia-like (DAZL; ab34139,Abcam). For dual staining of 5-mC and ASMA or 5-mC and5-HmC the above procedure was followed for the anti-mCstaining and subsequently slides were incubated with the anti-ASMA or anti-5-HmC antibody and developed. The intensity ofnuclear stainingwasmeasured using Image-J software (Image Jv 1.38) after outlining the nuclear area of epithelial or stromal

cells based on morphology, location, and ASMA or vimentinpositivity. The Ki67 index was calculated on the basis of 7 DACand PBS treated stage- or age-matched cases and expressed aspercentage of cells per high-power field averaged over 10 fieldsin each case.

Cytosine acceptor assay for DNA methylationA fluorometric assay that relies on differential cytosine

incorporation after digestion with the methylation-sensitiveHpaII restriction enzyme (Epigentek) was used to measure thepercentage of methylated cytosines in all genomic HpaIIrestriction sites. Results were normalized on the basis of acalibration curve ofmixtures of fully methylated and unmethy-lated DNA.

Methylation-sensitive bisulfite sequencing and cloningGenomic DNA was bisulfite-converted using the Epitech

Bisulfite Kit (Qiagen, S9104), followed by PCR with the follow-ing primer pairs designed using MethPrimer (21): STAT1amplicon A, forward: AGAGGTAGAAATAGGAGTTAGATTAA-T, reverse: AACAAAACCTCTTTACTTCTTCCTTTC; ampliconB, forward: AATTTTGTAAATATAATTTTTTGTTTG, reverse:ATCTCCAAAAAACTTTAACAAACTC; amplicon C, forward:ACATTTATTTGGGATATTGTTGAG, reverse: ATCATTTTAC-CCTAAAAATAAAAAC. The PCR cycling conditions were aninitial denaturation at 94�C for 1 minute, followed by 35 cycleswith denaturation for 30 seconds and extensions for 1 minuteat 72�C, with annealing temperatures starting at 60�C andtouching down 1� in each cycle until a plateau at 51�C. The PCRproducts were cloned in Escherichia coli (TOPO-TA CloningKit, Life Sciences) and multiple clones derived from eachoriginal genomic sample were sequenced to determine thepattern of methylated and unmethylated cytosines.

Gene expression profilingTotal RNA from cultured pancreatic cancer cells and CAFs

was prepared using Trizol reagent (Invitrogen). Microarrayanalysis for mRNA expression was conducted using theMouseWG-6 v2 Bead Chip Expression arrays (Illumina).Each array contained more than 45,000 probes based onRefSeq release 22, supplemented with MEEBO and RIKENFANTOM2 content. Normalized expression data were ana-lyzed in dChip (22) by ANOVA and supervised hierarchicalclustering after first applying a minimum 1.5-fold changecriterion. The probe-level data have been deposited atNational Center for Biotechnology Information (NCBI) GeneExpression Omnibus (GEO; accession GSE42364 andGSE42365). Gene set enrichment analysis was conductedusing Gene set enrichment analysis (GSEA) software (www.broadinstitute.org/gsea).

Isolation of primary cells and pretreatmentmixing-allografting experiments

We derived primary PDAC cells from the KPC-Brca1 mousetumors. From animals at an advanced stage of disease, a smallpiece of mouse pancreatic tumor tissue was dissected awayfrom normal pancreatic parenchyma, then minced, trypsi-nized, and passed through 18- and 21-gauge needles to

Shakya et al.

Cancer Res; 73(2) January 15, 2013 Cancer Research886

dissociate the cells. The tumor tissue was cultured on gelati-nized plates in Dulbecco's Modified Eagle's Medium supple-mented with pencillin/streptomycin, glutamine, and 10% FBSfor 1 to 2 weeks until the tumor cell lines were established.Primary CAFs were isolated from pancreata of 3.5-week-oldKPC-Brca1 mice, which displayed advanced pancreatic intrae-pithelial neoplasia lesions and microinvasive tumors only, tominimize the possibility of contamination with tumor cells.Briefly, half of each pancreas was minced, whereas the otherhalf was fixed in 10% buffered-formalin for histology. Mincedpancreata were trypsinized and passed through 18 G and 21 Gneedles to further disssociate the tissue, followed by plating inRPMI media supplemented with penicillin/streptomycin,L-glutamine and 10% FBS. Cultures were maintained undis-turbed for over a week until the CAFs had migrated out of thetissue. The purity of the CAF and PDAC epithelial cell pre-parations was determined morphologically and by antigenexpression: the short-term CAF culture was verified as mostlyconsisting of ASMA- and vimentin-positive cells (>90%),whereas the PDAC cell culture contained less than 10% ASMA-and vimentin-positive cells andmore than 80%CK-19–positivecells by immunofluorescence. To prevent clonal selection ofCAFs, primary cultures were passaged not more than once ortwice before being used for allograft experiments. Tumor cellsor myofibroblasts (CAFs) were pretreated for 24 hours (epi-thelial tumor cells) or 48 hours (myofibroblasts; slower divid-ing) with DAC (0.25 mmol/L for the tumor cells and 2 mmol/Lfor the CAFs) in cell culture dishes, followed by 24 hoursrecovery. Six- to 8-week-old immunocompromised mice[severe combined immunodeficient (SCID), Jackson Labora-tories) were used for subcutaneous injection of DAC exposedor unexposed control pancreatic tumor cells (2.5 � 105) withand without a mixture of DAC exposed or unexposed controlprimary pancreatic myofibroblasts (CAFs; 5 � 104). Tumorcells or myofibroblasts were pretreated for 24 hours (epithelialcells) or 48 hours (myofibroblasts; slower dividing) withDAC (1mmol/L) in cell culture dishes, followed by 1 day without thedrug. Cell mixtures were then injected into both flanks of eachrecipient mouse. Tumor size was monitored every 3 days for 4weeks. Tumors were dissected at 4 weeks and tumor diameterswere measured.

Testing DAC plus IFN-g as a drug combination in PDACcellsTwo of the primary PDAC cell lines were used for the IFN-g

experiments with or without concomitant DAC exposure. Ingelatin-coated 96-well plates, tumor cells were seeded at adensity of 1,000 cells per well; 48 hours postseeding of cells,treatment was begun; cells were either left untreated, ortreated with the following; mouse IFN-g (100 ng/mL) only, ormouse IFN-g (100 ng/mL) with DAC, or DAC only. IFN-gtreatment was done for 4 days; whenever DAC was includedin the treatment, it was applied at a concentration of 0.5 mmol/L for the first 24 hours followed by 0.05 mmol/L concentrationfor additional 24 hours, then removed. At the end of thetreatment period, cells were incubated with MTT (0.5 mg/mL)for 4 hours followed by solubilization in HCl/SDS-containinglysis buffer. Colorimetric measurements indicating relative

numbers of viable cells were taken on a microplate reader(BioTek Instruments).

ResultsTheKrasLSL-G12D; p53LSL-R270H/þ; Pdx1-cre;Brca1flex2/flex2;(KPC-Brca1) mouse model produces a rapid onsetstroma-rich form of PDAC

Mice carrying conditional oncogenic alleles of K-ras(KrasLSL-G12D; abbreviatedKras�) and the p53 tumor suppressorgene (p53LSL-R270H or p53flex7; abbreviated p53�), both driven bya pancreatic epithelial cell-specific Pdx1-Cre recombinase,have provided one of the most useful experimental models ofPDAC (9, 23, 24). However, while the tumor penetrance is highin thesemice, the tumor latency is long [Fig. 1A; T50¼ 150 daysin our mouse colony; consistent with published data fromHingorani and Tuveson (25, 26)]. To generate a fastermodel fordrug treatment studies, we added to this model a conditionalallele of the Brca1 tumor suppressor gene (abbreviated Brca1-flex2). Adding the Brca1flex2 allele is biologically reasonable, asBRCA1 is known to be downmodulated through nonmuta-tional pathways in human PDAC, and individuals with sometypes of BRCA1 germline mutations are at increased risk forPDAC (27–31). As shownby the survival curves in Fig. 1A, PDACformation in the resulting triple genetically modified KPC-Brca1 (KrasLSL-G12D/þ; p53LSL-R270H/þ; Pdx1-cre; Brca1flex2/flex2)mice is significantly accelerated (T50 ¼ 88 days with thep53R270H point-mutant allele and T50 ¼ 84 days with theconditional p53flex7 null allele). Development of pancreatictumors in thesemice hinges on stochastic loss of the remainingwild-type allele of p53, accelerated by genomic instability fromthe Brca1-deficiency, so the tumors are oligo-clonal. At sacri-fice, the tumors encompass the entire pancreas, animalsaccumulate ascites and they frequently show metastasis todistant organs including liver, lung, and diaphragm, similar tothe metastatic lesions in Kras�; p53R172H/þ; Pdx1-cre; (KPC)animals (25), but with faster onset.

These tumors show many features of the histopathology ofhuman PDAC, including a step-wise progression from PanINprecursor lesions of increasing dysplasia to invasive adeno-carcinoma with malignant glands surrounded by abundantstromalmyofibroblasts (Fig. 1B–E). Stabilizedmutant p53R270H

protein is already easily detectable by IHC in epithelial cellnuclei in early to mid-stage PanIN lesions (Fig. 1B) and, asexpected, is absent from the nonneoplastic stromal myofi-broblasts, which proliferate to form cellular cuffs around everyPanIN lesion (Fig. 1B–D). An additional histopathology in theKPC-Brca1 mice, which is not prominent in the original KPCmodel, is the occurence of pancreatic cysts of various sizes, alllined by epithelial cells that are malignant by cytology and bythe criterion of mutant p53 expression. This feature overlapswith the broader spectrum of human pancreatic histopathol-ogy, as cystic growth patterns can be seen in a subgroup ofhuman pancreatic neoplasms with K-ras mutations (32–34).On the basis of histologic examination inmice necropsied overa range of time points, the invasive PDACs seem to evolve fromPanIN lesions in small ductules, similar to the KPC model,whereas the cystic lesions seem to arise from dysplasticproliferative lesions in the larger pancreatic ducts.

DAC Therapy in PDAC

www.aacrjournals.org Cancer Res; 73(2) January 15, 2013 887

Global 5-mC content decreases in malignant epithelialcells and stromal myofibroblasts early in tumorprogression

The normal pancreas contains only rare ASMA-positivemyofibroblasts, which are thought to derive from pancreaticstellate cells. In early PanIN lesions and all later stages, thesecells express both ASMA and a general mesenchyme marker,the intermediate filament vimentin. Using IHC and immuno-fluorescence with a widely used monoclonal antibody against5-mC, we found a decrease in nuclear 5-mC content both in thestromal and malignant epithelial compartments during theprogression from early PanIN to late PanIN and invasive cancer(Fig. 2A–C; quantitation by image analysis in SupplementaryFig. S1). We previously showed in another type of cancer,gastric adenocarcinoma, that there are differences in 5-mCcontent and amount of immunoreactive DNMT1methyltrans-ferase in CAFs compared with malignant epithelial cells(16, 29). Similar to those data, the intensity of nuclear andcytoplasmic DNMT1 staining of epithelial cells versus CAFs inthe mouse PDACs was very distinct, with less DNMT1, nearlyundetectable at the level of sensitivity of IHC, in thefibroblasticstromal cells (Fig. 2D). These observations by IHC are strikingboth in themousemodel and in human PDAC (SupplementaryFig. S2A). DNMT1 expression is known to be highest inproliferating cells, and a difference in proliferation mayexplain, at least in part, the lower expression of DNMT1 inCAFs. This explanation is supported by our Western blotanalysis comparing DNMT1 protein levels in normal humanfibroblast lines grown in varying concentrations of serum(Supplementary Fig. S2B), and further supported by our anal-ysis of proliferation indices in CAFs versusmalignant epithelialcells in the KPC-Brca1 tumors (see later).

As the BRCA1 protein can interact with DNMT1 (35), wefurther asked whether our findings of reduced global DNAmethylation in the early to late PanIN transition would gen-eralize to the other transgenic models of pancreatic cancer, inwhich the Brca1 gene is not modified (i.e., Kras/Ink4aArf). IHCon sections of PanIN and PDAC from this slower onset cancermodel showed a similar progressive reduction in 5-mC (datanot shown), indicating that the phenomenon is characteristicof Kras-driven PDAC.

CAFs accumulate 5-HmC early in tumor progressionGiven recent results suggesting that the modified base 5-

HmC is a precursor to active or passive DNA demethylation invivo, we used triple-color immonofluorescence to assessthe relative nuclear content of 5-mC and 5-HmC in epithelialand stromal cells during PanIN progression, with ASMA orvimentin visualized as cytoplasmic markers for stromal myo-fibroblasts/CAFs. As noted earlier, at the early PanIN stageASMA-positive myofibroblasts begin to proliferate, formingwell-defined cellular cuffs around PanIN lesions. Triple immu-nofluorescence revealed that a majority of the rare vimentin-and ASMA-positive stromal cells in the nonneoplastic pan-creatic parenchyma, and many of these cells in the earliestPanIN lesions, have strong 5-mC staining with low levels of5-HmC, whereas with progression from early to late PanIN, weobserved a strong and uniform accumulation of nuclear5-HmC in the CAFs (Fig. 2A–C and Supplementary Fig. S1).Together with our previous findings in human and mousegastric cancers, these new results in PanIN and PDAC showthat CAFs differ from normal resident myofibroblasts in termsof their global genomic 5-mC and, as shown here, 5-HmC, andare also strikingly epigenetically distinct from malignant

Figure 1. Survival andhistopathology in the KPC-Brca1rapid-onset model of PDAC. A,Kaplan–Meier survival curvefor KrasLSL-G12D; p53LSL-R270H/þ;Pdx1-cre; Brca1flex2/flex2

(KPC-Brca1) mice comparedwith KrasLSL-G12D; p53LSL-R270H/þ;Pdx1-cre (KPC) mice. The KPC-Brca1 model results in asignificantly faster progression ofcancerwith aT50 of 88days (n¼32)compared with a T50 of 150 days inKPC mice (n ¼ 28; P < 0.001). B,PanIN lesion consisting ofproliferating p53R270H-positivedysplastic epithelial cells,surrounded by a dense cuff ofreactive nonneoplastic stromalmyofibroblasts, which are p53-negative. C to E, progression fromearly PanIN to late PanIN andinvasive adenocarcinoma isassociatedwith an accumulationofASMA-positive CAFs (asterisks).

Shakya et al.


epithelial cells in the relative global content of these 2 importantmarks. As noted earlier, the DNMT1 methyltransferase wasdetected by IHC at higher levels in the epithelial cells of thePDAC tumors, both in the mouse model and in human PDACs.DNMT1 expression is known tobe proliferation-dependent, andthis difference in expression is at least in part accounted for byslower proliferation of the CAFs (Fig. 3E andSupplementary Fig.S3). From a therapeutic standpoint, the relatively low baselinelevels of 5-mC in genomic DNA of both CAFs and malignantepithelial cells might contribute to a good therapeutic index forhypomethylating drugs, with less adverse effects on normalcells, whose genomes are more replete with 5-mC.

Low doses of DAC significantly inhibit tumor PDACtumor progression with minimal side effectsWe next tested the effects of single-agent DAC in the

PDAC model, starting treatment of the mice either at anearly time point when PanIN was established and micro-

invasive lesions were beginning to form, without overt tumorformation, or later after tumor formation (Table 1). We firstcarried out an early intervention study with low-dose DAC orPBS control intraperitoneal injections starting at 3 weeks ofage, with necropsies conducted at a single time point after 5weeks of treatment. We then repeated the same protocol as asurvival experiment, carrying out necropsies on the miceonly after they developed large tumors and preterminalmorbidity.

In our initial pilot experiment, starting at 3 weeks of age andfollowing 5 intraperitoneal injections of DAC (1 mg/g of bodyweight) or PBS once per week the animals were sacrificed fornecropsies. Analysis of the treated KPC-Brca1 mice showedthat DAC markedly reduced the PDAC tumor burden (Fig. 3Aand B) and reduced the amount of invasive cancer comparedwith littermate KPC-Brca1mice injected with PBS. At this timepoint, the majority of the animals in the PBS-treated controlgroup had large invasive tumors and concomitantly reducednormal pancreas. In contrast, the DAC-treated animals hadmuch smaller tumor nodules, with abundant remaining nor-mal pancreas and either absent or reduced tumor invasion(Table 1).

In the early treatment to survival experiment, we treatedgroups ofmice starting at the same early time point (3 weeks ofage) using the same dose and schedule of intraperitoneal DACversus PBS vehicle control. The results from measurements ofglobal 5-mC in tumors at necropsy indicated a significantpharmocodynamic effect of the drug, with substantial reduc-tions in genomic 5-mC in theDAC-treated versus sham-treatedtumors aswell as in spleenDNA from the same animals (Fig. 3Cand data not shown), and there was a marked survival benefitfor the mice, with a prolongation of the T50 from 87 to 128 days(P < 0.0001; Table 1 and Fig. 3D). In a third experiment, we useda later treatment paradigm, starting DAC at 8 weeks of age,which may be more reflective of its potential use in advancedhuman PDACs. In a group of mice sacrificed at day 0 oftreatment (n ¼ 6), all had invasive carcinomas. As shownin Table 1 and Fig. 3D, even when started at this later timethe single-agent DAC therapy substantially prolonged survival,increasing the T50 from 87 to 110 days (P < 0.0001). IHC for theproliferation marker Ki67 revealed that the CAFs were prolif-erating in the untreated tumors, albeit at a slower rate than themalignant epithelial cells, and that there was significant sup-pression of proliferation of both cell types by DAC treatment(Fig. 3E and Supplementary Fig. S3). From these experiments,we conclude that low-dose single-agent DAC has promisinganticancer activity in this mouse model of aggressive stroma-rich PDAC.

Importantly, no adverse effects on a group of 6 normalcontrol mice were noted with this dose regimen of DAC; wefound that the standard blood counts were not affected by thistreatment, with average hemoglobin (g/dL) 9.2 in DAC versus9.0 in control (reference range, 11–15.1), and white blood cells(WBC; �103/mL) 4.7 in DAC; 6.6 in control (reference range,1.8–10.7). In addition, each of 2 male wild-type mice given thisdrug regimen and then allowed tomate with wild-type femalesmaintained their reproductive function as evidenced by pro-duction of normal offspring.

Figure 2. Stage-specific and cell type–specific levels of genomic5-mC and 5-HmC and DNMT1 in stromal and epithelial cells duringtumor progression. A, the section of early PanIN and adjacenthistologically normal pancreatic parenchyma was subjected to tripleimmunofluorescence with antibodies recognizing vimentin (blue), 5-mC(green), and 5-HmC (red). Most of the proliferating epithelial cells in theearly PanIN lesion (asterisk) stain strongly for 5-mCandonly weakly for 5-HmC.The5-mCsignal is partly punctate inmouse cell nuclei, presumablyreflecting concentrations of heterochromatin, whereas the 5-HmC signalismore diffuse. The reactive stromalmyofibroblasts (arrows) are primarilystaining for 5-mC but are starting to accumulate 5-HmC. B, late PanINlesion in which the majority of vimentin-positive CAFs (arrows) showa stronger immunofluorescence signal for 5-HmC than for 5-mC. Thereis a moderate reduction in the average 5-mC immunofluorescencesignal in the proliferating dysplastic epithelial cells of the PanIN lesions(asterisk) but, contrasting with the stromal cells, this is not accompaniedby a strong increase in 5-HmC. C, in PDAC there is reduced 5-mCsignal intensity in the malignant epithelial cells (quantitative immuno-fluorescence measurements in Supplementary Fig. S1), with only a weakaccumulation of 5-HmC, whereas the proliferating CAFs show even lessstaining for 5-mC, contrastingwith a very strong signal for 5-HmC.D, IHCshowing greater DNMT1 expression in PDAC epithelial cells comparedwith the surrounding stromal CAFs.

DAC Therapy in PDAC


Mixing-allografting shows an additive antiproliferativeeffect of DAC on malignant epithelial cells and CAFs inreconstituted tumors

Because in vivo treatment in the KPC-Brca1 mouse modelcannot dissect whether the efficacy of DAC in inhibitingoverall tumor growth is due to its effects on tumor cells or ontheir supportive stromal cells or both, we next designed atumor reconstitution-allografting experiment to address thisissue. In this protocol, only pancreatic CAFs or only themalignant epithelial cells were pretreated with DAC fol-lowed by mixing of the 2 types of cells and allografting inimmunodeficient nude mice. We derived the primary PDACcell lines as well as the pancreatic CAFs from the KPC-Brca1mouse model; the pancreatic CAFs were maintained inculture for a short time (<1 week) to minimize clonalselection and possible epigenetic changes induced by pro-longed culture. In these short-term cultures, we found thatthe epithelial cells were growth-inhibited within 2 days byvery low amounts of DAC (0.25 mmol/L), whereas the CAFswere more resistant and continued to proliferate slowly inculture for up to a week in the presence of 2 mmol/L DAC.The CAFs, or separately the malignant epithelial cells, werepretreated (or untreated for control cells) with these con-centrations of DAC for a brief period of 24 hours in culture,

followed by continued cell culture for 1 day without drug,and were then injected together (carcinoma cells:CAFs in a10:1 ratio), or the carcinoma cells separately, subcutaneouslyinto the flanks of nonobese diabetic/severe combined immu-nodeficient (NOD/SCID) mice. Following injection, we mon-itored the allograft tumors by external examination for 3weeks, then sacrificed the animals and processed the tumorsfor size measurements and histology. The results (Fig. 4A)allowed several conclusions: (i) while the malignant epithe-lial cells by themselves can form tumors, adding CAFs has apositive effect on tumor growth; (ii) pretreating the carci-noma cells alone with DAC significantly reduces the growthof the tumor allografts; and (iii) treating the CAFs well beforemixing them with the malignant epithelial cells leads to aneven greater net antitumor effect.

DAC upregulates IFN-inducible genes in PDAC cells andIFN-g potentiates its antiproliferative effect

To begin to study the biologic mechanisms of DAC againstPDAC, and to gain insights that could lead to rational choices ofsecond agents to combine with DAC therapy, we conductedmicroarray-based gene expression profiling on short-termcultures of maligant epithelial cells and, separately, CAFsisolated from the KPC-Brca1 mouse tumors, with or without

Figure 3. Low doses of DAC inhibit tumor PDAC tumor progression in KPC-Brca1 mice. A, results of the initial pilot experiment in which DAC or PBS controlinjections were started at 3 weeks of age, with necropsies conducted at a single time point after 5 weeks of treatment. Pancreatic weight is mostlyaccounted for by tumor mass (see B). B, gross photographs of representative pancreata from the pilot experiment. C, levels of 5-mC in genomicDNA measured by the cytosine incorporation assay in tumors from the experiments in which mice were treated with DAC or PBS vehicle control for6 to 7 weeks. D, Kaplan–Meier survival curves showing the highly significant antitumor effects of DAC when started either at 3 weeks (all mice with PanINat the start of therapy; few with invasive lesions) or at 8 weeks of age (all mice with invasive PDAC at the start of therapy). E, reduction in proliferationindex (% Ki67-positive cells per high power field in 10 fields in 7 cases) by DAC in both CAFs and malignant epithelial cells.

Shakya et al.


DAC exposure. When we queried these data for differentiallyexpressed genes a large set of known IFN-inducible genes,including the STAT1 (>5-fold at the mRNA level in epithelialand 2-fold in CAFs) and STAT2 (>2-fold increase of mRNA inboth epithelial and CAF cultures treated with DAC vs. PBS)genes among others, were found to be upregulated by the DACtreatment (Supplementary Table S1). GSEA confirmed that IFNresponse and other cytokine signaling–induced genes werestatistically over-represented in the DAC-induced transcripts(Supplementary Fig. S4 and Supplementary Table S2). Becausewe postulated that the difference in their expressionwas due toaltered CpGmethylation, we conducted bisulfite sequencing ofthe extended STAT1 promoter region, which revealed demeth-ylation of an upstream region in DAC-treated primary cultures

of the PDAC epithelial cells, as well as in whole tumors treatedin vivo (Fig. 4B). We have not studied histone modifications intumor cells from the KPC-Brca1 mice, but alignment of ourbisulfite amplicons with ENCODE data (31) from a normalendodermal tissue (mouse fetal liver) suggests that the DAC-responsive region is marked by a block of histone modification(H3K4m1) that may be able to act in a regulatory fashion bybeing the first step in formation of the fully repressive H3K4m3mark.

Importantly, analysis of tissue sections by IHC confirmedthat this transcriptional upregulation led to markedlyincreased expression of STAT1 at the protein level, and amodest increase in STAT2, in the malignant epithelial cellsof the PDAC tumors that had been treated with DAC in vivo

Table 1. Effects of single-agent DAC in the KrasLSL-G12D; p53LSL-R270H/þ; Pdx1-cre; Brca1flex2/flex2

(KPC-Brca1) mouse model of PDAC

Treatmentstarted(age in wks)

Age atnecropsy(wks)

Treatment(number of mice)a Clinical efficacy

Pancreatic tumorhistopathology

Untreated 3 wk (n ¼ 4) Untreated N/A 4 of 4 animals had high-grade PanIN lesions;2 of 4 animals hadmicroinvasive cancers

Untreated 8 wk (n ¼ 6) Untreated N/A 6 of 6 animals had one ormore invasive cancers

3 wk 8 wk PBS (n ¼ 3) Control 3 of 3 animals had one ormore invasive cancers

DAC (n ¼ 3) Tumor size reduced by�90% (PBS-treated tumors,average size ¼ 12.5 mm;DAC-treated tumors,average size ¼ 1.3 mm)b

1 of 3 animals had a singleinvasive cancer

3 wk 9 wk PBS (n ¼ 3) Control 3 of 3 animals had one ormore invasive cancers

DAC (n ¼ 4) Tumor size reduced by �75%(PBS-treated tumors,average size ¼ 10.7 mm;DAC-treated tumors,average size ¼ 2.7 mm)c

2 of 4 animals had oneinvasive cancer peranimal

3 wk Survival study PBS (n ¼ 19) vs.DAC (n ¼ 12)

DAC extends survival by40.5 d; PBS (87 d) vs. DAC(127.5 d); P < 0.0001(log-rank test)

(See below)

8 wk Survival study PBS (n ¼ 19) vs.DAC (n ¼ 21)

DAC extends survival by23 days; PBS (87 days) vs.DAC (110 days); P < 0.0001(log-rank test)

Sarcomatoid change inmalignant epithelial cellsand loss of CAFs intumors escaping fromDAC

aThe treated animals received 1 mg of DAC/g of body weight or PBS vehicle control by intraperitoneal injection once per week.bPancreas weight reduced by 75% compared with PBS-treated pancreata (PBS-treated pancreata, average weight ¼ 0.8 g;DAC-treated pancreata, average weight ¼ 0.2 g).cPancreas weight reduced by 88% (PBS-treated pancreata, average weight ¼ 1.72 g; DAC-treated pancreata, average weight ¼0.21 g).

DAC Therapy in PDAC


(Fig. 5A). We also addressed whether DAC activates genes atboth the mRNA and protein level in the CAFs in vivo. From ourexpression profiling data, we selected those genes that showedsignificant upregulation following DAC treatment in CAFs andhad suitable antibody reagents available for IHC and/or immu-nofluorescence.DAZL is one of the top such genes and IHC andimmunofluorescence for DAZL protein expression in factshowed a marked increase in DAC-treated tumors in the CAFs(Fig. 5B).

Finally, on the basi of these results of gene expressionprofiling, IHC, and bisulfite sequencing, we postulated thatDAC-treatment might sensitize PDAC cells to antiproliferativecytokine signaling, including signaling by IFNs (36–38). Totest for additive or synergistic interactions between DAC andIFN-g , we conducted proliferation assays followingDAC, IFN-g ,or the combination of both agents, in short-term cultures ofPDAC cells from the KPC-Brca1 mice. These experimentsshowed an additive antiproliferative effect of IFN-g plus DAC(Fig. 6).

PDAC tumors that escape from DAC treatment havereduced stroma and sarcomatoid transformation ofmalignant epithelial cells

We next examined the effects of DAC treatment on thehistopathology of the PDAC tumors that ultimately escapedfrom inhibition by the drug and progressed to kill themice. Forthis purpose, we compared the histology of the DAC-exposedtumors to the histology of similar-sized tumors from the sham-treated (PBS injected) KPC-Brca1 animals. Two differences,namely, zonal necrosis and large areas of sarcomatoid changein the malignant component, were evident on low-powerexamination of the histologic sections from the DAC-treatedtumors comparedwith the vehicle control tumors. To unequiv-ocally distinguish between sarcomatoid change in the tumorcells, versus possible atypia in stromal cells mimicking sarco-matoid change, we used anti-p53 and anti-ASMA antibodies todistinguish cells of neoplastic epithelial origin from reactivemyofibroblasts. The results, from examining and quantifying5 DAC and 5 PBS control cases, were quite striking: in the

Figure4. Tumor reconstitution-allografting experiment and reduced promoter CpGmethylation of theSTAT1gene inPDACcells after DAC treatment. A, short-term cultures of pancreaticCAFsor PDACepithelial cells from theKPC-Brca1 tumorswere pretreatedwithDACor grownwithout the drug, followed bymixingof the 2 types of cells and allografting into the flanks of nude mice. Mice were sacrificed after 3 weeks and the tumor diameters were measured. Asindicated by comparing condition A to condition B, the presence of CAFs promotes tumor growth. As indicated by comparing condition B to conditionC, pretreating the carcinoma cells alone significantly reduces the growth of the tumor allografts. As indicated by comparing condition D to conditionE, treating theCAFs aswell beforemixing themwith themalignant epithelial cells leads to the greatest net antitumor effect. B, bisulfite sequencing of genomicDNA from PDAC tumor cells exposed to 0 or 0.5 mmol/L DAC in short-term culture, and whole PDAC tumors from mice receiving PBS vehicle controlor DAC shows partial demethylation of a potential regulatory region overlapping a block of histone H3K4m1 chromatin modification (from ENCODE/LICRfetal liver ChIP-Seq data) extending from approximately 0.6 to 1 kb upstream of the STAT1 first exon. The bar graph summarizes the bisulfitesequencing data (% methylated CpG's in amplicon a) for the indicated number of in vivo–treated tumors analyzed. Bisulfite sequencing of the 2 amplicons(b and c) overlapping the STAT1 CpG island (light grey) showed nearly complete lack of CpG methylation in both DAC-treated and untreated tumorsamples (data not shown).

Shakya et al.


DAC-treated tumors the ratio of CAFs (ASMA-positive cells) tomutant p53-positive cells (tumor cells of malignant epithelialorigin) was markedly reduced (Supplementary Fig. S5). Thisfinding in the escaping tumors is consistent with the signif-icant reduction in proliferation of CAFs that we had noted in

more acutely treated tumors (Fig. 3E). As the spindled mor-phology of the anaplastic escaping tumor cells in the DAC-treated animals was strongly suggestive of a sarcomatoidchange in the escaping tumors, we evaluated these tumorareas for the general mesenchymal marker vimentin. Wide-spread costaining for vimentin and mutant p53 in the samecells confirmed that these areas indeed consisted almostentirely of malignant epithelial cells that had undergonesarcomatoid change (Supplementary Fig. S5).

DiscussionAlteredDNAmethylation is a fundamental feature of human

cancers: promoter hypermethylation leading to silencing oftumor suppressor genes and tumor antigens, aberrant meth-ylation at specific imprinting control elements, and globallyreduced genomicmethylation affecting repetitive elements aregeneral findings (39). Furthermore, genetic data in mouseembryos leave no doubt that when nuclear DNA methylationis lost to below a critical level cells become growth arrested ornonviable (40). In several previously studied mouse models ofcancer, loss of Dnmt1 through genetic deletion or inhibition ofDNMT enzymes by DAC can delay tumor formation, withmostof these reports using chemopreventive protocols but with afew studies now starting to assess DAC in combination withother agents for treatment of established tumors (19, 20, 41–44). This evidence from mouse models and the molecularanalysis of human tumors, together with clinical findings oftherapeutic benefit from hypomethylating drugs in acutemyeloid leukemias andmyelodysplastic syndrome, have raised

Figure 5. DAC produces strong increases of STAT1 in PDAC epithelial cells and DAZL in CAFs in the KPC-Brca1 tumors in vivo. A, immunostaining for STAT1and STAT2 proteins in PDAC tumors from mice that received DAC or PBS vehicle control shows that DAC induces both proteins, with particularlystrong activation of STAT1 expression. B, IHC and immunofluorescence showing strong induction of the testis antigen DAZL in the stromal myofibroblasts ofthe DAC-treated tumors.

Figure 6. DAC and the JAK-STAT activator IFN-g have an additiveantiproliferative effect on PDAC cells. Primary PDAC epithelial cellsisolated from the KPC-Brca1 mice and grown in short-term culture weretreated for 4dayswith the indicated single agents or combinationsof low-dose DAC (0.5 mmol/L) and IFN-g (100 ng/mL) as described in MaterialsandMethods, followedbyMTTassays for relative numbers of viable cells.The results show an additive antiproliferative effect of DAC plus IFN-g .

DAC Therapy in PDAC


the question of whether hypomethylating drugs might beuseful against common types of human carcinomas. Here, wehave shown that single-agent DAC treatment has a significanttherapeutic benefit in KPC-Brca1 triple genetically modifiedmice that develop an aggressive stroma-rich form of PDAC.Wehave further shown that this effect is mediated both throughinhibition of themalignant epithelial cells and their supportiveCAFs.

How do our results compare with other treatment protocolsin PDAC models? Comparing relative therapeutic efficacy inmouse models of cancer can be difficult because of theheterogeneity of the models and treatment protocols. Inparticular, due to the slow kinetics of the KPC model, whichleads to awide range in the time of tumor onset,most studies inthese mice have used tumor imaging to identify a relativelylarge tumor size (5–7mm) as the starting point of intervention.The KPC-Brca1 model that we have used here is advantageousin that there is progression to invasive cancer over a narrowertimewindow, thus allowing treatment protocols to be initiatedat a specific age of the animals, before the tumor have becomevery advanced. On the other hand, our study carries thelimitation that there is some heterogeneity in tumor size atthe start of treatment. Despite these differences in studydesign, the fractional extension in survival that we haveobserved with single-agent DAC in the KPC-Brca1 mice seemsto be comparable with or greater than that seen in studieswith other agents in the related KPC model. Neither thedirect stromal acting hedgehog pathway inhibitors, nor thestromal-acting enzyme hyaluronidase, led to single-agentsurvival benefits, but these agents had promising activitywhen used in combination with cytotoxic agents. In ourstudy, DAC led to an increase in overall survival from 87 to128 days, which compares favorably to those other studies,even when compounds were being used in combination withgemcitabine. For example, the survival benefit observed bytargeting the sonic hedgehog (SHH) pathway was 25 days ascompared with 11 days in one experiment in whichIPI926þgemcitabine was compared with gemcitabine alonein KPC mice (9). Cyclopamine, in another transgenic modelof Kras/Ink4Arf had a modest survival benefit of 65 days ascompared with 61 days in controls (45). The pegylatedhyaluronidase had a more significant effect in combinationwith gemcitabine. In the KPC model it prolonged survivalfrom 55 to 91 days when started early (12) and from 9 to28 days when started at larger tumor sizes (i.e., later timepoint; ref. 11).

Although it seems unlikely that single-agent DAC treat-ment would be sufficient in the treatment of human PDAC,particularly if started after tumor progression on cytotoxictherapy, it is important to recognize certain unique andpotentially advantageous features of this drug in comparisonwith other drugs studied in PDAC. Most importantly, as itacts to demethylate DNA only over the course of several celldivisions, DAC is deliberately used in low doses, which havelow systemic toxicity. In this respect, DAC is a strongcandidate drug for use in combination with other anticanceragents. The over-representation of cytokine signaling path-way genes that we report here in the set of transcripts

induced by DAC has been seen previously in other cellsystems (46), and these observations point to the possibilityof combining DAC with biologic agents in combinationtherapy. Our experiments showing an additive effect of DACplus IFN-g on PDAC cells in culture directly support thisidea, and our finding that DAC induces strong expression ofthe testis antigen DAZL in CAFs suggests that future studiesshould test combinations of hypomethylating agents withboth cytokines and immunotherapy.

Another shift in paradigm that motivated us to investigateDAC against PDAC is the increasing recognition of the synergybetween tumor cells and their microenvironment, confirmedhere for the KPC-Brca1 model of PDAC in our mixing-allograftexperiment, and the realization that cancer-associated stromalcells can undergo epigenetic changes, including global reduc-tion in CpG methylation, that are analogous, though notidentical, to epigenetic alterations in the tumor cells them-selves. These already reduced baseline levels of 5-mC maycontribute to the therapeutic index of DAC against both PDACcells and their associated CAFs. We have also shown here thatthe distinctmodification 5-HmCdiffers very strikingly betweenCAFs (high 5-HmC/5-mC ratio) and malignant epithelial cells(lower 5-HmC/5-mC ratio). So, another question for futureresearch is whether genomic content of 5-HmC in cancer cellsand their supportive stromal cells will correlate with efficacy ofspecific epigenetically acting drugs. In summary, these pre-clinical results suggest that DAC or other hypomethylatingdrugs may be useful in the treatment of human PDAC, par-ticularly in combination with other rationally chosen drugs orbiologic agents.

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

Authors' ContributionsConception and design: R. Shakya, T. Gonda, B. Tycko, T. LudwigDevelopment of methodology: R. Shakya, T. Gonda, M. Quante, K. Olive, B.TyckoAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): R. Shakya, T. Gonda, M. Quante, M. Salas, S. Kim, J.Brooks, S. Hirsch, J. Davies, A. Cullo, K. Olive, M. Szabolcs, B. TyckoAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): R. Shakya, T. Gonda, M. Quante, S. Kim, J. Brooks, J.Davies, A. Cullo, K. Olive, M. Szabolcs, B. Tycko, T. LudwigWriting, review, and/or revision of themanuscript: R. Shakya, T. Gonda, M.Quante, T.C. Wang, M. Szabolcs, B. Tycko, T. LudwigAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): J. Davies, A. CulloStudy supervision: T. Gonda, B. Tycko, T. Ludwig

AcknowledgmentsThe authors thank the Molecular Pathology and Transgenic Mice Shared

Resources of the Herbert Irving Comprehensive Cancer Center for essentialservices.

Grant SupportThis work was supported by NIH grants U54CA163111 to B. Tycko and T.C.

Wang, U54CA126513 to B. Tycko, T.C. Wang, and T. Ludwig, R21CA125461 to B.Tycko, P01CA097403 to T. Ludwig, and Louis V. Gerstner Junior Scholar Award toT. Gonda.

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

Received May 14, 2012; revised October 1, 2012; accepted November 1, 2012;published OnlineFirst November 29, 2012.

Shakya et al.


References1. AllinenM, Beroukhim R, Cai L, Brennan C, Lahti-Domenici J, Huang H,

et al. Molecular characterization of the tumor microenvironment inbreast cancer. Cancer Cell 2004;6:17–32.

2. Hayward SW,Wang Y, Cao M, Hom YK, Zhang B, Grossfeld GD, et al.Malignant transformation in a nontumorigenic human prostatic epi-thelial cell line. Cancer Res 2001;61:8135–42.

3. Tsujino T, Seshimo I, Yamamoto H, Ngan CY, Ezumi K, Takemasa I,et al. Stromal myofibroblasts predict disease recurrence for colorectalcancer. Clin Cancer Res 2007;13:2082–90.

4. Orimo A, Weinberg RA. Stromal fibroblasts in cancer: a novel tumor-promoting cell type. Cell Cycle 2006;5:1597–601.

5. Desmouliere A, Guyot C, Gabbiani G. The stroma reaction myofibro-blast: a key player in the control of tumor cell behavior. Int J Dev Biol2004;48:509–17.

6. Mahadevan D, Von Hoff DD. Tumor-stroma interactions in pancreaticductal adenocarcinoma. Mol Cancer Ther 2007;6:1186–97.

7. Chung LW, Baseman A, Assikis V, Zhau HE. Molecular insights intoprostate cancer progression: the missing link of tumor microenviron-ment. J Urol 2005;173:10–20.

8. HwangRF,Moore T, ArumugamT,Ramachandran V, AmosKD,RiveraA, et al. Cancer-associated stromal fibroblasts promote pancreatictumor progression. Cancer Res 2008;68:918–26.

9. Olive KP, Jacobetz MA, Davidson CJ, Gopinathan A, McIntyre D,Honess D, et al. Inhibition of Hedgehog signaling enhances deliveryof chemotherapy in a mouse model of pancreatic cancer. Science2009;324:1457–61.

10. Gonda TA, Varro A, Wang TC, Tycko B. Molecular biology of cancer-associated fibroblasts: can these cells be targeted in anti-cancertherapy? Semin Cell Dev Biol 2010;21:2–10.

11. Provenzano PP, Cuevas C, Chang AE, Goel VK, Von Hoff DD, Hingor-ani SR. Enzymatic targeting of the stroma ablates physical barriers totreatment of pancreatic ductal adenocarcinoma. Cancer Cell2012;21:418–29.

12. Jacobetz MA, Chan DS, Neesse A, Bapiro TE, Cook N, Frese KK,et al. Hyaluronan impairs vascular function and drug delivery in amouse model of pancreatic cancer. Gut. 2012 Mar 30. [Epub aheadof print].

13. Kantarjian HM, O'Brien S, Huang X, Garcia-Manero G, Ravandi F,Cortes J, et al. Survival advantage with decitabine versus intensivechemotherapy in patients with higher risk myelodysplastic syn-drome: comparison with historical experience. Cancer 2007;109:1133–7.

14. Kantarjian H, Issa JP, Rosenfeld CS, Bennett JM, AlbitarM, DiPersio J,et al. Decitabine improves patient outcomes in myelodysplastic syn-dromes: results of a phase III randomized study. Cancer 2006;106:1794–803.

15. Shen H, Laird PW. In epigenetic therapy, less is more. Cell Stem Cell2012;10:353–4.

16. Jiang L, Gonda TA, Gamble MV, Salas M, Seshan V, Tu S, et al. Globalhypomethylation of genomic DNA in cancer-associated myofibro-blasts. Cancer Res 2008;68:9900–8.

17. Chen Z, Trotman LC, Shaffer D, Lin HK, Dotan ZA, Niki M, et al. Crucialrole of p53-dependent cellular senescence in suppression of Pten-deficient tumorigenesis. Nature 2005;436:725–30.

18. Ludwig T, Fisher P, Ganesan S, Efstratiadis A. Tumorigenesis inmice carrying a truncating Brca1 mutation. Genes Dev 2001;15:1188–93.

19. Laird PW, Jackson-Grusby L, Fazeli A, Dickinson SL, Jung WE, Li E,et al. Suppression of intestinal neoplasia byDNAhypomethylation.Cell1995;81:197–205.

20. Ecke I, Petry F, Rosenberger A, Tauber S, Monkemeyer S, Hess I, et al.Antitumor effects of a combined 5-aza-20deoxycytidine and valproicacid treatment on rhabdomyosarcoma and medulloblastoma in Ptchmutant mice. Cancer Res 2009;69:887–95.

21. Li LC, Dahiya R. MethPrimer: designing primers for methylation PCRs.Bioinformatics 2002;18:1427–31.

22. Amin SB, ShahPK, YanA, Adamia S,Minvielle S, Avet-LoiseauH, et al.The dChip survival analysis module for microarray data. BMC Bioin-formatics 2011;12:72.

23. Perez-Mancera PA, Tuveson DA. Physiological analysis of oncogenicK-ras. Methods Enzymol 2006;407:676–90.

24. Cook N, Olive KP, Frese K, Tuveson DA. K-Ras-driven pancreaticcancer mouse model for anticancer inhibitor analyses. Methods Enzy-mol 2008;439:73–85.

25. Hingorani SR, Wang L, Multani AS, Combs C, Deramaudt TB,Hruban RH, et al. Trp53R172H and KrasG12D cooperate topromote chromosomal instability and widely metastatic pan-creatic ductal adenocarcinoma in mice. Cancer Cell 2005;7:469–83.

26. Bardeesy N, Aguirre AJ, ChuGC, Cheng KH, Lopez LV, Hezel AF, et al.Both p16(Ink4a) and the p19(Arf)-p53 pathway constrain progressionof pancreatic adenocarcinoma in the mouse. Proc Natl Acad Sci U S A2006;103:5947–52.

27. Ferrone CR, Levine DA, Tang LH, Allen PJ, Jarnagin W, Brennan MF,et al. BRCA germline mutations in Jewish patients with pancreaticadenocarcinoma. J Clin Oncol 2009;27:433–8.

28. Al-SukhniW, RothenmundH, Borgida AE, Zogopoulos G, O'Shea AM,Pollett A, et al. Germline BRCA1 mutations predispose to pancreaticadenocarcinoma. Hum Genet 2008;124:271–8.

29. Skudra S, Staka A, Pukitis A, Sinicka O, Pokrotnieks J, NikitinaM, et al.Association of genetic variants with pancreatic cancer. Cancer GenetCytogenet 2007;179:76–8.

30. Habbe N, Langer P, Sina-Frey M, Bartsch DK. Familial pancreaticcancer syndromes. Endocrinol Metab Clin North Am 2006;35:417–30.

31. Beger C, Ramadani M, Meyer S, Leder G, Kruger M, Welte K,et al. Down-regulation of BRCA1 in chronic pancreatitis andsporadic pancreatic adenocarcinoma. Clin Cancer Res 2004;10:3780–7.

32. Yoshizawa K, Nagai H, Sakurai S, Hironaka M, Morinaga S, Saitoh K,et al. Clonality and K-ras mutation analyses of epithelia in intraductalpapillary mucinous tumor and mucinous cystic tumor of the pancreas.Virchows Arch 2002;441:437–43.

33. Jimenez RE, Warshaw AL, Z'Graggen K, Hartwig W, Taylor DZ,Compton CC, et al. Sequential accumulation of K-ras mutations andp53 overexpression in the progression of pancreatic mucinous cysticneoplasms to malignancy. Ann Surg 1999;230:501–9, discussion509–11.

34. YanagisawaA,KatoY,OhtakeK,KitagawaT,OhashiK,HoriM, et al. c-Ki-ras point mutations in ductectatic-typemucinous cystic neoplasmsof the pancreas. Jpn J Cancer Res 1991;82:1057–60.

35. Shukla V, Coumoul X, Lahusen T, Wang RH, Xu X, Vassilopoulos A,et al. BRCA1 affects global DNA methylation through regulation ofDNMT1. Cell Res 2011;20:1201–15.

36. Weber S, Maass F, Schuemann M, Krause E, Suske G, Bauer UM.PRMT1-mediated arginine methylation of PIAS1 regulates STAT1signaling. Genes Dev 2009;23:118–32.

37. Inamura K, Matsuzaki Y, Uematsu N, Honda A, Tanaka N, Uchida K.Rapid inhibition of MAPK signaling and anti-proliferation effect viaJAK/STAT signaling by interferon-alpha in hepatocellular carcinomacell lines. Biochim Biophys Acta 2005;1745:401–10.

38. Khabar KS, Al-Haj L, Al-Zoghaibi F,MarieM, DhallaM, Polyak SJ, et al.Expressed gene clusters associated with cellular sensitivity and resis-tance towards anti-viral and anti-proliferative actions of interferon. JMol Biol 2004;342:833–46.

39. Feinberg AP, Tycko B. The history of cancer epigenetics. Nat RevCancer 2004;4:143–53.

40. Li E, Bestor TH, Jaenisch R. Targeted mutation of the DNAmethyltransferase gene results in embryonic lethality. Cell 1992;69:915–26.

41. Trinh BN, Long TI, Nickel AE, Shibata D, Laird PW. DNA methyltrans-ferase deficiency modifies cancer susceptibility in mice lacking DNAmismatch repair. Mol Cell Biol 2002;22:2906–17.

42. Belinsky SA, Klinge DM, Stidley CA, Issa JP, Herman JG, March TH,et al. Inhibition of DNAmethylation and histone deacetylation preventsmurine lung cancer. Cancer Res 2003;63:7089–93.

43. Herranz M, Martin-Caballero J, Fraga MF, Ruiz-Cabello J, Flores JM,Desco M, et al. The novel DNA methylation inhibitor zebularine is

DAC Therapy in PDAC


effective against the development of murine T-cell lymphoma. Blood2006;107:1174–7.

44. Yoo CB, Chuang JC, Byun HM, Egger G, Yang AS, Dubeau L, et al.Long-term epigenetic therapy with oral zebularine has minimal sideeffects and prevents intestinal tumors in mice. Cancer Prev Res (Phila)2008;1:233–40.

45. Feldmann G, Habbe N, Dhara S, Bisht S, Alvarez H, Fendrich V, et al.Hedgehog inhibition prolongs survival in a genetically engineeredmouse model of pancreatic cancer. Gut 2008;57:1420–30.

46. Kulaeva OI, Draghici S, Tang L, Kraniak JM, Land SJ, Tainsky MA.Epigenetic silencing of multiple interferon pathway genes after cellularimmortalization. Oncogene 2003;22:4118–27.

Shakya et al.


Documents

Hypomethylating therapy in an aggressive stroma-rich model of pancreatic carcinoma