And Type the TITLE of YOUR WORK in All Caps

INVESTIGATION OF APPROACHES FOR OVERCOMING CHEMORESISTANCE

IN PANCREATIC CANCER

SAU WAI HUNG

(Under the Direction of Rajgopal Govindarajan)

ABSTRACT

Pancreatic cancer continues to be one of the most devastating malignances with a

5-year patient survival rate of less than 6%. One of the main reasons for therapeutic failure in patients is the development of resistance towards chemotherapeutic treatment, in particular, gemcitabine. In order to identify novel approaches for overcoming chemoresistance in pancreatic cancer, we focused on targeting three distinct tumor subpopulations that are highly implicated in conferring cancer drug resistance: 1) innately drug-resistant (DR)-phenotype cells, 2) cancer stem cells (CSCs), and 3) epithelial- mesenchymal transition (EMT)-phenotype cells. Enhanced expression of nucleoside transporters (NTs), necessary for gemcitabine uptake into cells, significantly increased gemcitabine efficacy in DR-phenotype cells. This was achieved directly via gene transfer as well as indirectly through the expression of a cell adhesion protein called E-cadherin.

Expression of E-cadherin increased the expression, activity, and stabilization of hENT1, allowing for greater drug uptake and efficacy. Similarly, expression of hCNT1 with another transmembrane protein, Cx32, was found to increase gemcitabine accumulation within a heterogeneous tumor cell population. With hCNT1 transporting the drug into

cells and Cx32 transferring the drug between cells, the so-called bystander cytotoxic effect could be enhanced. Using a novel epigenetic (i.e., histone methylation) reversal agent, DZNep, we reduced stemness in pancreatic cancer, sensitizing the resistant cancer cell population to gemcitabine without compromising for toxicity in normal pancreatic cells. After drug optimizations, we synthesized novel nanoparticle formulations to deliver the epigenetic-chemotherapeutic combination while mimicking the best dose and schedule. Lastly, we investigated the role of a novel oncoprotein, SET, in conferring

EMT and chemoresistance. SET isoform 2 expression promoted EMT through cadherin switching (i.e., from E-cadherin to N-cadherin) via the Rac1/JNK/c-Jun/AP-1 and

SPARC/Slug pathways. Overexpression of SET and cadherin switching were also observed in human pancreatic ductal adenocarcinoma tissues. In an orthotopic mouse model of human pancreatic cancer, SET isoform 2 expression facilitated metastasis whereas SET knockdown reduced metastatic tumor burden. These investigational approaches have vital therapeutic implications and may lead to the successful development of new therapeutic strategies for enhancing treatment efficacy in patients.

INDEX WORDS: Pancreatic cancer, Chemoresistance, Metastasis, Nucleoside transporter, Nucleoside analog, Gemcitabine, Cadherin, Gap junction, Connexin, Deazaneplanocin (DZNep), SET, Epithelial- mesenchymal transition (EMT), Nanoparticles

INVESTIGATION OF APPROACHES FOR OVERCOMING CHEMORESISTANCE

IN PANCREATIC CANCER

SAU WAI HUNG

BS, University of New Hampshire, 2009

A Dissertation Submitted to the Graduate Faculty of The University of Georgia in Partial

Fulfillment of the Requirements for the Degree

DOCTOR OF PHILOSOPHY

ATHENS, GEORGIA

2014

Sau Wai Hung

INVESTIGATION OF APPROACHES FOR OVERCOMING CHEMORESISTANCE

IN PANCREATIC CANCER

SAU WAI HUNG

Major Professor: Rajgopal Govindarajan

Committee: Huabei Guo Shelley Hooks Mandi Murph Jason Zastre

Electronic Version Approved:

Julie Coffield Interim Dean of the Graduate School The University of Georgia August 2014

DEDICATION

To my family and loved ones.

ACKNOWLEDGEMENTS

I present my utmost gratitude and appreciation to my mentor, Dr. Rajgopal

Govindarajan, for taking me on as his first graduate student, exposing me to the field of cancer biology, and providing financial support. I am sincerely grateful for the opportunity to work in his lab, learn under his expert guidance, and truly grow as an independent scientist. I would like to thank the other members of my graduate committee,

Dr. Shelley Hooks, Dr. Mandi Murph, Dr. Jason Zastre, and Dr. Huabei Guo, for their perpetual support and valuable suggestions throughout my graduate career. I would also like to thank our collaborators, Dr. Shanta Dhar, her graduate student, Sean Marrache,

Dr. Chung K. Chu, Dr. Michael Thomson, and Dr. Tamas Nagy for their expertise and contributions towards shared projects. I would like to acknowledge Dr. Isaiah Fidler, Dr.

Keith Johnson, Dr. Parmender Mehta, Dr. Ming Tsao, and Dr. Chung-Ming Tse for providing me with research materials and assistance. I thank the past and present members of the Govindarajan Laboratory - Hardik Mody, Shannon Cummins, Dylan

Lovin, Bhavi Patel, Dimal Patel, and Franky Davis for their scientific assistance, and especially Dr. Yangzom Bhutia for teaching me all the scientific methods from when I first walked in as a blank slate. I am exceedingly grateful for my undergraduate mentees,

Kimberly Proctor, Maddy Krentz, Haesung Lee, Caitlin Gilbert, Toan Hoang, and Kineta

Naidu, for going above and beyond in their efforts to propel my research projects. Their independence and abilities astounded me, and I am very appreciative for their devotion to my studies. Special thanks go to my family for their unconditional encouragement and

continual support. I am so grateful to the love of my life for being there for me every day, listening to all my tribulations and celebrating all my victories. Lastly, I’d like to acknowledge my financial support from the University of Georgia Department of

Pharmaceutical and Biomedical Sciences, the University of Georgia College of

Pharmacy, the University of Georgia Graduate School, the ARCS Foundation Atlanta

Chapter, the American Foundation for Pharmaceutical Education (AFPE), the American

Association of Pharmaceutical Scientists (AAPS), and the American Society for

Biochemistry and Molecular Biology (ASBMB).

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ...... v

LIST OF TABLES ...... x

LIST OF FIGURES ...... xi

LIST OF ABBREVIATIONS ...... xiv

CHAPTERS

1 INTRODUCTION AND LITERATURE REVIEW ...... 1

Literature Review...... 4

Purpose and Rationale of Study ...... 30

Objective and Hypothesis of Study...... 31

Expected Results and Significance of Study ...... 31

Section and Chapter Summaries ...... 32

Scope of Work and Limitations ...... 35

DRUG RESISTANT-PHENOTYPE CELLS AND CHEMOSENSITIVITY

2 E-CADHERIN INCREASES CHEMOSENSITIVITY THROUGH HUMAN

EQUILIBRATIVE NUCLEOSIDE TRANSPORTER 1 (HENT1) ACTIVITY

AND STABILIZATION ...... 42

Abstract ...... 43

Introduction ...... 44

Materials and Methods ...... 46

vii

Results ...... 50

Discussion ...... 56

3 CO-EXPRESSION OF HCNT1 WITH CX32 INCREASES NUCLEOSIDE

ANALOG TRANSPORT ...... 59

Abstract ...... 60

Introduction ...... 61

Materials and Methods ...... 64

Results ...... 66

Discussion ...... 71

CANCER STEM CELLS AND CHEMOSENSITIVITY

4 PHARMACOLOGICAL REVERSAL OF HISTONE METHYLATION

PRESENSITIZES PANCREATIC CANCER CELLS TO NUCLEOSIDE

DRUGS: IN VITRO OPTIMIZATION AND NOVEL NANOPARTICLE

DELIVERY STUDIES ...... 76

Abstract ...... 77

Introduction ...... 78

Materials and Methods ...... 80

Results ...... 88

Discussion ...... 101

EMT-PHENOTYPE CELLS AND CHEMOSENSITIVITY

5 EXPRESSION OF THE SET ONCOPROTEIN CONTRIBUTES TO THE

EPITHELIAL-MESENCHYMAL TRANSITION (EMT) OF PANCREATIC

CANCER ...... 111

viii

Abstract ...... 112

Introduction ...... 113

Materials and Methods ...... 115

Results ...... 120

Discussion ...... 133

6 DISCUSSION ...... 137

Major Conclusions and Future Directions ...... 137

Therapeutic Challenges in Overcoming Chemoresistance ...... 139

REFERENCES ...... 144

LIST OF TABLES

Page

Table 4.1: Restriction sites and sequences of primers used for cloning ...... 82

Table 4.2: Physiochemical characterization of NPs ...... 100

LIST OF FIGURES

Page

Figure 1.1: The links between drug resistant (DR)-phenotype cells, cancer stem cells

(CSCs), and epithelial-mesenchymal transition (EMT)-phenotype cells ...... 3

Figure 1.2: Novel therapeutic approaches for overcoming limitations in the gemcitabine

pathway ...... 17

Figure 2.1: E-cadherin expression increases gemcitabine sensitivity in pancreatic cancer

cells ...... 50

Figure 2.2: MIA PaCa-2/E-cad showed an increase in hENT1 expression, transport, and –

dependent gemcitabine cytotoxicity ...... 52

Figure 2.3: Expression of E-cadherin increased hENT1 expression, function, and

stabilization in a clean cadherin model cell line ...... 54

Figure 3.1: Stable expression of NTs and Cxs in MDCK cells increased total 3H-

thymidine cellular transport and gemcitabine cytotoxicity ...... 67

Figure 3.2: Stable expression of a combination of NTs and Cxs in MDCK cells altered

total 3H-thymidine cellular transport ...... 68

Figure 3.3: Statuses of each hENT1, hCNT1, Cx32, and Cx43 were identified in a panel

of pancreatic cell lines ...... 69

Figure 3.4: Proposed model of intracellular and intercellular nucleoside transport and

movement within a cell population ...... 70

Figure 4.1: DZNep and gemcitabine sensitivity, singly or in combination, and interactions

within a panel of pancreatic cell lines ...... 90

Figure 4.2: DZNep partially competes with the uptake of purine nucleosides by hENT1

and hCNT3 ...... 92

Figure 4.3: Acyl modifications of DZNep further enhance cytotoxicity ...... 94

Figure 4.4: DZNep alters histone lysine methylation and methyltransferase and

demethylase expressions in pancreatic cancer ...... 96

Figure 4.5: Short priming of DZNep demonstrated superior cytotoxicity and synergy with

gemcitabine than co-exposure of the two drugs ...... 98

Figure 4.6: Spatiotemporal release of DZNep and gemcitabine using engineered

nanoparticles reduced drug dose while potentiated chemosensitivity ...... 102

Figure 5.1: SET isoform 2 is highly overexpressed in pancreatic cancer cells ...... 122

Figure 5.2: SET isoform 2 is localized in the nucleus and at the cell surface in poorly-

differentiated pancreatic cancer cells ...... 124

Figure 5.3: SET isoform 2 induces EMT and promotes growth, migration, and invasion of

pancreatic cancer cells ...... 126

Figure 5.4: SET isoform 2 promotes cadherin switching from E-cadherin to N-cadherin to

undergo EMT ...... 127

Figure 5.5: SET-induced cadherin switching involves the Rac1/JNK/c-Jun/AP-1 and

SPARC/Slug signaling pathways ...... 129

Figure 5.6: SET isoform 2 overexpression at the cell surface and cadherin switching were

identified in patient-derived, poorly-differentiated pancreatic cancer tissues .....131

xii

Figure 5.7: SET isoform 2 expression increases tumor volume ...... 133

xiii

LIST OF ABBREVIATIONS

5-Aza-C 5-Azacytidine

5-Aza-dC 5-Aza-2’-deoxycytidine

5-FU 5-Fluorouracil

ABC ATP-binding cassette

ALDH Aldehyde dehydrogenase

ANOVA Analysis of variance

AP-1 Activator protein 1

ATP Adenosine triphosphate

BCA Bicinchoninic acid

BE Bystander effect

BMP Bone morphogenetic protein

CD Cluster of differentiation

CDA Cytidine deaminase

CDK6 Cyclin-dependent kinase 6

CDP Cytidine diphosphate

CDS Coding sequence

COL Collagen

CSC Cancer stem cell

Cx Connexin

xiv

CXCR4 CXC chemokine receptor type 4

DAPI 4’6-Diamidino-2-phenylindole dCDP Deoxycytidine diphosphate dCK Deoxycytidine kinase dCTP Deoxycytidine triphosphate dFdC Gemcitabine / 2’,2’-Difluorodeoxycytidine dFdCDP Gemcitabine diphosphate dFdCMP Gemcitabine monophosphate dFdCTP Gemcitabine triphosphate dFdU 2’,2’-Difluorodeoxyuridine

DLS Dynamic light scattering

DM Dimethylation

DMEM Dulbecco’s modified Eagle medium

DMSO Dimethyl sulfoxide

DNA Deoxyribonucleic acid dNTP Deoxynucleotide triphosphate

DR Drug resistant

DSPE 1,2-Distearoyl-sn-glycero-3-phosphoethanolamine

DTT Dithiothreitol

DZNep 3-Deazaneplanocin A

E-cad / E-cadherin Epithelial cadherin

EED Embryonic ectoderm development

EGFR Epidermal growth factor receptor

EGTA Ethylene glycol tetraacetic acid

EMT Epithelial-mesenchymal transition

EPR Enhanced permeability and retention

ERK Extracellular signal-regulated kinase

ESA Epithelial-specific antigen

EZH2 Histone-lysine N-methyltransferase

FBS Fetal bovine serum

FDA Food and Drug Administration

FDR Fixed dose rate

FEM Field emission microscopy

GCV Ganciclovir

GJ Gap junction

GPC Gel permeation chromatography

H3K27 Histone H3 lysine 27

H3K4 Histone H3 lysine 4

H3K9 Histone H3 lysine 9

H4K20 Histone H4 lysine 20

HA Hemagglutinin

HAT Histone acetyltransferase

HBSS Hank’s balanced salt solution hCNT Human concentrative nucleoside transporter

HDAC Histone deacetylase hENT Human equilibrative nucleoside transporter

xvi

HER2 Human epidermal growth factor receptor 2

HPDE Human pancreatic ductal epithelial

HPLC High performance liquid chromatography

HSV Herpes simplex virus

HSV-TK Herpes simplex virus thymidine kinase

HuR Human antigen R

I2PP2A Inhibitor of protein phosphatase 2A

IACUC Institutional Animal Care and Use Committee

IMAGE Integrated molecular analysis of genomes and their

expression

INHAT Inhibitor of histone acetyltransferase

JMJD Jumonji domain

JNK c-Jun N-terminal kinase

KPC KrasLSL-G12D/+;Trp53LSL-R172H/+;Cre

KRAS Kirsten rat sarcoma viral oncogene homolog

MAPK Mitogen-activated protein kinase

MDCK Madin-Darby canine kidney

MEK Mitogen-activated protein kinase kinase

MEM Minimum essential medium miRNA Micro-ribonucleic acid

MM Monomethylation

MMP Matrix metalloproteinase mRNA Messenger ribonucleic acid

xvii

MRP Multidrug resistance-associated protein

MTD Maximum tolerated dose mtDNA Mitochondrial deoxyribonucleic acid

MTT 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium

bromide

MW Molecular weight

Na2VO4 Sodium orthovanadate

NBMPR Nitrobenzylthioinosine

N-cad / N-cadherin Neural cadherin

NDRI National disease research interchange

NEM N-Ethylmaleimide

NFκB Nuclear factor kappa B

NMR Nuclear magnetic resonance

NP Nanoparticle

NT Nucleoside transporter

PBS Phosphate buffered saline

PcG Polycomb group

PCR Polymerase chain reaction

PDAC Pancreatic ductal adenocarcinoma

PDI Polydispersity index

PEG Polyethylene glycol

PEI Polyethylenimine

PET Polyethylene terephthalate

xviii

PI3K Phosphoinositide 3-kinase

PLGA Poly(lactic-co-glycolic acid)

PMSF Phenylmethylsulfonyl fluoride

PP2A Protein phosphatase 2A

PPAR-γ Peroxisome proliferator-activated receptor gamma

PRC2 Polycomb repressive complex 2

PTEN Phosphatase and tensin homolog

PVA Polyvinyl alcohol

Rac1 Ras-related C3 botulinum toxin substrate 1

Ras Rat sarcoma

RNA Ribonucleic acid

RPM Revolutions per minute

RR Ribonucleotide reductase

SDS Sodium dodecyl sulfate

SDS-PAGE Sodium dodecyl sulfate polyacrylamide gel electrophoresis

SHH Sonic hedgehog shRNA Short hairpin ribonucleic acid siRNA Small interfering ribonucleic acid

SLC Solute carrier

SPARC Secreted protein acidic and rich in cysteine

STAT3 Signal transducer and activator of transcription 3

SUZ12 Suppressor of zeste 12 homolog

TAF-I Template activating factor I

xix

TBAF Tetra-n-butylammonium fluoride

TBS Tris buffered saline

TEM Transmission electron microscopy

TGF Transforming growth factor

THU Tetrahydrouridine

TM Trimethylation

TPP Triphenylphosphonium

TS Thymidylate synthase

TSA Trichostatin A

UMK Uridine monophosphate kinase

ZEB Zinc finger E-box binding

CHAPTER 1

INTRODUCTION AND LITERATURE REVIEW

Cancer is the second most common cause of death in the United States [1], and the current risk of an American developing this condition is 44% and 38% in men and women, respectively [2]. Hence, approximately 1.7 million Americans are expected to be diagnosed this year with about 585,720 expected to die – almost 1,600 people per day [1].

In particular, pancreatic cancer continues to be one of the most devastating malignances with a 5-year patient survival rate of less than 6% [1]. Two main reasons for therapeutic failure in these patients are: (1) late diagnosis of the disease, when the cancer has already spread to distant sites, i.e., metastasized, and (2) the lack of effective treatment, including difficulties with drug delivery and the development of resistance towards chemotherapeutics. While chemotherapeutic drugs show initial efficacy by tumor shrinkage, their consistent use and exposure frequently causes the cancerous cells to adapt by changing their characteristics and becoming increasingly resistant to the drugs’ cytotoxic effects, otherwise known as chemotherapeutic refractoriness. This leads to the evolution of fully-resistant cancer cell colonies resulting in total chemotherapeutic failure, new cancer remissions, and deadly consequences.

Although new strategies against pancreatic cancer have recently emerged, there remains a heavy dependence on the use of gemcitabine (2’,2’-difluorodeoxycytidine, dFdC) in patients. Gemcitabine is often used as part of the first-line therapy, and it

remains the standard of care for adjuvant therapy [3, 4]. Nonetheless, the innate and acquired chemoresistance of patients to the drug has emerged as a major concern as observed by a very low patient response rate [5]. Innumerable clinical trials have ensued over the past decade investigating the efficacy of gemcitabine in chemotherapeutic combinations [6]. Unfortunately, this strategy has proven largely unsuccessful, leading to further evaluation of the molecular aspects of the disease biology as well as approaches for overcoming chemoresistance. Ultimately, a better understanding of cellular chemoresistance could lead to the successful development of new therapeutic strategies for enhancing treatment efficacy in patients.

The chemoresistance of gemcitabine in pancreatic cancer can be attributed to several anatomical (e.g., dense desmoplastic stroma), pathophysiological (e.g., amplification of growth, cell survival, and anti-apoptotic pathways), and pharmacological

(e.g., requirement of phosphorylation for drug activation) barriers. In order to identify novel approaches for overcoming chemoresistance in pancreatic cancer, this dissertation focuses on targeting three distinct tumor subpopulations that are highly implicated in conferring cancer drug resistance: 1) innately drug-resistant (DR)-phenotype cells, 2) cancer stem cells (CSCs), and 3) epithelial-mesenchymal transition (EMT)-phenotype cells (Fig. 1.1) [7, 8]. DR-phenotype cells are inherently chemoresistant due to cellular alterations limiting the transport, activation, and overall efficacy of chemotherapeutic agents. Among the various intrinsic and extrinsic factors implicated in determining drug efficacy, the reduction or complete loss of nucleoside drug transporters is thought to be critical for gemcitabine resistance. Aberrant expression of nucleoside transporters has been noted in pancreatic ductal adenocarcinomas as well as advanced-stage, metastatic

tumors, and it is now clear that the extent of cancer cell transportability often decides the preliminary response to nucleoside analog chemotherapy [9-14]. Pluripotent CSCs have the ability to renew into any cancer cell subtype and essentially repopulate the entire tumor upon initial eradication of the bulk of the cancerous cells by neoadjuvant chemotherapy [15-17]. This allows for subsequent clonal selection and development of chemoresistant cell populations. Indicative of advanced, metastatic cancer, EMT- phenotype cells have acquired morphological and structural changes that allow them to invade through the basement membrane and spread to distant parts of the body. Large scale tumor genomic studies have identified that these metastasized cells carry several genetic and epigenetic changes that occur during carcinogenesis as well as subclonal evolution [18-20].

Figure 1.1. The links between drug resistant (DR)-phenotype cells, cancer stem cells (CSCs), and epithelial-mesenchymal transition (EMT)-phenotype cells. These three cancer subpopulations are known to confer chemoresistance in tumors. They share many similar features and demonstrate plasticity for interconversions [21]. Figure modified from Wang, et al. [7].

Literature Review 1

For the purposes of this dissertation, the following Literature Review comprises of excerpts from the review paper, “Overcoming Nucleoside Analog Chemoresistance of

Pancreatic Cancer: A Therapeutic Challenge” published in Cancer Letters as well as additional background information not found in the review. Supplemental information was necessary to comprehensively introduce all the material presented in this dissertation.

Chemoresistant Pancreatic Cancer Cell Populations:

Since chemotherapy is often the first-line treatment strategy for pancreatic cancer patients, drug resistance is a major obstacle that needs further understanding. Drug resistance can be classified into two categories: intrinsic (or innate) and acquired. In patients with intrinsic drug resistance, chemotherapy is ineffective from the start of treatment because the cancer cells already exhibit a drug-resistant phenotype, preventing drugs from being transported, activated, or in general, effective. Acquired resistance develops only after prolonged exposure of anticancer drugs to the tumor cells. This continued treatment puts selective pressure on the tumor cells to survive and adapt, leading to recurrence and metastasis. The use of targeted therapies attempts to overcome these resistance mechanisms by specifically targeting an oncoprotein or aberrant signaling pathway contributing to drug resistance [8]. Throughout the past decade, studies have increasingly shown the critical roles of two pancreatic cancer cell populations, CSCs and EMT-phenotype cells, in conferring drug resistance.

1 SW Hung, H Mody, and R Govindarajan. 2012. Overcoming Nucleoside Analog Chemoresistance of Pancreatic Cancer: A Therapeutic Challenge. Cancer Letters. 320: 138-149. Reprinted here with permission of the publisher.

Cancer Stem Cells (CSCs) and Chemoresistance

Most current chemotherapies target and kill differentiated cancer cells, while neglecting a small but influential population of cells known as CSCs. CSCs have the ability to self-renew and produce differentiated cells to repopulate a tumor with invasive and metastatic properties. In other words, they are able to continually sustain tumorigenesis [16]. These stem cells have been identified and isolated from solid pancreatic tumors, with high expression of defining cell surface markers CD44, CD24, epithelial-specific antigen (ESA), CD133, and CXCR4 [7, 22, 23]. Elevated levels of aldehyde dehydrogenase (ALDH), another key indicator of CSCs, were also found [7,

22]. Human pancreatic CSCs have been shown to be exclusively tumorigenic and highly resistant to standard chemotherapy [24]. With greater tumorigenicity as well as metastatic potential, few current therapies can eliminate these cells [7]. These stem cells have also been shown to be highly gemcitabine-resistant and share many features with gemcitabine-resistant cancer cells such as high sphere-forming activity [7]. However, the mechanisms behind this connection remain unclear.

Epithelial-mesenchymal Transition (EMT)-phenotype Cells and Chemoresistance

During EMT, cells lose their epithelial cell-cell junctions and polarity and instead acquire the expressions of mesenchymal markers such as vimentin, fibronectin, and N- cadherin. They also have increased matrix metalloproteinase (MMP) activity which confers an invasive phenotype [7]. Several other gene families are also implicated in the

EMT process: BMP, COL, NOTCH, NODAL, SNAI, TGF, WNT, and ZEB, among others. Numerous studies have identified the close relationship between EMT and drug resistance in pancreatic cancer cells. Many of the pancreatic cancer cell lines that have

high expression of the epithelial marker, E-cadherin, and low expression of the mesenchymal marker, ZEB1, are sensitive to common chemotherapeutic agents [7]. In contrast, cell lines with opposite expression levels are chemoresistant [7]. Similarly, cell lines with acquired resistance to gemcitabine by continuous exposure exhibited increased expression of mesenchymal Snail and Twist [23]. Furthermore, downregulation of the

Notch signaling pathway was found to revert the EMT phenotype by decreasing the expressions of mesenchymal markers vimentin, ZEB1, Snail (SNAI1), Slug (SNAI2), and

NFκB, indicating a mechanistic tie between EMT and chemoresistance in pancreatic cancer cells [7]. Therefore, it is likely that EMT can lead to multidrug resistance as well as the rapid progression of the tumor [23].

Links between CSCs and EMT

EMT-phenotype cells share many key molecular characteristics with CSCs, and

CSCs exhibit a mesenchymal phenotype, suggesting the two drug-resistant populations are closely related [7]. For example, pancreatic CSCs have high tumorigenic potential and exhibit morphological features similar to that of EMT [7]. Treatment with transforming growth factor (TGF)-β can even induce pancreatic CSCs to undergo EMT

[7]. On the other hand, it has been shown that gemcitabine-resistant EMT-phenotype cells as well as metastatic cancer cells also exhibit CSC features and markers [7, 8]. For example, Mani et al. demonstrated that human mammary epithelial cells undergoing

EMT acquired both a fibroblastic mesenchymal phenotype as well as CSC characteristics including CD44high/CD24low, self-renewal, and the ability to form mammospheres [25].

The discovery of metastatic CSCs truly illustrates the close relationship between

CSCs and EMT. In 2007, Hermann et al. identified a distinct population of

CD133+/CXCR4+ cells that localized at the invasive edge of pancreatic carcinomas and exhibited strong migratory activity in vitro as well as metastatic activity in vivo [24]. It has also been proposed that under hypoxic conditions, cancer cells can both undergo

EMT as well as acquire CSC properties, suggesting a crossover between the roles of

EMT and CSCs in determining the properties of chemoresistance, invasiveness, metastasis, and recurrence [23].

Strategies for Overcoming Chemoresistance in These Populations

Selective, targeted elimination of these two cell populations could improve patient response to currently used drugs [7]. Novel inhibitors of EMT or even compounds that could revert the EMT phenotype could increase the drug sensitivity of pancreatic cancer cells [7]. This could be accomplished by targeting epigenetic regulators, specific non- coding RNAs, or signaling pathways that control this process. For example, several EMT transcription factors are controlled at the epigenetic level. E-cadherin (CDH1) is epigenetically regulated by Snail, Slug, zinc finger proteins (ZEB1 and ZEB2), and Twist

[18, 23]. SOX4 expression transcriptionally activates EZH2 which then trimethylates specific genes to promote EMT [18]. A recent burst of studies have shown microRNAs

(miRNAs) to be highly involved in the EMT process as well as in conferring gemcitabine resistance in pancreatic cancer. Among the most notable include the miR-200 family, miR-21, miR-221, and miR-126 [7, 26]. Furthermore, key signaling pathways known to be highly involved in EMT in pancreatic cancer (e.g., TGF-β, Ras/Raf/MEK/ERK, the

Wnt cascade, PI3K/Akt, Notch, and sonic hedgehog (SHH)) can also be targeted to alter the EMT process and induce chemosensitivity [23, 26].

The tumor microenvironment is also partly responsible for inducing and/or maintaining EMT. Conditions in the tumor microenvironment, such as hypoxia, may exert both EMT- and CSC-promoting effects [23, 26]. Other components, including the extracellular matrix, cancer-associated fibroblasts, immune cells, and soluble factors, may also be targeted to indirectly impair the EMT-phenotype and CSC populations and overcome chemoresistance [23].

Direct elimination of EMT-phenotype cells and CSCs from the tumor population is another viable approach for overcoming drug resistance, metastasis, and tumor recurrence. However, it may be difficult to effectively target CSCs without annihilating normal somatic stem cells as well. Likewise, molecular differences between EMT in embryological development and cancer progression will need to be better defined for specific targeting. These strategies will most likely require the use of combination therapy, with separate agents targeting the bulk tumor population, EMT-phenotype population, and the CSC population. This strategy will need to utilize the molecular differences between these populations of cells. Targeting specific transporters, cell surface markers, aberrant signaling pathways, or epigenetic processes may be useful [8].

Determinants of Chemoresistance in the Gemcitabine Pathway:

Transporters and Metabolic Enzymes

Due to its hydrophilicity (log P of -1.33) [27], gemcitabine relies on numerous transporters to pass through the cellular lipid bilayer and exert its cytotoxicity. Earlier studies by Mackey et al. identified key transporters involved in the uptake of gemcitabine and demonstrated the need for their activity in order to confer gemcitabine sensitivity [28,

29]. While they found that the human equilibrative transporters 1 and 2 (hENT1 and

hENT2) were able to mediate gemcitabine transport (Km of 160 and 740, respectively), the human concentrative nucleoside transporter 1 (hCNT1) had the greatest intrinsic transport activity (Vmax:Km of 0.24) [29]. Since then, both hENT1 and hCNT1 have been highly implicated in gemcitabine chemoresistance as well as patient outcome. Although the clinical correlation between hENT1 expression, both transcriptionally and immunohistochemically, and disease-free as well as overall survival of pancreatic cancer patients has been well-established [10-14], the correlation between hENT1 expression and innate and acquired chemoresistance in cultured pancreatic cancer cells remains somewhat incongruous and context-dependent [28, 30-33]. This may be due to the bidirectional nature of the transporter resulting in drug efflux at higher concentrations, the relative role of the transporter among the presence of other nucleoside transporters, and the expression characteristics of the transporter in normal versus tumor cells [31] including variations in cell cycle characteristics [9]. In pancreatic cells, mislocalization of hENT2, rather than changes in expression levels, proposes an alternative mechanism for gemcitabine chemoresistance [34]. Our lab has demonstrated that the loss of hCNT1 cell surface expression and activity frequently observed in pancreatic cancer cells was found to correlate directly with high gemcitabine chemoresistance [9, 31]. Structure-activity characterization of hCNT3 revealed electrostatic interaction by the 3’-hydroxyl position to play a major role in the transport of substrates, including gemcitabine [35]. While the exact role of hCNT3 in gemcitabine chemoresistance has yet to be evaluated, patients with elevated immunocytochemical expression of the transporter were observed to experience a lower risk of disease recurrence and longer overall survival [12].

In addition to influx transporters, a few efflux transporters have also been implicated in gemcitabine resistance of pancreatic cancer including the ATP-binding cassette (ABC) family of multidrug resistance-associated proteins (MRPs). Increased expression of MRP2 mRNA and protein in pancreatic cancer tissues has been associated with both intrinsic and acquired resistance to a regimen of gemcitabine plus cisplatin

[36]. Likewise, MRP7, expressed in both normal pancreas [37] and most established pancreatic carcinoma cell lines [38], has been demonstrated to be able to transport gemcitabine, although direct correlations with cytotoxicity have not yet been demonstrated [39]. This may be in part due to the ability of MRPs to efflux the active metabolite of gemcitabine, allowing for reduction of drug concentrations inside the cell

[40]. Although the overall significance of transporters in determining gemcitabine chemosensitivity has been well demonstrated, the precise mechanisms and alterations that occur in pancreatic tumors resulting in chemoresistance remain unclear.

Once inside the cell, gemcitabine is monophosphorylated by deoxycytidine kinase

(dCK) before further phosphorylation into its active diphosphate (dFdC-DP) and triphosphate (dFdC-TP) forms. Activated gemcitabine, in the form of dFdC-TP, induces masked chain termination during which the nucleotide is incorporated into DNA, followed by another deoxynucleotide, prohibiting further DNA polymerase action [41].

Since involved in the rate-limiting step for gemcitabine cellular activation, a deficiency in dCK expression and activity has been shown to be associated with gemcitabine resistance in pancreatic cancer. Loss of dCK mRNA in highly-resistant pancreatic cancer cells has been observed [30, 42, 43], and a clear correlation between dCK activity and gemcitabine sensitivity in murine and human pancreatic tumor xenografts has been demonstrated [44].

Overexpression of dCK was found to significantly enhance gemcitabine sensitivity in two of three pancreatic cancer cell lines [45]. Furthermore, simultaneous expression of dCK and p8 in a gemcitabine-resistant pancreatic cancer cell line, PANC-1, significantly decreased the cytotoxic IC50 of gemcitabine and enhanced apoptosis and caspase-3 activity; tumor growth inhibition was also noticeably improved in nude mice [46].

Clinically, low immunohistochemical expression of dCK was correlated with both decreased overall survival as well as older age of patients, suggesting a role of age- related methylation in patients [47]. Low expression of the RNA-binding protein HuR has also been shown to correlate with a 7-fold increase in the risk of mortality for pancreatic cancer patients due to its ability to regulate dCK protein levels and confer gemcitabine chemoresistance [48, 49].

Gemcitabine is rendered inactive by cytidine deaminase (CDA), which removes the NH2 group from the pyrimidine [50], allowing the uracil metabolite to be exported from the cell. Interestingly, gemcitabine-induced inhibition of cytidine deaminase activity leads to a decrease in dFdC-TP catabolism, hence propelling the self-potentiation of gemcitabine activity [51]. Although the focus of most studies involving CDA and gemcitabine in pancreatic cancer has been on generalized adverse effects (i.e., high-grade neutropenia) due to genetic polymorphisms rather than cancer cell cytotoxicity [52-56], a few studies have identified a dramatic increase in chemosensitivity, up to 54-fold, of cell lines with the inhibition of CDA [45, 57]. Although further studies are necessary, alterations in CDA expression levels in tumors could be a promising mechanism for improving gemcitabine sensitivity.

Other Molecular Targets

Additional candidates of interest for the manipulation of gemcitabine cytotoxicity include effectors of DNA synthesis and repair. In particular, ribonucleotide reductase subunits 1 and 2 (RRM1 and RRM2) are inhibited by dFdC-DP and dFdC-TP and are hindered from repairing flawed DNA. In addition to impeding DNA repair, inhibition of

RR, the rate-limiting enzyme in deoxyribonucleoside triphosphate (dNTP) synthesis, reduces the endogenous dNTP pool, lessening competition and indirectly facilitating dFdC-TP incorporation into DNA [58]. This secondary mechanism also contributes to the unique self-potentiating ability of gemcitabine. The transcriptional upregulation of the larger subunit, RRM1, has been consistently observed as pancreatic cancer cell lines acquire gemcitabine resistance [30, 59]. Consistently, low expression of RRM1 in tumors is correlated with enhanced response to gemcitabine specifically in recurrent cases [59,

60]. The subunit appears to have no correlation with disease-free or progression-free survival, and its correlation with overall survival is variable [60, 61]. Therefore, the expression of RRM1 seems to be more relevant to acquired rather than innate gemcitabine resistance in both pancreatic cancer cell lines and patients.

For RRM2, some studies have shown increased transcript and protein expressions in pancreatic cancer cell lines [30, 62], while others have additionally identified the significance of the alteration in determining gemcitabine sensitivity in pancreatic cancer.

Specifically, RNA interference of RRM2 attenuated chemoresistance and invasiveness in cells [42, 63], while it suppressed tumor growth, enhanced apoptosis, and inhibited metastasis in xenograft models [64]. Moreover, Ohhashi et al. found a reduction in cellular proliferation with the inhibition of RRM1 and RRM2 even in the absence of

gemcitabine treatment [42]. Clinically, low RRM2 mRNA expression levels correlated with significantly enhanced disease-free, median, and overall survival as well as overall response rate in gemcitabine-treated patients [65, 66]. Reduction of the dNTP pool by gemcitabine inhibition of RRM2 is clear in its effects on nuclear DNA (i.e., facilitating the incorporation of dFdC-TP into replicating DNA). However, this mechanism may also hold true for mitochondrial DNA (mtDNA) although evidence has not yet been provided.

Nevertheless, gemcitabine has been shown to directly affect mtDNA by inhibiting its γ- polymerase [67].

Approaches to Enhancing Gemcitabine Delivery:

Prodrugs by Chemical Modification

An apparent approach for enhancing gemcitabine delivery to cells is to bypass its dependence on transporters for entering the cell. This can be achieved by chemically modifying the drug and creating various prodrugs, such as acyl derivatives, with increased lipophilicity. For example, lipophilic prodrugs of troxacitabine, a nucleoside analog drug with an unnatural L-configuration, were created with the addition of linear aliphatic chains to the amino group; sensitivity of pancreatic cancer cells to the modified drugs was greater than 100-fold compared with troxacitabine [68]. Likewise, lipophilic prodrugs of gemcitabine were synthesized by linking the 4-amino group with acyl derivatives (i.e., valeroyl, heptanoyl, lauroyl, and stearoyl) [69, 70]. In addition, by masking the N-terminus, this modification also protects gemcitabine from rapid deamination and improves its half-life. In particular, 4-(N)-stearoyl-gemcitabine was found to decrease the cytotoxic IC50 of a cervix, breast, colorectal, and nasopharyngeal cell line by 1.5-5-fold [71, 72]. Furthermore, a novel gemcitabine-cardiolipin conjugate

was shown to induce cytotoxicity in several gemcitabine-resistant cell lines independent of nucleoside transporter activity. In vivo, the conjugate demonstrated less adverse effects compared with gemcitabine as measured by body weight and white blood cell count, while inhibiting tumor growth by 20% greater than gemcitabine. A high dose of the treatment increased the median survival of tumor-bearing mice by 73%, while the same dose and schedule of gemcitabine led to toxic death of all mice [73]. Another lipophilic prodrug of gemcitabine, CP-4126, was created by esterifying an elaidic fatty acid at the

5’ position. The compound was found to be as effective as gemcitabine in chemoresistant cell lines, but both were ineffective in cells devoid of dCK activity. However, CP-4126 maintained its efficacy in nucleoside transporter-inhibited cells, while the IC50 for gemcitabine increased 200-fold. In various xenograft models, both gemcitabine and CP-

4126 were equally effective, but the prodrug was able to be administered orally [74]. The potential of this approach has extended to a phase I study with a novel oral gemcitabine prodrug, LY2334737, evaluated in patients with advanced or metastatic solid tumors.

This prodrug protects the amine group of gemcitabine with a covalent bond to valproic acid, preventing extensive pre-systemic deamination. As the first human clinical trial for

LY2334737, the study demonstrated safe administration of the drug, despite the caveat of blood-brain barrier effects, with minor adverse effects and observations of anti-tumor activity [75].

An alternative approach for enhancing efficacy includes the creation of a phosphoramidate prodrug. By creating a variant of the monophosphate form (i.e., dFdC-

MP), the rate-limiting phosphorylation step is bypassed. Wu et al. demonstrated that a phosphoramidate prodrug of gemcitabine was approximately 4-fold more effective than

gemcitabine in dCK-deficient variants of cancerous cell lines. Furthermore, inhibition of transporter activity did not diminish the prodrug’s activity in the dCK variants, although the same did not hold true for the parental cell lines [76]. Similarly, another phosphoramidate prodrug, GemMP[10], was found to reduce thyroid cancer cell proliferation by arresting cells in S phase at concentrations 5-10-fold lower than gemcitabine [77]. In addition to increasing lipophilicity, reducing deamination, and bypassing the rate-limiting phosphorylation step, chemical modifications of gemcitabine can also be conducted in order to enhance its interaction with delivery vesicles such as liposomes [69, 70], which will be further discussed below.

Nanoparticle Drug Delivery

Nanoparticle drug delivery has been shown to be a promising approach for overcoming several anatomical, pathophysiological, and pharmacological barriers [78] and attenuating chemoresistance in pancreatic cancer. By encapsulating or adsorbing gemcitabine in nanoparticles, reduced pre-systemic metabolism, lower dosages, and sustained release are possible. For example, encapsulation of gemcitabine into chitosan and albumin nanoparticles produced sustained release profiles as well as improved antitumor activity in vitro compared with the administration of free drug [79]. Additional therapeutic advantages of nanoparticle drug delivery include altered pharmacokinetic parameters (e.g., decreased drug clearance), increased drug concentration in tumor tissues via the enhanced permeability and retention (EPR) effect (i.e., nanoparticles tend to greater accumulate in tumor rather than normal tissues) [80], increased local plasma T3 levels, and the potential for targeted delivery.

Furthermore, combination therapies can also be utilized with nanotechnology.

One study reported that a combination of gemcitabine and curcumin in nanoparticles enhanced the inhibition of tumor growth, abolished systemic metastases, and reduced the activation of NFκB in a pancreatic cancer xenograft model as compared with either agent alone [81]. Similarly, gemcitabine and paclitaxel preconjugated with a hydrolysable linker and subsequently loaded into a drug carrier were found to significantly improve the chemotherapeutic activity as compared with their free form [82].

As previously noted, decreasing the hydrophilicity of gemcitabine aids in the incorporation of the drug into liposomal formulations. In particular, stearoyl gemcitabine nanoparticles better managed tumor growth as compared with free gemcitabine, and results were further enhanced by PEGylation which increases drug half-life [83]. Such nanoparticles were also shown to overcome gemcitabine resistance related to the overexpression of RRM1 in both cell culture and mice. Furthermore, the enhanced cytotoxicity of these nanoparticles was additionally observed in dCK-deficient tumor cells with the induction of apoptosis through caspase activation [84]. Similarly, conjugation of gemcitabine with squalene, a precursor to cholesterol synthesis, resulted in amphiphilic molecules that were shown to have enhanced anticancer activity due to protection of the drug from rapid deamination [85]. Such nanoparticles were demonstrated to possess greater ability to induce S phase arrest and apoptosis of cancer cells as compared with free gemcitabine [86]. Furthermore, it has been reported that albumin enhanced the passive diffusion of squalenoyl gemcitabine nanoparticles rather than relying on membrane transporters [87]. This was confirmed by higher activity of the nanoparticles than free gemcitabine in transporter-deficient, gemcitabine-resistant tumor

Figure 1.2. Novel therapeutic approaches for overcoming limitations in the gemcitabine pathway. Gemcitabine enters the cell via nucleoside transporters (i.e., CNT1, CNT3, ENT1, ENT2). The drug is either phosphorylated into its active form (i.e., dFdC-phosphate) by dCK or deaminated into dFdU by CDA and eliminated from the cell. Activated gemcitabine can then terminate the cell by directly targeting DNA or inhibiting RRM1 and RRM2 to deplete the dNTP pool necessary for DNA replication. Descriptions in boxes (red) indicate known methods for targeting a particular determinant (green) of gemcitabine chemosensitivity in pancreatic cancer. Abbreviations: CNT, concentrative nucleoside transporter; ENT, equilibrative nucleoside transporter; dCK, deoxycytidine kinase; CDA, cytidine deaminase; RR, ribonucleotide reductase; dFdC, 2’,2’-difluorodeoxycytidine; dFdU, 2’,2’- difluorodeoxyuridine; CDP, cytidine diphosphate; dCDP, deoxycytidine diphosphate; dCTP, deoxycytidine triphosphate.

cells [88]. Furthermore, the penetration of the gemcitabine-squalene molecules was increased due to the presence of cholesterol in the monolayer, which may contribute to the increased anticancer activity since cancer cell membranes are rich in cholesterol [89].

Squalenoyl-based gemcitabine nanoparticles were found to overcome gemcitabine

resistance with increased cytotoxicity in gemcitabine-resistant cell lines deficient in both dCK and hENT1 [86].

Nanoparticles can also be modified for targeted delivery of gemcitabine to pancreatic cancer cells. The addition of monoclonal antibodies has been utilized as a targeting moiety for pancreatic tumors. A recent study demonstrated that Herceptin

(HER2)-conjugated chitosan nanoparticles loaded with gemcitabine led to an increase in antiproliferative activity in vitro along with enhanced S-phase arrest as compared with free gemcitabine. In addition, the targeted nanoparticles were efficiently taken up by the cells and prolonged intracellular retention was obtained. Sustained in vitro release of the drug from the nanoparticulate system indicated proper diffusion of the drug from the polymeric matrix [90]. Likewise, gold nanoparticles with anti-epidermal growth receptor antibodies demonstrated that a lower dose of gemcitabine was able to inhibit the proliferation of pancreatic cancer cells as well as orthotopic tumor growth [91].

An additional method for targeting includes the use of magnets which are directed to the desired site upon application of an external magnetic field. The drug can then be released with the help of various triggering factors like ultrasound or changes in physiological conditions (e.g., pH, temperature) or by simple diffusion [92, 93]. Such systems prepared by the inclusion of magnetic nanocrystals into squalenoyl gemcitabine bioconjugates have been reported. Upon injection into a mouse tumor model, magnetically-guided nanoparticles showed enhanced anticancer activity [94].

Nanogel formulations with tumor-specific molecules were also efficient in overcoming resistance by enhancing tumor growth inhibition with stable release of the drug over several days in nucleoside transporter-deficient and dCK-deficient

lymphogenic cancer cells [95]. Such nanogel formulations containing the active form of the nucleoside analog (i.e., dFdC-TP) demonstrated greater cytotoxicity and reduced resistance as compared with nucleoside analog prodrugs. The drug delivery systems aid in the protection of the active drugs and enhance intracellular retention [96, 97].

Dosage and Schedule Modifications

The mechanism of action of gemcitabine is atypical since the drug depends on the cell cycle and only a few key proteins for efficacy. The importance of these key players for gemcitabine is unique and evident by the significant influence on active metabolite concentration and drug efficacy with the modification of only a single target. Therefore, the efficacy of gemcitabine is dependent on a balance of both dosage and schedule.

Gemcitabine pharmacokinetics can be described by a linear, 2-compartment model with a half-life of 42-94 min (short infusion of <70 min) depending on age and gender. While the rate of clearance varies with age and gender, 92-98% of a typical gemcitabine dose (i.e., 1000 mg/m2/30 min infusion) is excreted, predominantly in the urine. The volume of distribution of gemcitabine is highly affected by infusion duration and gender, while effects of plasma protein binding are negligible [98].

Currently, the standard regimen consists of 1000 mg/m2 gemcitabine given weekly as a 30 min infusion for 3 weeks followed by one week of rest. While that dosage is believed to saturate dCK activity, pharmacokinetic studies have determined that a 1000 mg/m2/h infusion rate was optimum for intracellular phosphorylation of gemcitabine

[99]. Nonetheless, several studies have shown 1500 mg/m2 as the maximum tolerated dose (MTD), with high-grade neutropenia, granulocytopenia, and thrombocytopenia being dose-limiting factors [100, 101]. Trials comparing a fixed dose rate (FDR) of

gemcitabine (1500 mg/m2 at a rate of 10 mg/m2/min weekly for 3 weeks out of a 4-week cycle) with standard gemcitabine treatment indicated increased efficacy of the FDR but with much greater hematological toxicities (i.e., neutropenia, anemia, and thrombocytopenia) in patients [102-104]. Furthermore, although one study established the MTD of gemcitabine at 6500 mg/m2 with hematopoietic progenitor support, its efficacy was not superior than that reported with lower-dosage FDR schedules [105]. A high dose of gemcitabine was also investigated with 2200 mg/m2 administered as an infusion for 30 min on days 1 and 15 for 6 months. The regimen was deemed safe, tolerable, and effective for palliative treatment of advanced pancreatic cancer patients

[106]. Collectively, these data indicate that altering the exposure level of gemcitabine in patients is vital for obtaining specific desired effects.

When comparing schedules, 1000 mg/m2 gemcitabine as a 30 min infusion given weekly for either 3 consecutive weeks every 4 weeks or 2 consecutive weeks every 3 weeks indicated comparable efficacy although the 3-week schedule produced lower toxicity in patients [107]. Another schedule of 1000 mg/m2 gemcitabine given for 7 consecutive weeks followed by a week of rest and then weekly for 3 weeks out of 4 identified increased toxicities compared with the conventional regimen [108].

Metronomic dosing, or continuous and timed administration of low dosage, of gemcitabine (1 mg/kg/d for a month) was found to exert equal cytotoxicity in orthotopic models of human pancreatic carcinoma in nude mice compared with the conventional schedule of 100 mg/kg/days 0, 3, 6, and 9 post-implantation. However, an antiangiogenic effect was additionally observed with the metronomic regimen [109].

Infusion duration has also been identified as a key factor that influences gemcitabine efficacy. One study demonstrated that the chemotherapeutic retained its antitumor activity at doses as low as 300 mg/m2 when infusion time was prolonged [110].

Other trials increased the infusion rate of 100 mg/m2 to 24 h weekly for 3 weeks in a 28- day cycle and found an improvement in the quality of life of patients but also only marginal antitumor activity [111]. Furthermore, an investigational method of hypoxic abdominal stop-flow perfusion demonstrated increased efficacy of gemcitabine in patients compared with the standard treatment at doses up to 1125 mg/m2. However, this complicated modality remains in its infancy, and further studies are needed for its complete assessment [112].

Approaches to Directly Modifying Determinants of Gemcitabine Cytotoxicity:

Gene Therapy

Even though gene therapy has not yet been well-developed for the application of cancer (as compared with other disorders such as cystic fibrosis), the approach has become appealing for altering specific targets with its potential for producing a bystander effect within heterogeneous tumors. A fusion gene of dCK and uridine monophosphate kinase (UMK) (dCK::UMK) was found to sensitize pancreatic cancer cells to gemcitabine by markedly decreasing viability. Tumor volume was also reduced as an antitumoral bystander effect was observed due to apoptosis of untransduced cells.

Additionally, the use of a synthetic carrier (i.e., polyethyleneimine (PEI)) further induced tumor regression [113]. A combination of dCK::UMK with siRNA against RRM2 and thymidylate synthase (TS), whose inhibition activates hENT1, and gemcitabine promoted chemosensitivity even more with a 40-fold decrease in cytotoxic IC50 in PANC-1 cells.

Tumor volume was reduced dramatically and mouse survival prolonged significantly due to an increase in apoptosis and decrease in cellular proliferation [114]. Recently, it has also been found that the overexpression of dCK and knockdown of p8 with recombinant adenoviral vectors also significantly decreased gemcitabine resistance in PANC-1 cells and inhibited tumor growth with enhanced apoptosis and caspase-3 activity [46].

The herpes simplex virus 1 thymidine kinase (HSV-TK)/ganciclovir (GCV) strategy has thus far been successful up to the clinical level, and the same technique may be adapted to target increased RR in cancer cells. RR has been a target of interest with gene therapy since its overexpression is an attribute of mitotic cancerous cells. Therefore, vectors lacking the RR gene are attractive options due to their ability to complement with overexpressed mammalian RR and selectively target rapidly dividing cells. For example, a herpes simplex virus (HSV) with an RRM1 deletion mutant (i.e., ICP6Δ) was found to enhance the expression of adeno-associated viruses along with their kinase genes both in pancreatic cancer cells and xenografts, demonstrating potential for the combination therapy [115]. Similarly, an HSV vector lacking the RR gene (i.e., hrR3) was found to extend survival in 70% of mice receiving both the vector and ganciclovir, 40% of mice receiving the vector alone, and 0% of untreated mice [116].

Overexpression of hENT1 has also been conducted with a recombinant adenovirus (i.e., Ad-hENT1) in human pancreatic cancer cells. Although expression of the transporter is known to vary with cell cycle and cell type, treatment with the vector improved response to gemcitabine [117]. Other targets of gene therapy in pancreatic cancer include p53 and p16 [118] and Bax [119], although they are not directly related to gemcitabine metabolism.

Epigenetic Approaches

The field of epigenetics has been rapidly emerging since the identification of its influence on the initiation, progression, and resistance of various cancers. The ability to reverse aberrant epigenetic alterations renders this targeted approach highly attractive.

Among the various mechanisms, DNA methylation has been the most widely studied.

Abnormal promoter hypermethylation can silence tumor suppressor genes and have been implicated in the clinicopathological features and promotion of tumorigenic properties of pancreatic cancer. One study identified that low expression of dCK in pancreatic cancer patients correlated with overall survival as well as age, suggesting a role of age-related methylation of the dCK gene [47]. It was also found that the CDA gene was methylated in an entire cohort of colorectal cancer patient samples, although no correlation was observed between methylation status and clinicopathological parameters [120]. By utilizing microarrays, countless other aberrantly methylated genes have been identified in pancreatic cancer, particularly those related to growth, differentiation, angiogenesis, and apoptotic signaling [121-124]. However, further information about the methylation status of the targets of the gemcitabine pathway is limited.

FDA-approved inhibitors of DNA methylation are currently available with several more under investigation. For example, the nucleoside analogs 5-azacytidine (5-aza-C) and 5-aza-2’-deoxycytidine (5-aza-dC) have been approved for the treatment of high risk myelodysplastic syndromes [125] and are currently under study for acute and chronic myeloid leukemias and ovarian cancer [126]. Both compounds have been found to be substrates for hCNT1 [127] and rely on dCK for their phosphorylation [128].

Combination of 5-aza-dC with gemcitabine was shown to inhibit pancreatic cancer cell

growth to a greater extent than with gemcitabine alone [129]. The DNA methylation inhibitor cladribine, also a nucleoside analog, has been studied in conjunction with rituximab for the treatment of mantle cell lymphoma; high levels of activity with minimal toxicity were found [130].

In addition to DNA, epigenetic modifications, namely methylation and acetylation, can also occur at the histone level. Unlike DNA methylation, histone methylation is a mechanism that has yet to be thoroughly studied in the context of cancer.

Histone acetylation, on the other hand, is a mechanism well-studied in the development and progression of cancer since aberrant activity promotes the expression of various oncogenes. This process is reversibly catalyzed by histone acetyltransferases (HATs) and histone deacetylases (HDACs). Currently, FDA-approved HDAC inhibitors include vorinostat and romidepsin for the treatment of T cell lymphoma, although many others remain under study in clinical trials [126]. Romidepsin has also been demonstrated to have antiproliferative activity in five human pancreatic cancer cell lines (IC50: 1-500 nM) via cell cycle arrest at both G1 and G2/M phases followed by apoptosis [131].

Additionally, the HDAC inhibitor MS-27-275 was shown to decrease cells in S phase cells while increasing G1-phase cells. Antitumor efficacy was observed in both Capan-1 cells and mouse xenograft [132]. Trichostatin A (TSA) [133], another HDAC inhibitor, has been widely studied and shown to induce antitumor effects in pancreatic cancer cell lines at submicromolar concentrations [134, 135]. Furthermore, a novel compound derived from both MS-27-275 and TSA was shown to cause cell cycle arrest and subsequent apoptosis in pancreatic cancer cell lines [136]. Several studies have shown the increased efficacy of combining an HDAC inhibitor with gemcitabine in pancreatic

cancer cell lines including the synergistic apoptotic effects of TSA with gemcitabine in pancreatic ductal carcinoma cells [137-141]. With 50% reduction of pancreatic cancer xenografts and no observed toxicity, Donadelli et al. further identified that TSA did not affect the gene expressions of targets in the gemcitabine pathway [138]. Nonetheless, a phase II trial of CI994 (n-acetyl dinaline), an HDAC inhibitor, with gemcitabine demonstrated no additional advantage in survival rate compared with gemcitabine alone

[142], warranting further studies for the better understanding of this therapeutic approach.

The rapidly expanding field of miRNAs, short (~22 nt) RNA sequences that post- transcriptionally regulate gene expression by binding to complementary mRNA, is also appealing for its potential to influence cancer epigenetic therapy. Their ability to regulate expression of canonical oncogenes and tumor suppressor genes as well as the recent demonstration of their direct involvement in oncogenic processes emphasizes their potential to intervene in dysfunctional cancer pathways. There is also increasing evidence demonstrating that miRNAs are critical regulators of drug resistance in pancreatic cancer.

For example, re-expression of the miR-200 family or downregulation of miR-21 in gemcitabine-resistant cells led to sensitization of the cells to gemcitabine [7]. Since miRNA expression patterns are unique to each tumor type, aberrant expression profiles

(e.g., the downregulation of the let-7 family) have been shown to contribute to cancer pathogenesis and tumor development and progression [7, 143]. In addition to being epigenetic targets, miRNA activity can also be modulated by inhibition with antisense oligonucleotides or overexpression with synthetic miRNAs, as reviewed recently [7,

143]. The development of curcumin as a potential therapeutic for pancreatic cancer patients has been under investigation due to its interactions with DNA methyltransferase,

HDACs, and HATs and ability to alter miRNA expression, including upregulation of miR-200 and downregulation of miR-21 [144]. In this regard, several recent reviews identify key growth, proliferation, and differentiation (e.g., PI3K/Akt, MAPK, K-ras,

STAT-3, Notch-1, Notch-2), invasion (e.g., ABCG2, cadherin, ZEB1, vimentin), and apoptosis (e.g., CDK6, p53, caspase 3) proteins altered by miRNAs [7, 143, 145].

Therefore, targeting miRNAs offers great promise for tackling pancreatic cancer chemoresistance and providing diagnostic and therapeutic values.

Studies from our laboratory show miRNAs influencing the expression of direct targets of the gemcitabine pathway. We have identified that miR-214, miR-339-3p, and miR-650 overexpression significantly reduced hCNT1, but not hENT1, protein levels.

Furthermore, several miRNAs were found to markedly reduce both hCNT1 transport activity as well as gemcitabine cytotoxicity in a pancreatic cancer cell line [9]. Likewise, our recent study showed that the differential processing of the let-7 family of miRNAs altered gemcitabine chemosensitivity in pancreatic cancer cells through its activity on RR

(unpublished data). Both studies identify the potential for modulating miRNA expression to enhance the efficacy of gemcitabine, although the ability of pancreatic cancer cells to process miRNA precursors should be carefully considered while choosing precursor miRNA-based therapies. In addition to overcoming chemoresistance, epigenetic and miRNA alterations also have the potential to act as prognostic and predictive biomarkers for the guidance of therapy.

Molecular Therapeutic Agents

Several small and large molecule inhibitors demonstrate potential as therapeutic agents. For example, proteasomal inhibitors, such as bortezomib, have been shown to

increase cytotoxicity in combination with gemcitabine [9, 146]. Similarly, tetrahydrouridine (ThU), a competitive inhibitor of CDA, has been shown to sensitize pancreatic cancer cells to gemcitabine up to 54-fold [45]. Although pharmacokinetic studies have been performed on the compound, its therapeutic value as well as safety and toxicity profiles remain under study in humans [147]. RNA interference has also been commonly used for the study of specific targets in pancreatic cancer chemosensitivity.

While Ohhashi et al. noticed no change in gemcitabine sensitivity or cellular proliferation with hENT1-targeted siRNA [42], others found that knockout of TS with siRNA decreased chemoresistance in pancreatic cancer cell lines [148]. Furthermore, shRNA against hENT1 combined with 5-fluorouracil (5-FU) treatment attenuated cell viability and gemcitabine cytotoxic IC50 values. Given the dual treatment, cells were arrested in

G0/G1 phase, while mouse xenografts had diminished tumor volumes [35]. Furthermore, modulation of transporter activity and, consequently, gemcitabine uptake was recently conducted with siRNA against hCNT1 [9]. Similarly, silencing of dCK with siRNA in gemcitabine-resistant cells further reduced gemcitabine sensitivity without affecting cellular proliferation [42]. Treatment of gemcitabine-resistant PANC-1 cells with hydroxyurea led to a 4-fold increase in chemosensitivity [149]. RRM1- and RRM2- targeted siRNA in pancreatic carcinoma cell lines increased gemcitabine sensitivity, apoptosis, and caspase-3 activity while lessened cell proliferation and invasiveness [42,

63, 64, 150]. In an orthotopic xenograft model, synergism was observed between RRM2 silencing and gemcitabine, leading to reduced tumor growth, enhanced tumor apoptosis, and inhibition of metastasis [64].

Novel Therapeutic Combinations with Gemcitabine

In addition to radiosensitization, potentiating chemosensitization of gemcitabine by other pharmaceutical agents has been a longstanding area of interest. Although combination therapies with gemcitabine have been studied extensively both in vitro and in clinical trials, only a few have consistently resulted with significant improvements in patient survival. Since several mechanisms of drug resistance exist, combination therapy can simultaneously undertake multiple mechanisms and pathways, allowing the drugs to act in concert. This rationale is well-illustrated by the current approach towards breast cancer treatment. Nonetheless, combinations of gemcitabine with other chemotherapeutics such as other nucleoside analogs (i.e., 5-FU, capecitabine), platins

(i.e., cisplatin, oxaliplatin), and taxoids (i.e., docetaxel [133, 151], paclitaxel [152]) have shown variable results with no significant enhancement of patient survival [6].

With the interests geared towards targeted therapy, numerous novel therapeutics have been studied with gemcitabine: antibodies (i.e., bevacizumab, cetuximab), growth factor inhibitors (i.e., erlotinib, aflibercept, axitinib [153], saracatinib [154]), topoisomerase I inhibitors (i.e., irinotecan, exatecan, rubitecan), MMP inhibitors (i.e., marimastat, BAY 12-9566), tyrosine kinase inhibitors (i.e., masitinib, sorafenib, sunitinib

[155], ARQ 197 [156]), and other inhibitors (i.e., pemetrexed, tipifarnib, celecoxib, imitinib [157], PX-12 (a novel inhibitor of thioredoxin) [152] and enzastaurin [158]) [6].

Among the sizeable list, several agents show promise. However, erlotinib has been the only agent thus far to have consistently demonstrated equal or significantly enhanced median overall survival compared with gemcitabine (1-year survival rates of 23% and

17%, respectively) [6]. Therefore, the EGFR inhibitor is currently FDA-approved for use

in combination with gemcitabine for the treatment of pancreatic cancer. Other agents of interest include curcumin (a natural product) [6], leucovorin (an adjuvant agent) [6], tegafur (a prodrug of 5-FU) [98, 159], and MGCD0103 and valproic acid (HDAC inhibitors) [160]. In particular, notable combinations under study include GTX

(gemcitabine, docetaxel, and capecitabine) [161, 162], GemOx (gemcitabine, oxaliplatin)

[163], GemOxCet (gemcitabine, oxaliplatin, cetuximab), and Gemoxel (gemcitabine, oxaliplatin, capecitabine) [164]. Nonetheless, novel combinations not including gemcitabine such as Xeliri (capecitabine, irinotecan) [165], Folfiri (leucovorin, 5-FU, irinotecan) [165], and Folfirinox (leucovorin, 5-FU, irinotecan, oxaliplatin) [166] have also shown promise in combating pancreatic cancer, although more studies are needed to determine toxicity and efficacy profiles.

An alternative approach studied is the use of agents for modulating the tumor microenvironment. Several clinical trials have been conducted with prophylactic anticoagulants (i.e., nadroparin, enoxaparin, dalteparin) with gemcitabine, although minimal significant results for the advancement of pancreatic cancer therapy were obtained [6]. Similarly, the use of pomalidomide, an investigational immunomodulating drug that both inhibits angiogenesis as well as exerts antitumoral effects, was found to be feasible and safe in most patients [27]. When co-administered with gemcitabine, a drug depleting tumor-associated stromal tissue via inhibition of the Hedgehog cellular signaling pathway, IPI-926, was found to facilitate gemcitabine intratumoral delivery and extend survival of a spontaneous mouse pancreatic tumor model [167]. Furthermore, a novel ceramide analog, AL6, has been shown to increase the dCK/CDA gene expression ratio, leading to increased apoptosis in two pancreatic ductal adenocarcinoma cell lines

[168]. A study of gemcitabine in combination with imexon, a pro-oxidant small molecule, was found to significantly increase chemotherapeutic efficacy in PANC-1 xenograft tumors. This is thought to be due to the inhibition of RR by imexon, with inhibition effects even greater with the addition of gemcitabine [169]. Overall, combinations of a targeting agent with gemcitabine have shown to be well-tolerated with promise for enhanced efficacy. Further studies as well as meta-analysis of data from clinical trials are warranted for consistent efficacy results.

Summary

Gemcitabine is predicted to remain as a key treatment option for pancreatic cancer patients for the near future. Determinants of chemoresistance in the gemcitabine pathway are well-known, yet many challenges remain in pursuing the targets of interest to improve efficacy. Furthermore, other than the gemcitabine pathway, numerous other cellular and tumoral determinants pose as obstacles in overcoming chemoresistance. Nonetheless, novel therapeutic approaches are currently underway to circumvent limitations in gemcitabine transport, phosphorylation, and overall efficacy. Although therapeutic challenges are clear and present, countless studies, both at the benchtop and clinical levels, continue to provide clues for maximizing chemotherapeutic efficacy in patients.

Purpose and Rationale of Study

The purpose of this dissertation is to better understand how pancreatic cancer cells acquire chemoresistance and the ability to metastasize. Specifically, the study focuses on identifying key cellular, molecular, and epigenetic changes that allow these cells to survive in the presence of chemotherapeutics such as gemcitabine as well as spread to

distant sites in the body. The rationale of this dissertation is that the identification of determinants contributing to chemoresistance could lead to the discovery of novel drug targets for improved therapeutic management of pancreatic cancer.

Objective and Hypothesis of Study

The overall goal of the lab is to improve the chemotherapeutic management of pancreatic cancer. The overall objective of this dissertation is to determine ways to exploit the key characteristics that allow cancer cells to become chemoresistant and metastatic for improved sensitivity to therapeutics. The hypothesis is that targeting these cellular changes may lead to the discovery of novel treatment avenues that may be utilized in combination with current chemotherapies to enhance efficacy and reduce toxicity in pancreatic cancer patients.

Expected Results and Significance of Study

The expected results of this dissertation are the better understanding of cellular changes leading to chemoresistance, the investigation of key cancer genes and pathways as ‘druggable’ targets, and the evaluation of potential treatment approaches by utilizing improved pharmacological agents and novel drug delivery. These expected results are significant because they could lead to further clinical studies for a novel approach in treating pancreatic cancer. Moreover, these changes may be applicable to other solid tumors in which similar characteristics are found and the same chemotherapeutic drugs are used.

Section and Chapter Summaries

This manuscript-style dissertation consists of three sections with five research projects, each presented as a separate chapter. The first section focuses on targeting DR- phenotype cells in order to improve chemosensitivity. These cells include, but are not limited to, populations that have lost expression of nucleoside transporters necessary for drug uptake, intracellular connections necessary for drug transfer between cells, and drug metabolizing enzymes necessary for prodrug activation [7, 8].

Chapter 2 studies whether the expression of epithelial (E)-cadherin, a transmembrane protein that mediates cell-cell adhesion, can restore the gemcitabine activation pathway in order to improve chemosensitivity. We examined changes in the expression of gemcitabine transporters, gemcitabine phosphorylating and metabolizing enzymes, as well as other DNA replication and repair enzymes that affect gemcitabine cytotoxicity. We found that E-cadherin enhanced gemcitabine efficacy predominantly through increasing hENT1 expression and activity, leading to increased cellular transport and cytotoxicity of the drug. These results show that cell adhesion molecules, such as E- cadherin, can stabilize drug transport proteins at the cell surface in order to modulate drug sensitivity in pancreatic cancer. This study also shows potential for using E-cadherin as a biomarker for guiding nucleoside analog therapy in pancreatic cancer patients.

The aim of Chapter 3 is to characterize the intracellular (within the cell) and intercellular (cell-to-cell) movements of nucleosides in order to identify ways of improving drug targeting and chemotherapeutic efficacy. Our approach was to manipulate the expressions of nucleoside transporters, gap junctions (intracellular channels composed of connexins (Cxs)), or both to increase nucleoside drug

concentration within tumor cells, resulting in greater cytotoxicity. We found that increasing the expression of hCNT1 and Cx32, but not hENT1 and Cx43, resulted in greater nucleoside analog concentration within the cell population. However, no significant change was observed for gemcitabine cytotoxicity. These results indicate that different members of the nucleoside transporter and gap junction families can have disparate roles in nucleoside analog transport and efficacy in cells. Further studies are warranted to understand which combination of members can confer both enhanced drug uptake as well as efficacy.

The second section focuses on decreasing the stemness of advanced pancreatic cancer cells using pharmacological means. As a drug-resistant subpopulation, these pluripotent cells can survive chemotherapy and renew the tumor [15-17]. By targeting the stem cell characteristic, pancreatic cancer cells may become sensitized to chemotherapy before renewal can occur. In other words, it may be possible to overturn the chemotherapeutic refractoriness that tumors with CSC populations exhibit.

Chapter 4 evaluates the potential of an investigational histone methylation reversal agent, 3-deazaneplanocin A (DZNep), for decreasing the stemness associated with pancreatic cancer and improving the sensitivity of pancreatic cancer to gemcitabine.

Cellular proliferation assays identified that co-treatment of DZNep and gemcitabine enhanced cytotoxic effects in the cancerous pancreatic cell lines, while DZNep exerted antagonism with gemcitabine against the normal pancreatic cell line, HPDE. Further optimization studies revealed that short priming with DZNep followed by gemcitabine treatment, rather than co-treatment, produced a maximal chemosensitization response in the pancreatic cancer cells. Drug delivery using engineered nanoparticles that mimicked

this short priming further enhanced cytotoxicity. Ultimately, we found that histone methylation reversal by DZNep presensitizes pancreatic cancer cells to gemcitabine.

The third section examines whether reverting the EMT phenotype in cells could restore chemosensitivity. The epithelial-mesenchymal transition is a process by which epithelial cells lose their cell-cell contacts, polarity, and cuboidal shape to become mesenchymal cells with a spindle-like shape and migratory and invasive properties. As the cells gain metastatic ability, their response to chemotherapy decreases. This process has been well-characterized with several markers indicating an epithelial (e.g., E- cadherin, claudins, occludins) versus mesenchymal (e.g., N-cadherin, vimentin, fibronectin) phenotype [20, 170, 171]. In addition to cellular markers, epigenetic states have also been implicated in EMT [18, 19, 21, 172].

Chapter 5 describes how expression of the SET oncoprotein contributes to the progression of pancreatic cancer, particularly by regulating the EMT-phenotype subpopulation of tumor cells. This study was derived from a previous project in our lab which identified SET as a novel regulator of the let-7 miRNA family [173]. Let-7 is a tumor suppressor involved in differentiation and chemosensitivity. Among the two major isoforms of SET, we found that overexpressing the cell-surface isoform 2, rather than the better-known nuclear isoform 1, in the epithelial PANC-1 cell line induced EMT by causing a spindle-like, mesenchymal morphology and promoting cellular proliferation, colony formation, migration, and invasion. The induction of EMT in cells was determined to be a result of cadherin switching (i.e., from expressing epithelial E- cadherin to mesenchymal N-cadherin). Further mechanistic studies identified that SET- induced cadherin switching involved the SPARC/Slug and Rac1/JNK/c-Jun/AP-1

signaling pathways which inhibit E-cadherin and promote N-cadherin, respectively.

These findings have implications for the design and targeting of SET and its associated pathways for intervening pancreatic tumor progression.

The last chapter consists of a general discussion of the overall results as well as potential future directions.

Scope of Work and Limitations

The earlier chapters support the potential of E-cad to be used as a biomarker for early diagnosis, drug targeting, and guiding treatment. With the use of chemotherapeutics, this biomarker may avoid unnecessary gemcitabine usage and toxicities in gemcitabine-refractory pancreatic cancer patients and instead use an alternative medicine such as tyrosine kinase inhibitors or other molecular therapies currently approved for pancreatic cancer therapy. E-cad shows promise as a biomarker because of its role in regulating hENT1, an already employed biomarker for staging of differentiation and prediction of patient survival. Therefore, E-cad can be used as a surrogate marker and/or as part of a biomarker index to indicate preliminary patient response to nucleoside analog therapeutics.

Nonetheless, several limitations exist in our study. Firstly, there is not a perfect correlation between E-cad expression and hENT1 expression and activity. In fact, in cadherin-null A431D cells, we found hENT1 expressed at the cell surface and functional, suggesting that E-cad is not a requirement or absolute marker for hENT1 (see details in

Chapter 2). However, E-cad does augment hENT1 cell surface expression and activity and reduce the rate of degradation of the transporter. In a pancreatic cancer system,

numerous cadherins are expressed. Currently, most of their roles in conferring chemoresistance remain unknown, and it is possible that they are able to compensate for the loss of E-cad. Likewise, there are numerous transporters involved in gemcitabine uptake. In this study, we identified significant changes in hENT3 and hCNT1 transcript levels with E-cad expression. In a previous study, we found that hCNT1 correlates with differentiation and influences chemosensitivity [9]. Another limitation of this study is that we used pancreatic cancer cell lines as models for pancreatic cancer. While we focused on these parenchymal cells, there are several other cell types involved in pancreatic cancer: stromal cells, stellate cells, endothelial cells (blood vessels), and immune cells

(e.g., macrophages). The status of E-cad in these other cell types as well as their response to nucleoside analogs remains unknown. Furthermore, for this study, we were not able to acquire a large number of pancreatic ductal adenocarcinoma (PDAC) tissues, so we were not able to study the E-cad/hENT1 correlation with high statistical power. Overall, further studies are warranted - both retrospective (using human tissue samples) and prospective (conducting human clinical trials).

The study in Chapter 3 focuses on two key transporters of gemcitabine (hCNT1 and hENT1) and two prototypic Cxs (Cx32 and Cx43). Since Cxs can form hemichannels at the cell surface as well as GJs at cell-cell contacts, these proteins are fundamental determinants of preliminary response towards gemcitabine. With determinants of both nucleoside analog transport into cells and transfer between cells, these candidates can contribute to a meaningful biomarker index by compensating for inherent defects and eliciting a bystander effect. The NT/Cx combination can perhaps provide more valuable predictions than a single biomarker.

Similar to the previous study, there are more than 20 types of Cxs and not all are characterized or well-understood in pancreatic cancer. However, one recent study demonstrated the significant role of Cx26 in enhancing the gemcitabine bystander effect in pancreatic cancer [174]. In addition to Cxs, the role of pannexins, a family of proteins that form large transmembrane channels similar to GJs, has recently come under study in relation to cancer [175, 176].

Overall, our studies in the innately DR-phenotype pancreatic cancer cell population only focus on a few key determinants of gemcitabine. There are many other kinases, replication repair enzymes, etc. that together control the chemosensitivity process. Many of these candidates were profiled in Chapter 2, but not studied in-depth. It is likely that the collective roles of all these players will provide a more significant effect for overcoming chemoresistance than modulating only one determinant alone. In our studies, we found transporters to play a dominant role, and so we focused on the uptake aspect of gemcitabine efficacy. However, transporters are only the first rate-limiting step in the gemcitabine activation process; dCK is also a rate-limiting step due to its exclusive role in phosphorylating the gemcitabine prodrug. Currently, an increasing amount of studies are identifying the key role of dCK in the nucleoside salvage process (e.g., in lymphocyte development [177, 178], leukemia [179-185], and gynecological cancers

[186, 187]). Although alterations in dCK levels are not commonly seen in pancreatic cancer, it may be a candidate to be further studied in detail.

The focus of the work in Chapter 4 is sensitizing pancreatic cancer cells to gemcitabine by reverting stemness. This is a novel area in which we attempted to reduce stemness in pancreatic tumors and force them to respond to chemotherapeutics. This

approach, epigenetic modification plus the use of current chemotherapeutics, is quickly becoming of interest. In this study, we tested a drug candidate called DZNep. While this drug is not perfect, it showed potential in reverting histone methylation exclusively in cancer cells, increasing chemotherapeutic efficacy without compromising for toxicity. To optimize this drug candidate, we identified the most favorable time, dose, and exposure conditions. We found that priming the pancreatic cancer cells with DZNep for 8 h before gemcitabine treatment resulted in the greatest efficacy. Subsequently, we devised novel formulations to mimic this sequence of exposure and enhanced the delivery strategy to administer both drugs together with different time points for release.

The global effects of DZNep constitute some of its major limitations. Both our own studies as well as studies recently published implicate the possible role of DZNep on early embryonic and neural progenitor cells [188-190], which is expected since advanced cancers are a simple recapitulation of early embryonic development. Another study demonstrated that DZNep produces developmental and neuronal disorders in a zebrafish model (personal communications). To overcome these limitations, the subsequent research objective is to optimize the applicability of these investigational histone methylation reversal agents. One approach is to better understand the precise mechanisms of action of DZNep. A current fellow graduate student is now following up on this work, determining DZNep control of miRNAs. Preliminary results show that although DZNep has an enormous effect on reprogramming miRNA transcripts and is likely to reprogram other non-coding RNAs (e.g., lncRNA) as well as control determinants of chemosensitivity. A second approach is to identify the precise effectors of DZNep and design new molecules to target those effectors directly for only the intended benefits.

Another similar approach would be to synthesize novel DZNep analogs and test them in experimental systems. Although it poses as a difficult challenge, identifying cellular pathways unique to the carcinogenesis and tumorigenesis process but not to early embryonic development would be another strategy. This would segregate the two populations for exclusive drug targeting. Perhaps the most promising approach thus far is the use a polymeric delivery system to target compounds directly to cancer cells.

Significant advances in this strategy have been made recently, and the precise delivery of cytotoxic agents to cancer cells could very soon become a reality.

Our work in Chapter 5 was the first to identify SET overexpression in pancreatic tumors; a recent independent study corroborated our results [191]. In-depth analyses identified SET isoform 2 as the major form found at the cell surface, and that it facilitated

EMT through the switching of E-cad expression to N-cad. SET was also found to promote metastasis, suggesting it can be utilized as a biomarker for early diagnosis or for early prediction of metastasis. However, it is currently unclear when during the tumorigenic process the switch in cadherins occurs. Nonetheless, since SET isoform 2 is expressed at the cell surface, it could be an ideal candidate for easy drug targeting. If SET can be successfully inhibited, there is potential for anti-metastatic intervention in patients.

Manipulation of SET in our studies revealed that the oncoprotein reprograms a number of EMT genes (see details in Chapter 5). Therefore, it seems that SET could be a master regulator of genes involved in EMT. Additional studies need to be completed in order to identify exactly how SET regulates these genes. However, since the protein has been identified as part of the inhibitor of histone acetyltransferase (INHAT) complex

[192], it is possible that it exerts a chromatin effect on these EMT genes. More studies are

required to understand this potential chromatin remodeling process. Furthermore, the role of SET in directly determining chemosensitivity is not clear at this point. We only know that it is involved in metastasis and the regulation of let-7 which influences chemosensitivity [173]. Lastly, it may be possible that SET can be targeted by a different set of chemotherapeutics, other than nucleoside analogs, to produce a therapeutic response.

In summary, the scope of my work opens up both direct applicability (i.e., more therapeutic and diagnostic avenues) as well as more basic questions (e.g., how SET regulates genes). Although work up to the animal level (i.e., an orthotopic mouse model) was completed, there remains a large translational barrier that can only be overcome by further clinical studies. Overall, there remain several therapeutic challenges in overcoming chemoresistance in pancreatic cancer. While this dissertation attempts to examine the key characteristics of chemoresistance in three distinct tumor subpopulations, there may be additional subpopulations that also exist and need investigation. While the focus of this dissertation is on improving the gemcitabine treatment of pancreatic cancer, results may extend to other cancers with similar cellular and tumoral features and in which nucleoside analog therapy is used.

DRUG RESISTANT-PHENOTYPE CELLS AND CHEMOSENSITIVITY

CHAPTER 2

E-CADHERIN INCREASES CHEMOSENSITIVITY THROUGH HUMAN

EQUILIBRATIVE NUCLEOSIDE TRANSPORTER 1 (HENT1) ACTIVITY AND

STABILIZATION 2

2 SW Hung, YD Bhutia, K Naidu, B Patel, and R Govindarajan. To be submitted to Biochemical Journal.

Abstract

E-cadherin (E-cad), a cell-cell adhesion molecule expressed in well-differentiated epithelial cells, was shown to increase the cytotoxic response of gemcitabine (2’,2’- difluoro-2’-deoxycytidine). Since the precise mechanisms of this response are unclear, we hypothesized that E-cad acts on the gemcitabine activation pathway to improve chemosensitivity. To address this, we analyzed changes in expression of gemcitabine transporters, gemcitabine phosphorylating and metabolizing enzymes, and other DNA replication and repair enzymes that affect gemcitabine cytotoxicity in pancreatic cancer cells. Retroviral expression of human E-cad into MIA PaCa-2, a poorly-differentiated, E- cad-null pancreatic cancer cell line, showed maximal increase in human equilibrative nucleoside transporter 1 (hENT1) protein compared with the other players studied.

Consistently, E-cad expression steeply augmented hENT1-mediated 3H-gemcitabine cellular transport and cytotoxicity, which were ablated by pharmacological inhibition of hENT1 activity. Further mechanistic studies in A431D, a cadherin-null skin carcinoma cell line, showed that the increase in hENT1 transport activity was due to increased total cell hENT1 content as well as cell surface hENT1 expression and decreased protein degradation. These results suggest that E-cad enhancement of gemcitabine efficacy occurs at least in part through increased hENT1 expression, stability, and activity, leading to increased cellular transport and cytotoxicity of gemcitabine. These data demonstrate that E-cad may be a prospective biomarker for guiding nucleoside analog treatment in pancreatic cancer patients.

Introduction

Cadherins are homophilic, calcium-dependent transmembrane proteins vital for cell-cell adhesion, cellular morphology, and cell-cell recognition events. Of all cadherins, epithelial cadherin (E-cadherin or E-cad) has received the greatest attention over the years as its expression is often diminished or lost in cancer cells leading to cell dedifferentiation, epithelial-to-mesenchymal transition (EMT), and cell migration.

Currently, a large body of evidence in literature demonstrates E-cad as a suppressor of

EMT and metastasis and therefore a potential biomarker for the progression of cancer. In addition to these well-established roles, E-cad has also been shown to influence chemosensitization of cancer cells. Decreased E-cad expression is associated with poorer survival in several cancer types. As shown in pancreatic, breast, colon, and prostate cancer cell lines, this strong association occurs mainly via decreased sensitivity to chemotherapeutic agents [193-200]. On the other hand, overexpression or endogenous stimulation of E-cad expression has been shown to improve the chemosensitivity of anticancer agents including platinum drugs (e.g. cisplatin, oxaliplatin) [194, 200, 201], taxanes (e.g. paclitaxel, docetaxel) [200], and anthracyclines (e.g. doxorubicin, epirubicin) [200]. Naturally occurring cancer chemopreventive agents such as 3,3’- diindolylmethane (DIM), isoflavone, and retinoic acid [195, 202] have also been shown to improve E-cad expression, whereas several molecularly-targeted anti-cancer therapeutic agents have been shown to mediate its effects, at least partially, by increased

E-cad expression. The latter group include epithelial growth factor receptor (EGFR) inhibitors (e.g. cetuximab, sulindac sulfide, erlotinib) [197, 203-205], peroxisome proliferator-activated receptor-γ (PPAR-γ) activators (e.g. thiazolidinediones, ciglitazone)

[204], Hedgehog inhibitors (cyclopamine) [206], and topoisomerase II inhibitors

(etoposide) [207]. While it is likely that E-cad’s effect on cellular genes and signaling networks associated with growth, proliferation, and differentiation (MAPK,

PTEN/PI3K/Akt, Src, p53) [208-214] could enhance anti-cancer efficacy of chemotherapeutics, revealing the precise mechanisms by which E-cad improves chemosensitivity has proven to be difficult.

Recently, several reports have shown E-cad to potentiate the efficacy of a synthetic nucleoside analog 2’,2’-difluoro-2’-deoxycytidine (gemcitabine; dFdC) used in first-line treatment of pancreatic cancer [215, 216]. Although the relationship between E- cad expression and gemcitabine chemosensitivity in pancreatic cancer cells has been compelling [193-198], the mechanisms linking the two have not yet been understood.

Identifying such mechanisms could facilitate further improvement in chemo- or radiosensitization of pancreatic cancer cells or guide treatment by way of selection of patient groups and specific nucleoside analogs used in treatment (gemcitabine, troxacitabine, capecitabine etc.). Gemcitabine is a hydrophilic pro-drug that enters the cell via five primary nucleoside transporters: the Na+-independent human equilibrative nucleoside transporters (hENT1-3) and the Na+-dependent human concentrative nucleoside transporters (hCNT1 and hCNT3). Once inside the cell, gemcitabine is activated upon rate-limiting phosphorylation by deoxycytidine kinase (dCK).

Nonetheless, intracellular gemcitabine can also be deactivated into harmless compounds by cytidine deaminase (CDA). Further along, the active forms of gemcitabine can either halt DNA replication directly by incorporation into DNA or inhibit ribonucleotide reductase subunits M1 (RRM1) and M2 (RRM2) essential for the synthesis and

maintenance of deoxyribonucleotide pools. Inhibition of these two enzymes, critical for

DNA synthesis and repair, also contributes to suppression of cancer cell proliferation

[216-218]. We hypothesized that E-cad may increase chemosensitivity of nucleoside analogs in cancer cells through potentiating drug activation mechanisms by affecting one or more of the aforementioned players of gemcitabine transport, phosphorylation, and catabolism. Here we demonstrate that E-cad enhances gemcitabine efficacy in pancreatic cancer cells primarily through increasing hENT1 expression and activity, leading to increased cellular transport of gemcitabine and cytotoxicity. Further our data support that signaling through cell adhesion molecules can induce and recruit drug transport proteins to the cell surface in order to increase nucleoside drug sensitivity.

Materials and Methods

Materials. Gemcitabine, 3H-gemcitabine, 3H-adenosine, and 3H-thymidine were obtained from Chemie Tek (Indianapolis, IN), while 35S-labeled L-methionine was obtained from Moravek Radiochemicals (Brea, CA). FBS was acquired from PAA

Laboratories Inc. (Etobicoke, Ontario). Uridine, DAPI, DMSO, nitrobenzyl mercaptopurine riboside (NBMPR), and 3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide (MTT) were obtained from Sigma-Aldrich (St. Louis, MO).

The bicinchoninic acid (BCA) protein assay reagent was from Pierce Chemical

(Rockford, IL). The fluorescent anti-fade mounting reagent was obtained from Molecular

Probes (Invitrogen). All plastic wares for cell culture were obtained from Corning

(Corning, NY).

Cell Culture. All cell lines were obtained and cultured as previously described

[219]. The A431D and A431DE cell lines were kindly provided by Dr. Parmender Mehta at the University of Nebraska Medical Center (Omaha, NE), maintained in DMEM supplemented with 10% FBS, 100 units of penicillin/ml, and 2 µg streptomycin/ml

(Sigma-Aldrich), and subcultured at a 1:4 ratio.

Retroviral Expressions in Cell Lines. The retroviral expression of E-cadherin in the MIA PaCa-2 cell line was conducted with the expression vector (LZBOB-pac-E-cad), a kind gift from Dr. Keith Johnson at the University of Nebraska Medical Center. Control and E-cadherin-harboring retroviruses were generated as described previously [220], and

MIA PaCa-2 cells were infected. Clones expressing E-cad were selected with and maintained in 5 and 2.5 µg/ml of puromycin, respectively.

Western Blotting. Western blotting was conducted as previously described [219].

The monoclonal E-cad G10 (mouse), polyclonal hENT1 N-12 (goat), polyclonal hCNT1

H-70 (rabbit), polyclonal dCK L-19 (goat), and polyclonal RRM1 T-16 (goat) and RRM2

E-16 (goat) anti-human antibodies used were obtained from Santa Cruz Biotechnology

(Santa Cruz, CA). The polyclonal rabbit anti-human CDA antibody was from Abcam

(Cambridge, MA). The polyclonal rabbit anti-human Na+/K+-ATPase antibody was from

Cell Signaling Technology (Danvers, MA). The monoclonal mouse β-actin antibody was from Sigma-Aldrich. All secondary antibodies were purchased from Bethyl Laboratories

(Montgomery, TX).

MTT Cytotoxicity Assay. The MTT assays were performed as described earlier

[219]. Cytotoxic IC50 values were calculated using GraphPad Prism 5 software.

3H-Nucleoside Transport Study. Cells were seeded in 24-well culture plates

(5x104 cells/well) and cultured until cells reached 90-95% confluency. After aspirating the media, cells were incubated with Na+-free buffer [221] containing 0.02 µM of the 3H- nucleoside in the presence or absence of 10 nM NBMPR. Diffusional uptake was measure in the presence of 10 µM NBMPR and 20 mM uridine. Transport was arrested by immediately washing the cells with Na+-free buffer containing 20 mM cold uridine.

Cells were then lysed with 500 µL of 1% SDS and agitated in a cell shaker for 30 min.

Each well’s lysate was added to 5 mL of ScintiSafe Econo 1 (Fisher Scientific). The mixture was vortexed and counted using an LS 6500 Multipurpose Scintillation Counter.

An additional well for each cell type was cultured in parallel and subjected to protein estimation.

Real-time PCR. Real-time PCR methods were performed as described earlier

[219]. Validated TaqMan primers and probes for hENT1 (Hs00191940_m1), hENT2

(Hs00155426_m1), hENT3 (Hs00983219_m1), hCNT1 (Hs00188418_m1), hCNT3

(Hs00223220_m1), RRM1 (Hs01040698_m1), RRM2 (Hs01072069_m1), CDA

(Hs001560401_m1), dCK (Hs001040726_m1), and human GusB (internal control)

(Hs9999908_m1) were used (Applied Biosystems, Foster City, CA).

35S-Labeled Pulse-Chase Assay. The experiments were performed as described earlier [220], and hENT1 immunoprecipitation was performed using a rabbit anti-hENT1 antibody from Dr. Chung-Ming Tse at the Johns Hopkins School of Medicine (Baltimore,

MD) [221].

Cell Surface Protein Isolation. Isolation of the cell surface fraction of proteins was conducted using the Pierce® Cell Surface Protein Isolation Kit (Thermo Scientific,

Waltham, MA) as per the manufacturer’s instructions. Briefly, cells were washed twice with ice-cold PBS and subsequently biotinylated with 0.25 mg/ml Sulfo-NHS-SS-Biotin for 30 minutes on an orbital shaker at 4°C. The reaction was quenched, and cells were collected by scraping followed by centrifugation. Cells were lysed with 500 µl of Lysis

Buffer containing a protease inhibitor cocktail (Roche) for 30 minutes on ice with intermittent homogenization using a 23G needle and syringe. A sample of the total lysate was retained for analysis of total proteins. The biotinylated proteins were then bound to immobilized streptavidin-agarose beads (NeutrAvidin Agarose slurry) by a one-hour incubation at room temperature using an end-over-end rotator. The unbound proteins were then collected by centrifugation of the column at 1,000 x g for 2 minutes and retained for analysis of intracellular proteins. The biotinylated proteins were eluted form the beads using SDS-PAGE buffer containing 50 mM DTT. The collected samples were then separated on SDS-PAGE for immunoblot analysis.

Immunocytochemical Analysis. Immunostaining analysis was conducted as previously described [219] using the aforementioned antibodies. Immunostained cells were viewed with a Nikon Eclipse Ti fluorescence microscope and analyzed using NIS-

Elements AR 3.0 software.

Statistical Analysis. All experiments were performed in triplicate and repeated at least three times. Statistical significance was identified using the Student’s t test with

*p<0.05 and **p<0.01.

Results

E-cadherin expression increases gemcitabine sensitivity in pancreatic cancer cells. Our previous study [219] categorized a panel of pancreatic cancer cell lines as

Figure 2.1. E-cadherin expression increases gemcitabine sensitivity in pancreatic cancer cells. A. E-cad expression was greater in gemcitabine-sensitive cell lines (L3.6pl, BxPC-1, Capan-1) than gemcitabine-resistant cell lines (AsPC-1, PANC-1, MIA PaCa-2). The E-cad:β-actin signal intensity ratios of each protein band were quantified and normalized to non-cancerous HPDE. Fold change differences are presented above each representative band for comparison. B. Stable retroviral expression of E-cad in MIA PaCa-2 cells. E-cad expression in MIA PaCa-2 cells was localized to the cell periphery (arrows). The E-cad:β-actin ratio was used to compare the fold change in E-cad expression between mock- and E-cad-infected cells. C. Stable expression of E-Cad in MIA PaCa-2 cells increased gemcitabine cytotoxicity. E-cad expression decreased the gemcitabine cytotoxic IC50 by 12-32-fold (p<0.01). Bars, SD. n=3.

gemcitabine-sensitive (L3.6pl, BxPC-3, Capan-1) or -resistant (AsPC-1, PANC-1, MIA

PaCa-2). E-cad protein expression levels were found to correlate with the cell lines’ gemcitabine responsiveness (Fig. 2.1A), prompting us to investigate whether forced overexpression of E-cad in gemcitabine-resistant cell lines would significantly improve gemcitabine sensitivity. We chose the PANC-1 and MIA PaCa-2 cell lines since they are highly resistant to gemcitabine (IC50s >10 μM). Furthermore, PANC-1 expresses low levels of endogenous E-cad, while MIA PaCa-2 has undetectable levels of the cadherin via Western blotting. Stable retroviral gene transfer of E-cad in PANC-1 and MIA PaCa-

2 cells showed increased expression of E-cad protein as judged by Western blotting analysis. Further, E-cad was localized to the cell periphery in both cell lines as determined by immunocytochemical analysis. Data for MIA PaCa-2 were shown in Fig.

2.1B. MTT cytotoxicity analyses indicated that unlike control cells, PANC-1/E-cad and

MIA PaCa-2/E-cad exhibited profound increases in gemcitabine sensitivity (PANC-1 ~4-

8-fold and MIA PaCa-2 ~12-32-fold decrease in IC50) (Fig. 2.1C). These results demonstrate that exogenous E-cad expression can vastly increase gemcitabine sensitivity in normally resistant cells.

E-cadherin increases hENT1 expression and transport activity to influence gemcitabine sensitivity. In order to determine the mechanistic rationale for how E-cad expression increases gemcitabine sensitivity, the influence of E-cad expression on various players of gemcitabine transport, phosphorylation (activation), and catabolism was examined. By Western blotting, it was found that E-cad expression in the drug- resistant MIA PaCa-2 increased hENT1 levels the greatest (8.18-fold) followed by RRM2

(7.45-fold), dCK (6.94-fold), and CDA (1.82-fold); hCNT1 and RRM1 expressions

Figure 2.2. MIA PaCa-2/E-cad showed an increase in hENT1 expression, transport, and –dependent gemcitabine cytotoxicity. A. Compared with mock- infected MIA PaCa-2, MIA PaCa-2/E-cad displayed the highest fold increase in hENT1 protein expression followed by RRM2, dCK, and CDA. The mock to E-cad fold changes averaged from three independent experiments are presented above each representative band for comparison. B. MIA PaCa-2/E-cad displayed a moderate but significant increase in hENT1 mRNA levels (~2.5-fold). Changes in the other candidate genes are also shown. C. MIA PaCa-2/E-cad increased hENT1-mediated 3H- pyrimidine cellular transport. MIA PaCa-2/E-cad cells steeply augmented 3H- gemcitabine and 3H-thymidine cellular transport by hENT1 by 13-fold and 2-fold, respectively, compared with MIA PaCa-2/Mock. D. Pharmacological inhibition of hENT1 significantly reduced gemcitabine cytotoxicity in MIA PaCa-2/E-cad. Treatment of MIA PaCa-2/E-cad cells with 10 nM NBMPR, a selective inhibitor of hENT1, for 72 h reduced gemcitabine cytotoxicity by 5-27-fold (p<0.05). MIA PaCa- 2/Mock treated with 10 nM NBMPR or vehicle alone are shown as controls. Bars, SD. n=3. *p<0.05, **p<0.01.

decreased (0.61- and 0.15-fold, respectively) (Fig. 2.2A). Further investigation into the transcriptional levels of each gemcitabine candidate revealed significantly increased levels of hENT1 mRNA by approximately 2.5-fold (Fig. 2.2B). Significant changes in transcript levels were also noticed for hENT3 (1.926-fold; p<0.05), hCNT1 (0.455-fold; p<0.01), dCK (2.390-fold; p<0.05), and RRM1 (2.098-fold; p<0.05) (Fig. 2.2B). The transport activity of hENT1 was also increased with E-cad expression in MIA PaCa-2 as

3H-gemcitabine and 3H-thymidine cellular transport augmented by approximately 13-fold and 2-fold, respectively, after 10 min (p<0.01) (Fig. 2.2C). Similar results were found with 3H-adenosine uptake (~7-fold increase at 10 min; data not shown). Furthermore, pharmacological inhibition of hENT1 transport activity with 10 nM nitrobenzylthioinosine (NBMPR), a selective inhibitor of hENT1, abrogated the E- cadherin-mediated increase in gemcitabine sensitivity. In MIA PaCa-2/E-cad cells, gemcitabine IC50 was reduced by 5-27-fold (p<0.05) (Fig. 2.2D). MIA PaCa-2/Mock treated with 10 nM NBMPR or vehicle alone were used as controls. While we cannot disregard the effects of E-cad expression on several of the other gemcitabine players, the data suggest that E-cad enhances gemcitabine sensitivity partly through increasing hENT1 expression and activity, leading to increased cellular transport of gemcitabine and, ultimately, cytotoxicity.

E-cadherin increases hENT1 expression, function, and stabilization. In order to investigate the mechanisms by which E-cad expression increases hENT1 activity, we used a variant of a skin carcinoma cell line, A431D, and a subclone which exogenously expresses only E-cadherin, A431DE. Using this cadherin-null model system, we found that E-cad augments, but is not an absolute requirement for, hENT1 expression. While

Figure 2.3. Expression of E-cadherin increased hENT1 expression, function, and stabilization in a clean cadherin model cell line. A. E-cad expression augmented, but was not an absolute requirement for, hENT1 expression. While hENT1 protein was present in the cadherin-null A431D cells, its expression was further increased in the E- cad-expressing A431DE cells. Densitometry fold change differences are presented above each representative band for comparison. B. A431DE cells had increased total uptake as well as uptake by hENT1 of gemcitabine, thymidine, and adenosine compared with A431D cells. In particular, hENT1 uptake of gemcitabine increased over 18-fold with expression of E-cad. C. E-cad expression (red) led to greater hENT1 (green) expression in the cell, and particularly, at the cell surface. The two proteins co- localized at the cell surface (yellow). D. E-cad expression increased hENT1 stability at the cell surface and decreased hENT1 turnover compared with A431D cells. The rate of degradation of radiolabeled hENT1 was highly decreased in cells expressing E-cad.

hENT1 protein was expressed in the cadherin-null A431D cells, its expression was increased with E-cad in the A431DE cells (Fig. 2.3A). E-cad expression also increased total uptake as well as uptake by hENT1 of radiolabeled gemcitabine, thymidine, and adenosine compared with A431D cells (Fig. 2.3B). In particular, uptake by hENT1 increased transport by approximately 19-, 2-, and 3-folds, respectively (Fig. 2.3B). By immunocytochemistry, we confirmed that E-cad (red) was not an absolute requirement for the cell surface expression of hENT1 (green), but rather augmented both total cellular expression levels as well as cell surface expression levels of the transporter (Fig. 2.3C).

In A431D cells, hENT1 is predominantly cytoplasmic; with E-cad expression, greater membrane staining of hENT1 is evident, and the immunoreactivities of these two proteins overlapped at the cell surface (yellow) (Fig. 2.3C). Seeing that hENT1 staining was brighter in A431DE cells compared with A431D, we hypothesized that E-cad may increase the steady-state levels of the transporter either by increasing protein expression or decreasing protein turnover. Since hENT1 is expressed by the CMV promoter, the possibility of alterations in protein synthesis was unlikely, and we focused on changes in hENT1 degradation. Testing this hypothesis, we found that E-cad expression stabilized hENT1 at the cell surface and decreased its turnover. With E-cad expression, the rate of degradation of hENT1 decreased compared with A431D cells. At 14 hours, the approximate half-life of hENT1 [222], nearly 30% of the radiolabeled hENT1 population was degraded in the A431D cells, while only about 80% was degraded in A431DE cells

(Fig. 2.3D).

Discussion

E-cadherin is a cell-cell adhesion molecule that is highly expressed in well- differentiated epithelial cells. It has been highly implicated as a suppressor of EMT and metastasis, a potential biomarker for the progression of cancer, and a key factor in the chemosensitization of cancer cells. However, it has yet to be revealed the precise mechanisms by which E-cad improves chemosensitivity. While we and others [193-198] have found the relationship between E-cad expression and gemcitabine cytotoxicity, the precise mechanisms linking the two remain unclear. Therefore, this study attempts to determine the mechanisms behind E-cad’s influence on gemcitabine efficacy.

Since E-cad expression was found to enhance the cytotoxic response of two poorly-differentiated, drug-resistant pancreatic cancer cell lines to gemcitabine, we hypothesized that E-cad acted on the gemcitabine activation pathway to improve chemosensitivity. Stable expression of E-cad in the E-cad-null MIA PaCa-2 pancreatic cancer cell line identified hENT1 as a key player altered by E-cad expression. Both transcriptional and post-translational increases in hENT1 expression were found. Other genes that were modulated by E-cad and may have an effect on gemcitabine sensitivity included hCNT3, dCK, CDA, RRM1, and RRM2. Further studies are needed to identify whether these additional candidates, as well as other transporters that may be involved or affected, play a significant role in E-cad-mediated gemcitabine efficacy. hENT1 activity was also increased with E-cad expression, further supporting the functional involvement of the transporter in conferring chemosensitivity in MIA PaCa-2. Interestingly, hENT1 uptake of radiolabeled gemcitabine was much greater than that of radiolabeled thymidine.

That is, a moderate increase in protein expression led to a >18-fold increase in

gemcitabine uptake. To explain, previous reports have noted hENT1’s high affinity for gemcitabine (Km = 160 μM) [223, 224] and lower affinity for thymidine (Km = 240 μM)

[225]. While this property may contribute to the difference in substrate uptake, it may also be possible that expression of E-cad can change the parameters of the hENT1, although further studies are warranted.

Further mechanistic studies in a cadherin-null cell line, A431D, revealed that the increase in hENT1 transport activity was due to increased total cell hENT1 content as well as cell surface hENT1 expression concomitant to decreased protein degradation. It was also found that E-cadherin was not an absolute requirement for either hENT1 expression in general or hENT1 expression at the cell surface; instead, the cadherin simply augmented both those aspects. At this point, it remains unclear whether E-cad expression directly increases the recruitment, trafficking, and basolateral membrane insertion of hENT1. It is also unknown whether the lysosomal or proteasomal degradation pathway is inhibited via E-cad, although Nivillac et al. has shown the predominant involvement of lysosomes in normal hENT1 degradation [222].

Furthermore, whether E-cad prevents global turnover of membrane proteins needs to be investigated. For example, proteasomal inhibitors, such as the ground-breaking

Bortezomib, are currently used in clinics [226-228], and lysosomal inhibitors are becoming increasingly under study as anticancer agents [229-231].

This study suggests that E-cad enhancement of gemcitabine efficacy occurs at least in part through increased hENT1 expression, stability, and activity, leading to increased cellular transport and cytotoxicity of gemcitabine. While hENT1 is already employed as a biomarker for staging of differentiation and prediction of patient survival

[10-14], E-cad can be used as a surrogate marker or as part of a biomarker index to indicate preliminary patient response to nucleoside analog therapeutics, guiding subsequent treatment and avoiding unnecessary nucleoside analog toxicities. E-cad could also be used for drug targeting. Small molecule agents that can directly or indirectly induce E-cad expression could be therapeutically beneficial. For example, since E-cad is epigenetically silenced in pancreatic cancer, 5-aza-dC, a DNA methylation inhibitor, has been found to increase E-cad expression and even synergize with gemcitabine [126, 129,

232-238], suggesting that 5-aza-dC action may in part be mediated through hENT1 as found in this study.

In a pancreatic cancer system, numerous cadherins are expressed. Currently, most of their roles in conferring chemoresistance remain unknown, and it is possible that they are able to compensate for the loss of E-cad. It is also unknown whether N-cadherin, which is highly implicated in EMT and metastasis, plays an opposing role in hENT1- mediated gemcitabine efficacy. Likewise, there are numerous transporters involved in gemcitabine uptake. In addition to hENT1, we identified significant changes in hENT3 and hCNT1 transcript levels with E-cad expression. In a previous study, we found that hCNT1 correlates with differentiation and influences chemosensitivity [9]. Nonetheless, this study demonstrates that a cell adhesion molecule can induce transcription and stabilize proteins at the cell surface in order to improve chemotherapeutic drug sensitivity.

CHAPTER 3

CO-EXPRESSION OF HCNT1 WITH CX32 INCREASES NUCLEOSIDE ANALOG

TRANSPORT 3

3 SW Hung, K Proctor, M Krentz, H Lee, D Lovin, and R Govindarajan. To be submitted to Experimental Cell Research. 59

Abstract

Gemcitabine is a pyrimidine analog drug commonly used for the chemotherapeutic treatment of cancer. Due to its hydrophilicity, gemcitabine requires nucleoside transporters (NTs) for its movement across the cell membrane and entry into the cell. Gap junctions (GJs), composed of channel proteins called connexins (Cxs), have the ability to potentiate gemcitabine action by transferring the chemotherapeutic agent between normal and cancer cells, leading to a phenomenon known as the bystander effect. In order to identify the roles and interplay of NTs and GJs in affecting the distribution and efficacy of gemcitabine, we studied the contributions of each human equilibrative nucleoside transporter 1 (hENT1), human concentrative nucleoside transporter 1 (hCNT1), connexin 32 (Cx32), and connexin 43 (Cx43) both individually and in combinations. Using polarized MDCK cells as a model system, we found that

Cx32 operated in concert with hCNT1 to increase nucleoside uptake, while Cx43 with hENT1 decreased both. To relate this to pancreatic cancer, we found hENT1 and Cx43 to be overexpressed in a panel of pancreatic cancer cell lines known to be more chemoresistant than the normal pancreatic cell line. From these data, a working model for the intra- and intercellular transport of nucleosides in cancer chemotherapy was developed and the NT-GJ interplay in pancreatic cancer cells predicted. This study delves into the mechanism of action of nucleoside-analog chemotherapeutics as well as explores the potential for improvements in drug action.

Introduction

Nucleoside transporters (NTs), a family of transmembrane proteins at the cell surface, have diverse functions in the cell including carrying nucleoside substrates for nucleic acid synthesis [239]. Depending on the type of NT, they are also responsible for the movement of nucleoside-analog chemotherapeutic drugs in and out of cells. Two classes of NTs have been identified: the human equilibrative nucleoside transporters

(hENTs; SLC29) and the human concentrative nucleoside transporters (hCNTs; SLC28). hENTs operate bidirectionally and independent of Na+, while hCNTs operate unidirectionally and dependent of Na+. The hENT family consists of 4 members (hENT1-

4), and the hCNT family consists of 3 members (hCNT1-3). For gemcitabine, a standard chemotherapeutic of choice for pancreatic cancer, hENT1 and hCNT1 are the two most efficient transporters. Since nucleoside analog drugs are hydrophilic and rely on NTs for entry into cells, the transporters determine the drug’s selectivity, directionality, and rate of transport across the lipid bilayer [239].

Gap junctions (GJs) are the only intercellular channels that connect the cytoplasms of adjacent cells, allowing for the free passage of ions and small molecules between cells, including nucleosides and their analogs. GJs are composed of an assemblage of channel proteins, called connexins (Cxs), at the cell membrane and enable direct cell-cell communication [240]. Cxs can also assemble into large pore complexes at the cell surface, called hemichannels. The type of Cx confers certain properties to the GJ or hemichannel; for example, it has been recently shown that connexins 32 (Cx32) and 43

(Cx43) demonstrate permselectivity of adenosine (a nucleoside) and ATP (a nucleotide), respectively [241].

With the transport of nucleoside analog drugs into and between cells, there holds the potential for increasing cytotoxicity within a heterogeneous tumor cell population, a phenomenon termed as the “bystander effect.” The bystander effect (BE) was originally established in conjunction with prodrug-activating systems in suicide gene therapy [242,

243]. A BE occurs when a prodrug not only kills the tumor cells in which it has entered, but also neighboring tumor cells which did not directly take up the drug. In the case of gene therapy, this powerful event has been shown to produce therapeutically significant results with even only 1-2% of cells genetically modified [244]. In general, cell-cell contacts are not required for the BE to occur either in vitro or in vivo. However, purine and pyrimidine nucleosides and their metabolites are unable to passively diffuse across cell membranes, so cell-cell connections, specifically gap junctions, are required to elicit a BE. It has also been shown that pharmacological manipulation of gap junction expression and communication may be used to alter the BE [245]. Although there seems to be potential for eliciting a BE with nucleoside analog drug therapy, the precise roles of

NTs and GJs and their interplay in determining cellular chemosensitivity remain to be elucidated.

With 7 different types of NTs and over 20 types of connexins, the functions and expressions of each varies between normal and cancerous cells as well as within a range of cancer stages, leading to complex heterogeneity. Currently, it is known that hENT expressions increase with a variety of cancers, including gynecological carcinomas

(ovarian, uterine, cervix, and endometrial [246]) and mantle cell lymphoma [247] among several others. Their expressions have also been found to correlate with gemcitabine sensitivity and patient survival [247], particularly in pancreatic cancer [14, 248]. Contrary

to hENTs, hCNT expressions are decreased in several cancers [246]. For example, hCNT1 transcript and protein levels are significantly decreased in both pancreatic cancer cell lines as well as matched pair tissue samples as shown in our recently published paper

[9]. Earlier work conducted by Dr. Govindarajan using a spontaneous epithelial-to- mesenchymal transition (EMT) model (i.e., rat liver cells), showed that while Cx43 protein is expressed equally in both early and late stage EMT, the GJ protein was completely internalized during the late stage [249]. These results suggest the inactivation of Cx43 GJs in metastatic cancer cells that have undergone EMT.

In this study, we identified the roles and interplay of NTs and GJs in affecting the distribution and efficacy of the nucleoside analog drug, gemcitabine. We studied the contributions of hENT1 and hCNT1 because they are known to be the most efficient transporters of gemcitabine. We also chose to study Cx32 and Cx43 since Cx32 can transport gemcitabine in its inactive nucleoside form, and Cx43 can transport gemcitabine-triphosphate in its active nucleotide form. We hypothesized that the manipulation of these naturally occurring cellular channels and transporters will increase nucleoside drug concentrations in heterogeneous tumor cell populations, resulting in greater chemotherapeutic efficacy. By expressing these proteins both individually and in combinations, we hoped to better understand the mechanism of action of nucleoside analog chemotherapeutics as well as explore the potential for improvements in drug action.

Materials and Methods

Materials. Gemcitabine and 3H-thymidine were obtained from Chemie Tek

(Indianapolis, IN). FBS was acquired from PAA Laboratories Inc. (Etobicoke, Ontario).

Uridine, DAPI, DMSO, nitrobenzyl mercaptopurine riboside (NBMPR), and 3-(4,5- dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) were obtained from

Sigma-Aldrich (St. Louis, MO). All secondary antibodies were purchased from Bethyl

(Montgomery, TX). The bicinchoninic acid (BCA) protein assay reagent was from Pierce

Chemical (Rockford, IL). The fluorescent anti-fade mounting reagent and Vybrant

DyeCycle green were both obtained from Molecular Probes (Invitrogen). All plastic wares for cell culture were obtained from Corning, NY.

Cell Culture. MDCK and all pancreatic cell lines were obtained and cultured as previously described [9, 190]. HPAF-II was kindly provided by Dr. Parmender Mehta of the University of Nebraska Medical Center (Omaha, NE), grown and maintained in MEM supplemented with 10% FBS, and subcultured at a 1:4 ratio.

Expression of Nucleoside Transporters and Connexins in MDCK. Cells stably expressing each NT or Cx or both were generated by retroviral transduction of respective cDNAs [9]. Multiple clones expressing the desired protein(s) were selected with and maintained in 800 and 200 µg/ml of G418 antibiotic, respectively. Final clones for subsequent experiments were determined based on a preliminary functional transport assay (hENT1 Cl-2, hCNT1 Cl-3, Cx32 Cl-4, and Cx43 Cl-4 were chosen; data not shown).

Immunocytochemical Analysis. Immunostaining was conducted as previously described [9]. Antibodies were the same as those used for Western blotting with the

exception of anti-hENT1 (rabbit) which was a kind donation from Dr. Chung-Ming Tse of the Johns Hopkins University School of Medicine (Baltimore, MD). Immunostained cells were viewed with a Nikon TM Eclipse fluorescence microscope and analyzed using

Nikon TiE software.

Western Blotting. Western blotting was conducted as previously described [9].

Anti-human Cx43 (rabbit), anti-hENT1 (goat), and anti-hCNT1 (rabbit) antibodies were used (Santa Cruz Biotechnology; Santa Cruz, CA). The anti-human Cx32 (mouse) antibody was a generous gift from Dr. Mehta. The monoclonal mouse anti-human β-actin antibody was from Sigma-Aldrich.

MTT Cytotoxicity Assay. The MTT assays were performed as described earlier

[9]. Briefly, 1x103 cells/well in a 96-well plate were treated 36 h after seeding with increasing concentrations of gemcitabine (0.1 nM-100 μM). After 72 h, the percent viable cells were quantified using an MTT assay. Cytotoxic IC50 values were calculated using

GraphPad Prism 5 software.

3H-Nucleoside Transport Study. The transport assays were performed as previously described [221]. Briefly, cells were seeded in 24-well culture plates (5x104 cells/well) and cultured until cells reached 90-95% confluency. After aspirating the media, cells were incubated with sodium buffer (pH 7.4) [221] containing 0.02 µM of 3H- thymidine in the presence or absence of 10 nM NBMPR. Diffusional uptake was measure in the presence of 10 µM NBMPR and 20 mM uridine. After 5 min, permeant fluxes were terminated using 20 mM cold uridine. Cells were then lysed with 1% SDS and agitated in a cell shaker for 30 min. Each well’s lysate was added to 5 mL of ScintiSafe

Econo 1 (Fisher Scientific). The mixture was vortexed and counted using an LS 6500

Multipurpose Scintillation Counter. A parallel cultured well was subjected for protein estimation.

Statistical Analysis. All experiments were performed in triplicate and repeated at least three times. Statistical significance was identified using the Student’s t test with

*p<0.05 and **p<0.01.

Results

Stable expression of NTs or Cxs in MDCK cells increased total 3H-thymidine cellular transport and gemcitabine cytotoxicity. Using retroviral infection, we created

NT- and Cx-null MDCK cells stably expressing only hENT1, hCNT1, Cx32, or Cx43.

Expression of the desired protein was confirmed using immunocytochemistry. hENT1 was seen at the basolateral membrane, hCNT1 at the apical membrane, and the Cxs at the cell-to-cell contacts (Fig. 3.1A). None of the four proteins were seen in the control

MDCK cells (Fig. 3.1A).

Function of the expressed proteins was validated using a radiolabeled transport assay. 3H-thymidine uptake was found to be significantly increased in the cells expressing hENT1, hCNT1, Cx32, or Cx43 compared with control (Fig. 3.1B). As expected, hCNT1 expression led to the greatest nucleoside uptake, followed by hENT1,

Cx43, and Cx32, correspondingly (Fig. 3.1B). As a consequence of this increase, MDCK cells with hENT1 or hCNT1 expression also had greater sensitivity towards gemcitabine

(Fig. 3.1C). The cytotoxic IC50 values of the NT-expressing cells were 5-6-fold less than the control cells (Fig. 3.1C).

Figure 3.1. Stable expression of NTs and Cxs in MDCK cells increased total 3H- thymidine cellular transport and gemcitabine cytotoxicity. A. Stable retroviral expression of hENT1 (green, top), hCNT1 (red, top), Cx32 (green, bottom), and Cx43 (red, bottom) in MDCK cells. Immunostaining analysis showed strong hENT1, hCNT1, Cx32, and Cx43 expression in the respective infected cells while the proteins are barely visible in the WT cells. DAPI, blue. B. Expression of the NTs and Cxs in MDCK cells increased total radiolabeled thymidine uptake compared with control MDCK. C. Expression of hENT1 and hCNT1 augmented gemcitabine chemosensitivity by decreasing the cytotoxic IC50 approximately 6- and 5-fold, respectively. IC50 values are indicated. *p<0.05, **p<0.01.

Expression of hCNT1 with Cx32 synergistically increased nucleoside uptake, while hENT1 with Cx43 antagonistically decreased it, in MDCK cells. To determine

whether expressing a combination of the NTs and Cxs would produce an even greater

Figure 3.2. Stable expression of a combination of NTs and Cxs in MDCK cells altered total 3H-thymidine cellular transport. A. Stable retroviral expression of a combination of hENT1 or hCNT1 (green) with Cx32 or Cx43 (red) in MDCK cells. Immunostaining analysis showed hENT1 and hCNT1 protein expression predominantly at the cell periphery, while Cx32 and Cx43 were located at cell-cell contacts. B. Expression of the NT/Cx combinations altered total 3H-thymidine cellular transport. Cx32 expression with hCNT1 increased total 3H-thymidine uptake, while Cx43 with hENT1 decreased total uptake compared with expression of the NT alone. Although expression of hCNT1 with Cx43 and hENT1 with Cx32 also altered uptake, changes were not significant. *p<0.05, **p<0.01.

effect, we created MDCK cells that express one NT and one Cx (i.e., hENT1 + Cx32, hENT1 + Cx43, hCNT1 + Cx32, and hCNT1 + Cx43). Expression of the desired proteins was confirmed using immunocytochemistry. After merging the immunostained images, it is clear that the NTs and Cxs were both localized predominantly at the cell-to-cell contacts (yellow; Fig. 3.2A). Out of the combinations, hCNT1 with Cx32 produced an even greater, significant increase in 3H-thymidine uptake compared with hCNT1 expression alone (Fig. 3.2B). Interestingly, expression of hENT1 with Cx43 produced a significantly decreased level of 3H-thymidine uptake compared with the NT alone (Fig.

3.2B).

Overexpression of hENT1 and Cx43 in most pancreatic cancer cell lines. To translate the preliminary, model-system results to pancreatic cancer, the expressions of

Figure 3.3. Statuses of each hENT1, hCNT1, Cx32, and Cx43 were identified in a panel of pancreatic cell lines. Most of the pancreatic cancer cell lines overexpressed hENT1 and Cx43 compared with the normal human pancreatic ductal epithelial (HPDE) cell line. hCNT1 and Cx32 expressions were also noticed, but at a similar or lower level compared with HPDE. Western blotting analysis with 75 μg of total cell lysates was conducted.

hENT1, hCNT1, Cx32, and Cx43 were investigated in a panel of pancreatic cell lines.

Compared with the normal human pancreatic ductal epithelial (HPDE) cell line, most of the cancerous cells showed greater expression of hENT1 and Cx43 (Fig. 3.3). Three of the cell lines (L3.6pl, Capan-1, and PANC-1) also displayed both hCNT1 and Cx32 expressions (Fig. 3.3).

Proposed model of nucleoside and nucleotide movements within a cell population via NTs and Cxs. Based on the experimental evidence obtained with various transduced cell line combinations, we propose the following model for nucleoside analog handling by Cxs and NTs in tumor cells. Co-expression of hCNT1 and Cx32 leads to increased nucleoside uptake due to the rapid influx of nucleosides into cells via the high-affinity hCNT1 (high Km) and Cx32 hemichannels at the cell surface (Fig. 3.4). With a high

Figure 3.4. Proposed model of intracellular and intercellular nucleoside transport and movement within a cell population. Expression of hCNT1 with Cx32 increases nucleoside uptake due to the rapid transport of nucleosides into cells via Cx32 hemichannels and the unidirectional hCNT1. Rapid uptake causes less phosphorylation inside cells, and nucleosides are able to be transported to neighboring cells via Cx32 GJs. Expression of hENT1 with Cx43 decreases uptake due to the slow transport of nucleosides into cells by the bi-directional hENT1 alone. Slow uptake allows for more phosphorylation, and nucleotides are released from the cell via Cx43 hemichannels. Only a few nucleotides are transferred to neighboring cells by Cx43 GJs.

maximum velocity (Vmax) of the transporters, an abundance of nucleosides enter the cell, and the ability of kinases to phosphorylate the nucleosides into their active form becomes slow and saturated, having achieved their Vmax. The remaining un-phosphorylated nucleosides are then able to be transported to neighboring cells via Cx32 GJs, propelling the bystander effect. When hENT1 and Cx43 hemichannels are expressed at the cell surface, nucleoside uptake is less due to low transporter affinity (low km) and slow influx of nucleosides (Fig. 3.4). Furthermore, while Cx43 hemichannels are unable to transport nucleosides into the cell, hENT1 is bi-directional, and nucleosides are able to both enter and exit the cell. The slow influx of nucleosides causes most to be phosphorylated quickly into their active form since kinases are not reaching their Vmax. Once converted into nucleotides, they are then able to be exported from the cell by Cx43 hemichannels before being transferred to neighboring cells, hindering the bystander effect. Only a few nucleotides are transported to neighboring cells via Cx43 GJs.

Discussion

With the transport of nucleoside analog drugs into and between cells via NTs and

GJs, there holds potential for increasing the bystander effect. Therefore, this study attempts to identify the roles and interplay of NTs and GJs in affecting the distribution and efficacy of the nucleoside analog drug, gemcitabine. We hypothesized that the manipulation of these naturally occurring cellular channels and transporters will increase nucleoside drug concentrations in heterogeneous tumor cell populations, resulting in greater chemotherapeutic efficacy. Among the numerous different types of NTs and Cxs, we chose to study hCNT1, hENT1, Cx32, and Cx43. They are the prototypic candidates

in each respective family, and their expressions were confirmed in normal pancreatic tissues [31, 250, 251]. Furthermore, high-affinity hCNT1 and hENT1 have been shown to be key factors in determining gemcitabine efficacy in pancreatic cancer [9, 14, 246-248], while Cx32 and Cx43 show promise in potentiating their mechanism [241, 249]. By expressing these proteins both individually and in combinations, we aimed to elucidate the movements of nucleoside analog chemotherapeutics within tumor cell populations as well as explore the potential for improvements in drug action.

Since pancreatic cancer cells express one or more of the NTs or Cxs of interest, we first searched for a cell system that is deficient in these proteins. Our search identified

MDCK cells to possess only minimal levels of endogenous NTs or Cxs. Hence, we used the cell line to conduct mechanistic studies on the individual effects of the NTs and Cxs of interest on the distribution of gemcitabine within cell populations before advancing to pancreatic cancer cell models.

As expected, expression of the unidirectional, high-affinity, and rapid hCNT1 led to the greatest uptake of radiolabeled nucleosides followed by the bidirectional, lower- affinity, and slower hENT1. Consequences of this increased uptake were validated when expression of hENT1 and hCNT1 led to a 6- and 5-fold increase in gemcitabine cytotoxicity (leading to an IC50 of 13.87 and 17.61 μM), respectively, compared with control cells. Increased uptake with Cx32 and Cx43 expressions may be due to increased hemichannel transport or increased movement of radiolabeled thymidine into cells as the nucleoside is better distributed throughout the cell population by GJs.

Co-expression of hCNT1 with Cx32 led to a synergistically increased level of nucleoside transport, suggesting that since Cx32 preferentially transfers

unphosphorylated molecules (nucleosides), it is acting in cooperation with the unidirectional hCNT1 to increase uptake of the gemcitabine prodrug. Once entered, gemcitabine is phosphorylated and unable to exit the cell through Cx32 hemichannels.

On the other hand, cells expressing hENT1 and Cx43 displayed an antagonistically decreased level of nucleoside transport. Cx43 hemichannels, preferring phosphorylated molecules, does not aid in the uptake of gemcitabine but rather the expulsion of its phosphorylated (activated) forms from the cell.

This study provides the first evidence that there is an interplay between NTs and

Cxs. Between expressing hCNT1/Cx32 and hENT1/Cx43, there was a >10-fold difference in nucleoside uptake in our model cell line. Therefore, it is possible that nucleoside analog cytotoxicity is regulated by both NTs and Cxs (GJs and hemichannels) and that their favorable manipulations may provide a mechanism for improving gemcitabine chemosensitivity in pancreatic cancer cells.

There is a large number of possible combinations involving NT and Cx expressions, and hence the observed difference in nucleoside uptake in reality could far exceed the estimated value. Further studies are needed to determine which specific types and combinations of NTs and Cxs are most effective in heterogeneous cell populations.

For example, it was recently shown that Cx26 significantly contributes to the gemcitabine bystander effect in pancreatic cancer [174].

In preliminary studies with our cell culture system, we identified some NT/Cx combinations to produce increased chemosensitivity even with only a slight increase in transport, whereas other combinations did not improve chemosensitivity even with high levels of transport. This is likely due to the properties of the transporters and channel

proteins as explained by our hypothetical model. In an in vivo context, a small increase in drug uptake may actually substantially increase cytotoxicity due to the BE in a 3D model.

In future studies, co-culture (i.e., using mixed populations of cells expressing various

NTs and GJs), 3D culture (i.e., cell culture utilizing a 3D matrix), and animal tumors would need to be employed to mimic the heterogeneity of tumors and for testing the potential BE in more translational models.

The proposed model of intracellular and intercellular nucleoside transport could be used to explain current drug exposure limitations in heterogeneous tumor cells and how they can be manipulated to overcome chemoresistance in pancreatic cancer. This understanding could bring about diagnostic value and therapeutic guidance (e.g., drug selection) in patients. Overall, in contrast to the conventional paradigm suggesting that all transporters and channel proteins aid in drug uptake and chemosensitivity, these findings for the first time shed light on the remarkable functional selectivities of NTs and Cxs, which can be utilized for improved cancer treatments and predictions.

CANCER STEM CELLS AND CHEMOSENSITIVITY

CHAPTER 4

PHARMACOLOGICAL REVERSAL OF HISTONE METHYLATION

PRESENSITIZES PANCREATIC CANCER CELLS TO NUCLEOSIDE DRUGS: IN

VITRO OPTIMIZATION AND NOVEL NANOPARTICLE DELIVERY STUDIES 4

4 SW Hung, H Mody, S Marrache, YD Bhutia, F Davis, JH Cho, J Zastre, S Dhar, CK Chu, and R Govindarajan. 2013. PLOS ONE. 8(8): e71196. Reprinted here with permission of the publisher. 76

Abstract

We evaluated the potential of an investigational histone methylation reversal agent, 3-deazaneplanocin A (DZNep), in improving the chemosensitivity of pancreatic cancer to nucleoside analogs (i.e., gemcitabine). DZNep brought delayed but selective cytotoxicity to pancreatic cancer cells without affecting normal human pancreatic ductal epithelial (HPDE) cells. Co-exposure of DZNep and gemcitabine induced cytotoxic additivity or synergism in both well- and poorly-differentiated pancreatic cell lines by increased apoptosis. In contrast, DZNep exerted antagonism with gemcitabine against

HPDE cells with significant reduction in cytotoxicity compared with the gemcitabine- alone regimen. DZNep marginally depended on purine nucleoside transporters for its cytotoxicity, but the transport dependence was circumvented by acyl derivatization. Drug exposure studies revealed that a short priming with DZNep followed by gemcitabine treatment rather than co-treatment of both agents to produce a maximal chemosensitization response in both gemcitabine-sensitive and gemcitabine-resistant pancreatic cancer cells. DZNep rapidly and reversibly decreased trimethylation of histone

H3 lysine 27 but increased trimethylation of lysine 9 in an EZH2- and JMJD1A/2C- dependent manner, respectively. However, DZNep potentiation of nucleoside analog chemosensitization was found to be temporally coupled to trimethylation changes in lysine 27 and not lysine 9. Polymeric nanoparticles engineered to chronologically release

DZNep followed by gemcitabine produced pronounced chemosensitization and dose- lowering effects. Together, our results identify that an optimized DZNep exposure can presensitize pancreatic cancer cells to anticancer nucleoside analogs through the reversal

of histone methylation, emphasizing the promising clinical utilities of epigenetic reversal agents in future pancreatic cancer combination therapies.

Introduction

Polycomb group proteins (PcGs) can remodel chromatin by influencing the degree of compaction, leading to epigenetic gene silencing. Polycomb Repressive

Complex 2 (PRC2), one of the two classes of PcGs, induces histone methyltransferase activity primarily by trimethylating histone H3 at lysine 27 (H3K27me3), mediating silencing of tumor suppressor genes. The catalytic subunit of PRC2 is Enhancer of Zeste

Homolog 2 (EZH2), in which the SET domain constitutes the active site for histone

H3K27 methylation [252]. Studies support EZH2 as a key player in the development and progression of tumors due to its ability to alter gene expressions including those involved in cell cycle control, cell migration, and DNA repair [253]. EZH2 is crucial in the chromatin control of genetic reprogramming of cancer stem cell self-renewal and differentiation that have been implicated in chemoresistance [254-257].

As a marker of advanced and metastatic disease in many solid tumors, EZH2 overexpression has been reported in pancreatic cancers, particularly those that are poorly differentiated [257, 258]. EZH2 was found to be upregulated by oncogenic RAS through

MEK-ERK signaling, leading to the downregulation of tumor suppressors such as

RUNX3 and p27 (Kip1) [257, 259, 260]. EZH2 depletion led to cell cycle arrest at the

G1/S transition, suggesting the protein may repress the tumor suppressing p27 gene

[261]. Similarly, knockdown of EZH2 resulted in a significant decrease in cellular proliferation and invasiveness [257, 258, 262] and sensitized pancreatic cancer cells to

doxorubicin and gemcitabine, revealing the potential of an EZH2 inhibitor- chemotherapeutic combination therapy [257]. In vivo, suppressing EZH2 diminished tumorigenicity and inhibited pancreatic cancer metastasis [258]. Clinically, positive correlations have been observed between EZH2 expression and advanced pancreatic cancer stage and grade in patients [262]. In many cases, high levels of EZH2 in cancer were also significantly associated with decreased E-cadherin expression and highly aggressive disease. In gemcitabine-treated patients, significantly longer survival was observed in patients with low rather than high EZH2 expression [263]. Consequently,

EZH2 may be a significant prognostic value for overall survival in pancreatic cancer patients [264].

Recently, it has been shown that a potent chemical inhibitor of S- adenosylhomocysteine hydrolase, 3-deazaneplanocin A (DZNep), modulates chromatin through indirect (i.e., reducing methyl group availability) inhibition of histone methyltransferases including EZH2 [265, 266]. DZNep, a carbocyclic analog of adenosine, depletes cellular levels of the PRC2 components while inhibiting the associated H3K27me3 [267]. While the mechanisms and effects of DZNep have been studied in numerous solid tumors and leukemia [253, 254, 265, 266, 268-272], less is known about the potential of this compound for pancreatic cancer treatment.

Nevertheless, its current potential for reducing EZH2 levels, reverting epithelial-to- mesenchymal transition (EMT), and preventing tumor progression, makes it a highly promising antimetastatic agent [256]. The therapeutic potential of DZNep in combination with other agents, such as polyphenols and histone deacetylase inhibitors, has begun to emerge with encouraging results [265, 266]. Since increasing evidence suggests that

future cancer therapies will take advantage of the synergistic effects achieved from different combinations of epigenetic reversal and conventional antitumor agents [273], we investigated the potential of the DZNep-gemcitabine combination for improving anticancer activity in pancreatic cancer. Our results identified that histone methylation reversal by DZNep presensitizes pancreatic cancer cells to gemcitabine. Optimization of the drug combination through dosage and delivery methods was further conducted.

Materials and Methods

Reagents. Radiolabeled (3H) gemcitabine, adenosine, thymidine, and guanosine were obtained from Moravek Biochemicals and Radiochemicals (Brea, CA), while cold gemcitabine was from ChemieTek (Indianapolis, IN). Adenosine and guanosine were kindly provided by Dr. Chung K. Chu (University of Georgia). Thymidine, uridine, nitrobenzyl mercaptopurine riboside (NBMPR), and 3-(4,5-dimethylthiazol-2-yl)-2,5- diphenyltetrazolium bromide (MTT) were obtained from Sigma-Aldrich (St. Louis, MO).

Dimethylsulfoxide (DMSO) was purchased from Macron Chemicals (Center Valley, PA), and the bicinchoninic acid (BCA) protein assay reagent was from Thermo Scientific

Pierce (Rockford, IL). Plastic wares for cell culture were obtained from Corning

(Corning, NY).

Cell Culture. The pancreatic cancer cell lines (AsPC-1, BxPC-3, Capan-1, MIA

PaCa- 2, and PANC-1) and MCF-10A cells were received from the American Type

Culture Collection (ATCC; Manassas, VA) cell bank. These cell lines were propagated, expanded, and frozen immediately after arrival. The cells revived from the frozen stock were used within 10-20 passages, not exceeding a period of 2-3 months. The ATCC uses

morphological, cytogenetic and DNA profile analysis for characterization of cell lines.

Human pancreatic ductal epithelial (HPDE) cells [216] were kindly received from Dr.

Ming Tsao of the Ontario Cancer Institute (Toronto, Canada). The L3.6pl cell line [274] was kindly received from Dr. Isiah D. Fidler at The University of Texas MD Anderson

Cancer Center (Houston, TX). The HPDE and L3.6pl cell lines were handled as other cell lines and were genotyped by DNA fingerprinting (PowerPlex 16, Promega, Inc.) as per the manufacturer’s instructions. The 293T cell line was kindly received from Dr. J.

Michael Thomson of The University of Georgia (Athens, GA). The growth conditions of cell lines were performed as described previously [275].

MTT Cytotoxicity Assay. Cells were seeded at a density of 3x103 cells/well in a

96-well microtiter plate and grown to 90-95% confluency. After treatment, 50 μl of MTT solution (5 mg/ml in PBS) were added to each well, and the plates were incubated for 2 h.

MTT formazan crystals were dissolved in 100 μl/well of DMSO by shaking the plates on a rocking platform. A 96-well scanner was used to measure the spectrophotometric absorbance at 490 nm. The absorbance at 650 nm was used for background subtraction.

The 50% inhibitory concentration (IC50) was determined using GraphPad Prism 5.0 software.

Caspase 3 Assay. Apoptosis was measured using the Fluorimetric Caspase 3

Assay Kit from Sigma-Aldrich (Cat. No. CASP3F) as per the manufacturer’s instructions.

Briefly, 104 cells/well in a 96-well plate were treated with DZNep and/or gemcitabine for

72 h. Cells were then lysed and incubated on ice for 15-20 min. After adding assay buffer containing the Ac-DEVD-AMC substrate, samples were transferred to a black 96-well plate, and fluorescence was read every 10 min for 1 h at room temperature (360 nm

excitation, 460 nm emission). The appropriate blank (reaction mixture), positive control

(caspase 3), and negative control (caspase 3 + caspase 3 inhibitor) were also conducted.

Drug Interaction Studies. The combination index plots for DZNep and gemcitabine were estimated using the method developed by Chou and Talalay and the

CalcuSyn software [276].

Nucleoside Uptake in Cells and Xenopus Oocytes. These procedures were performed as previously described [221, 277].

Generation of Xenopus Oocyte Expression Constructs. The full-length IMAGE cDNA clones of the transporters (hENT1: clone ID 3010092, accession BC008954; hENT2: clone ID 9051840, accession BC143335; hCNT1: clone ID 8991920, accession

BC 126204; hCNT3: clone ID 7939668, accession BC093823) were obtained from Open

Biosystems (Huntsville, AL). Subcloning of the genes into the Xenopus oocyte expression vector, pOX, was completed using the primer sets designated in Table 4.1.

Table 4.1. Restriction sites and sequences of primers used for cloning. Gene Forward Reverse Restriction Sequence Restriction Sequence Site Site pOX-hENT1 SalI 5’-GCACGTCGACCAATA XbaI 5’-CGTGTTCTAGATC ATGACAACCAGTCAC-3’ ACACAATTGCCCG-3’ pOX-hENT2 SalI 5’-GATATAGTCGACCAAT XbaI 5’-CTATCTCTAGATCAGA AATGGCGCGAGGAGAC-3’ GCAGCGCCTTGAAGAG-3’ pOX-hCNT2 HindIII 5’-GCCCGAAGCTTCAATAA XbaI 5’-CGTCGTCTAGATTAG TGGAGAAAGCAAGTGG-3’ GCACAGACGGTATTG-3’ pOX-hCNT3 HindIII 5’-CGCACAAGCTTCAAT SpeI 5’-CCGCGCACTAGTTCAA AATGGAGCTGAGGAG-3’ AATGTATTAGAGATCCC-3’

Western Blotting. Western blotting was conducted as previously described [275].

The rabbit polyclonal anti-histone H3K4TM, H3K9MM, H3K9DM, H3K27MM,

H3K27DM, H4K20DM, and H4K20TM were from Millipore (Billerica, MA), as well as the rabbit polyclonal anti-EED antibody. Also obtained from Millipore were the mouse

monoclonal anti-histone H3K9TM, H3K27TM, and H4K20MM antibodies, as well as the mouse monoclonal anti-EZH2 (clone BD43) and anti-SUZ12 (clone 3C1.2) antibodies.

The rabbit polyclonal anti-JMJD1A and anti-JMJD2C antibodies were acquired from

Sigma-Aldrich. The mouse monoclonal anti-β-actin antibody was also purchased from

Sigma-Alrich, and the HRP-conjugated secondary antibodies were from Bethyl

Laboratories (Montgomery, TX).

Synthesis of Parent Nucleosides and Derivatives. Troxacitabine and its lipophilic prodrug, as previously described (compound 2K [278]; compound 6h [279];

C24H41N3O5), were obtained from Dr. Chung K. Chu (University of Georgia). The synthesis of a DZNep analog began from D-ribose. D-ribose was treated with 2,2- dimethoxypropane in the presence of a catalytic amount of p-toluenesulfonic acid to give a isopropylidine derivative, followed by the protection of the primary alcohol with triphenylmethyl chloride. The protected lactol was then reacted with vinyl magnesium bromide to give a single diastereomericdiol, which was subsequently protected with

TBDMS only at the allylic hydroxyl position to afford silyldienol. To incorporate another double bond for the ring-closing metathesis reaction, the protected secondary alcohol was oxidized to a ketone by Swern oxidation, followed by a Wittig reaction with methyltriphenylphosphonium bromide and n-BuLi to provide the diene. The silyl group from the diene was removed with TBAF to give a less sterically demanding dienol, which was then converted to cyclopentenol with a second-generation Grubbs catalyst in good yield. In order to carry out the Mitsunobu coupling, bis-Boc-3-deaza-adenine was prepared. Mitsunobu coupling provided the desired N-9 isomer as a major product.

Removal of protection groups using 2N-HCl in methanol produced DZNep. Three

alcohol groups of DZNep were protected with TBS which was substituted to the two

6 amide prodrugs (carbon chains of C6 to C8) by acylation of the N -NH2 group. The compound was finally fully deprotected by TBAF to give the DZNep prodrug 1

(C20H29ClN4O4) and DZNep Prodrug 2 (C18H24N4O4). The syntheses procedures and physicochemical characterizations of the compounds will be published elsewhere.

Nanoparticle Formulation. All chemicals were purchased from Sigma Aldrich and used without further purification unless otherwise noted. PLGA-COOH with an intrinsic viscosity of 0.18 dL/g was purchased from Lactel. The (5- carboxypentyl)triphenylphosphonium (TPP) cation was synthesized according to literature methods [280]. HO-PEG-OH with a MW of 3,350 g/mol was purchased from

Sigma Aldrich. DSPE-PEG-COOH with a MW of 2,000 g/mol was purchased from

Avanti Polar Lipids. Polyvinyl alcohol (PVA) with an average molecular weight of

10,000-26,000 was purchased from Alfa Aesar. Distilled water was purified by passage through a Millipore Milli-Q Biocel water purification system (18.2 MΩ) with a 0.22 μm filter. HPLC analyses were carried out using an Agilent 1200 series instrument. Size and zeta potential measurements were carried out on a Malvern Zetasizer Nanoseries instrument. Gel permeation chromatography (GPC) data was collected on a Shimadzu

LC20-AD Prominence liquid chromatographer. All NMR spectra were recorded on a

Varian 400 MHz NMR. TEM images were taken on a Tecnai 20 FEM microscope.

Ultrasonication was performed on a Misonix S-4000 Ultrasonic liquid processor.

Synthesis of PLGA-b-PEG-OH. Synthesis of PLGA-PEG-b-OH was performed as previously described [281].

Synthesis of PLGA-b-PEG-TPP. Synthesis of PLGA-PEG-b-TPP was performed as previously described [281].

Synthesis of Gemcitabine Encapsulated Nanoparticles (Gem-PLGA-b-PEG-OH-

NPs). Gemcitabine (10 mg/mL in nanopure H2O) was emulsified with PLGA-b-PEG-OH

(5 mg in CH2Cl2) using probe sonication for one minute (Amplitude: 40% of 600 W power, Pulse: 1 sec on, 1 sec off). The primary emulsion was emulsified again by adding the water-in-oil emulsion to 2 mL of nanopure water containing 0.5% polyby 8 mL of nanopure water containing 0.05% PVA and sonicated using the conditions mentioned above. Organic solvent was removed by washing three times using an Amicon filtration membrane with a 100 kDa cut-off. The NPs were resuspended in nanopure water and stored at 4°C until further use. Dynamic light scattering (DLS) measurements were carried to determine size, polydispersity index (PDI), and zeta potential of the NPs. NPs were characterized using TEM at an acceleration voltage of 200 kV. The TEM samples were prepared by depositing 8 μL of the NPs (5 mg/mL) onto a 200-mesh carbon-coated copper grid. Samples were blotted away after 15 min and grids were negatively stained with sterile 2% (w/v) uranyl acetate aqueous solution for 15 min.

Synthesis of DZNep Encapsulated Nanoparticles (DZNep-PLGA-b-PEG-OH-

NPs). DZNep (10 mg/mL in nanopure H2O) was emulsified with PLGA-b-PEG-OH (5 mg in CH2Cl2) using probe sonication for one minute (Amplitude 40 % of 600 W power,

Pulse: 1 sec on, 1 sec off). The primary emulsion was emulsified again by adding the above formed water-in-oil emulsion to 2 mL of nanopure water containing 0.5% PVA and sonicated using similar conditions. Finally, this emulsion was re-emulsified using 8 mL of nanopure water containing 0.05% PVA under sonication using the similar

conditions as before. Organic solvent was removed by washing three times using an

Amicon filtration membrane with a 100 kDa cut-off. The NPs were resuspended in nanopure water and stored at 4°C until further use. DLS measurements were carried to determine size, polydispersity index (PDI), and zeta potential of the NPs. NPs were characterized using TEM as mentioned above.

Synthesis of Gemcitabine and DZNep Encapsulated NPs (Gem-DZNep-PLGA-b-

PEG-OH-NPs). Gemcitabine and DZNep co-encapsulated PLGA-b-PEG-OH NPs were following the exact procedure mentioned above except the first step where DZNep (10 mg/mL in nanopure H2O) and Gemcitabine (10 mg/mL in nanopure H2O) was emulsified with PLGA-PEG-OH (5 mg in CH2Cl2). Organic solvent was removed by washing three times using an Amicon filtration membrane with a 100 kDa cut-off. The NPs were resuspended in nanopure water and stored at 4°C until further use. DLS measurements were carried to determine size, polydispersity index (PDI), and zeta potential of the NPs.

NPs were characterized using TEM as mentioned above.

Synthesis of Controlled Release Gem-DZNep-PLGA-b-PEG-TPP-NPs.

Gemcitabine (10 mg/mL in nanopure H2O) was emulsified with PLGA-b-PEG-TPP (5 mg in CH2Cl2) using probe sonication for one minute (Amplitude: 40% of 600 W power,

Pulse: 1 sec on, 1 sec off). The primary emulsion was emulsified again by adding the water-in-oil emulsion to 2 mL of nanopure water containing 0.5% PVA and sonicated using similar conditions. Finally this emulsion was emulsified using 8 mL of nanopure water containing 0.05% PVA and DZNep (10 mg/mL in nanopure H2O) under sonication using conditions as mentioned above. Organic solvent was removed by washing three times using an Amicon filtration membrane with a 100 kDa cut-off. The NPs were

resuspended in nanopure water and stored at 4°C until further use. DLS measurements were carried to determine size, polydispersity index (PDI), and zeta potential of the NPs.

NPs were characterized using TEM as mentioned above.

Synthesis of Controlled Release Gem-DZNep-DSPE-PEG-OH-NPs. Gemcitabine

(10 mg/mL in nanopure H2O) was emulsified with PLGA-b-PEG-OH (5 mg in CH2Cl2) using probe sonication for one minute (Amplitude: 40% of 600 W power, Pulse: 1 sec on,

1 sec off). The primary emulsion was emulsified again by adding the water-in-oil emulsion to 2 mL of nanopure water containing 0.5% PVA and sonicated using the conditions mentioned above. This emulsion was further emulsified using 8 mL of nanopure water containing 0.05% PVA, 0.1% DSPE-PEG-COOH, and DZNep (10 mg/mL in nanopure H2O) under sonication using the same conditions as mentioned above. Organic solvent was removed by washing three times using an Amicon filtration membrane with a 100 kDa cut-off. The NPs were resuspended in nanopure water and stored at 4°C until further use. DLS measurements were carried to determine size, polydispersity index (PDI), and zeta potential of the NPs. NPs were characterized using

TEM as mentioned above.

Loading and Encapsulation Efficiency of Gemcitabine and DZNep in NPs.

Gemcitabine and DZNep loading and encapsulation efficiency (EE) were determined by dissolving the polymeric core in 0.1 M NaOH and quantifying the amount of therapeutic in the NPs using HPLC (Mobile phase: acetonitrile with 20% H2O and 0.1% trifluoroacetic acid, Wavelength used: 270 nm.

Release Kinetic Studies of DZNep/Gemcitabine NPs. NPs were placed in Slide-A-

Lyzer® MINI dialysis units with a MW cutoff of 10,000 g/mol. Samples were dialyzed

against 1x PBS (6.0 L) in a constant temperature shaker at 37°C and 35 RPM. At various pre-determined time points, samples were removed and the gemcitabine/DZNep content was determined by dissolving the polymeric core in 0.1 mMol NaOH and quantifying by

HPLC analysis (Mobile phase: acetonitrile with 20% H2O and 0.1% trifluoroacetic acid,

Wavelength used: 270 nm).

Statistical Analyses. For instances where only two conditions were compared, the student's t test was used to identify significant differences. In cases where three or more conditions were compared, one-way ANOVA was conducted followed by Tukey’s

Multiple Comparison Test. Each experiment was repeated at least three times. Unless otherwise indicated, p<0.05 and p<0.01 compared with control conditions are represented by one and two asterisks, respectively.

Results

DZNep selectively augments gemcitabine cytotoxicity in cancerous but not normal pancreatic cells. We began by studying the cytotoxicity of DZNep on normal and cancerous pancreatic cell lines. Treatment with increasing concentrations of DZNep (0.1 nM-100 μM) showed a significant reduction in cellular viability in all six pancreatic cancer cell lines tested (i.e., BxPC-3, Capan-1, L3.6pl, AsPC-1, MIA PaCa-2, PANC-1)

(Fig. 4.1A). Cytotoxicity was observed starting at approximately 0.5-1 μM DZNep, depending on the cell line, and increased gradually with higher concentrations thereafter.

Cytotoxicity did not begin until ~48 h of treatment (data not shown). Maximal reduction in cell viability (~50% reduction with 10 μM DZNep for 72 h) was observed in poorly- differentiated MIA PaCa-2 which is only marginally gemcitabine-sensitive (Fig. 4.1A).

In subconfluent cells, the effect of DZNep on MIA PaCa-2 was almost similar to that of gemcitabine (Fig. 4.1C). DZNep showed much lower cytotoxicity in well-differentiated, gemcitabine-sensitive pancreatic cancer cell lines (i.e., BxPC-3, Capan-1, L3.6pl) when compared with gemcitabine (Fig. 4.1C). Of note, no significant changes were observed in the proliferation and cellular viability of normal human pancreatic ductal epithelial

(HPDE) cells (up to 10 μM DZNep for 72 h) which are highly gemcitabine-sensitive, as well as two other normal cell lines (293T (human embryonic kidney) and MCF-10A

(human breast epithelial)) (Figs. 4.1A-B; supplemental data in full manuscript [190]).

These results demonstrate that DZNep selectively imparts cytotoxic effects to (poorly- differentiated) pancreatic cancer cells without significant impairment of proliferation in normal pancreatic epithelial cells.

Encouraged by the potential ability of DZNep to compensate for nucleoside analog refractoriness in pancreatic cancer, we co-treated cells with DZNep and gemcitabine at equimolar ratios and compared the cellular viabilities with those obtained when cells were treated with either DZNep or gemcitabine alone. Interestingly, the co- treatment showed significantly higher reductions in cellular viabilities of most of the gemcitabine-sensitive and -insensitive cancerous cells than when the compounds were used as stand-alone agents (Fig. 4.1C). Conversely, the co-treatment reduced cytotoxicity

(>2-fold at 100 nM concentration) in HPDE (Fig. 4.1B), suggesting a cytoprotective role of DZNep only on normal cells.

We estimated the combination index (CI) plots [276] to identify the type of interaction between DZNep and gemcitabine. These estimations identified a synergistic or additive interaction between DZNep and gemcitabine (100 nM-10 μM) in all

Figure 4.1. DZNep and gemcitabine sensitivity, singly or in combination, and interactions within a panel of pancreatic cell lines. A. All cancerous cell lines excluding the normal HPDE are DZNep-responsive and reduced cellular viability. B. DZNep and gemcitabine displayed antagonistic effects in HPDE. C. DZNep and gemcitabine displayed additive or synergistic effects in many of the cancerous pancreatic cell lines. Twenty-four hours after 3x103 cells/well were seeded in a 96-well plate, cells were treated with either DZNep, gemcitabine, or a combination of both at an equimolar ratio for 72 h. Cellular viability was measured using an MTT assay. Cytotoxic IC50 values are indicated. Significances between gemcitabine and DZNep as well as DZNep + Gemcitabine and DZNep were identified using one-way ANOVA followed by Tukey’s post-hoc test. Combination index (CI) plots (insets) show the interactions between the two drugs. CI>1, antagonism; CI=1, additivity; CI<1, synergism. Bar graphs to the right indicate the relative caspase-3 activity (RCA) of each treatment as measured by fluorescence intensity. Values were background- subtracted and are presented as fold-change from the control. Significance between a single drug versus the drug combination was identified via one-way ANOVA followed by Tukey’s post-hoc analysis. Cells were treated with 1 μM DZNep, 100 nM gemcitabine, or both. Bars, SD. n=3. *p<0.05, **p<0.01.

pancreatic cancer cell lines with the only exception of AsPC-1, as well as a stark antagonistic interaction in normal HPDE (Fig. 4.1C). Co-treatment of DZNep (1 nM-10

µM) with differing gemcitabine concentrations (1-100 nM) showed that DZNep

potentiates cytotoxicity in a gemcitabine dose-dependent fashion in marginally gemcitabine-sensitive MIA PaCa-2, whereas such a potentiation occurred in a largely gemcitabine dose-independent fashion in highly gemcitabine-sensitive Capan-1 cells

(supplemental data in full manuscript [190]). HPDE remained cytotoxic to gemcitabine, confirming antagonism between the two compounds (supplemental data in full manuscript [190]). Further interaction analysis with various ratios of gemcitabine:DZNep

(1:1, 1:4, 4:1, 1:10, 10:1) identified a maximally synergistic and cytotoxic response in

MIA PaCa-2 using the 1:10 ratio (supplemental data in full manuscript [190]). Finally, in order to understand the mechanism of reduction in cellular viability, we conducted caspase-3 based apoptosis assays. These results (Fig. 4.1C) identified a significant induction of apoptosis as a major contributing mechanism for DZNep-induced increase in

gemcitabine cytotoxicity (Fig. 4.1C).

Nucleoside transporters facilitate cellular entry of DZNep. Since DZNep is a hydrophilic nucleoside analog with a log P of -1.38 [256], we hypothesized that its entry into cells would depend on the expression of nucleoside transporters, and therefore, the lack of sufficient carrier expression would limit its cellular activity. To test this, we performed a competitive assay by assessing the level of inhibition in the cellular transport of nucleosides with DZNep. We chose the PANC-1 and MIA PaCa-2 cell lines for these studies because they have been reported to express multiple nucleoside transporters

[282]. Transport analysis identified a significant inhibition of 3H-adenosine and 3H- guanosine (purines) transport but not 3H-thymidine (pyrimidine) (Fig. 4.2A).

Gemcitabine transport in both cell types was only inhibited at a higher concentration (200

Figure 4.2. DZNep partially competes with the uptake of purine nucleosides by hENT1 and hCNT3. A. DZNep hindered the uptake of radiolabeled purine nucleosides in PANC-1 and MIA PaCa-2. Twenty-four hours after 5x104 cells/well were seeded in a 24-well plate, cells were allowed to uptake the indicated radiolabeled nucleoside in the presence of DZNep or its respective unlabeled nucleoside. B. Inhibition of adenosine transport in Xenopus oocytes with DZNep. C. Pharmacological inhibition of hENT1 and excess uridine decreased the cytotoxicity of DZNep in MIA PaCa-2. Twenty-four hours after 3x103 cells/well were seeded in a 96-well plate, cells were treated with increasing concentrations of DZNep in the presence of DMSO (control), 10 µM NBMPR, or 200 µM uridine. Cellular viability was measured using an MTT assay. IC50 values are indicated. Significant differences between the control and each treatment were determined using the Student’s t test. Bars, SD. n=3. *p<0.05, **p<0.01.

μM) (Fig. 4.2A). These data suggest that DZNep cellular influx is facilitated by purine, but not pyrimidine, nucleoside transporters.

To further characterize the identity of the purine nucleoside transporters mediating DZNep influx, we examined the ability of DZNep to inhibit 3H-adenosine

transport in Xenopus oocytes expressing individual transporters. A significant inhibition of hENT1- and hCNT3-mediated 3H-adenosine transport was observed, whereas a significant inhibition of hCNT2-mediated 3H-adenosine transport was noticed only at the higher (200 μM) concentration (Fig. 4.2B). No significant reduction in adenosine transport by DZNep was observed in hENT2-expressing oocytes.

Next, to study the impact of hENT1 and hCNT3 on DZNep-mediated cytotoxicity in pancreatic cancer cells, we examined the proliferation of MIA PaCa-2 in the presence of a pharmacological inhibitor of hENT1 (10 nM NBMPR) or excess uridine (200 μM) which competitively inhibits all hENTs and hCNTs. Our results show that both conditions can significantly reduce DZNep cytotoxicity of MIA PaCa-2, as judged by

increases in cytotoxic IC50 estimates (2-8-fold) (Fig. 4.2C).

Acyl derivatization of DZNep augments sensitization of pancreatic cells to gemcitabine. Since DZNep at least partially utilizes transporters for exerting its fullest potential, it is likely that transporter-deficient patients may respond poorly to DZNep. To circumvent this issue, we generated polar acyl derivatives of DZNep (Fig. 4.3A) using a synthetic procedure described in Materials and Methods. We substituted DZNep at the

N6-NH2 group with acyl side chains to generate two lipophilic prodrugs (Prodrug 1:

C20H29ClN4O4; Prodrug 2: C18H24N4O4). The synthesized prodrugs were evaluated for their anti-proliferative activities, both alone and in combination with gemcitabine, in a normal (HPDE), gemcitabine-sensitive (Capan-1), and gemcitabine-resistant (MIA PaCa-

2) cell line. Both prodrugs showed higher cytotoxicity than the parent DZNep in MIA

PaCa-2 (Fig. 4.3B). Further, the levels of cytotoxicity produced by both of the prodrugs were unaffected in the presence of 10 nM NBMPR or 100 μM uridine since the estimated

IC50 values were not significantly different with the treatments. In normal HPDE, cytotoxicity of the prodrugs did not differ from that of DZNep (>10 µM) (Fig. 4.3B).

When cells were treated with both gemcitabine and a DZNep prodrug at the previously

Figure 4.3. Acyl modifications of DZNep further enhance cytotoxicity. A. The chemical structures of DZNep and its two acyl prodrugs (Prodrug 1: C20H29ClN4O4, and Prodrug 2: C18H24N4O4). B. Cytotoxicity of DZNep versus its prodrugs in HPDE and MIA PaCa-2. IC50 values are designated in each legend. Significance between each prodrug and DZNep was identified using the Student’s t test. C. Average IC50 values of the various drug combinations in HPDE, Capan-1, and MIA PaCa-2. Twenty-four hours after 3x103 cells/well were seeded in a 96-well plate, cells were treated for 72 h. Cellular viabilities were measured using MTT assays. IC50 values are plotted. Significance of each prodrug combination was compared with Gem + DZNep using one-way ANOVA followed by Tukey’s post-hoc test. Bars, SD. n=3. *p<0.05, **p<0.01.

calibrated 1:10 ratio, both prodrugs showed a distinct potentiation of gemcitabine cytotoxicity in the cancerous Capan-1 and MIA PaCa-2 (IC50 reduction of 12-fold in

Capan-1 and 7-fold in MIA PaCa-2) (Fig. 4.3C). Despite anticipated inertness, both prodrug combinations with gemcitabine increased cytotoxicity in HPDE (Fig. 4.3C).

Overall, cytotoxicity increased from gemcitabine as a single agent, to a combination with

DZNep, to a combination with DZNep prodrugs (Fig. 4.3C). Since gemcitabine is well- known to rely heavily on nucleoside transporters for cellular activity, we examined the effects of an acyl derivative (C24H41N3O5) of a less hydrophilic nucleoside analog, troxacitabine (log P of -0.66) [278, 279], in combination with DZNep Prodrug 1 on the same cell lines. Cytotoxicity was further enhanced with the use of the troxacitabine prodrug in combination with DZNep Prodrug 1 (Fig. 4.3C). While apoptosis is identified as a contributing mechanism for DZNep cytotoxicity, other possible mechanisms including autophagy or senescence may be responsible for the large reduction in cellular viability with multiple drugs. Taken together, these results indicate that acyl derivatives of one or both nucleosides can substantially potentiate the anti-proliferative effects of pancreatic cancer cells by overcoming dependence on cellular transport barriers.

However, these results also indicate that the prodrugs may compromise the selective antiproliferative effects seen in parent DZNep since cytotoxicity also increased in HPDE.

DZNep rapidly and reversibly decreases H3K27 and H4K20, but increases H3K9, trimethylations in pancreatic cancer cells. To investigate whether DZNep influences chemosensitivity in pancreatic cancer by affecting histone methylation-dependent chromatin states, we profiled methylation marks in key lysine residues in histones H3 and

H4. Western blotting analyses indicated that in untreated conditions, H3K9 was

Figure 4.4. DZNep alters histone lysine methylation and methyltransferase and demethylase expressions in pancreatic cancer. A. Changes in methylation levels of H3K4, H3K9, H3K27, and H4K20 in HPDE and MIA PaCa-2 treated with DZNep (0- 100 µM). Cells were treated with DZNep for 24 h, and whole cell lysates (50 µg) were subjected to Western blotting analysis. β-actin, the internal loading control, is shown with a representative blot. B. Western blotting analysis of histone lysine methyltransferases and demethylases in MIA PaCa-2 treated with DZNep (1 µM) for up to 48 h. C. Histone methylation dynamics in MIA PaCa-2 treated with DZNep (1 µM) for up to 72 h.

predominantly monomethylated, H3K27 was trimethylated, and H4K20 was dimethylated in MIA PaCa-2 (Fig. 4.4A). Treatment with increasing concentrations of

DZNep (1-100 μM) for 24 h clearly reversed at least two of these marks viz., increased

H3K9me3 and decreased H3K27me3 (Fig. 4.4A). Further, the DZNep-induced increase in H3K9me3 was accompanied by a decrease in H3K9me1/2 forms; conversely, the

DZNep-induced decrease in H3K27me3 was accompanied by an increase in

H3K27me1/2 forms (Fig. 4.4A). DZNep treatment in MIA PaCa-2 also slightly decreased

H4K20, but not H3K4, trimethylation states. Further time course analyses of the two distinct changes in the histone trimethylation states (i.e., H3K27 and H3K9) identified decreases in H3K27me3 to occur as early as 8 h of DZNep treatment, whereas the increase in H3K9me3 peaked only after 24 h of treatment (Fig. 4.4C). Unlike MIA PaCa-

2, DZNep treatment in HPDE cells produced no obvious changes in methylation states of any of the histone lysines tested, suggesting that DZNep-induced chromatin alterations were selective to cancerous pancreatic cells (Fig. 4.4A).

To test the mechanisms by which these methylation changes occurred, we profiled the expressions of putative histone lysine methyltransferases and demethylases implicated in these processes. Specifically, we tested changes in the levels of the PRC2 subunits (EZH2, SUZ12, and EED) and JMJD histone demethylases (JMJD1A, JMJD1B, and JMJD2C) which have been reported to selectively target the methylation and demethylation reactions in H3K27 and H3K9 residues, respectively. Western blotting analyses of whole cell lysates prepared from DZNep-treated MIA PaCa-2 showed a moderate decrease in EZH2 histone methyltransferase and significant decreases in

JMJD1A and JMJD2C histone demethylases correlating to changes observed in H3K27 and H3K9 methylations states, respectively (Fig. 4.4B). No significant changes were

observed in SUZ12, EED, and other JMJD subunits upon DZNep treatment (Fig. 4.4B).

Short DZNep priming prior to gemcitabine treatment shows superior chemosensitizing effects compared with simultaneous exposure of both drugs: Inhibition of H3K27me3 as a putative mark for DZNep chemosensitization. To test whether the temporal differences in various lysine trimethylation changes correspond with changes in gemcitabine chemosensitization, we next performed gemcitabine cytotoxicity analyses after DZNep priming of Capan-1 and MIA PaCa-2 for various time periods (4, 8, and 12 h). Consistent with early and rapid inhibition of H3K27me3, maximal gemcitabine chemosensitization was also observed in dosing schedules where the cells were primed with DZNep for only a short 4-8 h period prior to gemcitabine treatment, and the changes

Figure 4.5. Short priming of DZNep demonstrated superior cytotoxicity and synergy with gemcitabine than co-exposure of the two drugs. A. Short exposure with DZNep for 4-8 h produced maximal cytotoxic effects. Cells were exposed with DZNep at 1 µM for varying time intervals followed by increasing concentrations of gemcitabine (0-0.1 µM). Significance between 0 and 4 h is indicated. B. Superior cytotoxicity and synergism between gemcitabine and DZNep were observed when cells were primed with DZNep, as opposed to cotreatment with gemcitabine. Representative growth inhibition curves are shown. Twenty-four hours after 3x103 cells/well were seeded in a 96-well plate, cells were exposed to gemcitabine and DZNep concentrations at a 1:10 ratio either as a co-treatment for 72 h (C) or a primed treatment (with DZNep for 8 h followed by gemcitabine for 72 h) (P). Cellular viabilities were measured using MTT assays. Significance between co-treatment and priming is indicated. Combination index (CI) plots (insets) show the interactions between the two drugs. CI>1, antagonism; CI=1, additivity; CI<1, synergism. Bars, SD. n=3. *p<0.05, **p<0.01. C. Apoptosis levels were significantly greater in Capan-1 and MIA PaCa-2 cells with priming compared with co-treatment, while apoptosis levels in HPDE decreased with priming. Cells were either co-treated with 10 μM DZNep and 1 μM gemcitabine or primed with 10 μM DZNep for 8 h followed by 1 μM gemcitabine. Fluorescence values were background-subtracted and are indicated as fold-change from co-treatment to priming. Significant differences between co-treatment and priming were identified using the Student’s t test. Bars, SD. n=3. *p<0.05, **p<0.01. D. Maximal reduction in H3K27 trimethylation was seen with priming schedules at 1:10 DZNep:gemcitabine. MIA PaCa-2 was treated with vehicle, gemcitabine for 72 h,

DZNep for 8 h, DZNep and gemcitabine for 72 h, or DZNep for 8 h followed by gemcitabine for 72 h. 100 µg of whole cell lysates were subjected to Western blotting analysis. Blots were stripped and re-probed for β-actin, the internal loading control. Densitometry ratios are indicated.

did not significantly increase any further with a later priming time (12 h) (Fig. 4.5A).

Such increases in synergistic drug responses with short DZNep priming were also greater than that observed when DZNep was co-treated with gemcitabine for the entire 72 h (Fig.

4.5B). A decrease in cellular viability in MIA PaCa-2 was evident throughout, but only statistically significant at the highest concentration. For Capan-1, no statistical differences were observed; however, synergism was distinctly increased with priming.

Intriguingly, short priming of DZNep for 8 h augmented its antagonistic interaction with gemcitabine in HPDE causing a further reduction in the extent of cytotoxicity (Fig. 4.5B).

Statistically significant increases in caspase-3 levels further support apoptosis as a major contributing factor to the observed reduced cellular viability in pancreatic cancer cells

(Fig. 4.5C). Taken together, these data suggest that short DZNep priming rather than continuous co-treatment may chemosensitize pancreatic cancer cells to gemcitabine both in an effective and a selective manner.

Finally, we tested the H3K27 and H3K9 trimethylation statuses during the entire period of treatment with both DZNep and/or gemcitabine at varying schedules (short priming or co-exposure) and ratios (1:10 or 10:1). Maximal reductions in H3K27me3 were noticed with the DZNep short priming schedule followed by gemcitabine treatment at a 10:1 ratio (Fig. 4.5D). Although H3K9me3 was increased in the aforementioned schedule, it was essentially less differentiable from the co-exposure schedule as well as when the DZNep:gemcitabine ratios were different. Furthermore, the highest increase in

H3K9me3 was noted with the gemcitabine-alone schedule for the entire 72 h period, where cytotoxicity observed was the minimal among all schedules tested (Fig. 4.5D).

These data further provide evidence that the loss of H3K27me3, and not increase in

H3K9me3, is the closest predictor of DZNep potentiation of gemcitabine

chemosensitivity in pancreatic cancer cells.

An engineered nanoparticle for spatiotemporal release of DZNep and gemcitabine reduces DZNep dose while bringing maximal chemosensitivity. To enhance tumor cell delivery of the drugs, we encapsulated DZNep and gemcitabine individually into biodegradable poly(lactide-co-glycolide)-b-polyethyleneglycol (PLGA-b-PEG) nanoparticle (NP) formulations (Fig. 4.6A). Transmission electron microscopy (TEM) studies showed a uniform size distribution of both DZNep and gemcitabine NPs ranging between 100-200 nM (Fig. 4.6B). In vitro release kinetics suggested a similar release profile for both drugs with almost comparable physicochemical characteristics (Fig. 4.6C;

Table 4.2). Cytotoxicity analyses with drug-encapsulated NPs indicated that DZNep entrapment within NPs can produce the same level of cytotoxicity as that of free DZNep

Table 4.2. Physiochemical characterization of NPs Nanoparticle Size (nm) PDI Zeta Potential Loading (%) EE (%) (mV) Gemcitabine DZNep Gemcitabine DZNep Empty-PLGA- 172 ± 2 0.13 -23.5 ± 1.3 - - - - b-PEG-OH-NPs Gem-PLGA-b- 200 ± 4 0.16 -27.6 ± 0.7 1.4 ± 0.1 - 6.6 ± 0.2 - PEG-OH-NPs DZNep-PLGA- 206 ± 2 0.16 -29.6 ± 0.1 - 1.9 ± 0.1 - 9.1 ± 0.6 b-PEG-OH-NPs Gem-DZNep- 200 ± 2 0.17 -24.3 ± 4.9 1.3 ± 0.1 1.7 ± 0.2 6.2 ± 0.5 8.2 ± 0.7 PLGA-b-PEG- OH-NPs Gem-DZNep- 157 ± 1 0.20 16.4 ± 0.9 1.2 ± 0.2 1.6 ± 0.5 5.5 ± 0.8 7.7 ± 2.4 PLGA-b-PEG- TPP-NPs Gem-DZNep- 167 ± 2 0.23 -17.9 ± 2.8 1.3 ± 0.2 1.9 ± 1.0 6.3 ± 0.8 8.9 ± 4.9 DSPE-PEG-NPs

100

although at a much lower (~10-fold) concentration (Fig. 4.6D). A similar reduction in dose requirement was also obtained with gemcitabine when used in NP formulation (Fig.

4.6D).

Since superior chemosensitization effects were seen when cells were initially primed with DZNep for a short period of time, we engineered two different NP formulations such that gemcitabine was encased in the core while DZNep was loaded within the outer layer (Fig. 4.6A). The first strategy used DSPE-PEG-OH to create a lipid layer to accommodate DZNep alone, and the second used a PLGA-b-PEG-TPP polymer to keep the two drugs separated by a cationic charge (Fig. 4.6A). TEM studies once again showed a uniform size distribution of both NP formulations (Fig. 4.6E). However, comparison of gemcitabine and DZNep release kinetics showed a much more rapid release of DZNep (~80% within 8 h) than gemcitabine in the former and a slightly greater release of DZNep (~50% within 8 h) than gemcitabine in the latter (Fig. 4.6F). Treatment of MIA PaCa-2 with each controlled-release NP type showed a superior cytotoxicity response than gemcitabine and DZNep co-encapsulated NPs at the same ratios (IC50 reductions for DSPE formulation of 1.2-2.5-fold (p=0.48) and TPP formulation of 1.5-

4.9-fold (p=0.12)) (Fig. 4.6G). In both cases, a significant induction of apoptosis consequent to inhibition of H3K27me3 was identified as a key mechanism for the observed cytotoxicity (Figs. 4.6G-H).

Discussion

Advantages of gemcitabine for pancreatic cancer therapy include its mechanistic attributes such as self-potentiation and masked chain termination [218], ability to expose

101

Figure 4.6. Spatiotemporal release of DZNep and gemcitabine using engineered nanoparticles reduced drug dose while potentiating chemosensitivity. A. Spatial distribution of DZNep and gemcitabine within NPs. Co-encapsulating double-emulsion formulations were created using PLGA-b-PEG-OH (left), DSPE-PEG-OH (middle), and PLGA-b-PEG-TPP (right). B and E. TEM illustrates the inner core and outer shell of all the double-emulsion NPs created. Insets show the NPs at higher magnification.

102

Bars, SD. n=3. C. Release kinetics indicates the similarity between DZNep and gemcitabine release using the PLGA-b-PEG-OH formulation. D. Gemcitabine (top) and DZNep (bottom) in individual PLGA-b-PEG-OH formulations distinctly increased cytotoxicity in MIA PaCa-2. Significance between nanoparticles and free formulation is shown. F. HPLC analyses demonstrate the rapid and sequential release of DZNep compared with gemcitabine in both DSPE (top) and TPP (bottom) formulations.G. Both engineered DSPE (top) and TPP (bottom) delayed-release NPs increased the cytotoxicity of MIA PaCa-2 even further compared with PLGA-b-PEG-OH NPs. Twenty-four hours after 5x103 cells/well were seeded in a 96-well plate, cells were treated for 72 h. Cellular viabilities were measured using MTT assays. Cytotoxic IC50 values are indicated. Significance between simultaneous and delayed NPs using the Student’s t test is shown. Bar graphs to the right indicate the relative levels of caspase- 3 activity as measured by fluorescence intensity. Values were background-subtracted and are indicated as fold-change from simultaneous NPs to delayed NPs. Bars, SD. n=3. *p<0.05, **p<0.01. H. H3K27 trimethylation decreases with simultaneous and delayed-release NPs. MIA PaCa-2 was treated with empty, simultaneous, or delayed NPs as above and total cell lysates were analyzed for H3K27 trimethylation using Western Blotting. A clear reduction in H3K27 trimethylation was noticed in both simultaneous and delayed NPs, with DSPE (left) and TPP (right) delayed NPs producing a slightly greater reduction compared with simultaneous NPs.

cancer cell antigens to trigger immune reactions, and overall well-tolerability in patients.

Nonetheless, the use of gemcitabine monotherapy for pancreatic cancer only produces modest treatment and survival responses in patients. Several studies have observed potential issues associated with the continuous usage of this treatment, including increased development of stem cell characteristics and EMT behavior, enhancing chemoresistance [283]. Furthermore, innumerable studies evaluating the potential of gemcitabine in combination drug regimens with other conventional chemotherapeutics

(e.g., platins, taxols) displayed a lack of synergism or even antagonistic effects [6].

Therefore, our study attempts to increase the effectiveness of gemcitabine in pancreatic cancer by reducing its disadvantageous effects while retaining beneficial anticancer responses.

103

DZNep by itself exerted poor cytotoxicity in pancreatic cancer cells. Delayed and moderate results were observed with effects at the most equaling gemcitabine only in poorly-differentiated pancreatic cancer cell types. However, the drug was found to increase differentiation and E-cadherin expression in other cancer types as well as potentially reprogram gene expression [253]. Therefore, we predicted that DZNep could compensate for the refractoriness acquired from gemcitabine treatment. Although the exact mechanism of DZNep cellular action is unclear, based on its hydrophilicity (log P of -1.38) [256], we speculated that DZNep cellular entry will depend on a transport process. As noted in this report, DZNep appears to only partially utilize hENT1 and hCNT3, which is relatively moderate considering its nucleoside analog structure. This suggests that the drug may also utilize other classes of transporters capable of transporting nucleosides in order to gain entry into cells. Whether diffusion also plays a role is unclear; however, it is likely. Nevertheless, the dependence on transporters for drug uptake could partially explain a previous study which noted the limited absorption, reduced biodistribution, and rapid elimination of DZNep in mice [284]. More importantly, since earlier studies identified that patients with low nucleoside transporter expressions, especially hENT1 and hCNT3, exhibit poor treatment and survival outcomes when treated with nucleoside analogs, we synthesized acyl prodrugs of DZNep to maximize its cytotoxic response. The results support increased lipophilicity as a means to bypass transport requirements and as a preferred strategy for enhancing nucleoside analog efficacy in pancreatic cancer cells [278, 279]. Second, unlike gemcitabine, DZNep is most likely not phosphorylated into an active form. Earlier studies have reported the drug to be equally effective in cells expressing or lacking adenosine kinase [265],

104

suggesting it unlikely that phosphorylated metabolites of DZNep act as the active units for cytotoxicity. Furthermore, the structure of DZNep does not indicate the likelihood for phosphate group addition [285]. Consistently, and as demonstrated in this study, epigenetic mechanisms such as alterations in histone methylations could aid in the mechanism of DZNep action in pancreatic cancer cells.

The use of DZNep in combination with gemcitabine was investigated in detail for potential enhancement of efficacy via advantageous interactions between the two compounds. Further, since gemcitabine has the ability to enhance chemoresistance, it was crucial to consider the pattern of dose exposure in pancreatic tumors especially to minimize clonal expansion of populations with increased stemness. In this study, a prior short exposure (i.e., priming) of DZNep was found to have pronounced overall cytotoxic effects with less drug exposure compared with conventional dosing (i.e., co-treatment).

This illustrated the potential of DZNep to act as a potent chemosensitizing agent, rather than a combination cytotoxic agent, for nucleoside analogs in treating pancreatic cancer.

While the avoidance of transport-based drug interactions (i.e., competition for cellular entry, especially for broadly specific nucleoside transporters (hENT1, hCNT3)) could contribute to the superiority of the priming schedule, DZNep also profoundly alters histone methylation states in cancer cells. Therefore, it is speculated that the genetic reprogramming consequent to post-translational modifications of histone lysines could trigger downstream alterations of chemosensitizing factors advantageous to gemcitabine’s mechanism of action. Consistently, DZNep effects in this study were observed as early as 4 h, correlating with the approximate time needed for the transcription of new genes. DZNep-responsive genes have been examined [267, 286];

105

however, their roles in chemosensitivity remain unknown, warranting further studies.

Since gemcitabine is a pyrimidine (cytidine) nucleoside analog and DZNep is a purine

(adenosine) nucleoside analog, it is also likely that the two could impact distinct endogenous nucleotide pools and cellular targets, allowing their functions to synergize.

Conversely, our studies revealed an antagonistic interaction between DZNep and gemcitabine in HPDE. While the mechanism for a cytoprotective response in normal cells is presently unclear, the opposing action of DZNep on normal and cancerous pancreatic cancer cells presents broader clinical implications for improving chemotherapeutic efficacy in pancreatic tumors.

We found DZNep to rapidly and reversibly decrease H3K27me3 and increase

H3K9me3. In addition, both events appeared to be temporally separated during DZNep treatment with the loss of H3K27me3 preceding the gain of H3K9me3. While some studies show that an increase in H3K9me3 is associated with transcriptional repression, the significance of the increase during DZNep treatment remains unclear, including whether it favors or opposes chemosensitivity. However, a study by Rogenhofer et al. identified overexpressed H3K9 methylation in benign renal tissue when compared with cancerous, as well as a correlation between H3K9 and H4K20 methylation levels with renal cell carcinoma progression [287]. Furthermore, the methylation statuses of H3K9 and H3K4 (both shown to be correlated with drug sensitivity [288, 289]) with respect to

DZNep treatment may simply be governed by their mutually-exclusive properties (i.e.,

H3K9 demethylation as a requirement for subsequent H3K4 methylation [290]). Since both decreased H3K27 and increased H3K9 trimethylations are distinctly noted upon

DZNep treatment [270], we investigated these lysine alterations with cytotoxic response.

106

Clearly, maximal reduction of H3K27me3 was noticed in the DZNep short priming schedule followed by gemcitabine treatment, corroborating the sustained inhibition of

H3K27me3, and not increase in H3K9me3, as a favorable histone lysine mark for augmented nucleoside analog chemosensitization of pancreatic cancer cells.

Our data identified inhibition of JMJD1A and JMJD2C with DZNep treatment, correlating with increased H3K9me3 and consistent with the specificity of these demethylases to H3K9. However, our investigation into methylation changes in H3K27 demonstrated only moderately inhibited EZH2 protein, since it is known that EZH2 displays highest catalytic activity for the first monomethylation of H3K27 but relatively weak capability for the subsequent di- and tri-methylations [291]. Despite the common desire for a specific EZH2 inhibitor (since DZNep is a global methylation inhibitor, not specific to EZH2) [256, 270], we believe this may actually be an advantage of the drug.

One study observed defects in organ development or function in mice with the inactivation of EZH2 in adult stem cells [292], suggesting austere side effects in silencing the protein entirely. In addition, reducing both mRNA and protein levels of EZH2 with

RNAi has been shown to result in different patterns of PRC2 target genes expressed as compared with the pharmacological effects of DZNep on EZH2 [267]. Since DZNep alters more than just the silencing of EZH2 (such as the hypomethylation of other genes

[284]), it seems the collective consequences may be what are contributing to its preferred chemosensitizing effects.

While our studies address many of the concerns previously reviewed (i.e., increasing the hydrophilicity of DZNep and investigating its potential in combination with a conventional chemotherapeutic agent [256]), the effects of DZNep on normal adult

107

stem and progenitor cells, remain unknown [292]. As noted by Crea et al., toxicokinetics have not yet been conducted in humans, and the drug is still very investigational for its use in cancer. Further, since some of our treatments were cytotoxic to even HPDE cells

(e.g., Troxacitabine Prodrug + DZNep Prodrug), the utility of DZNep may be best increased by using a targeted method. In addition, DZNep augmentation of gemcitabine chemosensitization in pancreatic cancer cell lines occurred only at a very high (≥1-20

μM) dose range. Furthermore, earlier in vivo mouse pharmacokinetic studies have reported that DZNep is eliminated with a short half-life (12.8 min) and only poorly distributes into peripheral tissues. Since more favorable pharmacokinetic profiles for

DZNep such as dose reduction, longer circulating half-life, and improved tumor targeting due to enhanced permeation and retention (EPR) effects could all be obtained with nanoparticle drug delivery, we further investigated the delivery of DZNep in nanoparticle formulation into pancreatic cancer cells. As demonstrated in this study, engineered NPs co-encapsulating both drugs and sequentially releasing DZNep followed by gemcitabine led to improved efficacy in vitro, revealing the potential of the optimized epigenetic- chemotherapeutic combination with targeted drug delivery.

In summary, we have developed in vitro optimization procedures for the DZNep- gemcitabine combination against human pancreatic cancer cells. While gemcitabine alone only produces modest effects in the cancer, the addition of DZNep synergizes the two drugs to enhance overall efficacy in poorly-differentiated cancer cells but not normal epithelial cells. By altering structure (DZNep acyl derivatives), dose (10:1 DZNep to gemcitabine), exposure (4 h DZNep priming prior to 72 h gemcitabine treatment), and formulation (engineered NPs), these results demonstrate potential for the use of this

108

epigenetic-chemotherapeutic combination approach for further studies in the treatment of pancreatic cancer.

109

EMT-PHENOTYPE CELLS AND CHEMOSENSITIVITY

110

CHAPTER 5

EXPRESSION OF THE SET ONCOPROTEIN CONTRIBUTES TO THE

EPITHELIAL-MESENCHYMAL TRANSITION (EMT) OF PANCREATIC CANCER 5

5 SW Hung, K Naidu, H Lee, CA Gilbert, TT Hoang, T Nagy, and R Govindarajan. To be submitted to Cancer Research. 111

Abstract

One of the key reasons for why pancreatic cancer has such a devastating prognosis is due to 80-90% of diagnostic cases occurring when metastasis has already presented. At the cellular level, activation of the epithelial-mesenchymal transition

(EMT) is a prerequisite for metastasis because it allows for the dissemination of primary tumor cells to distant sites. In this study, we sought to determine the role of a novel oncoprotein, an inhibitor of protein phosphatase 2A (PP2A) called SET, in EMT and pancreatic tumor progression. Among the two major isoforms of SET (isoform 1 corresponding to TAFIα or I2PP2Aβ and isoform 2 corresponding to TAFIβ or

I2PP2Aα), higher protein levels of SET isoform 2 were identified in all aggressive pancreatic cancer cell lines as well as in a subset of pancreatic tumors, especially the poorly-differentiated pancreatic ductal adenocarcinomas (PDACs). Overexpressing this cell-surface isoform 2, rather than the better-known nuclear isoform 1, in the epithelial

PANC-1 cell line induced EMT by causing a cuboidal, epithelial morphology and promoting cellular proliferation, colony formation, migration, and invasion. Consistently, knockdown of SET isoform 2 in the mesenchymal MIA PaCa-2 cell line reverted EMT and all associated characteristics. The induction of EMT in cells by SET isoform 2 was determined to be a result of cadherin switching (i.e., from expressing epithelial E- cadherin to mesenchymal N-cadherin). Furthermore, mechanistic studies identified SET- induced cadherin switching involved the SPARC/Slug and Rac1/JNK/c-Jun/AP-1 signaling pathways which inhibit E-cadherin and promote N-cadherin, respectively. In vivo, SET isoform 2 overexpression in an orthotopic tumor model led to increased tumor volume and metastatic ability, while SET isoform 2 knockdown reduced both features.

112

These findings have implications for the design and targeting of SET and its associated pathways for intervening pancreatic tumor progression.

Introduction

Pancreatic cancer is currently the fourth leading cause of cancer-related deaths in the U.S. [1] and is expected to become the second leading cause by as early as 2020

[293]. In other words, the annual incidence of pancreatic cancer is almost equal to the mortality rate [1]. One of the key reasons for why pancreatic cancer has such a poor prognosis is due to 80-90% of diagnostic cases occurring at stage IV, when metastasis has already occurred and the median survival is only 4.5 months [1]. Metastasis is an early event in pancreatic cancer and occurs after cells have undergone the epithelial- mesenchymal transition (EMT). Activation of EMT allows for the dissemination of primary tumor cells; hence, EMT is a prerequisite for metastasis [294]. EMT is a well- characterized event in cancer progression during which cells transform from a cuboidal shape with polarity to a squamous shape lacking polarity. This de-differentiation allows cells to invade into the bloodstream and cause metastasis by forming a secondary tumor on a distant site. Previously, we have shown that the SET oncoprotein is involved in cellular differentiation [173]. SET expression was found to lead to a poorly- differentiated, mesenchymal phenotype in cells, while knockdown of the protein in a pancreatic cancer cell line, MIA PaCa-2, led to profound reductions in cellular proliferation and colony-forming capacities.

Overexpression of the SET oncoprotein has been observed in numerous cancers, including leukemias, lymphomas, nephroblastoma, hepatoma, and choriocarcinoma, with

113

poor patient outcome also correlating with the protein’s expression [295-301]. SET was first discovered as template-activating factor-I (TAF-I) [295] and is also known as an inhibitor of the tumor suppressor protein phosphatase 2A (PP2A); hence, SET is also known as I2PP2A [302]. Although there are four protein isoforms of SET (Accessions:

(1) NP_001116293, (2) NP_003002, (3) NP_001234929, and (4) NP_001234930), each differing only at the N-terminus (Fig. 5.1A), isoforms 1 and 2 are the most well-known and correspond to I2βPP2A or TAF-1α and I2αPP2A or TAF-Iβ, respectively. In this study, we investigated all SET isoforms as a whole as well as isoforms 1 and 2 alone.

While the numerous roles of SET are still being explored, it is known that SET is also an inhibitor of histone H4 acetylation [303]. Additional roles of SET include its relationships with Rac1 (i.e., Rac1 signaling is required for membrane recruitment of

SET leading to increased cell migration [304, 305]), the MEK/ERK pathway (i.e., SET negatively regulates cellular proliferation via this pathway [306]), and the JNK pathway

(i.e., SET induces c-Jun/AP-1 activity through changes in c-Jun phosphorylation [307]).

Furthermore, both Shintani et al. and Park et al. have independently shown that inhibition of JNK increases PP2A expression and decreases N-cadherin (N-cad) expression, a key protein involved in promoting the mesenchymal phenotype in cells [308, 309].

Meanwhile, direct inhibition of PP2A through pharmaceutical means has already shown promise in pancreatic cancer cells, with inhibition leading to overactivation of the c-Jun

N-terminal kinase pathway and inducing apoptosis [310-312].

With the numerous and various roles of SET in cancer development, the oncoprotein may be a viable target for discovering a multi-pathway approach towards cancer therapy [313]. With some clues about this novel protein, we sought to determine

114

the exact role of SET in pancreatic tumor development and progression. Here, we elucidated a novel role for SET isoform 2 in inducing EMT through cadherin switching and revealed the molecular pathways involved. SET isoform 2 expression also had significant consequences on the metastatic ability of pancreatic cancer cells in vivo.

Materials and Methods

Cell Culture. The pancreatic cancer cell lines AsPC-1, BxPC-3, Capan-1, HPAF-

II, MIA PaCa-2, and PANC-1 were received from the American Type Culture Collection

(ATCC; Manassas, VA) cell bank. These cell lines were propagated, expanded, and frozen immediately after arrival. The cells revived from the frozen stock were used within 10-20 passages, not exceeding a period of 2-3 months. The ATCC uses morphological, cytogenetic, and DNA profile analyses for characterization of cell lines.

Human pancreatic ductal epithelial (HPDE) cells were kindly received from Dr. Ming

Tsao of the Ontario Cancer Institute (Toronto, Canada). The L3.6pl cell line was kindly received from Dr. Isiah D. Fidler at The University of Texas MD Anderson Cancer

Center (Houston, TX). Both HPDE and L3.6pl cell lines were handled as other cell lines and were genotyped by DNA fingerprinting (PowerPlex 16; Promega, Madison, WI) as per the manufacturer’s instructions. The growth conditions of all cell lines were performed as described previously [219].

Reagents and Antibodies. Fetal bovine serum (FBS), 4′,6′-diamidino-2- phenylindole (DAPI), propidium iodide, ethylene glycol bis(2-aminoethyl ether) tetraacetic acid (EGTA), phenylmethanesulfonyl fluoride (PMSF), N-ethyl maleimide

(NEM), sodium orthovanadate (Na2VO4), and iodoacetamide were obtained from Sigma

115

Aldrich (St. Louis, MO). The bicinchoninic acid (BCA) protein assay reagent and West

Pico Chemiluminiscent substrate were from Pierce Chemical (Rockford, IL). Fluorescent anti-fade mounting reagent was obtained from Molecular Probes (Invitrogen, Life

Technologies, Carlsbad, CA). Plasticware for cell culture was obtained from Corning

(Corning, NY). All cell culture media were purchased from Mediatech (Manassas, VA) except for the human keratinocyte basal medium which was procured from Molecular

Probes.

The anti-human goat polyclonal I2PP2A (E-15), rabbit polyclonal N-cad (H-63) for immunocytochemistry, c-Jun (H-79), and JNK2 (N-18), and mouse monoclonal JNK1

(F-3) and p-c-Jun (KM-1) antibodies were obtained from Santa Cruz Biotechnology

(Santa Cruz, CA). The anti-human rabbit polyclonal E-cad for Western blotting

(ab53033), SET isoform 1 (ab97596), SET isoform 2 (ab1183), and AP-1+2β (ab97434) antibodies were from Abcam (Cambridge, MA). The anti-human rabbit polyclonal

SPARC (5420) and Na+/K+ ATPase (3010) and rabbit monoclonal Slug (C19G7; 9585), phospho-SAPK/JNK (Thr183/Tyr185; 81E11; 4668), vimentin (D21H3; 5741), and lamin B1 (D9V6H; 13435) antibodies were purchased from Cell Signaling Technology

(Danvers, MA). The anti-human mouse N-cad for Western blotting (610920) and Rac1

(610651) antibodies were obtained from BD Biosciences (San Jose, CA). The mouse monoclonal anti-β-actin antibody (A1978) was purchased from Sigma-Aldrich. The rabbit polyclonal anti-HA antibody and HRP-conjugated secondary antibodies were from

Bethyl Laboratories (Montgomery, TX). The anti-human mouse monoclonal E-cad antibody for immunocytochemistry was kindly provided by Dr. Parmender Mehta of the

University of Nebraska Medical Center (Omaha, NE).

116

Tumor RNA and Tissues. Total RNA from 19 PDAC tissues and 4 normal pancreatic tissues were acquired from OriGene (Rockville, MD). Five PDAC tissue samples along with their matched normal adjacent tissues were procured from the

National Disease Research Interchange (Philadelphia, PA) [219]. The NDRI obtains written consents from the sources. The procurement and use of these human tissues was done in accordance with the University of Georgia Institutional Review Board. The

Board has determined that the use of human biological tissues in this research does not meet the criteria for research involving human subjects per 45 CFR 46.102, and therefore does not require human subject approval by the Board.

Quantitative Real-time PCR, Western Blot Analysis, Immunocytochemistry,

Cellular Proliferation Assay, and Colony Formation Assay. These procedures were performed as described earlier [173, 219, 221]. TaqMan primers and probes for SET

(Hs00853870_g1), SET isoform 1 (custom synthesized to nucleotides 28-49 of the CDS),

SET isoform 2 (Hs01118851_m1), CDH1 (Hs01013959_m1), CDH2 (Hs00983062_m1),

SPARC (Hs00234160_m1), SNAI2 (Hs00950344_m1), and GusB (Hs99999908_m1) were obtained from Applied Biosystems (Life Technologies).

Plasmid Construction and Retroviral Gene Transfer. The GIPZ Lentiviral Mouse

SET shRNA construct (Thermo Scientific, Waltham, MA) was obtained and utilized as previously described [173]. The pcDNA-SET-FLAG-HA construct was obtained from

Dr. Judy Lieberman via AddGene (Plasmid 24998). The insert was then subsequently cut out of the pcDNA3.1 vector using BamHI and XhoI and ligated to the pLNCX2 vector at the BglII and SalI sites. The empty vector was used as a control. Transfection into the packaging cell line, Phoenix-H, and subsequent infection of the viral particles into

117

PANC-1 were performed as described earlier [219]. Cells were selected and maintained with G418 (800 and 200 µg/ml, respectively).

Cell Surface Protein Isolation. Isolation of the cell surface fraction of proteins was conducted using the Pierce® Cell Surface Protein Isolation Kit (Thermo Scientific) as per the manufacturer’s instructions and as previously described [219]. Briefly, cells were washed twice with ice-cold PBS and subsequently biotinylated with 0.25 mg/ml

Sulfo-NHS-SS-Biotin for 30 minutes on an orbital shaker at 4°C. The reaction was quenched, and cells were collected by scraping followed by centrifugation. Cells were lysed with 500 µl of Lysis Buffer containing a protease inhibitor cocktail (Roche) for 30 minutes on ice with intermittent homogenization using a 23G needle and syringe. A sample of the total lysate was retained for analysis of total proteins. The biotinylated proteins were then bound to immobilized streptavidin-agarose beads (NeutrAvidin

Agarose slurry) by a one-hour incubation at room temperature using an end-over-end rotator. The unbound proteins were then collected by centrifugation of the column at

1,000 x g for 2 minutes and retained for analysis of intracellular proteins. The biotinylated proteins were eluted form the beads using SDS-PAGE buffer containing 50 mM DTT. The collected samples were then separated on SDS-PAGE for immunoblot analysis.

Wound Healing Assay. 30x104 cells were seeded in a 6 cm dish and grown until

90-95% confluency. A single scratch across the dish was then made using a 200 µl pipette tip. Images of the wound were taken at 0, 24, 48, and 72 h at 8x magnification using a Nikon AZ100 Multizoom microscope.

118

EMT PCR Array Analysis. The human Epithelial to Mesenchymal Transition RT2

ProfilerTM PCR Array (PAHS-090ZC-2) was purchased from Qiagen (Venlo,

Netherlands). The recommended RT2 First Strand Kit (330401) and RT2 SYBR® Green

ROCTM qPCR Mastermix (330520) were used as per the manufacturer’s instructions.

Transwell Assay. 2.5x104 cells in serum-free medium were seeded into the upper chamber of a 24-well transwell insert with 8.0 μm PET membrane pores and a BD

BioCoatTM MatrigelTM coating (Corning). Bottom wells were filled with complete medium. Cells were allowed to invade through the pores for 48 h under cell culture conditions. After incubation, the invaded cells that adhered to the bottom of the membrane were then fixed with methanol and stained with crystal violet. Cells from the upper surface of the filter were removed by scraping with a cotton swab. The number of cells that penetrated the membrane was determined by counting the mean cell number of five randomly selected high-power fields.

Animal Studies. All animal experiments were performed in accordance with the

Animal Care and Use Procedures at the University of Georgia, and the experimental protocol was approved by the Institutional Animal Care and use Committee (IACUC) at the University of Georgia. Immunocompromised mice were maintained in sterile conditions, housed four to a cage, and allowed to acclimate for one week before the start of the study. Food and water were provided ad libitum at all times. Mice were handled under aseptic conditions including the wearing of gloves, gowns, and shoe coverings.

Eight-week old female NU/J athymic nude mice were purchased from The

Jackson Laboratory (Bar Harbor, ME). The mice underwent an established surgical method for an orthotopic injection into the pancreas [314]. In brief, mice were

119

anesthetized with ketamine-xylazine (100 and 20 mg/kg, respectively), and the abdominal skin and muscle layers were incised (approximately 1 cm) slightly medial to the splenic silhouette. The pancreas was gently retracted and positioned to allow for the injection of

5x106 cells in 50 µl of sterile HBSS containing 1% (v/v) Matrigel. After solidification of the Matrigel (approximately 2 min), the pancreas was placed back into the abdominal cavity, and both the muscle and skin layers were closed with interrupted sutures.

Following recovery from surgery, mice were monitored daily for 45 days.

As per the protocol, tumor volumes were measured and calculated as the mean of the three dimensions (length x width x depth). Signs of macrometastases were observed, and the primary tumor, liver, and spleen were harvested for micrometastic identification.

All tissues were snap frozen, sectioned for slide preparation, and examined histologically.

Statistical Analysis. The student's t test was used to identify significant differences, and each experiment was repeated at least three times. Unless otherwise indicated, p<0.05 and p<0.01 compared with control conditions were represented by one and two asterisks, respectively.

Results

SET isoform 2 is highly overexpressed in poorly-differentiated pancreatic cancer cells. Since our earlier study showed involvement of SET in pancreatic cancer cellular growth suppression [173], we evaluated in further detail the role of SET in pancreatic cancer tumor progression in this study. We began by identifying total SET expression in a panel of pancreatic cancer cell lines which we have previously categorized based on drug sensitivity and extent of differentiation [219]. Total SET expression was

120

investigated using a TaqMan probe for a 100 nt region in exon 8 as well as an antibody

(E-15) which binds to an internal region of SET, both tools common to all SET isoforms.

The results showed total SET transcript and protein levels increased in the majority of pancreatic cancer cell lines (4/7 at the transcript level; 5/7 at the protein level) compared with normal human pancreatic ductal epithelial cells (HPDE) (Fig. 5.1B-C). In particular, significant increases were seen in the highly metastatic L3.6pl (transcript) as well as the poorly-differentiated PANC-1 (protein) and MIA PaCa-2 (transcript and protein).

To investigate the precise isoform involvement, we utilized additional TaqMan probes: one custom synthesized to probe a 22 nt region (nts 28-49 of the coding sequence) in exon 1 specific to isoform 1 and another pre-synthesized to probe for a 125 nt region (nts 68-193 of the coding sequence) in exons 1-2 specific to isoform 2.

Likewise, antibodies for specific isoforms were used: one with an epitope for amino acids

1-11 of isoform 1 (97596) and another with an epitope for amino acids 3-18 of isoform 2

(1183). The protein levels of isoforms 3 and 4 were not able to be identified since commercially available antibodies are not available; furthermore, qRT-PCR results demonstrated them to be minor isoforms (i.e., expressed at much lower levels than isoforms 1 and 2) with no significant change between normal and cancerous cell lines

(data not shown). Compared with normal HPDE, SET isoform 1 was significant increased in four cell lines at the transcript level but only in one cell line (L3.6pl) at the protein level (Fig. 5.1B-C). While SET isoform 2 was significantly increased in only three cell lines at the transcript level (L3.6pl, Capan-1, and PANC-1), it was significantly increased in all of the cancerous cell lines at the protein level compared with normal

HPDE (Fig. 5.1B-C).

121

Figure 5.1. SET isoform 2 is highly overexpressed in pancreatic cancer cells. A. SET isoforms 1 and 2 only differ at the N-terminal end. Both isoforms possess a putative region towards the N-terminal end for PP2A inhibitory activity and protein dimerization, a nuclear localization signal sequence in the middle region, and a putative region towards the C-terminal end for chromatin remodeling activity. The two isoforms differ only in the first 38 (isoform 1) and 25 (isoform 2) amino acids of the protein. B. SET transcript levels (all isoforms, isoform 1, and/or isoform 2) are significantly increased in the L3.6pl, Capan-1, PANC-1, and MIA PaCa-2 cell lines compared with normal HPDE. C. Protein levels of SET isoform 2 were significantly increased in all the cancerous cell lines. Total SET was increased in all but one cell line, and SET isoform 1 showed no significant increases except for L3.6pl. Whole cell lysates (50 µg) were subjected to Western blotting analysis. β-actin, the internal loading control, is shown with a representative blot. Bars, SD. n=3. *p<0.05, **p<0.01.

In summary, these results suggest that SET isoform 2 is overexpressed in at least a subset of pancreatic cancer cell lines, particularly those that are poorly-differentiated

(PANC-1 and MIA PaCa-2) or highly metastatic (L3.6pl) [315-317]. Further, the increase

122

in total SET identified in the cancerous samples is likely due to an increase in isoform 2, and not isoform 1, expression.

SET isoform 2 is localized in the nucleus and at the cell surface in poorly- differentiated pancreatic cancer cells. SET was originally identified as a nuclear oncoprotein. However, recent studies have also shown SET isoform 2 to be able to shuttle between the nucleus and cell surface in a Rac1-dependent manner to facilitate cellular migration [304, 305]. Since we found SET isoform 2 overexpressed in many metastatic pancreatic cancer cell lines, we further investigated the localization of SET in the panel of cells. Conducting immunocytochemical analysis using the SET E-15 antibody, the distribution of SET was found to be cell type-dependent. In normal HPDE and two well- differentiated pancreatic cancer cell lines (L3.6pl and Capan-1), the protein was predominantly nuclear (Fig. 5.2A). In contrast, SET was located in the nucleus, cytoplasm, and cell surface in poorly-differentiated PANC-1 and MIA PaCa-2 (Fig.

5.2A).

To investigate differences in isoforms, we used the SET isoform 1 (97596) and isoform 2 (1183) antibodies for immunocytochemical analysis. SET isoform 1 was found to be only nuclear in all cell lines, whereas SET isoform 2 was mainly nuclear in HPDE and other well-differentiated cell lines and mainly at the cell surface in PANC-1 and MIA

PaCa-2 (Fig. 5.2A). Co-localization analysis clearly showed isoform 1 to be predominantly in the nucleus and isoform 2 to be predominantly at the cell surface (Fig.

5.2A). To further corroborate these findings, we subjected whole cell lysates for cellular fractionation. With the nuclear and cell surface fractions separated, Western blotting analysis confirmed the predominant nuclear localization of SET isoform 1 and

123

Figure 5.2. SET isoform 2 is localized in the nucleus and at the cell surface in poorly-differentiated pancreatic cancer cells. A. The distribution of the two SET isoforms is cell type-dependent. SET isoform 1 (green) is nuclear in all cell lines, while isoform 2 (red) is mainly nuclear in HPDE, L3.6pl, and Capan-1 and mainly at the cell surface in PANC-1 and MIA PaCa-2. Merging with DAPI (blue) clearly indicates the locations of each isoform as well as their colocalization in HPDE cells (yellow). B. SET isoform 1 is predominantly located in the nucleus, while SET isoform 2 is predominantly located in the cytosol and at the cell surface. The nuclear fraction was separated from whole cell lysates and, along with the cytosolic fraction, subjected to Western blotting analysis. β-actin was used as the internal loading control, while lamin B1 was used as the nuclear loading control. n=3.

124

predominant cell surface localization of SET isoform 2 (Fig. 5.2B). These results illustrate the cell surface localization of SET isoform 2 in poorly-differentiated pancreatic cancer cell lines.

SET isoform 2 induces EMT and promotes growth, migration, and invasion of pancreatic cancer cells. To identify the role of SET isoform 2 in poorly-differentiated pancreatic cancer cells, we retrovirally overexpressed SET isoform 2 in the inherently epithelial PANC-1 (Fig. 5.3A). We also knocked down endogenous SET expression using an shRNA construct against mouse SET, which has 85 and 86% protein homology to human SET isoforms 1 and 2, respectively, in the inherently mesechymal MIA PaCa-2

(Fig. 5.3A). Compared with controls, PANC-1 cells with SET overexpression (PANC-1-

SET-HA) were found to have a more mesenchymal morphology, while MIA PaCa-2 cells with SET knockdown (MIA PaCa-2-SET-shRNA) were found to have a more epithelial morphology (Fig. 5.3B). Furthermore, expression of vimentin and the formation of actin stress fibers were also found to increase with increasing SET expression and mesenchymal morphology (Fig. 5.3A-B), suggesting the induction and progression of

EMT with SET isoform 2 expression. With further investigation of the effects of SET isoform 2 on EMT, we found increased SET expression to promote cellular growth, colony formation, migration, and invasion in PANC-1 (Fig. 5.3C-F). Consistently, knockdown of SET isoform 2 in MIA PaCa-2 reverted these EMT characteristics (Fig.

5.3C-F).

SET isoform 2 promotes cadherin switching from E-cadherin to N-cadherin to undergo EMT. With the arrangement of the pancreatic cell panel in decreasing order of differentiation (HPDE to MIA PaCa-2), we found SET to moderately correlate in the

125

Figure 5.3. SET isoform 2 induces EMT and promotes growth, migration, and invasion of pancreatic cancer cells. A. Overexpression of SET isoform 2 in PANC-1 increases vimentin protein levels while knockdown of SET in MIA PaCa-2 decreases vimentin levels. Whole cell lysates (50 µg) were subjected to Western blotting analysis. β-actin, the internal loading control, is shown with a representative blot. B. SET isoform 2 expression induces EMT. Cells with overexpression of SET have a more mesenchymal morphology, and cells with SET knockdown have a more cuboidal, epithelial morphology. SET isoform 2 expression also correlated with increased actin stress fibers (red). DAPI is shown (blue). C. SET isoform 2 expression promotes cellular proliferation. Growth was measured using a colorimetric assay. D. Expression of SET isoform 2 promotes cellular proliferation and colony formation. Cells were grown for 96 h. E. SET isoform 2 expression promotes migration. Scratched wounds were allowed to heal for up to 72 h. F. Expression of SET isoform 2 promotes invasion. Cells were allowed to invade through the pores of a MatrigelTM-coated transwell insert for 48 h. Bars, SD. n=3. *p<0.05, **p<0.01.

reverse order (i.e., SET being highly expressed in the less differentiated cells). Therefore, we hypothesized that SET expression at the cell surface influenced the expression of cadherins to control EMT. To test this hypothesis, we first verified E-cad and N-cad

126

transcript levels in the pancreatic panel and found that E-cad levels were indeed higher in well-differentiated HPDE, BxPC-3, HPAF-II, L3.6pl, and Capan-1, whereas N-cad levels were highest in poorly-differentiated PANC-1 (Fig. 5.4A). Immunocytochemical analysis of N-cad expression showed the protein co-localizing with SET at the cell surface in

Figure 5.4. SET isoform 2 promotes cadherin switching from E-cadherin to N- cadherin to undergo EMT. A. E-cad and N-cad are oppositely expressed at the transcript level in the pancreatic cell panel. E-cad levels are higher in well- differentiated cell lines including HPDE, while N-cad levels are higher in poorly- differentiated cell lines. B. SET isoform 2 expression (red) co-localized with N-cad expression (green) at the cell surface (yellow). C. Cell surface expression of SET isoform 2 correlates with N-cad expression and the lack of E-cad expression at the cell surface. At the protein level, E-cad expression in cells decreases with poorer differentiation, and N-cad expression is only found in three cell lines. The cell surface fraction of whole cell lysates was isolated through biotinylation. D. SET isoform 2 expression correlates with increased N-cad (green) and decreased E-cad (red) expressions. E. Overexpression of SET in PANC-1 increases N-cad and decreases E- cad protein levels, while knockdown of SET in MIA PaCa-2 shows the opposite. Whole cell lysates (50 µg) were subjected to Western blotting analysis. β-actin, the internal loading control, is shown with a representative blot. Bars, SD. n=3. *p<0.05, **p<0.01.

127

PANC-1 (Fig. 5.4B).

To identify E- and N-cad expression at the cell surface, we isolated cell surface fractions of lysates and subjected them to Western blotting analysis. This clearly showed

E-cad expression decreasing with poorer differentiation and N-cad expression only in

HPAF-II, Capan-1, and PANC-1 (Fig. 5.4C). While HPAF-II, Capan-1, and PANC-1 expressed both cadherins, MIA PaCa-2 expressed neither cadherin (Fig. 5.4C). With the cell surface expressions of total SET, SET isoform 1, and SET isoform 2, it was also found that SET isoform 2 was expressed in N-cad-expressing cell lines (i.e., HPAF-II,

Capan-1, and PANC-1) and the cell surface cadherin-null cell line (i.e., MIA PaCa-2)

(Fig. 5.4C). Some cell surface expression of SET isoform 1 was also noticed in PANC-1 and MIA PaCa-2 (Fig. 5.4C). These studies suggest that the presence of SET at the cell surface may induce cadherin switching from E-cad to N-cad.

To confirm this, we examined SET overexpression or knockdown, as aforementioned, in a cell line expressing both E- and N-cad (PANC-1) and neither E- nor

N-cad (MIA PaCa-2). Overexpression of SET isoform 2 in PANC-1 increased N-cad expression robustly and decreased E-cad (Fig. 5.4D-E). Similarly, knockdown of SET in

MIA PaCa-2 restored E-cad expression (Fig. 5.4D-E). Expressions of both cadherins were seen at the cell surface (Fig. 5.4D). These studies confirm that expression of SET isoform 2 at the cell surface promotes cadherin switching for the induction of EMT in pancreatic cancer cells.

SET-induced cadherin switching involves the Rac1/JNK/c-Jun/AP-1 and

SPARC/Slug signaling pathways. Since N-cad has been shown to act downstream of

Rac1 and JNK1 [308, 318], we hypothesized that SET would act through this pathway to

128

Figure 5.5. SET-induced cadherin switching involves the Rac1/JNK/c-Jun/AP-1 and SPARC/Slug signaling pathways. A. SET isoform 2 expression alters the cellular organizations and expressions of players in the Rac1/JNK/c-Jun/AP-1 pathway. B. SET isoform 2 expression alters the protein levels of players in the Rac1/JNK/c-Jun/AP-1 pathway. Oppositely, a reduction in SET led to decreases in Rac1, JNK1, phospho- JNK, c-Jun, phospho-c-Jun, and AP-1 levels. Whole cell lysates (100 µg) were subjected to Western blotting analysis. GAPDH, the internal loading control, is shown with a representative blot. C. In an RT-PCR profiler array for 84 EMT-related genes, 18 genes were up- or down-regulated >0.5-fold in SET knockdown cells compared with WT. The genes with the greatest fold-change (down-regulated) included SPARC and SNAI2. D. In independent RT-PCR experiments, significant decreases in both SPARC and SNAI2 transcript levels were confirmed in MIA PaCa-2-SET-shRNA, while transcript levels of the two candidates were increased in PANC-1-SET-HA. E. At the protein level, Slug was also decreased in SET knockdown cells. SPARC was not detectable via Western blotting. Whole cell lysates (50 µg) were subjected to Western blotting analysis, and β-actin was used as the internal loading control. F. E-cad and Slug were oppositely expressed. Slug (red) is expressed in the nucleus, while E-cad (green) is expressed on the cell surface. Bars, SD. n=3. *p<0.05, **p<0.01.

inhibit N-cad expression. To test this hypothesis, we compared the expression levels of key players involved in this pathway in both PANC-1-SET-HA and MIA PaCa-2-SET-

129

shRNA compared with control. Along with an increase in SET at the cell surface, Rac1,

JNK1, c-Jun, phospho-c-Jun, and AP-1 were also increased in PANC-1-SET-HA (Fig.

5.5A-B). Only a minor change was observed for JNK2. Oppositely, with SET knockdown in MIA PaCa-2, Rac1 expression was decreased with significant reorganization as judged by immunocytochemical analysis (Fig. 5.5A). JNK1, c-Jun, phospho-c-Jun, and AP expressions were also reduced without significant changes in JNK2 (Fig. 5.5A-B). These results suggest that SET expression increases N-cad through the Rac1/JNK/c-Jun/AP-1 pathway.

In order to understand SET regulation of E-cad expression, we performed a PCR array for 84 genes known to be involved in EMT. These included, but were not limited to, genes involved in cell adhesion, migration, cellular differentiation, morphogenesis, development, growth, and proliferation. With the knockdown of SET in MIA PaCa-2, we found 9 genes significantly upregulated and 9 genes significantly downregulated (fold- change >3) with a maximal change in DSC2 (desmocollin 2; 76-fold downregulation).

Based on this analysis, we also found significant downregulation of both SPARC

(osteonectin; 74-fold) and SNAI2 (slug; 54-fold) (Fig. 5.5C), two transcriptional repressors of E-cad. These results were validated separately using individual TaqMan probes for SPARC and SNAI2 (Fig. 5.5D) as well as by Western blotting with protein- specific antibodies (Fig. 5.5E). Expression of E-cad at the cell surface in contrast to the expression of the transcription factor Slug in the nucleus was seen by immunocytochemical analysis (Fig. 5.5F). These results suggest that SET inhibition of E- cad expression is mediated through transcriptional repression of SPARC and SNAI2.

130

SET isoform 2 overexpression and cadherin switching were identified in patient- derived, poorly-differentiated pancreatic cancer tissues. To determine the clinical significance of the results, we looked at 19 pancreatic tumor tissues of varying degrees of differentiation and compared them with 4 normal pancreatic tissues. SET and N-cad were

Figure 5.6. SET isoform 2 overexpression at the cell surface and cadherin switching were identified in patient-derived, poorly-differentiated pancreatic cancer tissues. A. Both SET isoform 2 and N-cad are overexpressed at the transcript level in most pancreatic cancer tissue samples. SET isoform 1 mRNA was overexpressed in a few samples, while E-cad mRNA levels were mainly unchanged or decreased. B. SET transcript levels in the tissue samples correlated well with N-cad expression (R=0.8). C. SET isoform 1, SET isoform 2, and N-cad transcripts were overexpressed in all of the PDAC tissue samples. SET isoform 2 and N-cad were also overexpressed in approximately half of the NET tissue samples. Only a low percentage of tissue samples had a significantly increased amount of E-cad mRNA. D. SET isoform 2 transcripts were overexpressed in all five normal-tumor tissue pairs. SET isoform 1 and N-cad transcript levels either remained unchanged or decreased. PDAC, pancreatic ductal adenocarcinoma. ACC, acinar cell carcinoma. NET, neuroendocrine tumor. Bars, SD.

131

both increased while E-cad was decreased in the majority of tumor samples compared with normal (Fig. 5.6A). Furthermore, overexpression of SET correlated with increased

N-cad expression within the same tumor samples (R=0.8) (Fig. 5.6B). When the tumor samples were segregated by type, both SET isoforms as well as N-cad were overexpressed in all pancreatic ductal adenocarcinoma (PDAC) samples (Fig. 5.6C).

Both SET isoforms were also overexpressed in approximately half of the neuroendocrine tumor (NET) samples (Fig. 5.6C). To corroborate this further, we utilized five matched normal-tumor tissue pairs from five different patients and examined the expression of

SET as well as the cadherins. Once again, SET isoform 2 transcripts were significantly increased in all of the tumor samples (Fig. 5.6D).

SET isoform 2 expression promotes pancreatic tumor growth. The ability of SET isoform 2 expression to influence tumor progression was investigated using orthotopic mouse models of pancreatic cancer. Tumors were allowed to grow for 45 days, during which the injected mice gained weight and the control mouse did not (Fig. 5.7A). The injected mice gained approximately 1-2 g during the course of the experiment (Fig.

5.7A). Ultimately, tumors were identified in two PANC-1/LNCX2 mice (mice #3 and #4) and one PANC-1/SET+ mouse (mouse #2). Overall, we found primary tumor volume to increase with SET isoform 2 expression (Fig. 5.7B). The tumor expressing SET was larger, denser, and 86% heavier than the PANC-1/LNCX2 tumors (Fig. 5.7B). Tumor diameter also increased with SET expression, by approximately 47% or 5 mm (Fig.

5.7C). These results confirm that SET isoform 2 expression promotes tumor growth in vivo.

132

Figure 5.7. SET isoform 2 expression increases tumor volume. A. Mice with orthotopic tumor implantation gained weight post-surgery, whereas the control mouse did not. The injected mice gained approximately 1-2 g over 45 days. B. The mouse injected with PANC-1/SET+ cells developed a larger, denser, and heavier tumor than the PANC-1/LNCX2 mouse. C. SET isoform 2 expression led to a larger tumor in terms of diameter. The PANC-1/LNCX2 tumors were an average of approximately 5 mm smaller than the PANC-1/SET+ tumor.

Discussion

As a prerequisite for metastasis, the role of EMT in cancer progression is well- established. However, many aspects regarding the regulation of EMT remain unknown.

Overexpression of the SET oncoprotein has been observed in numerous cancers, and our previous study suggests its involvement in cellular differentiation and EMT [173]. In this study, we continued our investigation of SET and elucidated its role in EMT and pancreatic tumor progression.

133

Although there are four known isoforms of human SET, isoforms 1 and 2 are the most studied. The two isoforms differ only at the N-terminal end, away from areas relating to PP2A inhibitory activity or chromatin remodeling activity. In a panel of pancreatic cell lines, SET isoform 2 was found to be highly overexpressed. Interestingly, significant changes were seen in the highly metastatic L3.6pl and the poorly- differentiated PANC-1 and MIA PaCa-2. It was also evident that the isoform was expressed in the nucleus, in the cytoplasm, and at the cell surface rather than only in the nucleus and cytoplasm as traditionally described. While SET is often regarded as a nuclear oncoprotein, recent studies have shown its trafficking between the nucleus and cell surface in a Rac1-dependent manner [304, 305].

By modulating the expression of SET, we observed the promotion and regression of EMT in epithelial PANC-1 and mesenchymal MIA PaCa-2 cells, respectively. EMT was characterized by morphology, the expression of mesenchymal markers, growth, and colony formation, migration, and invasion abilities. While the relationship was not absolute, we noticed that SET isoform 2 was expressed at the cell surface in cell lines with N-cad cell surface expression and also a cell line with neither E-cad nor N-cad at the cell surface. Therefore, the mechanism behind SET regulation of EMT was identified to be cadherin switching from E-cad to N-cad. SET regulation of cadherin switching was determined to be two-fold: the Rac1/JNK/c-Jun/AP-1 pathway promoting N-cad expression and the SPARC/Slug pathway inhibiting E-cad expression. Significant upregulation of these genes with SET expression was identified using an EMT PCR profiling array; however, numerous additional EMT genes were found to be altered by

SET expression as well. Exactly how SET regulates these other genes remains unknown.

134

In a panel of patient-derived pancreatic cancer tissues, we confirmed cadherin switching and SET isoform 2 overexpression. However, changes were most consistent in the pancreatic ductal adenocarcinoma (PDAC) samples rather than the neuroendocrine tumor (NET) samples, suggesting a more predominant role of SET in exocrine rather than endocrine tumors.

In an orthotopic mouse model, we found SET isoform 2 expression to promote pancreatic tumor growth. With SET expression, the primary tumor developed larger, denser, and heavier than the primary tumors of the control counterpart. However, it is unknown whether SET also promotes pancreatic tumor progression. While macrometastases were not observed in the animals, further studies are needed to examine secondary sites for micrometastases. At the cellular level, the expressions and localizations of SET, N-cad, and E-cad also remain unidentified. Staining of tumor cryosections will be necessary.

This study elucidates a novel role for SET isoform 2 in inducing EMT through cadherin switching and revealed the molecular pathways involved. SET isoform 2 expression also had significant consequences on pancreatic primary tumor growth in vivo, suggesting it can be utilized as a biomarker for early diagnosis or for early prediction of metastasis, though additional studies are required. It is also unclear when during the tumorigenic process the activation of SET and the switch in cadherins occur. Moreover, the role of SET in directly determining chemosensitivity is not clear at this point, although the oncoprotein is known to regulate the let-7 miRNA which influences chemosensitivity [173].

135

SET’s involvement in EMT and cancer progression suggests that the oncoprotein may also be a viable target for discovering a multi-pathway approach towards cancer therapy. In addition to the genes and pathways examined in this study, SET reprogrammed a number of additional EMT genes, implicating its role as a master regulator. Furthermore, since the protein is part of the inhibitor of histone acetyltransferase (INHAT) complex [192], it is possible that it exerts a chromatin effect on EMT genes. More studies are required to understand this potential chromatin remodeling process. Since SET isoform 2 is expressed at the cell surface, it could be an ideal candidate for simple drug targeting, potentially allowing for anti-metastatic intervention in pancreatic cancer patients.

136

CHAPTER 6

DISCUSSION

Major Conclusions and Future Directions

Although there are several other determinants, we have identified that E-cad enhancement of gemcitabine efficacy occurs at least in part through hENT1. Expression of E-cad in a cadherin-null cell line increased hENT1 expression, activity, and stability, leading to increased cellular transport and cytotoxicity of gemcitabine. At this point, it remains unclear whether E-cad expression also increases the recruitment, trafficking, and basolateral membrane insertion of hENT1. Further, it is unknown whether N-cadherin, which is highly implicated in EMT and metastasis, plays an opposing role in hENT1- mediated gemcitabine efficacy. Identifying the direct mechanism between E-cad and hENT1 as well as the status of N-cad in this scenario are two of our future aims.

The major conclusion from the second study is that co-expression of hCNT1 with

Cx32 increased gemcitabine accumulation within a population of heterogeneous cells.

Although it did not lead to a significantly enhanced gemcitabine response, there is potential for increasing the bystander effect through favorable manipulations of certain

NTs and Cxs (GJs and hemichannels). However, further studies are needed to determine which specific types and combinations of NTs and GJs are the most effective in doing so.

One possibility is the manipulation of Cx26 which has been recently shown to significantly enhance the gemcitabine bystander effect in pancreatic cancer [174]. Our

137

future studies would include the use of co-culture experiments (i.e., using mixed populations of cells expressing various NTs and GJs) to mimic the heterogeneity of tumors and for testing the potential bystander effect.

In the third project, we have developed in vitro optimization procedures for the

DZNep-gemcitabine combination exclusively against human pancreatic cancer cells. By altering structure, dose, exposure, and formulation, these results demonstrate potential for the use of this epigenetic-chemotherapeutic combination approach for further studies in the treatment of pancreatic cancer. Future directions will be aimed at identifying and targeting the miRNA changes induced by DZNep. Since miRNAs have diverse targets, preliminary studies by a current graduate student using miRNA knockdown models revealed unprecedented phenotypes. Future studies in the lab will be focused on the delivery of miRNAs in animals.

The last study elucidated the role of SET isoform 2 in promoting EMT and metastasis. While we found SET to influence cadherin expression via the Rac1/JNK//c-

Jun/AP-1 and SPARC/Slug pathways, numerous additional EMT genes were found to be modulated by the oncoprotein as well. Additional studies will be necessary to clarify exactly how SET regulates these genes. Furthermore, the direct role of SET in determining chemosensitivity remains unclear. Future aims for this project include revealing additional genes and pathways altered by SET and the exact connection between SET, let-7, and chemosensitivity.

138

Therapeutic Challenges in Overcoming Chemoresistance 6

Despite the abundance of common and novel chemotherapies, gemcitabine remains the superior drug of choice for the treatment of pancreatic cancer due to its ability to extend survival, although only by a mere few weeks, and improve the quality of life with better tolerance. As one of the many enduring nucleoside analog drugs, gemcitabine has been predicted to remain as the drug of choice for pancreatic cancer for the near future even though its effects are truly suboptimal. In spite of considerable efforts to improve the drug of choice, several other approaches (e.g., targeting stromal pathways in tumors) for overcoming chemoresistance in pancreatic cancer have been underway, bringing with them the new challenges they pose.

It is inarguable the abundance of consistent genetic and molecular alterations observed in panIN, primary, and metastatic pancreatic neoplasms. Numerous large-scale genomic sequencings and proteomic screenings have been conducted with novel discoveries identified from the experiments. However, the direct applicability of these potential biomarkers in patient populations remains unknown. For example, perhaps the most recognized alteration noted in pancreatic cancer is the mutation of Kras. Although targeting this player has thus far been unsuccessful due to the redundancy of cellular signaling pathways and perhaps also the lessened dependency of cells on this pathway as the disease progresses, a novel small molecule inhibitor of the Ras oncoprotein may develop with promise [319]. While the utilization and targeting of such genetic and molecular modifications remain a challenge in overcoming chemoresistance, the identification of precise biomarkers is vital since they may be used for targeting tumor

6 SW Hung, H Mody, and R Govindarajan. 2012. Overcoming Nucleoside Analog Chemoresistance of Pancreatic Cancer: A Therapeutic Challenge. Cancer Letters. 320: 138-149. Reprinted here with permission of the publisher. 139

cells and for guiding therapy. Identifying cell surface markers could allow for the targeting of tumor cells alone, leaving surrounding healthy cells in the heterogeneous tumor microenvironment to thrive. Also, utilizing these markers could stratify patient groups for individualized and tailored treatments. In addition to earlier diagnosis, biomarkers also have the potential for therapeutic monitoring of patients, especially since pancreatic cancer is known to undergo genetic mutations at a high frequency and rapid changes in disease biology. Currently, challenges arise in both identifying appropriate biomarkers for use in patient populations as well as in the practicality of screening a patient in the narrow time period between initial diagnosis and start of treatment. Clinical trial inconsistencies pose another challenge as the patient populations are often low in number and heterogeneous (e.g., with some patients already under treatment and some chemo-naïve). Well-annotated, patient-derived tumor samples remain a poorly available resource for further research.

In addition to identifying genetic and molecular alterations, it is crucial to understand when the changes occur during the progression of the disease. Pancreatic cancer, like any other solid tumor, consists of multiple defined stages, yet all patients are treated with the same drug (mainly due to diagnoses occurring at advance-stage disease).

With better comprehension of the changes that occur throughout the cancer, including variations between the primary and secondary tumors and the primary and recurrent tumors, tailored drug therapies can be developed. Furthermore, management of genetic and molecular changes and disease monitoring are vital with disease progression.

Identification of genetic and molecular signatures that can segregate pancreatic tumors into categories, such as those from Collison et al. (i.e., classical, quasimesenchymal, and

140

exocrine-like) [320], could lead to separate and appropriate treatment and management strategies. Therefore, it is imperative to understand how the chemotherapeutic drugs exert their efficacy as this information may be particularly useful when considering the heterogeneity of pancreatic tumors.

Pancreatic tumors are well-known for their dense desmoplastic stroma and sparse, collapsed vasculature, all characteristics impeding effective chemotherapy. Although the tumors consist of both stromal and epithelial tissues, most earlier studies have focused mainly on epithelial components. However, currently, an increased number of studies are investigating the stroma of pancreatic tumors, including methods to target eminent stromal pathways such as Hedgehog, TGF-β, and Wnt/PI3K as well as deplete stromal tissues in order to stimulate angiogenesis and increase drug delivery. The former approach (targeting the TGF-β pathway using the histone methylation reversal agent,

DZNep) is currently under investigation by a fellow graduate student in the lab, while the latter is already being studied in numerous Phase I-III clinical trials. Furthermore, our collaborative work with engineered nanoparticles attempts to overcome the many drug delivery and exposure limitations in pancreatic cancer. Enhanced knowledge of the heterogeneity of pancreatic tumors may lead to improved methods for overcoming chemoresistance due to tumor diversity. Furthermore, accurate methods for diagnosis as well as disease and therapeutic monitoring can be developed.

Significant advances have been made in the study of pancreatic cancer, including the development of in vitro cell lines and in vivo mouse models. Although a plethora of pancreatic cancer cell lines exist for study, their ability to mimic the wide range of pancreatic tumor characteristics is lacking. For example, Collison et al. identified cells

141

lines that represented the classical and quasimesenchymal subtypes of pancreatic cancer but not the exocrine-like subtype [320]. Although the use of cell lines for cancer study remains valid, it is important to note the potential flaws of the model systems.

Furthermore, although novel mouse models such as the orthotopic and K-rasLSL-

G12D/+;Trp53LSL-R172H/+;Cre (KPC) have been developed, there remains the issue of interspecies differences in the development of chemoresistance. It is not yet fully determined how well these murine models represent human pancreatic tumors in both development and treatment outcomes. For example, the orthotopic model does not generate enough stroma to mimic human tumors, and the KPC model does not contain the entire spectrum of genetic mutations observed in human tumors. Although developmental issues are evident, the relative significance of these differences on influencing therapeutic outcome remains unknown. Ultimately, although great strides have been made with the development and use of novel mouse models, there remains a lack of model that has the ability to completely recapitulate human disease.

In addition to identifying new and far more improved drugs, an integrated approach is needed to overcome chemoresistance. For example, the rational combinations of the aforementioned therapies (Fig. 1.2), such as epigenetic agents, gene therapy, antibodies, and signaling inhibitors, along with the use of chemotherapeutics still need further explication. Our work with combining the epigenetic agent, DZNep, with the chemotherapeutic, gemcitabine, attempts to delve into this unknown. Still, the direct applicability of these therapies in combination (e.g., precise biomarker combinations to target cancer cells of all stages) in guiding treatment remains unknown. Furthermore, disconnect between benchtop, clinical, and regulatory science propels an additional,

142

complex challenge in overcoming pancreatic cancer. Currently, numerous clinical trials remain at early investigational stages and are difficult to escalate to the level of approved patient-use. Although divisions between specialized fields still exist, the process of streamlining translational research is gaining momentum, and its persistence may ultimately lead to overcoming the therapeutic challenges of pancreatic cancer.

143

REFERENCES

1. Cancer Facts & Figures 2014. 2014, American Cancer Society: Atlanta.

2. Cancer Basics: Lifetime Risk of Developing or Dying from Cancer. 2013; Available from: http://www.cancer.org/cancer/cancerbasics/lifetime-probability- of-developing-or-dying-from-cancer.

3. Stage IV Pancreatic Cancer Treatment. 2014; Available from: http://www.cancer.gov/cancertopics/pdq/treatment/pancreatic/HealthProfessional/ page7.

4. Oettle, H., Neuhaus, P., et al., Adjuvant chemotherapy with gemcitabine and long- term outcomes among patients with resected pancreatic cancer: the CONKO-001 randomized trial. Journal of the American Medical Association, 2013. 310(14): p. 1473-81.

5. Burris, H.A., Moore, M.J., et al., Improvements in survival and clinical benefit with gemcitabine as first-line therapy for patients with advanced pancreas cancer: a randomized trial. Journal of Clinical Oncology, 1997. 15(6): p. 2403.

6. Stathis, A. and Moore, M.J., Advanced pancreatic carcinoma: current treatment and future challenges. Nature Reviews Clinical Oncology, 2010. 7(3): p. 163-172.

7. Wang, Z., Li, Y., et al., Pancreatic cancer: understanding and overcoming chemoresistance. Nature Reviews Gastroenterology and Hepatology, 2011. 8(1): p. 27-33.

8. Singh, A. and Settleman, J., EMT, cancer stem cells and drug resistance: an emerging axis of evil in the war on cancer. Oncogene, 2010. 29(34): p. 4741- 4751.

9. Bhutia, Y.D., Hung, S.W., et al., CNT1 Expression Influences Proliferation and Chemosensitivity in Drug-Resistant Pancreatic Cancer Cells. Cancer Research, 2011. 71(5): p. 1825.

10. Farrell, J., Garcia, M., et al., Human ENT1 Is Predictive of Response in Patients With Pancreatic Cancer Treated With Gemcitabine: Results From the Rtog 9704 Prospective Randomized Trial. Pancreas, 2007. 35(4): p. 401.

11. Giovannetti, E., Del Tacca, M., et al., Transcription analysis of human equilibrative nucleoside transporter-1 predicts survival in pancreas cancer patients treated with gemcitabine. Cancer Research, 2006. 66(7): p. 3928.

144

12. Maréchal, R., Mackey, J.R., et al., Human equilibrative nucleoside transporter 1 and human concentrative nucleoside transporter 3 predict survival after adjuvant gemcitabine therapy in resected pancreatic adenocarcinoma. Clinical Cancer Research, 2009. 15(8): p. 2913.

13. Robins, M.J., Peng, Y., et al., Improved Syntheses of 5′-S-(2-Aminoethyl)-6-N-(4- nitrobenzyl)-5′-thioadenosine (SAENTA), Analogues, and Fluorescent Probe Conjugates: Analysis of Cell-Surface Human Equilibrative Nucleoside Transporter 1 (hENT1) Levels for Prediction of the Antitumor Efficacy of Gemcitabine. Journal of Medicinal Chemistry, 2010.

14. Spratlin, J., Sangha, R., et al., The absence of human equilibrative nucleoside transporter 1 is associated with reduced survival in patients with gemcitabine- treated pancreas adenocarcinoma. Clinical Cancer Research, 2004. 10(20): p. 6956.

15. Dean, M., Fojo, T., et al., Tumour stem cells and drug resistance. Nature Reviews Cancer, 2005. 5(4): p. 275-284.

16. Visvader, J.E. and Lindeman, G.J., Cancer stem cells in solid tumours: accumulating evidence and unresolved questions. Nature Reviews Cancer, 2008. 8(10): p. 755-68.

17. Easwaran, H., Tsai, H.C., et al., Cancer Epigenetics: Tumor Heterogeneity, Plasticity of Stem-like States, and Drug Resistance. Molecular Cell, 2014. 54(5): p. 716-727.

18. Bedi, U., Mishra, V.K., et al., Epigenetic plasticity: A central regulator of epithelial-to-mesenchymal transition in cancer. Oncotarget, 2014. 5(8): p. 2016- 29.

19. Cheng, H., Shi, S., et al., microRNA signature for human pancreatic cancer invasion and metastasis. Experimental and Therapeutic Medicine, 2012. 4(2): p. 181-187.

20. Meng, F. and Wu, G., The rejuvenated scenario of epithelial-mesenchymal transition (EMT) and cancer metastasis. Cancer and Metastasis Reviews, 2012. 31(3-4): p. 455-67.

21. Tam, W.L. and Weinberg, R.A., The epigenetics of epithelial-mesenchymal plasticity in cancer. Nature Medicine, 2013. 19(11): p. 1438-49.

22. Fitzgerald, T.L. and McCubrey, J.A., Pancreatic cancer stem cells: Association with cell surface markers, prognosis, resistance, metastasis and treatment. Advances in Biological Regulation, 2014.

23. Iwatsuki, M., Mimori, K., et al., Epithelial-mesenchymal transition in cancer development and its clinical significance. Cancer Science, 2010. 101(2): p. 293-9.

145

24. Hermann, P.C., Huber, S.L., et al., Distinct populations of cancer stem cells determine tumor growth and metastatic activity in human pancreatic cancer. Cell Stem Cell, 2007. 1(3): p. 313-23.

25. Mani, S.A., Guo, W., et al., The epithelial-mesenchymal transition generates cells with properties of stem cells. Cell, 2008. 133(4): p. 704-15.

26. Hamada, S., Satoh, K., et al., Regulators of epithelial mesenchymal transition in pancreatic cancer. Frontiers in Physiology, 2012. 3: p. 254.

27. Infante, J.R., Jones, S.F., et al., A phase I, dose-escalation study of pomalidomide (CC-4047) in combination with gemcitabine in metastatic pancreas cancer. European Journal of Cancer, 2011. 47(2): p. 199-205.

28. Mackey, J.R., Mani, R.S., et al., Functional nucleoside transporters are required for gemcitabine influx and manifestation of toxicity in cancer cell lines. Cancer Research, 1998. 58(19): p. 4349.

29. Mackey, J.R., Yao, S.Y.M., et al., Gemcitabine transport in xenopus oocytes expressing recombinant plasma membrane mammalian nucleoside transporters. Journal of the National Cancer Institute, 1999. 91(21): p. 1876.

30. Nakano, Y., Tanno, S., et al., Gemcitabine chemoresistance and molecular markers associated with gemcitabine transport and metabolism in human pancreatic cancer cells. British Journal of Cancer, 2007. 96(3): p. 457-463.

31. Garcia-Manteiga, J., Molina-Arcas, M., et al., Nucleoside transporter profiles in human pancreatic cancer cells: role of hCNT1 in 2', 2'-difluorodeoxycytidine- induced cytotoxicity. Clinical Cancer Research, 2003. 9(13): p. 5000-5008.

32. Rauchwerger, D.R., Firby, P.S., et al., Equilibrative-sensitive nucleoside transporter and its role in gemcitabine sensitivity. Cancer Research, 2000. 60(21): p. 6075.

33. Murata, Y., Hamada, T., et al., Human equilibrative nucleoside transporter 1 expression is a strong independent prognostic factor in UICC T3–T4 pancreatic cancer patients treated with preoperative gemcitabine-based chemoradiotherapy. Journal of Hepato-Biliary-Pancreatic Sciences, 2011: p. 1-13.

34. Nishio, R., Tsuchiya, H., et al., Disrupted plasma membrane localization of equilibrative nucleoside transporter 2 in the chemoresistance to gemcitabine (dFdCyd) of human pancreatic cancer cells. Cancer Science, 2011. 102(3): p. 622-9.

35. Hu, X., Chen, W., et al., Downregulation of Human Equilibrative Nucleoside Transporter 1 by RNAi Enhances 5--fluorouracil Response in Pancreatic Cancer. Hepato-Gastroenterology, 2010. 57(104): p. 1567-1572.

146

36. Noma, B., Sasaki, T., et al., Expression of multidrug resistance-associated protein 2 is involved in chemotherapy resistance in human pancreatic cancer. International Journal of Oncology, 2008. 33(6): p. 1187.

37. Kruh, G.D., Guo, Y., et al., ABCC10, ABCC11, and ABCC12. Pflügers Archiv - European Journal of Physiology, 2007. 453(5): p. 675-684.

38. Hagmann, W., Jesnowski, R., et al., ATP-binding cassette C transporters in human pancreatic carcinoma cell lines. Pancreatology, 2008. 9(1-2): p. 136-144.

39. Hopper-Borge, E., Xu, X., et al., Human multidrug resistance protein 7 (ABCC10) is a resistance factor for nucleoside analogues and epothilone B. Cancer Research, 2009. 69(1): p. 178.

40. Rudin, D., Li, L., et al., Gemcitabine Cytotoxicity: Interaction of Efflux and Deamination. Journal of Drug Metabolism and Toxicology, 2011. 2(107): p. 1.

41. Huang, P. and Plunkett, W., Induction of apoptosis by gemcitabine. Seminars in Oncology, 1995. 22(4 Suppl 11): p. 19-25.

42. Ohhashi, S., Ohuchida, K., et al., Down-regulation of deoxycytidine kinase enhances acquired resistance to gemcitabine in pancreatic cancer. Anticancer Research, 2008. 28(4B): p. 2205.

43. Yao, J., Feng, F., et al., The mechanism of resistance to 2', 2'- difluorodeoxycytidine (gemcitabine) in a pancreatic cancer cell line. Chinese Journal of Oncology, 2005. 27(12): p. 721.

44. Kroep, J.R., Loves, W.J.P., et al., Pretreatment Deoxycytidine Kinase Levels Predict in Vivo Gemcitabine Sensitivity. Molecular Cancer Therapeutics, 2002. 1(6): p. 371.

45. Funamizu, N., Okamoto, A., et al., Is the resistance of gemcitabine for pancreatic cancer settled only by overexpression of deoxycytidine kinase? Oncology Reports, 2010. 23(2): p. 471-475.

46. Tang, K., Zhang, Z., et al., Enhancement of gemcitabine sensitivity in pancreatic cancer by co-regulation of dCK and p8 expression. Oncology Reports, 2011. 25(4): p. 963-970.

47. Sebastiani, V., Ricci, F., et al., Immunohistochemical and genetic evaluation of deoxycytidine kinase in pancreatic cancer: relationship to molecular mechanisms of gemcitabine resistance and survival. Clinical Cancer Research, 2006. 12(8): p. 2492.

48. Williams, T.K., Costantino, C.L., et al., pp32 (ANP32A) Expression Inhibits Pancreatic Cancer Cell Growth and Induces Gemcitabine Resistance by Disrupting HuR Binding to mRNAs. PloS One, 2010. 5(11): p. e15455.

147

49. Costantino, C.L., Witkiewicz, A.K., et al., The role of HuR in gemcitabine efficacy in pancreatic cancer: HuR up-regulates the expression of the gemcitabine metabolizing enzyme deoxycytidine kinase. Cancer Research, 2009. 69(11): p. 4567.

50. Shipley, L.A., Brown, T.J., et al., Metabolism and disposition of gemcitabine, and oncolytic deoxycytidine analog, in mice, rats, and dogs. Drug Metabolism and Disposition, 1992. 20(6): p. 849-855.

51. Heinemann, V., Xu, Y.Z., et al., Cellular elimination of 2′, 2′- difluorodeoxycytidine 5′-triphosphate: a mechanism of self-potentiation. Cancer Research, 1992. 52(3): p. 533.

52. Mercier, C., Raynal, C., et al., Toxic death case in a patient undergoing gemcitabine-based chemotherapy in relation with cytidine deaminase downregulation. Pharmacogenetics and Genomics, 2007. 17(10): p. 841.

53. Farrell, J., Bae, K., et al., Cytidine deaminase single-nucleotide polymorphism is predictive of toxicity from gemcitabine in patients with pancreatic cancer: RTOG 9704. The Pharmacogenomics Journal, 2011.

54. Tanaka, M., Javle, M., et al., Gemcitabine metabolic and transporter gene polymorphisms are associated with drug toxicity and efficacy in patients with locally advanced pancreatic cancer. Cancer, 2010. 116(22): p. 5325-5335.

55. Sugiyama, E., Kaniwa, N., et al., Population pharmacokinetics of gemcitabine and its metabolite in Japanese cancer patients: impact of genetic polymorphisms. Clinical Pharmacokinetics, 2010. 49(8): p. 549-558.

56. Ueno, H., Kaniwa, N., et al., Homozygous CDA*3 is a major cause of life- threatening toxicities in gemcitabine-treated Japanese cancer patients. British Journal of Cancer, 2009. 100(6): p. 870-873.

57. Eda, H. and Ura, M., The antiproliferative activity of DMDC is modulated by inhibition of cytidine deaminase. Cancer Research, 1998. 58(6): p. 1165.

58. Pereira, S., Fernandes, P.A., et al., Mechanism for ribonucleotide reductase inactivation by the anticancer drug gemcitabine. Journal of Computational Chemistry, 2004. 25(10): p. 1286-1294.

59. Nakahira, S., Nakamori, S., et al., Involvement of ribonucleotide reductase M1 subunit overexpression in gemcitabine resistance of human pancreatic cancer. International Journal of Cancer, 2007. 120(6): p. 1355-1363.

60. Akita, H., Zheng, Z., et al., Significance of RRM1 and ERCC1 expression in resectable pancreatic adenocarcinoma. Oncogene, 2009. 28(32): p. 2903-2909.

148

61. Kim, R., Tan, A., et al., Prognostic roles of human equilibrative transporter 1 (hENT1) and ribonucleoside reductase subunit M1 (RRM1) in resected pancreatic cancer. Cancer, 2011.

62. Duxbury, M.S., Ito, H., et al., Inhibition of SRC tyrosine kinase impairs inherent and acquired gemcitabine resistance in human pancreatic adenocarcinoma cells. Clinical Cancer Research, 2004. 10(7): p. 2307.

63. Duxbury, M.S., Ito, H., et al., Retrovirally mediated RNA interference targeting the M2 subunit of ribonucleotide reductase: A novel therapeutic strategy in pancreatic cancer. Surgery, 2004. 136(2): p. 261-269.

64. Duxbury, M.S., Ito, H., et al., RNA interference targeting the M2 subunit of ribonucleotide reductase enhances pancreatic adenocarcinoma chemosensitivity to gemcitabine. Oncogene, 2004. 23(8): p. 1539-1548.

65. Fujita, H., Ohuchida, K., et al., Gene expression levels as predictive markers of outcome in pancreatic cancer after gemcitabine-based adjuvant chemotherapy. Neoplasia, 2010. 12(10): p. 807.

66. Itoi, T., Sofuni, A., et al., Ribonucleotide reductase subunit M2 mRNA expression in pretreatment biopsies obtained from unresectable pancreatic carcinomas. Journal of Gastroenterology, 2007. 42(5): p. 389-394.

67. Fowler, J.D., Brown, J.A., et al., Kinetic investigation of the inhibitory effect of gemcitabine on DNA polymerization catalyzed by human mitochondrial DNA polymerase. Journal of Biological Chemistry, 2008. 283(22): p. 15339.

68. Adema, A., Radi, M., et al., Troxacitabine prodrugs for pancreatic cancer. Nucleosides, Nucleotides, and Nucleic Acids, 2007. 8(9): p. 1073-1077.

69. Castelli, F., Sarpietro, M.G., et al., Characterization of lipophilic gemcitabine prodrug-liposomal membrane interaction by differential scanning calorimetry. Molecular Pharmaceutics, 2006. 3(6): p. 737-744.

70. Brusa, P., Immordino, M.L., et al., Antitumor activity and pharmacokinetics of liposomes containing lipophilic gemcitabine prodrugs. Anticancer Research, 2007. 27(1A): p. 195.

71. Stella, B., Arpicco, S., et al., Encapsulation of gemcitabine lipophilic derivatives into polycyanoacrylate nanospheres and nanocapsules. International Journal of Pharmaceutics, 2007. 344(1-2): p. 71-77.

72. Immordino, M.L., Brusa, P., et al., Preparation, characterization, cytotoxicity and pharmacokinetics of liposomes containing lipophilic gemcitabine prodrugs. Journal of Controlled Release, 2004. 100(3): p. 331-346.

149

73. Chen, P., Chien, P.Y., et al., In-vitro and in-vivo anti-cancer activity of a novel gemcitabine-cardiolipin conjugate. Anti-Cancer Drugs, 2006. 17(1): p. 53.

74. Bergman, A.M., Adema, A.D., et al., Antiproliferative activity, mechanism of action and oral antitumor activity of CP-4126, a fatty acid derivative of gemcitabine, in in vitro and in vivo tumor models. Investigational New Drugs, 2011: p. 1-11.

75. Koolen, S., Witteveen, P., et al., Phase I study of oral gemcitabine prodrug (LY2334737) alone and in combination with erlotinib in patients with advanced solid tumors. Clinical Cancer Research, 2011. 17(18): p. 6071-82.

76. Wu, W., Sigmond, J., et al., Synthesis and biological activity of a gemcitabine phosphoramidate prodrug. Journal of Medicinal Chemistry, 2007. 50(15): p. 3743-3746.

77. Kotchetkov, R., Groschel, B., et al., Antineoplastic activity of a novel multimeric gemcitabine-monophosphate prodrug against thyroid cancer cells in vitro. Anticancer Research, 2000. 20(5A): p. 2915-2922.

78. Jain, R.K. and Stylianopoulos, T., Delivering nanomedicine to solid tumors. Nature Reviews Clinical Oncology, 2010. 7(11): p. 653-664.

79. Arias, J.L., Reddy, L.H., et al., Superior preclinical efficacy of gemcitabine developed as chitosan nanoparticulate system. Biomacromolecules, 2011. 12(1): p. 97-104.

80. Maeda, H., Tumor-selective delivery of macromolecular drugs via the EPR effect: background and future prospects. Bioconjugate Chemistry, 2010. 21(5): p. 797- 802.

81. Bisht, S., Mizuma, M., et al., Systemic administration of polymeric nanoparticle- encapsulated curcumin (NanoCurc) blocks tumor growth and metastases in preclinical models of pancreatic cancer. Molecular Cancer Therapeutics, 2010. 9(8): p. 2255-64.

82. Aryal, S., Hu, C.M., et al., Combinatorial drug conjugation enables nanoparticle dual-drug delivery. Small, 2010. 6(13): p. 1442-8.

83. Sloat, B.R., Sandoval, M.A., et al., In vitro and in vivo anti-tumor activities of a gemcitabine derivative carried by nanoparticles. International Journal of Pharmaceutics, 2011. 409(1-2): p. 278-88.

84. Chung, W.G., Sandoval, M.A., et al., Stearoyl gemcitabine nanoparticles overcome resistance related to the over-expression of ribonucleotide reductase subunit M1. Journal of Controlled Release, 2011.

150

85. Rejiba, S., Reddy, L.H., et al., Squalenoyl gemcitabine nanomedicine overcomes the low efficacy of gemcitabine therapy in pancreatic cancer. Nanomedicine, 2011.

86. Reddy, L.H., Dubernet, C., et al., A new nanomedicine of gemcitabine displays enhanced anticancer activity in sensitive and resistant leukemia types. Journal of Controlled Release, 2007. 124(1-2): p. 20-7.

87. Reddy, L.H., Khoury, H., et al., Squalenoylation favorably modifies the in vivo pharmacokinetics and biodistribution of gemcitabine in mice. Drug Metabolism and Disposition, 2008. 36(8): p. 1570-7.

88. Bildstein, L., Dubernet, C., et al., Transmembrane diffusion of gemcitabine by a nanoparticulate squalenoyl prodrug: an original drug delivery pathway. Journal of Controlled Release, 2010. 147(2): p. 163-70.

89. Ambike, A., Rosilio, V., et al., Interaction of self-assembled squalenoyl gemcitabine nanoparticles with phospholipid-cholesterol monolayers mimicking a biomembrane. Langmuir, 2011. 27(8): p. 4891-9.

90. Arya, G., Vandana, M., et al., Enhanced antiproliferative activity of Herceptin (HER2)-conjugated gemcitabine-loaded chitosan nanoparticle in pancreatic cancer therapy. Nanomedicine, 2011.

91. Patra, C.R., Bhattacharya, R., et al., Targeted delivery of gemcitabine to pancreatic adenocarcinoma using cetuximab as a targeting agent. Cancer Research, 2008. 68(6): p. 1970-8.

92. Brigger, I., Dubernet, C., et al., Nanoparticles in cancer therapy and diagnosis. Advanced Drug Delivery Reviews, 2002. 54(5): p. 631-51.

93. Arias, J.L., Reddy, L.H., et al., Magnetoresponsive squalenoyl gemcitabine composite nanoparticles for cancer active targeting. Langmuir, 2008. 24(14): p. 7512-9.

94. Arias, J.L., Reddy, L.H., et al., Squalene based nanocomposites: a new platform for the design of multifunctional pharmaceutical theragnostics. ACS Nano, 2011. 5(2): p. 1513-21.

95. Galmarini, C.M., Warren, G., et al., Efficient overcoming of drug resistance to anticancer nucleoside analogs by nanodelivery of active phosphorylated drugs. International Journal of Pharmaceutics, 2010. 395(1-2): p. 281-9.

96. Hillaireau, H., Le Doan, T., et al., Hybrid polymer nanocapsules enhance in vitro delivery of azidothymidine-triphosphate to macrophages. Journal of Controlled Release, 2006. 116(3): p. 346-52.

151

97. Saiyed, Z.M., Gandhi, N.H., et al., AZT 5'-triphosphate nanoformulation suppresses human immunodeficiency virus type 1 replication in peripheral blood mononuclear cells. Journal of NeuroVirology, 2009. 15(4): p. 343-7.

98. Nakamori, S., Endo, W., et al., Multicenter Phase II Study of Pre-Administered Uracil/Tegafur (UFT) Plus Gemcitabine for Unresectable/Recurrent Pancreatic Cancer. Japanese Journal of Cancer and Chemotherapy, 2011. 38(5): p. 789.

99. Rivera, F., López-Tarruella, S., et al., Treatment of advanced pancreatic cancer: from gemcitabine single agent to combinations and targeted therapy. Cancer Treatment Reviews, 2009. 35(4): p. 335-339.

100. Furuse, J., Ishii, H., et al., Phase I study of fixed dose rate infusion of gemcitabine in patients with unresectable pancreatic cancer. Japanese Journal of Clinical Oncology, 2005. 35(12): p. 733.

101. Brand, R., Capadano, M., et al., A phase I trial of weekly gemcitabine administered as a prolonged infusion in patients with pancreatic cancer and other solid tumors. Investigational New Drugs, 1997. 15(4): p. 331-341.

102. Tempero, M., Plunkett, W., et al., Randomized phase II comparison of dose- intense gemcitabine: thirty-minute infusion and fixed dose rate infusion in patients with pancreatic adenocarcinoma. Journal of Clinical Oncology, 2003. 21(18): p. 3402.

103. Poplin, E., Levy, D., et al., Phase III trial of gemcitabine (30-minute infusion) versus gemcitabine (fixed-dose-rate infusion [FDR]) versus gemcitabine+ oxaliplatin (GEMOX) in patients with advanced pancreatic cancer (E6201). Journal of Clinical Oncology, 2006. 24(18 suppl): p. 4004.

104. Mané, J.M., Sancho, A., et al., Fixed-dose-rate gemcitabine infusion in patients with advanced pancreatic or biliary tree adenocarcinoma. Tumori, 2010. 96(3): p. 405-410.

105. Bengala, C., Guarneri, V., et al., Prolonged fixed dose rate infusion of gemcitabine with autologous haemopoietic support in advanced pancreatic adenocarcinoma. British Journal of Cancer, 2005. 93(1): p. 35-40.

106. Ulrich Pur, H., Kornek, G.V., et al., A phase II trial of biweekly high dose gemcitabine for patients with metastatic pancreatic adenocarcinoma. Cancer, 2000. 88(11): p. 2505-2511.

107. Hirao, K., Kawamoto, H., et al., A 4-week versus a 3-week schedule of gemcitabine monotherapy for advanced pancreatic cancer: a randomized phase II study to evaluate toxicity and dose intensity. International Journal of Clinical Oncology, 2011: p. 1-9.

152

108. Okada, S., Ueno, H., et al., Phase I trial of gemcitabine in patients with advanced pancreatic cancer. Japanese Journal of Clinical Oncology, 2001. 31(1): p. 7.

109. Laquente, B., Lacasa, C., et al., Antiangiogenic effect of gemcitabine following metronomic administration in a pancreas cancer model. Molecular Cancer Therapeutics, 2008. 7(3): p. 638.

110. Pollera, C.F., Ceribelli, A., et al., Prolonged infusion gemcitabine: a clinical phase I study at low-(300 mg/m 2) and high-dose (875 mg/m 2) levels. Investigational New Drugs, 1997. 15(2): p. 115-121.

111. Eckel, F., Schmelz, R., et al., Phase II trial of a 24-hour infusion of gemcitabine in previously untreated patients with advanced pancreatic adenocarcinoma. Cancer Investigation, 2003. 21(5): p. 690-694.

112. Kuemmerle, A., Decosterd, L.A., et al., A phase I pharmacokinetic study of hypoxic abdominal stop-flow perfusion with gemcitabine in patients with advanced pancreatic cancer and refractory malignant ascites. Cancer Chemotherapy and Pharmacology, 2009. 63(2): p. 331-341.

113. Vernejoul, F., Ghénassia, L., et al., Gene therapy based on gemcitabine chemosensitization suppresses pancreatic tumor growth. Molecular Therapy, 2004. 14(6): p. 758-767.

114. Réjiba, S., Bigand, C., et al., Gemcitabine-based chemogene therapy for pancreatic cancer using Ad-dCK:: UMK GDEPT and TS/RR siRNA strategies. Neoplasia, 2009. 11(7): p. 637.

115. Kasuya, H., Mizuno, M., et al., Combined effects of adeno associated virus vector and a herpes simplex virus mutant as neoplastic therapy. Journal of Surgical Oncology, 2000. 74(3): p. 214-218.

116. Kasuya, H., Nishiyama, Y., et al., Intraperitoneal delivery of hrR3 and ganciclovir prolongs survival in mice with disseminated pancreatic cancer. Journal of Surgical Oncology, 1999. 72(3): p. 136-141.

117. Pérez-Torras, S., Garcia-Manteiga, J., et al., Adenoviral-mediated overexpression of human equilibrative nucleoside transporter 1 (hENT1) enhances gemcitabine response in human pancreatic cancer. Biochemical Pharmacology, 2008. 76(3): p. 322-329.

118. Halloran, C.M., Ghaneh, P., et al., 5 Fluorouracil or gemcitabine combined with adenoviral mediated reintroduction of p16INK4A greatly enhanced cytotoxicity in Panc 1 pancreatic adenocarcinoma cells. The Journal of Gene Medicine, 2004. 6(5): p. 514-525.

153

119. Xu, Z.W., Friess, H., et al., Overexpression of Bax sensitizes human pancreatic cancer cells to apoptosis induced by chemotherapeutic agents. Cancer Chemotherapy and Pharmacology, 2002. 49(6): p. 504-510.

120. Prabhu, J.S., Korlimarla, A., et al., Gene-specific methylation: potential markers for colorectal cancer. The International Journal of Biological Markers, 2009. 24(1): p. 57-62.

121. Iacobuzio-Donahue, C.A., Maitra, A., et al., Exploration of global gene expression patterns in pancreatic adenocarcinoma using cDNA microarrays. American Journal of Pathology, 2003. 162(4): p. 1151.

122. Fry, L.C., Mönkemüller, K., et al., Molecular markers of pancreatic cancer: development and clinical relevance. Langenbeck's Archives of Surgery, 2008. 393(6): p. 883-890.

123. Habbe, N., Bert, T., et al., Identification of Methylation-Associated Gene Expression in Neuroendocrine Pancreatic Tumor Cells. Pancreatology, 2007. 7(4): p. 352-359.

124. Yu, X.J., Long, J., et al., Analysis of gene expression profiles in pancreatic carcinoma by using cDNA microarray. Hepatobiliary and Pancreatic Diseases Internaional, 2003. 2(3): p. 467-470.

125. Qin, T., Castoro, R., et al., Mechanisms of Resistance to Decitabine in the Myelodysplastic Syndrome. PloS One, 2011. 6(8): p. e23372.

126. Kelly, T.K., De Carvalho, D.D., et al., Epigenetic modifications as therapeutic targets. Nature Biotechnology, 2010. 28(10): p. 1069-1078.

127. Rius, M., Stresemann, C., et al., Human concentrative nucleoside transporter 1- mediated uptake of 5-azacytidine enhances DNA demethylation. Molecular Cancer Therapeutics, 2009. 8(1): p. 225.

128. Geutjes, E.J., Tian, S., et al., Deoxycytidine kinase is overexpressed in poor outcome breast cancer and determines responsiveness to nucleoside analogs. Breast Cancer Research and Treatment, 2012: p. 1-10.

129. Missiaglia, E., Donadelli, M., et al., Growth delay of human pancreatic cancer cells by methylase inhibitor 5-aza-2-deoxycytidine treatment is associated with activation of the interferon signalling pathway. Oncogene, 2005. 24(1): p. 199- 211.

130. Yu, M. and Epner, E., The epigenetics of mantle cell lymphoma. Current Treatment Options in Oncology, 2007. 8(5): p. 375-381.

154

131. Sato, N., Ohta, T., et al., FR901228, a novel histone deacetylase inhibitor, induces cell cycle arrest and subsequent apoptosis in refractory human pancreatic cancer cells. International Journal of Oncology, 2004. 24(3): p. 679-685.

132. Saito, A., Yamashita, T., et al., A synthetic inhibitor of histone deacetylase, MS- 27-275, with marked in vivo antitumor activity against human tumors. Proceedings of the National Academy of Sciences, 1999. 96(8): p. 4592.

133. Xenidis, N., Chelis, L., et al., Docetaxel plus gemcitabine in combination with capecitabine as treatment for inoperable pancreatic cancer: a phase II study. Cancer Chemotherapy and Pharmacology, 2011: p. 1-8.

134. Donadelli, M., Costanzo, C., et al., Trichostatin A, an inhibitor of histone deacetylases, strongly suppresses growth of pancreatic adenocarcinoma cells. Molecular Carcinogenesis, 2003. 38(2): p. 59-69.

135. Moore, P.S., Barbi, S., et al., Gene expression profiling after treatment with the histone deacetylase inhibitor trichostatin A reveals altered expression of both pro-and anti-apoptotic genes in pancreatic adenocarcinoma cells. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research, 2004. 1693(3): p. 167-176.

136. Ryu, J.K., Lee, W.J., et al., SK-7041, a new histone deacetylase inhibitor, induces G2-M cell cycle arrest and apoptosis in pancreatic cancer cell lines. Cancer Letters, 2006. 237(1): p. 143-154.

137. Cecconi, D., Donadelli, M., et al., Proteomic analysis of pancreatic endocrine tumor cell lines treated with the histone deacetylase inhibitor trichostatin A. Proteomics, 2007. 7(10): p. 1644-1653.

138. Donadelli, M., Costanzo, C., et al., Synergistic inhibition of pancreatic adenocarcinoma cell growth by trichostatin A and gemcitabine. Biochimica et Biophysica Acta (BBA) - Molecular Cell Research, 2007. 1773(7): p. 1095-1106.

139. Gahr, S., Ocker, M., et al., The combination of the histone-deacetylase inhibitor trichostatin A and gemcitabine induces inhibition of proliferation and increased apoptosis in pancreatic carcinoma cells. International Journal of Oncology, 2007. 31(3): p. 567-576.

140. Piacentini, P., Donadelli, M., et al., Trichostatin A enhances the response of chemotherapeutic agents in inhibiting pancreatic cancer cell proliferation. Virchows Archiv, 2006. 448(6): p. 797-804.

141. Arnold, N.B., Arkus, N., et al., The histone deacetylase inhibitor suberoylanilide hydroxamic acid induces growth inhibition and enhances gemcitabine-induced cell death in pancreatic cancer. Clinical Cancer Research, 2007. 13(1): p. 18.

142. Richards, D., Boehm, K., et al., Gemcitabine plus CI-994 offers no advantage over gemcitabine alone in the treatment of patients with advanced pancreatic

155

cancer: results of a phase II randomized, double-blind, placebo-controlled, multicenter study. Annals of Oncology, 2006. 17(7): p. 1096.

143. Rachagani, S., Kumar, S., et al., MicroRNA in pancreatic cancer: pathological, diagnostic and therapeutic implications. Cancer Letters, 2010. 292(1): p. 8-16.

144. Fu, S. and Kurzrock, R., Development of curcumin as an epigenetic agent. Cancer, 2010. 116(20): p. 4670-4676.

145. Dhayat, S., Mardin, W.A., et al., Epigenetic markers for chemosensitivity and chemoresistance in pancreatic cancer—A review. International Journal of Cancer, 2011.

146. Fahy, B.N., Schlieman, M.G., et al., Schedule-dependent molecular effects of the proteasome inhibitor bortezomib and gemcitabine in pancreatic cancer. Journal of Surgical Research, 2003. 113(1): p. 88-95.

147. Beumer, J.H., Eiseman, J.L., et al., Plasma pharmacokinetics and oral bioavailability of 3, 4, 5, 6-tetrahydrouridine, a cytidine deaminase inhibitor, in mice. Cancer Chemotherapy and Pharmacology, 2008. 62(3): p. 457-464.

148. Komori, S., Osada, S., et al., Contribution of thymidylate synthase to gemcitabine therapy for advanced pancreatic cancer. Pancreas, 2010. 39(8): p. 1284.

149. Funamizu, N., Kamata, Y., et al., Hydroxyurea Decreases Gemcitabine Resistance in Pancreatic Carcinoma Cells With Highly Expressed Ribonucleotide Reductase. Pancreas, 2011.

150. Mitsuno, M., Kitajima, Y., et al., Tranilast strongly sensitizes pancreatic cancer cells to gemcitabine via decreasing protein expression of ribonucleotide reductase 1. International Journal of Oncology, 2010. 36(2): p. 341-349.

151. Binder, D., Hübner, R.H., et al., Pulmonary toxicity among cancer patients treated with a combination of docetaxel and gemcitabine: A meta-analysis of clinical trials. Cancer Chemotherapy and Pharmacology, 2011: p. 1-9.

152. Von Hoff, D.D., Ramanathan, R.K., et al., Gemcitabine Plus nab-Paclitaxel Is an Active Regimen in Patients With Advanced Pancreatic Cancer: A Phase I/II Trial. Journal of Clinical Oncology, 2011.

153. Spano, J.P., Moore, M.J., et al., Phase I study of axitinib (AG-013736) in combination with gemcitabine in patients with advanced pancreatic cancer. Investigational New Drugs, 2011: p. 1-9.

154. Renouf, D.J., Moore, M.J., et al., A phase I/II study of the Src inhibitor saracatinib (AZD0530) in combination with gemcitabine in advanced pancreatic cancer. Investigational New Drugs, 2010: p. 1-8.

156

155. Awasthi, N., Schwarz, M.A., et al., Antitumour activity of sunitinib in combination with gemcitabine in experimental pancreatic cancer. HPB, 2011.

156. Adjei, A.A., Schwartz, B., et al., Early Clinical Development of ARQ 197, a Selective, Non–ATP-Competitive Inhibitor Targeting MET Tyrosine Kinase for the Treatment of Advanced Cancers. The Oncologist, 2011. 16(6): p. 788-799.

157. Starling, N., Hawkes, E., et al., A dose escalation study of gemcitabine plus oxaliplatin in combination with imatinib for gemcitabine-refractory advanced pancreatic adenocarcinoma. Annals of Oncology, 2011.

158. Richards, D.A., Kuefler, P.R., et al., Gemcitabine plus enzastaurin or single-agent gemcitabine in locally advanced or metastatic pancreatic cancer: results of a phase II, randomized, noncomparative study. Investigational New Drugs, 2011. 29(1): p. 144-153.

159. Pericay Pijaume, C., Escudero Emperador, P., et al., Open-label trial on efficacy and security of treatment with gemcitabine and oral modulation with tegafur and levofolinic acid (GEMTG) in patients with advanced pancreatic cancer. Clinical and Translational Oncology, 2011. 13(1): p. 61-66.

160. Sung, V., Richard, N., et al., Histone deacetylase inhibitor MGCD0103 synergizes with gemcitabine in human pancreatic cells. Cancer Science, 2011.

161. De Jesus-Acosta, A., Oliver, G.R., et al., A multicenter analysis of GTX chemotherapy in patients with locally advanced and metastatic pancreatic adenocarcinoma. Cancer Chemotherapy and Pharmacology, 2011: p. 1-10.

162. Hill, M.E., Li, X., et al., A phase I study of the biomodulation of capecitabine by docetaxel and gemcitabine (mGTX) in previously untreated patients with metastatic adenocarcinoma of the pancreas. Cancer Chemotherapy and Pharmacology, 2011. 67(3): p. 511-517.

163. Sahora, K., Kuehrer, I., et al., NeoGemOx: Gemcitabine and oxaliplatin as neoadjuvant treatment for locally advanced, nonmetastasized pancreatic cancer. Surgery, 2011. 149(3): p. 311-320.

164. Hess, V., Pratsch, S., et al., Combining gemcitabine, oxaliplatin and capecitabine (GEMOXEL) for patients with advanced pancreatic carcinoma (APC): a phase I/II trial. Annals of Oncology, 2010. 21(12): p. 2390.

165. Cereda, S., Reni, M., et al., XELIRI or FOLFIRI as Salvage Therapy in Advanced Pancreatic Cancer. Anticancer Research, 2010. 30(11): p. 4785.

166. Conroy, T., Desseigne, F., et al., FOLFIRINOX versus gemcitabine for metastatic pancreatic cancer. New England Journal of Medicine, 2011. 364(19): p. 1817- 1825.

157

167. Olive, K.P., Jacobetz, M.A., et al., Inhibition of Hedgehog signaling enhances delivery of chemotherapy in a mouse model of pancreatic cancer. Science, 2009. 324(5933): p. 1457.

168. Giovannetti, E., Leon, L., et al., Study of Apoptosis Induction and Deoxycytidine Kinase/Cytidine Deaminase Modulation in the Synergistic Interaction of a Novel Ceramide Analog and Gemcitabine in Pancreatic Cancer Cells. Nucleosides, Nucleotides and Nucleic Acids, 2010. 29(4-6): p. 419-426.

169. Roman, N.O., Samulitis, B.K., et al., Imexon enhances gemcitabine cytotoxicity by inhibition of ribonucleotide reductase. Cancer Chemotherapy and Pharmacology, 2011. 67(1): p. 183-192.

170. Maier, H.J., Wirth, T., et al., Epithelial-mesenchymal transition in pancreatic carcinoma. Cancers, 2010. 2(4): p. 2058-83.

171. Thiery, J.P. and Sleeman, J.P., Complex networks orchestrate epithelial- mesenchymal transitions. Nature Reviews Molecular Cell Biology, 2006. 7(2): p. 131-42.

172. Kiesslich, T., Pichler, M., et al., Epigenetic control of epithelial-mesenchymal- transition in human cancer. Molecular and Clinical Oncology, 2013. 1(1): p. 3- 11.

173. Bhutia, Y.D., Hung, S.W., et al., Differential processing of let-7a precursors influences RRM2 expression and chemosensitivity in pancreatic cancer: Role of LIN-28 and SET oncoprotein. PLOS ONE, 2013. 8(1): p. e53436.

174. Garcia-Rodriguez, L., Perez-Torras, S., et al., Connexin-26 is a key factor mediating gemcitabine bystander effect. Molecular Cancer Therapeutics, 2011. 10(3): p. 505-17.

175. Lai, C.P., Bechberger, J.F., et al., Tumor-suppressive effects of pannexin 1 in C6 glioma cells. Cancer Research, 2007. 67(4): p. 1545-54.

176. Penuela, S., Gyenis, L., et al., Loss of pannexin 1 attenuates melanoma progression by reversion to a melanocytic phenotype. Journal of Biological Chemistry, 2012. 287(34): p. 29184-93.

177. Choi, O., Heathcote, D.A., et al., A deficiency in nucleoside salvage impairs murine lymphocyte development, homeostasis, and survival. Journal of Immunology, 2012. 188(8): p. 3920-7.

178. Toy, G., Austin, W.R., et al., Requirement for deoxycytidine kinase in T and B lymphocyte development. Proceedings of the National Academy of Sciences, 2010. 107(12): p. 5551-6.

158

179. Bhalla, K. and Grant, S., Effect of deoxycytidine on the in vitro response of human leukemia cells to inhibitors of de novo pyrimidine biosynthesis. Cancer Chemotherapy and Pharmacology, 1987. 19(3): p. 226-32.

180. Colly, L.P., Richel, D.J., et al., Increase in Ara-C sensitivity in Ara-C sensitive and -resistant leukemia by stimulation of the salvage and inhibition of the de novo pathway. Annals of Hematology, 1992. 65(1): p. 26-32.

181. Eriksson, S., Arner, E., et al., Properties and levels of deoxynucleoside kinases in normal and tumor cells; implications for chemotherapy. Advances in Enzyme Regulation, 1994. 34: p. 13-25.

182. Nathanson, D.A., Armijo, A.L., et al., Co-targeting of convergent nucleotide biosynthetic pathways for leukemia eradication. Journal of Experimental Medicine, 2014. 211(3): p. 473-86.

183. Smal, C., Cardoen, S., et al., Activation of deoxycytidine kinase by protein kinase inhibitors and okadaic acid in leukemic cells. Biochemical Pharmacology, 2004. 68(1): p. 95-103.

184. White, J.C. and Capizzi, R.L., A critical role for uridine nucleotides in the regulation of deoxycytidine kinase and the concentration dependence of 1-beta-D- arabinofuranosylcytosine phosphorylation in human leukemia cells. Cancer Research, 1991. 51(10): p. 2559-65.

185. Stegmann, A.P., Honders, M.W., et al., Role of deoxycytidine kinase in an in vitro model for AraC- and DAC-resistance: substrate-enzyme interactions with deoxycytidine, 1-beta-D-arabinofuranosylcytosine and 5-aza-2'-deoxycytidine. Leukemia, 1993. 7(7): p. 1005-11.

186. Kunos, C.A., Ferris, G., et al., Deoxynucleoside salvage facilitates DNA repair during ribonucleotide reductase blockade in human cervical cancers. Radiation Research, 2011. 176(4): p. 425-33.

187. Singhal, R.L., Yeh, Y.A., et al., Increased deoxycytidine kinase activity in cancer cells and inhibition by difluorodeoxycytidine. Oncology Research, 1992. 4(11- 12): p. 517-22.

188. Azghadi, S. and Clark, A.T., Epigenetically reprogramming of human embryonic stem cells by 3-deazaneplanocin a and sodium butyrate. International Journal of Preventative Medicine, 2011. 2(2): p. 73-8.

189. Diaz Perez, S.V., Kim, R., et al., Derivation of new human embryonic stem cell lines reveals rapid epigenetic progression in vitro that can be prevented by chemical modification of chromatin. Human Molecular Genetics, 2012. 21(4): p. 751-64.

159

190. Hung, S.W., Mody, H., et al., Pharmacological Reversal of Histone Methylation Presensitizes Pancreatic Cancer Cells to Nucleoside Drugs: In Vitro Optimization and Novel Nanoparticle Delivery Studies. PLoS One, 2013. 8(8): p. e71196.

191. Farrell, A.S., Allen-Petersen, B., et al., Targeting Inhibitors of the Tumor Suppressor PP2A for the Treatment of Pancreatic Cancer. Molecular Cancer Research, 2014. 12(6): p. 924-39.

192. Seo, S.B., McNamara, P., et al., Regulation of histone acetylation and transcription by INHAT, a human cellular complex containing the set oncoprotein. Cell, 2001. 104(1): p. 119-30.

193. Shah, A., Summy, J., et al., Development and characterization of gemcitabine- resistant pancreatic tumor cells. Annals of Surgical Oncology, 2007. 14(12): p. 3629-3637.

194. Arumugam, T., Ramachandran, V., et al., Epithelial to mesenchymal transition contributes to drug resistance in pancreatic cancer. Cancer Research, 2009. 69(14): p. 5820.

195. Li, Y., VandenBoom, T., et al., Up-regulation of miR-200 and let-7 by natural agents leads to the reversal of epithelial-to-mesenchymal transition in gemcitabine-resistant pancreatic cancer cells. Cancer Research, 2009. 69(16): p. 6704.

196. Schaeffer, D., Assi, K., et al., Tumor expression of Integrin-linked kinase (ILK) correlates with the expression of the E-cadherin repressor Snail: an immunohistochemical study in ductal pancreatic adenocarcinoma. Virchows Archiv, 2010: p. 1-8.

197. Wang, F., Sloss, C., et al., Membrane-bound heparin-binding epidermal growth factor like growth factor regulates E-cadherin expression in pancreatic carcinoma cells. Cancer Research, 2007. 67(18): p. 8486.

198. Yin, T., Wang, C., et al., Expression of snail in pancreatic cancer promotes metastasis and chemoresistance. Journal of Surgical Research, 2007. 141(2): p. 196-203.

199. Nakamura, T., Kato, Y., et al., E-cadherin-dependent intercellular adhesion enhances chemoresistance. International Journal of Molecular Medicine, 2003. 12(5): p. 693.

200. Koo, J., Jung, W., et al., The Predictive Role of E-cadherin and Androgen Receptor on In Vitro Chemosensitivity in Triple-negative Breast Cancer. Japanese Journal of Clinical Oncology, 2009.

160

201. Yang, A., Fan, F., et al., Chronic oxaliplatin resistance induces epithelial-to- mesenchymal transition in colorectal cancer cell lines. Clinical Cancer Research, 2006. 12(14): p. 4147.

202. Matarrese, P., Giandomenico, V., et al., Antiproliferative activity of interferon &agr; and retinoic acid in SiHa carcinoma cells: The role of cell adhesion. International Journal of Cancer, 1998. 76(4): p. 531-540.

203. Black, P., Brown, G., et al., Sensitivity to epidermal growth factor receptor inhibitor requires E-cadherin expression in urothelial carcinoma cells. Clinical Cancer Research, 2008. 14(5): p. 1478.

204. Wick, M., Hurteau, G., et al., Peroxisome proliferator-activated receptor-gamma is a target of nonsteroidal anti-inflammatory drugs mediating cyclooxygenase- independent inhibition of lung cancer cell growth. Molecular Pharmacology, 2002. 62(5): p. 1207.

205. Witta, S., Gemmill, R., et al., Restoring E-cadherin expression increases sensitivity to epidermal growth factor receptor inhibitors in lung cancer cell lines. Cancer Research, 2006. 66(2): p. 944.

206. Feldmann, G., Dhara, S., et al., Blockade of hedgehog signaling inhibits pancreatic cancer invasion and metastases: a new paradigm for combination therapy in solid cancers. Cancer Research, 2007. 67(5): p. 2187.

207. Sasaki, C., Lin, H., et al., Expression of E-cadherin reduces bcl-2 expression and increases sensitivity to etoposide-induced apoptosis. International Journal of Cancer, 2000. 86(5): p. 660-666.

208. Aleskandarany, M.A., Negm, O.H., et al., Epithelial mesenchymal transition in early invasive breast cancer: an immunohistochemical and reverse phase protein array study. Breast Cancer Research and Treatment, 2014. 145(2): p. 339-48.

209. Calvisi, D.F., Pascale, R.M., et al., Dissection of signal transduction pathways as a tool for the development of targeted therapies of hepatocellular carcinoma. Reviews on Recent Clinical Trials, 2007. 2(3): p. 217-36.

210. Dong, H.M., Liu, G., et al., Dominant-negative E-cadherin inhibits the invasiveness of inflammatory breast cancer cells in vitro. Journal of Cancer Research and Clinical Oncology, 2007. 133(2): p. 83-92.

211. Han, L., Peng, B., et al., Indometacin ameliorates high glucose-induced proliferation and invasion via modulation of e-cadherin in pancreatic cancer cells. Current Medicinal Chemistry, 2013. 20(33): p. 4142-52.

212. Miotti, S., Tomassetti, A., et al., Simultaneous expression of caveolin-1 and E- cadherin in ovarian carcinoma cells stabilizes adherens junctions through

161

inhibition of src-related kinases. American Journal of Pathology, 2005. 167(5): p. 1411-27.

213. Wang, D., Su, L., et al., Downregulation of E-Cadherin enhances proliferation of head and neck cancer through transcriptional regulation of EGFR. Molecular Cancer, 2011. 10: p. 116.

214. Wang, Y., Zhang, Y.X., et al., Loss of P53 facilitates invasion and metastasis of prostate cancer cells. Molecular and Cellular Biochemistry, 2013. 384(1-2): p. 121-7.

215. Mini, E., Nobili, S., et al., Cellular pharmacology of gemcitabine. Annals of Oncology, 2006. 17(Supplement 5): p. v7.

216. Wong, A., Soo, R.A., et al., Clinical pharmacology and pharmacogenetics of gemcitabine. Drug Metabolism Reviews, 2009. 41(2): p. 77-88.

217. Lu, J., Getz, G., et al., MicroRNA expression profiles classify human cancers. Nature, 2005. 435(7043): p. 834-838.

218. Plunkett, W., Huang, P., et al., Gemcitabine: metabolism, mechanisms of action, and self-potentiation. Seminars in Oncology, 1995. 22.

219. Bhutia, Y.D., Hung, S.W., et al., CNT1 expression influences proliferation and chemosensitivity in drug-resistant pancreatic cancer cells. Cancer Research, 2011. 71(5): p. 1825-1835.

220. Kang, N., Jun, A.H., et al., Human Equilibrative Nucleoside Transporter-3 (hENT3) Spectrum Disorder Mutations Impair Nucleoside Transport, Protein Localization, and Stability The Journal of Biological Chemistry, 2010. 285(36): p. 28343-28352.

221. Govindarajan, R., Endres, C.J., et al., Expression and hepatobiliary transport characteristics of the concentrative and equilibrative nucleoside transporters in sandwich-cultured human hepatocytes. American Journal of Physiology - Gastrointestinal and Liver Physiology, 2008(295): p. G570-G580.

222. Nivillac, N.M., Bacani, J., et al., The life cycle of human equilibrative nucleoside transporter 1: from ER export to degradation. Experimental Cell Research, 2011. 317(11): p. 1567-79.

223. Mackey, J.R., Yao, S.Y.M., et al., Gemcitabine Transport in Xenopus Oocytes Expressing Recombinant Plasma Membrane Mammalian Nucleoside Transporters. Journal of the National Cancer Institute, 1999. 91(21): p. 1876- 1881.

162

224. Young, J.D., Yao, S.Y.M., et al., The human concentrative and equilibrative nucleoside transporter families, SLC28 and SLC29. Molecular Aspects of Medicine, 2013. 34(2–3): p. 529-547.

225. Ward, J.L., Sherali, A., et al., Kinetic and Pharmacological Properties of Cloned Human Equilibrative Nucleoside Transporters, ENT1 and ENT2, Stably Expressed in Nucleoside Transporter-deficient PK15 Cells: ENT2 EXHIBITS A LOW AFFINITY FOR GUANOSINE AND CYTIDINE BUT A HIGH AFFINITY FOR INOSINE. Journal of Biological Chemistry, 2000. 275(12): p. 8375-8381.

226. Crawford, L.J., Walker, B., et al., Proteasome inhibitors in cancer therapy. Journal of Cell Communication and Signaling, 2011. 5(2): p. 101-110.

227. Goldberg, A.L., Development of proteasome inhibitors as research tools and cancer drugs. The Journal of Cell Biology, 2012. 199(4): p. 583-588.

228. Joazeiro, C.A., Anderson, K.C., et al., Proteasome inhibitor drugs on the rise. Cancer Research, 2006. 66(16): p. 7840-2.

229. Fehrenbacher, N. and Jäättelä, M., Lysosomes as Targets for Cancer Therapy. Cancer Research, 2005. 65(8): p. 2993-2995.

230. Maclean, K.H., Dorsey, F.C., et al., Targeting lysosomal degradation induces p53-dependent cell death and prevents cancer in mouse models of lymphomagenesis. The Journal of Clinical Investigation, 2008. 118(1): p. 79-88.

231. Yang, Z.J., Chee, C.E., et al., The Role of Autophagy in Cancer: Therapeutic Implications. Molecular Cancer Therapeutics, 2011. 10(9): p. 1533-1541.

232. Cheng, J.C., Auersperg, N., et al., Inhibition of p53 represses E-cadherin expression by increasing DNA methyltransferase-1 and promoter methylation in serous borderline ovarian tumor cells. Oncogene, 2011. 30(37): p. 3930-42.

233. Hong, C.Q., Ran, Y.G., et al., Study on promoter methylation status of E-cadherin gene in nasopharyngeal carcinoma cell lines. Chinese Journal of Pathology, 2010. 39(8): p. 532-6.

234. Liu, Y.Y., Han, J.Y., et al., Effect of CDH1 gene methylation on transforming growth factor (TGF-beta)-induced epithelial-mesenchymal transition in alveolar epithelial cell line A549. Genetic and Molecular Research, 2014. 13(Aop).

235. Manavalan, T.T., Teng, Y., et al., Reduced expression of miR-200 family members contributes to antiestrogen resistance in LY2 human breast cancer cells. PLoS One, 2013. 8(4): p. e62334.

236. Nalls, D., Tang, S.N., et al., Targeting epigenetic regulation of miR-34a for treatment of pancreatic cancer by inhibition of pancreatic cancer stem cells. PLoS One, 2011. 6(8): p. e24099.

163

237. Ran, Y., Wu, S., et al., Demethylation of E-cadherin gene in nasopharyngeal carcinoma could serve as a potential therapeutic strategy. The Journal of Biochemistry, 2011. 149(1): p. 49-54.

238. Wang, G., Hu, X., et al., Promoter-hypermethylation associated defective expression of E-cadherin in primary non-small cell lung cancer. Lung Cancer, 2008. 62(2): p. 162-72.

239. Baldwin, S.A., Mackey, J.R., et al., Nucleoside transporters: molecular biology and implications for therapeutic development. Molecular Medicine Today, 1999. 5(5): p. 216-224.

240. Beyer, E.C., Paul, D.L., et al., Connexin family of gap junction proteins. Journal of Membrane Biology, 1990. 116(3): p. 187-194.

241. Goldberg, G.S., Moreno, A.P., et al., Gap junctions between cells expressing connexin 43 or 32 show inverse permselectivity to adenosine and ATP. Journal of Biological Chemistry, 2002. 277(39): p. 36725.

242. Mesnil, M. and Yamasaki, H., Bystander Effect in Herpes Simplex Virus- Thymidine Kinase/Ganciclovir Cancer Gene Therapy: Role of Gap-junctional Intercellular Communication1. Cancer Research, 2000. 60(15): p. 3989.

243. Fick, J., Barker, F., et al., The extent of heterocellular communication mediated by gap junctions is predictive of bystander tumor cytotoxicity in vitro. Proceedings of the National Academy of Sciences, 1995. 92(24): p. 11071.

244. Tanaka, T., Duflot-Dancer, A., et al., Bystander effect from cytosine deaminase and uracil phosphoribosyl transferase genes in vitro: A partial contribution of gap junctions. Cancer Letters, 2009. 282(1): p. 43-47.

245. He, B., Tong, X., et al., Tramadol and Flurbiprofen Depress the Cytotoxicity of Cisplatin via Their Effects on Gap Junctions. Clinical Cancer Research, 2009. 15(18): p. 5803.

246. Farré, X., Guillén Gómez, E., et al., Expression of the nucleoside derived drug transporters hCNT1, hENT1 and hENT2 in gynecologic tumors. International Journal of Cancer, 2004. 112(6): p. 959-966.

247. Marcé, S., Molina-Arcas, M., et al., Expression of human equilibrative nucleoside transporter 1 (hENT1) and its correlation with gemcitabine uptake and cytotoxicity in mantle cell lymphoma. Haematologica, 2006. 91(7): p. 895.

248. Farrell, J.J., Elsaleh, H., et al., Human equilibrative nucleoside transporter 1 levels predict response to gemcitabine in patients with pancreatic cancer. Gastroenterology, 2009. 136(1): p. 187-195.

164

249. Govindarajan, R., Chakraborty, S., et al., Assembly of connexin43 into gap junctions is regulated differentially by E-cadherin and N-cadherin in rat liver epithelial cells. Molecular Biology of the Cell, 2010. 21(23): p. 4089-107.

250. Neuhaus, I.M., Bone, L., et al., The human connexin32 gene is transcribed from two tissue-specific promoters. Bioscience Reports, 1996. 16(3): p. 239-48.

251. Tai, M.H., Olson, L.K., et al., Characterization of gap junctional intercellular communication in immortalized human pancreatic ductal epithelial cells with stem cell characteristics. Pancreas, 2003. 26(1): p. e18-26.

252. Simon, J.A. and Kingston, R.E., Mechanisms of polycomb gene silencing: knowns and unknowns. Nature Reviews Molecular Cell Biology, 2009. 10(10): p. 697- 708.

253. Crea, F., Fornaro, L., et al., EZH2 inhibition: targeting the crossroad of tumor invasion and angiogenesis. Cancer and Metastasis Reviews, 2012: p. 1-9.

254. Suvà, M.L., Riggi, N., et al., EZH2 is essential for glioblastoma cancer stem cell maintenance. Cancer Research, 2009. 69(24): p. 9211-9218.

255. Bao, B., Ali, S., et al., Curcumin Analogue CDF Inhibits Pancreatic Tumor Growth by Switching on Suppressor microRNAs and Attenuating EZH2 Expression. Cancer Research, 2012. 72(1): p. 335-345.

256. Crea, F., Paolicchi, E., et al., Polycomb genes and cancer: Time for clinical application? CRC Critical Reviews in Oncology/Hematology, 2012. 83(2): p. 184-193.

257. Ougolkov, A.V., Bilim, V.N., et al., Regulation of pancreatic tumor cell proliferation and chemoresistance by the histone methyltransferase enhancer of zeste homologue 2. Clinical Cancer Research, 2008. 14(21): p. 6790-6796.

258. Chen, Y., Xie, D., et al., RNAi targeting EZH2 inhibits tumor growth and liver metastasis of pancreatic cancer in vivo. Cancer Letters, 2010. 297(1): p. 109-116.

259. Fujii, S., Fukamachi, K., et al., RAS oncogenic signal upregulates EZH2 in pancreatic cancer. Biochemical and Biophysical Research Communications, 2012. 417(3): p. 1074-1079.

260. Fujii, S., Ito, K., et al., Enhancer of zeste homologue 2 (EZH2) down-regulates RUNX3 by increasing histone H3 methylation. Journal of Biological Chemistry, 2008. 283(25): p. 17324-17332.

261. Fussbroich, B., Wagener, N., et al., EZH2 depletion blocks the proliferation of colon cancer cells. PLoS One, 2011. 6(7): p. e21651.

165

262. Qazi, A.M., Aggarwal, S., et al., Laser Capture Microdissection of Pancreatic Ductal Adeno-Carcinoma Cells to Analyze EzH2 by Western Blot Analysis. Methods in Molecular Biology, 2011. 755(1): p. 245-256.

263. Toll, A.D., Dasgupta, A., et al., Implications of enhancer of zeste homologue 2 expression in pancreatic ductal adenocarcinoma. Human Pathology, 2010. 41(9): p. 1205-1209.

264. Wei, Y., Xia, W., et al., Loss of trimethylation at lysine 27 of histone H3 is a predictor of poor outcome in breast, ovarian, and pancreatic cancers. Molecular Carcinogenesis, 2008. 47(9): p. 701-706.

265. Hayden, A., Johnson, P.W.M., et al., S-adenosylhomocysteine hydrolase inhibition by 3-deazaneplanocin A analogues induces anti-cancer effects in breast cancer cell lines and synergy with both histone deacetylase and HER2 inhibition. Breast Cancer Research and Treatment, 2011. 127(1): p. 109-119.

266. Choudhury, S.R., Balasubramanian, S., et al., (-)-Epigallocatechin-3-gallate and DZNep reduce polycomb protein level via a proteasome-dependent mechanism in skin cancer cells. Carcinogenesis, 2011. 32(10): p. 1525-1532.

267. Tan, J., Yang, X., et al., Pharmacologic disruption of Polycomb-repressive complex 2-mediated gene repression selectively induces apoptosis in cancer cells. Genes and Development, 2007. 21(9): p. 1050-1063.

268. Cheng, L.L., Itahana, Y., et al., TP53 Genomic Status Regulates Sensitivity of Gastric Cancer Cells to the Histone Methylation Inhibitor 3-Deazaneplanocin A (DZNep). Clinical Cancer Research, 2012. 18(15): p. 4201-4212.

269. Crea, F., Hurt, E.M., et al., Pharmacologic disruption of Polycomb Repressive Complex 2 inhibits tumorigenicity and tumor progression in prostate cancer. Molecular Cancer, 2011. 10(1): p. 40.

270. Chase, A. and Cross, N.C.P., Aberrations of EZH2 in cancer. Clinical Cancer Research, 2011. 17(9): p. 2613-2618.

271. Zhou, J., Bi, C., et al., The histone methyltransferase inhibitor, DZNep, upregulates TXNIP, increases ROS production, and targets leukemia cells in AML. Blood, 2011. 118(10): p. 2830-2839.

272. Puppe, J., Drost, R., et al., BRCA1-deficient mammary tumor cells are dependent on EZH2 expression and sensitive to Polycomb Repressive Complex 2-inhibitor 3- deazaneplanocin A. Breast Cancer Research, 2009. 11(4): p. R63.

273. Yang, X., Lay, F., et al., Targeting DNA methylation for epigenetic therapy. Trends in Pharmacological Sciences, 2010. 31(11): p. 536-546.

166

274. Bayraktar, S., Bayraktar, U.D., et al., Recent developments in palliative chemotherapy for locally advanced and metastatic pancreas cancer. World Journal of Gastroenterology, 2010. 16(6): p. 673.

275. Bhutia, Y.D., Hung, S.W., et al., CNT1 expression influences proliferation and chemosensitivity in drug-resistant pancreatic cancer cells. Cancer Res, 2011. 71(5): p. 1825-1835.

276. Chou, T.C. and Talalay, P., Quantitative analysis of dose-effect relationships: the combined effects of multiple drugs or enzyme inhibitors. Advances in Enzyme Regulation, 1984. 22(1): p. 27-55.

277. Kang, N., Jun, A.H., et al., Human equilibrative nucleoside transporter-3 (hENT3) spectrum disorder mutations impair nucleoside transport, protein localization, and stability. Journal of Biological Chemistry, 2010. 285(36): p. 28343-28352.

278. Adema, A., Radi, M., et al., Troxacitabine prodrugs for pancreatic cancer. Nucleos Nucleot Nucl, 2007. 26(8-9): p. 1073-1077.

279. Radi, M., Adema, A.D., et al., In vitro optimization of non-small cell lung cancer activity with troxacitabine, L-1, 3-dioxolane-cytidine, prodrugs. Journal of Medicinal Chemistry, 2007. 50(9): p. 2249-2253.

280. Deroose, F.D. and Clercq, P.J.D., Novel enantioselective syntheses of (+)-biotin. Journal of Organic Chemistry, 1995. 60(2): p. 321-330.

281. Marrache, S. and Dhar, S., Engineering of blended nanoparticle platform for delivery of mitochondria-acting therapeutics. Proceedings of the National Academy of Sciences, 2012. 109: p. 16288-16293.

282. Paproski, R.J., Ng, A.M.L., et al., The role of human nucleoside transporters in uptake of 3′-deoxy-3′-fluorothymidine. Molecular Pharmacology, 2008. 74(5): p. 1372-1380.

283. Avan, A., Crea, F., et al., Molecular Mechanisms Involved in the Synergistic Interaction of the EZH2 Inhibitor 3-Deazaneplanocin A with Gemcitabine in Pancreatic Cancer Cells. Molecular Cancer Therapeutics, 2012. 11(8): p. 1735- 1746.

284. Coulombe, R.A., Sharma, R., et al., Pharmacokinetics of the antiviral agent 3- deazaneplanocin A. European Journal of Drug Metabolism and Pharmacokinetics, 1995. 20(3): p. 197-202.

285. Glazer, R.I., Hartman, K.D., et al., 3-Deazaneplanocin: a new and potent inhibitor of S-adenosylhomocysteine hydrolase and its effects on human promyelocytic leukemia cell line HL-60. Biochemical and Biophysical Research Communications, 1986. 135(2): p. 688-694.

167

286. Mallen-St Clair, J., Soydaner-Azeloglu, R., et al., EZH2 couples pancreatic regeneration to neoplastic progression. Genes and Development, 2012. 26(5): p. 439-444.

287. Rogenhofer, S., Kahl, P., et al., Decreased levels of histone H3K9me1 indicate poor prognosis in patients with renal cell carcinoma. Anticancer Research, 2012. 32(3): p. 879-886.

288. Hou, J., Wu, J., et al., Genomic amplification and a role in drug-resistance for the KDM5A histone demethylase in breast cancer. American Journal of Translational Research, 2012. 4(3): p. 247.

289. Toth, M., Boros, I.M., et al., Elevated level of lysine 9-acetylated histone H3 at the MDR1 promoter in multidrug-resistant cells. Cancer Science, 2012. 104(4): p. 659-669.

290. Shi, L., Sun, L., et al., Histone demethylase JMJD2B coordinates H3K4/H3K9 methylation and promotes hormonally responsive breast carcinogenesis. Proceedings of the National Academy of Sciences, 2011. 108(18): p. 7541-7546.

291. Chang, C. and Hung, M., The role of EZH2 in tumour progression. British Journal of Cancer, 2011. 106(2): p. 243-247.

292. Abdel-Wahab, O. and Levine, R.L., EZH2 mutations: Mutating the epigenetic machinery in myeloid malignancies. Cancer Cell, 2010. 18(2): p. 105-107.

293. Matrisian, L.M., Aizenberg, R., et al., The Alarming Rise of Pancreatic Cancer Deaths in the United States: Why We Need to Stem the Tide Today. 2012, Research and Scientific Affairs Department of the Pancreatic Cancer Action Network.

294. Rhim, A.D., Mirek, E.T., et al., EMT and dissemination precede pancreatic tumor formation. Cell, 2012. 148(1–2): p. 349-361.

295. Nagata, K., Kawase, H., et al., Replication factor encoded by a putative oncogene, set, associated with myeloid leukemogenesis. Proceedings of the National Academy of Sciences, 1995. 92(10): p. 4279-4283.

296. Christensen, D.J., Chen, Y., et al., SET oncoprotein overexpression in B-cell chronic lymphocytic leukemia and non-Hodgkin lymphoma: a predictor of aggressive disease and a new treatment target. Blood, 2011. 118(15): p. 4150- 4158.

297. Cristóbal, I., Garcia-Orti, L., et al., Overexpression of SET is a recurrent event associated with poor outcome and contributes to protein phosphatase 2A inhibition in acute myeloid leukemia. Haematologica, 2012. 97(4): p. 543-550.

168

298. Neviani, P., Santhanam, R., et al., The tumor suppressor PP2A is functionally inactivated in blast crisis CML through the inhibitory activity of the BCR/ABL- regulated SET protein. Cancer Cell, 2005. 8(5): p. 355-368.

299. Carlson, S.G., Eng, E., et al., Expression of SET, an inhibitor of protein phosphatase 2A, in renal development and Wilms' tumor. Journal of the American Society of Nephrology, 1998. 9(10): p. 1873-1880.

300. Fukukawa, C., Shima, H., et al., Up-regulation of I-2PP2A/SET gene expression in rat primary hepatomas and regenerating livers. Cancer Letters, 2000. 161(1): p. 89-95.

301. Chao, A., Tsai, C.L., et al., Decreased expression of microRNA-199b increases protein levels of SET (protein phosphatase 2A inhibitor) in human choriocarcinoma. Cancer Letters, 2010. 291(1): p. 99-107.

302. Li, M., Makkinje, A., et al., The myeloid leukemia-associated protein SET is a potent inhibitor of protein phosphatase 2A. Journal of Biological Chemistry, 1996. 271(19): p. 11059-11062.

303. Seo, S.B., McNamara, P., et al., Regulation of histone acetylation and transcription by INHAT, a human cellular complex containing the set oncoprotein. Cell, 2001. 104(1): p. 119-130.

304. ten Klooster, J.P., v Leeuwen, I., et al., Rac1-induced cell migration requires membrane recruitment of the nuclear oncogene SET. The EMBO Journal, 2007. 26(2): p. 336-345.

305. Daniel Lam, B., Anthony, E.C., et al., Cytoplasmic targeting of the proto- oncogene SET promotes cell spreading and migration. FEBS Letters, 2012.

306. Fukukawa, C., Shima, H., et al., The oncoprotein I-2PP2A/SET negatively regulates the MEK/ERK pathway and cell proliferation. International Journal of Oncology, 2005. 26(3): p. 751.

307. Al-Murrani, S., Woodgett, J., et al., Expression of I2PP2A, an inhibitor of protein phosphatase 2A, induces c-Jun and AP-1 activity. Biochemical Journal, 1999. 341: p. 293-298.

308. Shintani, Y., Hollingsworth, M.A., et al., Collagen I promotes metastasis in pancreatic cancer by activating c-Jun NH2-terminal kinase 1 and up-regulating N-cadherin expression. Cancer Research, 2006. 66(24): p. 11745-11753.

309. Park, M.K., You, H.J., et al., Transglutaminase-2 induces N-cadherin expression in TGF-β1-induced epithelial mesenchymal transition via c-Jun-N-terminal kinase activation by protein phosphatase 2A down-regulation. European Journal of Cancer, 2013. 49(7): p. 1692-1705.

169

310. Li, W., Xie, L., et al., Cantharidin, a potent and selective PP2A inhibitor, induces an oxidative stress-independent growth inhibition of pancreatic cancer cells through G2/M cell-cycle arrest and apoptosis. Cancer Science, 2010. 101(5): p. 1226-1233.

311. Li, W., Chen, Z., et al., PP2A inhibitors induce apoptosis in pancreatic cancer cell line PANC-1 through persistent phosphorylation of IKKα and sustained activation of the NF-κB pathway. Cancer Letters, 2011. 304(2): p. 117-127.

312. Li, W., Chen, Z., et al., Growth of the pancreatic cancer cell line PANC-1 is inhibited by protein phosphatase 2A inhibitors through overactivation of the c- Jun N-terminal kinase pathway. European Journal of Cancer, 2011. 47(17): p. 2654-2664.

313. Switzer, C.H., Cheng, R.Y., et al., Targeting SET/I2PP2A oncoprotein functions as a multi-pathway strategy for cancer therapy. Oncogene, 2011. 30(22): p. 2504- 2513.

314. Kim, M.P., Evans, D.B., et al., Generation of orthotopic and heterotopic human pancreatic cancer xenografts in immunodeficient mice. Nature Protocols, 2009. 4(11): p. 1670-1680.

315. Lieber, M., Mazzetta, J., et al., Establishment of a continuous tumor-cell line (panc-1) from a human carcinoma of the exocrine pancreas. International Journal of Cancer, 1975. 15(5): p. 741-7.

316. Winter, J.M., Ting, A.H., et al., Absence of E-cadherin expression distinguishes noncohesive from cohesive pancreatic cancer. Clinical Cancer Research, 2008. 14(2): p. 412-8.

317. Bruns, C.J., Harbison, M.T., et al., In vivo selection and characterization of metastatic variants from human pancreatic adenocarcinoma by using orthotopic implantation in nude mice. Neoplasia, 1999. 1(1): p. 50-62.

318. Shintani, Y., Wheelock, M.J., et al., Phosphoinositide-3 Kinase–Rac1–c-Jun NH2-terminal Kinase Signaling Mediates Collagen I–induced Cell Scattering and Up-Regulation of N-Cadherin Expression in Mouse Mammary Epithelial Cells. Molecular Biology of the Cell, 2006. 17(7): p. 2963-2975.

319. Maurer, T., Garrenton, L., et al., Drugging the Undruggable: Small-Molecule Inhibition of the Ras Oncoprotein. Molecular Biology of the Cell, 2011. 22: p. Abstract No. 24.

320. Collisson, E.A., Sadanandam, A., et al., Subtypes of pancreatic ductal adenocarcinoma and their differing responses to therapy. Nature Medicine, 2011. 17(4): p. 500-503.

170