I V Characterization of Gene-Environment Interactions

Characterization of Gene-Environment Interactions That Govern Metabolic Adaptation

Sydney Morgan Sanderson

Department of Pharmacology and Cancer Biology Duke University

Date:______Approved:

______David MacAlpine, Supervisor

______Jason Locasale

______Christopher Newgard

______Matthew Hirschey

______Kris Wood

Dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Pharmacology and Cancer Biology in the Graduate School of Duke University

2019

ABSTRACT

Characterization of Gene-Environment Interactions That Govern Metabolic Adaptation

Sydney Morgan Sanderson

Department of Pharmacology and Cancer Biology Duke University

Date:______Approved:

______David MacAlpine, Supervisor

______Jason Locasale

______Christopher Newgard

______Matthew Hirschey

______Kris Wood

An abstract of a dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Pharmacology and Cancer Biology in the Graduate School of Duke University

2019

Abstract

Metabolism is known to be driven by intrinsic genetic programs as well as contextual factors within the environment. Individual genetic and environmental determinants of metabolic state have been extensively characterized, both within normal physiological processes as well as in the context of disease states such as cancer. However, it is becoming increasingly appreciated that the inevitable interaction between these differential sources of metabolic regulation can dramatically influence cellular phenotypes, a phenomenon commonly referred to as gene- environment interaction. These interactions can create substantial heterogeneity between individuals, particularly in the context of tumor metabolism which can ultimately impede the development and efficacy of many clinical therapies. Characterization of these relationships can therefore improve the predictive applicability of targeted therapeutic approaches, as well as contribute to the identification of novel treatment strategies that can circumvent the biological limitations imposed by gene-environment interactions. Using metabolomic, genetic, and pharmacological approaches, in this dissertation I examine the metabolic consequences of environmental alterations in defined genetic settings. I provide in-depth characterization of the relative predictability of cellular responsiveness to nutrient availability in the context of genetic deletion of the metabolic enzyme MTAP, results of which demonstrate potential implications in previously-identified metabolic vulnerabilities in MTAP-deleted cancers. I additionally examine how perturbation of energetic demand, via either pharmacological inhibition of the Na+/K+

ATPase or with the physiological stimulus of exercise, impacts metabolic processes in diverse biological contexts. This work collectively illustrates the exceptional heterogeneity in metabolic adaptation to environmental alterations, and provides support for the future development of

lifestyle modifications and repurposing of common pharmacological agents as therapeutic modalities in cancer treatment.

Contents

Abstract ...... iv

List of Tables ...... xii

List of Figures ...... xiii

Acknowledgements ...... xvi

1. Introduction ...... 1

1.1 Fundamental characteristics of cancer metabolism ...... 1

1.1.1 The Warburg Effect ...... 1

1.1.2 Alterations in amino acid metabolism ...... 2

1.2 Metabolic heterogeneity in human cancer patients ...... 4

1.3 Environmental determinants of metabolic programming in cancer ...... 6

1.4 Methionine metabolism in health and cancer ...... 8

1.4.1 Methionine cycle and related pathways ...... 8

1.4.2 Role of methionine in biological processes ...... 9

1.4.3 Health phenotypes associated with dietary methionine availability ...... 11

1.4.4 Dietary methionine in cancer ...... 12

1.4.5 MTAP deletions in cancer ...... 13

1.4.6 Other methionine-associated alterations in cancer ...... 14

1.5 Targeting metabolic vulnerabilities in cancer ...... 16

1.5.1 Antimetabolite chemotherapies ...... 16

1.5.2 Targeting methionine metabolism in cancer ...... 18

1.5.3 Repurposing of common metabolic agents ...... 22

1.6 Scope of dissertation ...... 26

2. Nutrient Availability Shapes Methionine Metabolism in p16/MTAP-deleted Cells ...... 29

2.1 Background and Context ...... 29

2.2 Results ...... 30

2.2.1 MTAP status has a defined metabolic signature ...... 30

2.2.2 Responsiveness to methionine availability is not predicted by MTAP status ...... 34

2.2.3 Responsiveness to alterations in other one-carbon nutrient availability is largely MTAP status independent ...... 37

2.2.4 Restoration of MTAP expression has heterogeneous effects on metabolism ...... 41

2.2.5 MTAP status remains nonpredictive of responsiveness to nutrient restriction in a panel of tissue-matched cell lines ...... 44

2.2.6 Defining the quantitative impact of MTAP deletion and environmental factors on metabolism ...... 50

2.3 Discussion ...... 53

2.4 Materials and Methods ...... 55

2.4.1 Cell culture and reagents ...... 55

2.4.2 Nutrient restriction experiments ...... 55

2.4.3 Cell viability assays ...... 56

2.4.4 Lentiviral transfection and transduction for ectopic MTAP expression ...... 56

2.4.5 Immunoblotting methods ...... 57

2.4.6 Quantitative PCR methods ...... 58

2.4.7 Metabolite extraction ...... 59

2.4.8 Liquid chromatography ...... 59

2.4.9 Mass spectrometry ...... 60

2.4.10 Peak extraction and metabolomics data analysis ...... 60

2.4.11 Statistical analysis ...... 61

vii

2.4.12 Network analysis ...... 61

3. Methionine restriction synergizes with 5-FU therapy by disrupting nucleotide metabolism and redox balance ...... 63

3.1 Background and Context ...... 63

3.2 Results ...... 64

3.2.1 MR exhibits anti-cancer properties and synergizes with 5-FU in vivo ...... 64

3.2.2 MR as a single agent and in combination with 5-FU exerts cytotoxic effects through disruption of redox and nucleotide metabolism ...... 66

3.3 Discussion ...... 69

3.4 Materials and Methods ...... 70

3.4.1 Animals, diets, and tissue collection ...... 70

3.4.2 PDX models of colorectal cancer ...... 71

3.4.3 Colorectal cancer cell lines ...... 71

3.4.4 Metabolite rescue experiments ...... 72

3.4.5 Metabolite profiling and isotope tracing ...... 72

3.4.6 Metabolite extraction ...... 73

3.4.7 Liquid chromatography ...... 73

3.4.8 Mass spectrometry ...... 74

3.4.9 Peak extraction and metabolomics data analysis ...... 74

4. Digoxin Disrupts Central Carbon Metabolism and Reprograms the Tumor Microenvironment ...... 76

4.1 Background and Context ...... 76

4.2 Results ...... 77

4.2.1 Digoxin disrupts central carbon metabolism and energy-related processes in a temporal- and dose-dependent manner ...... 77

4.2.2 Digoxin exerts its metabolic effects via on-target inhibition of the Na+/K+ ATPase ... 82

viii

4.2.3 Digoxin impacts energy metabolism in a tissue-specific and antineoplastic manner .. 84

4.2.4 Acute digoxin treatment induces a defined shift in the tumor microenvironmental landscape ...... 88

4.2.5 Digoxin is associated with transcriptional reprogramming of metabolic processes in tumor cells and immune infiltrates...... 91

4.3 Discussion ...... 94

4.4 Materials and Methods ...... 96

4.4.1 Cell culture and reagents ...... 96

4.4.2 Digoxin IC-50 measurements ...... 96

4.4.3 Nutrient rescue experiments ...... 97

4.4.4 Isotope tracing ...... 97

4.4.5 Lentiviral transfection and transduction ...... 98

4.4.6 Sarcoma allograft mouse studies ...... 99

4.4.7 Immunohistochemistry and microscopy ...... 100

4.4.8 Metabolite extraction ...... 100

4.4.9 Liquid chromatography ...... 100

4.4.10 Mass spectrometry ...... 101

4.4.11 Peak extraction and metabolomics data analysis ...... 101

4.4.12 Single-cell RNA sequencing ...... 102

4.4.13 Single-cell RNA sequencing data processing ...... 102

4.4.14 Classification of single cells into malignant and non-malignant cells ...... 103

4.4.15 Classification of cell populations within non-malignant cells ...... 104

4.4.16 Differential gene expression and pathway enrichment analysis ...... 105

4.4.17 Quantification and statistical analysis ...... 105

5. Exercise inhibits tumor growth and central carbon metabolism in patient-derived xenograft models of colorectal cancer ...... 106

5.1 Background and Context ...... 106

5.2 Results ...... 107

5.2.1 Characterization of six colorectal cancer PDX models ...... 107

5.2.2 Exercise induces alterations in central carbon metabolism in exercise-responsive tumors ...... 111

5.2.3 Exercise-responsive tumors demonstrate distinct metabolic profiles ...... 115

5.2.4 Exercise-responsive tumors exhibit differential metabolic responses to exercise compared to non-responsive tumors...... 116

5.3 Discussion ...... 118

5.4 Materials and Methods ...... 121

5.4.1 Collection of patient tumors ...... 121

5.4.2 Generation of patient-derived xenografts ...... 121

5.4.3 Physical activity studies ...... 122

5.4.4 Metabolite extraction ...... 122

5.4.5 Liquid chromatography ...... 123

5.4.6 Mass spectrometry ...... 123

5.4.7 Peak extraction and metabolomics data analysis ...... 123

6. Conclusion ...... 125

6.1 Summary ...... 125

6.2 Place in Current Research ...... 127

6.3 Future Considerations ...... 128

6.4 Final Remarks ...... 130

References ...... 131

Biography ...... 157

List of Tables

Table 1: Top 50 differential metabolites between MTAP+/+ and MTAP-/- cells ...... 33

Table 2: Top 50 differential metabolites between MTAP+/+ and MTAP-/- glioma lines...... 46

Table 3: Demographics of patient-derived xenografts of CRC240, CRC282, CRC344, CRC361, CRC370, and BRPC12-146...... 108

xii

List of Figures

Figure 1: Metabolic heterogeneity between tumor subtypes in human cancer patients...... 6

Figure 2: The methionine cycle and related metabolic pathways...... 9

Figure 3: Role of methionine in biological processes...... 11

Figure 4: Mechanism of action of common antimetabolite chemotherapies...... 18

Figure 5: MTA is the most significantly altered metabolite in MTAP-deleted cells...... 31

Figure 6: MTAP status accounts for approximately 5% of metabolic variation...... 32

Figure 7: Methionine restriction impacts global metabolism and reduces MTA levels...... 35

Figure 8: Responsiveness to methionine availability is heterogeneous and is not predicted by MTAP status...... 37

Figure 9: Responsiveness to restriction of other one-carbon nutrients is largely MTAP status- independent...... 38

Figure 10: Serine or cysteine restriction has heterogeneous consequences on methionine metabolism and related metabolic pathways...... 40

Figure 11: Responsiveness to one-carbon nutrient availability is heterogeneous...... 41

Figure 12: Restoration of MTAP expression has heterogeneous effects on metabolism...... 42

Figure 13: Re-expression of MTAP has heterogeneous consequences on responsiveness to nutrient restriction...... 43

Figure 14: MTAP status remains nonpredictive of responsiveness to nutrient availability in a panel of tissue-matched cell lines...... 45

Figure 15: MTAP status remains non-predictive of responsiveness to nutrient availability in a tissue-matched panel of cell lines...... 48

Figure 16: Cell heterogeneity continues to dominate MTAP status in predictiveness of metabolic responsiveness to restriction of one-carbon nutrients...... 49

Figure 17: Integration of responsiveness to nutrient restriction determines quantitative impact of MTAP deletion on global metabolic networks...... 52

Figure 18: Graphical representation of relative impact of environmental and genetic factors on the metabolome...... 53

xiii

Figure 19: MR inhibits tumor growth in PDX models of colorectal cancer...... 65

Figure 20: MR synergizes with 5-FU in PDX models of colorectal cancer...... 66

Figure 21: Supplementation of nucleosides and antioxidants is partially cytoprotective against MR and 5-FU...... 67

Figure 22: Disruption to nucleotide synthesis is one mechanism by which MR and 5-FU synergize...... 68

Figure 23: Comparison of digoxin IC50 values and basal metabolic flux demonstrates association between digoxin and energy metabolism...... 78

Figure 24: Digoxin disrupts central carbon metabolism and energy-related processes in a temporal- and dose-dependent manner...... 79

Figure 25: Digoxin alters flux through central carbon metabolism...... 81

Figure 26: Nucleoside and antioxidant supplementation partially block digoxin-induced cytotoxicity...... 82

Figure 27: Digoxin exerts its metabolic effects via on-target inhibition of the Na+/K+ ATPase. 84

Figure 28: Digoxin impacts central carbon metabolism in primary murine sarcoma cells...... 85

Figure 29: Acute digoxin treatment impacts energy metabolism in both healthy and malignant tissue...... 86

Figure 30: Digoxin impacts intratumoral energy metabolism in an antineoplastic manner...... 88

Figure 31: Digoxin induces a shift in cell populations within the tumor microenvironment...... 90

Figure 32: Acute digoxin treatment is associated with transcriptional reprogramming of central carbon metabolism within tumor cells and immune infiltrates...... 92

Figure 33: Transcriptional profiles of macrophage populations suggest their activation is associated with digoxin treatment...... 93

Figure 34: Histological and quantitative analysis of exercise in six diverse CRC PDX models. 108

Figure 35: Exercise has differential effects on tumor growth across six CRC PDX models...... 110

Figure 36: The effect of exercise on tumor metabolism in six different CRC PDX models...... 111

Figure 37: Tumors from exercised mice exhibit global alterations in metabolism compared to tumors from sedentary mice...... 114

xiv

Figure 38: Exercise-responsive tumors exhibit distinct metabolic profiles...... 116

Figure 39: Exercise-responsive tumors exhibit differential metabolic responses to exercise compared to non-responsive tumors...... 118

Acknowledgements

First and foremost, I would like to express my gratitude to my PhD advisor, Dr. Jason

Locasale. When my previous PhD advisor left Duke following my preliminary exam, Dr.

Locasale welcomed me into his laboratory without reservation and continued to provide reassurance and guidance as I familiarized myself with the expansive field of metabolism.

Without his scientific mentorship and support, I would not have been able to accomplish the achievements that I have made throughout my graduate career. I would also like to thank my committee members Dr. David MacAlpine, Dr. Christopher Newgard, Dr. Matthew Hirschey, and

Dr. Kris Wood.; I am extremely grateful for their support during my doctoral training, and I will always be appreciative of their ability to make me feel like a fellow colleague despite their relatively intimidating scientific backgrounds.

My time spent during my graduate training would not have been nearly as fulfilling or entertaining if not for my colleagues in the Locasale laboratory (both past and present), most especially Maria Liberti, Peter Mikhael, Xiaojing Liu, and more recently Vijyendra Ramesh and

Shree Bose; I can only attribute it to pure serendipity that I have been fortunate enough to be surrounded by such inspiring, hilarious, and genuine people who also happen to be bonafide geniuses. It has been an absolute pleasure to work with such an incredible group of individuals, and I am positive I will forever be nostalgic for our time spent together commiserating over scientific or general philosophical conundrums.

I would like to further express my eternal appreciation for the love and support that my parents, Robert and Polly, and my brothers, Cameron and Chris, have provided me throughout my educational career. I would also like to acknowledge Laura, Brent, Tucker, and Cain Philbrook

(i.e. my “second family”); words cannot express how appreciative I am for their unwavering

xvi

support. I am additionally grateful for the invaluable friendships and support that I have received throughout my graduate training; in particular, I’d like to acknowledge Maria Liberti, who has throughout our friendship enthusiastically been my wingman for practically all of my scientific and personal endeavors, as well as Andrea Walens, whom I’m constantly in awe of and has always been my guide for effectively functioning as both a scientist and human being in general.

Finally, I would like to acknowledge my funding sources. I was supported by grants from the National Institute of Health (NIH) including the Pharmacological Science Training Program

(PSTP) grant (5T32GM007105) during my first two years of study, as well as the National

Cancer Institution (NCI) Predoctoral Fellowship grant (F31CA224973).

xvii

1. Introduction

1.1 Fundamental characteristics of cancer metabolism

Metabolic reprogramming is a hallmark feature of malignancy, whereby cancer cells commonly differ from their normal counterparts both in their nutritional dependencies and catabolic activities to support uncontrolled proliferation 1. This chapter subsection will provide an overview of characteristic alterations within central carbon metabolism (Section 1.1.1) and amino acid requirements (Section 1.1.2) that have been identified in cancer contexts.

1.1.1 The Warburg Effect

Perhaps the greatest advance in our understanding of cancer metabolism was the observation made by Otto Warburg and colleagues nearly a century ago that many tumors exhibit elevated rates of glucose uptake, even in the presence of sufficient oxygen, with a concomitant increase in lactate production 2. In conditions with sufficient oxygen, the majority of differentiated cells fully metabolize glucose through three highly regulated sequences of enzymatic processes: the cytosolic pathway termed glycolysis, followed by the mitochondrial processes referred to as the tricarboxylic acid (TCA) cycle and the electron transport chain (ETC)

3,4. One molecule of glucose can be completely oxidized to generate approximately 36 units of adenosine 5´-triphosphate (ATP), which serves as the major molecular currency of cellular energy, as well as other energy-rich molecules such as reduced nicotinamide adenine dinucleotide

(NADH) 3,4. However, under conditions of oxygen limitation such as in muscle tissue during strenuous activity, the end product of glycolysis (pyruvate) is diverted away from the mitochondria to produce the fermentation product lactate 4; this process, referred to as anaerobic glycolysis, generates only two net ATP molecules per molecule of glucose while consuming

NADH, but it can be beneficial to cells exhibiting high energetic demand due to relatively faster rates of glycolytic ATP production than through mitochondrial respiration 5,6.

Warburg hypothesized that the metabolic reprogramming of cancer cells towards enhanced rates of anaerobic glycolysis was primarily due to an acquired defect in mitochondrial function, and was an essential biological event for tumorigenesis 7. After many iterations over the years, this hypothesis (now referred to as the Warburg Effect, or aerobic glycolysis) has become conceptually defined as a switch from oxidative glucose metabolism, which is primarily comprised of mitochondrial respiration, to fermentation from glycolysis in which lactate rather than pyruvate is generated as the primary energy source 8. However, human patient tumors are now appreciated to exhibit remarkably variable degrees of reprogramming within central carbon metabolism (Figure 1), which is discussed in more detail in Section 1.2.

With Warburg’s discovery, intratumoral glucose utilization has become the most commonly used biomarker for tumor detection and monitoring. Positron emission topography

(PET) coupled with 2-deoxy-2-[fluorine-18]fluoro-D-glucose, or 18F-FDG-PET, is a technique that utilizes a radioactively-labeled glucose analog to measure glucose uptake, allowing for the anatomical localization of tumors when coupled with computed tomography (CT). The glucose analog 18F-FDG is phosphorylated by hexokinase similarly to glucose, but is not further metabolized, resulting in the accumulation of 18F-FDG in regions with high glucose uptake rates such as the liver, brain, and the majority of tumors 9. This technique thus allows for both the initial identification of tumors, as well as tumor monitoring for a large majority of tumor subtypes.

1.1.2 Alterations in amino acid metabolism

Cancer cells have also been found to extensively rely on amino acid catabolism to meet their increased metabolic demands. Genetic alterations that are typically thought to promote

tumorigenesis (commonly referred to as “oncogenes”) have been shown to impact amino acid uptake through cell membrane transport channels and catabolism 10-13. Additionally, cancer cells can also obtain amino acids via de novo synthesis, as well as through autophagic digestion of cellular proteins 14,15. This subsection provides an overview of the observed cancer cell dependencies on various amino acids, with the exception of the essential amino acid methionine which is discussed in detail in Section 1.4.

The amino acid glutamine serves as a source of both carbon and nitrogen, contributing to maintenance of cellular redox balance through glutathione (GSH) synthesis as well as to anabolic metabolism including the biosynthesis of lipids 16,17. Glutamine also maintains nitrogen pools for the synthesis of nonessential amino acids and nucleotides 18. Importantly, glutamine can also be critical for regulation of mitochondrial respiration through a process referred to as anaplerosis, whereby glutamine is converted to the amino acid glutamate to ultimately generate alpha- ketoglutarate (αKG), which can maintain the flow of the TCA cycle and thereby support ATP production 19. Numerous studies have demonstrated the importance of glutamine in certain cancer contexts, as inhibition of glutamine metabolism prevents tumor growth in a multitude of murine cancer models 20-23.

The utilization and recycling of serine and glycine, along with methionine (discussed in subsection 1.4), are collectively referred to as one-carbon metabolism, and serve as essential integrators of nutritional status. Conversion of serine to glycine via serine hydroxymethyltransferase (SHMT) generates one-carbon molecules within the folate cycle 24

(Figure 2), which serves as a major source of nucleotides and thus is a critical component of cell proliferation. Beyond contributing to nucleotide pools, one-carbon metabolism also supports the production of other types of biomass such as phospholipids, as well as the generation of redox factors such as NADPH and glutathionine. Serine and glycine additionally contribute to redox

balance through the transsulfuration pathway for cysteine synthesis, which is required for glutathione (GSH) production 25 (Figure 2). Further highlighting the importance of serine in cancer was the discovery that PHGDH, the gene encoding the enzyme that catalyzes the rate- limiting step of serine synthesis, is amplified in melanoma and breast cancer 26, and PHGDH inhibitors have shown efficacy in a handful of preclinical cancer models 27,28; interestingly, our group recently found that PHGDH activity couples serine-mediated nucleotide synthesis with central carbon metabolism, thereby disrupting global metabolic networks upon its inhibition 29.

Finally, the amino acids arginine, tryptophan, and the branch chain amino acids leucine, isoleucine, and valine have also been extensively implicated in cancer. Arginine can act as a nitrogen donor for nitric oxide (NO) signaling, a fuel for the urea cycle, or to maintain pools of nonessential amino acids, and tumors lacking the arginosuccinate synthase enzyme highly depend on exogenous arginine for growth 30. Tryptophan catabolism by tumors has additionally been shown to create an immunosuppressive tumor microenvironment 31,32, and high levels of the branched-chain amino acids leucine, isoleucine, and valine are found in the serum of early stage pancreatic cancer patients33. Additionally, leucine and arginine are potent activators of mammalian target of rapamycin (mTOR), a regulator of protein translation and cell growth that is often found to be dysregulated in tumorigenic settings 34.

1.2 Metabolic heterogeneity in human cancer patients i

Even within the initial work from Otto Warburg and others demonstrating increased rates of glycolysis in cancer, it was found that many tumors still undergo normal mitochondrial

i This chapter subsection was adapted and modified from a published commentary: Sanderson S.M. and Locasale J.W. “Revisiting the Warburg Effect: Some Tumors Hold Their Breath.” Cell Metabolism (2018). This text was reproduced in accordance with the CC-BY license. All text included in this section was written by S.M.S. 4

respiration. Indeed, in recent years, with the advent of in vivo analysis of cancer metabolism, it has been shown that non-small cell lung (NSCLC) tumors 35, liver cancers 36, as well as primary and metastatic brain tumors 37 exhibit significant and even elevated levels of complete glucose oxidation, as indicated by stable isotope tracing of mouse tumors or human patients infused with

13C-glucose (Figure 1). Certain cancers remain dependent on mitochondrial function in vivo 38,39, while others specifically require aerobic glycolysis 40. These phenomena have been largely attributed to oncogene-induced metabolic programming (which has been primarily characterized in cell culture), although intriguing evidence from in vivo studies suggests that the environment may play an equal if not larger role as these genetic alterations in shaping metabolic state 36,41

(detailed in Section 1.3). Overall, these observations have led to an evolved understanding of the

Warburg Effect, with the currently predominant view being that tumors acquire the ability to undergo enhanced aerobic glycolysis while simultaneously maintaining if not also upregulating mitochondrial metabolism.

Interestingly, one study recently demonstrated some of the first experimental evidence in human patients that the metabolic switch proposed by Warburg indeed occurs in certain contexts such as in clear cell renal cell carcinoma (ccRCC), in which tumors exhibit a metabolic phenotype characterized by reduced utilization of glucose oxidation in favor of aerobic glycolysis

42 (Figure 1). These findings illustrate the increasing complexity of characterizing cancer-specific metabolic phenotypes, and indicate that factors beyond oncogenic genetic events contribute to the heterogeneity in metabolic reprogramming across different cancer subtypes.

Figure 1: Metabolic heterogeneity between tumor subtypes in human cancer patients.

1.3 Environmental determinants of metabolic programming in cancer

Despite the identification of oncogene-associated metabolic phenotypes 43, it has become increasingly appreciated that the exceptional degree of metabolic heterogeneity found in tumors is overwhelmingly non-attributable to genetics alone. Indeed, a growing body of evidence suggests that environmental factors, such as originating tissue and nutrient availability, can have a significant if not greater impact on cancer cell metabolism than genetic status. As metabolic demands increase to support continual tumor growth, nutrients in the microenvironment become exceedingly low 44, and many tumors undergo extensive metabolic reprogramming to adapt to nutrient-scarce conditions. For instance, although cancer cells are known to exhibit enhanced dependence on glutamine, it has been shown that some cancer cells can adapt to limited glutamine availability by altering their glucose consumption 45,46, while others have shown that the degree of glutamine dependence is directly impacted by environmental availability of cystine

(the oxidized form of cysteine) via activity of the cystine/glutamate antiporter 47. Furthermore, the availability of histidine and asparagine can in certain contexts mediate the response to methotrexate 48 or the progression of breast cancer metastasis 49, respectively. Studies have also demonstrated that dietary nutrient restriction as a therapeutic modality itself can even exert 6

antineoplastic consequences, with the dietary removal of serine and glycine shown to directly modulate cancer outcome 50.

Gene-environment interactions further add to the complexity of metabolic programming in cancer whereby genetic status can impact tumor responsiveness to the environment, and environmental factors can similarly drive specific genetic programs. For example, while the oncogenic signaling protein Ras has been shown to upregulate glucose transporters and increase cellular glucose consumption 51, it was subsequently discovered that the metabolic dependencies of Ras-driven tumors are highly dependent on the microenvironment of the residing tissue 41.

These results are additionally supported by multiple findings that the same oncogenic event can induce entirely distinct metabolic programs in tumors originating from different tissues 36,52.

Furthermore, glutamine deficiency has been shown to promote de-differentiation through inhibition of histone demethylation 53, and diets deficient in methionine or other one-carbon nutrients have also been shown to directly impact histone and DNA methylation profiles 54-56.

Indeed, methionine in particular has been extensively linked to cancer cell biology which is discussed in-depth in the following section.

These collective observations illustrate the exceptional difficulty in defining cancer- specific metabolic programs, particularly between diverse cancer subsets and in various environmental contexts; however, given the broad implications that gene-environment interactions are increasingly found to have on tumor biology, increased characterization of these relationships will be crucial for improving clinical outcomes by identifying novel therapeutic strategies as well as specific patient subpopulations that would benefit from them.

1.4 Methionine metabolism in health and cancer ii

1.4.1 Methionine cycle and related pathways

Methionine is an essential sulfur-containing amino acid that is catabolized and recycled in a series of metabolic reactions termed the methionine cycle (Figure 2). Briefly, methionine is converted to the universal methyl donor S-adenosyl-methionine (SAM), which upon donation of its methyl group is converted to S-adenosyl-homocysteine (SAH). SAH is hydrolyzed to generate homocysteine, which is then converted to cysteine via the transsulfuration pathway or, with a methyl donation from the folate cycle, back into methionine. As discussed previously (section

1.1.2), this cycle is closely linked to the folate cycle (fueled in large part by serine and glycine), collectively forming the two major components of what is referred to as one-carbon metabolism

25; this metabolic network allows for the integration of nutritional carbon units in a diverse set of critical cellular processes (Figure 3). Additionally, methionine can also be recycled from the

SAM-dependent polyamine biosynthesis byproduct methylthioadenosine (MTA) and further processed by the enzyme methylthioadenosine phosphorylase (MTAP) via the methionine salvage pathway. As with most metabolic processes, each of these biochemical reactions is tightly regulated and coupled to the other reactions in one-carbon metabolism, and aberrant activity within one node of this network can lead to drastic dysregulation of cellular function as is commonly observed in disease contexts such as cancer.

ii This chapter subsection was adapted and modified from a published review: Sanderson S.M., Gao X., Dai Z., and Locasale J.W. “Methionine Metabolism in Health and Cancer: a Nexus of Diet and Precision Medicine.” Nature Reviews Cancer (Epub Sept 12, 2019). This text was reproduced in accordance with the CC-BY license. All text included in this section was written by S.M.S. 8

Figure 2: The methionine cycle and related metabolic pathways.

1.4.2 Role of methionine in biological processes

One of the most prominent functions of methionine is in the contribution to intracellular methylation by serving as the sole source of the universal methyl donor SAM; SAM is a necessary substrate for all methylation reactions, including those that modulate gene expression

(via methylation of DNA, RNA, and histones), phospholipid integrity, activity of signaling pathways, and polyamine biosynthesis. Due to the intracellular concentration of SAM relative to

57 the Km values of cellular methyltransferases , SAM availability can directly impact these processes thereby serving as a metabolic link between one-carbon nutritional status and cellular behavior. Further illustrating its biological importance, the recently discovered protein SAMTOR was found to specifically function as a SAM sensor by inhibiting mTORC1 activity under

conditions of low methionine 58. SAM is also necessary for the biosynthesis of polyamines

(including putrescine, spermidine, and spermine), which function to maintain protein, DNA, and

RNA stability, protect against oxidative stress, and regulate activity of ion channels 59.

Importantly, major disruptions in a subset of SAM-consuming reactions can significantly drive aberrant methylation patterns in other methylation-dependent processes 60.

Beyond mediating these SAM-dependent reactions, methionine also contributes to essential metabolic pathways that regulate nucleotide biosynthesis and intracellular redox balance. Via contribution of homocysteine, methionine directly contributes to the folate cycle which provides multiple inputs to both purine and pyrimidine biosynthesis (Figure 2); furthermore, the methionine salvage pathway produces adenine from the polyamine metabolic byproduct methylthioadenosine (MTA), creating an additional substrate for purine metabolism.

Methionine also contributes to maintenance of cellular redox status by providing homocysteine as a substrate for the transsulfuration pathway, which ultimately produces the antioxidant glutathione (GSH). The subsequent reversible oxidation of GSH to GSSG effectively combats cellular damage caused by reactive oxygen species (ROS) 61. Further contributing to its role as a regulator of cellular oxidative stress, methionine also functions as a source of sulfur for the

62 production of the critical signaling molecule hydrogen sulfide (H2S) . Finally, methionine plays a particularly significant role in autophagic processes63 and in protein synthesis via multiple mechanisms 64 (Figure 3).

Figure 3: Role of methionine in biological processes.

1.4.3 Health phenotypes associated with dietary methionine availability

A number of studies have demonstrated a connection between dietary methionine restriction (MR), which reduces but does not completely eliminate methionine, and improvement of health as well as reversal of a myriad of pathological phenotypes. Following a key finding that

MR can extend lifespan in yeast65, numerous studies have further demonstrated the evolutionary conservation of lifespan extension in Drosophila66, C. elegans67, mouse68,69, and rat70. These observations provided one of the first definitive links between specific dietary amino acid composition and longitudinal health maintenance, and further supported the scarcely-tested theory that dietary MR could potentially provide a therapeutic benefit in diseases such as cancer.

Lending further support to this hypothesis are the health benefits that have been observed in numerous studies of MR in mice, most notably in reduced adiposity71, improved cardiac function72, and increased insulin sensitivity73. It is currently unclear whether these health- promoting properties of MR are driven by cell-autonomous changes in methionine metabolism

(Figure 3) or a result of systemic alterations in metabolic regulation. Nevertheless, these MR- mediated benefits in age-related disorders illustrate the notion that methionine metabolism is intrinsically linked to cancer biology.

1.4.4 Dietary methionine in cancer

Given the expansive role of methionine in normal biological processes, it is perhaps unsurprising that alterations within its metabolic network have been widely implicated in cancer.

The clinical applicability of cancer-specific alterations in methionine consumption and utilization is perhaps most readily apparent by the observation that intratumoral methionine uptake, as evidenced by PET imaging of 11C-methionine, is at least in certain contexts more indicative of therapeutic response and overall survival than glucose uptake 74,75. However, the therapeutic potential of targeting this cancer-specific phenotype remains an active area of investigation.

The antineoplastic effect of the complete removal of methionine from the diet (i.e. dietary methionine deprivation) was first reported in Sprague-Dawley rats carrying the Walker-256 carcinosarcoma, whereby animals were fed diets lacking individual amino acids and were subsequently shown to exhibit significantly reduced tumor growth under a methionine-deprived diet 76. Following this observation, a number of additional animal studies have reported similar findings in various settings. For instance, MR was shown to effectively induce a cell cycle blockade and overall tumor regression in Yoshida sarcoma-bearing nude mice 77 as well as in a xenograft model of glioma 78. Other reports further demonstrated that depleting dietary methionine could induce sensitivity to cytotoxic agents such as cisplatin79 and doxorubicin 79,80 in

drug-resistant xenograft tumors in mice. A more recent study additionally observed enhanced efficacy of lexatumumab when combined with dietary MR in an orthotopic triple-negative breast cancer (TNBC) mouse model 81, an observation which was followed by another study that found reduction in dietary methionine alone was effective in suppressing lung metastasis in a TNBC xenograft mouse model 82. Although the consistency of the antineoplastic effects found with MR is promising, many of these studies have yet to draw definitive conclusions about what aspects of molecular metabolism are driving the observed effects and whether these outcomes may extend to more advanced preclinical models.

1.4.5 MTAP deletions in cancer

Genetic deletions of methylthioadenosine phosphorylase (MTAP) are commonly found in tumors due to the proximity to the CDKN2A locus on chromosome 9p21, which encodes one of the most frequently altered tumor suppressors p16 83; ~15% of all cancers (most notably glioblastoma 84, melanoma 85, mesothelioma 86, and pancreatic 87) exhibit deletions of this chromosomal locus, with MTAP codeletions occurring in ~80-90% of this subset 88. Alterations in

MTAP expression have also been found independent of CDKN2A deletion, due to hypermethylation of the MTAP promoter 89,90 or (in some rare cases) selective homozygous deletion of MTAP 91, suggesting that MTAP could potentially function as a tumor suppressor in addition to its established role as a passenger event 92, although more studies are needed to definitively establish this. As mentioned previously, MTAP is an enzyme in the methionine salvage pathway that converts the polyamine biosynthesis byproduct methylthioadenosine (MTA) ultimately into methionine and adenine (Figure 2); loss of MTAP expression has thus been found to consistently induce intracellular accumulation of its substrate MTA 93-95.

Due to the difficulty in therapeutically targeting tumor-suppressor pathways, MTAP deletion has gained considerable interest as a modifier of cancer-specific metabolic

vulnerabilities. Some preliminary work has found that MTAP-deleted cells exhibit enhanced sensitivity to inhibitors of de novo purine metabolism 96-98. Interestingly, recent studies have also identified a specific arginine methyltransferase, PRMT5, that exhibits a high degree of sensitivity to intracellular MTA levels due to high affinity of MTA for the SAM binding pocket of PRMT5

95 and is selectively required for cell growth across diverse MTAP-deficient cell lines 93,95.

Importantly, in these studies the supplementation of MTA in MTAP-expressing cells was found to induce sensitivity to PRMT5 inhibition, demonstrating that the metabolic vulnerability induced by MTAP deletion is specifically a result of MTA accumulation. Another study has also shown evidence that the reduction of PRMT5 activity found in multiple MTAP-deficient cell lines creates additional vulnerabilities in methionine metabolism that can be targeted therapeutically 94.

Although an understanding of how PRMT5-mediated arginine methylation regulates cellular processes, and by extension why MTAP-deleted cells exhibit enhanced dependency on its activity, is currently lacking, aberrations in PRMT5 activity and expression have been implicated in numerous cancer types 99-101 thereby providing support for the therapeutic potential of targeting its activity.

1.4.6 Other methionine-associated alterations in cancer

It has recently been shown that tumor-initiating cells exhibit elevated activity of enzymes within the methionine cycle, most notably an upregulation of MAT2A expression and activity 102.

Additionally, activity of the methyltransferase nicotinamide N-methyltransferase (NNMT), in addition to its cell autonomous cancer promoting function103, was found to be a driver of the oncogenic behavior of cancer-associated fibroblasts 104. In this context NNMT, which uses SAM to convert nicotinamide into NAD+ and the metabolically inert byproduct 1-methylnicotinamide

(1-MNA) 105, was shown to consume the available SAM pool in these cancer-associated

fibroblasts, thereby diverting SAM from DNA and histone methylation processes (a phenomenon referred to as a “methyl sink”), ultimately leading to metastasis and overall cancer progression 104.

Elevated activity of polyamine metabolism, which directly branches from the methionine cycle (Figure 2), has also been long associated with rapid cell proliferation 106. Since this discovery, genetic alterations that lead to changes in expression or activity of enzymes that regulate polyamine metabolism have been found to be highly prevalent in cancer cells.

Overexpression of the polyamine biosynthesis enzyme ornithine decarboxylase (ODC) is frequently observed across numerous cancer types 107-109, and has been shown to promote cancer cell growth in preclinical studies 110-112. Furthermore, ODC expression has been shown to be predictive of the degree of tumorigenicity as well as therapy resistance 113-115. Of particular interest is the finding that ODC activity is regulated by 2-keto-4-methylthiobutyrate (MTOB), an intermediate in the MTAP-mediated conversion of MTA to methionine 87; it was later found that

ODC overexpression is frequently associated with concurrent deletion of MTAP in pancreatic cancers 116.

Another polyamine enzyme found to be dysregulated in cancer, adenosylmethionine decarboxylase (AMD1), has gained considerable interest. AMD1 serves as an enzymatic link between the methionine cycle and polyamine biosynthesis via decarboxylation of the universal methyl donor SAM, ultimately controlling the interconversion of the polyamine putrescine to spermidine (Figure 2). One investigation identified AMD1 as a putative tumor suppressor in a murine model of lymphoma, and subsequently discovered that heterozygous deletions of AMDI are prevalent in human lymphomas 117. It was more recently shown that the expression of AMD1 is frequently upregulated in mTORC1-driven prostate cancers, and that tumors excised from patients treated with an mTORC1 inhibitor exhibit decreased AMD1 expression and reduced proliferation 118.

1.5 Targeting metabolic vulnerabilities in cancer iii

1.5.1 Antimetabolite chemotherapies

Folates, commonly supplemented in the diet in their more stable oxidized form as folic acid, are a major component of one-carbon metabolism and thus are essential for major biosynthetic reactions such as amino acid metabolism and cell proliferation 25. For instance, 5- methyl-tetrahydrofolate (5-methylTHF) is used along with vitamin B12 to generate methionine 119

(Figure 2). In these folate-mediated reactions, dihydrofolate (DHF) is produced, and is then reduced by dihydrofolate reductase (DHFR) to regenerate THF cofactor pools.

The selective dependence on folates for cancer cell proliferation (discussed in Section

1.1.2) can be therapeutically exploited with a class of drugs known as antifolates. Methotrexate, a potent and well-characterized polyglutamatable antifolate compound, mimics folic acid but has an additional methyl group at the N10 position; this allows the compound to have an exceptionally high affinity for DHFR, but the subtle structural methyl modification antagonizes the enzyme’s ability to reduce DHF 120,121 (Figure 4). However, methotrexate can exhibit a toxicity profile against normal proliferating tissues such as bone marrow and intestinal tissue 122; to increase the therapeutic window, usually high doses of methotrexate are administered followed by low doses of leucovorin (5-formylTHF) to competitively inhibit intracellular transport of methotrexate in tissues with close proximity to blood vessels (referred to as the “leucovorin rescue”) 123,124.

Methotrexate remains a primary treatment option for lymphoma, osteosarcoma, and leukemias120.

Other polyglutamatable antifolates that inhibit DHFR activity include pralatrexate and

iii The chapter subsections 1.5.1, 1.5.2, and 1.5.4 were adapted and modified from a published textbook chapter: Reid M.A., Sanderson S.M., and Locasale J.W. (2019) “Chapter 9: Cancer metabolism. In Niederhuber J.E. (Ed.) et al. Abeloff’s Clinical Oncology (pp. 127-138).” Philadelphia, PA: Elsevier. This text was reproduced in accordance with the CC-BY license. All text included in this section was written by S.M.S. 16

pemetrexed, while the non-polyglutamatable antifolates that act on DHFR include trimetrexate, piritrexim, and talotrexin 125. Pemetrexed remains the first line treatment for non-small cell lung cancer (NSCLC) as a dual treatment with cisplatin (a DNA cross-linking compound) 126, and remains under clinical investigation for the treatment of various other cancers 127.

Another class of antimetabolites includes compounds that directly interfere with DNA or

RNA synthesis. The original and most well known of these compounds is 5-flurouracil (5-FU), a pyrimidine analog that bears structural similarity to uracil but with a fluoride atom at the 5C position of its ring128. The TS enzyme converts uracil to thymine via a methylation reaction at the

5C carbon position, but is unable to execute this methylation of 5-FU due to the presence of the fluoride atom, effectively inhibiting TS (Figure 4). 5-FU remains a commonly used agent clinically and is regularly used as a frontline agent for the treatment of colorectal cancer.

Gemcitabine (2’,2’-diflurodeoxycytidine, or dFdC) is another popular pyrimidine analog used as a chemotherapeutic; gemcitabine is a deoxycytidine analog that has fluoride atoms in place of the hydrogen atoms at the 2C position of its ring 129 (Figure 4). Gemcitabine (both alone and in combination with other therapies such as cisplatin) has been approved for the treatment of numerous cancers including breast, ovarian, bladder, NSCLC, and pancreatic cancer, and is being investigated as an adjuvant to 5-FU therapy for metastatic colorectal cancer130.

Finally, the pyrimidine analog decitabine (also referred to as aza-dCyd) is converted by deoxycytidine kinase to aza-dCTP (Figure 4) and is rapidly incorporated into DNA, ultimately preventing the DNA methylation. Normally, cytosine residues of DNA can be methylated to prevent transcription; aza-dCTP inhibits DNA methyltransferases, thereby inducing enhanced gene expression 131. Decitabine is approved for the treatment of myelodysplastic syndromes, including acute myeloid leukemia in elderly cancer patients 132.

Figure 4: Mechanism of action of common antimetabolite chemotherapies.

1.5.2 Targeting methionine metabolism in canceriv

Efforts have been made to develop various methionine analogs in hopes of selectively targeting cancer cells as had been done with other efficacious antimetabolite therapies 120. While one of these analogs, ethionine, appeared to demonstrate preclinical efficacy 133, it was ultimately found to be toxic and clinical investigations of its use have subsequently been abandoned. A similar approach involved the administration of methioninase (and its more stable recombinant form rMETase), which degrades methionine to α-ketobutyrate, methanethiol, and ammonia rather than SAM 134. Although a phase I clinical trial demonstrated that its administration was tolerable and effectively lowered serum methionine levels 135,136, over 20 years have passed since this

iv This chapter subsection (1.5.3) was adapted and modified from a published review: Sanderson S.M., Gao X., Dai Z., and Locasale J.W. “Methionine Metabolism in Health and Cancer: a Nexus of Diet and Precision Medicine.” Nature Reviews Cancer (Epub Sept 12, 2019). This text was reproduced in accordance with the CC-BY license. All text included in this chapter was written by S.M.S. 18

initial phase I trial and it has yet to advance to subsequent clinical development. It is worth noting, however, that one group has recently published a number of studies demonstrating efficacy of rMETase in patient-derived xenograft mouse models of melanoma137 and sarcoma51, as well as in an orthotopic model of osteosarcoma138. Additionally, a very recent independent pilot phase I clinical trial of rMETase was conducted with no toxicities reported, suggesting a possible resurgence of interest in this therapeutic approach139. Nonetheless, a number of novel strategies to therapeutically target methionine metabolism are also currently under active investigation.

Therapies targeting the methionine salvage pathway: The inability of MTAP-deleted cells to synthesize adenine (a purine derivative) from MTA (Figure 2) provides an attractive potential therapeutic approach of targeting purine synthesis in this subset of tumors. Initial studies focused on identifying ways to take advantage of this vulnerability, primarily via administration of adenine analogs that inhibit formation of critical intermediates for nucleotide synthesis 140,141; however, as noted previously, a phase II trial investigating the adenine analog L-alanosine (also referred to as SDX-102) failed to show efficacy in advanced-stage tumors exhibiting MTAP deletion 142. Although efforts using this approach have been mostly abandoned, preclinical investigations of MTAP-associated vulnerabilities within purine metabolism remain ongoing.

The recent discoveries of additional metabolic vulnerabilities induced by MTAP deletion create a multitude of potential avenues for clinical investigation. Upregulation of PRMT5 expression and/or activity has been extensively implicated in numerous cancer types 99-101, which has resulted in three ongoing phase I clinical trials with one phase II clinical trial already approved (NCT03614728, NCT03854227, NCT03573310, clinicaltrials.gov). Although these investigations are not specific to tumors with MTAP deletion, it is likely that the applicability of those compounds to MTAP-deleted tumors will be an active area of investigation in the near

future. Another therapeutic approach that has gained considerable attention recently is an inhibitor of MAT2A, which was shown to exhibit substantial efficacy in preclinical models 94.

However, given the enhanced methionine dependency of tumors independent of MTAP status, it is likely that other tumors may also exhibit sensitivity to MAT2A inhibition; it will be interesting to see the future clinical applications this therapeutic approach may have.

Therapies targeting polyamine metabolism: Given the role of polyamines in cancer (Section

1.4.6), targeting the metabolic processes associated with their regulation is another potentially promising therapeutic strategy. Difluoromethylornithine (DFMO), an irreversible inhibitor of ornithine decarboxylase (ODC) was developed shortly after the requirement of polyamines for cellular growth was discovered 143. This compound was considered to be a particularly attractive therapeutic agent due to its apparent selective cytotoxicity against malignant cells 144; however, it failed to demonstrate efficacy as a single agent in early clinical trials 145,146, potentially due to upregulation of polyamine uptake from the microenvironment 147, although a phase II clinical trial recently demonstrated its efficacy as a chemopreventative agent in reducing incidence of relapse in neuroblastoma 148.

A resurgence of interest in DFMO as an adjuvant chemotherapeutic agent has recently occurred, particularly in neuroblastomas that are characterized by MYC overexpression 149 as

ODC expression has been found to be directly regulated by c-Myc activation in murine fibroblasts

150, with ongoing phase I clinical trials examining DFMO co-administration with either cyclophosphamide or the polyamine transporter inhibitor AMXT-1501 (NCT03536728, clinicaltrials.gov). Interestingly, preclinical evidence suggests that the combination of DFMO with AMXT-1501 may exert antitumor effects in an immune-dependent manner by preventing T- cell immune repression 151, providing further support for future investigations of this therapeutic approach. Finally, as mentioned previously (Section 1.4.6), MTAP-deleted cells have also been

shown to exhibit enhanced ODC activity 87. Although preclinical studies have failed to demonstrate efficacy of DFMO in MTAP-deleted cells 152, ongoing efforts to identify contextual factors that could potentially enhance tumor sensitivity in MTAP-deleted cells may provide novel therapeutic strategies for clinical use of DFMO as an anti-neoplastic agent.

Therapies targeting the methionine cycle: Substantial efforts have been made to identify other novel approaches to chemically disrupt methionine metabolism and the processes that are reliant on its activity. Two additional compounds targeting MAT2A have recently been described in preclinical studies. One of these compounds, Acetyl-11-keto-b-boswellic acid (AKBA), is a natural MAT2A inhibitor that demonstrated activity in keratinocytes 153, while the other is a small-molecule allosteric modulator (PF-9366) which inhibits MAT2A when methionine or SAM levels are high, and activates MAT2A when levels of these metabolites are low; however, chronic

PF-9366 treatment can result in the compensatory upregulation of MAT2A expression154. This feedback mechanism, although poorly characterized and only reported in PF-9366, could potentially reduce the therapeutic potential of these compounds.

Another novel approach under investigation is the inhibition of the pro-angiogenic protein methionine aminopeptidase (MetAP2), a metalloenzyme responsible for removal of N- methionine residue from nascent proteins, thereby effectively impairing protein synthesis 155 A

MetAP2 inhibitor, M8891, is currently in phase I clinical trials for advanced solid tumors

(NCT03138538, clinicaltrials.gov) although an earlier MetAP2 inhibitor (ZGN-440, or beloranib), which was previously investigated for its ability to promote metabolic health in patients with obesity, as evidenced by weight loss and increased insulin sensitivity, was pulled from clinical trials due to vascular toxicity 156. These approaches further illustrate both the high therapeutic potential as well as the appreciable challenges in pharmacologically targeting methionine-related processes and other cancer-specific metabolic vulnerabilities.

1.5.3 Repurposing of common metabolic agents

The chemotherapeutic repurposing of agents that have been traditionally used to treat various disorders is another major area of investigation due to the relatively reduced toxicity, increased tolerability, and lower cost of these agents relative to conventional chemotherapies

157,158. Additionally, these FDA-approved compounds have the potential to enter phase III clinical trials more readily than novel drugs as their safety profiles have already been well established.

The following section will summarize some of the current drugs that are being investigated as repurposed cancer treatments.

Metformin: Metformin is the leading therapy prescribed for the management of type II diabetes 159. This compound has been shown to reduce hepatic gluconeogenesis 160 and increase insulin sensitivity 161, thereby increasing peripheral glucose uptake and reducing circulating levels of glucose. While the exact mechanism of action that mediates its efficacy remains unknown 162, metformin has been observed to inhibit complex I (NADH dehydrogenase) of the mitochondrial respiratory chain 163, inhibit mitochondrial glycerophosphate dehydrogenase 160, and activate

AMP kinase (AMPK) signaling through these mechanisms in certain settings 164. Metformin also inhibits insulin production and insulin growth factor (IGF) signaling 165, which can also result in antineoplastic effects.

Substantial evidence for metformin directly affecting tumor cell autonomous metabolism is also apparent. As mitochondria are known to have pleiotropic functions in maintaining tumorigenesis166, mitochondrial disruption by metformin has been observed to have global effects on altering mitochondrial metabolism in tumor cells. For instance, although it was originally believed that metformin exerted its antitumor effects primarily via its ability to activate AMPK signaling, it was found that cancer cell lines that demonstrated high sensitivity to metformin tended to exhibit mitochondrial mutations in complex I genes; this sensitivity was abolished by

the exogenous expression of an enzyme (NDI1, derived from yeast) that enables cells to undergo oxidative phosphorylation without a functional complex I 167, implying an AMPK-independent antineoplastic mechanism. Similar results were found in murine models, in which tumors of

NDI1 transgenic mice failed to show sensitivity to metformin compared to tumors harbored by mice that did not express NDH1163. This is also consistent with metabolic profiles that are observed in the biopsies of ovarian tumors obtained from patients taking metformin, in which the level of metformin accumulation within tumors was found to correlate with the degree of mitochondrial metabolic dysfunction including accumulation of both NADH and reactive oxygen species (ROS) 38.

Both preclinical and epidemiological studies have demonstrated the antitumor potential of metformin. Some meta-analysis studies have found an approximate 30% reduced cancer incidence among type II diabetic metformin users compared to non-users 168. Dozens of clinical trials examining metformin as either a monotherapy or in combination with various chemotherapies are currently ongoing, especially as a repurposed treatment for breast, endometrial, pancreatic, and prostate cancers 169.

Statins: Inhibitors of the hydroxymethylglutaryl Coenzyme A (HMG-CoA) reductase enzyme, commonly referred to as statins, are a class of cholesterol-lowering drugs that are used clinically as a preventative measure against (as well as treatment for) cardiovascular disease

170,171. A handful of statins are currently FDA-approved, including lovastatin (the most commonly prescribed), simvastatin, fluvastatin, atorvastatin, rosuvastatin, and pravastatin. Statins are known to inhibit lipogenesis via inhibition of the mevalonate pathway 172. Interestingly, preclinical studies have demonstrated that statins are able to effectively induce cell death in various types of human cancer cells 172. Numerous clinical trials have investigated the efficacy of statins in the prevention and/or treatment of various types of cancer, yet the majority of these trials have shown

inconclusive results 173; a recent meta-analysis study found that statin treatment was negatively correlated with cancer incidence, but this effect was mostly limited to lung and prostate cancer and was concluded to be nonsignificant 174. It is believed that the inconclusive results found in the large number of observational investigations are due to multiple confounding factors that differed between studies, including timing of when treatment was initiated, dosage regimen, and heterogeneity of other risk factors (such as obesity and smoking). However, the use of statins in the prevention and treatment of cancer still remains an attractive prospect and continues to be under investigation.

Aspirin: The nonsteroidal anti-inflammatory drug (NSAID) aspirin, a cyclooxygenase

(COX) inhibitor that reduces prostaglandin and thromboxane production, is an over-the-counter medication used to relieve pain and reduce fever. Aspirin is also commonly recommended by physicians to reduce the risk of heart attack and stroke, as well as the recurrence of pulmonary embolism and deep vein thrombosis, due to its ability to reduce blood clotting 175. Interestingly, recent evidence appears to also link long-term aspirin use to protection against cancer development and mortality, particularly colorectal cancer176. An observational study investigated aspirin use in 135,965 women and men over the course of 32 years, and found that regular low- dosing (weekly administration of 0.5-1.5 standard tablets) of aspirin over the course of 6 years reduced overall cancer risk by 3%, with a 15% reduction in gastrointestinal cancers 177. While it has been demonstrated that aspirin can activate AMPK signaling 178, it is believed that aspirin may exert its antitumor effects via both off-target (COX-independent) and on-target (COX- dependent) mechanisms 179. For instance, COX enzymes have been found to play a significant role in the development of intestinal tumors in animal models 180. Aspirin may also reduce the incidence of tumor metastasis via its antiplatelet activity 181, as platelets have been shown to promote secondary lesion development by shielding tumor cells from immune responses 182.

Nevertheless, more work is needed to understand the molecular mechanisms that may contribute to these antineoplastic effects.

Vitamins C and D: Recent evidence suggests that some commonly used compounds contained in diet and used as nutritional supplements, particularly Vitamin C and Vitamin D, may play a much more substantial role in combating cancer progression than was originally believed.

For instance, a recent preclinical study found that deletion of the vitamin D receptor accelerated tumor growth and promoted metastasis in a murine model of breast cancer 183, implying that vitamin D could act as a negative regulator of tumor progression. Another study found that vitamin C was able to induce oxidative stress in cells that could inactivate the glycolytic enzyme glyceraldehyde 3-phosphate dehydrogenase (GAPDH), thereby effectively inhibiting glycolysis; this finding extended to mouse models, in which vitamin C administration significantly impaired tumor growth 184. Interestingly, recent studies have also demonstrated that vitamin C can boost the efficacy of DNA methyltransferase inhibitors, such as 5-Azacytidine and 5-aza-2’- deoxycytidine, by enhancing the catalytic activity of DNA demethylases that oxidize methylated cytosine residues to hydroxymethyl cytosine 185,186. Given the tolerability of these vitamins, these preclinical findings are an exciting new field of investigation and remain under clinical investigation as an adjuvant therapy187.

Exercise: Given the increased metabolic and energetic demands in cancer, a significant field of research is currently investigating the role of energy expenditure in reducing tumor development and recurrence. Physical activity has been linked to prolonged disease-free survival and quality of life amongst cancer patients 188, and recent studies have observed a positive effect of exercise on the reduction of proangiogenic 189 and hypoxic 190 factors, resulting in reduced tumor vascularity. Furthermore, physical activity remains the primary mode for weight loss in both healthy and obese individuals. Obesity has been shown to increase risk of numerous cancers

including breast, prostate, pancreatic, and colon 191 with some estimates concluding that up to one-in-five cancers are directly attributed to obesity 192. Obesity is characterized by excessive amounts of adipose tissue resulting in excessive nutrient storage in the form of lipids. Adipose tissue is important for energy storage, cell signaling through hormone secretion, and inflammatory responses through cytokine secretion; all of which are altered in the obese state 193.

It has been shown in ovarian cancer that fatty acids produced by local adipose tissue can fuel tumor cell metastasis 194, thus suggesting that fat depots in themselves can promote tumorigenesis

195. Therefore, a role for exercise in inhibiting or reversing tumor growth is a promising current area of investigation, including in my own work that I describe in Chapter 5.

1.6 Scope of dissertation

In this dissertation, I investigate the impact of environmental factors on metabolic programming, and the extent to which their effects are modulated by genetic status. I use a metabolomic, pharmacological, genetic, behavioral, and systematic approach to characterize these potential gene-environment interactions in the context of cancer, with nutrient availability

(chapters 2 and 3) and energetic demand (chapters 4 and 5) as environmental variables of interest.

These studies examine lifestyle intervention (i.e. diet or exercise) as well as cardiac drug repurposement in the context of metabolic reprogramming in cancer; however, although I utilize various cancer models for these investigations, many of the principles of gene-environment interactions that I discuss also exhibit basic applications to cell biology and human physiology.

In chapter 2, I examine how the genetic deletion of a metabolic enzyme impacts cellular metabolism relative to environmental composition, with MTAP deletion and one-carbon nutrient availability as variables of interest. Using global metabolite profiling, I demonstrate that while

MTAP-deleted cells exhibit an observable metabolic signature across diverse cell types, neither

MTAP status or tissue of origin are predictive of metabolic responsiveness to nutrient restriction. 26

With the use of multiple isogenic cell pairs, I show that re-expression of MTAP has heterogeneous consequences on metabolism. Finally, I utilize a machine learning classifier to quantify the relative degree to which the global metabolic profile of a cell line is predicted by

MTAP status compared to nutrient availability.

In chapter 3, I survey a range of nutrient supplementations to elucidate potential molecular mechanisms that mediate the cytotoxic effects of MR both as a single agent and in combination with 5-FU treatment. I characterize the sensitivity to MR with and without 5-FU in colorectal cell lines, and I provide evidence to support in vivo observations that nucleotide metabolism and redox balance are primary factors contributing to the synergistic effects of MR and 5-FU.

In chapter 4, I demonstrate that pharmacological inhibition of the Na+/K+ ATPase with the cardiac glycoside digoxin impacts intracellular energy demand and energy-producing processes. Through global metabolite profiling and isotope tracing I show that central carbon and mitochondrial metabolism, as well as other energy-associated metabolites such as taurine and creatine, are altered by acute digoxin treatment and are specific to Na+/K+ ATPase blockade. I find that these metabolic processes are similarly impacted in cardiac tissue with acute digoxin administration in mice, and that long-term digoxin treatment at sublethal doses significantly delays tumor growth in an allograft mouse model of sarcoma. Furthermore, I provide evidence that acute digoxin treatment can shift the tumor microenvironmental landscape and reprogram multiple tumor cell populations.

Lastly, in chapter 5 I use exercise as a physiological stimulus to alter energetic demand in six patient-derived xenograft (PDX) mouse models of colorectal cancer. I find that although these different models exhibit varying sensitivity to exercise as evidenced by tumor growth rates, exercise significantly impacts energetic processes within tumors independent of growth

inhibition. Furthermore, I show that exercise-responsive tumors exhibit metabolic profiles that are distinct from exercise-nonresponsive tumors, most notably in alterations to creatine-associated metabolites, providing potential avenues for future characterization of patient populations that may benefit from exercise as an adjuvant therapy.

Overall, this dissertation aims to elucidate the degree of cross-talk between intrinsic and extrinsic factors that ultimately shapes global metabolic state. The described work accomplishes this by surveying contexts in which environmental composition impacts genetically-driven phenotypes, as well as characterizing heterogeneity in cell-autonomous responsiveness to alterations in metabolic demand. Together, these findings significantly contribute to our understanding of gene-environment interactions in cancer and cell biology, and provide additional support for the chemotherapeutic potential of targeting metabolic processes through lifestyle modifications or drug repurposing in cancer.

2. Nutrient Availability Shapes Methionine Metabolism in p16/MTAP-deleted Cellsv

2.1 Background and Context

As described in Chapter 1, metabolic phenotypes arise from a complex interaction between genes and the environment. Determinants of these phenotypes include the genomic encoding of metabolic genes and their sequence variants, transcriptional and allosteric regulation of metabolic enzyme activity, and nutrient availability. Despite this complexity, the prospect of targeting metabolism for therapy is attractive because of both the relative drugability of metabolic enzymes and the numerous metabolic alterations observed in pathological conditions such as in cancer. Nevertheless, principled strategies that define context-specific metabolic differences are desired. For example, a number of studies have investigated MTAP deletion as a possible collateral lethality and have identified vulnerabilities in this subset of cancers 87,96,140,152,196, with several investigating some of the effects of MTAP deletion on methionine metabolism 93-95.

However, a systematic comparison of the relative effects of MTAP deletion as it relates to other variables that have been shown to shape metabolism is lacking.

Utilizing a metabolomics approach 197-200, we quantified the impact of MTAP deletion on metabolism in the context of cell type and the availability of nutrients related to one-carbon

v This chapter was adapted and modified from published work: Sanderson S.M., Mikhael P.G., Ramesh V., Dai Z., and Locasale J.W. “Nutrient availability shapes methionine metabolism in p16/MTAP-deleted cells.” Science Advances (2019). This text was reproduced in accordance with the CC-BY license. Author contributions: Conceptualization, S.M.S and J.W.L.; Cell culture experiments, S.M.S.; Metabolomics and data analysis, S.M.S. and P.G.M.; mathematical modeling, P.G.M. and Z.D.; all other experiments, S.M.S. and V.R. All text in this chapter was written by S.M.S.

metabolism. We find that while MTAP deletion produces a defined metabolic signature, this signature is diminished upon consideration of the changes to metabolism that result from the availability of nutrients related to methionine and one-carbon metabolism. Furthermore, these changes vary widely across individual cell lines and are not predicted by MTAP status. Thus, upon consideration of other variables that shape metabolic processes, MTAP status alone appears to exert a relatively modest effect on cellular metabolism.

2.2 Results

2.2.1 MTAP status has a defined metabolic signature

MTAP uses the substrate methylthioadenosine (MTA) to allow for the recycling of methionine back into the methionine cycle (Figure 2). To investigate the impact of its deletion on metabolism, we first assembled a panel of ten genetically diverse tissue-matched cancer cell lines, with each pair composed of one cell line characterized by homozygous deletions of CDKN2a and

MTAP (MTAP -/- ) and the other exhibiting no alterations in this chromosomal locus (MTAP +/+); the presence or absence of MTAP protein was verified by immunoblotting, while Cdkn2a mRNA expression was assessed using RNA sequencing data 201 for these cell lines (Figure 5A). Using liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS), we analyzed the levels of more than 200 metabolites between the cell lines in standard culture conditions to assess global metabolic profiles of each line. Extending on previous studies that found that MTAP status could predict differential MTA levels 93-95, we found that MTA was the most differentially abundant metabolite between the two groups, with MTAP-deleted lines exhibiting significantly higher levels of MTA compared to MTAP-expressing lines (Figure 5B).

Figure 5: MTA is the most significantly altered metabolite in MTAP-deleted cells.

(A) Cancer cell line panel of 10 lines from 5 different tissues, exhibiting either wild-type or homozygous deletion of p16/MTAP. Below, immunoblot of MTAP expression (left) and mRNA expression of Cdkn2a (right). (B) Relative MS intensity values of MTA in MTAP+/+ vs MTAP-/- cell lines. P value was obtained using Student’s t-test. (C) Heat map of top 50 differential metabolites between MTAP+/+ and MTAP-/- cell lines. Top impacted pathways are indicated below.

We next identified the top 50 differential metabolites between the MTAP +/+ and MTAP -/- lines (Table 1). When these metabolites were visualized using unsupervised hierarchical clustering, a pattern (i.e., MTAP “signature”) was evident (Figure 5C). Network-based pathway analysis of these metabolites indicated that the metabolic profiles of MTAP -/- lines differed from those of MTAP +/+ lines predominantly via differential activity of cysteine and methionine metabolism as well as glycolysis (Figure 5C).

We then performed principal components analysis (PCA) to determine the extent to which MTAP status segregated the metabolic profiles of the cell lines (Figure 6A). We found that 31

the seventh principal component (PC7) best separated the two groups, accounting for 4.3% of the overall variance (Figure 6B). Comparatively, PC3 (accounting for ~8% of the variance) best separated the cell lines by tissue type. Of note, most of the variance was linked to the individual cell line (i.e., each cell line distinctly clustered from each other). These findings indicate that while MTAP deletion does induce a reproducible metabolic shift, heterogeneity between cell lines is a larger source of variation compared to tissue type, which accounted for almost twice as much variation as MTAP status (Figure 6B).

Figure 6: MTAP status accounts for approximately 5% of metabolic variation.

(A) Principal component analysis of variance between MTAP+/+ and MTAP-/- groups. (B) Assignment of principal component that displays greatest variance between the indicated groups.

Table 1: Top 50 differential metabolites between MTAP+/+ and MTAP-/- cells

Average integrated mass spectrometry (MS) intensity values for MTAP+/+ and MTAP-/- groups are displayed, as well as the relative fold differences in intensities between the two groups.

2.2.2 Responsiveness to methionine availability is not predicted by MTAP status

In basal conditions, an MTAP-associated metabolic signature is readily apparent; however, as described in Section 1.3, metabolism has been shown to be heavily influenced by nutrient availability and other environmental factors 47,52,54. Therefore, we next sought to determine whether MTAP-deleted lines would exhibit enhanced responsiveness to nutrient availability, particularly methionine restriction. We hypothesized that under these conditions, disruption of methionine-related metabolic processes would be more pronounced in MTAP -/- due to their inability to salvage methionine from MTA. To test this, we cultured each cell line in either complete (100 µM methionine, or Met+ ) or methionine-restricted (3 µM methionine, or

Met− ) medium for 24 hours and used LC-HRMS to define the metabolic consequences of methionine restriction across the cell lines (Figure 7A). Importantly, previous work has shown that 3 µM is the lower limit of what is observed in human plasma 54, whereas 100 µM is the value in typical cell culture conditions.

Following unsupervised hierarchical clustering of both the individual samples and the fold changes of metabolites, pathway analysis demonstrated that particular metabolic pathways tended to cluster together (such as fatty acid metabolism with tryptophan and phenylalanine metabolism) as well as pathways involved in one-carbon metabolism (including serine, glycine, cysteine, and methionine metabolism); however, cell lines did not cluster by MTAP status or tissue type, demonstrating that metabolic responsiveness to methionine restriction is not defined or predicted by these factors in cell culture (Figure 7B).

MTA levels were significantly diminished in all except one (Sk-Mel-5) of the ten cell lines upon methionine restriction (Figure 7C). When we analyzed the number of significantly altered metabolites that were methionine-related (i.e., within five biochemical reactions of

methionine) or methionine-unrelated (i.e. greater than five reactions away), MTAP -/- lines were not found to exhibit a greater degree of altered methionine metabolism compared to MTAP +/+ lines upon methionine restriction (Figure 7D). Furthermore, PCA conducted for fold change values of each metabolite demonstrated that although cell lines tended to cluster by MTAP status to a greater extent than by tissue of origin, most of the variance was accounted for by cell line– to–cell line variability (Figure 7E).

Figure 7: Methionine restriction impacts global metabolism and reduces MTA levels.

(A) Experimental set-up, and validation of methionine restriction. Each dot corresponds to the integrated intensity of detected methionine in each cell line for the indicated culture conditions. (B) Heat map of fold changes (FC) in global metabolite levels upon methionine restriction, hierarchically clustered by cell lines and metabolites. Top impacted pathways are indicated. (C) FC values of MTA metabolite levels either in complete (100µM Met, or Met+) or restricted (~3µM Met, or Met-) media. (D) Fraction of significantly altered (p<0.05, Student’s t-test) metabolites that are related (i.e. within 5 reactions) or unrelated to methionine. (E) Assignment of principal component that displays greatest variance between the indicated groups.

Of particular interest was the finding that when fold change values of each metabolite were averaged across the five cell lines within each group, methionine restriction appeared to

have little to no effect on global metabolism (Figure 8A); in contrast, methionine restriction appeared to have a substantial yet highly variable effect when fold changes were analyzed in each cell line (Figure 7B,D). To further illustrate this finding, we examined alterations in key metabolites of the transsulfuration pathway, which is downstream of the methionine cycle and provides glutathione for intracellular redox reactions (Figure 2). Independent of MTAP status, some cell lines demonstrated significant alterations throughout this pathway, while others appeared to show no changes as a result of methionine restriction. Furthermore, the alterations that were induced were unexpectedly variable between cell lines; for instance, glutathione was found to increase twofold in the MTAP +/+ melanoma line UACC-257 upon methionine restriction, while it decreased by almost half in the MTAP -/- bladder line RT4 and was nonsignificantly altered in most of the other cell lines (Figure 8B). Together, these findings illustrate that the responsiveness to alterations in methionine availability is not predicted by

MTAP status and is notably heterogeneous between cell lines, and that accumulation of MTA, which is the key feature of the MTAP metabolic signature, can be abrogated by methionine restriction.

Figure 8: Responsiveness to methionine availability is heterogeneous and is not predicted by MTAP status.

(A) Volcano plot of fold change values (FC, Met-/Met+) of metabolites, averaged across either MTAP+/+ or MTAP-/- groups. P values obtained using Student’s t-test. (B) Example of heterogeneous responsiveness to methionine restriction, as indicated by FC in metabolites involved in the transsulfuration pathway.

2.2.3 Responsiveness to alterations in other one-carbon nutrient availability is largely MTAP status independent

Although MTAP status did not appear to be predictive of responsiveness to methionine availability, we further hypothesized that the MTAP-deleted lines may have adapted to the inability to salvage methionine by differentially regulating the utilization of nutrients that feed into and out of the methionine cycle. To test this, we cultured the cell lines either in serine- restricted (~16 µM from 10% serum compared to ~280 µM) medium for 24 hours or in cysteine- restricted (~6 µM from 10% serum compared to ~200 µM) medium for 6 hours and validated the reduction in levels of the respective amino acid (Figure 9A). Of note, a shorter culture time for 37

the cysteine restriction experiments was used because of the highly variable toxicity of the 24- hour cysteine restriction across the panel of cell lines independent of MTAP status (Figure 9B).

As observed with methionine restriction, cell lines did not appear to globally respond to either serine or cysteine restriction in an MTAP status–dependent manner, and the new metabolic state that arose from altered nutrient availability was independent of MTAP status (Figure 9C).

Figure 9: Responsiveness to restriction of other one-carbon nutrients is largely MTAP status-independent.

(A) Validation of serine (280µM or 16µM) and cysteine (200µM or 6µM) restriction. (B) Viability of each cell line cultured for 24 hours in media containing the respective concentration of cysteine. P values obtained using Student’s t-test. (C) Heat map of fold changes in global metabolite levels upon serine (left) and cysteine (right) restriction, hierarchically clustered by cell lines and metabolites. Top impacted pathways are indicated. 38

Serine restriction appeared to induce an increase in MTA levels in three of five MTAP +/+ lines and in one MTAP -/- line while producing no significant change in MTA levels in the other lines (Figure 10A); conversely, cysteine restriction tended to reduce MTA levels in both groups

(Figure 10B). Similar to what was found with methionine restriction, when the fold changes of metabolites were averaged across the five cell lines for each group, serine restriction was found to have no significantly reproducible and general impact on global metabolism (Figure 10C) although each cell line demonstrated responsiveness to the restriction when examined individually (Figure 10D). In contrast, averaging fold changes across the respective cell lines showed that both groups responded similarly to cysteine restriction, as evidenced by significant reductions in cysteine-related metabolites such as cysteate and oxidized glutathione (Figure 10E); however, each cell line still exhibited a unique metabolic response to the restriction as was found in the other nutrient restriction conditions (Figure 10F). Furthermore, the tissue of origin was found to account for more variance in the data than the MTAP status, although both were again found to account for substantially less of the variance compared to heterogeneity between cell lines (Figure 10G).

Figure 10: Serine or cysteine restriction has heterogeneous consequences on methionine metabolism and related metabolic pathways.

(A,B) FC values of MTA metabolite levels either in complete, (A) serine-restricted, or (B) cysteine- restricted media. (C,E) Volcano plot of FC (C, Ser-/Ser+) (E, Cys-/Cys+) values of metabolites, averaged across either MTAP+/+ or MTAP-/- groups. P values obtained using Student’s t-test. (D,F) Fraction of significantly altered (p<0.05, Student’s t-test) metabolites that are related (i.e. within 5 reactions) or unrelated to (D) serine or (F) cysteine. (G) Assignment of principal component that displays greatest variance between the indicated groups upon serine (left) or cysteine (right) restriction.

Cysteate (an intermediate in taurine biosynthesis) levels were found to be consistently depleted across all cell lines in response to cysteine restriction; however, closer examination of this pathway demonstrated that with the exception of one cell line (SW1573), taurine and hypotaurine levels tended to be altered less in MTAP -/- cell lines compared to MTAP +/+ cell lines

(Figure 11). This observation could imply that MTAP -/- cells adapt to deficiencies in recycling methionine by becoming less dependent on cysteine availability, although MTAP -/- lines on average did not show decreased cytotoxicity in cysteine-restricted conditions compared to MTAP 40

+/+ lines (Figure 9B). Together, these results show that, in terms of the factors mediating responsiveness to one-carbon nutrient availability, the resulting metabolic state is largely MTAP status independent and is better predicted by other contextual factors.

Figure 11: Responsiveness to one-carbon nutrient availability is heterogeneous.

Alterations in relative metabolite levels upon cysteine restriction in metabolites involved in taurine biosynthesis.

2.2.4 Restoration of MTAP expression has heterogeneous effects on metabolism

We next sought to determine whether re-expression of MTAP protein in MTAP-deleted cell lines would induce similar metabolic effects both on responsiveness to nutrient restriction and on overall metabolism. We ectopically expressed MTAP or a green fluorescent protein (GFP) control in four MTAP-deleted cell lines (Figure 12A) and demonstrated that the resulting protein was functional as evidenced by a reduction in MTA levels in MTAP-expressing cells compared to control cells (Figure 12B). The expression of MTAP in these cell lines resulted in markedly heterogeneous changes to both global and methionine metabolism in each of the cell lines; in fact, only one metabolite (MTA) was significantly altered in all four lines (Figure 12C).

Figure 12: Restoration of MTAP expression has heterogeneous effects on metabolism.

(A) Western blot validation of lentiviral ectopic expression of MTAP (compared to GFP control) in MTAP- /- cell lines. (B) Integrated intensity values of relative MTA metabolite levels in MTAP- or GFP-infected cell lines. (C) Volcano plots of each cell line, showing FC values of metabolites between MTAP-infected and GFP-infected isogenic pairs. P values obtained using Student’s t-test.

Ectopic MTAP expression did not significantly affect responsiveness of MTA levels to methionine or serine restriction, although cysteine restriction resulted in a significant drop in

MTA levels exclusively in MTAP -/- lines (Figure 13A). MTAP re-expression also largely did not appear to significantly alter overall responsiveness to methionine, cysteine, or serine availability compared to controls as evidenced by global changes in metabolism (Figure 13B). In addition,

MTAP re-expression also did not seem to increase the proportion of significantly altered metabolites that were related versus unrelated to the respective nutrient that was restricted (Figure

13C). These results imply that while ectopic MTAP expression in MTAP-deleted cancer cell lines can induce significant metabolic alterations, these effects are highly variable between cell lines and are not sufficient to override the programmed responsiveness to nutrient availability that has been adapted by individual cell lines.

Figure 13: Re-expression of MTAP has heterogeneous consequences on responsiveness to nutrient restriction.

(A) Validation of nutrient restriction using integrated intensity values of the respective nutrient (left) and FC values of MTA metabolite levels in complete/restricted culture conditions (right). (B) Number of significantly altered metabolites (p<0.05, Student’s t-test) in MTAP-expressing lines compared to GFP controls. (C) Fraction of significantly altered (p<0.05, Student’s t-test) metabolites that are related (i.e. within 5 reactions) or unrelated to the indicated nutrient.

2.2.5 MTAP status remains nonpredictive of responsiveness to nutrient restriction in a panel of tissue-matched cell lines

To further characterize how MTAP deletion affects one-carbon metabolism in a tissue- matched set of cell lines, we next examined a panel of seven brain-derived cancer cell lines (three

MTAP +/+ and four MTAP -/-); the presence or the absence of MTAP protein was verified by immunoblotting (Figure 14A, left), and Cdkn2a transcript levels were validated using quantitative polymerase chain reaction (qPCR) (Figure 14A, right). Of note, as a result of the relative scarcity of p16/MTAP +/+ glioma cell lines due to the high prevalence of p16 alterations in brain cancers 93, one of three MTAP -/- cell lines in this panel was determined to be Cdkn2a null (Figure 14A, right). Analysis of global metabolic profiles of these cell lines again identified MTA as the most differentially abundant metabolite (Figure 14B), with cysteine and methionine metabolism being one of the most differential metabolic pathways between the two groups (Figure 14C and Table

2). PCA demonstrated that MTAP status only accounted for slightly more of the variation (7.7% in PC5) in the metabolic profiles of these cell lines (Fig. 14D) than was found in our original panel (Figure 6B).

Figure 14: MTAP status remains nonpredictive of responsiveness to nutrient availability in a panel of tissue-matched cell lines.

(A) Western blot validation of MTAP expression (left). Cdkn2a transcript levels shown as [2^(-dCq) ± SEM] x 1E7 between Cdkn2a and 18S rRNA housekeeping control (n = 3 technical replicates per cell line). For U138MG Cdkn2a mRNA levels, n = 2 because only two replicates amplified. (B) Relative metabolite abundance of MTA in MTAP+/+ vs MTAP-/- cell lines. P values were obtained from a student’s t-test.(C) Heat map of top 50 differential metabolites between MTAP+/+ and MTAP-/- cell lines. Top impacted pathways are indicated. (D) Principal component analysis (PCA) of variance between MTAP+/+ and MTAP- /- groups.

Table 2: Top 50 differential metabolites between MTAP+/+ and MTAP-/- glioma lines.

Average integrated mass spectrometry (MS) intensity values for MTAP+/+ and MTAP-/- groups are displayed, as well as the relative fold differences in intensities between the two groups.

We then conducted methionine (Figure 15A), cysteine (Figure 15B), and serine (Figure

15C) restrictions for each cell line; consistent with our previous results, we found that these cell lines all responded to each nutrient restriction in a unique and heterogeneous manner and that these metabolic shifts were not primarily associated with MTAP status as evidenced by the clustering patterns of the resulting metabolic profiles (Figure 15A-C). Interestingly, compared to our previous cell line panel we found that methylated metabolites were more consistently altered by methionine restriction in these tissue-matched cell lines, independent of MTAP status (Figure

15D); furthermore, MTAP +/+ cell lines exhibited less variable responsiveness and greater sensitivity to cysteine restriction (Figure 15E), supporting our previous observations that MTAP -/- cell lines tend to exhibit relatively less sensitivity to cysteine restriction within cysteine-related metabolic pathways (Figure 11). Additionally, cell lines did not demonstrate any similarity in responsiveness to serine restriction (Figure 15F) as was found in our previous cell line panel

(Figure 10C).

Figure 15: MTAP status remains non-predictive of responsiveness to nutrient availability in a tissue-matched panel of cell lines.

(A-C) Relative metabolite abundance of (A) methionine, (B) cysteine, and (C) serine and the resulting heat maps of global metabolic profiles generated in each cell line from the indicated nutrient restriction. (D-F) Volcano plots of log-transformed fold change values of metabolites upon (D) methionine, (E) cysteine, or (F) serine restriction, averaged across either MTAP+/+ or MTAP-/- groups. P values obtained using Student’s t-test.

Methionine restriction effectively reduced MTA levels in MTAP -/- cells to those found in

MTAP +/+ cells in methionine-rich culture conditions (Figure 16A), while cysteine restriction 48

significantly reduced MTA levels solely in MTAP +/+ cells (Figure 16B) and serine restriction did not significantly alter MTA levels in any cell line (Figure 16C). However, we found that the fraction of methionine-related metabolites (Figure 16D), as well as cysteine-related (Figure 16E) or serine-related (Figure 16F) metabolites, that were significantly altered by the respective nutrient restriction did not differ between MTAP +/+ and MTAP -/- groups. Lastly, PCA demonstrated that methionine restriction continued to account for more of the variance in global metabolic profiles compared to MTAP status (Figure 16G). Together, the results in this tissue matched panel of brain-derived cell lines are in line with those found in the previous panel of tissue-variable cell lines and illustrate that MTAP status is insufficient to shape how a given cell line will respond to alterations in the environment.

Figure 16: Cell heterogeneity continues to dominate MTAP status in predictiveness of metabolic responsiveness to restriction of one-carbon nutrients. 49

(A-C) Relative metabolite abundance of MTA in cell lines cultured in complete or (A) methionine-, (B) cysteine-, or (C) serine-restricted media. (D-F) Fraction of significantly altered (p<0.05, Student’s t-test) metabolites that are related (i.e. within 5 reactions) or unrelated to (D) methionine, (E) cysteine, or (F) serine. (G) Assignment of principal component that displays greatest variance between the indicated groups upon methionine restriction.

Of note, while linoleic acid metabolism was found to be significantly affected both between MTAP +/+ and MTAP -/- groups (Figure 5C) and upon each of the nutrient restrictions

(Figure 7B and 9C) in our panel of ten tissue-diverse cell lines, this pathway was not found to be significantly affected upon nutrient restriction in this panel of brain-derived cell lines; upon further examination, we found that this was specifically due to alterations in linoleic acid (which is the central metabolite within this metabolic process and thus is highly weighted in the pathway analysis), while the other metabolites in this pathway were not significantly affected. Although the functional significance of this observation is presently unclear, this may represent a tissue- specific phenotype and provides an interesting avenue for future investigations of gene- environment interactions in the context of fatty acid metabolism.

2.2.6 Defining the quantitative impact of MTAP deletion and environmental factors on metabolism

Our results thus far reveal that heterogeneity across cell types and nutrient availability exert dominant effects on metabolism relative to MTAP status. For example, each cell line exhibited marked differences in the number of altered metabolites in response to nutrient restriction independent of MTAP status or tissue identity (Figure 17A). Furthermore, when the numbers of unique significantly altered metabolites were averaged within MTAP +/+ and MTAP -/- groups, we found that the two groups displayed similar degrees of responsivity to methionine restriction, while the MTAP +/+ cell lines were, on average, more responsive to serine and cysteine restriction (Figure 17B).

As cancer cell lines are known to individually exhibit a unique combination of both genetic and environmentally driven transcriptional programs, and that previous studies suggest that MTAP deletion may elicit epigenetic consequences 93,95, we next sought to determine whether the MTAP-deleted cell lines included in our panel exhibit differential expression of other enzymes involved in one-carbon metabolism compared to MTAP-expressing cell lines. Correlation analysis of global RNA sequencing data 201 showed that these cell lines did not cluster by either MTAP status or tissue type, consistent with our metabolomics analysis (Figure 17C); additionally, although we correctly predicted that the expression of one-carbon metabolic enzymes is highly variable between cell lines (Figure 17D), this variability was comparable within MTAP +/+ and

MTAP -/- groups and is thus not likely to be the determinant of the observed metabolic differences.

Figure 17: Integration of responsiveness to nutrient restriction determines quantitative impact of MTAP deletion on global metabolic networks.

(A) Number of significantly altered (p<0.05, Student’s t-test) metabolites for each cell line under each of the three nutrient restriction conditions. (B) Venn diagrams depicting overlap of average number of significantly altered related metabolites within a cell line for MTAP+/+ and MTAP-/- groups. (C) Spearman correlations comparing global RNA sequencing datasets for each cell line. (D) Normalized levels of transcripts per million (TPM) of metabolic enzymes involved in methionine metabolism or related metabolic pathways. TPM was normalized to the average TPM value (averaged across all 10 cell lines) for the indicated transcript.

Thus, these collective results demonstrate that while MTAP deletion does confer a reproducible metabolic signature within a single environment, the availability of nutrients related to MTAP (i.e., methionine, serine, and cysteine) has a larger effect both on MTAP-related

methionine metabolism and on the global metabolic network, which is further confounded by metabolic variation across cell lines that span different tissues of origin (Figure 18).

Figure 18: Graphical representation of relative impact of environmental and genetic factors on the metabolome.

2.3 Discussion

Cells have been thought to exhibit similar metabolic phenotypes based on their genetic status 202-205. In line with this, it has recently been proposed that MTAP deletion creates a state of disordered methionine metabolism resulting in targetable liabilities, in both metabolism and the epigenetic regulation that is connected to methionine metabolism 54,93-95,206. However, while the proposed liabilities induced by MTAP deletion may indeed be promising, our study reveals that methionine metabolism can be largely shaped by environmental factors thus supporting the conclusion that contextual factors beyond a single genetic event should be considered when characterizing cellular metabolism. Using MTAP status and the pathway its enzyme resides in

(i.e. methionine and by extension one-carbon metabolism) as a model to quantitatively and systematically study the relevant variables that shape metabolic outcomes, we found that while perhaps expectedly global metabolism was not overall predictive of MTAP status, methionine

metabolism as well was also shaped to a greater extent by the nutrients that are used in the one- carbon network and by cell-to-cell variability.

Our results demonstrate that while metabolic signatures of MTAP status across diverse cell types are reproducible in nutrient-rich culture media, these signatures are lost upon consideration of nutrient availability. For instance, we found that the accumulation of MTA, which was essential for mediating the collateral lethality induced by MTAP deletion 93-95, was abrogated when cells were subjected to methionine or cysteine restriction. Furthermore, metabolic variation in one-carbon metabolism across cell types was far greater than the differences observed in a pan tissue analysis of MTAP wild-type and homozygous deleted cells; surprisingly, each cell type exhibited dramatically different metabolic outcomes upon nutrient restriction which, importantly, were not predicted by MTAP status. Additionally, re-expression of functional MTAP protein produced extremely heterogeneous consequences on various metabolic pathways beyond methionine metabolism across a diverse panel of cell lines. Nevertheless, these results by no means exclude the possibility of targeting methionine metabolism (or MTAP-deleted cancers) by for example inhibiting MAT2a, but instead further illustrate the complexities of gene- environment interactions that may influence the efficacy of therapeutic approaches in highly variable patient populations.

Considering that many studies of metabolism are conducted in nutrient-rich media designed for optimal cell proliferation or protein production (which can exhibit methionine concentrations up to three-fold higher than of those found in human plasma) and that tissues exhibit significantly variable metabolic requirements and nutrient availabilities in a physiological setting 52, our results contribute to our understanding of the limitations inherent to current experimental approaches using cell culture; it will therefore be essential in future investigations to consider how dietary nutrient composition may influence genetically-driven phenotypes,

particularly in a physiological setting. Other examples of collateral lethality that depend on genetically-defined altered metabolisms in general, such as the case of malic enzyme or enolase

207,208, may likely encounter similar considerations. Together, these findings indicate greater complexities in the factors that shape metabolism beyond individual genetic alterations.

2.4 Materials and Methods

2.4.1 Cell culture and reagents

Cells were cultured at 37°C, with 5% atmospheric CO2 in RPMI 1640 (GIBCO), 10% heat-inactivated fetal bovine serum (FBS), penicillin (100 U/ml), and streptomycin (100 mg/ml).

All cell lines were obtained from the American Tissue Culture Collection except UACC-257

(obtained from the National Cancer Institute, National Institutes of Health), U373MG (obtained from the Duke Cell Culture Facility), and JHOS2 (obtained from A. Berchuck, Duke University).

2.4.2 Nutrient restriction experiments

For all nutrient restriction experiments, cells were plated at a density of 3.0 × 105 cells per well in triplicate in a six-well plate and were allowed to adhere for 24 hours. Conditional medium was prepared using RPMI 1640 lacking amino acids, glucose, and glutamine (GIBCO).

Dropout medium (i.e., Met− , Ser− , or Cys− ) was supplemented with 5 mM glucose, 2 mM L- glutamine, 10% heat-inactivated FBS, penicillin (100 U/ml), streptomycin (100 mg/ml), and 19 of

20 individual amino acids (excluding either methionine, serine, or cystine) at concentrations found in full RPMI 1640 medium. For full media conditions in these experiments, the respective nutrient of interest was individually added back to the medium (i.e., Met+ , Ser+ , or Cys+ ). For all experiments, three technical replicates per culture condition for each cell line were used.

2.4.3 Cell viability assays

For all cell viability measurements, cells were plated at a density of 5.0 × 103 cells per well in triplicate in a 96-well plate and were allowed to adhere in full RPMI 1640 medium for 24 hours. Conditional medium, containing five increasing concentrations of cystine was prepared as described above. After 24 hours, cells were briefly washed with phosphate-buffered saline (PBS) and then incubated in the conditional medium for 24 hours. After 24 hours, the medium was aspirated and replaced with 100 µl phenol red–free RPMI 1640 (GIBCO) and 12 mM 3-[4,5- dimethylthiazol-2-yl]-2,5-diphenyltetrazolium (MTT) (M6494, Thermo Fisher Scientific). After 4 hours, the MTT-containing medium was aspirated, and 50-µl dimethyl sulfoxide was added to dissolve the formazan. After 5 minutes, absorbance was read at 540 nm.

2.4.4 Lentiviral transfection and transduction for ectopic MTAP expression

Human embryonic kidney 293T cells were plated at a density of 1.0 × 106 cells/10 cm plate in RPMI 1640 (GIBCO) supplemented with 10% heat-inactivated FBS, penicillin (100

U/ml), and streptomycin (100 mg/ml) and were allowed to adhere and reach 70% confluency. 15

µg of MTAP (EX-A3221-Lv105, GeneCopoeia) or GFP control (EX-EGFP-Lv105,

GeneCopoeia) plasmid, 10 µg of PsPAX2 packaging vector (no. 12260, Addgene), and 5 µg of

PMD2. G envelope– expressing plasmids (no. 12259, Addgene) were diluted in 500 µl of jetPRIME buffer (no. 114-07, Polyplus-transfection) and vortexed. Next, 60 µl of the jetPRIME transfection reagent (no. 114-07, Polyplus-transfection) was added to the mixture, vortexed for

10 seconds, and left to incubate for 10 minutes at room temperature. The medium in the plate was replaced with fresh medium, and the transfection mix was then added to the 10-cm plate dropwise. After 24 hours, the transfection medium was replaced with fresh RPMI 1640 medium.

After an additional 24 hours, the medium was collected and filtered through a 0.45-µm filter for

virus collection. SW1573, U118MG, JHOS2, and Sk-Mel-5 cells were plated in 10-cm plates, and when they reached 30-50% confluency, virus-containing medium (1:1 with fresh RPMI 1640 medium) was added to the plates along with 4 µg/µl polybrene (#107689, Sigma Aldrich). After

24 hours, the virus-containing medium was removed and replaced with fresh RPMI 1640 medium. Cells were incubated with 1 µg/ml puromycin for 48 hours, and MTAP expression in the four isogenic cell pairs (eight lines total) were verified by immunoblotting.

2.4.5 Immunoblotting methods

Samples were homogenized in 100 µl of 1x radioimmunoprecipitation assay (RIPA) buffer (VWR International) supplemented with 100 µM phenylmethylsulfonyl fluoride (Sigma

Aldrich), 2 µg/µl aprotinin (Sigma Aldrich), 1x phosphatase inhibitor cocktail (Roche), and 2 mM dithiothreitol (Sigma Aldrich). Cell lysates were centrifuged at 14,000 rpm for 30 minutes at

4°C. The resulting supernatant was transferred to a clean tube, and a bicinchoninic acid assay

(Thermo Fisher Scientific) was carried out to quantify protein concentration. Protein samples were loaded onto TGX Stain-Free precast gels (Bio-Rad) and transferred to polyvinylidene difluoride membranes. Membranes were blocked in 5% dry nonfat milk in either tris-buffered or phosphate-buffered saline with 0.1% Tween-20 (TBST or PBST, respectively) and incubated in anti-MTAP [1:7500 (Figure 5A and 12A) or 1:1333 (Figure 14A) in 5% BSA in TBST; no.

4158S, Cell Signaling] and anti-actin [1:2000 (Figure 5A and 12A) or 1:1000 (Figure 14A) in 5% dry nonfat milk in TBST/ PBST; no. MA5-15739, Thermo Fisher Scientific] overnight at 4°C.

Membranes were washed 3 times for 10 minutes (Figure 5A and 12A) or 5 minutes (Figure 14A) each with TBST/PBST at room temperature with gentle shaking. Horseradish peroxidase- conjugated anti-mouse (no. 610-1302, Rockland) and anti-rabbit (no. 611-1302, Rockland), both

1:2000 (Figure 5A and 12A) or 1:3000 (Figure 14A), were used as secondary antibodies in 5% dry nonfat milk in TBST/PBST at room temperature for 20 minutes (Figure 5A and 12A) or 57

60 minutes (Figure 14A). Membranes were washed 3 times in TBST/PBST once more, and then chemiluminescent signals were detected with the Clarity Western ECL Detection Kit (no.

1705061, Bio-Rad) (Figure 5A and 12A) or SuperSignal West Pico PLUS Chemiluminescent

Substrate (no. 34579, Thermo Fisher Scientific) (Figure 14A). The membranes were then imaged using the ChemiDoc Touch Imaging System (Bio-Rad).

2.4.6 Quantitative PCR methods

Confluent 10-cm plates of each cell line were washed with ice-cold PBS, and 1 ml of

TRIzol (no. 15596026, Amibon by Life Technologies) was added to each plate. Cells were scraped into the TRIzol solution and transferred to tubes. 200 µl chloroform (C2432, Sigma

Aldrich) was then added to each tube, and samples were briefly vortexed and then incubated at room temperature for 3 minutes. Samples were then centrifuged at 12,000g at 4°C for 15 minutes.

The clear upper aqueous layer was transferred to a fresh tube, and 500-µl isopropanol was added to precipitate RNA. The samples were incubated at room temperature for 10 minutes and then centrifuged at 12,000g at 4°C for 10 minutes. The supernatant was removed, and the RNA pellet was washed twice with 1 ml of 75% ethanol; between and after washes, samples were vortexed and then centrifuged at 7500g at 4°C for 5 minutes. The resulting RNA pellets were then airdried for 10 minutes and resuspended in 100 µl of nuclease-free water. RNA quantity/quality was assessed by NanoDrop Spectrophotometer measurements using a Bioanalyzer 2100 (Agilent

Technologies). One microgram of total RNA was then converted to single-strand complementary

DNA using the Reverse Transcription System (no. A3500, Promega). Expression levels of

Cdkn2a, MTAP, and 18-S transcripts were assessed by qPCR on a CFX384 Touch Real-Time

PCR Detection System (Bio-Rad) using iTaq Universal SYBR Green (Bio-Rad) and the following primers (Integrated DNA Technologies):

Cdkn2a-forward: (5′-TGTGCCACACATCTTTGACCT-3′) 58

Cdkn2a-reverse: (5′-AGGACCTTCGGTGACTGATGA-3′)

MTAP-forward: (5′-TGGAATAATTGGTGGAACAGGC-3′)

MTAP-reverse: (5′-TGGCACACTCCTCTGGCAC-3′)

18-S–forward: (5′-CTTAGAGGGACAAGTGGCG-3′)

18-S–reverse: (5′-ACGCTGAGCCAGTCAGTGTA-3′)

Samples were run in three technical replicates. The difference between the highest and lowest quantification cycles (dCq) values for Cdkn2a and MTAP were calculated as follows:

2−[Cq(Cdkn2a or MTAP) − Cq(18S)]. Data were plotted as (mean dCq ± SEM) × 107.

2.4.7 Metabolite extraction

Medium was quickly aspirated, and 1 ml of extraction solvent (80% methanol/water, cooled to −80°C) was added to each well of the 6-well plates which were then transferred to

−80°C for 15 minutes. Plates were removed and cells were scraped into the extraction solvent on dry ice, and the supernatant was transferred to Eppendorf tubes. All metabolite extracts were centrifuged at 20,000g at 4°C for 10 minutes. The resulting supernatant for each sample was split into two equal volumes in Eppendorf tubes and evaporated in a speed vacuum. The resulting pellets were stored in −80°C until resuspension. For polar metabolite analysis, the cell extract was dissolved in 15 µl of water and 15 µl of methanol/acetonitrile (1:1, v/v) [liquid chromatography– mass spectrometry (LC-MS) optima grade, Thermo Fisher Scientific]. Samples were centrifuged at 20,000g for 2 minutes at 4°C, and the supernatants were transferred to LC vials.

2.4.8 Liquid chromatography

UltiMate 3000 HPLC (Dionex) with an XBridge Amide column [100 mm × 2.1 mm inner diameter (i.d.), 3.5 µm; Waters] was coupled to a Q Exactive (QE) mass spectrometer (Thermo

Fisher Scientific) for metabolite separation and detection at room temperature. The mobile phase

A reagent was composed of 20 mM ammonium acetate and 15 mM ammonium hydroxide in 3% acetonitrile in high-performance liquid chromatography (HPLC)–grade water (pH 9.0), while the mobile phase B reagent was acetonitrile. All solvents were of LC-MS grade, purchased from

Thermo Fisher Scientific. The flow rate used was 0.15 ml/min from 0 to 10 min and 15 to 20 min, and 0.3 ml/min from 10.5 to 14.5 min. The linear gradient was as follows: 0 min 85% B, 1.5 min

85% B, 5.5 min 35% B, 10 min 35% B, 10.5 min 25% B, 14.5 min 35% B, 15 min 85% B, and

20 min 85% B.

2.4.9 Mass spectrometry

The QE mass spectrometer (Thermo Fisher Scientific) was outfitted with a heated electrospray ionization probe (HESI) with the following parameters: evaporation temperature,

120°C; sheath gas, 30; auxiliary gas, 10; sweep gas, 3; and spray voltage, 3.6 kV for positive mode and 2.5 kV for negative mode. Capillary temperature was set at 320°C, and S-lens was 55.

A full scan range was set at 60 to 900 mass/charge ratio (m/z) with the resolution set to 70,000.

The maximum injection time was 200 ms. Automated gain control was targeted at 3 million ions.

2.4.10 Peak extraction and metabolomics data analysis

Data collected from LC-QE mass spectrometer was processed using commercially available software Sieve 2.0 (Thermo Fisher Scientific). For targeted metabolite analysis, the method “peak alignment and frame extraction” was applied. An input file (“frame seed”) of theoretical m/z (width set at 10 ppm) and retention time of ~260 known metabolites were used for positive mode analysis, while a separate frame seed file of ~200 metabolites was used for negative mode analysis. To calculate the fold changes between different experimental groups, integrated peak intensities generated from the raw data were used. Hierarchical clustering and heat maps were generated using Morpheus software (Broad Institute, https://software.broadinstitute.org/morpheus/). For hierarchical clustering, Spearman correlation 60

parameters were implemented for row and column parameters. Pathway enrichment analysis was conducted using MetaboAnalyst 3.0 software (www.metaboanalyst.ca/faces/home.xhtml); briefly, metabolite identifications from the human metabolome database (HMDB IDs) from the metabolites that were significantly enriched [greater than log2 (fold change) with p < 0.05] were inputted. The pathway library used was Homo sapiens and Fisher’s exact test was used for over- representation analysis. Other quantitation and statistics were calculated using GraphPad Prism software.

2.4.11 Statistical analysis

For the control data from the three nutrient restriction experiments, raw intensities of metabolites were log-transformed (log(x+1)), corrected for batch effect using the removeBatchEffect() function in the LIMMA R package 209, and then z-score normalized. Other metabolomic data and fold changes between conditions were also z-score normalized before the following analysis. PCA was performed using the PCA() function from the FactoMineR R package 210. Correlation (r2) between each PC and each of the three categorical variables (MTAP status, Cell Line, Tissue of Origin) was also calculated by the PCA() function from FactoMineR.

ANOVA was done on each of the three categorical variables and the first 10 PCs using the aov() function in R. Unpaired student’s t-test was used for comparison between two groups unless otherwise stated. Unpaired Student’s t test was used for comparison between two groups unless otherwise stated. p < 0.05 was considered statistically significant.

2.4.12 Network analysis

The genome-scale reconstruction of human metabolism, Recon2 211, was used as the metabolic network model in defining the nutrient-related and nutrient-unrelated metabolites.

Metabolites that were five or fewer reactions away from the nutrient (methionine, serine, or cysteine) were categorized as “related”; otherwise, they were categorized as “unrelated.” Only 61

enzyme-catalyzed reactions with the highest confidence score were considered. Cofactors associated with more than 100 reactions were also excluded from the analysis.

3. Methionine restriction synergizes with 5-FU therapy by disrupting nucleotide metabolism and redox balance vi

3.1 Background and Context

The previous chapter demonstrated that nutrient composition in growth media can have marked effects on cancer cell metabolism. However, the extent to which diet, through its influence on levels of circulating metabolites, alters metabolic pathways in tumors and affects therapeutic outcomes is largely unknown. As discussed in Section 1.3, the availability of certain amino acids have been shown to modulate cancer outcome 48-50,212,213, although it remains to be explored how these dietary alterations broadly impact metabolism or whether they exert targeted effects on nutrient-specific processes.

As discussed in detail in Section 1.4, methionine is essential for the regulation of a multitude of diverse cellular functions as a result of its location in one-carbon metabolism, and is one of the most variable metabolites found in human plasma 54 (Section 1.4.1 and 1.4.2). As many cancer cells have been shown to exhibit enhanced methionine dependence 102,214,215, our group reasoned that dietary methionine restriction (MR) could have broad antineoplastic effects by targeting specific metabolic processes. Furthermore, given the role of one-carbon metabolism in redox balance and nucleotide metabolism (sections 1.1.2 and 1.4.2), in collaboration with Dr.

David Hsu’s laboratory we investigated two patient-derived xenograft (PDX) models of

vi This chapter was partially adapted and modified from published work: Gao X., Sanderson S.M.*, Dai Z.*, Reid M.A., Cooper D.E., Lu M., Richie J.P., Ciccarella A., Calcagnotto A., Mikhael P.G., Mentch S.J., Liu J., Ables G., Kirsch D.G., Hsu D.S., Nichenametla S.N., and Locasale J.W. “Dietary methionine influences therapy in mouse cancer models and alters human metabolism.” Nature (2019). This text was reproduced in accordance with the CC-BY license. Author contributions to the adapted content of this chapter: Colorectal PDX models, X.G., M.L., D.E.C., and G.A.; Metabolomics and data analysis, X.G.; Growth curves, S.M.S.; Metabolite rescue experiments, S.M.S.; Isotope tracing, M.A.R. All text in this chapter was written by S.M.S. * = equal contribution. 63

colorectal cancer to test the hypothesis that MR could synergize with frontline cancer chemotherapies that also target these processes such as 5-fluorouracil (5-FU).

3.2 Results

3.2.1 MR exhibits anti-cancer properties and synergizes with 5-FU in vivo

My colleagues first considered two RAS-driven colorectal PDX models bearing a

KRASG12A (CRC119) or NRASQ61K (CRC240) mutation. Mice were subjected to a control (0.86% methionine w/w) or methionine-restricted (MR) diet (0.12% methionine w/w) when the tumor was palpable (treatment) or two weeks prior to inoculation (prevention) (Figure 19A). MR was found to significantly inhibited tumor growth in CRC119 and showed an effect in CRC240

(Figure 19B, left); furthermore, these antineoplastic effects were significantly more pronounced in both groups when the MR diet was initiated two weeks prior to tumor engraftment, suggesting

MR may exhibit preventative effects (Figure 19B, right). We then examined the resulting metabolic profiles of tumors and found that MR altered methionine and sulfur-related metabolism in both groups (Figure 19C), which was additionally replicated when subjecting the primary

CRC119 and CRC240 cells were cultured in methionine-restricted medium (Figure 19D).

Figure 19: MR inhibits tumor growth in PDX models of colorectal cancer.

(A) Schematic of experimental design using colorectal PDXs. Treatment, n = 7 mice per group (4 female and 3 male). Prevention, n = 8 mice per group (4 female and 4 male). P1, passage 1; P2, passage 2; P3, passage 3; P4, passage 4. (B) Tumor growth curve and images of tumors at the end point from (A). Mean +/- s.e.m., * p<0.05, Student’s t-test. (C) Volcano plots of metabolites in tumors from prevention study. P values were obtained using Student’s t-test. (D) Volcano plots of metabolites in primary cells cultured in control (100 µM) or methionine-restricted (0 µM) media for 24 hours. P values were obtained using Student’s t-test.

As discussed in Section 1.5.1, 5-FU targets thymidylate synthase (Figure 4) and is a frontline chemotherapy for colorectal cancer with therapeutic strategies achieving modest (~60-

65%) responses 25,26. We therefore tested whether MR could synergize with 5-FU in CRC119

PDX tumors, in which we fed mice either a control or MR diet and delivered a tolerable low dose of 5-FU that alone showed no effect on tumor growth (Figure 20A,B). My colleagues found that

MR synergized with 5-FU treatment, leading to a marked inhibition of tumor growth (Figure

20B) and broad metabolic effects with the most prominent changes related to nucleotide metabolism and redox state (Figure 20C), consistent with the mechanistic actions of both 5-FU and MR. Together, these results demonstrate that dietary MR can exert antineoplastic effects

alone as well as in synergy with 5-FU in CRC PDX tumors, with a substantial disruption to nucleotide metabolism and redox balance.

Figure 20: MR synergizes with 5-FU in PDX models of colorectal cancer.

(A) Schematic of experimental design. (B) Tumor growth curves, quantification and images at the end point. Mean +/- s.e.m., * p<0.05 using Student’s t-test. N=8 mice per group (4 female and 4 male). (C) Volcano plots of fold changes (FC) in metabolites in tumors. P values were obtained using Student’s t-test.

3.2.2 MR as a single agent and in combination with 5-FU exerts cytotoxic effects through disruption of redox and nucleotide metabolism

To gain more insight into the metabolic effects exerted by MR alone and with 5-FU, we first verified that we could reproduce the cytotoxic effects of MR alone and in synergy with 5-FU

(Figure 21A), and then supplemented nutrients related to one-carbon and nucleotide metabolism to primary CRC119 (Figure 21B) and HCT116 (Figure 21C) colorectal cells in the presence of control or MR media, treated with either vehicle or with 5-FU. We found that nucleosides and the

antioxidant N-acetyl-cysteine (NAC), alone and in combination with related supplements such as homocysteine and the folate vitamin B12, partially alleviated the inhibition of cell proliferation due to MR or MR plus 5-FU in CRC119 cells (Figure 21B), which was largely replicated in

HCT116 cells (Figure 21C).

Figure 21: Supplementation of nucleosides and antioxidants is partially cytoprotective against MR and 5-FU.

(A) The synergistic cytotoxic effects of MR and 5-FU in primary CRC119 (left) and HCT116 (right) colorectal cells, evaluated by cell counting. Mean +/- s.e.m., n = 3 biological replicates. * p < 0.05 using Student’s t-test. (B,C) Effect of nutrient supplementation on MR alone or with 5-FU-inhibited cell 67

proliferation in (B) CRC119 or (C) HCT116 colorectal cells.. Mean +/- s.e.m., n = 9 replicates from 3 independent experiments. * p < 0.05 vs control, ^ p < 0.05 vs MR, # p < 0.05 vs 5-FU, † p < 0.05 vs MR + 5-FU using Student’s t-test.

Finally, my colleagues conducted [U-13C]-serine tracing to determine the extent to which folate-mediated nucleotide metabolism was impacted (Figure 22A). MR and 5-FU (both alone and together) led to a significant reduction in labeling of dTTP, with a concomitant increase in labeling of methionine when MR and 5-FU were combined (Figure 22B). These results suggest that the synergistic effect of MR and 5-FU is at least partially due to a compensatory increase in methionine synthesis, thereby diverting dTMP synthesis by competing for the serine-derived one- carbon unit 5,10-methylene-tetrahydrofolate (Figure 2 and Figure 22A).

Figure 22: Disruption to nucleotide synthesis is one mechanism by which MR and 5-FU synergize.

(A) U-13C-serine tracing for nucleotide and methionine synthesis. (B) Mass intensity for [M+1] dTTP and [M+1] methionine in CRC119 cells. Mean +/- s.d., n = 3 replicates. * p < 0.05 vs control, # p < 0.05 vs MR by two-tailed Student’s t-test.

3.3 Discussionvii

Considering the multifaceted functions of methionine (Figure 3), the anti-tumor effects of

MR are likely to be manifold. One hypothesis for this dependence on exogenous methionine is an increased reliance on transmethylation reactions 102,215, although neither SAM nor SAH were significantly altered by MR in several in vivo settings 216. However, the epigenetic consequences of dietary MR within tumors remain to be fully elucidated and may indeed be contributing to cancer in some contexts. It is also possible that altered protein synthesis could play a role in the observed antineoplastic phenotype, but studies directly examining this possibility are lacking. As

MR-mediated extension of lifespan appears to be autophagy-dependent in both yeast and progeroid mouse models 65,69, these metabolic differences may be due to differential activation of autophagic processes. Additionally, it is possible that nutrient-sensing signaling pathways might also play a part in shaping metabolic responsiveness to MR 58,73,217,218. This is supported by the finding that hydrogen sulfide (H2S) production mediates the stress resistance phenotype imparted by methionine restriction in hepatic tissue, which was abrogated by mTORC1 hyperactivation 62.

The mechanisms driving the anti-tumor effects of MR are also likely dependent on the contextual factors (i.e. gene-environment interactions) that shape individual tumors. For example, the metabolic dependencies of KRAS-driven tumors have been previously shown to be dependent on the originating tissue 52. Therefore, as an extension of this concept, it is possible that MR may primarily exert antineoplastic effects via differential mechanisms which can be dependent on both intrinsic and extrinsic factors that are specific to individual tumors. For instance, PI3K mutations have been shown to enhance the methionine dependence phenotype via differential activity of the

vii This chapter subsection was adapted and modified from a published review: Sanderson S.M., Gao X., Dai Z., and Locasale J.W. “Methionine Metabolism in Health and Cancer: a Nexus of Diet and Precision Medicine.” Nature Reviews Cancer (Epub Sept. 12, 2019). This text was reproduced in accordance with the CC-BY license. All text included in this subsection was written by S.M.S. 69

cystine-glutamate antiporter219, suggesting that this subset of tumors may exhibit sensitivity to

MR due to the subsequent alterations in the availability of other amino acids. Interestingly, we observed that MR achieved a stronger inhibitory effect on tumor growth when given two weeks prior to tumor engraftment in a colorectal PDX model compared with dietary interventions initiated only after tumor formation (Figure 19B) suggesting a potential protective effect on host physiology as well. These considerations illuminate the degree to which our mechanistic understanding of methionine dependence in cancer is currently lacking, but the results included in our study demonstrating the significant MR-mediated disruption to intratumoral nucleotide metabolism and redox balance provide further characterization and support for the therapeutic potential of dietary intervention.

3.4 Materials and Methods

3.4.1 Animals, diets, and tissue collection

All animal procedures and studies were approved by the Institutional Animal Care and

Use Committee (IACUC) at Duke University. All experiments were performed in accordance with relevant guidelines and regulations. All mice were housed at 20 ± 2°C with 50 ± 10% relative humidity and a standard 12-hour dark–12-hour light cycle. The special diets with defined methionine levels that have previously been used 54,71 were purchased from Research Diets; the control diet contained 0.86% methionine (w/w, catalogue no. A11051302) and methionine- restricted diet contained 0.12% methionine (w/w, catalogue no. A11051301). For all animal studies, mice were randomized to the control or methionine-restricted diet, and investigators were not blinded to allocation during experiments or outcome assessment.

3.4.2 PDX models of colorectal cancer

PDX models of colorectal cancer with liver metastasis were developed as previously described 220,221 under an IRB-approved protocol (Pro00002435). In brief, CRC119 and CRC240 tumors were resected, washed and minced, and then passaged through JAX NOD.CB17-

PrkdcSCID-J mice 2–5 times. For the dietary studies, CRC119 and CRC240 PDX tumors were minced in PBS at 150 mg/ml and 200 μl of tumor suspension was subcutaneously injected into the flanks of NOD.Cg-PrkdcscidIl2rgtm1Wjl/SzJ mice from the Jackson Laboratory. Mice (four female and three or four male) were subjected to the control or methionine-restricted diet, either two weeks before the tumor injection or from when the tumor was palpable until the end point (a tumor volume of about 1,500 mm3). Tumor size was monitored two to three times per week until the end point. For the combination therapy with the standard chemotherapy drug 5-FU, mice were subjected to the control or the methionine-restricted diet from two weeks before the tumor injection until the end point. When tumors were palpable, mice (four female and four male) were randomized to treatment of 5-FU (NDC 63323-117-10, 12.5 mg/kg three times per week) or vehicle (saline) through intraperitoneal injection. To minimize toxicity, an established low dose of 5-FU was used 222. Tumor size was monitored two to three times per week until the end point.

3.4.3 Colorectal cancer cell lines

Early-passage primary CRC119 and CRC240 colorectal cancer cell lines were developed from the PDXs. PDXs were collected and homogenized, and the homogenates were grown in

RPMI 1640 medium (GIBCO) with addition of 10% fetal bovine serum (FBS), 100,000 U/l penicillin and 100 mg/l streptomycin at 5% CO2. A single-cell clone was isolated using an O ring. The HCT116 cell line was a gift from the laboratory of L. Cantley, and was maintained in

RPMI 1640 supplemented with 10% FBS and 100,000 U/l penicillin and 100 mg/l streptomycin.

Cells were grown at 37°C with 5% CO2. Cell lines were authenticated and tested for mycoplasma 71

at the Duke University DNA Analysis Facility by analyzing DNA samples from each cell lines for polymorphic short tandem repeat markers using the GenePrint 10 kit (Promega). All cell lines were negative for mycoplasma contamination.

3.4.4 Metabolite rescue experiments

10 μM methionine (approximately 10% of the concentration used in complete RPMI

1640 medium) was used to model dietary methionine restriction. CRC119, CRC240, and

HCT116 cells were plated at a density of 5.0 × 103 cells per well in triplicate in a 96-well plate, and were allowed to adhere in full RPMI 1640 medium (GIBCO) for 24 hours. Medium was then aspirated, upon which the cells were briefly washed with PBS and 100 μl conditional media containing either 10 μM or 100 μM methionine, 2.5 μM 5-FU (F6627, Sigma Aldrich) or vehicle

(DMSO), as well as the indicated nutrient supplementations was added to each well. The following concentrations of nutrients were used: 400 μM homocysteine (H6010, Sigma Aldrich),

20 μM vitamin B12 (V2876, Sigma Aldrich), 1x Embryomax nucleosides (ES-008-D, Millipore),

1mM choline (C7017, Sigma Aldrich)), 0.5mM formate (67253, Sigma Aldrich), 1mM NAC

(A9165, Sigma Aldrich) and 2.5 μM 5-FU. After 48 hours, the medium was aspirated and replaced with 100 µl phenol red–free RPMI 1640 (GIBCO) and 12 mM 3-[4,5-dimethylthiazol-2- yl]-2,5-diphenyltetrazolium (MTT) (M6494, Thermo Fisher Scientific). After 4 hours, the MTT- containing medium was aspirated, and 50-µl dimethyl sulfoxide was added to dissolve the formazan. After 5 minutes, absorbance was read at 540 nm.

3.4.5 Metabolite profiling and isotope tracing

PDX primary cell lines were plated in 6-well plates at a density of 2.0 × 105 cells per well. For overall polar metabolite profile, after overnight incubation cells were washed once with

PBS and cultured for an additional 24 hours with 2 ml of conditional RPMI medium containing 3

μM or 100 μM methionine plus 10% FBS. Cellular metabolites were extracted after incubation.

For U-13C-serine isotope tracing, both primary CRC119 cells and HCT116 cells were seeded in 6- well plates at a density of 2.0 × 105 cells per well. Cells were washed once with PBS after overnight incubation, and cultured for an additional 24 hours with 2 ml of conditional RPMI medium containing 3 μM or 100 μM methionine with or without addition of 5-FU (3.4 or 10 μM) plus 10% FBS. Medium was then replaced with fresh conditional RPMI medium (0 μM or 100

μM methionine) prepared with 10% dialysed FBS, with or without addition of 5-FU (3.4 or 10

μM) containing the tracer [U-13C]-serine (CLM-1572-PK, Cambridge Isotope Laboratories). Cells were traced for 6 hours, and tracing was followed by cellular metabolite extraction.

3.4.6 Metabolite extraction

Medium was quickly aspirated, and 1 ml of extraction solvent (80% methanol/water, cooled to −80°C) was added to each well of the 6-well plates which were then transferred to

3.4.7 Liquid chromatography

UltiMate 3000 HPLC (Dionex) with an XBridge Amide column [100 mm × 2.1 mm inner diameter (i.d.), 3.5 µm; Waters] was coupled to a Q Exactive (QE) mass spectrometer (Thermo

Fisher Scientific) for metabolite separation and detection at room temperature. The mobile phase

Thermo Fisher Scientific. The flow rate used was 0.15 ml/min from 0 to 10 min and 15 to 20 min, and 0.3 ml/min from 10.5 to 14.5 min. The linear gradient was as follows: 0 min 85% B, 1.5 min

85% B, 5.5 min 35% B, 10 min 35% B, 10.5 min 25% B, 14.5 min 35% B, 15 min 85% B, and

20 min 85% B.

3.4.8 Mass spectrometry

The QE mass spectrometer (Thermo Fisher Scientific) was outfitted with a heated electrospray ionization probe (HESI) with the following parameters: evaporation temperature,

120°C; sheath gas, 30; auxiliary gas, 10; sweep gas, 3; and spray voltage, 3.6 kV for positive mode and 2.5 kV for negative mode. Capillary temperature was set at 320°C, and S-lens was 55.

A full scan range was set at 60 to 900 mass/charge ratio (m/z) with the resolution set to 70,000.

The maximum injection time was 200 ms. Automated gain control was targeted at 3 million ions.

3.4.9 Peak extraction and metabolomics data analysis

mean +/- s.d. or mean +/- s.e.m. as indicated. P values were calculated using Student’s t-test unless otherwise noted.

4. Digoxin Disrupts Central Carbon Metabolism and Reprograms the Tumor Microenvironment viii

4.1 Background and Context

Cardiac glycosides (CGs) are commonly prescribed for the treatment of cardiac arrythmias or congestive heart failure, and are thought to primarily act as inhibitors of the sodium-potassium pump (also referred to as the Na+/K+ ATPase)223. This transmembrane enzyme imports two potassium ions while exporting three sodium ions in an ATP-dependent manner, thereby maintaining the electrochemical gradient across the cell membrane224. Activity of the

Na+/K+ ATPase additionally contributes to regulation of intracellular pH225, glucose uptake226, and Ca2+ levels227; indeed, the propensity of CGs to enhance heart contractility is primarily attributed to the intracellular accumulation of Ca2+ caused by indirect activation of the Na+/Ca2+ exchanger223, as well as the isoform-specific tissue distribution of the Na+/K+ ATPase228,229.

Interestingly, CGs have also extensively been shown to exhibit anticancer activity230,231.

Numerous mechanisms have been proposed for this cytotoxicity, including intracellular acidification232, inhibition of hypoxic factors 233 or transcriptional machinery234,235, Ca2+ signaling232,236,237, induction of apoptosis via NF-kB pathway disruption238,239, and Na+/K+

ATPase-mediated activation of tyrosine kinases independent of its ion-pumping activity (i.e.

Na+/K+ ATPase “signalosome”)240-243. However, an intriguing theory proposed nearly half a century ago postulated that a partially defective Na+/K+ ATPase could be a primary tumorigenic

viii This chapter was adapted and modified from a manuscript titled “Digoxin Disrupts Central Carbon Metabolism and Reprograms the Tumor Microenvironment” by Sanderson S.M.*, Xiao Z.*, Wisdom A.J., Bose S., Liberti M.V., Reid M.A., Hocke E., Gregory S.G., Kirsch D.G., and Locasale J.W., which will be submitted for peer review in the near future. Author contributions: Conceptualization, S.M.S. and J.W.L.; Computational Data Analysis, Z.X.; Animal Experiments, A.J.W., S.M.S., M.E.R., and D.G.K.; Single-cell RNA Sequencing, E.H. and S.G.G.; Metabolomics, S.M.S.; Isotope Tracing, S.M.S. and S.B.; NCI-60 Correlations, M.V.L. and S.M.S.; All Other Experiments, S.M.S.; Supervision, S.M.S. and J.W.L. * these authors contributed equally. All text included in this chapter was written by S.M.S. 76

event by disrupting the overall energetic state of a cell244; given that the Na+/K+ ATPase accounts for anywhere between 20-70% of a cell’s ATP demand depending on the cell type245,246, it was suggested that sub-optimal activity of the enzyme could induce dysregulation of ATP-producing processes and eventually reprogram the cell towards an oncogenic state244. While this theory has remained relatively unexplored, it suggests Na+/K+ ATPase inhibition by CGs could effectively target central carbon metabolism.

In this study, we show that an early and direct consequence of the CG digoxin is disruption of central carbon metabolism and other energy-associated metabolic processes. We demonstrate that these metabolic consequences of digoxin are specifically mediated through on- target inhibition of the Na+/K+ ATPase, and that digoxin treatment can impact these metabolic processes in both healthy and malignant tissue. Finally, we use single-cell RNA sequencing to show that acute digoxin treatment induces a shift in the tumor microenvironment and exerts metabolic consequences in multiple intratumoral cell populations.

4.2 Results

4.2.1 Digoxin disrupts central carbon metabolism and energy-related processes in a temporal- and dose-dependent manner

To determine whether cell sensitivity to digoxin is associated with intrinsic metabolic state, we compared basal metabolic uptake and excretion rates with digoxin IC50 values of the

NCI-60 cancer cell line panel. This analysis demonstrated that digoxin treatment most strongly correlated with metabolic flux of TCA intermediates (Figure 23A) such as malate (Spearman correlation, r = -0.38, p = .0025) and citrate (Spearman correlation, r = -0.35, p = 0.0056) (Figure

23B), as well as the ATP-recycling metabolite creatine (Spearman correlation, r = 0.29, p = 0.02)

(Figure 23C). This initial finding suggested that Na+/K+ ATPase activity may indeed be intrinsically linked to central carbon metabolism.

Figure 23: Comparison of digoxin IC50 values and basal metabolic flux demonstrates association between digoxin and energy metabolism.

(A) Spearman rank correlations of digoxin IC50 values and basal metabolic flux in individual metabolites in NCI-60 cancer cell panel. (B) Correlation of basal metabolic flux with digoxin IC50 for malate and citrate, which exhibited the highest positive correlation coefficients in (A). (C) Correlation of basal metabolic flux with digoxin IC50 for creatine, which exhibited the highest negative correlation coefficient in (A). Each dot is representative of a cell line from the NCI-60 cell panel.

To assess the metabolic consequences of digoxin treatment, we generated global metabolite profiles of colorectal HCT116 cells both temporally (using the digoxin IC50 concentration of 100 nM) (Figure 24A) as well as acutely (3 hours) using incremental concentrations of digoxin (Figure 24B). Pathway analyses demonstrated that central carbon metabolism, as well as processes associated with its activity including aspartate/glutamate metabolism and taurine metabolism, were among the most significantly impacted metabolic pathways by digoxin in both a temporal- and dose-dependent manner (Figure 24B, below).

Closer examination of these metabolite profiles revealed an increase in upper glycolytic intermediates as well as a decrease in TCA cycle metabolites in both a temporal (Figure 24C) and dose-dependent (Figure 24D) manner. Importantly, accumulation of fructose 1,6-bisphosphate

(F1,6BP) has been shown to be predictive of glycolytic flux 6, indicating that the observed

increases in F1,6BP levels (Figure 24C,D) are demonstrative of substantial central carbon disruption. Additionally, we consistently observed a significant decrease in the levels of the metabolites taurine and creatine (Figure 24C,D). Taurine has been suggested to support mitochondrial electron transport chain (ETC) activity 247, while creatine enables the rapid anaerobic recycling of ATP under high energetic demand 248, providing additional evidence of disrupted cellular energetic state.

Figure 24: Digoxin disrupts central carbon metabolism and energy-related processes in a temporal- and dose-dependent manner.

(A) Experimental measurement of digoxin IC50 in HCT-116 cells after 48 hours. (B) Heat maps of relative integrated intensity MS values of global metabolites upon digoxin treatment at indicated times (left) and doses (right). Pathways impacted by digoxin treatment, determined from significantly-altered (p < 0.05, Student’s t-test) metabolites, denoted by indicated regions on heat maps. (C,D) Representative central- carbon and energy metabolites impacted by digoxin treatment with respect to time (C) and dose (D).

To study how digoxin impacts the regulation of central carbon pathways, we then measured steady-state glucose incorporation into glycolysis and the downstream TCA cycle using media supplemented with uniformly labeled glucose ([U-13C]-glucose) (Figure 25A). We found significant reductions in labeling of both glycolytic and TCA intermediates (Figure 25B). 79

Additional assessment of kinetic glycolytic flux into the mitochondria further revealed significantly reduced labeling of TCA intermediates as early as 2 hours after [U-13C]-glucose administration (Figure 25C). Finally, we used [U-13C]-glutamine to measure steady-state incorporation of glutamine into the TCA cycle via conversion to a-ketoglutarate (a-KG) (Figure

25D). Similar to our previous results, we found significantly reduced labeling of TCA intermediates (Figure 25E) further indicating dysregulation of mitochondrial activity. Altogether, these findings demonstrate that digoxin substantially disrupts multiple nodes of central carbon metabolism which further extends to other intracellular energetic processes.

Figure 25: Digoxin alters flux through central carbon metabolism.

(A) Diagram of isotopologue labeling of [U-13C] glucose through glycolysis into the TCA cycle. (B) Fraction of [U-13C] glucose labeled isotopologues in glycolytic and TCA intermediates relative to total abundance. (C) Relative MS intensity of biologically-relevant [U-13C] glucose isotopologues of lactate and succinate at indicated time points. p < 0.05, *** p < 0.001, Student’s t-test. (D) Diagram of isotopologue labeling from [U-13C] glutamine through the TCA cycle. (E) Fraction of [U-13C] glutamine-labeled isotopologues in TCA intermediates relative to total abundance.

4.2.2 Digoxin exerts its metabolic effects via on-target inhibition of the Na+/K+ ATPase

Upon consideration of the extensive metabolic reprogramming induced by digoxin, particularly within processes that are critical for supporting the high proliferation rates of cancer cells, we assessed whether the cytotoxic effects of digoxin could be blocked by metabolic intervention. We cultured HCT116 cells in media containing 100 nM digoxin as well as supplementations of nutrients related to central carbon metabolism or redox balance; surprisingly, we found that the majority of these supplementations were insufficient to rescue cells from digoxin treatment (Figure 27A), although the supplementation of either nucleosides or the antioxidant N-acetyl-cysteine (NAC) were modestly cytoprotective (Figure 27B), in line with previous reports of marginal antioxidant protection against digoxin cytotoxicity 249. These results suggest that the metabolic disruption induced by digoxin likely exerts extensive downstream consequences on cellular homeostasis and function.

Figure 26: Nucleoside and antioxidant supplementation partially block digoxin-induced cytotoxicity.

(A) Relative cell viabilities of HCT-116 cells treated with vehicle or 100 nM digoxin in regulator media or with supplementation of the indicated nutrient for 72 hours. B) Relative cell viabilities of HCT-116 cells treated with vehicle or 100 nM digoxin in regulator media or with supplementation of the indicated nutrient for 72 hours. * p < 0.05, ** p < 0.01, *** p < 0.001, Student’s t-test.

Given the multitude of proposed mechanisms for the cytotoxic activity of CGs, the possibility remained that the metabolic consequences of digoxin were due to factors beyond specific inhibition of Na+/K+ ATPase activity. It is well-established that murine Na+/K+ ATPase enzymes are substantially less sensitive to CGs, with the mouse isoform exhibiting roughly 1000- fold lower affinity for CGs than the human counterpart 250,251; indeed, it has been shown that ectopic expression of the a-subunit of the mouse ATPase (mATP1a1) in human cells can be sufficient to rescue cell viability upon digitoxin treatment 232. Therefore, to determine whether the observed metabolic effects were specifically attributable to the on-target cytotoxic activity of digoxin, we ectopically expressed the mouse mATP1a1 subunit in HCT116 cells and demonstrated a rescue of cell viability (Figure 27A). We then treated them with the digoxin IC50 concentration exhibited by parental cells, and compared the resulting metabolite profiles between the two groups (Figure 27B). We found that mATP1a1 expression effectively blocked the metabolic consequences of digoxin (Figure 27C), thereby restoring central carbon metabolism

(Figure 27D,E) as well as taurine (Figure 27F) and creatine (Figure 27G) levels. This finding establishes that the cytotoxic activity of digoxin is intrinsically linked to disrupted central carbon metabolism, which collectively are a direct result of Na+/K+ ATPase inhibition.

Figure 27: Digoxin exerts its metabolic effects via on-target inhibition of the Na+/K+ ATPase.

(A) Relative cell viability of HCT-116 cells transfected with mATP1a1 digoxin-resistant subunit, cultured in incremental concentrations of digoxin. (B) Heat map of fold changes in global metabolite levels with or without 48-hour 100 nM digoxin treatment in HCT-116 transfected cells. (C) Volcano plots of log- transformed fold changes in metabolites after digoxin treatment in HCT-116 transfected with GFP control or mATP1a1 vectors. P values obtained using Student’s t-test. GSH, reduced glutathione; GSSG, oxidized glutathione. (D-G) Relative fold changes in (D) TCA cycle metabolite levels, (E) lactate levels, (F) taurine levels, and (G) creatine levels upon 100nM digoxin treatment.

4.2.3 Digoxin impacts energy metabolism in a tissue-specific and antineoplastic manner

While our results thus far illustrate the significant impact of digoxin on central carbon metabolism and related energetic processes in cell culture, we next sought to determine whether

these effects could be achieved in a more variable in vivo setting. Numerous findings of CGs effectively inhibiting tumor growth in xenograft studies have been reported 232,233,249; however, given the extremely differential CG sensitivity of the ATPase isoforms found in human-derived cells and the murine host species, these findings are limited by the possibility that the therapeutic window needed to achieve tumor growth inhibition may be prohibitively toxic in a setting where the host and tumor both express similar ATPase isoforms 252. Additionally, it remains to be explored whether Na+/K+ ATPase inhibition impacts metabolism in healthy tissue as well.

To investigate these important considerations we cultured mouse sarcoma cells generated from primary sarcoma tumors driven by an oncogenic KRAS mutation and p53 deletion (KrasLSL-

G12D/+ flox/flox ;Trp53 , or KP), and found that they exhibited an IC50 of 100 µM in line with their murine isoform expression (Figure 28A). As proof of principle, we verified that administration of this concentration reliably impacted central carbon metabolism in cell culture (Figure 28B).

Figure 28: Digoxin impacts central carbon metabolism in primary murine sarcoma cells.

(A) Experimental determination of digoxin IC-50 in cultured primary murine sarcoma cells. (B) Relative fold changes in glycolytic (left) and TCA cycle (right) intermediates after 48 hours of 100 µM digoxin treatment.

We then orthotopically injected these cells into the upper right thigh muscle of syngeneic mice; upon tumor palpation (approximately 11 days after injection), we treated mice with a previously-reported dose253 of 2mg/kg digoxin every 24 hours and collected tumors as well as healthy tissue after administration of the fourth dose (Figure 29A). Of note, while this treatment 85

regime appeared to trend towards tumor growth inhibition (Figure 29B), treated mice exhibited partial weight loss although no other physiological or behavioral signs of toxicity were observed.

Metabolite profiling revealed cardiac tissue was the most significantly impacted tissue type

(Figure 29C), with substantial alterations in glycolytic and TCA intermediates (Figure 29D).

Surprisingly, these effects were even more pronounced in cardiac than tumor tissue in this context

(Figure 29E). A number of similar changes were also found in muscle, brain, liver, and kidney tissue (Figure 29F), although to a much lesser extent than was found in cardiac tissue or the cultured sarcoma cells.

Figure 29: Acute digoxin treatment impacts energy metabolism in both healthy and malignant tissue.

(A) Diagram of digoxin treatment schedule(s) in allograft sarcoma model. N=6 per group. i.m., intramuscular; i.p., intraperitoneal. (B) Quantification of difference in tumor volume from day 0 (first day of treatment) to day 3 (final day of treatment). P-value obtained using Student’s t-test. (C) Volcano plot of log-transformed fold changes in metabolite levels between vehicle- and digoxin-treated cardiac tissue. P- values obtained using Student’s t-test. (D) Relative fold changes in glycolytic (left) and TCA (right) 86

intermediates in cardiac tissue. * p < 0.05, ** p < 0.01, Student’s t-test. (E) Volcano plot of log- transformed fold changes in metabolite levels between vehicle- and digoxin-treated tumors. P-values obtained using Student’s t-test. (F) Volcano plot of log-transformed fold changes in metabolite levels from muscle, brain, liver, and kidney tissue extracted from vehicle- or digoxin-treated mice. P-values obtained using Student’s t-test.

To examine the antineoplastic potential of digoxin in this setting, we additionally treated a group of orthotopically-engrafted syngeneic mice with 2mg/kg every 48 hours and monitored the resulting longitudinal tumor growth. There were no discernible histological differences between vehicle- and digoxin-treated groups, with the majority of tumors exhibiting regions of both low and high vascularity (Figure 30A). Regression analyses demonstrated that digoxin treatment significantly delayed tumor growth as measured by time to tumor quintupling (Figure

30B). Furthermore, the most prominent metabolic alterations found in tumors exposed to chronic digoxin treatment were consistent with dysregulation of energy-related metabolic processes

(Figure 30C), with disruptions to mitochondrial metabolism (Figure 30D) as well as taurine

(Figure 30E) and creatine (Figure 30F) levels. Collectively, these findings demonstrate that digoxin can impact metabolic processes in both healthy and malignant tissue, and that these metabolic perturbations are associated with antineoplastic effects in a physiological setting.

Figure 30: Digoxin impacts intratumoral energy metabolism in an antineoplastic manner.

(A) H&E sections of a representative tumor, showing regions of both low (left) and high (right) vascularity. Scale bars represent 20 µm. (B) Kaplan-Meier survival curve (left) and quantification of time for tumors to quintuple in volume (right). P-values obtained using Student’s t-test. (C) Volcano plot of log- transformed fold changes in metabolite levels between vehicle- and digoxin-treated tumors. P-values obtained using Student’s t-test. (D-F) Relative fold changes in levels of (D) TCA intermediates, (E) taurine metabolites, and (F) creatine. * p < 0.05, ** p < 0.01, *** p < 0.001, Student’s t-test.

4.2.4 Acute digoxin treatment induces a defined shift in the tumor microenvironmental landscape

An additional feature of allograft models is the presence of an intact immune system, which is necessarily lacking in human xenograft models to enable the introduction of foreign species material 254,255. Given that induction of central carbon metabolic reprogramming has become increasingly appreciated in immune cell function 256-258, we performed single-cell RNA sequencing (scRNA-seq) to explore the effects of digoxin treatment on the tumor microenvironment. Briefly, we treated mice with vehicle or 2 mg/kg digoxin (using two mice per group) with the daily treatment regimen described previously (Figure 29A) and harvested the tumors after the final fourth dose, at which point we dissociated the cells into a single-cell 88

suspension and performed 10x scRNA-seq without cell type sorting or purification using the

Chromium drop-seq platform (10x Genomics) (Figure 31A). We analyzed approximately 10,000 cells from each tumor, with roughly 100,000 reads from each cell that covered nearly 5,000 genes

(Figure 31B), and used previously published scRNA-seq data of healthy murine muscle tissue 259 to distinguish between malignant and non-malignant cells based on relative gene copy number

(Figure 31C). Acute digoxin treatment trended towards a reduction in the relative population of malignant cells (Figure 31D), consistent with our previous observations of early tumor growth inhibition (Figure 29B).

Upon further analysis of the single-cell transcriptomes we identified fifteen immune cell populations, including five distinct T-cell populations (Figure 31E). Myeloid cells (i.e. type I and type II macrophages, M1 and M2 respectively) were the most abundant immune cell population, consistent with previous reports of the primary KP sarcomas 260 (Figure 31F). While the populations of intratumoral T cell and neutrophil infiltrates did not appear to be appreciably skewed by this short-term digoxin treatment, the relative populations of dendritic cell and B cell infiltrates were increased and decreased, respectively (Figure 31G). These results demonstrate that acute digoxin treatment induces an immediate shift in the tumor microenvironment in this orthotopic sarcoma allograft model.

Figure 31: Digoxin induces a shift in cell populations within the tumor microenvironment.

(A) Diagram of single-cell RNA sequencing workflow, from tumor generation and harvesting to gene profiling. (B) The number of cells, median number of genes, and mean transcript reads obtained from individual cells for each indicated sample. (C) Representation of the transcriptome of single cells, classified as either malignant or non-malignant as determined by relative copy number to normal mouse muscle tissue. Represented here by the Replicate 1 tumor sample from vehicle-treated group. (D) Relative abundance of malignant and non-malignant cell populations in the indicated tumor sample. (E) t-SNE plots of scRNA-seq data demonstrating distinct clusters of immune cells from all tumor samples, either full immune populations (left) or specific T-cell populations (right). (F) Relative abundance of indicated cell populations in the total immune cell pool. (G) Relative abundance of indicated cell populations between the 4 tumors.

4.2.5 Digoxin is associated with transcriptional reprogramming of metabolic processes in tumor cells and immune infiltrates.

Finally, to determine whether digoxin treatment exerts differential effects on metabolic programming between cell types, we analyzed the scRNA-seq transcriptomes of individual tumor cell populations. Oxidative phosphorylation was the most extensively altered metabolic program in malignant cells, with significant increases in expression of Atp5k transcripts corresponding to the ATP synthase gene (Figure 32A). ATP synthase is particularly known to be intrinsically linked to glycolysis with its expression and activity shown to inversely correlate with glycolytic rate 261, thereby providing additional evidence that disruption of central carbon metabolism is an immediate and direct consequence of digoxin treatment.

Upon examination of the immune cell populations, we found that both M1 and M2 macrophage populations exhibited the most substantial transcriptional alterations in metabolic genes compared to the other immune infiltrates as a result of digoxin treatment (Figure 32B).

These macrophage populations exhibited surprisingly unique metabolic reprogramming specifically within central carbon metabolism, with M1 macrophages characterized by upregulation of glycolysis (Figure 32C) and M2 macrophages characterized by an upregulation in oxidative phosphorylation similar to malignant cells (Figure 32D).

Figure 32: Acute digoxin treatment is associated with transcriptional reprogramming of central carbon metabolism within tumor cells and immune infiltrates.

(A) Volcano plot of log-transformed fold changes in metabolic gene transcript levels between vehicle and digoxin-treated tumor samples (left), with KEGG pathway analysis demonstrating the most impacted metabolic pathways (right). P-values obtained using Wilcoxon Rank Sum test. (B) Distribution of metabolic gene transcripts for each distinct cell population, with significantly-altered (p < 0.1) genes marked in red. Adjusted p-values calculated using Bonferroni correction. (C, D) Volcano plot of log- transformed fold changes in metabolic gene transcript levels in M1 (C) and M2 (D) macrophage populations derived from vehicle and digoxin-treated tumor samples; below, KEGG pathway analysis of impacted metabolic pathways, determined using significantly-upregulated gene transcripts. P-values obtained using one-tailed Fisher’s exact test.

Interestingly, previous studies of macrophage behavior have also observed similar metabolic shifts in these two populations upon their activation 262,263. To examine this possibility

more closely, we performed KEGG pathway analysis on their full transcriptomes, which were similarly most significantly impacted in comparison to other immune infiltrate populations

(Figure 33A) and additionally found transcriptional signatures consistent with their polarization

(Figure 33B), including upregulation of components in the phagosome 264 and endoplasmic reticulum 265,266. These observations illustrate that digoxin is associated with broad transcriptional consequences, including within central carbon metabolic processes, across multiple intratumoral cell populations.

Figure 33: Transcriptional profiles of macrophage populations suggest their activation is associated with digoxin treatment.

(A) Distribution of gene transcripts from full transcriptome for each distinct cell population, with significantly-altered (p < 0.1) genes marked in red. Adjusted p-values calculated using Bonferroni correction. (B) KEGG pathway analysis of most significantly altered pathways as determined by significantly upregulated (left) and downregulated (right) transcripts in M1 (left) and M2 (right) macrophages. P values obtained using one-tailed Fisher’s exact test.

4.3 Discussion

Metabolic programming, on both a cellular and physiological level, is known to be highly regulated by contextual factors which are extrinsic to discrete metabolic reactions 41,53,267-269. In line with this, many conventional therapies have been shown to exert substantial metabolic effects beyond their understood mechanism of action 160,270,271. As many of the metabolic vulnerabilities inherent to cancer cells are difficult to target without inducing toxic consequences on healthy tissue, identification of agents that can be repurposed to exploit these processes in an antineoplastic manner remains an active area of investigation 38,163,172,272.

Cardiac glycosides have been shown to exhibit antineoplastic effects in numerous settings 230,231,253, which have been attributed to a myriad of sources such as disruption of proton gradients 232 and activation of kinases that physically interact with the sodium-potassium pump

(i.e. Na+/K+ ATPase “signalosome”) 273-275. However, fluctuations in energetic demand from membrane transport activity have been shown to impact glycolytic rate 276, and it has been historically hypothesized that subtle alterations in Na+/K+ ATPase activity may be a contributing factor to enhanced dependence on glycolysis in cancer 244. Our results demonstrate that digoxin, through on-target inhibition of the Na+/K+ ATPase, induces broad metabolic disruptions in a temporal- and dose-dependent manner; these disruptions were most prominent within central carbon and energy-related metabolic processes, including taurine and creatine metabolism.

Interestingly, taurine has been suggested to act as a mild cardiac glycoside in cardiac tissue through its modulation of mitochondrial ROS 247, and taurine loss has been associated with Na+ efflux thereby reducing the degree of Ca2+ overload 277. It’s therefore tempting to speculate that the dual loss of taurine and creatine upon digoxin treatment are indirect compensatory responses to ion imbalance and reduced ATP production, respectively.

Our results further provide the first characterization of the global metabolic consequences of digoxin on healthy tissues to allow direct comparison with metabolic profiles in tumors. We found that acute digoxin treatment effectively impacted central carbon metabolism or other energy-related metabolites (i.e. taurine and creatine) in diverse healthy tissues, with the most observable metabolic consequences found in cardiac tissue. Although the higher degree of central carbon metabolic disturbance in cardiac compared to tumor tissue after short-term digoxin treatment was relatively unexpected, it is likely a result of the relative tissue-specific affinity for

CGs; indeed, the human cardiac-specific a-subunit isoform is known to exhibit a slightly higher affinity for CGs compared to the isoforms present in most other tissues 229, thereby providing a potential explanation for the observed effects. Importantly, despite these prominent metabolic alterations in cardiac tissue, we did not observe any signs of cardiotoxicity and these alterations were considerably more prominent in tumors with the antineoplastic long-term digoxin treatment, indicating the enhanced dependency of tumor cells on these processes.

We have recently demonstrated that scRNA metabolic gene transcriptomes can be used to determine overall metabolic activity and plasticity within individual cells 278. Our results additionally contribute to our understanding of unique cell-autonomous responses within the tumor microenvironment by using scRNA-seq to examine diverse cell populations and metabolic programs within tumors following acute cytotoxic treatment. Our findings of both altered representation of immune infiltrates, as well as transcriptional programs of metabolic processes featuring clear distinctions between malignant and immune cell populations, in response to short- term digoxin treatment provide a novel glimpse into cell-specific intratumoral heterogeneity. It remains unclear whether the altered metabolic programs and possible activation of different myeloid populations are a direct result of exposure to digoxin, or are instead an indirect response to signals from adjacent malignant cells; furthermore, it is possible that the observed variations in

cell populations might become more pronounced with longer exposure to digoxin or at different stages of tumor development. Indeed, the potential dual polarization of M1 and M2 macrophages

(which are commonly characterized as pro-inflammatory or anti-inflammatory, respectively)279 is intriguing and warrants further investigation. It will be interesting in future studies to determine the functional consequences of these variations in metabolic reprogramming between immune cell populations, especially upon consideration of previous studies reporting consequences of digoxin on immune cell activity in some noncancerous settings 280,281. Additional studies of these interactions could potentially illuminate synergistic mechanisms whereby the clinical efficacy of digoxin could be enhanced when combined with targeted immune therapies, as has been found with other compounds 260,282,283.

4.4 Materials and Methods

4.4.1 Cell culture and reagents

Cells were cultured at 37°C, with 5% atmospheric CO2 in RPMI-1640 (GIBCO), 10% heat-inactivated fetal bovine serum (FBS; F2442, Sigma), 100 U/mL penicillin , and 100 mg/mL streptomycin. HCT116 cells were obtained from the American Tissue Culture Collection

(ATCC), and murine sarcoma cells were generated and cultured from primary murine sarcomas described previously260 and below (Section 4.4.5).

4.4.2 Digoxin IC-50 measurements

Cells were plated at a density of 5.0 x103 cells/well in triplicate in a 96-well plate and were allowed to adhere in full RPMI 1640 media for 24 hours. Cells were briefly washed with

PBS and then incubated in medium containing vehicle (DMSO; #97061-250, VWR) or indicated concentrations of digoxin (#D6003, Sigma Aldrich). After 48 hours, the media was aspirated and replaced with 100µl phenol-red free RPMI-1640 (Gibco) and 12mM 3-[4,5-Dimethylthiazol-2-

yl]-2,5-diphenyltetrazolium (MTT) (Thermo Fisher Scientific, #M6494). After 4 hours, the MTT- containing media was aspirated and 50µl DMSO was added to dissolve the formazan. After 5 minutes, absorbance was read at 540nm. For all experiments, three technical replicates per culture condition were used.

4.4.3 Nutrient rescue experiments

For all nutrient rescue experiments, cells were plated at a density of 5.0 x105 cells/well in triplicate in a 6-well plate and were allowed to adhere for 24 hours. Cells were briefly washed with PBS and then incubated with vehicle (DMSO) or 100nM digoxin, as well as one or more of the following supplementations: 5mM N-acetyl-cysteine (NAC; A9165, Sigma), 2µM Trolox

(#238813, Sigma Aldrich), 100µM adenosine 5’-triphosphate disodium salt hydrate (ATP;

A26209, Sigma Aldrich), 10mM creatine (C3630, Sigma Aldrich), 10mM Taurine (T8691, Sigma

Aldrich), 2µM dimethyl-a-ketoglutarate (cell-permeable a-KG; 349631, Sigma Aldrich), 5mM sodium pyruvate (sc-208397A, Santa Cruz Biotechnology), 500uM nicotinamide (N0636, Sigma

Aldrich), 100µM beta-nicotinamide adenine dinucleotide (NAD+; 160047, MP Biomedical), and

1x Embryomax Nucleosides (ES-008-D, Millipore). For glucose-restricted medium, RPMI 1640 glucose- and glutamine-free medium was supplemented with 2mM glutamine, 10% heat- inactivated FBS, 100 U/mL penicillin, and 100mg/mL streptomycin; indicated concentrations of glucose were then serially added to the medium. After 48 hours, an MTT cell viability assay was performed as described above. For all experiments, three technical replicates per culture condition were used.

4.4.4 Isotope tracing

[U-13C] glucose (CLM-1396-10) and [U-13C] glutamine (CLM-1822-H-PK) were purchased from Cambridge Isotope Laboratories. [U-13C] glucose was added to RPMI 1640

glucose-free medium at a concentration of 11mM, while [U-13C] glutamine was added to glucose- and glutamine-free RPMI-1640 medium (supplemented with 11mM glucose) to a concentration of 2 mM [U-13C] glutamine. Cells were plated at a density of 3.0 x105 cells/well in a 6-well plate, and after 24 hours were treated with either vehicle (DMSO) or 100nM digoxin. After 4 hours of drug treatment, the medium was quickly aspirated and briefly rinsed with glucose-free RPMI

1640 medium, and either [U-13C] glucose- or [U-13C] glutamine-containing medium was added to each well. Cells were then collected for metabolite extraction either 20 hours later or at the indicated time points. For all experiments, three technical replicates per culture condition were used.

4.4.5 Lentiviral transfection and transduction

HEK-293T cells were plated at a density of 1.0 × 106 cells/10 cm plate in RPMI 1640

(Gibco) supplemented with 10% heat-inactivated FBS, penicillin (100 U/ml), and streptomycin

(100 mg/ml) and were allowed to adhere and reach 70% confluency. 15 µg of mATP1a1 (EX-

Mm01329-Lv105, GeneCopoeia) or GFP control (EX-EGFP-Lv105, GeneCopoeia) plasmid, 10

µg of PsPAX2 packaging vector (no. 12260, Addgene), and 5 µg of PMD2.G envelope– expressing plasmids (no. 12259, Addgene) were diluted in 500 µl of jetPRIME buffer (no. 114-

07, Polyplus-transfection) and vortexed. Next, 60 µl of the jetPRIME transfection reagent (no.

114-07, Polyplus-transfection) was added to the mixture, vortexed for 10 seconds, and left to incubate for 10 minutes at room temperature. The medium in the plate was replaced with fresh

RPMI 1640 medium, and the transfection mix was then added to the 10-cm plate dropwise. After

24 hours, the transfection medium was replaced with fresh RPMI 1640 medium. After an additional 24 hours, the medium was collected and filtered through a 0.45-µm filter for virus collection. HCT116 cells were plated in 10-cm plates, and when they reached 30-50% confluency, virus-containing medium (1:1 with fresh RPMI 1640 medium) was added to the 98

plates along with 4 µg/µl polybrene. After 24 hours, the virus-containing medium was removed and replaced with fresh RPMI 1640 medium. Cells were incubated with 1 µg/ml puromycin for

48 hours for positive selection.

4.4.6 Sarcoma allograft mouse studies

All animal studies were performed in accordance with protocols approved by the Duke

University Institutional Animal Care and Use Committee (IACUC) and adhere to the NIH Guide for the Care and Use of Laboratory Animals. Cultured murine sarcoma cells were generated from primary KrasLSL-G12D/+;Trp53flox/flox (KP) murine sarcoma tumors as previously described 260. All mice were maintained on a pure 129/SvJae genetic background. Cultured KP murine cells were suspended in DMEM medium at a concentration of 5.0 x 106 cells/mL. An intramuscular (i.m.) injection of approximately 5.0 x 104 cells was administered to the flank region of mice from the same strain as those used to generate the primary tumors. When tumors reached 70-150 mm3 as determined by caliper measurement in two dimensions, the sarcomas were randomized to vehicle or digoxin groups. Mice were administered an intraperitoneal (i.p.) injection of either vehicle

(PBS) or 2 mg/kg digoxin (prepared in DMSO and then diluted in PBS) with a volume not exceeding 250 µL. For short-term treatments, injections were administered every 24 hours, until mice were euthanized via cervical dislocation 3 hours after administration of the fourth dose. For long-term treatments, injections were administered every 48 hours with tumor growth measured

3x weekly until sarcomas exceeded 13 mm3 in any dimension in accordance with IACUC guidelines at Duke University, at which point mice were euthanized via cervical dislocation.

Tumor, heart, kidney, liver, brain, muscle, and plasma samples were collected and immediately snap-frozen in liquid nitrogen. For the longitudinal treatment study, sections of tumors were preserved for immunohistochemical analysis.

4.4.7 Immunohistochemistry and microscopy

Fresh tumor samples were harvested after euthanasia, fixed in 4% PFA overnight, and preserved in 70% ethanol until paraffin embedding. All paraffin embedding, tissue sectioning, and immunohistochemical (IHC) staining was carried out by the Duke University Pathology

Research Histology and Immunohistochemistry Laboratory shared resource facility. For tissue embedding and sectioning, previously established protocols were used 284. For hematoxylin and eosin (H&E) staining of tissue sections, standard protocols were used 285. Representative images of each H&E section were captured using a Leica DM IL LED microscope equipped with a Leica

MC170HD camera with a 20x objective using LAS EZ software (Leica). Scale bars = 20 µm.

4.4.8 Metabolite extraction

Media was quickly aspirated and 1 mL of extraction solvent (80% methanol/water, cooled to -80°C) was added to each well of the 6-well plates, and were then transferred to -80°C for 15 minutes. Plates were removed and cells scraped into the extraction solvent on dry ice. All metabolite extracts were centrifuged at 20,000g at 4°C for 10 minutes. Finally, the solvent in each sample was evaporated in a speed vacuum, and the resulting pellets were stored in -80°C until resuspension. For polar metabolite analysis, the cell extract was dissolved in 15µl of water and

15µl of methanol/acetonitrile (1:1, v/v) (LC-MS optima grade, Thermo Fisher Scientific).

Samples were centrifuged at 20,000g 4°C for 2 minutes, and the supernatants were transferred to liquid chromatography (LC) vials.

4.4.9 Liquid chromatography

Ultimate 3000 HPLC (Dionex) with an Xbridge amide column (100 x 2.1 mm i.d., 3.5

μm; Waters) is coupled to Q Exactive-Mass spectrometer (QE-MS, Thermo Scientific) for metabolite separation and detection at room temperature. The mobile phase A reagent is

100

composed of 20 mM ammonium acetate and 15 mM ammonium hydroxide in 3% acetonitrile in

HPLC-grade water (pH 9.0), while the mobile phase B reagent is acetonitrile. All solvents are

LC-MS grade, purchased from Fischer Scientific. The flow rate used was 0.15 mL/min from 0-10 minutes and 15-20 minutes, and 0.3 mL/min from 10.5-14.5 minutes. The linear gradient was as follows: 0 minutes 85% B, 1.5 minutes 85% B, 5.5 minutes 35% B, 10 minutes 35% B, 10.5 minutes 25% B, 14.5 minutes 35% B, 15 minutes 85% B, and 20 minutes 85% B.

4.4.10 Mass spectrometry

The QE-MS is outfitted with a heated electrospray ionization probe (HESI) with the following parameters: evaporation temperature, 120°C; sheath gas, 30; auxiliary gas, 10; sweep gas, 3; spray voltage, 3.6 kV for positive mode and 2.5 kV for negative mode. Capillary temperature was set at 320°C and S-lens was 55. A full scan range was set at 60 to 900 (m/z), with the resolution set to 70,000. The maximum injection time (max IT) was 200 ms. Automated gain control (AGC) was targeted at 3,000,000 ions.

4.4.11 Peak extraction and metabolomics data analysis

Data collected from LC-Q Exactive MS was processed using commercially available software Sieve 2.0 (Thermo Scientific). For targeted metabolite analysis, the method “peak alignment and frame extraction” was applied. An input file (“frame seed”) of theoretical m/z

(width set at 10 ppm) and retention time of ~260 known metabolites was used for positive mode analysis, while a separate frame seed file of ~200 metabolites was used for negative mode analysis. To calculate the fold changes between different experimental groups, integrated peak intensities generated from the raw data were used. Hierarchical clustering and heatmaps were generated using Morpheus software (The Broad Institute, https://software.broadinstitute.org/morpheus/). For hierarchical clustering, spearman correlation parameters were implemented for row and column parameters. Pathway enrichment analysis was 101

conducted using MetaboAnalyst 3.0 software (http://www.metaboanalyst.ca/faces/home.xhtml); briefly, HMDB IDs from the metabolites that were significantly enriched (greater than log2fold change with P<0.05) were inputted. The pathway library used was Homo sapiens and Fishers’

Exact test was employed for over-representation analysis.

4.4.12 Single-cell RNA sequencing

Tumors were dissected and minced following the manufacturer’s protocol using MACS

C tubes and the mouse Tumor Dissociation Kit (Miltenyi Biotec) After tumor dissociation, the cells were filtered through a 40 µM strainer. Red blood cells were lysed using ACK Lysing

Buffer (Lonza) and washed with flow buffer made of HBSS (cat 13175-095, Gibco), 5 mM

EDTA (E7899, Sigma-Aldrich), and 2.5% FBS (Gibco). Tumors were washed twice more in

0.04% bovine serum albumin (BSA) in PBS, then resuspended at 1000 cells per µL. Cell suspensions were loaded on the 10x Genomics Chromium Controller Single-Cell Instrument (10x

Genomics) using the Chromium Single Cell 3’ Reagent v3 Kit (PN-1000092, 10x Genomics).

Cells were mixed with reverse transcription reagents, gel beads, and oil to generate single-cell gel beads in emulsions (GEM) for reverse transcription (RT). After RT, GEMs were broken and the single-stranded cDNA was purified with DynaBeads. cDNA was amplified by PCR and the cDNA product was purified with the SPRIselect Reagent Kit (B23318, Beckman Coulter).

Sequencing libraries were constructed using the reagents provided in the Chromium Single-Cell

3’ Library Kit following the user guide. Sequencing libraries were sequenced with the Illumina

Novaseq 6000 platform at the Duke GCB Sequencing and Genomic Technologies Core.

4.4.13 Single-cell RNA sequencing data processing

The sequencing data produced by the 10x Chromium platform was processed into gene expression table using the pipelines from Cell Ranger v3.0.2 (https://support.10xgenomics.com/).

102

Briefly, raw base call files were first demultiplexed into sample specific FASTQ files with the cellranger mkfastq pipeline. The FASTQ files for each sample were then aligned to mouse reference genome (mm10) using STAR 286. The aligned reads for each gene were further counted by the cellranger count pipeline based on unique molecular identifier (UMI) and the cell barcode which were devoted to distinguishing different mRNA and the source cells. Quality control and filtering steps were performed to remove the low-quality cells and uninformative genes. Cells with fewer than 1800 expressed genes were discarded. The genes that were detected in fewer than

10 cells were excluded from the downstream analyses.

4.4.14 Classification of single cells into malignant and non-malignant cells

Since malignant cells typically harbor large-scale copy number alteration (i.e. gains or deletions of whole chromosomes or large chromosomal regions) that distinguish them from nonmalignant cells 287-290, we performed the copy number variations (CNVs) analysis for each sample to classify the single cells into malignant and nonmalignant cells. The copy number profiles were estimated based on the average expression of large sets of genes in each chromosomal region using inferCNV v0.99.7 (https://github.com/broadinstitute/inferCNV). An external normal cells from limb muscle of two mouse that were processed on the 10X genomics platform 259 served as the reference for CNVs calling. The same filtering procedures were applied to reference dataset so that the cells with less than 1800 expressed genes, and the genes that were expressed in fewer than 10 cells were excluded. Genes with average expression value of at least

0.1 in reference cells were included in the following analyses. The CNVs were estimated by sorting the genes according to their chromosomal position and using a moving average window with length 101 within each chromosome. The default hierarchical clustering was performed on the CNV scores and split the cells of each sample into two clusters. We assigned the cluster with

103

low-frequency CNVs like the reference cells as the non-malignant, while the other cluster with high-frequency CNVs was considered as the malignant.

4.4.15 Classification of cell populations within non-malignant cells

The Seurat v3.0.2 (http://satijalab.org/seurat/) R package was used for non-malignant cell type identification as described below. The gene expression data was firstly log-normalized and scaled with default parameters. The top 2000 most variable genes selected by Seurat were used in the principle component analysis (PCA). The first 85 principal components (PCs) that selected based on the built-in jackstraw analysis were used for downstream clustering analysis and t-SNE analysis. Cell clusters were defined using FindClusters functions implemented in Seurat with default parameters and resolution=0.15. The t-SNE analysis was used to visualize the clustering results with perplexity setting to 1% of cell number whenever it was larger than 30 and learning rate setting to 1/12 of cell number whenever it was above 200, as previously described 291. Based on the known marker genes from literatures and differential expressed genes identified by

FindAllMarkers function, we annotated the cell type of each cluster. T cells were further separated into different subtypes based on following procedures: cells were firstly classified as

CD8+ and CD4+ based on the expression levels of gene Cd8a and Cd4. The T cells with Cd8a expression level larger than 0.1 were considered as the CD8+. Similarly, those with Cd4 larger than 0.1 were considered as CD4+ type. While the remaining cells with both Cd4 and Cd8a expression below than 0.1 were tentatively labeled as the double negative T cells. CD4+ T cells with the total expression level of Foxp3 and Il2ra higher than 0.2 were further labeled as Tregs, while other CD4+ T cells were labeled as Ths. Hence the T cells were initially classified as CD8+,

CD4+ Tregs, CD4+ Ths and double negative T cells based on expression level of Cd8, Cd4, Foxp3 and Il2ra. Differential expression analysis on these four group of T cells was performed using the

FindAllMarkers function in Seurat. The differentially-expressed genes (Bonferroni-corrected p-

104

value < 0.1) were used for PCA. The top four PCs that selected based on jackstraw analysis were used for next clustering analysis. We identified five cell populations with setting resolution to 0.1.

Finally, based on known gene markers we annotated these T cells populations as CD8+, Tregs,

Ths, natural killer T cell (NKT) and other one cell population that cannot be clearly defined.

4.4.16 Differential gene expression and pathway enrichment analysis

The Wilcoxon Rank Sum test was performed on metabolic genes to identify differences in metabolism between single cells in control and digoxin treatment groups. Lists of metabolic genes and pathways were obtained from the KEGG database (https://www.kegg.jp/). The metabolic genes with Bonferroni-corrected p-value smaller than 0.1 and absolute average log2(fold change) larger than 0.01 were considered statistically significant and included in the enrichment analysis. The p-value calculated by one-tailed Fisher’s exact test was used to evaluate the enrichment significance of each metabolic pathway.

4.4.17 Quantification and statistical analysis

All error bars were reported as +/- s.e.m. with n=3 independent biological replicates and statistical tests resulting in p-value computations were obtained using a two-tailed Student’s t-test unless otherwise noted. All statistics were computed using Graphpad Prism 6 (GraphPad, http://graphpad.com/scientific-software/prism/).

105

5. Exercise inhibits tumor growth and central carbon metabolism in patient-derived xenograft models of colorectal cancer ix

5.1 Background and Context

As discussed in Section 1.5.4, epidemiological studies have shown that exercise or physical activity can reduce both cancer incidence and in colorectal, breast, and prostate cancer

292,293, and that the impact of exercise may differ as a function of tumor phenotype 294,295.

Numerous molecular mechanisms driving the effects of structured exercise on tumorigenesis have been proposed, with most relating to systemic effects such as inflammation and changes in circulating hormone levels 296-298. However, given that exercise is known to impact both cellular and systemic metabolism in healthy subjects, including altering mitochondrial dynamics in skeletal muscle 299, it remains to be explored whether exercise can exert similar effects on these metabolic processes within tumors.

To examine these considerations, we generated six colorectal cancer patient-derived xenograft (CRC PDX) models to determine the effects of exercise on tumor growth and intratumoral metabolic alterations. Although the use of preclinical murine models to study effects of therapeutic intervention can be a powerful tool to identify and characterize potential therapies that may be clinically beneficial, the majority of xenograft models have been based on immortalized cancer cell lines engrafted in mice. Some of these cancer cell lines have been

ix This chapter was adapted and modified from published work: Lu M.*, Sanderson S.M.*, Zessin A., Ashcraft K.A., Jones L.W., Dewhirst M.W., Locasale J.W., and Hsu D.S. “Exercise inhibits tumor growth and central carbon metabolism in patient-derived xenograft models of colorectal cancer.” Cancer & Metabolism (2018). This chapter was reproduced in accordance with the CC-BY license. Author contributions: Conceptualization, J.W.L. and D.S.H., with intellectual input contributed by L.W.J. and M.W.D.; Study and experimental design, M.L. and S.M.S.; Animal experiments; M.L., A.Z., K.A.A., L.W.J., M.W.D., D.S.H.; Metabolomics sample preparation and analysis, S.M.S. * = these authors contributed equally. All text in this chapter was written by S.M.S.

106

established for decades in cell culture, and thus may not fully capture the heterogeneity within an individual tumors 300. In contrast, we and others have demonstrated that rapid engraftment of patient tumor samples into immunodeficient mice to develop PDXs may provide a more clinically applicable murine model to study potential therapeutic strategies 301-306. Specifically, we have shown that many elements of the biology of PDXs are similar to that of the corresponding patient tumors, at both the histologic and molecular levels even after multiple passages 220,221. Using our established PDX approach coupled with metabolomics analysis of tumor samples, our current study provides the first characterization of the effect of exercise on PDX tumor growth as well as a global analysis of the metabolic alterations induced by exercise in both exercise-responsive and exercise-non-responsive tumors.

5.2 Results

5.2.1 Characterization of six colorectal cancer PDX models

Six colorectal cancer (CRC) patient-derived xenograft (PDXs) were developed: CRC240,

CRC282, CRC344, CRC361, CRC370, and BRPC12-146, as previously described 220,221. Tumor samples were derived from patients who underwent resection of their CRC liver metastasis, peritoneal implants, or primary CRC tumor (Table 3). Similar to our previous work, pathological features between matched PDX and patient tumor were observed (Figure 34A). We subsequently used these six CRC PDX models to study the effects of exercise on tumor metabolism and growth.

107

Table 3: Demographics of patient-derived xenografts of CRC240, CRC282, CRC344, CRC361, CRC370, and BRPC12-146.

MSS, microsatellite-stable; MSI, microsatellite-instable.

Control mice were singly housed with an enrichment hut, while mice assigned to the exercise group were singly housed with an exercise wheel and ran approximately 6 km per day

(Figure 34B,C). As the PDXs displayed different growth curves, a tumor size of approximately

1000 mm3 was used as the primary endpoint.

Figure 34: Histological and quantitative analysis of exercise in six diverse CRC PDX models. 108

(A) H&E slides of the six CRC PDX tumors. (B) Daily time (minutes) spent on the exercise wheel in the light and dark. (C) Average distance time run on the wheel by 3 representative PDX groups; each group ran approximately 5-8 km/day.

Three of the PDX models (CRC282, CRC344, and CRC370) were responsive to exercise as demonstrated by reduced tumor growth, while the remaining three models (CRC240, CRC361, and BRPC12-146) were non-responsive (Figure 35). Based on these results, we categorized the mice into four subcategories: PDXs (CRC282, CRC344, and CRC370) with response to exercise

(control - group 1, exercise - group 2) and PDXs (CRC240, CRC361, and BRPC12-146) with no response to exercise (control - group 3, and exercise – group 4).

109

Figure 35: Exercise has differential effects on tumor growth across six CRC PDX models.

(A-F) Longitudinal tumor growth (left) and body weight (right) for six PDX tumor models, in which three were exercise-responsive (A-C, CRC240, BRPC12–146 and CRC361) were exercise-responsive, while the remaining three (D-F, CRC282, CRC370 and CRC344) were not responsive to exercise treatment had no response to exercise treatment as determined by tumor growth.

To identify metabolic alterations induced by exercise treatment, we performed liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS) in our four groups of PDX samples to profile the levels of 205 metabolites. We performed an unsupervised hierarchical clustering analysis, which revealed widespread heterogeneity in metabolic

110

programming between the six PDX tumors and no distinct clustering that separated the four groups (Figure 35).

Figure 36: The effect of exercise on tumor metabolism in six different CRC PDX models.

Heat map of integrated intensity values of 205 metabolites that were detected in tumors from all 6 PDX groups, with metabolites grouped by unsupervised hierarchical clustering.

5.2.2 Exercise induces alterations in central carbon metabolism in exercise- responsive tumors

We next wanted to determine changes in metabolites between the exercise (groups 2 and

4) and sedentary (groups 1 and 3) tumors. Supervised analysis revealed 47 metabolites were

111

significantly altered by exercise; when these significantly altered metabolites were subject to unsupervised hierarchical clustering, a more defined pattern of metabolic alterations induced by exercise was observed (Figure 37A). Pathway analysis revealed significant alterations in nucleotide (purine and pyrimidine), vitamin B6, and amino acid metabolism, as well as the TCA cycle (Figure 37B).

Based on these results, we examined representative metabolites of each significantly altered pathway to determine which metabolites were contributing to the observed shifts. The

TCA cycle was observably altered by exercise, with the majority of intermediates (with the exception of succinate and glutamate) significantly reduced (Figure 37C). We additionally found a slight increase in glucose levels in PDX tumors extracted from exercised mice, while other glycolytic intermediates were either decreased or unchanged (Figure 37D). Interestingly, intratumoral pyruvate and lactate (the major byproducts of glycolysis) were not significantly elevated despite an increase in glucose uptake in the exercise group. It is possible that these intermediates were excreted from the tumor; alternatively, excess pyruvate and lactate can be converted into alternative energy sources such as acetyl-CoA (which feeds into the TCA cycle as well as fatty acid production), or into carbohydrates (such as glucose via gluconeogenesis) 307,308.

Overall, these results indicate that exercise induces global alterations in intratumoral metabolic pathways involved in central carbon metabolism.

Nucleotide metabolism was also found to be significantly impacted, as evidenced by alterations in both purine and pyrimidine metabolism (Figure 37B). Within purine synthesis intermediates, both adenine and the adenine-derivative ADP-ribose were increased while IMP and hypoxanthine were decreased (Figure 37E). Furthermore, metabolites involved in pyrimidine metabolism tended to be reduced, including uracil, the uracil-derived molecule UDP, and ureidosuccinic acid (Figure 37F). However, we also saw a significant increase in dihydrothymine,

112

while thymine and thymidine were not significantly altered (Figure 37F). While the functional significance of these data remains to be explored, these results indicate that exercise induces subtle but significant alterations in nucleotide precursor molecules, with generally opposite effects on purine- and pyrimidine-derived metabolites.

Finally, we found that exercise resulted in a small but significant decrease in tyrosine and proline, while levels of the other amino acids were not significantly impacted (Figure 37G).

Tyrosine is synthesized from phenylalanine and acts as a substrate for the production of catecholamines as well as the TCA intermediate fumarate, while proline can be synthesized from glutamate and has been shown to increase in response to tissue repair 309. Similarly, the acylcarnitines L-palmitoylcarnitine and stearoyl carnitine were found to be significantly reduced in tumors excised from exercised mice (Figure 37H); acylcarnitines play an important role in providing substrates for fatty acid β-oxidation and have been explored as biomarkers for altered mitochondrial function 310, providing additional evidence for alterations in the TCA cycle and the electron transport chain.

113

Figure 37: Tumors from exercised mice exhibit global alterations in metabolism compared to tumors from sedentary mice.

(A) Heat map of significantly altered metabolites between exercise and control groups in the six CRC PDX models. The significantly altered metabolites are displayed using unsupervised hierarchal clustering. Significantly altered metabolites (p < 0.05) determined using paired Student’s t-test. (B) Corresponding impacted pathways as determined by the list of 47 significantly altered metabolites. (C- F) Key metabolic pathways broken down by relative levels of individual metabolites. (G) Relative levels of individual amino acids. (H) Relative levels of individual acylcarnitines. Mean +/- s.e.m., * p < 0.05, paired Student’s t-test.

114

5.2.3 Exercise-responsive tumors demonstrate distinct metabolic profiles

To more closely examine the metabolic programming induced by exercise in responsive

PDX tumors, we analyzed the metabolic profiles of tumors from sedentary and exercise groups in the exercise-responsive PDXs (Groups 1 and 2) and identified significantly differential metabolites (Figure 38A) which primarily corresponded to amino acid and nucleotide metabolism

(Figure 38B). Tumors from exercised mice exhibited significantly increased levels of pyridoxal

(Figure 38C), a vitamin B6 molecule that plays a role in regulating intracellular homocysteine levels, and methylnicotinamide (Figure 38D), a niacin derivative that has been shown to inhibit choline transport 311. Surprisingly, while the TCA cycle was found to be highly impacted when exercise and sedentary groups were compared across all six PDX models, analysis of the three exercise-responsive models showed only a modest effect on the TCA cycle. Indeed, succinate was found to be the only significantly-altered TCA intermediate (Figure 38E); however, succinate serves as a proton donor to the electron transport chain, so alterations in intracellular levels can directly impact other critical energetic processes. We also observed similar patterns in nucleotide metabolism that were found in earlier analyses, such that adenine and ADP-ribose were elevated while uracil and dUMP were reduced (Figure 38F,G), suggesting that intrinsic variation in nucleotide metabolism may track with responsiveness to exercise in the tumors.

115

Figure 38: Exercise-responsive tumors exhibit distinct metabolic profiles.

(A) Heat map of integrated intensity values of the significantly altered metabolites between responsive control and exercise groups. Significantly altered metabolites (p < 0.05) determined using paired Student’s t-test. (B) Corresponding impacted pathways as determined by the 15 significantly altered metabolites. (C- G) Key metabolic pathways broken down by individual metabolites. Mean +/- s.e.m.; * p < 0.05, paired Student’s t-test.

5.2.4 Exercise-responsive tumors exhibit differential metabolic responses to exercise compared to non-responsive tumors.

Finally, we examined the differences in metabolic alterations induced by exercise between responsive (group 2) and non-responsive (group 4) PDX models and identified 58 metabolites that were differentially altered between the two groups (Figure 39A). Many of these metabolites were involved in pathways that regulate energy and redox balance, such as cysteine

116

metabolism, glycolysis, and fatty acid β-oxidation. Closer examination of cysteine-related metabolites showed that taurine and hypotaurine were slightly increased in exercise-responsive tumors, while they were reduced in their non-responsive counterparts (Figure 39B); the significance of this differential alteration in cysteine metabolism is unclear, but could indicate that exercise-non-responsive tumors are able to use these downstream metabolites to help shuttle intracellular energy stores more efficiently.

ADP and ATP tended to be reduced in both exercise-responsive and exercise-non- responsive tumors, although this reduction was found to only reach significance in non- responsive tumors (Figure 39C). We also found that both exercise groups exhibited increased glucose levels, potentially reflective of changes to mitochondrial metabolism (Figure 39D).

However, when we compared glycolytic intermediates, we observed that while both exercise- responsive and exercise-non-responsive tumors showed reductions in similar glycolytic intermediates, these reductions were more pronounced in the exercise-responsive tumors (Figure

39D). Oxidative stress, as assessed by the ratio of reduced to oxidized glutathione, was unchanged by exercise in both exercise-responsive and exercise-non-responsive tumors (Figure

39E). Interestingly, phosphocreatine was significantly reduced in the exercise-responsive group while essentially unchanged in the exercise-non-responsive group (Figure 39F). While the functional implications of these alterations remain to be elucidated, together these preliminary data suggest that exercise-responsive tumors manifest through differential responses in energy metabolism with creatine metabolites possibly leading to biomarkers that may identify this response.

117

Figure 39: Exercise-responsive tumors exhibit differential metabolic responses to exercise compared to non-responsive tumors.

(A) Heat map composed of fold changes (exercise versus control) of the 24 significantly altered metabolites between exercise-responsive and exercise non-responsive PDX models, displayed using unsupervised hierarchical clustering. Significantly altered metabolites (p < 0.05) determined using paired Student’s t-test. (B-D) Key metabolic pathways broken down by individual metabolites, segregated into four experimental subgroups. (E) Oxidative stress as measured by ratio of reduced to oxidized glutathione levels. Mean +/- s.e.m. (F) Relative fold changes in phosphocreatine levels between control and exercise groups, compared between exercise-responsive and exercise-nonresponsive tumors. Mean +/- s.e.m. * p < 0.05, paired Student’s t-test.

5.3 Discussion

Previous investigations of the effects of exercise on cancer patient outcome have mostly focused on reducing the risk or recurrence of tumor formation. These studies have found that having moderate to high levels of physical activity reduced the risk of developing colorectal cancer by age 30–40 years old, compared to those who led a more sedentary lifestyle 312,313;

118

similar findings have also been shown for the breast, lung, and prostate cancers 314-318.

Furthermore, the majority of these studies have suggested that this beneficial effect is independent of body mass index (BMI) and overall physical fitness. However, little information is known about how exercise may exert these antineotplastic effects in patients.

Preclinical murine models of cancer have previously shown that exercise can activate p53 and increase apoptosis 319, as well as reduce tumor growth by increasing microvessel density, vessel maturity, and perfusion, thereby decreasing intratumoral hypoxia 190. Interestingly, these findings were not found to be general, as some xenograft models did not display these alterations

320. Using our patient-derived xenograft model system to study the effects of exercise on tumor growth, our study provides the first molecular evidence that exercise can induce intratumoral metabolic disruptions independent of tumor growth in vivo.

Although our study exhibits notable limitations, including the use of a relatively low sample size of PDXs as well as potential drawbacks inherent to steady-state metabolomics methods, our initial findings suggest that not all cancer patients may benefit from vigorous exercise, as only three of the six PDX models (specifically CRC240, BRPC12-146, and CRC361) showed significant reductions in tumor growth. Overall responsiveness to exercise therapy was heterogeneous across these models, and metabolism in each of these models in the presence and absence of exercise was markedly different, lending further appreciation to our understanding of the metabolic diversity in tumors. Our metabolomics analysis revealed that exercise induced changes in intratumoral central carbon metabolism in each model, independent of whether exercise was able to inhibit tumor growth. Nevertheless, such an effect on mitochondrial metabolism was not able to produce therapeutic results in all models, consistent with current thinking that the requirements of mitochondrial metabolism vary widely across patient populations. These limitations provide further insight into the challenges of considering exercise

119

as a monotherapy and indicate the need for identification of potential biomarkers in patients that would benefit from this therapy.

Interestingly, metabolic alterations that were notably associated with responsiveness to exercise were related to nucleotide metabolism. Due to the limiting dependence of proliferating cells on nucleotide synthesis, antimetabolite chemotherapies targeting tumor metabolism remain some of the most effective anti-cancer treatments including 5-FU which is a frontline agent for advanced stage CRC (Section 1.5.1). Given that mitochondrial metabolism generates nucleotides through several branching pathways (such as aspartate, serine, and glutamine metabolism) 321-326, it is tempting to speculate that this link is differentially altered in the responsive tumors.

We further observed a significant decrease in phosphocreatine levels selectively in exercise-responsive tumors, further indicating altered energy metabolism. While it is possible that these observed differences may be a result of differential physiological host responsiveness to exercise independent of the inherent metabolism of the original PDX, this possibility is unlikely as mice in each PDX group performed similar levels of exercise (Figure 35). Therefore, the findings in this study suggest that intrinsic metabolic programming of energy balance could have a larger than previously appreciated role for the mitochondria in mediating tumor heterogeneity.

In summary, these observations suggest that exercise appears to act on mitochondrial metabolism in PDX models and may in some tumor contexts to have antineoplastic properties.

This may provide a therapeutic advantage for a subset of cancer patients, and could potentially be further enhanced when combined with other therapeutics or precision diets. Future studies aimed at defining the mechanism of how exercise interacts with other therapeutic modalities would be of great clinical interest.

120

5.4 Materials and Methods

5.4.1 Collection of patient tumors

Tissues were collected from patients with histologically confirmed colorectal cancer

(CRC) who had undergone complete surgical resections at the Duke University Medical Center between October 24, 2007 and June 9, 2011 under a Duke IRB protocol (Pro0002435). All participants provided written informed consent to participate in the study. Tumor tissues frozen in optimal cutting temperature (OCT) medium at the time of surgery were processed as hematoxylin and eosin (H&E, Sigma Aldrich) stained sections as previously described 284,285 and histologically characterized by a board-certified pathologist.

5.4.2 Generation of patient-derived xenografts

All animal studies were performed at Duke University under an Institutional Animal Care and Use Committee (IACUC) approved protocol. Patient-derived xenografts (PDXs) of colorectal cancer explants were generated as described previously 220,221. Six surgically resected colorectal cancer samples (CRC240, CRC282, CRC361, CRC370, and BRPC12-146) were minced and injected subcutaneously into the flanks of 8-week-old NOD.CB17-PrkdcSCID-J mice (The

Jackson Laboratory). Tumors were measured 2–3 times per week using a vernier caliper, and volumes were calculated using the formula, �=(�2×�)2V=(L2×W)2 (L = longest diameter, W = shortest diameter)) until tumors reached ~ 1000 mm3. Tumors were then harvested and serially re-implanted until stable PDXs were established with a minimum of three generations. Harvested tumor tissues were frozen in optimal cutting temperature (OCT) medium and processed as H&E stained sections as previously described 284,285 and histologically characterized by a board-certified pathologist.

121

5.4.3 Physical activity studies

Patient-derived xenografts (PDXs) of colorectal cancer used in the studies were developed by subcutaneous injection of homogeneous suspensions of previously developed PDXs and monitored until tumor formation began as described above. Mice in the control group were individually housed with an enrichment hut. Mice in the exercise group were singly housed with a wheel with a magnetic sensor to determine the distance each mouse traveled daily during the course of the experiment. Tumors were measured 2-3 times per week, and volumes were calculated by the equation �=(�2×�)2V=(L2×W)2. For every independent PDX study, each group (control group or exercise group) contained 4–10 mice (Figure 35). Endpoint of the study was determined using time to reach tumor volume of ~ 1000 mm3 or tumor ulceration. When tumors reached the endpoint, mice were euthanized under CO2 for 5 minutes. Tumors were harvested immediately (within 1 minute after euthanization), snap frozen in liquid nitrogen, and stored at − 80 °C.

5.4.4 Metabolite extraction

Tumor tissue (4 mg per tumor) was homogenized using a magnetized automatic homogenizer in 700 μL of ice-cold extraction solvent (80% methanol/water) and centrifuged for

10 minutes at 20,000 g at 4 °C. The resulting supernatants were transferred and split into two new

Eppendorf tubes, and the solvent for each sample was evaporated in a Speed Vacuum. For polar metabolite analysis, the metabolite extracts were dissolved in 15 μL HPLC-grade water and

15 μL methanol/acetonitrile (1:1 v/v) (LC-MS optima grade, Thermo Scientific). Samples were then centrifuged for 10 minutes at 20,000 g at 4 °C and supernatants were transferred to liquid chromatography vials. An injection volume of 5 μL was used for polar metabolite analysis.

122

5.4.5 Liquid chromatography

Ultimate 3000 HPLC (Dionex) with an Xbridge amide column (100 × 2.1 mm i.d.,

3.5 μm; Waters) is coupled to Q Exactive-Mass spectrometer (QE-MS, Thermo Scientific) for metabolite separation and detection at room temperature. The mobile phase A reagent is composed of 20 mM ammonium acetate and 15 mM ammonium hydroxide in 3% acetonitrile in

HPLC-grade water (pH 9.0), while the mobile phase B reagent is acetonitrile. All solvents are

LC-MS grade, purchased from Fischer Scientific. The flow rate used was 0.15 ml/min from 0 to

10 minutes and 15–20 minutes, and 0.3 ml/min from 10.5–14.5 minutes. The linear gradient was as follows: 0 minutes 85% B; 1.5 minutes 85% B, 5.5 minutes 35% B; 10 minutes 35% B,

10.5 minutes 25% B, 14.5 minutes 35% B, 15 minutes 85% B, and 20 minutes 85% B.

5.4.6 Mass spectrometry

The Q Exactive MS (Thermo Scientific) is outfitted with a heated electrospray ionization probe (HESI) with the following parameters: evaporation temperature, 120°C; sheath gas, 30; auxiliary gas, 10; sweep gas, 3; spray voltage, 3.6 kV for positive mode and 2.5 kV for negative mode. Capillary temperature was set at 320 °C and S-lens was 55. A full scan range was set at 60 to 900 (m/z), with the resolution set to 70,000. The maximum injection time (max IT) was 200 ms. Automated gain control (AGC) was targeted at 3.0 x 103 ions.

5.4.7 Peak extraction and metabolomics data analysis

integrated peak intensities generated from the raw data were used. Hierarchical clustering and heat maps were generated using Morpheus software (Broad Institute, https://software.broadinstitute.org/morpheus/). For hierarchical clustering, Spearman correlation parameters were implemented for row and column parameters. Pathway enrichment analysis was conducted using MetaboAnalyst 3.0 software (www.metaboanalyst.ca/faces/home.xhtml); briefly, metabolite identifications from the human metabolome database (HMDB IDs) from the metabolites that were significantly enriched [greater than log2(fold change) with p < 0.05] were inputted. The pathway library used was Homo sapiens and Fisher’s exact test was used for over- representation analysis. Other quantitation and statistics were calculated using GraphPad Prism software. P-values were calculated using Student’s t-test.

124

6. Conclusion

6.1 Summary

Within this dissertation, I have described how environmental factors contribute to metabolic reprogramming in relation to defined genetic profiles. I characterize how cancer cells from diverse genetic origins respond to environmental nutrient availability, as well as how environmentally-driven alterations in energetic demand impact metabolic reprogramming in healthy and malignant tissue.

Cancer cell metabolism and behavior are known to be highly driven by both individual genetic events 327 as well as tissue-of-origin 52 to varying degrees between tumors. However, it has become increasingly appreciated that nutrient availability can also exert substantial if not greater consequences on these processes 50,212,328. Given recent reports of targetable vulnerabilities induced by the deletion of the methionine salvage enzyme MTAP 93-95, I examined a panel of diverse cell lines with or without homozygous deletion of MTAP in altered environmental compositions. I demonstrated that metabolic responsiveness to environmental restriction of methionine or other one-carbon nutrients is highly heterogeneous, and is surprisingly not predicted by either MTAP deletion or tissue of origin. Furthermore, restriction of methionine as well as cysteine was sufficient to abrogate the metabolic signature that was reported to be necessary for the targetable vulnerability in MTAP-deleted cells. Given the high variability of methionine in human plasma 54, these findings further illuminate the growing importance of examining environmental context during the development of potential clinical therapies. This work was recently published in Science Advances 329.

Providing additional support for the significance of nutrient availability on cancer cell biology, I contributed to the characterization of metabolic consequences induced by methionine restriction (MR). In collaboration with multiple groups, fellow members of my laboratory found 125

that MR alone or in combination with the chemotherapy 5-FU significantly inhibited tumor growth in PDX colorectal cancer murine models. I conducted extensive nutrient supplementation experiments in cell culture to ultimately demonstrate that MR in combination with 5-FU exerts cytotoxic effects in colorectal cancer cells through disruption of nucleotide synthesis and redox balance, findings which supported metabolomic profiles generated from PDX tumors. This work is published in Nature 216, and I subsequently led efforts for an invited review article on dietary methionine in health and cancer in Nature Reviews Cancer that was recently published 330 to accompany our research article.

I further examined how environmental factors beyond dietary composition could impact metabolic processes in diverse settings. Using either pharmacological or physiological intervention, I investigated metabolic reprogramming in response to environmentally-driven alterations in energetic demand. I demonstrated that the cardiac glycoside digoxin, through on- target inhibition of the Na+/K+ ATPase, exerts substantial disruption to central carbon metabolism and other energy-related metabolic processes in both healthy and malignant tissue. Furthermore, I showed that the antineoplastic effects of digoxin are associated with a shift in the tumor microenvironmental landscape, as well as differential transcriptional programming of metabolic processes in both tumor cells and immune infiltrates. This work, which will hopefully lead to new initiatives for clinical repurposing of digoxin in cancer treatment, is in the final stages of preparation for initial peer review.

Finally, I additionally demonstrated that exercise impacts these energy-related metabolic processes in six distinct PDX models of colorectal cancer, although antineoplastic effects were achieved in only three of these models. My results suggest that although these central carbon metabolic pathways are impacted in both responsive and non-responsive PDX tumors, the ATP- recycling metabolite phosphocreatine was significantly altered solely in exercise-responsive

126

tumors and thus may potentially serve as a novel determinant of therapeutic response to exercise in patients. This work is published in Cancer & Metabolism 331.

This collective body of work provides a unique characterization of the heterogeneity in cancer resulting from a multitude of gene-environment interactions, and sheds light on potential therapeutic approaches of circumventing this metabolic heterogeneity.

6.2 Place in Current Research

As discussed in detail in Section 1.4.2, methionine plays an important role in a host of essential cellular functions including methylation reactions, redox balance, autophagy, macromolecule and protein synthesis, and nucleotide synthesis via donation to folate metabolism

(Figure 3). My colleagues and I have extensively characterized the metabolic consequences of

MR in cancer, as well as provided evidence that its cytotoxic effects are at least in part mediated through disruption to redox and nucleotide maintenance, which are even more pronounced in combination with 5-FU. However, the specific cellular components mediating this activity still remain to be determined.

To examine this more closely, in collaboration with members of Dr. Kris Wood’s laboratory I performed genome-wide CRISPR (clustered regularly interspaced short palindromic repeats) knockout screens in HCT116 and CRC119 colorectal cells, using the previously- described two-vector GeCKO v2 libraries 332. I cultured cells in four different culture conditions: control or methionine-restricted (100 or 10 µM, respectively) medium, treated with either vehicle

(DMSO) or 2.5 µM 5-FU. After five-six weeks, I collected the cells and had their DNA sequenced the Duke GCB Sequencing and Genomic Technologies Core to compare the relative barcode expression profiles of the remaining cells with those expressed in pre-treated cells. While the analysis of these experiments remains ongoing in collaboration with fellow members of my laboratory, specifically Dr. Zhengtao Xiao and Dr. Xia Gao, we believe these analyses will shed 127

light on which (if any) specific factors are essential for mediating the cytotoxic effects of MR, and whether any of the factors essential for cytotoxic activity are overlapping between MR and 5-

FU. It is our hope that these results will not only provide additional mechanistic characterization of MR in cancer cell biology, but potentially identify novel therapeutic targets as well.

6.3 Future Considerations

The work I have detailed thus far illustrates the substantial role of the environment in mediating genetically-driven programs of cellular metabolism, which contributes to our knowledge of the extensive heterogeneity in cancer and creates additional avenues of exploration for future characterization and development of novel treatment strategies.

Using MTAP deletion and one-carbon nutrient restriction as a model to investigate gene- environment interaction, I have demonstrated that cancer cell responsiveness to nutrient availability is exceptionally variable, and that this heterogeneity is largely not predicted by either the status of metabolic genes or tissue-of-origin in cell culture. Importantly, I found that both methionine and cysteine restriction were sufficient to abrogate the accumulation of MTA in these cells, thereby eliminating the targetable vulnerability (i.e. PRMT5) that had been previously identified 93,95. Given that methionine is one of the most variable amino acids in human plasma and that plasma methionine levels functionally correlate with methylation-dependent metabolic products 54, it will need to be determined in physiological settings whether variations in dietary methionine can impact the efficacy of PRMT5 inhibitors. Furthermore, it has yet to be determined whether MTAP-deleted patient tumors exhibit similar metabolic profiles as those found in preclinical studies; in particular, it will be important to examine whether factors such as originating tissue or dietary composition are associated with physiologically different metabolic programs in these tumors as has been found in other contexts 42,52. Additionally, MTA was

128

recently identified as a plasma biomarker for sepsis 333, indicating its potential to also serve as a biomarker for MTAP-deleted tumors and should be examined in further clinical investigations.

My findings that digoxin alters metabolism in both healthy and malignant tissue as well as reprograms the tumor microenvironment provides a myriad of interesting avenues to explore.

For example, my study provides the first characterization of non-toxic metabolic disruption in diverse tissues (most notably cardiac tissue) from digoxin treatment; given that digoxin is widely prescribed for heart arrythmia or failure, this finding is of particular interest to the cardiology field and warrants further characterization of tissue-specific metabolic adaptation to acute as well as chronic digoxin treatment. Additionally, while our results showing that acute digoxin treatment impacts intratumoral immune infiltrate cell populations in our allograft model are intriguing, it is presently unclear what the functional consequences of this shift may be and whether it may play a causal role in mediating tumor growth inhibition. To gain a better understanding of these immunological consequences, additional studies will need to investigate the longitudinal impact of digoxin on immune cell populations at multiple stages of tumor development. It still remains to be determined whether the observed metabolic reprogramming of distinct myeloid populations is a result of adjacent tumor cell death or is instead a direct consequence of immune cell- autonomous exposure to digoxin; furthermore, it is also unknown what the functional implications of either of these two possibilities might be in patient populations. These considerations will be important to investigate before determining the contexts in which digoxin would be clinically beneficial.

Lastly, although the observation that exercise exhibits markedly variable antineoplastic activity across six different colorectal PDX models was relatively unexpected, these results are ultimately in line with my findings in Chapter 2 illustrating the exceptional heterogeneity of cancer cell metabolism, which largely extends beyond originating tissue or individual alterations

129

in oncogenic drivers. The finding that central carbon metabolism was impacted by exercise in both responsive and non-responsive tumors suggests that the intrinsic metabolic programming of individual patient tumors create variable metabolic dependencies and thus differential responsiveness to environmental stressors. It is intriguing that one of the most differential alterations between the responsive and non-responsive groups in response to exercise was the substantial decrease in phosphocreatine levels, which was also found to be a major metabolic consequence of digoxin-induced cytotoxicity. An interesting question for future investigations would thus be whether tumor sensitivity to exercise correlates with Na+/K+ ATPase subunit expression and/or activity, or by extension if digoxin could sensitize non-responsive tumors to exercise. Finally, from a physiological perspective it would be fascinating to determine either clinically or epidemiologically whether exercise ability is impacted by digoxin treatment.

6.4 Final Remarks

Although genetics are known to play a role in shaping cancer cell metabolic reprogramming, it is becoming increasingly appreciated that numerous factors such as tissue type and nutrition, can also have significant consequences on the metabolic state of cancer cells. The intersection of these factors creates a high degree of variability between tumors, making the identification of patient populations that would benefit from specific therapies challenging.

Elucidating how inherent characteristics such as genetic status impact responsiveness of tumors as well as healthy tissue to environmental factors (such as dietary composition and lifestyle behaviors) will be critical in the development of novel precision treatment strategies to optimize patient outcome as well as to promote overall health.

130

References

1 Pavlova, N. N. & Thompson, C. B. The Emerging Hallmarks of Cancer Metabolism. Cell Metab 23, 27-47, doi:10.1016/j.cmet.2015.12.006 (2016).

2 Warburg, O., Wind, F. & Negelein, E. The Metabolism of Tumors in the Body. J Gen Physiol 8, 519-530, doi:10.1085/jgp.8.6.519 (1927).

3 Berg, J. M., Tymoczko, J. L., Stryer, L. & Stryer, L. Biochemistry. 6th edn, (W.H. Freeman, 2007).

4 Lehninger, A. L., Nelson, D. L. & Cox, M. M. Lehninger principles of biochemistry. 6th edn, (W.H. Freeman, 2013).

5 Pfeiffer, T., Schuster, S. & Bonhoeffer, S. Cooperation and competition in the evolution of ATP-producing pathways. Science 292, 504-507, doi:10.1126/science.1058079 (2001).

6 Shestov, A. A. et al. Quantitative determinants of aerobic glycolysis identify flux through the enzyme GAPDH as a limiting step. Elife 3, doi:10.7554/eLife.03342 (2014).

7 Warburg, O. On the origin of cancer cells. Science 123, 309-314, doi:10.1126/science.123.3191.309 (1956).

8 Liberti, M. V. & Locasale, J. W. The Warburg Effect: How Does it Benefit Cancer Cells? Trends Biochem Sci 41, 211-218, doi:10.1016/j.tibs.2015.12.001 (2016).

9 Berger, A. How does it work? Positron emission tomography. BMJ 326, 1449, doi:10.1136/bmj.326.7404.1449 (2003).

10 Wise, D. R. et al. Myc regulates a transcriptional program that stimulates mitochondrial glutaminolysis and leads to glutamine addiction. Proc Natl Acad Sci U S A 105, 18782- 18787, doi:10.1073/pnas.0810199105 (2008).

11 Gao, P. et al. c-Myc suppression of miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism. Nature 458, 762-765, doi:10.1038/nature07823 (2009).

12 Son, J. et al. Glutamine supports pancreatic cancer growth through a KRAS-regulated metabolic pathway. Nature 496, 101-105, doi:10.1038/nature12040 (2013).

13 Boroughs, L. K. & DeBerardinis, R. J. Metabolic pathways promoting cancer cell survival and growth. Nat Cell Biol 17, 351-359, doi:10.1038/ncb3124 (2015).

131

14 Degenhardt, K. et al. Autophagy promotes tumor cell survival and restricts necrosis, inflammation, and tumorigenesis. Cancer Cell 10, 51-64, doi:10.1016/j.ccr.2006.06.001 (2006).

15 Yang, S. et al. Pancreatic cancers require autophagy for tumor growth. Genes Dev 25, 717-729, doi:10.1101/gad.2016111 (2011).

16 Welbourne, T. C. Ammonia production and glutamine incorporation into glutathione in the functioning rat kidney. Can J Biochem 57, 233-237 (1979).

17 Jiang, L. et al. Reductive carboxylation supports redox homeostasis during anchorage- independent growth. Nature 532, 255-258, doi:10.1038/nature17393 (2016).

18 DeBerardinis, R. J. et al. Beyond aerobic glycolysis: transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. Proc Natl Acad Sci U S A 104, 19345-19350, doi:10.1073/pnas.0709747104 (2007).

19 DeBerardinis, R. J., Lum, J. J., Hatzivassiliou, G. & Thompson, C. B. The biology of cancer: metabolic reprogramming fuels cell growth and proliferation. Cell Metab 7, 11- 20, doi:10.1016/j.cmet.2007.10.002 (2008).

20 Gross, M. I. et al. Antitumor activity of the glutaminase inhibitor CB-839 in triple- negative breast cancer. Mol Cancer Ther 13, 890-901, doi:10.1158/1535-7163.MCT-13- 0870 (2014).

21 Xiang, Y. et al. Targeted inhibition of tumor-specific glutaminase diminishes cell- autonomous tumorigenesis. J Clin Invest 125, 2293-2306, doi:10.1172/JCI75836 (2015).

22 Kung, H. N., Marks, J. R. & Chi, J. T. Glutamine synthetase is a genetic determinant of cell type-specific glutamine independence in breast epithelia. PLoS Genet 7, e1002229, doi:10.1371/journal.pgen.1002229 (2011).

23 Herranz, D. et al. Metabolic reprogramming induces resistance to anti-NOTCH1 therapies in T cell acute lymphoblastic leukemia. Nat Med 21, 1182-1189, doi:10.1038/nm.3955 (2015).

24 MacFarlane, A. J. et al. Cytoplasmic serine hydroxymethyltransferase regulates the metabolic partitioning of methylenetetrahydrofolate but is not essential in mice. J Biol Chem 283, 25846-25853, doi:10.1074/jbc.M802671200 (2008).

25 Locasale, J. W. Serine, glycine and one-carbon units: cancer metabolism in full circle. Nat Rev Cancer 13, 572-583, doi:10.1038/nrc3557 (2013).

26 Locasale, J. W. et al. Phosphoglycerate dehydrogenase diverts glycolytic flux and contributes to oncogenesis. Nat Genet 43, 869-874, doi:10.1038/ng.890 (2011).

27 Wang, Q. et al. Rational Design of Selective Allosteric Inhibitors of PHGDH and Serine Synthesis with Anti-tumor Activity. Cell Chem Biol 24, 55-65, doi:10.1016/j.chembiol.2016.11.013 (2017). 132

28 Mullarky, E., Lairson, L. L., Cantley, L. C. & Lyssiotis, C. A. A novel small-molecule inhibitor of 3-phosphoglycerate dehydrogenase. Mol Cell Oncol 3, e1164280, doi:10.1080/23723556.2016.1164280 (2016).

29 Reid, M. A. et al. Serine synthesis through PHGDH coordinates nucleotide levels by maintaining central carbon metabolism. Nat Commun 9, 5442, doi:10.1038/s41467-018- 07868-6 (2018).

30 Dillon, B. J. et al. Incidence and distribution of argininosuccinate synthetase deficiency in human cancers: a method for identifying cancers sensitive to arginine deprivation. Cancer 100, 826-833, doi:10.1002/cncr.20057 (2004).

31 Grohmann, U. & Bronte, V. Control of immune response by amino acid metabolism. Immunol Rev 236, 243-264, doi:10.1111/j.1600-065X.2010.00915.x (2010).

32 Mellor, A. L. & Munn, D. H. IDO expression by dendritic cells: tolerance and tryptophan catabolism. Nat Rev Immunol 4, 762-774, doi:10.1038/nri1457 (2004).

33 Mayers, J. R. et al. Elevation of circulating branched-chain amino acids is an early event in human pancreatic adenocarcinoma development. Nat Med 20, 1193-1198, doi:10.1038/nm.3686 (2014).

34 Laplante, M. & Sabatini, D. M. mTOR signaling in growth control and disease. Cell 149, 274-293, doi:10.1016/j.cell.2012.03.017 (2012).

35 Hensley, C. T. et al. Metabolic Heterogeneity in Human Lung Tumors. Cell 164, 681- 694, doi:10.1016/j.cell.2015.12.034 (2016).

36 Yuneva, M. O. et al. The metabolic profile of tumors depends on both the responsible genetic lesion and tissue type. Cell Metab 15, 157-170, doi:10.1016/j.cmet.2011.12.015 (2012).

37 Maher, E. A. et al. Metabolism of [U-13 C]glucose in human brain tumors in vivo. NMR Biomed 25, 1234-1244, doi:10.1002/nbm.2794 (2012).

38 Liu, X., Romero, I. L., Litchfield, L. M., Lengyel, E. & Locasale, J. W. Metformin Targets Central Carbon Metabolism and Reveals Mitochondrial Requirements in Human Cancers. Cell Metab 24, 728-739, doi:10.1016/j.cmet.2016.09.005 (2016).

39 Weinberg, S. E. & Chandel, N. S. Targeting mitochondria metabolism for cancer therapy. Nat Chem Biol 11, 9-15, doi:10.1038/nchembio.1712 (2015).

40 Liberti, M. V. et al. A Predictive Model for Selective Targeting of the Warburg Effect through GAPDH Inhibition with a Natural Product. Cell Metab 26, 648-659 e648, doi:10.1016/j.cmet.2017.08.017 (2017).

41 Davidson, S. M. et al. Environment Impacts the Metabolic Dependencies of Ras-Driven Non-Small Cell Lung Cancer. Cell Metab 23, 517-528, doi:10.1016/j.cmet.2016.01.007 (2016). 133

42 Courtney, K. D. et al. Isotope Tracing of Human Clear Cell Renal Cell Carcinomas Demonstrates Suppressed Glucose Oxidation In Vivo. Cell Metab, doi:10.1016/j.cmet.2018.07.020 (2018).

43 Levine, A. J. & Puzio-Kuter, A. M. The control of the metabolic switch in cancers by oncogenes and tumor suppressor genes. Science 330, 1340-1344, doi:10.1126/science.1193494 (2010).

44 Kamphorst, J. J. et al. Human pancreatic cancer tumors are nutrient poor and tumor cells actively scavenge extracellular protein. Cancer Res 75, 544-553, doi:10.1158/0008- 5472.CAN-14-2211 (2015).

45 Tran, T. Q. et al. Tumor-associated mutant p53 promotes cancer cell survival upon glutamine deprivation through p21 induction. Oncogene 36, 1991-2001, doi:10.1038/onc.2016.360 (2017).

46 Reid, M. A. et al. The B55alpha subunit of PP2A drives a p53-dependent metabolic adaptation to glutamine deprivation. Mol Cell 50, 200-211, doi:10.1016/j.molcel.2013.02.008 (2013).

47 Muir, A. et al. Environmental cystine drives glutamine anaplerosis and sensitizes cancer cells to glutaminase inhibition. Elife 6, doi:10.7554/eLife.27713 (2017).

48 Kanarek, N. et al. Histidine catabolism is a major determinant of methotrexate sensitivity. Nature 559, 632-636, doi:10.1038/s41586-018-0316-7 (2018).

49 Knott, S. R. V. et al. Asparagine bioavailability governs metastasis in a model of breast cancer. Nature 554, 378-381, doi:10.1038/nature25465 (2018).

50 Maddocks, O. D. K. et al. Modulating the therapeutic response of tumours to dietary serine and glycine starvation. Nature 544, 372-376, doi:10.1038/nature22056 (2017).

51 Murakami, T. et al. Recombinant methioninase effectively targets a Ewing's sarcoma in a patient-derived orthotopic xenograft (PDOX) nude-mouse model. Oncotarget 8, 35630- 35638, doi:10.18632/oncotarget.15823 (2017).

52 Mayers, J. R. et al. Tissue of origin dictates branched-chain amino acid metabolism in mutant Kras-driven cancers. Science 353, 1161-1165, doi:10.1126/science.aaf5171 (2016).

53 Pan, M. et al. Regional glutamine deficiency in tumours promotes dedifferentiation through inhibition of histone demethylation. Nat Cell Biol 18, 1090-1101, doi:10.1038/ncb3410 (2016).

54 Mentch, S. J. et al. Histone Methylation Dynamics and Gene Regulation Occur through the Sensing of One-Carbon Metabolism. Cell Metab 22, 861-873, doi:10.1016/j.cmet.2015.08.024 (2015).

134

55 Du, Y. P. et al. Assessment of the effect of betaine on p16 and c-myc DNA methylation and mRNA expression in a chemical induced rat liver cancer model. BMC Cancer 9, 261, doi:10.1186/1471-2407-9-261 (2009).

56 Mehedint, M. G., Niculescu, M. D., Craciunescu, C. N. & Zeisel, S. H. Choline deficiency alters global histone methylation and epigenetic marking at the Re1 site of the calbindin 1 gene. FASEB J 24, 184-195, doi:10.1096/fj.09-140145 (2010).

57 Mentch, S. J. & Locasale, J. W. One-carbon metabolism and epigenetics: understanding the specificity. Ann N Y Acad Sci 1363, 91-98, doi:10.1111/nyas.12956 (2016).

58 Gu, X. et al. SAMTOR is an S-adenosylmethionine sensor for the mTORC1 pathway. Science 358, 813-818, doi:10.1126/science.aao3265 (2017).

59 Pegg, A. E. Functions of Polyamines in Mammals. J Biol Chem 291, 14904-14912, doi:10.1074/jbc.R116.731661 (2016).

60 Ye, C., Sutter, B. M., Wang, Y., Kuang, Z. & Tu, B. P. A Metabolic Function for Phospholipid and Histone Methylation. Mol Cell 66, 180-193 e188, doi:10.1016/j.molcel.2017.02.026 (2017).

61 Hosios, A. M. & Vander Heiden, M. G. The redox requirements of proliferating mammalian cells. J Biol Chem 293, 7490-7498, doi:10.1074/jbc.TM117.000239 (2018).

62 Hine, C. et al. Endogenous hydrogen sulfide production is essential for dietary restriction benefits. Cell 160, 132-144, doi:10.1016/j.cell.2014.11.048 (2015).

63 Sutter, B. M., Wu, X., Laxman, S. & Tu, B. P. Methionine inhibits autophagy and promotes growth by inducing the SAM-responsive methylation of PP2A. Cell 154, 403- 415, doi:10.1016/j.cell.2013.06.041 (2013).

64 Mazor, K. M. et al. Effects of single amino acid deficiency on mRNA translation are markedly different for methionine versus leucine. Sci Rep 8, 8076, doi:10.1038/s41598- 018-26254-2 (2018).

65 Ruckenstuhl, C. et al. Lifespan extension by methionine restriction requires autophagy- dependent vacuolar acidification. PLoS Genet 10, e1004347, doi:10.1371/journal.pgen.1004347 (2014).

66 Lee, B. C. et al. Methionine restriction extends lifespan of Drosophila melanogaster under conditions of low amino-acid status. Nat Commun 5, 3592, doi:10.1038/ncomms4592 (2014).

67 Cabreiro, F. et al. Metformin retards aging in C. elegans by altering microbial folate and methionine metabolism. Cell 153, 228-239, doi:10.1016/j.cell.2013.02.035 (2013).

68 Miller, R. A. et al. Methionine-deficient diet extends mouse lifespan, slows immune and lens aging, alters glucose, T4, IGF-I and insulin levels, and increases hepatocyte MIF

135

levels and stress resistance. Aging Cell 4, 119-125, doi:10.1111/j.1474- 9726.2005.00152.x (2005).

69 Barcena, C. et al. Methionine Restriction Extends Lifespan in Progeroid Mice and Alters Lipid and Bile Acid Metabolism. Cell Rep 24, 2392-2403, doi:10.1016/j.celrep.2018.07.089 (2018).

70 Orentreich, N., Matias, J. R., DeFelice, A. & Zimmerman, J. A. Low methionine ingestion by rats extends life span. J Nutr 123, 269-274, doi:10.1093/jn/123.2.269 (1993).

71 Ables, G. P., Perrone, C. E., Orentreich, D. & Orentreich, N. Methionine-restricted C57BL/6J mice are resistant to diet-induced obesity and insulin resistance but have low bone density. PLoS One 7, e51357, doi:10.1371/journal.pone.0051357 (2012).

72 Ables, G. P. et al. Dietary methionine restriction in mice elicits an adaptive cardiovascular response to hyperhomocysteinemia. Sci Rep 5, 8886, doi:10.1038/srep08886 (2015).

73 Stone, K. P., Wanders, D., Orgeron, M., Cortez, C. C. & Gettys, T. W. Mechanisms of increased in vivo insulin sensitivity by dietary methionine restriction in mice. Diabetes 63, 3721-3733, doi:10.2337/db14-0464 (2014).

74 Luckerath, K. et al. 11C-Methionine-PET: a novel and sensitive tool for monitoring of early response to treatment in multiple myeloma. Oncotarget 6, 8418-8429, doi:10.18632/oncotarget.3053 (2015).

75 Glaudemans, A. W. et al. Value of 11C-methionine PET in imaging brain tumours and metastases. Eur J Nucl Med Mol Imaging 40, 615-635, doi:10.1007/s00259-012-2295-5 (2013).

76 Sugimura, T., Birnbaum, S. M., Winitz, M. & Greenstein, J. P. Quantitative nutritional studies with water-soluble, chemically defined diets. VII. Nitrogen balance in normal and tumor-bearing rats following forced feeding. Arch Biochem Biophys 81, 439-447 (1959).

77 Guo, H. et al. Therapeutic tumor-specific cell cycle block induced by methionine starvation in vivo. Cancer Res 53, 5676-5679 (1993).

78 Poirson-Bichat, F. et al. Methionine deprivation and methionine analogs inhibit cell proliferation and growth of human xenografted gliomas. Life Sci 60, 919-931 (1997).

79 Poirson-Bichat, F., Goncalves, R. A., Miccoli, L., Dutrillaux, B. & Poupon, M. F. Methionine depletion enhances the antitumoral efficacy of cytotoxic agents in drug- resistant human tumor xenografts. Clin Cancer Res 6, 643-653 (2000).

80 Goseki, N. et al. Antitumor effect of methionine-depleting total parenteral nutrition with doxorubicin administration on Yoshida sarcoma-bearing rats. Cancer 69, 1865-1872 (1992).

136

81 Strekalova, E., Malin, D., Good, D. M. & Cryns, V. L. Methionine Deprivation Induces a Targetable Vulnerability in Triple-Negative Breast Cancer Cells by Enhancing TRAIL Receptor-2 Expression. Clin Cancer Res 21, 2780-2791, doi:10.1158/1078-0432.CCR- 14-2792 (2015).

82 Jeon, H. et al. Methionine deprivation suppresses triple-negative breast cancer metastasis in vitro and in vivo. Oncotarget 7, 67223-67234, doi:10.18632/oncotarget.11615 (2016).

83 Beroukhim, R. et al. The landscape of somatic copy-number alteration across human cancers. Nature 463, 899-905, doi:10.1038/nature08822 (2010).

84 Parsons, D. W. et al. An integrated genomic analysis of human glioblastoma multiforme. Science 321, 1807-1812, doi:10.1126/science.1164382 (2008).

85 Behrmann, I. et al. Characterization of methylthioadenosin phosphorylase (MTAP) expression in malignant melanoma. Am J Pathol 163, 683-690, doi:10.1016/S0002- 9440(10)63695-4 (2003).

86 Illei, P. B., Ladanyi, M., Rusch, V. W. & Zakowski, M. F. The use of CDKN2A deletion as a diagnostic marker for malignant mesothelioma in body cavity effusions. Cancer 99, 51-56, doi:10.1002/cncr.10923 (2003).

87 Subhi, A. L. et al. Methylthioadenosine phosphorylase regulates ornithine decarboxylase by production of downstream metabolites. J Biol Chem 278, 49868-49873, doi:10.1074/jbc.M308451200 (2003).

88 Zhang, H., Chen, Z. H. & Savarese, T. M. Codeletion of the genes for p16INK4, methylthioadenosine phosphorylase, interferon-alpha1, interferon-beta1, and other 9p21 markers in human malignant cell lines. Cancer Genet Cytogenet 86, 22-28 (1996).

89 Ishii, N. et al. Frequent co-alterations of TP53, p16/CDKN2A, p14ARF, PTEN tumor suppressor genes in human glioma cell lines. Brain Pathol 9, 469-479 (1999).

90 Hellerbrand, C. et al. Promoter-hypermethylation is causing functional relevant downregulation of methylthioadenosine phosphorylase (MTAP) expression in hepatocellular carcinoma. Carcinogenesis 27, 64-72, doi:10.1093/carcin/bgi201 (2006).

91 Schmid, M. et al. Homozygous deletions of methylthioadenosine phosphorylase (MTAP) are more frequent than p16INK4A (CDKN2) homozygous deletions in primary non-small cell lung cancers (NSCLC). Oncogene 17, 2669-2675, doi:10.1038/sj.onc.1202205 (1998).

92 Christopher, S. A., Diegelman, P., Porter, C. W. & Kruger, W. D. Methylthioadenosine phosphorylase, a gene frequently codeleted with p16(cdkN2a/ARF), acts as a tumor suppressor in a breast cancer cell line. Cancer Res 62, 6639-6644 (2002).

137

93 Kryukov, G. V. et al. MTAP deletion confers enhanced dependency on the PRMT5 arginine methyltransferase in cancer cells. Science 351, 1214-1218, doi:10.1126/science.aad5214 (2016).

94 Marjon, K. et al. MTAP Deletions in Cancer Create Vulnerability to Targeting of the MAT2A/PRMT5/RIOK1 Axis. Cell Rep 15, 574-587, doi:10.1016/j.celrep.2016.03.043 (2016).

95 Mavrakis, K. J. et al. Disordered methionine metabolism in MTAP/CDKN2A-deleted cancers leads to dependence on PRMT5. Science 351, 1208-1213, doi:10.1126/science.aad5944 (2016).

96 Chen, Z. H., Olopade, O. I. & Savarese, T. M. Expression of methylthioadenosine phosphorylase cDNA in p16-, MTAP- malignant cells: restoration of methylthioadenosine phosphorylase-dependent salvage pathways and alterations of sensitivity to inhibitors of purine de novo synthesis. Mol Pharmacol 52, 903-911 (1997).

97 Li, W. et al. Status of methylthioadenosine phosphorylase and its impact on cellular response to L-alanosine and methylmercaptopurine riboside in human soft tissue sarcoma cells. Oncol Res 14, 373-379 (2004).

98 Hori, H. et al. Methylthioadenosine phosphorylase cDNA transfection alters sensitivity to depletion of purine and methionine in A549 lung cancer cells. Cancer Res 56, 5653-5658 (1996).

99 Chung, J., Karkhanis, V., Baiocchi, R. A. & Sif, S. Protein Arginine Methyltransferase 5 (PRMT5) Promotes Survival of Lymphoma Cells via Activation of WNT/beta- CATENIN and AKT/GSK3beta Proliferative Signaling. J Biol Chem, doi:10.1074/jbc.RA119.007640 (2019).

100 Amano, Y. et al. Expression of protein arginine methyltransferase-5 in oral squamous cell carcinoma and its significance in epithelial-to-mesenchymal transition. Pathol Int 68, 359-366, doi:10.1111/pin.12666 (2018).

101 Serio, J. et al. The PAF complex regulation of Prmt5 facilitates the progression and maintenance of MLL fusion leukemia. Oncogene 37, 450-460, doi:10.1038/onc.2017.337 (2018).

102 Wang, Z. et al. Methionine is a metabolic dependency of tumor-initiating cells. Nat Med, doi:10.1038/s41591-019-0423-5 (2019).

103 Ulanovskaya, O. A., Zuhl, A. M. & Cravatt, B. F. NNMT promotes epigenetic remodeling in cancer by creating a metabolic methylation sink. Nat Chem Biol 9, 300- 306, doi:10.1038/nchembio.1204 (2013).

104 Eckert, M. A. et al. Proteomics reveals NNMT as a master metabolic regulator of cancer- associated fibroblasts. Nature, doi:10.1038/s41586-019-1173-8 (2019).

138

105 Kraus, D. et al. Nicotinamide N-methyltransferase knockdown protects against diet- induced obesity. Nature 508, 258-262, doi:10.1038/nature13198 (2014).

106 Pohjanpelto, P. Putrescine transport is greatly increased in human fibroblasts initiated to proliferate. J Cell Biol 68, 512-520 (1976).

107 Mohan, R. R. et al. Overexpression of ornithine decarboxylase in prostate cancer and prostatic fluid in humans. Clin Cancer Res 5, 143-147 (1999).

108 Hoshino, Y. et al. Ornithine decarboxylase activity as a prognostic marker for colorectal cancer. Fukushima J Med Sci 53, 1-9 (2007).

109 Deng, W. et al. Role of ornithine decarboxylase in breast cancer. Acta Biochim Biophys Sin (Shanghai) 40, 235-243 (2008).

110 He, W. et al. Targeting ornithine decarboxylase (ODC) inhibits esophageal squamous cell carcinoma progression. NPJ Precis Oncol 1, 13, doi:10.1038/s41698-017-0014-1 (2017).

111 O'Brien, T. G., Megosh, L. C., Gilliard, G. & Soler, A. P. Ornithine decarboxylase overexpression is a sufficient condition for tumor promotion in mouse skin. Cancer Res 57, 2630-2637 (1997).

112 Dai, F. et al. Extracellular polyamines-induced proliferation and migration of cancer cells by ODC, SSAT, and Akt1-mediated pathway. Anticancer Drugs 28, 457-464, doi:10.1097/CAD.0000000000000465 (2017).

113 Koseki, J. et al. A Trans-omics Mathematical Analysis Reveals Novel Functions of the Ornithine Metabolic Pathway in Cancer Stem Cells. Sci Rep 6, 20726, doi:10.1038/srep20726 (2016).

114 Hayashi, K. et al. Visualization and characterization of cancer stem-like cells in cervical cancer. Int J Oncol 45, 2468-2474, doi:10.3892/ijo.2014.2670 (2014).

115 Adikrisna, R. et al. Identification of pancreatic cancer stem cells and selective toxicity of chemotherapeutic agents. Gastroenterology 143, 234-245 e237, doi:10.1053/j.gastro.2012.03.054 (2012).

116 Subhi, A. L. et al. Loss of methylthioadenosine phosphorylase and elevated ornithine decarboxylase is common in pancreatic cancer. Clin Cancer Res 10, 7290-7296, doi:10.1158/1078-0432.CCR-04-0972 (2004).

117 Scuoppo, C. et al. A tumour suppressor network relying on the polyamine-hypusine axis. Nature 487, 244-248, doi:10.1038/nature11126 (2012).

118 Zabala-Letona, A. et al. mTORC1-dependent AMD1 regulation sustains polyamine metabolism in prostate cancer. Nature 547, 109-113, doi:10.1038/nature22964 (2017).

119 Fox, J. T. & Stover, P. J. Folate-mediated one-carbon metabolism. Vitam Horm 79, 1-44, doi:10.1016/S0083-6729(08)00401-9 (2008). 139

120 Visentin, M., Zhao, R. & Goldman, I. D. The antifolates. Hematol Oncol Clin North Am 26, 629-648, ix, doi:10.1016/j.hoc.2012.02.002 (2012).

121 Fabre, I., Fabre, G. & Goldman, I. D. Polyglutamylation, an important element in methotrexate cytotoxicity and selectivity in tumor versus murine granulocytic progenitor cells in vitro. Cancer Res 44, 3190-3195 (1984).

122 Koizumi, S. et al. Reversal of methotrexate cytotoxicity to human bone marrow cells and leukemic K562 cells by leucovorin: methotrexate polyglutamates formation as a possible important factor. Jpn J Cancer Res 81, 1162-1167 (1990).

123 Pratt, C. B., Roberts, D., Shanks, E. C. & Warmath, E. L. Clinical trials and pharmacokinetics of intermittent high-dose methotrexate-"leucovorin rescue" for children with malignant tumors. Cancer Res 34, 3326-3331 (1974).

124 Howard, S. C., McCormick, J., Pui, C. H., Buddington, R. K. & Harvey, R. D. Preventing and Managing Toxicities of High-Dose Methotrexate. Oncologist 21, 1471-1482, doi:10.1634/theoncologist.2015-0164 (2016).

125 Gonen, N. & Assaraf, Y. G. Antifolates in cancer therapy: structure, activity and mechanisms of drug resistance. Drug Resist Updat 15, 183-210, doi:10.1016/j.drup.2012.07.002 (2012).

126 Cohen, M. H., Justice, R. & Pazdur, R. Approval summary: pemetrexed in the initial treatment of advanced/metastatic non-small cell lung cancer. Oncologist 14, 930-935, doi:10.1634/theoncologist.2009-0092 (2009).

127 Jarmula, A. Antifolate inhibitors of thymidylate synthase as anticancer drugs. Mini Rev Med Chem 10, 1211-1222 (2010).

128 Longley, D. B., Harkin, D. P. & Johnston, P. G. 5-fluorouracil: mechanisms of action and clinical strategies. Nat Rev Cancer 3, 330-338, doi:10.1038/nrc1074 (2003).

129 Ewald, B., Sampath, D. & Plunkett, W. Nucleoside analogs: molecular mechanisms signaling cell death. Oncogene 27, 6522-6537, doi:10.1038/onc.2008.316 (2008).

130 Elnaggar, M., Giovannetti, E. & Peters, G. J. Molecular targets of gemcitabine action: rationale for development of novel drugs and drug combinations. Curr Pharm Des 18, 2811-2829 (2012).

131 Stresemann, C. & Lyko, F. Modes of action of the DNA methyltransferase inhibitors azacytidine and decitabine. Int J Cancer 123, 8-13, doi:10.1002/ijc.23607 (2008).

132 Malik, P. & Cashen, A. F. Decitabine in the treatment of acute myeloid leukemia in elderly patients. Cancer Manag Res 6, 53-61, doi:10.2147/CMAR.S40600 (2014).

133 Poirson-Bichat, F., Gonfalone, G., Bras-Goncalves, R. A., Dutrillaux, B. & Poupon, M. F. Growth of methionine-dependent human prostate cancer (PC-3) is inhibited by ethionine combined with methionine starvation. Br J Cancer 75, 1605-1612 (1997). 140

134 Kreis, W. & Hession, C. Isolation and purification of L-methionine-alpha-deamino- gamma-mercaptomethane-lyase (L-methioninase) from Clostridium sporogenes. Cancer Res 33, 1862-1865 (1973).

135 Tan, Y., Zavala, J., Sr., Xu, M., Zavala, J., Jr. & Hoffman, R. M. Serum methionine depletion without side effects by methioninase in metastatic breast cancer patients. Anticancer Res 16, 3937-3942 (1996).

136 Tan, Y. et al. Recombinant methioninase infusion reduces the biochemical endpoint of serum methionine with minimal toxicity in high-stage cancer patients. Anticancer Res 17, 3857-3860 (1997).

137 Kawaguchi, K. et al. Recombinant methioninase (rMETase) is an effective therapeutic for BRAF-V600E-negative as well as -positive melanoma in patient-derived orthotopic xenograft (PDOX) mouse models. Oncotarget 9, 915-923, doi:10.18632/oncotarget.23185 (2018).

138 Igarashi, K. et al. Effective Metabolic Targeting of Human Osteosarcoma Cells In Vitro and in Orthotopic Nude-mouse Models with Recombinant Methioninase. Anticancer Res 37, 4807-4812, doi:10.21873/anticanres.11887 (2017).

139 Hoffman, R. M. et al. Pilot Phase I Clinical Trial of Methioninase on High-Stage Cancer Patients: Rapid Depletion of Circulating Methionine. Methods Mol Biol 1866, 231-242, doi:10.1007/978-1-4939-8796-2_17 (2019).

140 Lubin, M. & Lubin, A. Selective killing of tumors deficient in methylthioadenosine phosphorylase: a novel strategy. PLoS One 4, e5735, doi:10.1371/journal.pone.0005735 (2009).

141 Tang, B., Lee, H. O., An, S. S., Cai, K. Q. & Kruger, W. D. Specific Targeting of MTAP- Deleted Tumors with a Combination of 2'-Fluoroadenine and 5'-Methylthioadenosine. Cancer Res 78, 4386-4395, doi:10.1158/0008-5472.CAN-18-0814 (2018).

142 Kindler, H. L., Burris, H. A., 3rd, Sandler, A. B. & Oliff, I. A. A phase II multicenter study of L-alanosine, a potent inhibitor of adenine biosynthesis, in patients with MTAP- deficient cancer. Invest New Drugs 27, 75-81, doi:10.1007/s10637-008-9160-1 (2009).

143 Metcalf, B. W. et al. Catalytic irreversible inhibition of mammalian ornithine decarboxylase (E.C.4.1.1.17) by substrate and product analogs. Journal of the American Chemical Society 100, 2551-2553, doi:10.1021/ja00476a050 (1978).

144 Luk, G. D., Civin, C. I., Weissman, R. M. & Baylin, S. B. Ornithine decarboxylase: essential in proliferation but not differentiation of human promyelocytic leukemia cells. Science 216, 75-77 (1982).

145 Meyskens, F. L., Kingsley, E. M., Glattke, T., Loescher, L. & Booth, A. A phase II study of alpha-difluoromethylornithine (DFMO) for the treatment of metastatic melanoma. Invest New Drugs 4, 257-262 (1986).

141

146 Abeloff, M. D. et al. Phase II trials of alpha-difluoromethylornithine, an inhibitor of polyamine synthesis, in advanced small cell lung cancer and colon cancer. Cancer Treat Rep 70, 843-845 (1986).

147 Seiler, N., Delcros, J. G. & Moulinoux, J. P. Polyamine transport in mammalian cells. An update. Int J Biochem Cell Biol 28, 843-861 (1996).

148 Sholler, G. L. S. et al. Maintenance DFMO Increases Survival in High Risk Neuroblastoma. Sci Rep 8, 14445, doi:10.1038/s41598-018-32659-w (2018).

149 Hogarty, M. D. et al. ODC1 is a critical determinant of MYCN oncogenesis and a therapeutic target in neuroblastoma. Cancer Res 68, 9735-9745, doi:10.1158/0008- 5472.CAN-07-6866 (2008).

150 Wagner, A. J., Meyers, C., Laimins, L. A. & Hay, N. c-Myc induces the expression and activity of ornithine decarboxylase. Cell Growth Differ 4, 879-883 (1993).

151 Hayes, C. S. et al. Polyamine-blocking therapy reverses immunosuppression in the tumor microenvironment. Cancer Immunol Res 2, 274-285, doi:10.1158/2326-6066.CIR-13- 0120-T (2014).

152 Tang, B., Kadariya, Y., Chen, Y., Slifker, M. & Kruger, W. D. Expression of MTAP inhibits tumor-related phenotypes in HT1080 cells via a mechanism unrelated to its enzymatic function. G3 (Bethesda) 5, 35-44, doi:10.1534/g3.114.014555 (2014).

153 Bai, J. et al. Identification of a natural inhibitor of methionine adenosyltransferase 2A regulating one-carbon metabolism in keratinocytes. EBioMedicine 39, 575-590, doi:10.1016/j.ebiom.2018.12.036 (2019).

154 Quinlan, C. L. et al. Targeting S-adenosylmethionine biosynthesis with a novel allosteric inhibitor of Mat2A. Nat Chem Biol 13, 785-792, doi:10.1038/nchembio.2384 (2017).

155 Warder, S. E. et al. Discovery, identification, and characterization of candidate pharmacodynamic markers of methionine aminopeptidase-2 inhibition. J Proteome Res 7, 4807-4820, doi:10.1021/pr800388p (2008).

156 Burkey, B. F. et al. Preclinical Efficacy and Safety of the Novel Antidiabetic, Antiobesity MetAP2 Inhibitor ZGN-1061. J Pharmacol Exp Ther 365, 301-313, doi:10.1124/jpet.117.246272 (2018).

157 Jiao, M., Liu, G., Xue, Y. & Ding, C. Computational drug repositioning for cancer therapeutics. Curr Top Med Chem 15, 767-775 (2015).

158 Ashburn, T. T. & Thor, K. B. Drug repositioning: identifying and developing new uses for existing drugs. Nat Rev Drug Discov 3, 673-683, doi:10.1038/nrd1468 (2004).

159 Inzucchi, S. E. et al. Management of hyperglycaemia in type 2 diabetes, 2015: a patient- centred approach. Update to a position statement of the American Diabetes Association

142

and the European Association for the Study of Diabetes. Diabetologia 58, 429-442, doi:10.1007/s00125-014-3460-0 (2015).

160 Madiraju, A. K. et al. Metformin suppresses gluconeogenesis by inhibiting mitochondrial glycerophosphate dehydrogenase. Nature 510, 542-546, doi:10.1038/nature13270 (2014).

161 Matthaei, S. & Greten, H. Evidence that metformin ameliorates cellular insulin-resistance by potentiating insulin-induced translocation of glucose transporters to the plasma membrane. Diabete Metab 17, 150-158 (1991).

162 Pernicova, I. & Korbonits, M. Metformin--mode of action and clinical implications for diabetes and cancer. Nat Rev Endocrinol 10, 143-156, doi:10.1038/nrendo.2013.256 (2014).

163 Wheaton, W. W. et al. Metformin inhibits mitochondrial complex I of cancer cells to reduce tumorigenesis. Elife 3, e02242, doi:10.7554/eLife.02242 (2014).

164 Zhou, G. et al. Role of AMP-activated protein kinase in mechanism of metformin action. J Clin Invest 108, 1167-1174, doi:10.1172/JCI13505 (2001).

165 Xie, Y. et al. Regulation of insulin-like growth factor signaling by metformin in endometrial cancer cells. Oncol Lett 8, 1993-1999, doi:10.3892/ol.2014.2466 (2014).

166 Porporato, P. E., Filigheddu, N., Pedro, J. M. B., Kroemer, G. & Galluzzi, L. Mitochondrial metabolism and cancer. Cell Res 28, 265-280, doi:10.1038/cr.2017.155 (2018).

167 Birsoy, K. et al. Metabolic determinants of cancer cell sensitivity to glucose limitation and biguanides. Nature 508, 108-112, doi:10.1038/nature13110 (2014).

168 Libby, G. et al. New users of metformin are at low risk of incident cancer: a cohort study among people with type 2 diabetes. Diabetes Care 32, 1620-1625, doi:10.2337/dc08- 2175 (2009).

169 Chae, Y. K. et al. Repurposing metformin for cancer treatment: current clinical studies. Oncotarget 7, 40767-40780, doi:10.18632/oncotarget.8194 (2016).

170 Newby, L. K. et al. Early statin initiation and outcomes in patients with acute coronary syndromes. JAMA 287, 3087-3095 (2002).

171 Nissen, S. E. et al. Effect of very high-intensity statin therapy on regression of coronary atherosclerosis: the ASTEROID trial. JAMA 295, 1556-1565, doi:10.1001/jama.295.13.jpc60002 (2006).

172 Chan, K. K., Oza, A. M. & Siu, L. L. The statins as anticancer agents. Clin Cancer Res 9, 10-19 (2003).

173 Boudreau, D. M., Yu, O. & Johnson, J. Statin use and cancer risk: a comprehensive review. Expert Opin Drug Saf 9, 603-621, doi:10.1517/14740331003662620 (2010). 143

174 Ahn, J. et al. Prediagnostic total and high-density lipoprotein cholesterol and risk of cancer. Cancer Epidemiol Biomarkers Prev 18, 2814-2821, doi:10.1158/1055-9965.EPI- 08-1248 (2009).

175 Jankowski, J., Attwood, S., deCaestecker, J. & Barr, H. Aspirin in the primary prevention of vascular disease. Lancet 374, 877-878; author reply 879, doi:10.1016/S0140- 6736(09)61621-4 (2009).

176 Rothwell, P. M. et al. Short-term effects of daily aspirin on cancer incidence, mortality, and non-vascular death: analysis of the time course of risks and benefits in 51 randomised controlled trials. Lancet 379, 1602-1612, doi:10.1016/S0140-6736(11)61720-0 (2012).

177 Cao, Y. et al. Population-wide Impact of Long-term Use of Aspirin and the Risk for Cancer. JAMA Oncol 2, 762-769, doi:10.1001/jamaoncol.2015.6396 (2016).

178 Hawley, S. A. et al. The ancient drug salicylate directly activates AMP-activated protein kinase. Science 336, 918-922, doi:10.1126/science.1215327 (2012).

179 Thun, M. J., Jacobs, E. J. & Patrono, C. The role of aspirin in cancer prevention. Nat Rev Clin Oncol 9, 259-267, doi:10.1038/nrclinonc.2011.199 (2012).

180 Oshima, M. & Taketo, M. M. COX selectivity and animal models for colon cancer. Curr Pharm Des 8, 1021-1034 (2002).

181 Rothwell, P. M. et al. Effect of daily aspirin on risk of cancer metastasis: a study of incident cancers during randomised controlled trials. Lancet 379, 1591-1601, doi:10.1016/S0140-6736(12)60209-8 (2012).

182 Gay, L. J. & Felding-Habermann, B. Contribution of platelets to tumour metastasis. Nat Rev Cancer 11, 123-134, doi:10.1038/nrc3004 (2011).

183 Williams, J. D. et al. Tumor Autonomous Effects of Vitamin D Deficiency Promote Breast Cancer Metastasis. Endocrinology 157, 1341-1347, doi:10.1210/en.2015-2036 (2016).

184 Yun, J. et al. Vitamin C selectively kills KRAS and BRAF mutant colorectal cancer cells by targeting GAPDH. Science 350, 1391-1396, doi:10.1126/science.aaa5004 (2015).

185 Sajadian, S. O. et al. Vitamin C enhances epigenetic modifications induced by 5- azacytidine and cell cycle arrest in the hepatocellular carcinoma cell lines HLE and Huh7. Clin Epigenetics 8, 46, doi:10.1186/s13148-016-0213-6 (2016).

186 Liu, M. et al. Vitamin C increases viral mimicry induced by 5-aza-2'-deoxycytidine. Proc Natl Acad Sci U S A 113, 10238-10244, doi:10.1073/pnas.1612262113 (2016).

187 in PDQ Cancer Information Summaries (2002).

188 Courneya, K. S. Exercise in cancer survivors: an overview of research. Med Sci Sports Exerc 35, 1846-1852, doi:10.1249/01.MSS.0000093622.41587.B6 (2003). 144

189 Jones, L. W. et al. Modulation of circulating angiogenic factors and tumor biology by aerobic training in breast cancer patients receiving neoadjuvant chemotherapy. Cancer Prev Res (Phila) 6, 925-937, doi:10.1158/1940-6207.CAPR-12-0416 (2013).

190 Betof, A. S. et al. Modulation of murine breast tumor vascularity, hypoxia and chemotherapeutic response by exercise. J Natl Cancer Inst 107, doi:10.1093/jnci/djv040 (2015).

191 Bhaskaran, K. et al. Body-mass index and risk of 22 specific cancers: a population-based cohort study of 5.24 million UK adults. Lancet 384, 755-765, doi:10.1016/S0140- 6736(14)60892-8 (2014).

192 Calle, E. E. & Kaaks, R. Overweight, obesity and cancer: epidemiological evidence and proposed mechanisms. Nat Rev Cancer 4, 579-591, doi:10.1038/nrc1408 (2004).

193 Font-Burgada, J., Sun, B. & Karin, M. Obesity and Cancer: The Oil that Feeds the Flame. Cell Metab 23, 48-62, doi:10.1016/j.cmet.2015.12.015 (2016).

194 Nieman, K. M. et al. Adipocytes promote ovarian cancer metastasis and provide energy for rapid tumor growth. Nat Med 17, 1498-1503, doi:10.1038/nm.2492 (2011).

195 Nieman, K. M., Romero, I. L., Van Houten, B. & Lengyel, E. Adipose tissue and adipocytes support tumorigenesis and metastasis. Biochim Biophys Acta 1831, 1533- 1541, doi:10.1016/j.bbalip.2013.02.010 (2013).

196 de Oliveira, S. F. et al. Characterization of MTAP Gene Expression in Breast Cancer Patients and Cell Lines. PLoS One 11, e0145647, doi:10.1371/journal.pone.0145647 (2016).

197 Liu, X., Ser, Z., Cluntun, A. A., Mentch, S. J. & Locasale, J. W. A strategy for sensitive, large scale quantitative metabolomics. J Vis Exp, doi:10.3791/51358 (2014).

198 Liu, X., Ser, Z. & Locasale, J. W. Development and quantitative evaluation of a high- resolution metabolomics technology. Anal Chem 86, 2175-2184, doi:10.1021/ac403845u (2014).

199 Liu, X. & Locasale, J. W. Metabolomics: A Primer. Trends Biochem Sci 42, 274-284, doi:10.1016/j.tibs.2017.01.004 (2017).

200 Gao, X. et al. Serine Availability Influences Mitochondrial Dynamics and Function through Lipid Metabolism. Cell Rep 22, 3507-3520, doi:10.1016/j.celrep.2018.03.017 (2018).

201 Petryszak, R. et al. Expression Atlas update--an integrated database of gene and protein expression in humans, animals and plants. Nucleic Acids Res 44, D746-752, doi:10.1093/nar/gkv1045 (2016).

145

202 Vander Heiden, M. G., Cantley, L. C. & Thompson, C. B. Understanding the Warburg effect: the metabolic requirements of cell proliferation. Science 324, 1029-1033, doi:10.1126/science.1160809 (2009).

203 Nagarajan, A., Malvi, P. & Wajapeyee, N. Oncogene-directed alterations in cancer cell metabolism. Trends Cancer 2, 365-377, doi:10.1016/j.trecan.2016.06.002 (2016).

204 Commisso, C. et al. Macropinocytosis of protein is an amino acid supply route in Ras- transformed cells. Nature 497, 633-637, doi:10.1038/nature12138 (2013).

205 Le Li, Y. M., Lina Zhao, Lijia Li, Jinjun Wu, Mengjia Zhao, Wenjing Du, Li Yu, Peng Jiang. p53 regulation of ammonia metabolism through urea cycle controls polyamine biosynthesis. Nature, 1476-4687 (2019).

206 Reid, M. A., Dai, Z. & Locasale, J. W. The impact of cellular metabolism on chromatin dynamics and epigenetics. Nat Cell Biol 19, 1298-1306, doi:10.1038/ncb3629 (2017).

207 Dey, P. et al. Genomic deletion of malic enzyme 2 confers collateral lethality in pancreatic cancer. Nature 542, 119-123, doi:10.1038/nature21052 (2017).

208 Muller, F. L., Aquilanti, E. A. & DePinho, R. A. Collateral Lethality: A new therapeutic strategy in oncology. Trends Cancer 1, 161-173, doi:10.1016/j.trecan.2015.10.002 (2015).

209 Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43, e47, doi:10.1093/nar/gkv007 (2015).

210 Le, S., Josse, J. & Husson, F. FactoMineR: An R package for multivariate analysis. J Stat Softw 25, 1-18 (2008).

211 Thiele, I. et al. A community-driven global reconstruction of human metabolism. Nat Biotechnol 31, 419-425, doi:10.1038/nbt.2488 (2013).

212 Maddocks, O. D. et al. Serine starvation induces stress and p53-dependent metabolic remodelling in cancer cells. Nature 493, 542-546, doi:10.1038/nature11743 (2013).

213 Gravel, S. P. et al. Serine deprivation enhances antineoplastic activity of biguanides. Cancer Res 74, 7521-7533, doi:10.1158/0008-5472.CAN-14-2643-T (2014).

214 Agrawal, V., Alpini, S. E., Stone, E. M., Frenkel, E. P. & Frankel, A. E. Targeting methionine auxotrophy in cancer: discovery & exploration. Expert Opin Biol Ther 12, 53- 61, doi:10.1517/14712598.2012.636349 (2012).

215 Judde, J. G., Ellis, M. & Frost, P. Biochemical analysis of the role of transmethylation in the methionine dependence of tumor cells. Cancer Res 49, 4859-4865 (1989).

216 Gao, X. S., SM; Dai, Z; Reid, MA; Cooper, DE; Lu, M; Richie, JP; Ciccarella, A; Calcagnotto, A; Mikhael, PG; Mentch, SJ; Liu, J; Ables, G; Kirsch, DG; Hsu, DS; Nichenametla, SN; Locasale, JW. Dietary methionine restriction targets one carbon 146

metabolism in humans and produces broad therapeutic responses in cancer. Nature (2019).

217 Wanders, D. et al. Role of GCN2-Independent Signaling Through a Noncanonical PERK/NRF2 Pathway in the Physiological Responses to Dietary Methionine Restriction. Diabetes 65, 1499-1510, doi:10.2337/db15-1324 (2016).

218 Wanders, D. et al. FGF21 Mediates the Thermogenic and Insulin-Sensitizing Effects of Dietary Methionine Restriction but Not Its Effects on Hepatic Lipid Metabolism. Diabetes 66, 858-867, doi:10.2337/db16-1212 (2017).

219 Lien, E. C., Ghisolfi, L., Geck, R. C., Asara, J. M. & Toker, A. Oncogenic PI3K promotes methionine dependency in breast cancer cells through the cystine-glutamate antiporter xCT. Sci Signal 10, doi:10.1126/scisignal.aao6604 (2017).

220 Kim, M. K. et al. Characterization of an oxaliplatin sensitivity predictor in a preclinical murine model of colorectal cancer. Mol Cancer Ther 11, 1500-1509, doi:10.1158/1535- 7163.MCT-11-0937 (2012).

221 Uronis, J. M. et al. Histological and molecular evaluation of patient-derived colorectal cancer explants. PLoS One 7, e38422, doi:10.1371/journal.pone.0038422 (2012).

222 Udofot, O. et al. Pharmacokinetic, biodistribution and therapeutic efficacy of 5- fluorouracil-loaded pH-sensitive PEGylated liposomal nanoparticles in HCT-116 tumor bearing mouse. J Nat Sci 2 (2016).

223 Barry, W. H., Hasin, Y. & Smith, T. W. Sodium pump inhibition, enhanced calcium influx via sodium-calcium exchange, and positive inotropic response in cultured heart cells. Circ Res 56, 231-241, doi:10.1161/01.res.56.2.231 (1985).

224 Kaplan, J. H. Biochemistry of Na,K-ATPase. Annu Rev Biochem 71, 511-535, doi:10.1146/annurev.biochem.71.102201.141218 (2002).

225 Clausen, T. Potassium and sodium transport and pH regulation. Can J Physiol Pharmacol 70 Suppl, S219-222, doi:10.1139/y92-265 (1992).

226 Wright, E. M., Hirayama, B. A. & Loo, D. F. Active sugar transport in health and disease. J Intern Med 261, 32-43, doi:10.1111/j.1365-2796.2006.01746.x (2007).

227 Sibarov, D. A., Bolshakov, A. E., Abushik, P. A., Krivoi, II & Antonov, S. M. Na+,K+- ATPase functionally interacts with the plasma membrane Na+,Ca2+ exchanger to prevent Ca2+ overload and neuronal apoptosis in excitotoxic stress. J Pharmacol Exp Ther 343, 596-607, doi:10.1124/jpet.112.198341 (2012).

228 Clausen, M. V., Hilbers, F. & Poulsen, H. The Structure and Function of the Na,K- ATPase Isoforms in Health and Disease. Front Physiol 8, 371, doi:10.3389/fphys.2017.00371 (2017).

147

229 Katz, A. et al. Selectivity of digitalis glycosides for isoforms of human Na,K-ATPase. J Biol Chem 285, 19582-19592, doi:10.1074/jbc.M110.119248 (2010).

230 Kepp, O. et al. Anticancer activity of cardiac glycosides: At the frontier between cell- autonomous and immunological effects. Oncoimmunology 1, 1640-1642, doi:10.4161/onci.21684 (2012).

231 Slingerland, M., Cerella, C., Guchelaar, H. J., Diederich, M. & Gelderblom, H. Cardiac glycosides in cancer therapy: from preclinical investigations towards clinical trials. Invest New Drugs 31, 1087-1094, doi:10.1007/s10637-013-9984-1 (2013).

232 Eskiocak, U. et al. Synergistic effects of ion transporter and MAP kinase pathway inhibitors in melanoma. Nat Commun 7, 12336, doi:10.1038/ncomms12336 (2016).

233 Zhang, H. et al. Digoxin and other cardiac glycosides inhibit HIF-1alpha synthesis and block tumor growth. Proc Natl Acad Sci U S A 105, 19579-19586, doi:10.1073/pnas.0809763105 (2008).

234 Lopez-Lazaro, M. et al. Digitoxin inhibits the growth of cancer cell lines at concentrations commonly found in cardiac patients. J Nat Prod 68, 1642-1645, doi:10.1021/np050226l (2005).

235 Bielawski, K., Winnicka, K. & Bielawska, A. Inhibition of DNA topoisomerases I and II, and growth inhibition of breast cancer MCF-7 cells by ouabain, digoxin and proscillaridin A. Biol Pharm Bull 29, 1493-1497, doi:10.1248/bpb.29.1493 (2006).

236 Raynal, N. J. et al. Targeting Calcium Signaling Induces Epigenetic Reactivation of Tumor Suppressor Genes in Cancer. Cancer Res 76, 1494-1505, doi:10.1158/0008- 5472.CAN-14-2391 (2016).

237 Arispe, N., Diaz, J. C., Simakova, O. & Pollard, H. B. Heart failure drug digitoxin induces calcium uptake into cells by forming transmembrane calcium channels. Proc Natl Acad Sci U S A 105, 2610-2615, doi:10.1073/pnas.0712270105 (2008).

238 Yang, Q. et al. Cardiac glycosides inhibit TNF-alpha/NF-kappaB signaling by blocking recruitment of TNF receptor-associated death domain to the TNF receptor. Proc Natl Acad Sci U S A 102, 9631-9636, doi:10.1073/pnas.0504097102 (2005).

239 Ye, J., Chen, S. & Maniatis, T. Cardiac glycosides are potent inhibitors of interferon-beta gene expression. Nat Chem Biol 7, 25-33, doi:10.1038/nchembio.476 (2011).

240 Haas, M., Wang, H., Tian, J. & Xie, Z. Src-mediated inter-receptor cross-talk between the Na+/K+-ATPase and the epidermal growth factor receptor relays the signal from ouabain to mitogen-activated protein kinases. J Biol Chem 277, 18694-18702, doi:10.1074/jbc.M111357200 (2002).

148

241 Wu, J. et al. Cell signaling associated with Na(+)/K(+)-ATPase: activation of phosphatidylinositide 3-kinase IA/Akt by ouabain is independent of Src. Biochemistry 52, 9059-9067, doi:10.1021/bi4011804 (2013).

242 Prassas, I. et al. Digitoxin-induced cytotoxicity in cancer cells is mediated through distinct kinase and interferon signaling networks. Mol Cancer Ther 10, 2083-2093, doi:10.1158/1535-7163.MCT-11-0421 (2011).

243 Xie, Z. & Cai, T. Na+-K+--ATPase-mediated signal transduction: from protein interaction to cellular function. Mol Interv 3, 157-168, doi:10.1124/mi.3.3.157 (2003).

244 Racker, E. Bioenergetics and the problem of tumor growth. Am Sci 60, 56-63 (1972).

245 Howarth, C., Gleeson, P. & Attwell, D. Updated energy budgets for neural computation in the neocortex and cerebellum. J Cereb Blood Flow Metab 32, 1222-1232, doi:10.1038/jcbfm.2012.35 (2012).

246 Milligan, L. P. & McBride, B. W. Energy costs of ion pumping by animal tissues. J Nutr 115, 1374-1382, doi:10.1093/jn/115.10.1374 (1985).

247 Jong, C. J., Azuma, J. & Schaffer, S. Mechanism underlying the antioxidant activity of taurine: prevention of mitochondrial oxidant production. Amino Acids 42, 2223-2232, doi:10.1007/s00726-011-0962-7 (2012).

248 Kurosawa, Y. et al. Creatine supplementation enhances anaerobic ATP synthesis during a single 10 sec maximal handgrip exercise. Mol Cell Biochem 244, 105-112 (2003).

249 Kim, N. et al. Cardiac glycosides display selective efficacy for STK11 mutant lung cancer. Sci Rep 6, 29721, doi:10.1038/srep29721 (2016).

250 Calderon-Montano, J. M., Burgos-Moron, E. & Lopez-Lazaro, M. The in vivo antitumor activity of cardiac glycosides in mice xenografted with human cancer cells is probably an experimental artifact. Oncogene 33, 2947-2948, doi:10.1038/onc.2013.229 (2014).

251 Gupta, R. S., Chopra, A. & Stetsko, D. K. Cellular basis for the species differences in sensitivity to cardiac glycosides (digitalis). J Cell Physiol 127, 197-206, doi:10.1002/jcp.1041270202 (1986).

252 Calderon-Montano, J. M. et al. Evaluating the cancer therapeutic potential of cardiac glycosides. Biomed Res Int 2014, 794930, doi:10.1155/2014/794930 (2014).

253 Huang, L. et al. Systems biology-based drug repositioning identifies digoxin as a potential therapy for groups 3 and 4 medulloblastoma. Sci Transl Med 10, doi:10.1126/scitranslmed.aat0150 (2018).

254 Morton, C. L. & Houghton, P. J. Establishment of human tumor xenografts in immunodeficient mice. Nat Protoc 2, 247-250, doi:10.1038/nprot.2007.25 (2007).

149

255 Buque, A. & Galluzzi, L. Modeling Tumor Immunology and Immunotherapy in Mice. Trends Cancer 4, 599-601, doi:10.1016/j.trecan.2018.07.003 (2018).

256 Sukumar, M. et al. Inhibiting glycolytic metabolism enhances CD8+ T cell memory and antitumor function. J Clin Invest 123, 4479-4488, doi:10.1172/JCI69589 (2013).

257 Chang, C. H. et al. Posttranscriptional control of T cell effector function by aerobic glycolysis. Cell 153, 1239-1251, doi:10.1016/j.cell.2013.05.016 (2013).

258 Patel, C. H., Leone, R. D., Horton, M. R. & Powell, J. D. Targeting metabolism to regulate immune responses in autoimmunity and cancer. Nat Rev Drug Discov 18, 669- 688, doi:10.1038/s41573-019-0032-5 (2019).

259 Tabula Muris, C. et al. Single-cell transcriptomics of 20 mouse organs creates a Tabula Muris. Nature 562, 367-372, doi:10.1038/s41586-018-0590-4 (2018).

260 Wisdom, A. J. et al. Neutrophils promote tumor resistance to radiation therapy. Proc Natl Acad Sci U S A, doi:10.1073/pnas.1901562116 (2019).

261 Esparza-Molto, P. B. & Cuezva, J. M. The Role of Mitochondrial H(+)-ATP Synthase in Cancer. Front Oncol 8, 53, doi:10.3389/fonc.2018.00053 (2018).

262 Newsholme, P., Curi, R., Gordon, S. & Newsholme, E. A. Metabolism of glucose, glutamine, long-chain fatty acids and ketone bodies by murine macrophages. Biochem J 239, 121-125, doi:10.1042/bj2390121 (1986).

263 Newsholme, P., Gordon, S. & Newsholme, E. A. Rates of utilization and fates of glucose, glutamine, pyruvate, fatty acids and ketone bodies by mouse macrophages. Biochem J 242, 631-636, doi:10.1042/bj2420631 (1987).

264 Cantor, J. R. et al. Physiologic Medium Rewires Cellular Metabolism and Reveals Uric Acid as an Endogenous Inhibitor of UMP Synthase. Cell 169, 258-272 e217, doi:10.1016/j.cell.2017.03.023 (2017).

265 Gagnon, E. et al. Endoplasmic reticulum-mediated phagocytosis is a mechanism of entry into macrophages. Cell 110, 119-131, doi:10.1016/s0092-8674(02)00797-3 (2002).

266 Oh, J. et al. Endoplasmic reticulum stress controls M2 macrophage differentiation and foam cell formation. J Biol Chem 287, 11629-11641, doi:10.1074/jbc.M111.338673 (2012).

267 Erecinska, M. & Dagani, F. Relationships between the neuronal sodium/potassium pump and energy metabolism. Effects of K+, Na+, and adenosine triphosphate in isolated brain synaptosomes. J Gen Physiol 95, 591-616, doi:10.1085/jgp.95.4.591 (1990).

268 Bensaad, K. & Harris, A. L. Hypoxia and metabolism in cancer. Adv Exp Med Biol 772, 1-39, doi:10.1007/978-1-4614-5915-6_1 (2014).

150

269 Moghetti, P., Bacchi, E., Brangani, C., Dona, S. & Negri, C. Metabolic Effects of Exercise. Front Horm Res 47, 44-57, doi:10.1159/000445156 (2016).

270 Hennings, J. M. et al. Effect of mirtazapine on metabolism and energy substrate partitioning in healthy men. JCI Insight 4, doi:10.1172/jci.insight.123786 (2019).

271 Ballon, J. S. et al. Pathophysiology of drug induced weight and metabolic effects: findings from an RCT in healthy volunteers treated with olanzapine, iloperidone, or placebo. J Psychopharmacol 32, 533-540, doi:10.1177/0269881118754708 (2018).

272 Mills, E. J. et al. Low-dose aspirin and cancer mortality: a meta-analysis of randomized trials. Am J Med 125, 560-567, doi:10.1016/j.amjmed.2012.01.017 (2012).

273 Banerjee, M. et al. Na/K-ATPase Y260 Phosphorylation-mediated Src Regulation in Control of Aerobic Glycolysis and Tumor Growth. Sci Rep 8, 12322, doi:10.1038/s41598-018-29995-2 (2018).

274 Felippe Goncalves-de-Albuquerque, C., Ribeiro Silva, A., Ignacio da Silva, C., Caire Castro-Faria-Neto, H. & Burth, P. Na/K Pump and Beyond: Na/K-ATPase as a Modulator of Apoptosis and Autophagy. Molecules 22, doi:10.3390/molecules22040578 (2017).

275 Xie, Z. & Askari, A. Na(+)/K(+)-ATPase as a signal transducer. Eur J Biochem 269, 2434-2439, doi:10.1046/j.1432-1033.2002.02910.x (2002).

276 Epstein, T., Xu, L., Gillies, R. J. & Gatenby, R. A. Separation of metabolic supply and demand: aerobic glycolysis as a normal physiological response to fluctuating energetic demands in the membrane. Cancer Metab 2, 7, doi:10.1186/2049-3002-2-7 (2014).

277 Schaffer, S. W., Solodushko, V. & Kakhniashvili, D. Beneficial effect of taurine depletion on osmotic sodium and calcium loading during chemical hypoxia. Am J Physiol Cell Physiol 282, C1113-1120, doi:10.1152/ajpcell.00485.2001 (2002).

278 Xiao, Z., Dai, Z. & Locasale, J. W. Metabolic landscape of the tumor microenvironment at single cell resolution. Nat Commun 10, 3763, doi:10.1038/s41467-019-11738-0 (2019).

279 Najafi, M. et al. Macrophage polarity in cancer: A review. J Cell Biochem 120, 2756- 2765, doi:10.1002/jcb.27646 (2019).

280 Huh, J. R. et al. Digoxin and its derivatives suppress TH17 cell differentiation by antagonizing RORgammat activity. Nature 472, 486-490, doi:10.1038/nature09978 (2011).

281 Zhyvoloup, A. et al. Digoxin reveals a functional connection between HIV-1 integration preference and T-cell activation. PLoS Pathog 13, e1006460, doi:10.1371/journal.ppat.1006460 (2017).

151

282 Sen, T. et al. Targeting DNA Damage Response Promotes Antitumor Immunity through STING-Mediated T-cell Activation in Small Cell Lung Cancer. Cancer Discov 9, 646- 661, doi:10.1158/2159-8290.CD-18-1020 (2019).

283 Kuntz, E. M. et al. Targeting mitochondrial oxidative phosphorylation eradicates therapy- resistant chronic myeloid leukemia stem cells. Nat Med 23, 1234-1240, doi:10.1038/nm.4399 (2017).

284 Fischer, A. H., Jacobson, K. A., Rose, J. & Zeller, R. Paraffin embedding tissue samples for sectioning. CSH Protoc 2008, pdb prot4989, doi:10.1101/pdb.prot4989 (2008).

285 Fischer, A. H., Jacobson, K. A., Rose, J. & Zeller, R. Hematoxylin and eosin staining of tissue and cell sections. CSH Protoc 2008, pdb prot4986, doi:10.1101/pdb.prot4986 (2008).

286 Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21, doi:10.1093/bioinformatics/bts635 (2013).

287 Puram, S. V. et al. Single-Cell Transcriptomic Analysis of Primary and Metastatic Tumor Ecosystems in Head and Neck Cancer. Cell 171, 1611-1624 e1624, doi:10.1016/j.cell.2017.10.044 (2017).

288 Patel, A. P. et al. Single-cell RNA-seq highlights intratumoral heterogeneity in primary glioblastoma. Science 344, 1396-1401, doi:10.1126/science.1254257 (2014).

289 Venteicher, A. S. et al. Decoupling genetics, lineages, and microenvironment in IDH- mutant gliomas by single-cell RNA-seq. Science 355, doi:10.1126/science.aai8478 (2017).

290 Tirosh, I. et al. Dissecting the multicellular ecosystem of metastatic melanoma by single- cell RNA-seq. Science 352, 189-196, doi:10.1126/science.aad0501 (2016).

291 Kobak, D. & Berens, P. The art of using t-SNE for single-cell transcriptomics. bioRxiv (2019).

292 Heinonen, I. et al. Organ-specific physiological responses to acute physical exercise and long-term training in humans. Physiology (Bethesda) 29, 421-436, doi:10.1152/physiol.00067.2013 (2014).

293 Brown, J. C., Winters-Stone, K., Lee, A. & Schmitz, K. H. Cancer, physical activity, and exercise. Compr Physiol 2, 2775-2809, doi:10.1002/cphy.c120005 (2012).

294 Jones, L. W., Eves, N. D. & Scott, J. M. Bench-to-Bedside Approaches for Personalized Exercise Therapy in Cancer. American Society of Clinical Oncology educational book. American Society of Clinical Oncology. Meeting 37, 684-694, doi:10.14694/EDBK_173836 (2017).

152

295 Jones, L. W. Precision Oncology Framework for Investigation of Exercise As Treatment for Cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 33, 4134-4137, doi:10.1200/JCO.2015.62.7687 (2015).

296 Pedersen, L. et al. Voluntary Running Suppresses Tumor Growth through Epinephrine- and IL-6-Dependent NK Cell Mobilization and Redistribution. Cell Metab 23, 554-562, doi:10.1016/j.cmet.2016.01.011 (2016).

297 Hojman, P., Gehl, J., Christensen, J. F. & Pedersen, B. K. Molecular Mechanisms Linking Exercise to Cancer Prevention and Treatment. Cell Metab, doi:10.1016/j.cmet.2017.09.015 (2017).

298 Ruiz-Casado, A. et al. Exercise and the Hallmarks of Cancer. Trends Cancer 3, 423-441, doi:10.1016/j.trecan.2017.04.007 (2017).

299 Gan, Z., Fu, T., Kelly, D. P. & Vega, R. B. Skeletal muscle mitochondrial remodeling in exercise and diseases. Cell Res 28, 969-980, doi:10.1038/s41422-018-0078-7 (2018).

300 Kendall, S. D., Adam, S. J. & Counter, C. M. Genetically engineered human cancer models utilizing mammalian transgene expression. Cell Cycle 5, 1074-1079, doi:10.4161/cc.5.10.2734 (2006).

301 Kerbel, R. S. Human tumor xenografts as predictive preclinical models for anticancer drug activity in humans: better than commonly perceived-but they can be improved. Cancer Biol Ther 2, S134-139 (2003).

302 Fichtner, I. et al. Anticancer drug response and expression of molecular markers in early- passage xenotransplanted colon carcinomas. Eur J Cancer 40, 298-307 (2004).

303 Guenot, D. et al. Primary tumour genetic alterations and intra-tumoral heterogeneity are maintained in xenografts of human colon cancers showing chromosome instability. J Pathol 208, 643-652, doi:10.1002/path.1936 (2006).

304 Bertotti, A. et al. A molecularly annotated platform of patient-derived xenografts ("xenopatients") identifies HER2 as an effective therapeutic target in cetuximab-resistant colorectal cancer. Cancer discovery 1, 508-523, doi:10.1158/2159-8290.CD-11-0109 (2011).

305 Tentler, J. J. et al. Identification of predictive markers of response to the MEK1/2 inhibitor selumetinib (AZD6244) in K-ras-mutated colorectal cancer. Mol Cancer Ther 9, 3351-3362, doi:10.1158/1535-7163.MCT-10-0376 (2010).

306 Dangles-Marie, V. et al. Establishment of human colon cancer cell lines from fresh tumors versus xenografts: comparison of success rate and cell line features. Cancer Res 67, 398-407, doi:10.1158/0008-5472.CAN-06-0594 (2007).

153

307 Sonveaux, P. et al. Targeting lactate-fueled respiration selectively kills hypoxic tumor cells in mice. The Journal of clinical investigation 118, 3930-3942, doi:10.1172/JCI36843 (2008).

308 Pavlides, S. et al. The reverse Warburg effect: aerobic glycolysis in cancer associated fibroblasts and the tumor stroma. Cell cycle 8, 3984-4001, doi:10.4161/cc.8.23.10238 (2009).

309 Fukasawa, M., Bryant, S. M., Orita, H., Campeau, J. D. & DiZerega, G. S. Modulation of proline and glucosamine incorporation into tissue repair cells by peritoneal macrophages. J Surg Res 46, 166-171 (1989).

310 Muoio, D. M. Metabolic inflexibility: when mitochondrial indecision leads to metabolic gridlock. Cell 159, 1253-1262, doi:10.1016/j.cell.2014.11.034 (2014).

311 Koppen, A., Klein, J., Holler, T. & Loffelholz, K. Synergistic effect of nicotinamide and choline administration on extracellular choline levels in the brain. J Pharmacol Exp Ther 266, 720-725 (1993).

312 Slattery, M. L. Physical activity and colorectal cancer. Sports Med 34, 239-252 (2004).

313 Meyerhardt, J. A. et al. Physical activity and survival after colorectal cancer diagnosis. J Clin Oncol 24, 3527-3534, doi:10.1200/JCO.2006.06.0855 (2006).

314 Tardon, A. et al. Leisure-time physical activity and lung cancer: a meta-analysis. Cancer Causes Control 16, 389-397, doi:10.1007/s10552-004-5026-9 (2005).

315 Giovannucci, E. L., Liu, Y., Leitzmann, M. F., Stampfer, M. J. & Willett, W. C. A prospective study of physical activity and incident and fatal prostate cancer. Arch Intern Med 165, 1005-1010, doi:10.1001/archinte.165.9.1005 (2005).

316 Holmes, M. D., Chen, W. Y., Feskanich, D., Kroenke, C. H. & Colditz, G. A. Physical activity and survival after breast cancer diagnosis. JAMA 293, 2479-2486, doi:10.1001/jama.293.20.2479 (2005).

317 Pinto, B. M., Frierson, G. M., Rabin, C., Trunzo, J. J. & Marcus, B. H. Home-based physical activity intervention for breast cancer patients. J Clin Oncol 23, 3577-3587, doi:10.1200/JCO.2005.03.080 (2005).

318 Goh, J. et al. Exercise training in transgenic mice is associated with attenuation of early breast cancer growth in a dose-dependent manner. PLoS One 8, e80123, doi:10.1371/journal.pone.0080123 (2013).

319 Higgins, K. A., Park, D., Lee, G. Y., Curran, W. J. & Deng, X. Exercise-induced lung cancer regression: mechanistic findings from a mouse model. Cancer 120, 3302-3310, doi:10.1002/cncr.28878 (2014).

320 Glass, O. K. et al. Differential response to exercise in claudin-low breast cancer. Oncotarget 8, 100989-101004, doi:10.18632/oncotarget.21054 (2017). 154

321 Cardaci, S. et al. Pyruvate carboxylation enables growth of SDH-deficient cells by supporting aspartate biosynthesis. Nat Cell Biol 17, 1317-1326, doi:10.1038/ncb3233 (2015).

322 Sullivan, L. B. et al. Supporting Aspartate Biosynthesis Is an Essential Function of Respiration in Proliferating Cells. Cell 162, 552-563, doi:10.1016/j.cell.2015.07.017 (2015).

323 Birsoy, K. et al. An Essential Role of the Mitochondrial Electron Transport Chain in Cell Proliferation Is to Enable Aspartate Synthesis. Cell 162, 540-551, doi:10.1016/j.cell.2015.07.016 (2015).

324 Bao, X. R. et al. Mitochondrial dysfunction remodels one-carbon metabolism in human cells. Elife 5, doi:10.7554/eLife.10575 (2016).

325 Nikkanen, J. et al. Mitochondrial DNA Replication Defects Disturb Cellular dNTP Pools and Remodel One-Carbon Metabolism. Cell Metab 23, 635-648, doi:10.1016/j.cmet.2016.01.019 (2016).

326 Vander Heiden, M. G. & DeBerardinis, R. J. Understanding the Intersections between Metabolism and Cancer Biology. Cell 168, 657-669, doi:10.1016/j.cell.2016.12.039 (2017).

327 Kimmelman, A. C. Metabolic Dependencies in RAS-Driven Cancers. Clin Cancer Res 21, 1828-1834, doi:10.1158/1078-0432.CCR-14-2425 (2015).

328 Fontana, L. et al. Dietary protein restriction inhibits tumor growth in human xenograft models. Oncotarget 4, 2451-2461, doi:10.18632/oncotarget.1586 (2013).

329 Sanderson, S. M., Mikhael, P. G., Ramesh, V., Dai, Z. & Locasale, J. W. Nutrient availability shapes methionine metabolism in p16/MTAP-deleted cells. Sci Adv 5, eaav7769, doi:10.1126/sciadv.aav7769 (2019).

330 Sanderson, S. M., Gao, X., Dai, Z. & Locasale, J. W. Methionine metabolism in health and cancer: a nexus of diet and precision medicine. Nature Reviews Cancer, doi:10.1038/s41568-019-0187-8 (2019).

331 Lu, M. et al. Exercise inhibits tumor growth and central carbon metabolism in patient- derived xenograft models of colorectal cancer. Cancer Metab 6, 14, doi:10.1186/s40170- 018-0190-7 (2018).

332 Sanjana, N. E., Shalem, O. & Zhang, F. Improved vectors and genome-wide libraries for CRISPR screening. Nat Methods 11, 783-784, doi:10.1038/nmeth.3047 (2014).

333 Wang, L. et al. Human genetic and metabolite variation reveals that methylthioadenosine is a prognostic biomarker and an inflammatory regulator in sepsis. Sci Adv 3, e1602096, doi:10.1126/sciadv.1602096 (2017).

155

334 Sanderson, S. M. & Locasale, J. W. Revisiting the Warburg Effect: Some Tumors Hold Their Breath. Cell Metab 28, 669-670, doi:10.1016/j.cmet.2018.10.011 (2018).

335 Lee, J. et al. Effective breast cancer combination therapy targeting BACH1 and mitochondrial metabolism. Nature 568, 254-258, doi:10.1038/s41586-019-1005-x (2019).

336 Tang, S. et al. Methionine metabolism is essential for SIRT1-regulated mouse embryonic stem cell maintenance and embryonic development. EMBO J 36, 3175-3193, doi:10.15252/embj.201796708 (2017).

337 Rittiner, J. E. et al. Functional Genomic Analyses of Mendelian and Sporadic Disease Identify Impaired eIF2alpha Signaling as a Generalizable Mechanism for Dystonia. Neuron 92, 1238-1251, doi:10.1016/j.neuron.2016.11.012 (2016).

338 Reid M.A., S. S. M., Locasale J.W. Cancer Metabolism. . 6 edn, 127-138 (Elsevier, 2019).

156

Biography

Sydney M. Sanderson received her Bachelor of Science in Neuroscience from Duke

University in Durham, NC in Spring 2013. She performed undergraduate research in the laboratory of Dr. Miguel Nicolelis, under the supervision of the postdoctoral associate Dr. Hao

Zhang, studying the electrophysiological mechanisms driving memory consolidation during REM sleep cycles. In Fall 2014, Sydney joined the Department of Pharmacology and Cancer Biology

(PCB) as a doctoral student in the Pharmacology program at Duke University. She joined the laboratory of Dr. Oren Becher in the Department of Pediatrics, where she characterized the epigenetic consequences of a histone mutation found in rare pediatric brainstem gliomas. Upon completion of her preliminary examination, Dr. Becher accepted a position at Northwestern

University in Chicago, IL, at which time Sydney transferred to the laboratory of Dr. Jason

Locasale in the PCB Department.

During her time in Dr. Locasale’s laboratory, Sydney primarily worked on characterizing gene-environment interactions that contribute to metabolic heterogeneity in cancer. Throughout her career at Duke, she was a lead author on two scientific papers 329,331, a comprehensive review article 330, an invited commentary on human cancer patient tumor profiling 334, and contributed as a co-author on a number of additional manuscripts 200,216,335-337. Sydney additionally served as a co-author for chapter on cancer metabolism in the sixth edition of Abeloff’s Clinical Oncology textbook 338. During her graduate school career, Sydney received the Fitzgerald Award for

Academic Achievement, a Predoctoral Fellowship Award (F31) from the NCI-NIH, and the Paul and Lauren Ghaffari Fellowship. She was also nominated for the E. Bayard Halsted Scholarship and the Jo Rae Wright Fellowship for Outstanding Women in Science. She additionally participated as a scholar in the NIH-sponsored Pharmacological Sciences Training Program

(PSTP) during her first two years of study.

157