SLC22A4) in the Disposition of Nucleoside Analogs in AML

Home , Nucleoside transporter, SLC22A4

Role of OCTN1 (SLC22A4) in the Disposition of Nucleoside Analogs in AML

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

Jason T. Anderson, PharmD

Graduate Program in Pharmaceutical Sciences

The Ohio State University

2019

Dissertation Committee

Sharyn D. Baker, PharmD, PhD, Co-advisor

Alex Sparreboom, PhD, Co-advisor

Cynthia Carnes, PharmD, PhD, Committee Member

Christopher Coss, PhD, Committee Member

Jason T. Anderson

2019

1 Abstract

The nucleoside analog cytarabine (Ara-C) is among the most effective and widely prescribed anticancer agents, but its clinical use is associated with unpredictable pharmacodynamic efficacy profiles, and the agent is liable to drug-drug interactions. Due to its hydrophilic nature, Ara-C requires functional transporters to enter cells. Despite >20 years of clinical use, the transporter(s) contributing to Ara-C disposition remain poorly understood. Recently, we reported that Ara-C is a substrate of OCTN1 using overexpression models. Here, we explored the regulation of this nucleoside analog transporter, OCTN1, its impact in the context of AML (Chapter 2), and the characterization of this transporter using in vivo models (Chapter 3).

Chapter 1 focuses on introducing the disease state of Acute Myeloid Leukemia

(AML) and the many factors contributing to treatment failure in patients. Of the many causes, we and others in the field have highlighted the transport of nucleoside analog as a main contributor to treatment failure. More recently, we have provided evidence of a nucleoside analog transport system, OCTN1, which plays a role in transporting a component of the first-line regimen, Ara-C, used in AML patients. In order to better understand the regulation of this transport system, we explored how epigenetic regulation, specifically DNA methylation, can affect transporter expression.

Chapter 2 focuses on the regulation of OCTN1 and its impact in facilitating Ara-C accumulation within AML cells. First, we characterize the uptake and the sensitivities of

AML cell lines to first-line component, Ara-C. Using AML cell lines, the uptake of Ara-C (1

µM; 15 min) varied 6-fold (17-101 pmol/mg) and was highly sensitive to the classical

nucleoside transport inhibitor, NBMPR. Among other AML-directed drugs evaluated, we

found that the anthracyclines, daunorubicin, and idarubicin, inhibited Ara-C uptake in a concentration-dependent manner (P<0.05). In contrast, exposure to the hypomethylating agent 5-azacytidine or decitabine increased SLC22A4 expression, Ara-C uptake, and subsequent cytotoxicity. These findings are consistent with the notion that methylation of

SLC22A4 was inversely related to Ara-C uptake and that this transport can be modulated using hypomethylating agents such as 5-azacytidine and decitabine. Taken together, these results identify SLC22A4 (OCTN1) methylation status as a contributor to the expression of OCTN1, cellular uptake, and the efficacy of Ara-C.

Chapter 3 explores the transporter(s) contributing to systemic disposition of Ara-C using various transporter-deficient murine models. To better understand the pharmacokinetics of Ara-C, we explored a previously documented drug-drug interaction

(DDI) involving Ara-C and the classical nucleoside transport inhibitor, NBMPR. As previously shown, we were able to replicate the increase in exposure of Ara-C (2.5-fold) upon coadministration with NBMPR. To better understand the transporter(s) contributing to this increase in exposure, we utilized various transporter-deficient mouse models.

Interestingly, the interaction was unrelated to altered blood cell distribution, and subsequent studies indicated that the disposition of Ara-C was unaffected in mice with a deficiency of the postulated candidate transporters, including ENT1, OCTN1, OATP1B2, and MATE1. These studies indicate the involvement of an unknown NBMPR-sensitive

Ara-C transporter that impacts the systemic disposition of this clinically relevant agent.

Chapter 4 encompasses the future directions of this project with supporting preliminary evidence for each trajectory. Since the endogenous metabolites of OCTN1

iii have not been characterized, we wanted to address this gap in the literature using an unbiased metabolomic approach with samples from a genetically engineered knockout mouse model. Using this methodology, we found a multitude of hits (>1,600) and metabolic pathways that are affected by the genetic deletion of OCTN1 in the murine model. In parallel with metabolomics, we also characterized the changes in endogenous lipid species using this same mouse model. Ultimately, these two approaches have generated data that can be used to steer confirmative studies to identify novel endogenous metabolites of OCTN1. Lastly, Chapter 4 touches on other cancer types, specifically pancreatic cancer, in which an association between OCTN1 expression and survival was observed, similar to what we reported in AML.

2 Acknowledgments

I would like to acknowledge and thank my mentors, Dr. Sharyn Baker and Dr. Alex

Sparreboom for their unwavering support, patience, and assistance during my transition

from a student to a young investigator. I would also like to thank Dr. Cynthia Carnes, Dr.

Christopher Coss, Dr. Shuiying Hu, Dr. Daelynn Buelow and Alice Gibson for their guidance and advice throughout the PharmD/PhD program. In addition, I would like to thank Dr. Mamuka Kvaratskhelia and Dr. Ross Larue for their mentoring prior to joining the Experimental Cancer Pharmacology lab.

3 Vita

May 2012 ...... B.S., Chemistry, Youngstown State

University, Youngstown, Ohio

May 2016 ...... Pharm.D., The Ohio State University

College of Pharmacy, Columbus, Ohio

May 2016 to Present ...... Graduate Research Associate, Department

of Pharmaceutics, The Ohio State

University, Columbus, Ohio

Publications

Anderson JT, Buelow D, Pounds SB, Shi L, Lamba JK, Hu SY, Gibson A, Goodwin E,

Sparreboom A, and Baker SD. Epigenetic Regulation of OCTN1-mediated Cytarabine

Transport in Acute Myeloid Leukemia. Cancer Res. 2019. (Manuscript in preparation).

Anderson JT, Hu SY, Fu Q, Baker SD, and Sparreboom A. Role of ENT1 in the

Disposition of Cytarabine in Mice. Pharmacol Res Perspect. 2019. (Accepted).

Fu Q, Chen MQ, Anderson JT, Sun XX, Hu SY, Sparreboom A, and Baker SD. Interaction

Between Sex and Organic Anion-Transporting Polypeptide 1b2 on the Pharmacokinetics of Regorafenib and Its Metabolites Regorafenib-N-Oxide and Regorafenib-Glucuronide in

Mice. Clin Transl Sci. 2019;12(4):400-407.

Leblanc AF, Huang KM, Uddin ME, Anderson JT, Chen M, Hu S. Murine Pharmacokinetic

Studies. Bio-protocol. 2018;8(20).

Fields of Study

Major Field: Pharmaceutical Sciences

Specialization: Translational Sciences

vii

4 Table of Contents

1 Abstract ...... ii 2 Acknowledgments ...... v 3 Vita ...... vi 4 Table of Contents ...... viii 5 List of Tables ...... x 6 List of Figures ...... xi 1 Chapter 1. Introduction ...... 1 1.1 Acute Myeloid Leukemia ...... 1 1.2 Nucleoside Transporters ...... 3 1.3 Organic Cation Transporter, Novel, Type 1 (OCTN1) ...... 4 1.4 Epigenetics ...... 5 2 Chapter 2: Regulation of OCTN1 and its role in transporting nucleoside analogs in the context of AML...... 8 2.1 Introduction ...... 8 2.2 Materials and Methods ...... 11 2.3 Results ...... 15 2.3.1 Impact of transport on Ara-C sensitivity in AML cell lines...... 15 2.3.2 Impact of other AML-directed chemotherapies on the uptake of Ara-C .....16 2.3.3 AML cells are differentially methylated within the SLC22A4 promoter region 16 2.3.4 Methylation within the promoter region affects the downstream expression of SLC22A4 ...... 17 2.3.5 Pretreatment with methyltransferase inhibitors decreases methylation within the SLC22A4 promoter region ...... 18 2.3.6 Pretreatment with DNA methyltransferase inhibitors increases OCTN1 expression, uptake and subsequent sensitivity to Ara-C ...... 18 2.3.7 Methylation within SLC22A4 correlates with clinical outcomes in both adult and pediatric AML ...... 19 2.4 Discussion ...... 20 viii

2.4.1 Impact of transport on Ara-C sensitivity ...... 20 2.4.2 Regulation of OCTN1 ...... 21 2.5 Acknowledgements ...... 24 2.6 Figures ...... 25 3 Chapter 3: Exploring the in vivo role of OCTN1 and other transporters in the handling of Ara-C ...... 40 3.1 Introduction ...... 40 3.2 Materials and Methods ...... 41 3.3 Results ...... 45 3.3.1 Characterization of OCTN1(-/-) ...... 45 3.3.2 Influence of NBMPR and ENT1-deficiency on the disposition of Ara-C. ....45 3.3.3 Influence of OCTN1-, OATP1B2-, and MATE1-deficiency on the pharmacokinetic of Ara-C...... 46 3.4 Discussion ...... 48 3.5 Acknowledgements ...... 53 3.6 Figures ...... 54 4 Chapter 4: Future Directions ...... 59 4.1 Metabolomics ...... 59 4.2 Figures ...... 61 4.3 Lipidomics ...... 65 4.4 Figures ...... 66 4.5 Involvement of OCTN1 in Other Disease States ...... 68 4.6 Figures ...... 69 4.7 Acknowledgements ...... 69 5 Bibliography ...... 70 6 Appendix: Transcription Factor Binding Motifs in SLC22A4 ...... 79

5 List of Tables

Table 1: Potential causes of treatment failure in AML...... 7

Table 2: IC50 values of various AML cell lines...... 28 Table 3: Pharmacokinetic parameter estimates of Ara-C in mice ...... 47 Table 4: Pathway analysis of metabolomics features generated from female kidney samples from OCTN1(-/-) compared to wild-type samples...... 64

6 List of Figures

Figure 1: Cartoon depiction of intracellular accumulation and intracellular metabolism of Ara-C...... 25 Figure 2: NMBPR-sensitive Ara-C accumulation in various AML cell lines...... 26 Figure 3: Ara-C sensitivity of various AML cell lines...... 27

Figure 4: Correlation analysis comparing uptake of Ara-C and IC50 values...... 29

Figure 5: Correlation analysis of IC50 and mRNA expression...... 30 Figure 6: Inhibition of Ara-C uptake in HEK293 hOCTN1 overexpressor cells by various chemotherapeutic agents commonly used in AML...... 31 Figure 7: Inhibition of Ara-C uptake by AML chemotherapeutic agents in various AML cell lines...... 32 Figure 8: Cartoon depiction of epigenetic priming...... 33 Figure 9: AML cells are differentiated methylated within their SLC22A4 promoter region...... 34 Figure 10: Methylation within OCTN1 promoter region represses downstream transcription...... 35 Figure 11: Pretreatment with methyltransferase inhibitors decreases methylation within the SLC22A4 promoter region...... 36 Figure 12: Pretreatment with methyltransferase inhibitors increases OCTN1 expression and Ara-C uptake ...... 37

Figure 13: Fold change in IC50 values following epigenetic priming...... 38 Figure 14: Methylation within SLC22A4 correlates with clinical outcomes in both adult and pediatric AML...... 39 Figure 15: RT-PCR of various mouse tissues from wild-type and OCTN1(-/-) mice...... 54 Figure 16: Colony Formation Assay using lineage negative cells extracted from wild-type and OCTN1(-/-) mice...... 55 Figure 17: Influence of NBMPR and ENT1-deficiency on the pharmacokinetics of Ara-C...... 56 Figure 18: Influence of OCTN1-, OATP1B2-, and MATE1-deficiency on the pharmacokinetic of Ara-C...... 57 Figure 19: Measurement of the unchanged parent compound, Ara-C, in HEK293 cells transfected with OCTN1 using LC-MS/MS...... 58 xi

Figure 20: Principal component analysis plot for metabolomics data generated from OCTN1(-/-) or wild-type female kidney samples...... 61 Figure 21: Metabolomics data shown as a volcano plot displaying hits...... 62 Figure 22: Summary of metabolic pathway analysis of female kidney samples of OCTN1(-/-) compared to wild-type samples...... 63 Figure 23: PCA analysis of lipidomics hits...... 66 Figure 24: Volcano plot of top hits from lipidomics of bone marrow...... 67 Figure 25: Kaplan-Meier survival curves showing high and low expression of ENT1 and OCTN1 in pancreatic adenocarcinoma (PAAD) patients from the TGCA database...... 69

xii

1 Chapter 1. Introduction

1.1 Acute Myeloid Leukemia

Acute myeloid leukemia (AML) is a form of cancer that is classified by abnormal proliferation and differentiation of myeloid cells in the bone marrow compartment. In AML,

these myeloid cells rapidly expand, and the disease quickly advances; hence the term

“acute” compared to chronic leukemias such as chronic myeloid leukemia (CML). Patients

with AML release immature and abnormal cells (termed “myeloblasts”) into the

bloodstream, and their condition results in hematopoietic inefficiencies such as

neutropenia, anemia, and thrombocytopenia.1

In 2018, there were an estimated 11,000 deaths due to AML, which accounts for

the largest proportion of deaths resulting from leukemia in the US.2 Despite advances in supportive care, the backbone of therapy remains unchanged for many years consisting of a combination of cytarabine (Ara-C) and daunorubicin, termed “7+3”.3 The term 7+3 is derived from the duration of each component, with 7 days of Ara-C and the first 3 days receiving daunorubicin. Both medications are delivered via an intervenous (IV) route of

administration and thus require the administration in an inpatient setting. 7+3 is a modestly

efficacious first-line therapy for patients, who can tolerate it, that targets leukemic cells

with chemotherapeutics that have two distinct mechanisms of action. The first component,

Ara-C, is a nucleoside analog, with a similar structure to the endogenous nucleoside,

cytosine, that causes apoptosis of cells undergoing active DNA synthesis, whereas

daunorubicin intercalates between DNA and inhibits topoisomerase II activity thus causing 1 its cytotoxicity. In addition to Ara-C, additional nucleoside analogs are often utilized in AML such as fludarabine, clofarabine, and gemcitabine for use in salvage regimens (treatment after failure of first-line agents). In AML patients undergoing treatment, the overall outcome is poor with 45% long-term survival in patients aged <65 years and 10% in patients aged

>65 years.4 These disheartening survival rates can be attributed to intrinsic drug

insensitivity and/or the development of resistance, which remains a significant obstacle to

successful long-term treatment outcomes in AML.

The efficacy and response to a component of first-line therapy, Ara-C, varies

dramatically between individual AML patients. There are many potential causes of

treatment failure in AML (Table 1), and many of these mechanisms of resistance remain

poorly understood. Of the various mechanisms, Ara-C transport is the initial step to

intracellular accumulation and subsequent cytotoxicity and has been speculated to be the

major contributor to clinical resistance of nucleoside analogs therapy in AML patients.5

After transport into the cell, Ara-C must be phosphorylated to its active form, cytarabine triphosphate (Ara-CTP). Ara-CTP is the only phosphorylated intermediate that has antileukemic effects by interfering with DNA replication and causing downstream apoptosis. The formation of the active triphosphorylated form may be opposed by reversible dephosphorylation events along with irreversible deamination reactions.

Regarding dephosphorylation events, 5’ nucleotidase (NT5C2) is able to dephosphorylate

Ara-CMP and antagonize deoxycytidine kinase in the formation of the active triphosphorylated form. Cytidine deaminase (CDA) and deoxycytidylate deaminase

(DCTD) have the potential to irreversibly deaminate the cytosine ring of Ara-C and Ara-

CMP forming the inactive metabolite, Ara-Uracil (Ara-U) and Ara-U monophosphate respectively (summarized in Figure 1).

1.2 Nucleoside Transporters

There are two main solute carrier (SLC) families of transporters that have been well-characterized as nucleoside transporters, which are SLC28 and SLC29. The SLC28

family is referred to as concentrative nucleoside transporters (CNT) and are dependent

on sodium to transport nucleosides against their concentration gradient in a unidirectional

manner. There are three members of the human SLC28 family consisting of hCNT1

(SLC28A1), hCNT2 (SLC28A2), and hCNT3 (SLC28A3). CNT-type transporters are

mostly localized to the apical membrane in intestinal and renal epithelial barriers.6 In

contrast, SLC29 are equilibrative nucleoside transporters (ENT) that can transport

nucleosides in a bidirectional, sodium-independent manner and are mostly localized to the

basolateral membrane to facilitate the vectorial flux of nucleosides in concert with CNT-

type transporters.7 There are four members of the human SLC29 family consisting of

hENT1 (SLC29A1), hENT2 (SLC29A2), hENT3 (SLC29A3), and hENT4 (SLC29A4).

hENT1 and hENT2 can be functionally distinguished from the other ENT family by their

sensitivity to an inhibitor, nitrobenzylthioinosine (NBMPR), with ENT1 being much more

sensitive (>50-fold) to inhibition than ENT2.8 In addition to ENT1 being sensitive to

nanomolar concentrations of NBMPR, we have recently provided data for a novel

transporter, organic cation transporter novel type 1 (OCTN1), that is sensitive to this

classical nucleoside transport inhibitor.9

1.3 Organic Cation Transporter, Novel, Type 1 (OCTN1)

hOCTN1 was first cloned in 199710 is a 551-amino acid transporter, localized to

the plasma membrane, that contains 12 predicted transmembrane domains, with ~30%

sequence homology within other human organic cation transporters, e.g. OCT1 (31%) and

OCT2 (33%). Because of OCTN1’s ability to transport tetraethylammonium (TEA) in a pH-

dependent manner,11 it was grouped in the solute carrier family of organic cation

transporters (SLC22) despite its dissimilarities in commonly classified organic cation

substrates outside of TEA. OCTN1 has broad tissue distribution with the highest

expression shown in bone marrow, small intestine, kidney, prostate, and the appendix.10

hOCTN1 is localized to the apical surface in the small intestine12 and the proximal tubules

of the kidneys.13

Despite being cloned over twenty years ago, little is known about OCTN1’s

endogenous role. It was not until 2005 that Gründemann et al. were able to determine a

substrate specific to OCTN1 called ergothioneine, which is an amino acid derivative that

is contained in certain types of mushrooms and beans.14 Ergothioneine is not synthesized in mammals; therefore it is solely derived from the diet and not endogenously produced.

After ingestions, ergothioneine accumulates in most tissues with highest concentrations found in bone marrow, RBCs, liver, and kidney.15 Despite classifying ergothioneine’s disposition properties, very little is known about its role in cellular physiology. Some evidence has postulated ergothioneine to play an important physiologic role as a natural antioxidant,16 metal chelator17 or cytoprotective by inhibiting inflammation.18 Although no

definitive link has been established between OCTN1 and disease pathogenesis, a few

studies have found associations with functional variants of OCTN1 and susceptibility to

diseases such as Crohn’s19 or rheumatoid arthritis.20 4

To add to the complexity of OCTN1’s characterization, OCTN1 has been shown to act in various transport mechanism depending on the substrate of interest and membrane

localization. Human OCTN1 may act in a sodium-dependent mechanism of uptake of ergothioneine14 as previously described, but recent data has suggested OCTN1 may act in a sodium-independent manner for the uptake of gabapentin,21 organic cation/proton exchanger for TEA13 or acetylcholine.22 Little is known about the endogenous role of

OCTN1, but it is speculated that OCTN1’s main physiological functions are the uptake of the antioxidant ergothioneine and as a low-affinity transporter for the uptake of carnitine.

The impact of OCTN1 on xenobiotics transport has only been recently explored by the transporter field, and thus OCTN1 is not included on the Food and Drug Administration

(FDA) transporter guidance list for studying drug-drug interactions.23

Despite the in depth understanding of other organic cations transporter, the regulation of OCTN1 is still poorly understood. Gaining a better understanding of the regulation of OCTN1 will shed light on variable expression24 and subsequent drug accumulation seen in AML patients.25 Previous studies on OCTN1 have provided evidence that OCTN1 may be affected by circadian rhythm,26 circulating testosterone levels,27 and

various cytokines.28 Despite these earlier reports of OCTN1 regulation, the mechanism(s)

contributing to variable patient response to Ara-C remains poorly understood.

1.4 Epigenetics

Epigenetics is defined as the heritable changes in gene activity that alters the

chromatin, without modifying the underlying DNA sequence.29 There are many different

epigenetic processes that have been identified, such as methylation, acetylation,

phosphorylation, ubiquitylation, and sumoylation. These processes are facilitated by 5 epigenetic players that are categorized into writers, erasers, or readers.30 Writers, such as

DNA methyltransferase (DNMTs) and histone acetyltransferase (HATs), introduce various chemical modifications to DNA and histones to either signal repression or activation of a genomic region, whereas, erasers, ten-eleven translocation (TETs) and histone deacetylase (HDACs), remove these chemical tags. Lastly, readers, such as methyl CpG binding proteins (MBP), can identify and interpret these modifications and either activate or repress gene transcription. Of many types of epigenetic modifications, the best studied is the addition of a methyl group to the 5’ position of cytosine residues within the underlying

DNA sequence (termed “DNA methylation”).

Aging and exposure to carcinogens can have a direct impact on the epigenome, and act as a driver for genetic abnormalities, and can contribute to tumorigenesis.

Clinicians can modify this epigenetic landscape by using hypomethylating agents in cancer patients. These agents inhibit these DNA methylation events from occurring by causing degradation of the DNA methyltransferase (DNMT) enzyme. In terms of AML, two

FDA-approved hypomethylating agents, 5-azacytidine and decitabine, are utilized in the epigenetic modification of AML patients; mostly elderly patients who are not able to tolerate first-line “7+3” therapy.31 More recent clinical trials have looked at the sequential

combination of these hypomethylating agents (3-5 days of “epigenetic priming”) before starting first-line agents.32,33 These clinical trials have shown that epigenetic priming is well

tolerated in AML patients34 and show better clinical outcomes despite not knowing the specific mechanism of action contributing to these beneficial response rates. In addition to AML, epigenetic priming has been explored in a multitude of disease states such as gastric cancer,35 metastatic colorectal cancer,36 and T-cell lymphoma.37

Table 1: Potential causes of treatment failure in AML.

Cause Examples

Delivery of drug to leukemic cell Pharmacokinetics Penetration of drugs in sanctuary sites (e.g., CNS, testis)

Cellular uptake of drug* Intracellular drug metabolism Intracellular drug sequestration Cellular Drug Drug efflux Resistance Changes in the drug target Increased DNA repair Altered cell cycle checkpoints

Bone marrow microenvironment mediated drug resistance Regrowth of Residual Cells Drug target mutations

* Reduced uptake into tumor cells via nucleoside transporters is likely responsible for

most instances of clinical resistance to Ara-C5

2 Chapter 2: Regulation of OCTN1 and its role in transporting nucleoside analogs in the context of AML

2.1 Introduction

Ara-C was initially approved by the FDA for the use in leukemia patients in 1969 and during that time, required less rigorous preclinical data for FDA approval. Because of its efficacious profile and its ability to extend survival in patients,38 the drug was approved

by the FDA despite the lack of a clear understanding of its transport mechanism.

Early studies examining the transport of Ara-C in ex vivo AML patient samples

noted increased Ara-C uptake in blast cells that contained a higher number of NBMPR

binding sites.39 Additionally, myeloid cells had higher Ara-C transport rates, compared to lymphoid, alluding to the greater sensitivity of AML to Ara-C than other leukemias.39 Later, it was determined for lower Ara-C concentrations (<1 μM) that accumulation is primarily determined by transport and that this intracellular accumulation of Ara-C in AML cells is sensitive to inhibition by nanomolar concentrations of NBMPR.40

NBMPR was initially shown to have the ability to inhibit the family of equilibrative

transporters with varying concentrations, with ENT1 being more sensitive (0.4 nM IC50)

41 than ENT2 (2.8 µM IC50). Only until recently has it been shown that OCTN1-mediated

nucleoside analog uptake is also sensitive to nanomolar concentrations of the classical

nucleoside transporter inhibitor NBMPR.9 Since Ara-C transport in AML cells was sensitive to NBMPR inhibition, it has been assumed for decades that Ara-C accumulates in leukemic cells by an ENT1-dependent pathway. Despite the lack of robust data, ENT1 is 8 still speculated to be the primary transporter of Ara-C and other similar nucleoside analogs.42,43

Nucleoside transporters play an integral role in the disposition of anticancer nucleoside analogs and contribute to cellular sensitivities. Due to their polar nature, all nucleoside analogs require a transport-mediated process to enter cells prior to being activated via phosphorylation to their active triphosphorylated form. Reduced uptake into tumor cells via nucleoside transporters has been proposed as a process underlying most instances of clinical resistance to Ara-C, although the responsible mechanism remains poorly understood.5 Previous investigations demonstrated that low Ara-C uptake in AML

cells predicted poor response to therapy,39 and we hypothesize that this membrane transport is the major rate-limiting step in its conversion to Ara-CTP in leukemic blast cells.

This transport-mediated process contributing to clinical resistance can be derived

from inherent interpatient variability but also from external factors such as concurrent

administration of other therapies leading to drug-drug interactions (DDI). One such

instance is the interaction between both first-line therapies for AML, Ara-C and

daunorubicin. In vitro studies show a requirement of a specific 5:1 stoichiometric ratio of

Ara-C to daunorubicin for a synergistic effect, but in stark contrast, the 1:5 ratio (more

daunorubicin than Ara-C) produces an antagonistic effect.44,45 Furthermore, it is current practice to give both medications within the same timeframe on days 1-3 in the conventional 7+3 regimen, which may inadvertently be causing decreased efficacy with

drug-drug interaction by overlapping the administration. This fixed ratio dosing is not a

novel concept but has been previously described with the combination of camptothecin and doxorubicin with glioma cells.46 Using a transporter-based approach, we can start to

9 postulate the mechanism of this DDI and how one substrate has the ability to inhibit the uptake of another during concurrent administration.

In order to better understand the role of influx transporters in AML, we explored

correlations between solute carrier (SLC) expression profiles and overall survival. Using

mRNA expression from a previously completed frontline AML clinical trial, we were able

to show that a novel membrane transporter, SLC22A4 (OCTN1), played an integral role

in both overall survival and event-free survival in both adult and pediatric AML patients.9

We postulated that higher expression of OCTN1 would lead to increased intracellular

accumulation of drug and contribute to more favorable clinical outcomes. Additionally, our

cellular studies have confirmed this notion and showed that Ara-C and other structural-

related nucleosides are substrates of OCTN1 with a 70-fold increase in the uptake of Ara-

C compared to vector control cells. Interestingly, the published putative nucleoside

transporter for Ara-C, SLC29A1 (ENT1), showed only modest 2-fold increases in Ara-C

uptake compared to vector control cells. To further support the importance of OCTN1 and

its connection to AML, OCTN1 has high expression in the bone marrow and kidney,47

where it may contribute to the cellular and systemic disposition of nucleoside analogs

utilized in AML.

One hallmark of AML is dysregulation at the genetic and epigenetic level.48,49

Previously completed clinical trials have demonstrated that an increase in overall

methylation leads to worse patient outcomes in both adult50 and pediatric AML patients.51

Currently, clinical trials are underway to modify this basal methylation profile and

“epigenetically prime” AML patients using already FDA-approved DNA methyltransferase

inhibitors, such as 5-azacytidine and decitabine, prior to starting first-line therapy

(NCT03164057). These clinical trials continue to progress despite not having a clear

10 understanding of the mechanism(s) of action of the improved outcomes seen in AML patients.

In terms of epigenetic regulation of transporters, various studies have already demonstrated that altering the DNA methylation profile of organic cation transporters has the ability to affect the downstream expression of transporters such as OCT1,52 OCT2,53

OCT3,54 MATE1,55 and OCTN2.56 Surprisingly, this epigenetic regulation has not been

reported for OCTN1, which we have previously shown to be a predictor of survival in

patients with AML receiving first-line therapy with Ara-C.9 Here, we report that increased

DNA methylation within the promoter region of SLC22A4 contributes to downstream repression of OCTN1, decreased Ara-C accumulation, and subsequent decreased antileukemic effects (summarized in Figure 8).

2.2 Materials and Methods

Chemicals: Cytarabine, 3,4,5,6-tetrahydrouridine, and NBMPR were purchased from

Sigma-Aldrich (St. Louis, MO). 5-azacytidine and decitabine were purchased from LC

Laboratories (Woburn, MA). All other chemicals were purchased from Thermo Fisher

Scientific (Waltham, MA) unless otherwise specified.

Generation of stable expressing OCTN1: Full-length human OCTN1 cDNA (SC126568) was obtained from Origene, Fusion OCTN1-flag fragment were amplified by High Fidelity

PCR, and then subcloned into pIRES2-EGFP. Selected colonies were confirmed with sequencing. pIRES2-GFP-OCTN1-flag, and pIRES2-GFP were transfected using

Lipofectamine 2000 (Invitrogen) and selected using geneticin selective antibiotic (G418) and sorted using GFP. Engineered HEK293 cell were maintained in DMEM supplemented 11 with 10% fetal bovine serum (FBS), routinely checked for mycoplasma contamination

(MycoAlert Detection Kit) and cells lines were not used after 30 passages.

Cell lines: HEK293T (Invitrogen) was obtained commercially. M-07e, MOLM-13, ML-2,

and NB-4 cells were obtained from the Leibzniz-Institut (DSMZ); MV4-11, HL-60, KG-1,

THP-1, U937 from American Type Culture Collection; CHRF-288-11 provided by Tanja

Gruber (St. Jude Children’s Research Hospital); CMS from Yubin Ge (Karmanos Cancer

Institute); OCI-AML3 from Brian Sorrentino (St. Jude Children’s Research Hospital). All

cells maintained in RPMI 1640 medium supplemented with 10% FBS. All cell lines were

kept at 37 °C and 5% CO2, regularly tested for mycoplasma contamination and cell lines

were not used after 30 passages.

RNA Preparation, cDNA Generation and Reverse Transcriptase PCR: RNA was

extracted from various AML cell lines using an E.Z.N.A Total RNA Extraction

Kit (Omega Bio-tek). cDNA was generated from 2 µg of RNA using qScript XLT

cDNA Supermix (Quantabio). Real-time reverse transcriptase PCR (RT-PCR) was

performed on SLC22A4 and SLC28A1 using TaqMan probes predesigned by Thermo

Fisher Scientific with TaqMan Fast reagents (Thermo Fisher Scientific). Reactions were

carried out in triplicates, with normalization to GAPDH.

Cytotoxicity Assays: To determine cytotoxicity of various nucleoside analogs, AML cell

lines were plated in a 96-well plate with increasing concentrations of drug or vehicle and

incubated for 72 hours at 37 °C and 5% CO2. Cell viability was quantified and measured using tetrazolium (MTT) reagent and reading the absorbance at 590 nm. The

12 concentration of drug that produced 50% inhibition of cell proliferation (IC50) was

determined using GraphPad Prism software and normalized to vehicle control for each

respective cell type.

Cellular Uptake: Substrates of hOCTN1 were characterized using radioactive substrates

([3H] or [13C]) in serum-free, phenol red-free DMEM and quantified by liquid scintillation

counting using a LS 6500 Counter (Beckman). Drug uptake was normalized to total protein content using a commercially available BCA kit (ThermoFisher). Uptake experiments were repeated twice independently in triplicate and represented in figures as the mean values with SEM.

Epigenetic Priming of AML cells: AML cells were plated at a density of 2.0 x

105 cells/mL in a 6-well plate in media treated with varying concentrations of 5-azacytidine

or decitabine. The suspension media was changed daily due to the aqueous instability of

the demethylating agents.57 On day 3, AML cells were washed, subjected to uptake, MTT

assay, RT-PCR and/or the genomic DNA was harvested.

Bisulfite-Sequencing (BS-Seq): Genomic DNA was treated with bisulfite treatment using

an EZ DNA Methylation kit (Zymo Research). Primers used in the amplification and

sequencing of bisulfite-treated DNA were generated using the KAPA Library Preparation

Kit for Ion Torrent platforms and Ion Xpress Barcode Adapters. Next, library molecules

were purified using Agencourt AMPure XP beads and quantified using the

Qiagen QIAxcel Advanced System. Barcoded samples were then pooled in an equimolar

fashion before template preparation was performed on the Ion Chef System. Following

13 this, enriched, template-positive library molecules were sequenced on the Ion S5 sequencer using Ion 530 sequencing chips. FASTQ files from the Ion Torrent S5 server were aligned to the local reference database using open-source Bismark Bisulfite Read

Mapper with the Bowtie2 alignment algorithm. Methylation levels were calculated in

Bismark by dividing the number of methylated reads by the total number of reads.

Construction of pCpGL-OCTN1 Vector: The pCpGL-empty and pCpGL-CMV vectors

were obtained as a gift from the Rehli Lab.58 The 1 kb insert was PCR amplified from genomic DNA using the primers (PstI_F

5’-AAA CTGCAG AAGGGGCTCAAAATCAGAACC-3’ and BamHI_R 5’-AAA GGATCC

GAAACTTGCAACTACGGGTGA-3’). The pCpGL-OCTN1 plasmid was cloned by

inserting the PCR products using the cut sites PstI and BamHI. Proper orientation of the

1 kb sequence was confirmed by sequencing. Plasmids were then methylated, prior to transfection, using M.SssI methylase (New England Biolabs) and purified by ethanol precipitation. Completeness of the methylation was confirmed by digestion with a methylation-sensitive restriction enzyme, HpaII (New England Biolabs).

OCTN1 Dual Reporter Assay: Dual reporter assays were completed as previously described.59 In short, HEK293T cells (1.0 x 105/well) were plated overnight on polylysine- coated plates. Cells were transfected with Lipofectamine 3000 (Thermo Fisher Scientific) using either 30 ng of methylated or unmethylated promoter construct, along with 1 ng of

Cypridina TK control construct (Switchgear Genomics, SN0322S). After 24 hours, cells were lysed according to the manufacturer's protocol (Thermo Fisher Scientific) and read on a plate luminometer (SpectraMax). The constructed promoters and empty vector

14 encode a luciferase luminescent Firefly reporter gene, while the normalization TK vector encodes a Cypridina luciferase. The reporter readout of each well was normalized by taking the ratio of Firefly luciferase to Cypridina of each well. Individual replicate plates were combined by normalizing to 100% of the respective unmethylated within the same experiment.

Statistical methods: All data are presented as mean values ± SEM unless stated otherwise. Experiments were independently repeated at least twice on separate occasions. All experiments used α<0.05 to be considered significant. Correlation R values

(rs) were generated using Spearman’s rank correlation.

2.3 Results

2.3.1 Impact of transport on Ara-C sensitivity in AML cell lines

Previous reports have postulated that the initial uptake of Ara-C into leukemic cells

is the mechanism most contributing to clinical resistance in AML. With this in mind, we

wanted to test the hypothesis that uptake of Ara-C is a significant contributor to Ara-C

sensitivity. To test this hypothesis, we characterized the uptake of [3H]Ara-C, mRNA expression of OCTN1 and ENT1, and sensitivity to Ara-C in 15 AML cell lines.

In our uptake experiments, we were able to see variable uptake of Ara-C (1 µM;

15 min) in AML cell lines (6-fold; 17-101 pmol/mg), which was sensitive to NBMPR inhibition (Figure 2). Using these same AML cell lines, we saw a diverse range of IC50

values (Table 2), which we could group into “sensitive” and “insensitive” based upon their

sensitivities to Ara-C as measured with a 72-hour MTT assay (Figure 3). The cut off for

determining sensitivity was set at 300 nM, which was the concentration found at steady

15 state of patients receiving continuous infusion of 100 mg/m2 Ara-C.60 We found that overall,

sensitivity to Ara-C associated with the ability to accumulate [3H]Ara-C in our cell-based assay (Figure 4). Furthermore, we found that sensitive AML cell lines (<300 nM) had

increased Ara-C uptake (P = 0.019) and decreased uptake in insensitive cell lines (>300

nM) (Figure 4).

In addition to sensitivities, we were able to look at mRNA expression of ENT1 and

OCTN1 (Figure 5) and found no significant correlation between mRNA expression of

ENT1 and sensitivity to Ara-C (rs = 0.046, p>0.05). Surprisingly, we were able to determine a positive correlation between OCTN1 mRNA and IC50 values for Ara-C (rs = 0.55, P =

0.036), showing a survival advantage during Ara-C treatment with higher OCTN1 expression.

2.3.2 Impact of other AML-directed chemotherapies on the uptake of Ara-C

Previous evidence has shown the need for specific stoichiometric dosing when combining Ara-C with other agents such as daunorubicin.61 Based upon this notion, we

speculated that the OCTN1-mediated accumulation of Ara-C might be negatively impacted by daunorubicin. Using our hOCTN1 overexpression system, we were able to show that

anthracycline-based AML chemotherapeutics inhibited the uptake of Ara-C (Figure 6). In

addition, we were able to determine that both daunorubicin and idarubicin act in a

concentration-dependent manner to inhibit Ara-C uptake in various AML cell lines (Figure

7).

2.3.3 AML cells are differentially methylated within the SLC22A4 promoter region

In order to further understand the potential epigenetic regulation of OCTN1 through

DNA methylation, we first inspected the predicted CpG islands within proximity to 16

SLC22A4 using the UCSC Genome Browser. Of the three predicted CpG islands (Figure

9A), CpG122 was of the highest interest because it spanned the promoter area, which has been previously described as the methylation region regulating expression of other organic cation transporters.53,54,56

To extend our previously completed OCTN1 studies examining nucleoside analog uptake, we wanted to investigate the methylation of SLC22A4 in a disease-specific context, specifically AML. Using BS-Seq, we were able to compare the methylation status of the CpG island overlapping the promoter region (CpG122) in two phenotypic distinct

AML cell lines, OCI-AML3 and CHRF-288-11, which out of ten AML cell lines have been shown to have high and low Ara-C uptake, respectively (Figure 2). In our high uptake cell type, OCI-AML3, we observed an overall lower methylation profile within the promoter region, whereas our low uptake cell type, CHRF-288-11, shows higher methylation within this same region (Figure 9B).

2.3.4 Methylation within the promoter region affects the downstream expression of SLC22A4

To functionally characterize the SLC22A4 promoter activity, with and without methylation, we constructed a luciferase gene vector that contains a 1,085 bp fragment (-

928 from TSS) of the SLC22A4 promoter, which carries a proportion of CpG122, upstream of the Firefly luciferase. The specific Firefly luciferase construct was selected based on the fact it does not contain any CG dinucleotides within its luciferase sequence, which could potentially be methylated and erroneously hinder luciferase expression. The experimental Firefly luciferase plasmids were cotransfection with a normalization plasmid

(pTK), with constitutive promoter activity, to normalize transfection efficiency between methylated and unmethylated plasmids. 17

Using this dual luciferase expression system, we were able to determine that both our positive control (labeled CMV) and our OCTN1 promoter plasmid were able to drive the expression of the Firefly luciferase. After in vitro methylation with an M.SssI methylase,

OCTN1 showed a significant decrease in luciferase expression (39.1 fold; P < 0.001), with only modest changes seen in CMV (2.1 fold; P = 0.0074) and the empty constructs (2.3 fold; P = 0.0036) (Figure 10).

2.3.5 Pretreatment with methyltransferase inhibitors decreases methylation within the SLC22A4 promoter region

Since differences in methylation were observed in the tested AML cell lines and this methylation has the potential to repress downstream transcription, we explored the ability of hypomethylating agents to modify the methylation status of the SLC22A4 promoter in the context of AML cell lines. In agreement with a previously completed clinical trial,62 we treated AML cells for 72 hours prior to downstream analysis and sequencing.

Furthermore, pretreatment with decitabine (500 nM, 72 hours) decreased methylation within the SLC22A4 promoter region as determined by BS-Sequencing (Figure 11) in both

AML cell lines. These findings suggest that pretreatment with decitabine has the ability to decrease SLC22A4 methylation in AML cells.

2.3.6 Pretreatment with DNA methyltransferase inhibitors increases OCTN1 expression, uptake and subsequent sensitivity to Ara-C

Beyond changes in SLC22A4 promoter methylation, we next explored phenotypic changes that may impact cellular drug disposition following pretreatment with DNA methyltransferase inhibitors. Treatment of AML cells with nanomolar concentrations of 5- azacytidine (0-500 nM) was associated with a concentration-dependent increase in mRNA

18 expression of SLC22A4 compared to vehicle-treated cells (Figure 12A). Similarly,

concentration-dependent increases in SLC22A4 expression were observed in both AML

cell lines after treatment with decitabine (Figure 12C), although the changes in expression

were more pronounced following exposure to decitabine, in line with its increased DNA-

methyltransferase inhibitory properties.63 Of note, we observed higher fold changes in

CHRF-288-11 cells compared to OCI-AML3 cells, presumably due to the increased basal

methylation in the former cell line (Figure 9B). The observed changes in SLC22A4

expression were functionally significant in that higher uptake of Ara-C was seen in CHRF-

288-11 cells following pretreatment with 5-azacytidine (1.61-fold for CHRF-288-11 and

1.16 fold for OCI-AML3; Figure 12B) or decitabine (2.91 fold for CHRF-288-11 and 1.49

fold for OCI-AML3; Figure 12D).

Next, we evaluated the influence of increased intracellular accumulation of Ara-C in AML cells on changes in cellular sensitivity to Ara-C as measured by a 72-hour MTT assay. With 500 nM 5-azacytidine pretreatment, an increase in Ara-C sensitivity was observed in CHRF-288-11 cells (5.3 fold; Figure 13A), whereas only modest increases

(2.8 fold; Figure 13B) were noted in the low methylated OCI-AML3 cells. Similar observations were made with 500 nM decitabine in CHRF-288-11 cells (6.2 fold; Figure

13C) and OCI-AML3 cells (3.3 fold; Figure 13D). Taken together, these findings suggest that prior treatment with hypomethylating agents increases SLC22A4 expression, Ara-C uptake, and subsequent sensitivity to the antileukemic effects of Ara-C.

2.3.7 Methylation within SLC22A4 correlates with clinical outcomes in both adult and pediatric AML

Next, we wanted to determine if SLC22A4 methylation correlates with clinical outcomes in AML. Examining a pediatric AML cohort, we found that methylation values of 19

18 probe-sets were significantly correlated with the expression values of SLC22A4 (CC

R2 = 0.53; P = 0.0002). Furthermore, the rate of relapse was associated with SLC22A4

methylation (P = 0.016) and increased SLC22A4 expression was associated with a

reduced rate of relapse (P = 0.045), after adjusting for risk factors in each group. In

addition, using adult data from The Cancer Genome Atlas (TCGA) LAML dataset, we

found that increased SLC22A4 methylation correlated with lower survival in adult AML

patients, after adjusting for age and sex differences (Figure 14).

2.4 Discussion

2.4.1 Impact of transport on Ara-C sensitivity

Previous reports have eluted to the importance of Ara-C uptake to the contribution

of clinical resistance in AML patients.5 In our studies, we used a sampling of 15 different

AML cell lines to help characterize the contribution of uptake to the antileukemic effects of

Ara-C. Our sampling of AML cell lines yielded various phenotypes with variable uptake and sensitivity to Ara-C. Using this data, we were able to show that Ara-C uptake significantly correlated with IC50 values (rs = -0.60; P = 0.020) (Figure 4A) and that AML cells can be characterized into sensitive or insensitive based upon their uptake values

(Figure 4B).

Using these same sensitivity values, we were able to show no significant correlation between IC50 values and ENT1 expression (P > 0.05), which the body of literature asserts is the putative transporter contributing to Ara-C accumulation.

Surprisingly, we were able to determine a positive correlation between OCTN1 expression and IC50 values (rs = 0.55, P = 0.037). Interestingly, this data provides evidence of a

beneficial impact of OCTN1 expression during treatment of Ara-C. Although this is 20 counterintuitive, this notion is not unreasonable based on the lack of knowledge regarding

the endogenous substrates of OCTN1, including ergothioneine, that could be contributing

to a survival advantage by lowering reactive oxygen species in opposition to OCTN1- mediated Ara-C influx.64

In addition to cellular-specific factors, we saw, in a concentration-dependent manner, inhibition of Ara-C accumulation by both daunorubicin and idarubicin in our HEK

OCTN1 overexpression system (Figure 6) and AML cell lines (Figure 7). This data directly shows that the anthracycline component of the first-line treatment option for AML inhibits

the accumulation of the other first-line component, Ara-C, via an OCTN1-mediated mechanism. The notion that anthracyclines inhibit the uptake of Ara-C is supported by the specific requirement for (Ara-C:daunorubicin) ratiometric dosing, with 5:1 (more Ara-C) producing synergistic effects while 1:5 (more daunorubicin) producing antagonist effects during an in vitro screen against the leukemic cell line P388.61 Additionally, these findings

are supported by the development of an FDA-approved daunorubicin/cytarabine lipid formulation (Vyxeos®; Jazz Pharmaceuticals), which requires a specific 5:1 stoichiometric dosing for synergistic efficacy.

2.4.2 Regulation of OCTN1

The epigenetic regulation of transporters contributes to a dynamic interplay between intracellular substrate concentrations and the extracellular environment of various cells types. Disruption of this sensitive balance has the potential to modify intracellular accumulation of substrates and may contribute to variable drug resistance or toxicity. This epigenetic regulation can be exploited in cancer cells by restoring the expression of drug influx transporter in resistant cells and impacting the accumulation of cytotoxic drugs. 21

Although it was previously reported that SLC22A4 expression is impacted by cytokines28 and hormones,27 the contribution of DNA methylation in the epigenetic

regulation of SLC22A4 has remained unclear. As previously noted, various other organic cation transporters are regulated by the DNA methylation status of their corresponding upstream promoters. Using this prior knowledge, we hypothesized that OCTN1 is

regulated in this same manner. In particular, we speculated that methylation within the

CpG island (CpG122) located in the SLC22A4 promoter region could cause repression of

the downstream transcription of OCTN1 and subsequently lead to decreased intracellular

accumulation of transported substrates. In addition, we postulated that we can prevent

these methylation events using hypomethylating agents and restore transporter

expression in AML cell lines.

To confirm our hypothesis, we examined the DNA methylation profile of the CpG island (CpG122), which corresponds with the promoter region of SLC22A4, and found that cells exhibiting low intrinsic uptake of Ara-C (CHRF-288-11) show increased SLC22A4 methylation as compared with cells exhibiting high intrinsic uptake of Ara-C (OCI-AML3).

In addition, we found that a 1 kb region of the SLC22A4 promoter is highly sensitive to

DNA methylation, which can inhibit the downstream transcription of OCTN1, and that exposure to hypomethylating agents decreases DNA methylation within this region.

Indeed, nanomolar concentrations of decitabine were able to decrease DNA methylation in CHRF-288-11 cells, although restoration to basal levels of OCI-AML3 were not achieved. This inability to completely restore methylation levels to unmethylated cell types points to the existence of other mechanisms of epigenetic regulation beyond the scope of this work, such as histone methylation.65

Several prior reports have confirmed that DNA hypomethylating agents can

sensitize various cancer cells to cytotoxic agents66-68 although the mechanistic basis for the observed increases in cytotoxicity, compared with single-agent use, have remained

elusive. Our studies provide a plausible mechanistic explanation for these prior findings

and suggest that increase in cytotoxicity following epigenetic priming stems, in part, from

restoring the expression of transporters of relevance to the intracellular accumulation and

subsequent cytotoxicity of chemotherapeutic agents. These findings are particularly

relevant in light of the previous observation that high SLC22A4 expression in AML blasts

is associated with distinct survival advantages.24

The methylation of CpG dinucleotides within SLC22A4 and subsequent inhibition of expression are potentially occurring through two different mechanisms. First, the methylation of specific CpG dinucleotides may disrupt the binding of specific transcription factors that regulate OCTN1 such as Sp1 and RUNX1.28 In agreement with this notion, we observed these specific transcription factor binding motifs within our 1 kb OCTN1 region

(Appendix), which we found to have strikingly different methylation profiles in AML cell lines. Secondly, binding of methyl-CpG binding proteins may be altered by the DNA methylation status of SLC22A4.69 Further work is warranted to address the specific

mechanism contributing to the altered expression of SLC22A4 following DNA methylation.

It should be pointed out that changes in cell growth following epigenetic priming

may impact cellular sensitivities to drugs. Indeed, treatment with hypomethylating agents

has been shown to decrease doubling times in HL60 and T24 cells,70 which would make the cells less sensitive to the S-phase dependent cytotoxicity of agents like Ara-C.

Contrary to this, we found increases in sensitivity following epigenetic priming, suggesting

that the observed phenotypes are unrelated to possible effects on changes in cell growth.

Collectively, our findings suggest that DNA methylation within the promoter region of SLC22A4 has a direct impact on OCTN1 expression, function, and sensitivity to Ara-C.

These findings are in agreement with clinical data showing a distinct survival advantage

of low SLC22A4 methylation in both pediatric and adult AML patients. The reported

epigenetic regulation of OCTN1 may have immediate translational relevance, and

provides a mechanistic explanation for previously observed antileukemic properties

associated with the combined use of hypomethylating agents with Ara-C. In addition, the

described regulation pathway of OCTN1 may help explain the varying expression levels

of OCTN1 and subsequent variable pharmacodynamic responses seen in patients

receiving Ara-C based regimens.

2.5 Acknowledgements

We want to thank the Rehli lab (University of Regensburg, Germany) for the CpG-free

luciferase plasmid, and Dr. Moray Campbell for assistance with CpG island prediction.

I am grateful for Drs. Daelynn Buelow and Shuiying Hu for assistance in the BS-

Sequencing, Alice Gibson and Marissa Pioso for their contributions in uptake studies, Dr.

Stanley Pounds for his help in correlating clinical outcomes, Emily Goodwin with her

contribution to dual luciferase assay, and Elizabeth Muhowski for her participation with the

inhibition assays.

I want to thank the Sparreboom, Baker, and Hu lab for producing and providing reagents

for this study. This project was guided and conceived by Dr. Alex Sparreboom and Dr.

Sharyn Baker.

2.6 Figures

Figure 1: Cartoon depiction of intracellular accumulation and intracellular metabolism of Ara-C. Abbreviations: cytarabine (Ara-C), Ara-C monophosphate (Ara-CMP), Ara-C diphosphate (Ara-CDP), Ara-C triphosphate (Ara-CTP), Ara-uracil (Ara-U), deoxycytidine kinase (DCK), cytidine monophosphate kinase 1 (CMPK1), nucleotide-diphosphate kinase (NDPK)

Figure 2: NMBPR-sensitive Ara-C accumulation in various AML cell lines. [3H]-Labeled Ara-C uptake (1 µM; 15 mins) in various AML cell lines ± classic nucleoside transport inhibitor nitro-benzyl-mercapto-purine ribonucleoside (NBMPR; 10 µM). Data are shown as mean ± SD.

Figure 3: Ara-C sensitivity of various AML cell lines. Sensitivities measured by a 72-hour MTT assay. Cell lines grouped by sensitivities in the following groups: (A) sensitive (<200 nM), (B) intermediate (200-500 nM) and (C) resistant (>500 nM). Data shown as mean (n=6 each concentration) ± SD.

Table 2: IC50 values of various AML cell lines. Data shown as mean of two independent experiments conducted twice (n=6 each concentration) ± SEM.

AML Cell Line IC50(µM) CTS 0.014 ± 0.002 WSU-AML 0.0175 ± 0.002 HEL 0.0826 ± 0.08 ML-2 0.103 ± 0.03 U937 0.130 ± 0.03 KG-1 0.172 ± 0.06 Molm-13 0.207 ± 0.07 PL-21 0.26 ± 0.2 Mo7e 0.332 ± 0.1 NB4 0.482 ± 0.4 OCI-AML3 1.60 ± 0.2 MV4-11 2.15 ± 0.7 CHRF 2.92 ± 1.0 THP-1 9.75 ± 0.08 MEG-01 >10

3 Figure 4: Correlation analysis comparing uptake of Ara-C and IC50 values. (A) [ H]Ara-C and sensitivity to unlabeled Ara-C as measured by MTT assay. The correlation value (rs) was calculated using a two-sided Spearman test. (B) Grouping of the AML cells based on IC50 values into “Sensitive” or “Insensitive” based on plasma levels from patients receiving continuous infusion of Ara-C (100 mg/m2).60 p-value was calculated using a two-sided unpaired t test. Data are shown as averaged uptake (n=6) and IC50 values (n=12 each concentration) of two independent experiments.

Figure 5: Correlation analysis of IC50 and mRNA expression. (A) Correlation analysis of Ara-C sensitivities and SLC22A4 mRNA expression and (B) SLC28A1 mRNA expression. Data shown as averaged values of two independent experiments ± SD. Correlation analysis and p-values were calculated using a two-sided Spearman correlation (rs).

Figure 6: Inhibition of Ara-C uptake in HEK293 hOCTN1 overexpressor cells by various chemotherapeutic agents commonly used in AML. 15-minute preincubation of inhibitor (10 µM) then addition of 1 µM Ara-C for a 5-minute uptake. Data shown as mean values ± SEM; asterisks signify significance from DMSO-treated as determined by t test.

Figure 7: Inhibition of Ara-C uptake by AML chemotherapeutic agents in various AML cell lines. 15-minute preincubation of anthracycline inhibitor or DMSO then addition of 1 µM Ara-C for a 5-minute uptake. Data shown as mean values ± SEM.

Figure 8: Cartoon depiction of epigenetic priming. (A) Panel showing a representative cell with high SLC22A4 methylation and lower intracellular Ara-C accumulation. (B) Epigenetically primed cell showing decreased SLC22A4 methylation and increased OCTN1 expression, and higher intracellular Ara-C accumulation.

Figure 9: AML cells are differentiated methylated within their SLC22A4 promoter region. (A) Cartoon depiction of the OCTN1 promoter region with overlapping CpG island. (B) Methylation profiles as determine by BS-Sequencing of AML cell lines OCI-AML3 and CHRF-288-11, with “high” and “low” Ara-C uptake respectively.

Figure 10: Methylation within OCTN1 promoter region represses downstream transcription. (A) Cartoon depiction of the three transfected plasmids with and without in vitro methylation. (B) Relative luciferase readings shown as the ratio of Firefly luciferase (CpG-free plasmids) to Cypridina luciferase (pTK) expression. Data shown as the mean of two independent experiment ± SEM.

Figure 11: Pretreatment with methyltransferase inhibitors decreases methylation within the SLC22A4 promoter region. (A) Percent methylation as determined by BS-Seq of OCTN1’s promoter region following 3-day treatment with 500 nM decitabine or aqueous control.

Figure 12: Pretreatment with methyltransferase inhibitors increases OCTN1 expression and Ara-C uptake (A) Fold change in SLC22A4 mRNA expression (normalized to GAPDH) and (B) fold change in [3H]-Ara-C uptake (1 μM; 15 mins) after 72-hour 5- azacytidine treatment. (C) Fold change in SLC22A4 mRNA expression (normalized to GAPDH) and (D) fold change in [3H]-Ara-C uptake (1 μM; 15 mins) after 72-hour decitabine treatment.

Figure 13: Fold change in IC50 values following epigenetic priming. Sensitivities determined by 72-hour MTT assay with varying concentrations of Ara-C in (A) CHRF-288- 11 and (B) OCI-AML3 after 72-hour treatment with 500 nM 5-Aza. Fold change in (C) CHRF-288-11 and (D) OCI-AML3 after 72-hour treatment with 500 nM decitabine. Data are shown as mean ± SEM of representative experiment. Fold change in IC50 value shown as the mean of two independent experiments with standard deviation in parentheses.

Figure 14: Methylation within SLC22A4 correlates with clinical outcomes in both adult and pediatric AML. Kaplan-Meier curves of OCTN1 methylation status of blast samples and overall survival in (A) pediatric (cg26794221) AML patients enrolled in AML02 (NCT00136084). (B) Survival analysis examining SLC22A4 methylation (cg04470557) in adult AML patients from the TCGA database (LAML) using MethSurv platform.71

3 Chapter 3: Exploring the in vivo role of OCTN1 and other transporters in the handling of Ara-C

3.1 Introduction

Acute myeloid leukemia (AML) is a form of cancer that is classified by an abnormal proliferation and differentiation of myeloid cells within the bone marrow compartment.

Despite advances in supportive care, the backbone of therapy has remained unchanged for over 30 years consisting of cytarabine (Ara-C)-based combination regimens.3 The

efficacy and response to Ara-C varies dramatically between individual AML patients and

is dependent on uptake,72 intracellular activation,73 and deamination.74 The transport of

Ara-C is the initial step to intracellular accumulation and subsequent cytotoxicity and has been speculated to be the major contributor to clinical resistance of nucleoside analogs therapy in AML.5

Previous studies examining the transport of Ara-C in AML cells indicated a correlation between Ara-C uptake and the number of binding sites for nitrobenzylmercaptopurine ribonucleoside (NBMPR),39 and showed that intracellular accumulation of Ara-C in AML cells is sensitive to inhibition by nanomolar concentrations of NBMPR.40 Since Ara-C accumulation was sensitive to nanomolar concentration of

NBMPR, it was concluded that Ara-C accumulation in leukemic cells was facilitated by an

equilibrative nucleoside transporter rather than a concentrative nucleoside transporter,

which are not sensitive to nanomolar concentrations of NBMPR. Interestingly, in addition

to ENT1 being sensitive to nanomolar concentrations of NBMPR, we have recently

40 provided data for a different transporter, OCTN1, that is also sensitive to this classical nucleoside transport inhibitor.9 Based on these observations, it has been presumed for

decades now that Ara-C accumulates in leukemic cells by a mechanism that is dependent

on the nucleoside transporter ENT1.42,43

There have been many studies examining ENT1’s role in the cellular disposition of

Ara-C, but the transport mechanism(s) contributing to the systemic disposition of Ara-C remains poorly characterized. While no studies have directly looked at the transporter(s) contributing to systemic Ara-C disposition, reports have shown that prior administration of

NBMPR confers cytoprotection after a lethal dose of similar nucleoside analogs such as

fludarabine75 and tubercidin,76 which may be attributed to decreased uptake by nucleoside

transporters.

Although this thesis is consistent with the notion that concurrent administration of

NBMPR causes a possible pharmacokinetic drug-drug interaction (DDI) with Ara-C in

mice,77 the mechanistic details of this finding remains unclear. The aim of this current work was to re-examine the possible interaction between NBMPR and Ara-C using mouse models that are genetically deficient for ENT1 and other putative Ara-C carriers.

3.2 Materials and Methods

Chemicals. Cytarabine, 3,4,5,6-tetrahydrouridine, and NBMPR were purchased from

13 15 Sigma-Aldrich (St. Louis, MO). [ C, N2]-Cytarabine was purchased from Alsa Chim

(Illkirch Graffenstaden, France). All other chemicals were purchased from Thermo Fisher

Scientific (Waltham, MA) unless otherwise specified.

Murine Pharmacokinetic Studies. All pharmacokinetic studies were conducted as previously described78 and with controls that were matched for age and strain: C57BL/6J for OCTN1(-/-) and ENT1(-/-); DBA/1lacJ for OATP1B2(-/-); FVB/NJ for MATE1(-/-). All mice

were female, between 8-12 weeks of age, and were housed in a temperature-controlled

environment with a 12-hour light/dark cycle. All mice received a standard diet, water ad

libitum, and were housed and handled in accordance with the Institutional Animal Care

and Use Committee of The Ohio State University and following national guidelines and

regulations.

For in vivo studies, Ara-C (dose, 15 mg/kg unless otherwise stated) was dissolved in phosphate-buffered saline (PBS) for intraperitoneal (IP) and intravenous (IV)

injection. NBMPR (dose, 100 mg/kg) was suspended in normal saline for oral gavage (PO)

1 hour prior to Ara-C dosing; control animals received the same volume of normal saline

(PO) prior to Ara-C. At select time points after administration, blood was collected in

heparinized capillary tubes from individual mice via cheek bleeding, retro-orbital bleeding,

and cardiac puncture for the final time point. Collection tubes were coated with cytidine

deaminase inhibitor, 3,4,5,6-tetrahydrouridine (THU), to prevent the metabolic degradation of Ara-C during the collection period. Isoflurane was used as an anesthetic prior to retro-orbital bleeds. For plasma analysis, blood samples were centrifuged at

11,000 g for 5 minutes, and plasma was separated and stored at –80°C until analysis by a validated method based on liquid chromatography-tandem mass spectrometry (LC-

MS/MS).24 In brief, analytes of interest were extracted from plasma or whole blood using

a methanol-based method. Extracted samples were spiked with the internal standard

13 15 [ C, N2]-cytarabine and then diluted with HPLC grade water. Samples were then

analyzed on a Vanquish UHPLC system and a TSQ Quantum Ultra triple quadrupole

42 mass spectrometer (Thermo Fisher Scientific). Noncompartmental pharmacokinetic

parameter estimates were obtained using Phoenix WinNonlin 7.6 software (Certara,

Princeton, NJ). The concentration of Ara-C in erythrocytes was derived from the previously

reported relation between whole blood and plasma concentration,79 as follows:

Equation 1: = + (1 )

𝐶𝐶𝐶𝐶 𝐻𝐻 ∗ 𝐶𝐶𝐶𝐶𝐶𝐶 − 𝐻𝐻 ∗ 𝐶𝐶𝐶𝐶 In this equation, H, Cb, Cbc, and Cp represent hematocrit, blood, blood cell, and total

plasma concentration of Ara-C, respectively. Hematocrit levels were obtained from

reference values reported in The Jackson Laboratory Physiological Data Summary for

C57BL/6J mice.

Magnetic Cell Sorting (MACS), Flow Cytometry and Fluorescence Activated Cell

Sorting (FACS): Following bone marrow extraction80 and RBC lysis with ammonium chloride, hematopoietic stem cells were incubated with magnetic direct lineage depletion beads (Miltenyi Biotec) and loading onto a MACS LS column (Miltenyi Biotec) and washed with LS buffer (Milltenyi). Lineage negative cells were able to flow through untouched, whereas the lineage positive population was obtained by removing the column from the magnetic field and eluting with LS buffer.

After lineage depletion, lineage negative cells were stained using Sca1 and c-Kit antibodies (BD Biosciences). After 1-hour incubation with each stain, cells were washed with PBS, passed through a 35 µm filter, sorted on a BD FACS Aria III and gated for the

Lin-Sca1+c-Kit+ (LSK) population for collection in PBS.

Colony Formation Assays: Clonogenic assays were set up by plating lineage negative

cells derived from either WT or OCTN1(-/-) mice at a density of 10,000 cells/mL in 43 methylcellulose medium (MethoCult, Stem Cell Technologies) with or without cytarabine

(25-200 nM). After culturing for 7 days, the number of colonies were counted with a 2x

objective lens by two separate researchers and then averaged for the representative

count.

Metabolomics: Untargeted analysis was performed on an Agilent 6545 Quadrupole Time-

of-Flight (QTOF) with an Agilent 1290 Infinity HPLC system. In brief, tissue and plasma

samples were weighed, homogenized by sonication and then extracted using a 2:1

methanol/chloroform solution. After extraction, samples were vacuum died on a speedvac

and resuspended in 1:1 water/methanol solution before injecting.

For metabolomic profiling, samples were pooled and analyzed for total feature

alignment. Relative fold change was determined for each set of samples compared to wild-

type controls. P-values were determined using ANOVA between transporter-deficient and

wild-type samples. Features were annotated using KEGG database. Pathway analysis

was performed using the MS peak pathway finder of MetaboAnalyst using a 0.001 p-value

and a 0.1 ppm mass error cutoff.

Lipidomics: Bone marrow was harvested from female wild-type and OCTN1(-/-) mice and

purified for lineage negative bone marrow cells as previously described using the MACs

column. Untargeted metabolomics studies were conducted using a SelexION QTrap 5500

(Sciex) with a SelexION differential ion mobility spectrometry device for enabling the

Lipidyzer platform. Samples were normalized to cell counts during analysis.

Statistical Analysis. An unpaired two-tailed Student’s t-test was used to determine group differences, and P < 0.05 was considered a cutoff for statistical significance. Analyses were performed using GraphPad Prism (La Jolla, CA).

3.3 Results

3.3.1 Characterization of OCTN1(-/-)

Using mRNA extracted from various tissues, we were able to confirm the high expression of murine OCTN1 in both bone marrow and kidney, along with the absence of expression in OCTN1(-/-) mice (Figure 15). Using lineage negative cells extracted from

wild-type and OCTN1(-/-) mice, we were able to culture these cells in variable

concentrations of Ara-C for seven days. Compared to wild-type, OCTN1(-/-)-derived bone

marrow cells showed lower sensitivity to Ara-C as determined by increased colony-forming

units on day 7 compared to wild-type derived cells. (Figure 16).

3.3.2 Influence of NBMPR and ENT1-deficiency on the disposition of Ara-C.

To better understand the transporter(s) contributing to system disposition of Ara-

C, we utilized a two-pronged approach with pharmacologic inhibition of nucleoside

transport systems using the classical nucleoside inhibitor, NBMPR, along with genetic

knockout of the postulated carriers contributing to the cellular transport of Ara-C. Similar

to what was reported earlier by Cass et al., we found that NBMPR given to wild-type mice

prior to Ara-C resulted in 2.5-fold (P = 0.0042) increased concentrations in plasma (Figure

17A).

Since ENT1 is highly expressed in circulating erythrocytes, we next examined the

possibility that NBMPR may inhibit the distribution of Ara-C into erythrocytes and causes 45 altered whole blood distribution. In our murine studies, however, NBMPR only modestly increased Ara-C concentrations in whole blood (Figure 17B) and had a negligible impact on the distribution of Ara-C to erythrocytes (Figure 17C). Next, we wanted to evaluate the role ENT1 in the disposition of Ara-C using transporter deficient mice. After a 15 mg/kg IP dose of Ara-C, we found that the pharmacokinetic properties of Ara-C were not substantially altered by ENT1-deficiency, as evidenced by the unchanged concentration-

time profiles in plasma (Figure 17D), whole blood (Figure 17E) and resulting AUCs shown

in Table 3 (P = 0.61 and P =0.12, respectively), as compared with results obtained in wild-

type mice.

3.3.3 Influence of OCTN1-, OATP1B2-, and MATE1-deficiency on the pharmacokinetic of Ara-C.

To evaluate alternative transport mechanisms involved in the NBMPR-Ara-C interaction, we next considered a possible contribution by the ergothioneine transporter,

OCTN1. Similar to our ENT1 pharmacokinetic studies, we repeated this format in wild- type and OCTN1(-/-) mice but found that OCTN1-deficiency did not influence the levels of

Ara-C in plasma (Figure 18A) or whole blood (Figure 18B). Despite the major

involvement of the liver in regulating the systemic exposure to Ara-C, we found that the

plasma levels of Ara-C were not significantly altered by the deficiency of OATP1B2

(Figure 18C) or MATE1 (Figure 18D).

Table 3: Pharmacokinetic parameter estimates of Ara-C in mice

Dose Cmax AUC∞ Group ID Matrix T1/2 (h) (mg/kg) (µg/mL) (ng×h/mL) Wild-type 0.44 Plasma 15 7.79 (1.97) 7.02 (1.93) (Vehicle) (0.16) Wild-type 0.76 Plasma 15 21.0 (5.32) 17.5 (4.46) (NBMPR) (0.16) Wild-type Whole 0.52 5.48 15 5.63 (1.23) (Vehicle) Blood (0.03) (0.901) Wild-type Whole 0.80 15 8.54 (2.39) 9.19 (2.14) (NBMPR) Blood (0.21)

0.96 Wild-type Plasma 15 10.4 (1.01) 9.91 (2.13) (0.07) 0.92 10.7 ENT1(-/-) Plasma 15 10.6 (0.605) (0.08) (0.580) Whole 1.28 Wild-type 15 10.6 (2.33) 11.6 (2.41) Blood (0.31) Whole 1.63 9.82 ENT1(-/-) 15 14.0 (0.913) Blood (0.24) (0.826)

0.85 Wild-type Plasma 15 11.1 (2.13) 14.4 (4.26) (0.33) 0.48 OCTN1(-/-) Plasma 15 11.9 (2.44) 10.8 (2.03) (0.10) Whole 0.52 5.48 Wild-type 15 4.32 (1.12) Blood (0.03) (0.901) Whole 0.57 OCTN1(-/-) 15 6.18 (1.50) 5.24 (1.16) Blood (0.06)

1.19 Wild-type Plasma 10 14.5 (2.11) 18.0 (1.17) (0.11) 1.03 13.4 OATP1B2(-/-) Plasma 10 14.4 (1.24) (0.07) (0.484)

0.98 Wild-type Plasma 100 70.8 (2.89) 106 (21.1) (0.57) 0.66 MATE1(-/-) Plasma 100 68.5 (6.70) 92.9 (13.8) (0.17)

*Data shown as mean ± SD in parenthesis using 3-4 animals per group.

Abbreviations: T1/2, half-life of the terminal phase; Cmax, peak concentration; AUC, area under the

concentration-time curve. 47

3.4 Discussion

In agreement with previously published reports,12,81 we were able to confirm the

expression profile of OCTN1 in our wild-type (C57BL/6J) and the absence of expression

in our knockout mice (Figure 15). Of note, OCTN1 has high expression in kidney, where

it could potentially play a role in the handling of renally eliminated drugs, and bone marrow,

the site of action for antileukemic nucleoside analogs. Furthermore, the expression of

OCTN1 increases during the maturation of hematopoietic cells from immature (lineage

negative) to mature (lineage positive) blood components, which may signal a potential role

of OCTN1 in stem cell differentiation.

To evaluate the role of murine OCTN1 in the impact on intracellular accumulation

of Ara-C, we conducted ex vivo studies utilizing the stem cell-like (lineage negative

population) from wild-type and OCTN1(-/-) mice. Using varying concentrations of Ara-C, we were able to show that OCTN1(-/-)-derived bone marrow cells were less sensitive to Ara-C treatment than wild-type-derived cells (Figure 16). This is in alignment with the notion that removal of OCTN1 would decrease the intracellular accumulation of Ara-C. Additionally, cells deficient in OCTN1 still showed decreased colony formation at higher Ara-C concentrations alluding to the notion of other transport systems contributing to the intracellular accumulation of Ara-C beyond OCTN1.

Using our in vitro assays, we were able to show that cellular disposition of Ara-C is impacted by OCTN1, but it is still unclear if OCTN1 plays a role in the systemic disposition of Ara-C. Despite the many years of clinical use, the transporter(s) impacting the systemic disposition properties of Ara-C remain poorly understood. To gain insights into this field, we took advantage of a previously reported drug-drug interaction between

Ara-C and NBMPR, an agent with potent inhibitory properties toward the putative Ara-C 48 uptake transporter, ENT1, which we were able to recapitulate in both plasma (Figure 17A)

and whole blood (Figure 17B). We opted to examine both plasma and whole blood in this

study due to the notion that erythrocytes have a known ability to sequester similar

antimetabolites such as 5-fluorouracil,82 and that the antiviral nucleoside analog, ribavirin,

accumulates into erythrocytes via an ENT1-mediated mechanism.83 Interestingly, Cass et al. did report similar pharmacokinetic alterations in their mouse model but dismissed these

findings as insignificant due to its impact on therapeutic potentiation rather than a

mechanistic basis as is the aim of this study. Of note, we also noticed a slight increase in

half-life after the addition of NBMPR, which may have arisen from a decreased uptake into

a compartment of high Ara-C metabolism, such as the liver,84 rather than an intrinsic ability of NBMPR to inhibit the deamination reaction of Ara-C.

After confirming the presence of a DDI with the classical nucleoside inhibitor,

NBMPR and Ara-C, we wanted to test if this increase in exposure was due to the inhibition

of ENT1. Currently, there is a lack of publications directly addressing the nucleoside transporter(s) contributing to the systemic disposition of Ara-C using an in vivo model. In vitro data shows the role of ENT1 and to a lesser extent, ENT2, to contribute to cellular uptake of Ara-C.42 In regards to other nucleoside transporters contribution in our mouse

model, a dose of 15 mg/kg NBMPR (IP) generated NBMPR plasma levels >1 µM and was

shown to accumulate in RBCs.85 With this in mind, NBMPR could potentially inhibit other

nucleoside transporters, such as CNTs, but based upon in vitro data, ENTs contribute to

cellular disposition of Ara-C to a much greater extent than CNTs.42,43

In our study and contrary to the literature that suggests Ara-C is a substrate of

ENT1, removal of the transporter has no impact on Ara-C disposition in both plasma

(Figure 17D) and whole blood (Figure 17E). This apparent lack of erythrocytes

49 contributing to the in vivo blood distribution of Ara-C (Figure 17F) is consistent with the finding that Ara-C does not substantially interact with erythrocyte membranes in a nonspecific manner,86 and with results from a recent study indicating that the contribution

of ENT1 to the cellular uptake of Ara-C in vitro is minimal.24

Since we were unable to recapitulate the increase in Ara-C exposure seen in the

NBMPR DDI with ENT1 deficiency, we wanted to explore other transport systems contributing to Ara-C disposition, such as OCTN1. In support of this notion, OCTN1 is highly expressed in erythrocytes,81 has been previously linked with Ara-C transport in myeloid cells, and is sensitive to inhibition by NBMPR.24 Involvement of OCTN1 would

also be consistent with the previous finding that opossum kidney proximal tubular cells

express an (unidentified) organic cation transporter that recognizes tetraethylammonium,

and that is sensitive to inhibition by several nucleosides, including Ara-C.87 Interestingly,

we saw no difference in the systemic disposition in plasma (Figure 18A) or whole blood

(Figure 18B) between OCTN1(-/-) and wild-type mice. This is somewhat unexpected given that genetic deficiency of OCTN1 was previously associated with altered plasma levels and uptake of substrates such as ergothioneine in organs of elimination (e.g., kidney, liver) and distribution (e.g., heart).15 It is possible that the inability to translate documented in

vitro Ara-C uptake data to an in vivo scenario is caused by a species-dependent interaction

between human or mouse OCTN1. Such interspecies differences in uptake mechanisms

for xenobiotics have been previously recorded for various drug-transporter pairs, including

sorafenib88 and digoxin.89

In this context, it is worth pointing out that the intracellular accumulation of Ara-C

was recently found not to be influenced in a cell-based model engineered to overexpress

human OCTN1.90 In this analysis, the authors used an LC-MS/MS-based method to

50 measure the intracellular levels of unchanged Ara-C, whereas in our original studies, we used radiolabeled drug and analyzed total radioactivity [i.e., the total of parent drug and

metabolite(s)]. This is an important methodological difference as Ara-C can undergo rapid

enzyme-mediated metabolism once inside cells to form mono-, di-, and tri-phosphorylated

forms91 that may easily escape detection and results in underestimating the actual extent

of uptake. The extensive formation of phosphorylated Ara-C metabolites was previously

demonstrated in the HEK293 cells used in our experiments.24 Using the same model, we have now confirmed in a comparative analysis that intracellular levels of total radioactivity

originating from Ara-C in cells overexpressing OCTN1 is high, while levels of the

unchanged parent drug as measured by LC-MS/MS remains undetectable (Figure 19).

As a next step toward understanding the mechanisms underlying the NBMPR-Ara-

C interaction, we evaluated the contribution of the liver and kidney as vital organs of

elimination. OATP1B2, an uptake transporter localized to the liver, has been previously

described as a transporter of Ara-C using in vitro model systems.92 This supports the

possibility that OATP1B2 could act as a putative carrier of relevance to drug interactions

with Ara-C. In addition, we explored a potential connection with the renal transporter

MATE1, based on the previous observation that concurrent administration of the

nucleoside analog clofarabine, with either Ara-C or fludarabine, results in a marked

change in clofarabine clearance compared to clofarabine given alone, suggesting an

interaction at the level of a renal apically-localized transporter.93 A connection with MATE1

is further supported by the notion that certain antiviral nucleosides are transported by

organic cation transporters,94 and a direct connection has been suggested for rodent organic cation transporters in relation to Ara-C.95 In evaluating these alternative

mechanisms, only plasma samples were obtained in the pharmacokinetic studies involving

OATP1B2- or MATE1-deficiency due to the minimal expression of these transporters outside of the liver96 and the kidney,97 respectively. These findings support the notion that,

in addition to the Ara-C transporter ABCC4,92 MATE1 can now be discounted as a key

renal tubular secretion/reabsorption pathway for Ara-C that is liable to clinically relevant

drug-drug interactions.

One plausible rationale for the inability to translate our in vitro findings into in vivo

models could be explained by compensatory changes following the genetic deletion of a

specific transporter in mice. In regards to our ENT1 model, there are no significant

compensatory changes noted in gene expression of nucleoside transporter or metabolism

after genetic deletion of murine ENT1.98,99 Likewise, previous studies have indicated that

genetic deficiency of OATP1B2 is not associated with any pronounced compensatory

alterations in metabolic enzyme or transporter expression in the liver and kidney.100 Similar findings have been reported for MATE1-knockout mice,101 and OCTN1-knockout mice.102

Taken together, compensatory changes in our murine models cannot account for the

inability to detect changes in the systemic disposition in our putative Ara-C transporter-

deficient mice.

In conclusion, this study confirmed the existence of a possible NBMPR-mediated interaction with the nucleoside analog Ara-C in mice that appears to occur independently of two known NBMPR-sensitive Ara-C transporters (ENT1 and OCTN1) and is unlikely connected with two other transporters of suspected relevance (OATP1B2 and MATE1).

The discrepancy between previous in vitro observations and those observed here in mice support the possibility that one or more additional uptake transporters for Ara-C exist in mice that are highly sensitive to inhibition by NBMPR. Although the identity of the carrier- mediated mechanism(s) remains unconfirmed and requires further investigation, this

52 study provides direct in vivo evidence that ENT1 is a transporter of limited clinical relevance to the systemic pharmacokinetics of Ara-C.

3.5 Acknowledgements

We thank Dr. Raj Govindarajan (The Ohio State University, Columbus, OH) for providing

ENT1(-/-) mice, Dr. Yukio Kato (Kanazawa University, Kanazawa, Japan) for OCTN1(–/–)

mice, Drs. Richard B. Kim (Western University, London, Ontario, Canada) and Jeffrey L.

Stock (Pfizer, Groton, CT) for OATP2(-/-) mice, Dr. Yan Shu (University of Maryland,

Baltimore, MD) for MATE1(-/-) mice.

I would like to thank Dr. Qiang Fu for his contribution for Ara-C quantification via mass

spectrometry, Dr. Shuiying Hu for assistance in the colony formation assay and

pharmacokinetic studies, and the many graduate students that have assisted with the

pharmacokinetic sample collection (Mingqing Chen, Dr. Eric Eisenmann, Dominique

Garrison, Kevin Huang, and Muhammad Erfan Uddin).

I would like to thank the Sparreboom, Baker and Hu lab for producing and providing

reagents for this study. This project was guided and conceived by Drs. Alex Sparreboom and Sharyn Baker.

3.6 Figures

Figure 15: RT-PCR of various mouse tissues from wild-type and OCTN1(-/-) mice. mRNA expression, normalized to GAPDH, from multiple tissues from OCTN1-deficient or age- matched wild-type mice (N=3-6 of each gender per tissue type). Data shown as mean normalized expression values ± SEM.

Figure 16: Colony Formation Assay using lineage negative cells extracted from wild-type and OCTN1(-/-) mice. Lineage negative bone marrow cells, from OCTN1-deficent mice or age matched wild-type, plated with varying concentration of Ara-C. Colonies were counted on day 7 after plating. P-values were calculated using an unpaired t test.

Figure 17: Influence of NBMPR and ENT1-deficiency on the pharmacokinetics of Ara-C. (A) Plasma, (B) whole blood, and (C) erythrocyte concentration-time profiles of Ara-C in wild-type mice receiving vehicle (PO) (closed circles; n=4) or 100 mg/kg NBMPR (PO) (open circles; n=4) one hour prior to Ara-C dosing. (D) Plasma (n=3 each group), (E) whole blood (n=4 each group), and (F) erythrocyte concentration-time profiles of 15 mg/kg Ara- C (IP) in wild-type mice (closed circles; n=4) or ENT1(-/-) mice (open circles; n=3). Results are shown as mean values (symbols) and SEM (error bars).

Figure 18: Influence of OCTN1-, OATP1B2-, and MATE1-deficiency on the pharmacokinetic of Ara-C. (A) Plasma (n=8 for wild-type and n=7 for OCTN1(-/-)) and (B) whole blood (n=4 each group) concentration-time profiles of Ara-C in wild-type mice (closed circles) or OCTN1(-/-) mice (open circles) receiving Ara-C at a dose of 15 mg/kg (IP). (C) Plasma concentration-time profiles of Ara-C in wild-type mice (closed circles; n=4) or OATP1B2(-/-) mice (open circles; n=4) receiving Ara-C at a dose of 10 mg/kg (IV). (D) Plasma concentration-time profiles of Ara-C in wild-type mice (closed circles; n=4) or MATE1(-/-) mice (open circles; n=4) receiving Ara-C at a dose of 100 mg/kg (IP). Results are shown as mean values (symbols) and SEM (error bars).

Figure 19: Measurement of the unchanged parent compound, Ara-C, in HEK293 cells transfected with OCTN1 using LC-MS/MS. Chromatograms are shown for a sample of Ara-C in uptake media containing the drug at a concentration of 1 µM (black circles) or lysates of HEK293 cells overexpressing OCTN1 (red circles) following a 30-min incubation of the cells with Ara-C (extracellular concentration, 1 µM) at 37°C.

4 Chapter 4: Future Directions

4.1 Metabolomics

The physiologic role and classification of the endogenous substrates of OCTN1 have remained elusive. Initial metabolomic studies identified ergothioneine as a substrate specific for OCTN1.103 Additional studies looked at ergothioneine disposition in organs by

comparing the accumulation of ergothioneine in OCTN1(-/-) and wild-type mice following

oral administration of radiolabeled [3H]ergothioneine.15 Another more recent study

examined the unclassified substrates of OCTN1 using a targeted combination

metabolomics approach and a transfected overexpression model incubated with colon extracts from colitis-induced mice, mimicking ulcerative colitis.104 By selecting for

compound similar in structure to previously reported OCTN1 substrate TEA, the authors

concluded that spermine, an amino acid precursor, was the OCTN1-mediated substrate

associated with gastrointestinal inflammation. Unfortunately, the authors did not comment

on the role of OCTN1 outside of their indicated disease state. Based upon these previous

studies, there is a gap in the literature for a comprehensive, untargeted metabolic

approach to classify the endogenous substrates of OCTN1.

Using an untargeted approach, we harvested multiple tissue types, including

plasma and kidney, of both male and female OCTN1(-/-) mice to compare the endogenous

metabolomic profile to wild-type mice. The tissues were harvested, and flash frozen on

the same day to minimize external variables such as housing, diet, and age of mice. Using

this untargeted, mass spec-based approach, we were able to see distinct separation in

the datasets comparing OCTN1(-/-) and wild-type samples using principal component

analysis (PCA) plots (Figure 20). In addition, we are able to take these hits and focus on

59 the hits represent the most likely endogenous metabolites of OCTN1 (-logP>3 and

log2(fold change) >2.5) (Figure 21). The features matching these criteria are highlighted

in purple. Since the hits are metabolic features (m/z ratios and retention times), there are multiple possible metabolite matches per hit; therefore, the hits are unable to be properly annotated until confirmatory studies identify individual metabolites.

Furthermore, we can import these metabolomic hits into MetaboAnalyst to derive a pathway analysis of the compared metabolites from our OCTN1(-/-) or age-matched wild- type samples (Figure 22). We see there are many different metabolic pathways that are dysregulated upon genetic removal of OCTN1 in our murine model. One interesting disrupted pathway is the amino sugar and nucleotide sugar pathway (highlight in red in

Table 4). Disruptions in this nucleotide pathway would be in alignment with our initial hypothesis that OCTN1 is able to transport nucleosides and other analogs similar in structure, which we supported using in vitro data. Future studies would confirm the individual metabolites in this nucleotide and amino sugar pathway using a targeted and quantitative LC-MS/MS approach.

4.2 Figures

Figure 20: Principal component analysis plot for metabolomics data generated from OCTN1(-/-) or wild-type female kidney samples. The relative abundance of metabolomic features (m/z ratio and retention time) were compared between kidney samples harvested from transporter-deficient or age-matched wild-type female mice (n=3 mice per group, ran in triplicate).

Figure 21: Metabolomics data shown as a volcano plot displaying hits. Significant features are highlighted in purple and are filtered by >2.5-fold change and –LogP > 3.0.

Figure 22: Summary of metabolic pathway analysis of female kidney samples of OCTN1(- /-) compared to wild-type samples. Pathway analysis was generated using MetaboAnalyst 4.0 database.105 Enrichment Factor is calculated from the ratio of the number of significant pathway hits (P < 0.01) to the expected number of hits within that pathway.

Table 4: Pathway analysis of metabolomics features generated from female kidney samples from OCTN1(-/-) compared to wild-type samples.

Rank Pathway Number of Metabolites Hits Hits in Pathway (total) (p<0.01) 1 Aminoacyl-tRNA biosynthesis 69 22 11 2 Arginine and proline 44 25 11 metabolism 3 Lysine degradation 23 11 6 4 Glutathione metabolism 26 12 6 5 Tryptophan metabolism 40 14 6 6 Tyrosine metabolism 44 16 6 7 Valine, leucine and isoleucine 38 9 5 degradation 8 Butanoate metabolism 22 11 5 9 Glycine, serine and threonine 31 20 5 metabolism 10 Cysteine and methionine 27 15 4 metabolism 11 Alanine, aspartate and 24 17 4 glutamate metabolism 12 Amino sugar and nucleotide 37 18 4 sugar metabolism 13 Purine metabolism 68 24 4 14 Synthesis and degradation of 5 3 3 ketone bodies 15 Lysine biosynthesis 4 4 3

4.3 Lipidomics

During our investigation of which SLCs contributed to patient survival in AML, we noticed that there was a strong association of OCTN1 expression with various lipid metabolism pathways. Fatty acid metabolism has been shown to play a critical role in

AML106,107 and response to cytarabine,108 but very little is known about OCTN1’s role in

this process. It has been shown, in a concentration manner, that compounds involved in the lipid metabolism pathway, acetylcarnitine, and gamma-butyrobetaine, were able to inhibit the uptake of [14C]TEA when coincubation occurred during uptake in proteoliposomes, which points to similar uptake mechanism involving an organic cation.109

In order to address this gap in the literature, we harvested bone marrow from

OCTN1(-/-) and wild-type mice and then purified the bulk cells for the lineage negative

(stem-cell-like) population using a MACs purification described earlier. These cells were

submitted for lipidomic analysis on the Lipidyzer (Sciex) platform, which allows for quantification of over 13 classes of lipids, comprising of >1,100 lipid species. Previous studies have used this Lipidyzer platform for determining tissue-specific lipid profiles using mouse heart and liver samples.110

Examining the PCA plot of the different lipid species within our lineage negative bone

marrow, we do not see a distinct separation of wild-type and knockout samples as noted

by overlapping 95% confidence interval ellipses (Figure 23). This observation can stem

from two different issues: there are no detectable differences between the two data sets

or there are differences but unable to detect due to the low signal to noise ratio. In our case, using a specific subpopulation of bone marrow (<5% of total bone marrow cells) limited our detection because of a decreased signal to noise ratio. This notion is supported

65 by the lack of hits reaching the significant threshold (Figure 24). Furthermore, we see a bias in the ability to only detect lipid species with increased fold change compared to a decreased fold change, which would be indiscernible from baseline noise. Future studies would examine the lipidomics that would allow more sample input such as whole bone marrow or kidney.

4.4 Figures

Figure 23: PCA analysis of lipidomics hits. Concentrations of lipid species were compared between lineage negative bone marrow cells harvested from transporter-deficient or age- matched wild-type female mice. Text labels correlate with animal IDs. (n=4-5 mice per group signified with either red (OCTN1 KO) or blue (wild-type) coloring).

Figure 24: Volcano plot of top hits from lipidomics of bone marrow. Features are highlighted in purple are filtered by >2.0-fold change. Dotted lines show cut off values for ± 2 fold change whereas solid black line demarks –LogP value of 3.

4.5 Involvement of OCTN1 in Other Disease States

Since we were able to associate the expression of OCTN1 and survival in both adult and pediatric AML patients, we wanted to explore other disease states that may be impacted by the expression of OCTN1. Previous evidence has shown that uptake of nucleoside analog, gemcitabine, is sensitive to nanomolar concentrations of classical nucleoside inhibitor, NBMPR.111 In addition, we were able to show that gemcitabine is a substrate of hOCTN1 (>50-fold over vector control) in our HEK293 overexpression cell model while hENT1 only showed slight accumulation (1.5-fold) over vector control cells.24

We wanted to translate these in vitro findings that gemcitabine is a substrate of

OCTN1 by exploring cancer subtypes that utilize gemcitabine as first-line therapy. In pancreatic cancer, gemcitabine is utilized as a single agent or in combination with

capecitabine, a prodrug of 5-fluorouracil, as a first-line therapy. In agreement with our

previously proposed hypothesis, we postulated that higher expression of OCTN1 would

lead to higher intracellular accumulation of gemcitabine and subsequently lead to higher

cytotoxicity and better clinical outcomes. Based on mRNA expression data from the TCGA

database, we were able to show that higher expression of OCTN1 is significantly

correlated with survival in pancreatic patients (HR 0.42; P = 0.04), whereas, expression of

ENT1 was not significantly correlated (P = 0.27) (Figure 25).

4.6 Figures

Figure 25: Kaplan-Meier survival curves showing high and low expression of ENT1 and OCTN1 in pancreatic adenocarcinoma (PAAD) patients from the TGCA database. mRNA expression of pancreatic samples grouped based on “high” or “low” (A) ENT1 or (B) OCTN1 expression (25% quartiles; n=176 from the PAAD dataset). P-values calculated using a log-rank test while hazard ratio was calculated using a cox proportional model.

4.7 Acknowledgements

A part of the metabolomics work was supported by OSU’s Center for Clinical and

Translational Science (CCTS) core voucher.

5 Bibliography

1 Davis, A. S., Viera, A. J. & Mead, M. D. Leukemia: An overview for primary care. Am. Fam. Physician 89, 731-738, (2014).

2 Siegel, R. L., Miller, K. D. & Jemal, A. Cancer statistics, 2016. CA Cancer J. Clin. 66, 7-30, (2016).

3 Ferrara, F. & Schiffer, C. A. Acute myeloid leukaemia in adults. Lancet 381, 484- 495, (2013).

4 Estey, E. Acute myeloid leukemia: 2016 Update on risk‐stratification and management. Am. J. Hematol. 91, 824-846, (2016).

5 Obata, T., Endo, Y., Murata, D., Sakamoto, K. & Sasaki, T. The molecular targets of antitumor 2'-deoxycytidine analogues. Curr. Drug Targets 4, 305-313, (2003).

6 Mangravite, L. M., Lipschutz, J. H., Mostov, K. E. & Giacomini, K. M. Localization of GFP-tagged concentrative nucleoside transporters in a renal polarized epithelial cell line. Am. J. Physiol. Renal Physiol. 280, F879-885, (2001).

7 Mangravite, L. M., Xiao, G. & Giacomini, K. M. Localization of human equilibrative nucleoside transporters, hENT1 and hENT2, in renal epithelial cells. Am. J. Physiol. Renal Physiol. 284, F902-910, (2003).

8 Kong, W., Engel, K. & Wang, J. (Section A: Molecular, Structural, and Cellular Biology of Drug Transporters) Mammalian Nucleoside Transporters. Curr Drug Metab 5, 63-84, (2004).

9 Drenberg, C. D. et al. OCTN1 Is a High-Affinity Carrier of Nucleoside Analogues. Cancer Res. 77, 2102-2111, (2017).

10 Tamai, I. et al. Cloning and characterization of a novel human pH-dependent organic cation transporter, OCTN1. FEBS Lett. 419, 107-111, (1997).

11 Tamai, I. et al. Involvement of OCTN1 (SLC22A4) in pH-dependent transport of organic cations. Mol. Pharm. 1, 57-66, (2004).

12 Sugiura, T. et al. Functional expression of carnitine/organic cation transporter OCTN1/SLC22A4 in mouse small intestine and liver. Drug Metab. Dispos. 38, 1665-1672, (2010).

13 Tamai, I. et al. Involvement of OCTN1 (SLC22A4) in pH-Dependent Transport of Organic Cations. Mol. Pharm. 1, 57-66, (2004).

14 Gründemann, D. et al. Discovery of the ergothioneine transporter. Proc. Natl. Acad. Sci. U. S. A. 102, 5256-5261, (2005).

15 Kato, Y. et al. Gene knockout and metabolome analysis of carnitine/organic cation transporter OCTN1. Pharm. Res. 27, 832-840, (2010).

16 Mitsuyama, H. & May, J. M. Uptake and antioxidant effects of ergothioneine in human erythrocytes. Clin. Sci. (Lond.) 97, 407-411, (1999).

17 Hanlon, D. P. Interaction of ergothioneine with metal ions and metalloenzymes. J. Med. Chem. 14, 1084-1087, (1971).

18 Laurenza, I., Colognato, R., Migliore, L., Del Prato, S. & Benzi, L. Modulation of palmitic acid-induced cell death by ergothioneine: evidence of an anti- inflammatory action. Biofactors 33, 237-247, (2008).

19 Lin, Z. et al. OCTN1 variant L503F is associated with familial and sporadic inflammatory bowel disease. J Crohns Colitis 4, 132-138, (2010).

20 Lee, Y. H. et al. Meta-analysis of SLC22A4 and RUNX1 polymorphisms : Associations with rheumatoid arthritis susceptibility. Z. Rheumatol. 74, 351-358, (2015).

21 Urban, T. J. et al. Effects of genetic variation in the novel organic cation transporter, OCTN1, on the renal clearance of gabapentin. Clin. Pharmacol. Ther. 83, 416-421, (2008).

22 Pochini, L., Scalise, M., Galluccio, M. & Indiveri, C. Regulation by physiological cations of acetylcholine transport mediated by human OCTN1 (SLC22A4). Implications in the non-neuronal cholinergic system. Life Sci. 91, 1013-1016, (2012).

23 Food, U. & Administration, D. In vitro metabolism and transporter-mediated drug- drug interaction studies. Guidance for Industry, (2017).

24 Drenberg, C. D. et al. OCTN1 Is a High-Affinity Carrier of Nucleoside Analogues. Cancer Res. 77, 2102-2111, (2017).

25 Jamieson, G. P., Snook, M. B. & Wiley, J. S. Saturation of intracellular cytosine arabinoside triphosphate accumulation in human leukemic blast cells. Leuk. Res. 14, 475-479, (1990).

26 Wada, E. et al. Modulation of peroxisome proliferator-activated receptor-alpha activity by bile acids causes circadian changes in the intestinal expression of Octn1/Slc22a4 in mice. Mol. Pharmacol. 87, 314-322, (2015).

27 Meetam, P., Srimaroeng, C., Soodvilai, S. & Chatsudthipong, V. Regulatory role of testosterone in organic cation transport: in vivo and in vitro studies. Biol. Pharm. Bull. 32, 982-987, (2009).

28 Maeda, T., Hirayama, M., Kobayashi, D., Miyazawa, K. & Tamai, I. Mechanism of the regulation of organic cation/carnitine transporter 1 (SLC22A4) by rheumatoid 71

arthritis-associated transcriptional factor RUNX1 and inflammatory cytokines. Drug Metab. Dispos. 35, 394-401, (2007).

29 Weinhold, B. Epigenetics: the science of change. Environ. Health Perspect. 114, A160-167, (2006).

30 Biswas, S. & Rao, C. M. Epigenetic tools (The Writers, The Readers and The Erasers) and their implications in cancer therapy. Eur. J. Pharmacol. 837, 8-24, (2018).

31 Derissen, E. J., Beijnen, J. H. & Schellens, J. H. Concise drug review: azacitidine and decitabine. Oncologist 18, 619-624, (2013).

32 Gore, L. et al. A multicenter, randomized study of decitabine as epigenetic priming with induction chemotherapy in children with AML. Clin. Epigenetics 9, 108, (2017).

33 Ye, X. N. et al. Epigenetic priming with decitabine followed by low-dose idarubicin/cytarabine has an increased anti-leukemic effect compared to traditional chemotherapy in high-risk myeloid neoplasms. Leuk. Lymphoma 57, 1311-1318, (2016).

34 Scandura, J. M. et al. Phase 1 study of epigenetic priming with decitabine prior to standard induction chemotherapy for patients with AML. Blood 118, 1472-1480, (2011).

35 Schneider, B. J. et al. Phase I Study of Epigenetic Priming with Azacitidine Prior to Standard Neoadjuvant Chemotherapy for Patients with Resectable Gastric and Esophageal Adenocarcinoma: Evidence of Tumor Hypomethylation as an Indicator of Major Histopathologic Response. Clin. Cancer Res. 23, 2673-2680, (2017).

36 Lee, V. et al. A Phase I Trial of a Guadecitabine (SGI-110) and Irinotecan in Metastatic Colorectal Cancer Patients Previously Exposed to Irinotecan. Clin. Cancer Res. 24, 6160-6167, (2018).

37 Jain, S. et al. Decitabine Priming Enhances Mucin 1 Inhibition Mediated Disruption of Redox Homeostasis in Cutaneous T-Cell Lymphoma. Mol. Cancer Ther. 16, 2304-2314, (2017).

38 Ellison, R. R. et al. Arabinosyl cytosine: a useful agent in the treatment of acute leukemia in adults. Blood 32, 507-523, (1968).

39 Wiley, J. S., Jones, S. P., Sawyer, W. H. & Paterson, A. R. Cytosine arabinoside influx and nucleoside transport sites in acute leukemia. J. Clin. Invest. 69, 479- 489, (1982).

40 White, J. C., Rathmell, J. P. & Capizzi, R. L. Membrane transport influences the rate of accumulation of cytosine arabinoside in human leukemia cells. J. Clin. Invest. 79, 380-387, (1987).

41 Ward, J. L., Sherali, A., Mo, Z. P. & Tse, C. M. 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. J. Biol. Chem. 275, 8375-8381, (2000).

42 Clarke, M. L. et al. The role of human nucleoside transporters in cellular uptake of 4'-thio-beta-D-arabinofuranosylcytosine and beta-D-arabinosylcytosine. Mol. Pharmacol. 70, 303-310, (2006).

43 Zimmerman, E. I. et al. Identification of a novel point mutation in ENT1 that confers resistance to Ara-C in human T cell leukemia CCRF-CEM cells. FEBS Lett. 583, 425-429, (2009).

44 Mayer, L. D. et al. Ratiometric dosing of anticancer drug combinations: Controlling drug ratios after systemic administration regulates therapeutic activity in tumor-bearing mice. 5, 1854-1863, (2006).

45 Lancet, J. E. et al. Phase 2 trial of CPX-351, a fixed 5:1 molar ratio of cytarabine/daunorubicin, vs cytarabine/daunorubicin in older adults with untreated AML. Blood 123, 3239-3246, (2014).

46 Pavillard, V., Kherfellah, D., Richard, S., Robert, J. & Montaudon, D. Effects of the combination of camptothecin and doxorubicin or etoposide on rat glioma cells and camptothecin-resistant variants. Br. J. Cancer 85, 1077-1083, (2001).

47 Kobayashi, D. et al. Expression of organic cation transporter OCTN1 in hematopoietic cells during erythroid differentiation. Exp. Hematol. 32, 1156-1162, (2004).

48 Gutierrez, S. E. & Romero-Oliva, F. A. Epigenetic changes: a common theme in acute myelogenous leukemogenesis. J. Hematol. Oncol. 6, 57, (2013).

49 Network, C. G. A. R. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med. 368, 2059-2074, (2013).

50 Bullinger, L. et al. Quantitative DNA methylation predicts survival in adult acute myeloid leukemia. Blood 115, 636-642, (2010).

51 Ochs, M. et al. Genome Wide Promoter Methylation Patterns Predict AML Subtype Outcomes and Identify Novel Pathways Characterizing Diagnostic and Relapsed Disease in Children. Blood 120, (2012).

52 Schaeffeler, E. et al. DNA methylation is associated with downregulation of the organic cation transporter OCT1 (SLC22A1) in human hepatocellular carcinoma. Genome Med. 3, 82, (2011).

53 Liu, Y. et al. Epigenetic activation of the drug transporter OCT2 sensitizes renal cell carcinoma to oxaliplatin. Sci. Transl. Med. 8, 348ra397, (2016).

54 Chen, L. et al. Genetic and epigenetic regulation of the organic cation transporter 3, SLC22A3. Pharmacogenomics J. 13, 110-120, (2013).

55 Tanaka, T., Hirota, T. & Ieiri, I. Relationship between DNA Methylation in the 5' CpG Island of the SLC47A1 (Multidrug and Toxin Extrusion Protein, MATE1) Gene and Interindividual Variability in MATE1 Expression in the Human Liver. Mol. Pharmacol. 93, 1-7, (2018).

56 Qu, Q. et al. Different involvement of promoter methylation in the expression of organic cation/carnitine transporter 2 (OCTN2) in cancer cell lines. PLoS One 8, e76474, (2013).

57 Stresemann, C. & Lyko, F. Modes of action of the DNA methyltransferase inhibitors azacytidine and decitabine. Int. J. Cancer 123, 8-13, (2008).

58 Klug, M. & Rehli, M. Functional analysis of promoter CPG-methylation using a CpG-free luciferase reporter vector. Epigenetics 1, 127-130, (2006).

59 Buelow, D. R. & Baker, S. D. Hypoxia Reporter Element Assay. Bio Protoc 8, e2951, (2018).

60 Early, A. P., Preisler, H. D., Slocum, H. & Rustum, Y. M. A pilot study of high- dose 1-beta-D-arabinofuranosylcytosine for acute leukemia and refractory lymphoma: clinical response and pharmacology. Cancer Res. 42, 1587-1594, (1982).

61 Mayer, L. D. et al. Ratiometric dosing of anticancer drug combinations: controlling drug ratios after systemic administration regulates therapeutic activity in tumor-bearing mice. Mol. Cancer Ther. 5, 1854-1863, (2006).

62 Scandura, J. M. et al. Phase 1 study of epigenetic priming with decitabine prior to standard induction chemotherapy for patients with AML. Blood 118, 1472-1480, (2011).

63 Hollenbach, P. W. et al. A comparison of azacitidine and decitabine activities in acute myeloid leukemia cell lines. PLoS One 5, e9001, (2010).

64 Schulz, J. B. et al. Cooperative interception of neuronal apoptosis by BCL-2 and BAG-1 expression: prevention of caspase activation and reduced production of reactive oxygen species. J. Neurochem. 69, 2075-2086, (1997).

65 Yu, Q. et al. Histone H3 Lysine 4 Trimethylation, Lysine 27 Trimethylation, and Lysine 27 Acetylation Contribute to the Transcriptional Repression of Solute Carrier Family 47 Member 2 in Renal Cell Carcinoma. Drug Metab. Dispos. 45, 109-117, (2017).

66 Anzai, H., Frost, P. & Abbruzzese, J. L. Synergistic cytotoxicity with 2'-deoxy-5- azacytidine and topotecan in vitro and in vivo. Cancer Res. 52, 2180-2185, (1992).

67 Eramo, A. et al. Inhibition of DNA methylation sensitizes glioblastoma for tumor necrosis factor-related apoptosis-inducing ligand-mediated destruction. Cancer Res. 65, 11469-11477, (2005).

68 Kanda, T. et al. 5-aza-2'-deoxycytidine sensitizes hepatoma and pancreatic cancer cell lines. Oncol. Rep. 14, 975-979, (2005).

69 Harikrishnan, K. N. et al. Alleviating transcriptional inhibition of the norepinephrine SLC6A2 transporter gene in depolarized neurons. J. Neurosci. 30, 1494-1501, (2010).

70 Qiu, X. et al. Equitoxic doses of 5-azacytidine and 5-aza-2'deoxycytidine induce diverse immediate and overlapping heritable changes in the transcriptome. PLoS One 5, (2010).

71 Modhukur, V. et al. MethSurv: a web tool to perform multivariable survival analysis using DNA methylation data. Epigenomics 10, 277-288, (2018).

72 Gati, W. P., Paterson, A. R. P., Larratt, L. M., Turner, A. R. & Belch, A. R. Sensitivity of acute leukemia cells to cytarabine is a correlate of cellular es nucleoside transporter site content measured by flow cytometry with SAENTA- fluorescein. Blood 90, 346-353, (1997).

73 Bhalla, K., Nayak, R. & Grant, S. Isolation and characterization of a deoxycytidine kinase-deficient human promyelocytic leukemic cell line highly resistant to 1-beta-D- arabinofuranosylcytosine. Cancer Res. 44, 5029-5037, (1984).

74 Schroder, J. K., Seidelmann, M., Kirch, H. C., Seeber, S. & Schutte, J. Assessment of resistance induction to cytosine arabinoside following transfer and overexpression of the deoxycytidylate deaminase gene in vitro. Leuk. Res. 22, 619-624, (1998).

75 Adjei, A. A., Dagnino, L., Wong, M. M. & Paterson, A. R. Protection against fludarabine neurotoxicity in leukemic mice by the nucleoside transport inhibitor nitrobenzylthioinosine. Cancer Chemother. Pharmacol. 31, 71-75, (1992).

76 Kolassa, N., Jakobs, E. S., Buzzell, G. R. & Paterson, A. R. Manipulation of toxicity and tissue distribution of tubercidin in mice by nitrobenzylthioinosine 5'- monophosphate. Biochem. Pharmacol. 31, 1863-1874, (1982). 75

77 Cass, C. E., Muzik, H. & Paterson, A. R. Combination Therapy of Mouse Leukemia L1210 by 1-β-d-Arabinofuranosylcytosine and 6-[(4-Nitrobenzyl) thio]- 9-β-d-ribofuranosylpurine. Cancer Res. 35, 1187-1193, (1975).

78 Leblanc, A. F. et al. Murine Pharmacokinetic Studies. Bio-protocol 8, e3056, (2018).

79 Henningsson, A. et al. Mechanism-based pharmacokinetic model for paclitaxel. J. Clin. Oncol. 19, 4065-4073, (2001).

80 Amend, S. R., Valkenburg, K. C. & Pienta, K. J. Murine Hind Limb Long Bone Dissection and Bone Marrow Isolation. J Vis Exp, (2016).

81 Kobayashi, D. et al. Expression of organic cation transporter OCTN1 in hematopoietic cells during erythroid differentiation. Exp. Hematol. 32, 1156-1162, (2004).

82 Rostami-Hodjegan, A., Lennard, L. & Lilleyman, J. S. The accumulation of mercaptopurine metabolites in age fractionated red blood cells. Br. J. Clin. Pharmacol. 40, 217-222, (1995).

83 Endres, C. J. et al. The role of the equilibrative nucleoside transporter 1 (ENT1) in transport and metabolism of ribavirin by human and wild-type or Ent1-/- mouse erythrocytes. J. Pharmacol. Exp. Ther. 329, 387-398, (2009).

84 Chabot, G. G., Bouchard, J. & Momparler, R. L. Kinetics of deamination of 5-aza- 2'-deoxycytidine and cytosine arabinoside by human liver cytidine deaminase and its inhibition by 3-deazauridine, thymidine or uracil arabinoside. Biochem. Pharmacol. 32, 1327-1328, (1983).

85 Gati, W. P. & Paterson, A. R. Measurement of nitrobenzylthioinosine in plasma and erythrocytes: a pharmacokinetic study in mice. Cancer Chemother. Pharmacol. 40, 342-346, (1997).

86 Suwalsky, M., Hernández, P. L., Villena, F. & Sotomayor, C. P. The anticancer drug cytarabine does not interact with the human erythrocyte membrane. Z Naturforsch C. 58, 885-890, (2003).

87 Chen, R., Pan, B. F., Sakurai, M. & Nelson, J. A. A nucleoside-sensitive organic cation transporter in opossum kidney cells. Am. J. Physiol. 276, F323-328, (1999).

88 Zimmerman, E. I. et al. Contribution of OATP1B1 and OATP1B3 to the disposition of sorafenib and sorafenib-glucuronide. Clin. Cancer Res. 19, 1458- 1466, (2013).

89 Hagenbuch, B., Adler, I. D. & Schmid, T. E. Molecular cloning and functional characterization of the mouse organic-anion-transporting polypeptide 1 (Oatp1) and mapping of the gene to chromosome X. Biochem. J. 345, 115-120, (2000). 76

90 Tschirka, J., Kreisor, M., Betz, J. & Grundemann, D. Substrate Selectivity Check of the Ergothioneine Transporter. Drug Metab. Dispos. 46, 779-785, (2018).

91 Owens, J. K., Shewach, D. S., Ullman, B. & Mitchell, B. S. Resistance to 1-β-d- arabinofuranosylcytosine in human T-lymphoblasts mediated by mutations within the deoxycytidine kinase gene. Cancer Res. 52, 2389-2393, (1992).

92 Drenberg, C. D. et al. Inherited variation in OATP1B1 is associated with treatment outcome in acute myeloid leukemia. Clin. Pharmacol. Ther. 99, 651- 660, (2016).

93 Ajavon, A. D., Bonate, P. L. & Taft, D. R. Renal excretion of clofarabine: Assessment of dose-linearity and role of renal transport systems on drug excretion. Eur. J. Pharm. Sci. 40, 209-216, (2010).

94 Molina-Arcas, M. et al. Equilibrative nucleoside transporter-2 (hENT2) protein expression correlates with ex vivo sensitivity to fludarabine in chronic lymphocytic leukemia (CLL) cells. Leukemia 19, 64-68, (2005).

95 Chen, R. & Nelson, J. A. Role of organic cation transporters in the renal secretion of nucleosides. Biochem. Pharmacol. 60, 215-219, (2000).

96 Meyer Zu Schwabedissen, H. E., Ware, J. A., Tirona, R. G. & Kim, R. B. Identification, expression, and functional characterization of full-length and splice variants of murine organic anion transporting polypeptide 1b2. Mol. Pharm. 6, 1790-1797, (2009).

97 Otsuka, M. et al. A human transporter protein that mediates the final excretion step for toxic organic cations. Proc. Natl. Acad. Sci. U. S. A. 102, 17923-17928, (2005).

98 Graham, K. et al. Nucleoside transporter gene expression in wild-type and mENT1 knockout mice. Biochem. Cell Biol. 89, 236-245, (2011).

99 Rose, J. B. et al. Equilibrative nucleoside transporter 1 plays an essential role in cardioprotection. Am. J. Physiol. Heart Circ. Physiol. 298, H771-777, (2010).

100 Lancaster, C. S. et al. OATP1B1 polymorphism as a determinant of erythromycin disposition. Clin. Pharmacol. Ther. 92, 642-650, (2012).

101 Tsuda, M. et al. Targeted disruption of the multidrug and toxin extrusion 1 (MATE1) gene in mice reduces renal secretion of metformin. Mol. Pharmacol. 75, 1280-1286, (2009).

102 Nakamichi, N. et al. Involvement of carnitine/organic cation transporter OCTN1/SLC22A4 in gastrointestinal absorption of metformin. J. Pharm. Sci. 102, 3407-3417, (2013).

103 Grundemann, D. et al. Discovery of the ergothioneine transporter. Proc. Natl. Acad. Sci. U. S. A. 102, 5256-5261, (2005).

104 Masuo, Y. et al. Combination Metabolomics Approach for Identifying Endogenous Substrates of Carnitine/Organic Cation Transporter OCTN1. Pharm. Res. 35, 224, (2018).

105 Chong, J., Wishart, D. S. & Xia, J. Using MetaboAnalyst 4.0 for Comprehensive and Integrative Metabolomics Data Analysis. Curr Protoc Bioinformatics 68, e86, (2019).

106 Kreitz, J. et al. Metabolic Plasticity of Acute Myeloid Leukemia. Cells 8, (2019).

107 Loew, A., Kohnke, T., Rehbeil, E., Pietzner, A. & Weylandt, K. H. A Role for Lipid Mediators in Acute Myeloid Leukemia. Int. J. Mol. Sci. 20, (2019).

108 Tabe, Y. et al. Inhibition of FAO in AML co-cultured with BM adipocytes: mechanisms of survival and chemosensitization to cytarabine. Sci. Rep. 8, 16837, (2018).

109 Pochini, L., Scalise, M., Galluccio, M., Amelio, L. & Indiveri, C. Reconstitution in liposomes of the functionally active human OCTN1 (SLC22A4) transporter overexpressed in Escherichia coli. Biochem. J. 439, 227-233, (2011).

110 Cao, Z. et al. Evaluation of the Performance of Lipidyzer Platform and Its Application in the Lipidomics Analysis in Mouse Heart and Liver. J. Proteome Res., (2019).

111 Mackey, J. R. et al. Functional nucleoside transporters are required for gemcitabine influx and manifestation of toxicity in cancer cell lines. Cancer Res. 58, 4349-4357, (1998).

6 Appendix: Transcription Factor Binding Motifs in SLC22A4

Transcription Factor Binding Motifs within SLC22A4: RUNX1 SP1 E2F1

>SLC22A4_1kb plasmid AAGGGGCTCAAAATCAGAACCCAGCGTGACGACAGTGCCTCCATGACTGTGCTCC CGCAATGCGCAGGCTCCATGTCCTGGCGCCCGCCCGCCCATGGACCCGGCGGGG GCTTCCAGGCTGGGCTCAGCCATTACGCCGGCGTGCGGGGGAGGAAACTCGCCT CCCGGGCACTCGGTTGTCTCCTGCCCCCGCCCCTCCCTCCGATCCGGGCCCATCT CTGACGTAGTGTGACCTTGCTCATCCCTTCCAGGCTGTGGGCCTGTTTTCCCTGTG CAAGATGAGGGTCCTGGCTGTCCTGAGGACGCTGTCCGGGCGCCGCCAGGGGTG ACCGAATTCAGCTCTGCTAGGACTGTTGGGAAATGAGCTCCCTGTCGGCGTGTGC CAGCCGCCTGCGCGAGGCGCCACAGGAGAGGGCGCGCTCCTGTCGCTCTGCCGC CCCCAGAAGTTTCCCGGGAACCGACTCCACTGACTCGCCCCTCCGCGCCCCACGC GTGGCAGCCTAAGCTCAGCCTCCAGATTGGAGGAGACCGCGGAGGGAACCCTGCT GGGGTCTGGGCCCGGGGCCACGCGGCCCGAGCAGATCGAGGGCCGACCCCTCC GAGAACTCGCTCTCTGGCCTCGGCTCCTCCCTTGCGCCCGCCCTCCCACGTGGGG CCCAGGTCTGGGAATCAGCGCTCGGGGGTGGCTGGGGACAACCGAGAACGAGCT TCTTCCCCGGCACGCGGGCGGAATGGCTGAGCCCAGCCTGGAAGCCCCCGTCAG GTCCTTGGGGGCGGGCGGGCGCGCGAAGCACAGGGCGGAGACAGCCGGGAGCC CAGCCTCCCGGGCTGGGCCGCCCTCCCCTTCCCCGCGCCCGGCCGGGGATGGG GGTGTGGTCCCAAGTGTACAGTGGCATCAAGCTCAGCGCGAGCTCCCGGGAACGC TCCAACGCCTTCAGCCTGTTTCCCAGGAACGGTCCCCGGCTTCGCGCCCCAATTTC TAACAGCCTGCCTGTCCCCCGGGAACGTTCTAACATCCTTGGGGAGCGCCCCAGC TACAAGACACTGTCCTGAGAACGCTGTCATCACCCGTAGTTGCAAGTTTC