The Epithelial to Mesenchymal Transition in Breast Cancer Induces Alterations in Metabolism

The Texas Medical Center Library DigitalCommons@TMC

The University of Texas MD Anderson Cancer Center UTHealth Graduate School of The University of Texas MD Anderson Cancer Biomedical Sciences Dissertations and Theses Center UTHealth Graduate School of (Open Access) Biomedical Sciences

5-2018

The Epithelial To Mesenchymal Transition In Breast Cancer Induces Alterations In Metabolism

Esmeralda Ramirez-Peña

Follow this and additional works at: https://digitalcommons.library.tmc.edu/utgsbs_dissertations

Part of the Medicine and Health Sciences Commons

Recommended Citation Ramirez-Peña, Esmeralda, "The Epithelial To Mesenchymal Transition In Breast Cancer Induces Alterations In Metabolism" (2018). The University of Texas MD Anderson Cancer Center UTHealth Graduate School of Biomedical Sciences Dissertations and Theses (Open Access). 859. https://digitalcommons.library.tmc.edu/utgsbs_dissertations/859

This Dissertation (PhD) is brought to you for free and open access by the The University of Texas MD Anderson Cancer Center UTHealth Graduate School of Biomedical Sciences at DigitalCommons@TMC. It has been accepted for inclusion in The University of Texas MD Anderson Cancer Center UTHealth Graduate School of Biomedical Sciences Dissertations and Theses (Open Access) by an authorized administrator of DigitalCommons@TMC. For more information, please contact [email protected]. THE EPITHELIAL TO MESENCHYMAL TRANSITION IN BREAST CANCER

INDUCES ALTERATIONS IN METABOLISM

Esmeralda Quetzalcitlali Ramirez-Peña, B.S.

APPROVED:

______Sendurai A. Mani, Ph.D. Advisory Professor

______Richard Behringer, Ph.D.

______Mikhail Kolonin, Ph.D.

______Suzanne Fuqua, Ph.D.

______Arun Sreekumar, Ph.D.

APPROVED:

______Dean, The University of Texas MD Anderson Cancer Center UTHealth Graduate School of Biomedical Sciences i

THE EPITHELIAL TO MESENCHYMAL TRANSITION IN BREAST CANCER

INDUCES ALTERATIONS IN METABOLISM

DISSERTATION

Presented to the faculty of

The University of Texas

MD Anderson Cancer Center UTHealth

Graduate School of Biomedical Sciences

In Partial Fulfillment

For the Degree of

DOCTOR OF PHILOSOPHY

Esmeralda Quetzalcitlali Ramirez-Peña, B.S.

Houston, Texas

May, 2018

Dedication

To my parents and brother for the American dream that we embarked on together

Para mis padres y hermano por el sueño Americano en que navegamos juntos

iii

Acknowledgements

My dream of becoming a scientist was fulfilled with the support of many individuals and organizations that provided me with professional, educational, and emotional support. First, I would like to thank my parents, Mrs. Esperanza Peña-Rosales and Mr. Arturo Ramirez-Cuellar, for buying me my first microscope when I was six years old. Their sacrifices and perpetual support have allowed me to have the freedom to transform my education into the career of my dreams and encouraged me to fly as high as my heart desired. I would like to thank my brother,

Arturo Ramirez-Peña, whose strength and positivity has always filled me with joy and laughter. I am thankful for the lessons I have learned from my family, particularly to be fearless, humble, and hard working in pursuit of great goals.

I am eternally grateful to the MD Anderson Cancer Center UTHealth Graduate School of

Biomedical Sciences for nurturing me in an environment that is dedicated to students’ success.

The faculty and staff went above and beyond their job descriptions; they listened to my concerns, encouraged my curiosity, gave me the freedom to practice leadership and community service, and always made me feel welcome.

I would like to thank my mentor, Dr. Sendurai A. Mani for giving me the opportunity to develop my projects with my own creative freedom, for opening the door to many collaborations while challenging me to improve my scientific skills to the best of my potential. I will always remember our first conversation where he told me that the most important quality a scientist should have is passion. In his lab, I was blessed to meet wonderful colleagues and I cherish all those times where we engaged in scientific discussions that challenged and motivated me. I want to thank my “lab sisters” and current members of the Mani lab, Dr. Rama Soundararajan, Dr.

Petra Den Hollander, Dr. Robiya Joseph, Neeraja Bhangre, Dr. Sridevi Addanki, and Mr. Paul

Allegakoen. They showed me that hard work can be accomplished in a supportive and positive environment. Working by their side has been one of the best experiences in my career.

I would also like to acknowledge previous members of the Mani lab who mentored me and taught me many skills; Dr. Joe Taube, Dr. Tapasree Roy Sarkar, Dr. Anurag Paranjape, Dr.

Steve Werden, Dr. Mika Pietilä and Dr. Nathalie Sphyris. I would like to thank Dr. Geraldine Raja,

Dr. Rebeca Romero-Aburto, Dr. Andria Schibler, and Dr. Patty Dimarco-Hewitt for their friendship and companionship in our journey together as graduate students. I would like to thank Ms. Vinita

Shivakumar for sticking by my side for two years. She invigorated me with her curiosity and enthusiasm and mentoring her inspired me to strive to be a better teacher.

I would like to thank our administrative support from the Translational Molecular

Pathology (TMP) Department, Ms. Teresa Self, Ms. Barbara Sadeghi and Ms. Andrea Jones who always helped our lab run smoothly, and Dr. Celia Garcia-Prieto and Ms. Gabriela Mendoza for their assistance with my animal protocols. I am very grateful for the support of the TMP department led by Dr. Ignacio Wistuba whose dedication to trainee career development initiatives funded me with a travel award that gave me the opportunity to present my work at an international

Keystone Conference. I would like to thank the leadership and of the TMP Interdisciplinary

Translational Education and Research Training Program (iTERT), Dr. Ignacio Wistuba, Dr. Rama

Soundararajan, Dr. Ann Killary, and administrative support Ms. Sarah Fayle and Mr. Bert Shaw

I benefited immensely from the iTERT journal club, technical workshops, and seminar series and they always listened to my ideas with enthusiasm.

My project and progress was overseen by mentors who provided me with constructive suggestions and were dedicated to my success. I would like to thank the members of my advisory committee, Dr. Richard Behringer, Dr. Mikhail Kolonin, Dr. Suzanne Fuqua, and Dr. Arun

Sreekumar, for taking the time to listen to my concerns and making sure that I remained on the correct path. I would also like to thank my mentors from the Genes and Development graduate program: Dr. Swathi Arur, Dr. Michael Galko, Dr. Andrew Gladden, Dr. Russell Broaddus, Dr.

Jichao Chen, Dr. Sadhan Majumder, Dr. William Mattox, and Dr. Pierre McCrea and the G&D program manager Ms. Elisabeth Lindheim who always helped me with a caring attitude. v

The development of my projects would not have been possible without the support of our collaborators and those who generously shared their expertise, instruments, and reagents with us in the spirit of team science. I would like to thank members of Dr. Sreekumar’s lab, Dr. James

Arnold and Dr. Salil Bhowmik, for their patience and dedication to our EMS project that led to many future projects. Also, Dr. Serena Chan from Biolog, Dr. Arnold Levine and members of his lab for our insightful discussions during our joint lab meetings, Dr. Jeffrey Rosen and members of the Baylor Breast Center for giving me the opportunity to share my work, Dr. Elsa Flores, Dr.

Avinash Venkatanarayan, Dr. Michael Davies, members of Dr. Davies’ lab, Dr. Vashisht Yennu-

Nanda, Rishika Prasad, and Grant Fisher, Dr. Peng Huang, Dr. Helen Pelicano, Dr. Lynn

Zechiedrich, and Zach Conley were all extremely instrumental in guiding me with their knowledge of cancer metabolism and sharing their instruments for the metabolic assays in this thesis. I am extremely grateful to Dr. Li Ma and Dr. Kenneth Scott for providing us with vectors for this study, and Dr. José Matés and Dr. Javier Márquez from the University of Malaga in Spain for providing us with the GLS and GLS2 antibody.

I want to acknowledge the Association of Minority Biomedical Researchers (AMBR) and its leadership Dr. Marenda Wilson-Pham and Dr. Andy Bean for giving me the opportunity to exercise leadership and making me aware that my role as a minority in science is extremely valuable. I also want to thank Dr. Sina Safayi for believing in me and providing me with the tools that opened the doors to the future of my dreams. I also thank Dr. Thomas Goka, whose kind and reassuring words still resonate with me.

Last but not least, I want to thank my other Houston family who encouraged me and were a source of much happiness. I want to thank Dr. Marco Farina for sharing this journey with me and always reminding me that this sacrifice will be fruitful one day. I also want to thank my dear friends who never stopped cheering me on, understood all the times that I went “missing” because I was either too busy or too tired, and I could always count on to be ready to enjoy life with me. vi

THE EPITHELIAL TO MESENCHYMAL TRANSITION IN BREAST CANCER

INDUCES ALTERATIONS IN METABOLISM

Esmeralda Quetzalcitlali Ramirez-Peña, B.S.

Supervisory Professor: Sendurai A. Mani, Ph.D.

Metastasis is currently incurable. Recent studies suggest that metabolic reprogramming significantly impacts metastatic progression. Understanding the underlying biology of how metabolic reprogramming influences metastasis will shed light on developing novel ways to treat metastasis and is of urgent clinical need. A potent mediator of metastatic potential in breast cancer is the winged helix/Forkhead domain transcription factor FOXC2. We have previously shown that FOXC2 is necessary and sufficient to promote metastasis through induction of the epithelial to mesenchymal transition (EMT). EMT is a latent embryonic program that is aberrantly induced in breast cancer cells conferring migratory and invasive capabilities in addition to cancer stem cell (CSC) properties. In normal adipose tissue, FOXC2 regulates adipocyte metabolism by promoting the expression of genes associated with energy turnover, insulin sensitivity, differentiation, and mitochondrial biogenesis. I hypothesized that similar to its regulatory role of adipocyte metabolism, FOXC2 dictates metabolic functions during EMT.

Here we report that distinct metabolic alterations occur with the onset of EMT. We identified an EMT metabolite signature (EMS) composed of four elevated metabolites (glutamine, glutamate, beta-alanine, and glycyleucine) and found that the EMS is predictive of poor clinical outcome in breast cancer patients. In efforts to characterize the functional relevance of the metabolic alterations defined by the EMS, we examined genes associated with metabolism that change with the onset of EMT. From this analysis, we identified genes associated with purine nucleotide and glutamine metabolism that became downregulated with the onset of EMT. We measured the expression of these genes in EMT-induced cells with FOXC2 knockdown. We observed that depletion of FOXC2 resulted in upregulation of these purine and glutamine metabolism genes and a restoration of their associated metabolism. Moreover, we found that the

vii induction of EMT results in reduced mitochondrial metabolism and glutamine independence. This metabolic program is maintained by the downregulation of Glutaminase 2 (GLS2) expression, an enzyme that catalyzes the conversion of glutamine to glutamate. We also observed that GLS2 expression is FOXC2 dependent. Inhibition of FOXC2 signaling by treatment of cells with a small moleculae inhibitor of its upstream regulator p38MAPK with a chemical inhibitor enhanced GLS2 mRNA expression and mitochondrial metabolism. Moreover, we demonstrate that GLS2 over- expression alone increases mitochondrial metabolism, reduces the glycolytic metabolic program, and CSC mammosphere formation. Collectively these data are in line with an analysis of patient data that demonstrated that higher GLS2 expression is associated with better over-all survival.

Lastly, we tested a novel therapeutic strategy that combined a p38MAPK inhibitor, which inhibits the FOXC2 signaling pathway, with a standard-of-care chemotherapeutic to reduce the tumor bulk. We observed both a decrease in lung metastasis and tumor size in mice treated with the combination strategy compared to mice receiving monotherapy.

viii

Table of Contents

Approvals………………………………………………………………………………………………….i

Title ………………………………………………………………………………………………………..ii

Dedication………………………………………………………………………………………………..iii

Acknowledgements……………………………………………………………………………………..iv

Abstract…………………………………………………………………………………………………..vii

Table of Contents………………………………………………………………………………………..ix

List of Figures……………………………………………………………………………………………..x

Chapter 1: Introduction…………………………………………………………………………………..1

Chapter 2: Central Hypothesis and Objectives………………………………………………………19

Chapter 3: Materials and Methods…………………………………………………………………….22

Chapter 4: Metabolic alterations associated with EMT predict poor clinical outcome…………..36

Chapter 5: Investigating metabolic genes in the context of FOXC2-associated metabolic

Reprogramming……………………………………………………………………………51

Chapter 6: EMT directs metabolic reprogramming by suppressing mitochondrial metabolism and

Glutaminase 2……………………………………………………………………………..62

Chapter 7: Inhibition of FOXC2 signaling with p38MAPK inhibitor sensitizes breast cancer stem

cells to standard of care chemotherapeutics…………………………………………..97

Chapter 8: Discussion and Future Directions………………………………………………………113

Bibliography……………………………………………………………………………………………126

Vita……………………………………………………………………………………………………..146

List of Figures

Chapter 1

Figure 1: The epithelial to mesenchymal transition promotes metastasis

Figure 2: Glycolytic and oxidative metabolism

Chapter 2

Figure 3: Summary of objectives and chapters in this thesis in which each was addressed

Chapter 4

Figure 4: Analysis of metabolism-associated genes differentially expressed upon EMT

Figure 5: Heatmaps of significantly differential metabolites

Figure 6: Metabolites with significantly elevated abundances form the EMS

Figure 7: Principal component analysis of EMS in patient metabolomic profiles

Figure 8: The EMS is associated with aggressive breast cancer molecular subtypes

Figure 9: EMT expression within tumor samples from Group 1 and Group 2 samples

Chapter 5

Figure 10: Metabolic genes that are altered in EMT

Figure 11: FOXC2 regulates the expression of XDH, UPP1, and APRT in EMT

Figure 12: XDH and nucleotide metabolism influence EMT

Chapter 6

Figure 13: Induction of EMT results in repressed mitochondrial metabolism

Figure 14: EMT-enriched breast cancer cell lines exhibit reduced mitochondrial metabolism

Figure 15: Levels of glycolytic intermediates are higher and glutamine flux is decreased after

EMT

Figure 16: GLS2 is down-regulated in mesenchymal breast cancers

Figure 17: GLS2 is downregulated in cells that have undergone EMT and in metastatic breast

cancer cell lines

Figure 18: GLS2 protein is expressed in cells that have undergone EMT

Figure 19: EMT marker expression in HMLER cells x

Figure 20: FOXC2 is negatively correlated with GLS2 in breast cancer patient samples

Figure 21: Cells deficient in FOXC2 have higher GLS2 expression

Figure 22: ZEB1 induces FOXC2-mediated repression of GLS2

Figure 23: In the absence of FOXC2, ZEB1 does not repress GLS2

Figure 24: Inhibition of EMT by expression of shFOXC2 enhances mitochondrial metabolism

Figure 25: Inhibition of p38MAPK-FOXC2 signaling enhances GLS2 expression

Figure 26: Inhibition of p38MAPK-FOXC2 signaling enhances mitochondrial metabolism

Figure 27: Over-expression of GLS2 enhances mitochondrial metabolism

Figure 28: Loss of GLS2 does not result in a CSC phenotype

Figure 29: Over-expression of GLS2 enhances mitochondrial metabolism

Figure 30: Amino Acid utilization

Figure 31: Higher GLS2 expression is correlates with improved patient outcomes

Chapter 7

Figure 32: Survival outcomes decrease in women diagnosed with distant metastasis compared

to localized or regional breast cancer

Figure 33: Graphical model of the proposed advantage of a combination treatment that targets

CSCs and bulk tumor

Figure 34: Inhibiting p38MAPK with SB203580 reduces FOXC2 and sphere formation in 4T1

cells

Figure 35: Employing chemotherapeutics in combination with SB203580 in vitro

Figure 36: Schematic of in vivo treatment groups

Figure 37: The combination of SB203580 with carboplatin reduces tumor volume and lung metastasis

Chapter 1: Introduction

xii

Chapter 1: Introduction

1.1 Metastatic breast cancer

There is no cure for metastatic breast cancer, which affects women of all socio-economic backgrounds around the world. Metastasis is the leading cause of cancer death and results in a significant economic burden for patients and health care systems. Despite significant advances in our understanding of the fundamental biology of cancer, and improvements in surveillance and treatment strategies, the survival of women diagnosed with metastatic breast cancer has not significantly improved in ten years (Cardoso et al., 2016). Furthermore, 20-30% of women diagnosed with early stage breast cancer will progress to metastatic disease (O'Shaughnessy,

2005). Survival outcomes are lowest for patients living in low-income communities and countries where there is a lack of access to early detection, and treatment (DeSantis et al., 2015). To improve outcomes, increasing our understanding of the biology that governs the development of metastasis is an urgent public health concern.

Metastasis occurs when cells from the primary tumor acquire the ability to detach and invade surrounding tissue, intravasate and survive in the circulation. Metastatic cells have properties that allow them to evade the immune system; these cells are therefore not recognized as foreign or eliminated. The most aggressive cells complete the metastatic cascade by surviving in circulation and ultimately arrive at a distant organ and are able to grow in a foreign microenvironment. Breast cancer is described as a heterogeneous disease or “multiple diseases” that are highly diverse and are generally divided into subtypes based on the expression of estrogen

(ER), progesterone (PR), and growth factor receptor HER2 (Er-B2) (Brooks et al., 2015). Triple negative breast cancer (TNBC) is a subtype that does not express ER, PR, or HER2. Although it only represents approximately 11% of total cases of breast cancer in developed nations, TNBC is the most aggressive subtype (Howlader et al., 2014). Because of the absence of receptor expression, TNBC lacks actionable molecular targets leaving few therapeutic options.

Furthermore, patients with TNBC have relatively poorer outcomes that those patients with other subtypes due to the inherent aggressive features of TNBC (Malorni et al., 2012). About 95% of

TNBCs are histologically classified as invasive mammary carcinomas (Weigelt and Reis, 2009), and based on the intrinsic gene expression subtyping, 79% are classified as basal-like (Prat et al., 2013). Another method of classification describes 61-71% of TNBC tumors as claudin-low, indicating that these tumors resemble undifferentiated mammary epithelial stem cells (Prat et al.,

2010). The claudin-low subtype is enriched with undifferentiated cells with stem-like features that express genes associated with the epithelial to mesenchymal transition (EMT) and are associated with therapy resistance (Creighton et al., 2009; Li et al., 2008; Taube et al., 2010a).

The Pietenpol classification further breaks down TNBC into six subtypes. Two of which are mesenchymal and mesenchymal stem-like. These two subtypes are enriched in genes and pathways involved with cell motility (Rho pathway), cell differentiation (Wnt pathway), growth factor signaling (TGFβ signaling), stem cells (HOX genes, PDGFRB, VCAM1, ALDHA1), and

EMT-associated genes (MMP2, SNAI2, TWIST1, ZEB1) (Lehmann et al., 2011; Mayer et al.,

2014). A majority of the tumors within the mesenchymal stem-like subtype also fall into the claudin-low subtype (Lehmann and Pietenpol, 2014).

1.2 EMT and metastasis

The EMT program orchestrates molecular changes that enable the dynamic processes that leads to metastasis. EMT operates during embryogenesis to allow gastrulation and organogenesis and to enable epithelial cells to become mesenchymal and motile to form organ structures and tissues. The loss of apicobasal cell polarity, adherens junctions, and a cytoskeletal reorganization is required for cells to transition into a mesenchymal and migratory phenotype

(Fristrom, 1988). The reverse process, the mesenchymal to epithelial transition (MET) is also highly regulated during development, particularly in the formation of tubules. In normal adult tissue, EMT can be reactivated for wound healing and tissue regeneration (May et al., 2011). In the context of cancer, the EMT program can be “hijacked” by cancer cells and the acquisition of 14 mesenchymal features equips these cancer cells with aggressive characteristics. This phenotypic conversion is orchestrated by a network of EMT-inducing transcription factors (EMT-

TFs) including SNAIL, TWIST, GOOSECOID, ZEB1, and FOXC2 (Taube et al., 2010b). The expression of EMT-TFs leads to the direct and indirect repression of E-cadherin (CDH1), a critical component of adherens junctions that maintains cell-cell contacts. Not only does EMT direct epithelial cells to shed epithelial characteristics, it also causes the de novo expression of mesenchymal genes (CDH2, FN, SMA, VIM) (Figure 1).

Migratory and invasive capabilities facilitate the initial steps of the metastatic cascade; however, to complete the metastatic cascade, it has been proposed that the induction of MET must also take place. Distant metastases resembles the molecular subtype of the primary tumor, and reports have demonstrated that disseminated cancer cells are often epithelial in nature

(Dykxhoorn et al., 2009; van der Pluijm, 2011; Weigelt et al., 2005). The plasticity between

EMT/MET is key for efficient completion of the metastatic cascade (Brooks et al., 2015). The link between EMT and metastasis has been bridged with multiple functional studies that demonstrate how the acquisition of EMT properties occurs in parallel with the acquisition of cancer stem cell (CSC) properties (Cano et al., 2000; Vesuna et al., 2009; Yang et al., 2010).

Figure 1: The epithelial to mesenchymal transition promotes metastasis. Diagram of the process of EMT driven by EMT-TFs FOXC2, Snail, Twist, Goosecoid, TGFβ, and ZEB1. EMT causes epithelial cells to lose cell polarity, cell-to-cell contacts, and expression of E-cadherin.

The acquisition of mesenchymal properties, such as migration and invasion, allows a cell to enter into circulation, evade the immune system, colonize distant organs, and metastasize. These properties also lead to chemotherapeutic resistance and tumor relapse.

1.3 Cancer stem cells

The CSC model hypothesizes that tumors are organized by a hierarchy where cells at the apex of the lineage tree display stem cell properties. CSC properties include the ability to self- renew and generate differentiated cells that form the bulk tumor. This hierarchical organization was first demonstrated in human acute myeloid leukemia (Bonnet and Dick, 1997). Multiple lines of evidence have demonstrated that EMT/CSC properties equip cancer cells with the ability to migrate, survive in hostile and evolving microenvironments, and form metastases (Mani et al.,

2008; Morel et al., 2008). The subpopulation of breast cancer cells with tumor initiating capacity was identified based on cell surface marker expression. CD44+CD24-/low/Lineage- have the capacity to self-renew and to recapitulate the heterogeneity of the parental tumor (Al-Hajj et al.,

2003). Cell surface marker CD44 is an adhesion molecule that is expressed on early multipotent epithelial progenitor cells (Naor et al., 1997), therefore its expression in the tumor-initiating cells indicates a degree of stemness. Furthermore, EMT-TF TWIST1 down-regulates the expression of CD24 (Vesuna et al., 2009). Multiple lines of evidence indicate that breast CSCs express genes associated with the latent embryonic EMT program and that the expression of these genes predictive of metastatic competence and clinical outcome (Soundararajan et al., 2015). EMT-TF

ZEB1 suppresses expression of the miR-200 family (Wellner et al., 2009) and is evident in invasive breast cancers (Park et al., 2008). The miR-200 family suppresses stem-cell associated genes (BMI1, SOX2, KLF4, and therefore suppression by ZEB1 reinforces the stemness phenotype (Shimono et al., 2009).

1.4 Transcription factor FOXC2 and its role in CSCs and metastasis

EMT-TF FOXC2 is a winged helix/Forkhead domain transcription factor. Its role as a key regulator of metastasis and EMT/CSC properties was identified by gene expression analysis of murine cell lines (67NR, 168FARN, 4TO7, and 4T1) with different degrees of metastatic potential

(Mani et al., 2007a). FOXC2 expression was highest in the cell line 4T1 with the ability to

17 metastasize to the lung that therefore has the highest degree of metastatic capacity among the murine cell lines examined. Notably, FOXC2 expression is high in highly aggressive basal-like patient breast cancer samples and metastatic breast cancer cell lines. FOXC2 expression is sufficient to induce EMT and promote migration, invasion, and down-regulation of E-cadherin in otherwise non-invasive cells and inhibition of FOXC2 expression reduced the ability of 4T1 cells to metastasize to the lung (Mani et al., 2007a).

A follow-up study demonstrated that FOXC2 expression links EMT and CSC properties

(Hollier et al., 2013b). FOXC2 over-expression enriched human mammary epithelial cells with a stem cell CD44High/CD24low population, induced mammosphere formation, and resulted in resistance to the chemotherapeutic drug paclitaxel. A hallmark of CSCs is their de novo resistance to chemotherapeutic drugs (Fillmore and Kuperwasser, 2008; Lagadec et al., 2010).

FOXC2-expressing cells injected into the mammary fat pad of mice had tumor-initiation capacity and metastasized to the lung, bone, brain, and liver. Furthermore, this study demonstrated that decrease in FOXC2 levels decreased the percentage of the cells with the CD44High/CD24low phenotype, decreased mesenchymal markers N-cadherin, vimentin, and fibronectin. In contrast, expression of the epithelial marker E-cadherin decreased tumor initiation capacity and mammosphere formation and sensitized cells to the chemotherapy drug paclitaxel. Collectively, this study demonstrated that FOXC2 is a common mediator of multiple EMT programs where signaling instigated by Snail, Twist, and TFGβ converge and confer metastatic competence and self-renewal capacity to CSCs. Consistent with the critical role of FOXC2 in regulating breast

CSCs, FOXC2 expression is elevated in breast CSCs and in the residual surviving cells in patients treated with conventional chemotherapeutics (Creighton et al., 2009). These findings highlight the role of FOXC2 as a key EMT-TF that integrates CSC properties to confer metastatic potential in breast cancer.

1.5 The role of FOXC2 in adipocyte metabolism

In normal tissue, FOXC2 is only expressed in adipocytes. In adipocytes, FOXC2 regulates genes involved in adipocyte differentiation including PPAR-γ, which encodes peroxisome proliferator-activated receptor, C/EBPα and ADD1/SREBP1 which encode transcription factors. Over-expression of FOXC2 in adipocytes leads to a lean and insulin- sensitive phenotype in mice due to an increase in PKA-mediated signaling (Cederberg et al.,

2001). Furthermore, FOXC2 induces mitochondriogenesis by trans-activating the mitochondrial transcription factor TFAM (also known as mtTFA) which is a master regulator of mitochondrial biogenesis. TFAM regulates the expression of mitochondrial elongation and fusion genes Mfn1,

Mfn2, and Opa1 (Lidell et al., 2011). As a result, high FOXC2 activity correlates with increased mitochondrial fusion and enhanced aerobic capacity and palmitate oxidation. In an inflammatory model with inflammation induced by lipopolysaccharide (LPS), exogenous FOXC2 expression alleviated adipocyte inflammation by positive regulation of leptin and the STAT3-PRDM16 complex. This signaling promoted adipose browning in high fat diet-induced obese mice (Gan et al., 2018).

1.6 Metabolic reprogramming in EMT

For cancer cells to adapt, survive, and grow in an environment of fluctuating nutrient and/or oxygen availability, like that of the tumor microenvironment, metabolic reprogramming must occur. Metabolic reprogramming is the re-wiring of bio-energetic pathways that gives cancer cells a selective advantage during tumorigenesis. Reprogrammed bio-energetic pathways allow the highly proliferative cancer cells to support the production of biomass, confer redox balance, and enable growth under stresses like nutrient and oxygen deprivation

(DeBerardinis and Chandel, 2016).

In the 1920-s Otto Warburg observed that instead of using oxidative phosphorylation, cancer cells utilize glucose and produce lactate even in the presence of oxygen (aerobic glycolysis) (Figure 2). This metabolic pathway, called glycolysis, is a physiological response to 19 oxygen deprivation (hypoxia) in normal cells. When glycolysis is activated, glucose enters the cell (in a process facilitated by a glucose transporter) and is broken down by a series of reactions that will produce four molecules of ATP and two molecules of NADH for every one molecule of glucose (Vander Heiden et al., 2009). Although oxidative phosphorylation produces 36 molecules of ATP per molecule of glucose, a cancer cell benefits from glycolysis because intermediates of glycolysis are diverted into other metabolic pathways. These pathways support cancer cell growth. An example is the pentose phosphate pathway (PPP) that generates NADPH and pentose phosphates for ribonucleotide synthesis (Hay, 2016). Glycolysis intermediates also contribute to the hexosamine pathway that generates metabolites for glycosylation, and the serine biosynthesis pathway for amino acid production.

Figure 2: Oxidative and glycolytic metabolism. When cancer cells employ glycolysis, they utilize high levels of glucose to support the synthesis of nucleotides, lipids, and amino acids and lactate is secreted as a byproduct. Oxidative metabolism is mostly employed normal cells under homeostasis to create fuel in the form of ATP by breaking down glutamine, amino acids, and fatty acids. This process occurs in the mitochondria.

Metabolic re-programming occur during tumorigenic transformation because the expression of metabolic enzymes and nutrient transporters are regulated by oncogenes such as c-MYC, HIF-1α, and the PI3K-AKT-mTOR signaling network and tumor suppressor p53. In rare cases, mutations in metabolic enzymes such as fumarate hydratase (FH), succinate dehydrogenase (SDH), and iso-citrate dehydrogenase promote the accumulation of metabolites that affect cellular signaling to support cancer (Kaelin, 2009). Mutations that occur in the mitochondrial DNA and render the production of ATP defective require cancer cells to turn to glycolysis by necessity (Yeung et al., 2008). Furthermore, it has been reported that glycolytic genes are over-expressed in more than 70% of human cancers (Altenberg and Greulich, 2004).

Glycolysis results in the production of lactate and leads to extracellular lactic acidosis surrounding the tumor micro-environment.

This acidic micro-environment also gives cancer cells a growth advantage as they can utilize lactate as an alternative fuel and signaling molecule to stimulate activity of the transcription factor HIF1α. Targets of HIF1α are GLUT1, which encodes a glucose transporter, and HK2 which encodes a kinase that catalyzes glucose phosphorylation, the first committed step in glycolysis.

Expression of these proteins is augmented by the metabolic reprogramming toward glycolysis in cancer cells (Hong et al., 2004). It is of interest that cancer cells that exhibit high levels of glycolysis exhibit highly aggressive characteristics. It has also been shown that HIF1α induces the expression of genes that promote invasion and migration, features of EMT (Zhang et al.,

2013b). HIF1α expression is also induced in hypoxic conditions of the tumor microenvironment

(Rankin and Giaccia, 2016; Vaupel and Mayer, 2007).

In hypoxic conditions, metabolic reprogramming increases the utilization of lipids from the microenvironment by cancer cells and is considered to be a hallmark of cancer aggressiveness (Hanahan and Weinberg, 2011; Palm and Thompson, 2017). Fatty acids are the lipids predominantly used to generate energy in the form of ATP and produce almost twice the energy compared to glucose (39 KJ*g-1 and 15 KJg-1, respectively) (Rohrig and Schulze, 2016).

In cancer cells they can be synthesized de novo or acquired exogenously (Beloribi-Djefaflia et 22 al., 2016; Lengyel et al., 2018). In addition to providing ATP for energy, fatty acids have been shown to support EMT signaling, migration, invasion, and mitochondrial membrane lipid composition that influences chemotherapeutic resistance (Rohrig and Schulze, 2016). In breast cancer, exogenous fatty acids are mainly acquired from the surrounding adipocytes (Dirat et al.,

2011; Wang et al., 2017; Wang et al., 2012). The metabolic symbiosis that occurs between breast cancer cells and adipocytes promotes EMT in epithelial cancer cell lines by downregulating E- cadherin, re-distributing β-catenin, leading to the increased presence of lung metastatic nodules

(Dirat et al., 2011).

There is increasing evidence that the acquisition of EMT properties requires metabolic adaptation to survive in environments in which levels of nutrients and oxygen fluctuate. EMT-TFs such as Snail, Twist, Goosecoid, TGFβ, Zeb1, and FOXC2 operate as an interactome where the expression of one EMT-TF induces the expression of others (Taube et al., 2010b). The EMT- interactome maintains the signaling required to directly suppress expression of epithelial cell adhesion molecule E-cadherin (CDH1), downregulate the miR200 family, and induce mesenchymal markers N-cadherin (CDH2), vimentin, and the invasion-associated matrix metalloproteinase MMP2. There is increasing evidence that EMT-TFs not only regulate the classical EMT processes such as invasion, migration, chemotherapeutic resistance, and resistance to apoptosis but also regulate metabolic reprogramming.

A mesenchymal metabolic signature (MMS) was generated from gene expression data of mesenchymal-like cancer cell lines including breast cancer, hepatocellular carcinoma, and melanoma among others (Shaul et al., 2014). The MMS is enriched for glycan biosynthesis genes such as those encoding the glutamine-fructose-6-phosphate aminotransferase GFPT2 and UDP-

N-acetylglucosamine pyrophosphorylase UAP1. Other metabolic genes in the MMS with expression that correlates with highly aggressive features of migration and invasion, NT5E, and

ENPP2 which encode a nucleotidase and pyrophosphatase/phosphodiesterase. Using an RNAi screen to eliminate the expression of the MMS genes, the dihydropyrimidine dehydrogenase

DPYD was determined to be required for EMT induction. DPYD is the rate-limiting enzyme in the 23 pyrimidine degrading pathway and its expression is upregulated during EMT resulting in an accumulation of DYPD degradation products (dihydropyrimidines).

When DYPD levels were reduced, viability and proliferation were unaffected but induction of EMT was blocked as were sphere formation, and migration. Interestingly, the exogenous addition of dihydrouracil and 5,6-dihydrothymidine rescued the sphere formation in DYPD- deficient cells. It is yet unknown what the functional role of accumulated dihydropyrimidine products are in EMT but one possibility is that they could act as signaling molecules. The work by Shaul et al provides evidence that distinct metabolic reprogramming is regulated by EMT and supports EMT-associated properties.

There is further evidence that individual EMT-TFs regulate expression of specific metabolic genes. For example, Snail represses FBP1 expression and the repression of this fructose bisphophatase favors glucose uptake and diversion of carbons toward the PPP and inhibits oxidative phosphorylation (Dong et al., 2013). Furthermore, Wnt-Snail signaling inhibits the activity of the cytochrome C oxidase COX in the mitochondrial respiratory chain inducing metabolic reprogramming toward glycolysis (Lee et al., 2012). Twist has been shown to be a direct target of HIF-1α and over-expression of Twist results in the up-regulation of critical metabolism genes like LDHA, PKM2, HK2, and G6PD (Yang and Wu, 2008). The EMT-TF

Goosecoid plays a significant role in maintaining the EMT interactome by interacting with the other EMT-TFs ZEB1/2, Twist, Snail, and FOXC2 but its role in directly regulating metabolic reprogramming has not been described.

1.7 Glutamine metabolism in cancer

Glucose metabolism is central to cancer metabolism studies, but glutamine metabolism also contributes during tumor progression by supporting proliferation (DeBerardinis and Chandel, 2016). Highly proliferative tumors consume glucose and amino acids to sustain the metabolic demands that support biomass production. Glutamine is a conditionally essential amino acid that can be obtained by a cell from the microenvironment or synthesized de novo. In

24 circulation, it is highly abundant (~500 µM) and levels are relatively constant (Bergstrom et al.,

1974) Glutamine makes up more than 20% of the free amino acid pool in the blood and 40% in the muscle. Normal intestinal mucosal cells are highly dependent on glutamine and die rapidly after glutamine depletion (Lacey and Wilmore, 1990). This glutamine dependence and susceptibility to glutamine deprivation is observed in some cancers and is highly regulated (Wise and Thompson, 2010; Yuneva et al., 2007). As with glucose, cancer cells exploit the highly available glutamine to drive tumor progression.

The critical role of glutamine comes from its contribution of carbon and nitrogen to the tricarboxylic acid (TCA) cycle, which provides a source of intermediates necessary for macromolecule synthesis. Glutaminolysis is the process whereby glutamine is converted to glutamate and then to α-ketoglutarate, which enters the TCA cycle. Glutaminolysis supports oxidative phosphorylation and, energy generation, provides a carbon source for fatty acid synthesis, a nitrogen source for amino acids and nucleotides, and contributes to the production of reduced glutathione (GSH) which neutralizes reactive oxygen species (ROS) (Mates et al.,

2009).

1.8 Glutaminases in cancer

There are two glutaminase enzymes, GLS and GLS2 that catalyze the conversion of glutamine to glutamate. The two have very distinct expression patterns among different organ and tumor types (Marquez et al., 2016). The GLS gene is on chromosome 2 and encodes the kidney-type (K) isozyme (Katt et al., 2017). GLS contributes to glutamine addiction in primary tumors. Previous reports indicate that GLS is induced during EMT (Kung et al., 2011; Lee et al.,

2016b). GLS is expressed in many tissues and is over-expressed in several cancers as a result of direct regulation by oncogenes such as KRAS and MYC (Gao et al., 2009; Son et al., 2013a).

Tumors addicted to glutamine require glutamine to support growth, and this may result in selection for GLS over-expression in tumors.

Under conditions of nutrient deprivation, cancer cells are capable of acquiring glutamine by breaking down macromolecules. The macromolecules are obtained by a process called micropinocytosis during which the plasma membrane ruffles to create an arm-like protrusion that can capture nutrients from the extra cellular micro-environment (Recouvreux and Commisso,

2017). This process is highly regulated by the KRAS oncogene and is required for proliferation and tumor growth in pancreatic cancer (Commisso et al., 2013). Pharmacological agents that target GLS directly such as CB-839 are currently being tested in clinical trials (Galluzzi et al.,

2013). Although glutamine addiction has been observed in many cancers, recent studies employing 3D organoids and in vivo fluorinated glutamine (18F-FGln PET) labeling have demonstrated that not all tumor types metabolize glutamine (Venneti et al., 2015). The observed glutamine independence in these tumors could confer resistance to glutaminase inhibitors, which are used clinically (Altman et al., 2016a).

The role of GLS2 in tumor progression and metastasis is less well-understood than that of GLS. Whereas GLS promotes tumorigenesis, GLS2 seems to have tumor-suppressive functions. The GLS2 gene is on chromosome 12 and encodes the liver (L type) isozymes GAB and LGA (Marquez et al., 2016). GLS2 is expressed mostly in the liver, brain, and pancreas and is directly regulated by p53, p63, and p73 (Giacobbe et al., 2013; Suzuki et al., 2010; Velletri et al., 2013). p53-mediated induction of GLS2 expression enhances mitochondrial respiration and

ATP production by regulating the production of glutamate and α-ketoglutarate (Hu et al., 2010).

GLS2 regulates the cell’s antioxidant defense mechanism by directing the production of GSH and decreasing levels of ROS. Promoting a cell’s antioxidant defense is protective against oxidative stress-induced apoptosis and DNA damage. Hu et al. showed that GLS2 over- expression enhances mitochondrial respiration and increases GSH, intracellular glutamate, and

α-ketoglutarate levels in hepatocellular carcinoma. GLS2 expression also protects cells from

ROS-induced apoptosis. These metabolic functions led to the decrease of colony formation in vitro. Moreover, GLS2 expression is lost as a result of DNA methylation in glioblastoma and liver,

26 and colon cancers (Szeliga et al., 2016; Zhang et al., 2013a). Liver and colon cancer cell lines treated with DNA demethylating agent 5-aza-2’-deoxycytidine have increased GLS2 mRNA levels. GLS2 over-expression inhibits cell proliferation and induces cell cycle arrest in the G2/M phase. GLS2 over-expressing cells also have a significant reduction of p21; this supports the hypothesis that GLS2 is a tumor suppressor.

The role of GLS2 in metastasis was recently reported in a study that identified a non- enzymatic role for GLS2. A screen to identify novel GLS2-interacting proteins identified the small

Rho GTPase Rac1 (Zhang et al., 2016). Rac1 is a cytoskeletal modulator that regulates the formation of lamellipodia required by mesenchymal cells for migration and invasion (Bid et al.,

2013). It functions as a molecular switch by cycling between inactive GDP-bound and active

GTP-bound forms. In cancer, Rac1 is often over-expressed or activated, and its activity promotes metastasis. The C-terminal region of GLS2 binds to and inhibits the function of Rac1-GDP in a manner that is independent to its glutaminase domain. GLS2-mediated suppression of Rac1-

GDP activity inhibited migration and, invasion in vitro and lung metastasis of hepatocellular carcinoma cells in vivo. The notion that GLS2 has non-enzymatic functions was documented previously in neuronal cells of rat and monkey (Olalla et al., 2002). In this study, catalytically active GLS2 was found enriched in the nuclei of brain cells. The protein structure of the human

GLS2 isoform LGA has an LXXLL consensus motif for nuclear receptor interaction. GLS2 contains a PDZ interaction motif that interacts with α1-syntrophin and glutaminase-interacting protein (Olalla et al., 2001). PDZ domains in proteins mediate interactions with other proteins and direct signal transduction cascades (Lee and Zheng, 2010). Collectively, this suggests that the most significant difference between GLS and GLS2 can be traced to the differences in protein domains that dictate cellular localization and protein-protein interactions.

1.9 p38MAPK-FOXC2 signaling in metastasis

We previously determined that p38MAPK stabilizes FOXC2 protein enabling it to exert its functions that lead to the EMT/CSC phenotype (Paranjape et al., 2016a; Werden et al., 2016). In efforts to target cells with EMT/CSC properties, targeting FOXC2 expressing cells would be ideal because the cells with highly aggressive features could be eliminated. However, therapeutically targeting a transcription factor like FOXC2 is currently not feasible. We reasoned that by targeting p38MAPK (also known as MAPK14) signaling we could destabilize FOXC2 protein and attenuate its activity.

Activation of MAPK signaling, particularly p38α-mediated signaling is reportedly induced by cytokines, growth factors such as TGFβ, and environmental stress such as ROS. The signal transduction occurs by activation of MAP3Ks (MEKK4 and MEKK3) and MAP2Ks (MKK3 and

MKK6 kinases) that converge on p38MAPK leading to its activation by phosphorylation. In addition to stabilizing FOXC2, p38MAPK regulates other transcription factors such as p53, ATF2,

MEF2, ELK1, and C/EBPβ. These transcription factors regulate oncogenic processes such as invasion, inflammation, angiogenesis and cancer progression (Wagner and Nebreda, 2009).

Although many signaling pathways converge on p38MAPK we observed that inhibiting the p38MAPK-FOXC2 axis was sufficient to inhibit CSC properties and metastasis in models of metastatic breast cancer. This is particularly important in CSC TNBC for which no targeted therapeutic strategies available.

p38MAPK inhibitors have been used extensively in patients with chronic obstructive pulmonary disease (NCT01543919), rheumatoid arthritis (NCT00760864), systemic inflammation as safer alternative for aspirin (NCT0039557 and NCT002055478), and acute coronary syndrome (NCT01648192, NCT017564945, NCT02145468). More recently, p38MAPK inhibitor LY2228820 has been tested pre-clinically in models of non-small cell lung cancer, ovarian cancer, glioma, myeloma, breast cancer, and melanoma (Campbell et al., 2014).

Notably, LY2228820 was tested in combination with a chemotherapeutic agent in multiple myeloma and demonstrated improved efficacy in cytotoxicity (Ishitsuka et al., 2008). This 28 compound is also in phase I/II clinical trials in several cancers such as adult glioblastoma in combination with radiotherapy (NCT02364206), metastatic breast cancer in combination with tamoxifen (NCT02322853), platinum sensitive ovarian, and fallopian tube cancer in combination with gemcitabine and carboplatin (NCT01663857), and advanced cancer (NCT01393990). To target p38MAPK in vivo, we used SB203580, a pyridinyl imidazole that competitively binds to the

ATP pocket of p38MAPK. SB203580 is the first reported inhibitor of p38MAPKα and β at micromolar concentration in vitro that does not inhibit closely related homologues SAPK3 and

SAPK4 or JNKs or ERKs. (Gum et al., 1998). This compound is currently not approved for use in human patients but has been used in mice and demonstrated to be effective (Mader et al.,

2008).

Chapter 2: Central hypothesis and objectives

Cancer cells require distinct metabolic programs that provide energy and building blocks to maintain their highly proliferative phenotype. Although significant clinical advances have been made to target primary tumor metabolism, the bioenergetics of metastatic disease are currently not well characterized. However, the idea that metabolic reprogramming can significantly impact the progression of metastasis has recently emerged. A potent mediator of metastatic potential in breast cancer is the winged helix/Forkhead domain transcription factor FOXC2. We have previously shown that FOXC2 is necessary and sufficient to induce metastasis by inducing the epithelial to mesenchymal transition. EMT is a latent embryonic program that is aberrantly induced in breast cancer cells conferring migratory and invasive capabilities in addition to chemotherapeutic resistance. Notably, in normal tissue FOXC2 regulates adipocyte metabolism by promoting the expression of genes associated with energy turnover, insulin sensitivity, differentiation, and mitochondrial biogenesis. I hypothesized that similar to its regulatory role in adipocyte metabolism, FOXC2 dictates metabolic functions during EMT.

Overall, the goal of this thesis is to improve our understanding of metabolic reprogramming during EMT as well as suggest therapeutic strategies to target the unique bioenergetics pertaining to FOXC2 in metastatic breast cancer (Figure 3). To test this hypothesis my objectives are:

Objective 1: Identify distinct metabolic alterations that occur with the onset of EMT by metabolomics analysis of cells that have undergone EMT. Examine if these EMT-specific metabolic alterations can provide clinical prognostic information in breast cancer patients.

Objective 2: Identify genes associated with metabolism with expression changes at the onset of

EMT. Examine the role of FOXC2 in regulating the EMT-specific metabolic gene expression changes.

Objective 3: Characterize the functional implications of dysregulated mitochondrial metabolism during EMT. Examine the role of p38MAPK-FOXC2 signaling in regulating levels of the glutaminase GLS2. Define how GLS2 expression determines mitochondrial metabolism during

EMT.

Objective 4: Test a novel combination strategy in vivo that is designed to target p38MAPK-

FOXC2 signaling to eliminate CSCs in combination with a standard of care chemotherapeutic to simultaneously reduce tumor bulk.

Figure 3: Summary of objectives and respective chapters in this thesis in which each was addressed.

Chapter 3. Materials and Methods

Contents of this chapter are based on Bhowmik, S. K., E. Ramirez-Pena, J. M. Arnold, V. Putluri,

N. Sphyris, G. Michailidis, N. Putluri, S. Ambs, A. Sreekumar and S. A. Mani (2015). EMT- induced metabolite signature identifies poor clinical outcome. Oncotarget 6(40): 42651-42660.

Copyright permission was not required as Oncotarget journal states “authors retain ownership of the copyright for their article”

3.1 Model Systems

HMLE: Immortalized human mammary epithelial (HMLE) cells expressing SV40 large T-antigen, and the telomerase catalytic subunit (hTERT) as described in (Elenbaas et al., 2001)

HMLER: Immortalized human mammary epithelial (HMLE) cells expressing SV40 large T- antigen, the telomerase catalytic subunit (hTERT) and an oncogenic allele of H-ras V12H-Ras

(HMLER) as described in (Elenbaas et al., 2001)

MCF10A: Immortal nontransformed human mammary cells with transcriptional characteristics of the Basal B subtype (Neve et al., 2006)

MCF7: Transformed breast cancer cell line characteristic of the luminal subtype,

ER+/PR+/HER2- (Neve et al., 2006)

SUM159: Transformed breast cancer cell line characteristic of the Basal B, ER-/PR-/HER2-

(Neve et al., 2006)

MDA-MB-231: Transformed breast cancer cell line characteristic of the Basal B, ER-/PR-/HER2-

(Neve et al., 2006)

67NR: Non-metastatic, geneticin-resistant murine tumor cell line isolated from a single spontaneously arising mammary tumor from a BALB/cfC3H mouse (Aslakson and Miller, 1992)

4T1-RFP: Highly metastatic, thioguanine-resistant cell line derived from a single spontaneously arising mammary tumor from a BALB/cfC3H mouse that spontaneously metastasized to the lung

(Aslakson and Miller, 1992)

3.2 Methods

3.2.1 Cell Culture

All cell lines were cultured at 37°C in 5% CO2. All cell lines derived from HMLE cells were cultured in MEGM (Lonza CC-3051): DMEM F12 (Mediatech 10090CV) (1:1) supplemented with penicillin and streptomycin (Gibco/Life Technologies), insulin (Sigma I9278), hEGF (Sigma

9644), hydrocortisone (Sigma H0888), and bovine pituitary extract. MCF7 and MDA-MB-231 cells were cultured in DMEM/F12 (Fisher 10-090-cv) supplemented with 10% fetal bovine serum

(FBS) (Sigma), and penicillin and streptomycin. SUM159 cells were cultured in Ham’s F12

(Corning 10-090-cv) medium containing 10% FBS and penicillin and streptomycin. HEK293T cell lines were cultured in DMEM (Corning) supplemented with 10% FBS and penicillin and streptomycin. MCF10A cells were cultured in DMEM/F12 (Corning) supplemented with 5% horse serum (Sigma H1138), 20 ng/mL human epidermal growth factor (Sigma E9644), 0.5 µg/ml hydrocortisone, 100 ng/ml cholera toxin (Sigma C-8052), 5 µg/ml insulin, and penicillin and streptomycin.

HEK293T cells were plated at 1.5X106 cells per 10-cm dish in 10 mL DMEM with 10%

FBS and penicillin and streptomycin. After 24 hours, cells were transfected with 185 µl serum- free medium containing 1.8 µg of pUMVC (retrovirus) or pΔ8.2 (lentivirus), 0.2 µg pCMV-VSVG,

2.0 µg gene of interest, and 12 µl FuGENE transfection reagent (Promega E2691). After 24 hours, cells were fed with 7.5 mL of 4:1 target cell medium to 293T medium. The target cells were seeded at 30-50% confluency. At 48 and 72 hours after transfections, the spent 4:1 medium was removed and filtered through a 0.45-µm filter. Protamine sulfate was added to the 4:1 medium before infecting target cells, and 4:1 medium was replenished for 293T cells. Target

35 cells were incubated with viral 4:1 medium for 6 hours and then allowed to recover overnight in appropriate growth media. After at least two cycles of infection, target cells were selected with appropriate selection antibiotic.

3.2.2 Inhibitor IC50 determinations

To assess the effect of p38MAPK inhibitor SB203580 on 4T1-RFP cells, 0.3X104 cells were plated in triplicate in a 96-well plate in growth medium. After 24 hours, the medium was aspirated and replaced with growth medium containing vehicle (water) or 20, 40, 80,120, or 160

µl of a solution of 20 mM SB203580. To assess the effects of carboplatin on 4T1-RFP cells,

0.3X104 cells were plated in triplicate in a 96-well plate in growth medium. After 24 hours, the medium was removed by aspiration and replaced with growth medium containing vehicle

(DMSO) or 0.1, 10, 20, 50, 100, or 150 µM carboplatin. To assess the effects of paclitaxel on

4T1-RFP cells, 0.3X104 cells were plated in triplicate in a 96-well plate in growth medium. After

24 hours, the medium was removed by aspiration and replaced with growth medium containing vehicle (DMSO) or 0.1, 1.0, 10, 50, or 100 µM paclitaxel. To assess the effects of SB203850 in combination with carboplatin, 0.3X104 cells were plated in triplicate in a 96-well plate in growth medium. After 24 hours, the medium was removed by aspiration and replaced with growth medium containing vehicle (water) or 20, 40, 80,120, or 160 µl of a solution of 20 mM SB20358 and vehicle (DMSO) or 0.1, 10, 20, 50, 100, or 150 µM carboplatin. To assess the effects of

SB203850 in combination with paclitaxel, 0.3X104 cells were plated in triplicate in a 96-well plate in growth medium. After 24 hours, the medium was removed by aspiration and replaced with growth medium containing vehicle (water) or 20, 40, 80,120, or 160 µl of SB20358 and vehicle

(DMSO) or 0.1, 1.0, 10, 50, or 100 µM paclitaxel. All treatments were incubated for 72 hours. To measure the viability, 20 µl of CellTiter 96 Aqueous One Solution (Promega) was added to each well. After incubation at 37°C for 60 minutes, the absorbance at 490 nm was measured. Raw data were converted to relative percent viabilities, transformed to log, and the concentration at which viability was reduced by 50% (IC50) was calculated using GraphPad Prism software.

3.2.3 Glucose sensitivity assay

Cells were plated at 3,000 cells per well in 96-well plates in growth medium. After 24 hours, medium was removed by aspiration, and cells were washed with PBS. Medium was replenished with 100 µl of glucose and glutamine-free DMEM containing 12 mM, 3 mM, 1.5 mM,

0.8 mM, 0.4 mM, or 0 mM of glucose (Sigma). Each concentration was evaluated in four replicates. After 24 hours, 20 µl of CellTiter 96 Aqueous One Solution (Promega) was added to each well. After incubation at 37 °C for 60 minutes, absorbance at 490 nm was determined. Raw data were converted to relative percent viabilities, transformed to log, and the IC50 was calculated using GraphPad Prism software.

3.2.4 Glutamate determination

Samples were analyzed with the Glutamine/Glutamate Determination Kit (SIGMA GLN-

1) following the manufacturer’s recommendations. Briefly, 0.5X106 cells were plated in triplicate in 6-well plates. After 24 hours, growth medium was removed by aspiration, and cells were washed with PBS. Cells were lysed with 250 µl of RIPA buffer. A glutamate standard curve was prepared in a 96-well plate. Test wells contained 25 µl of sample, 175 µl of reaction buffer, and

2 µl of L-GLDH enzyme. The absorbance at 340 nm was measured at time 0 and in 40-minute increments until the signal plateaued.

3.2.5 Glutathione determination

Cells were plated at 10,000 cells per well in 6 replicates in a 96-well plate. After 24 hours, reduced glutathione was assayed with GSH-Glo Glutathione Assay (Promega V6911). Briefly, the growth medium was removed from each well, and cells were washed with PBS. A GSH standard curve was prepared, and cells were incubated with GSH-Glo reagent. After 30 minutes,

Luciferin detection reagent was added to each well and incubated for 15 minutes. The luminescence was measured using a luminometer.

3.2.6 Seahorse MitoStress Test

Cells were seeded at 15,000 cells per well in an XF96 cell culture plate with appropriate medium. The XF96 probes were calibrated with 200 µL of calibrant solution and incubated at 37

°C in a non-CO2 incubator. After 12 hours, the medium was changed to Seahorse base medium supplemented with 10 mM glucose, 2 mM glutamine, and 1 mM sodium pyruvate adjusted to pH

7.4. The probe cartridge was prepared to have 0.5 µM Oligomycin in port “A”, 0.25 µM FCCP in port “B”, and 0.25 µM Rotenone with 0.25 µM Antimycin in port “C”. Oxygen consumption rate and extracellular acidification rate were measured by the Seahorse XFe96 Analyzer.

3.2.7 Sample preparation for mass spectrometry

For metabolomic profiling, 5X106 cells were evaluated in triplicate. Cells were collected using trypsin. All cell pellets were stored at -80 °C until analysis. Cell pellets were thawed at 4

°C and subjected to freeze-thaw cycles in liquid nitrogen and ice three times to rupture the cell membrane. Following this, 750 µl of ice-cold methanol/water (4:1) containing 20 µl of internal standards were added to each cell extract. Next, ice-cold chloroform/water (3:1) was added.

Organic and aqueous layers were separated. The aqueous portion was deproteinized using a 3-

KDa filter (Amicon Ultracel-3K membrane, Millipore Corporation), and the filtrate containing metabolites was dried under vacuum (Genevac EZ-2plus). Prior to mass spectrometry, the dried extracts were re-suspended in water/methanol (1:1) with 0.1% formic acid and subjected to liquid chromatography mass spectrometry.

3.2.8 Liquid chromatography/mass spectrometry

Targeted metabolomics profiling was carried out with an Agilent 1290 Series LC and 6430

Triple Quadrupole Mass Spectrometer (Agilent Technologies). Reverse-phase (RP) and aqueous normal phase (ANP) chromatographic separations of metabolites were performed using liquid chromatography. The RP chromatographic separation was performed using a Zorbax

Eclipse XDB-C18 column (50 × 4.6 mm i.d.; 1.8 μm, Agilent Technologies) at both positive and negative polarity. The RP separation was also performed with Synergi™ 4 μm Max-RP 80 Å (100 38

× 4.6 mm, Phenomenex) with mass spectrometric negative polarity. ANP chromatographic separation was conducted with a Diamond Hydride column (4 µm, 100 Å, 2.1 × 150 mm,

MicroSolv Technology) and a Luna 3 μ NH2 column (4 µm, 100 Å, 2.00 × 150 mm, Phenomenex) at positive and negative polarity, respectively. The mixture of seven internal standard compounds was used to monitor the profiling process. Additionally, a characterized pool of mouse liver tissue was prepared and analyzed in tandem with the clinical samples. These controls were incorporated multiple times into the analysis in a randomized fashion such that sample preparation and analytical variability could be constantly monitored. Furthermore, one blank run was performed following the analysis of each clinical sample to prevent any carryover of metabolites. Single Reaction Monitoring (SRM) experiments were performed using a Triple

Quadrupole Mass Spectrometer. The optimized mass spectrometric operational parameters included the following source conditions: capillary voltage of 3000 V, source temperature of 350

°C, drying gas maintained at 10 ml/min, nebulizer pressure set at 35 psi, and fragmentor voltage set at 70 V. The collision energies used for fragmentation were set at 5–60 eV unless otherwise stated. Agilent MassHunter Workstation Data Acquisition software was used for the data acquisition. The mass spectrometric data was analyzed using QQQ Qualitative Analysis B.05and

QQQ Quantitative Analysis B 0.5 (Agilent MassHunter Qual and Qaunt).

3.2.9 U13C-Glutamine flux analysis

U-13C-Glutamine (17 mg) was added to 50 mL of HMLE medium and filtered. HMLER-FOXC2 and HMLER-vector cells were seeded into wells of 10 cm plates with at 70% confluency; each cell type was analyzed in four replicates. After 24 hours, medium was aspirated, and 5 mL of labeled glutamine solution was added to the cells. A sample was collected at Time 0 for each plate. After 6 hours, medium was removed by aspiration, and a sample was collected. Cells were washed with ice cold PBS and 1 mL of 1:1 methanol:water was added to the cells. Cells were scraped into 15-mL Falcon tubes and immediately flash frozen by submerging tubes into liquid nitrogen. Samples were stored at -80 C until metabolite extraction.

3.2.10 Statistical analyses

Gene expression analysis was performed on a previously published dataset (Taube et al., 2010b) as previously described in (Bhowmik et al., 2015). Analysis was performed using the

‘limma’ and “affy” packages in R. All p-values were adjusted for multiple testing using the

Benjamini Hochberg method, significance was set at an adjusted p-value < 0.001. Metabolic enzymes and transporters were identified using a previously published comprehensive gene set

(Possemato et al., 2011).

The EMT metabolic signature was determined by testing log-transformed metabolomics data via linear modeling with matrices specified for each group (HMLESNAIL, HMLETWIST,

HMLEGOOSECOID, and HMLEGFP) and comparisons for each EMT groups against control.

Analysis was performed using the ‘limma’ package in R. The EMS is composed of only those metabolites with >1.5 log fold change and adjusted p-value < 0.0001 in each EMT model relative to control. To test whether the EMS had potential prognostic value, we applied principle component analysis to previously published metabolomics datasets (Terunuma et al., 2014) utilizing either the EMS metabolites or metabolic profiles associated HMLESNAIL, HMLETWIST, or

HMLEGOOSECOID. Patient samples were stratified into two groups on the basis of principle components using the ‘pca3d’ package in R, PC1 and PC2 accounted for nearly 90% of the variation in all tests. Group 1 was composed of all samples with PC1 and PC2 scores ≥ 1. These groups were then compared for clinical and biological parameters relevant to EMT and metastasis including breast cancer subtype, lymph node invasion, EMT gene expression, and overall survival. Survival analysis was performed using an age-adjusted multivariate Cox proportional hazards model that included EMS stratification group, grade, stage, and ER status.

The model and Kaplan-Meier plot were generated using the ‘survival’ package in R. Frequency of lymph node invasion and ER status by EMS subgroup was determined by testing variables by

EMS groups using Fisher’s exact test. PAM50 subtype distribution was tested for significance using chi-square test. For gene expression analysis, the ‘affy’ and ‘limma’ packages were utilized

40 to perform differential gene expression analysis between Group 1 and Group 2 samples for which gene expression data was available, and a list of known EMT markers were selected for observation with significance set at p < 0.05.

Unless otherwise stated, all samples were assayed in triplicate. All in vitro experiments were repeated at least three independent times, and the animal study included 10 mice in each group. Unless otherwise indicated, data are represented as means ±s.e.m., and significance was calculated using the Student’s unpaired two-tailed t-test.

3.2.11 Biolog phenotypic microarray to measure amino acid utilization

Cells were plated in triplicate per cell line at 1X104 cells per well into PM-M2 (Biolog

13102) 96-well Biolog plates. Cells were cultured in 50 µl of Biolog M1 medium (Biolog 72301) supplemented with penicillin and streptomycin, 200 mM glutamine, and 5% FBS. After 24 hours,

10 µl of Biolog MA redox dye (Biolog 74351) was added to each well, and the plate was placed in the Biolog OmniLog2.3 plate reader to obtain kinetic data at 15-minute intervals for 24 hours.

Data were analyzed with Biolog software pacages PMM_Kinetic and PMM_Parametric to obtain the average of the initial rate values at 8 hours.

3.2.12 Mitochondria staining

Cells were plated on glass coverslips at 50% confluency. After 24 hours cells were stained with 20 nM MitoTracker RedCMXRos (Life Technologies) in culture conditions. After 30 minutes, cells were fixed with 4% paraformaldehyde. After washing with PBS, coverslips were treated with

5% glycine in PBS for 15 minutes. Actin was stained with 488 Phalloidin (Alexa Fluor) and the nuclei were stained with DAPI. Fluorescent images were obtained with a Zeiss fluorescent microscope. The fluorescence intensities of 100 cells were analyzed and quantified by ImageJ software to obtain the relative fluorescence intensity (RFI).

3.2.13 ATP measurement

Cells were plated in triplicate at 1X104 per well in a 96-well plate. After 24 hours, an ATP standard curve was prepared, and 100 µl of CellTiter-Glo reagent was added to each test sample and standard. After incubating at room temperature for 10 minutes, the luminescent signal was recorded with a luminometer.

3.2.14 Cell cycle analysis

Cells were synchronized cells by treatment with 100 nM nocodazole for 14 hours. After trypsinization and pelleting, cells were resuspended in 1 mL of ice cold PBS in a 15-mL Falcon tube. Cells were fixed by adding 3 mL of 70% ethanol in a dropwise manner while vortexing.

Samples were stored at -20 C until staining. To stain, samples were centrifuged and

PBS/ethanol was aspirated. Cell pellets were washed with ice cold PBS, 300 l of 40 g/ml propidium iodide (Sigma P4170) with RNAase A (Thermo Fisher) was added, and samples were placed on ice. Samples were run on a BD Accuri flow cytometer, and cell cycle was evaluated by FlowJo software.

3.2.15 Cell surface marker analysis

Cells were plated at 0.1X106 per well in 6-well plates in triplicate. After 24 hours, cells were harvested by trypsin and resuspended in 500 l FACs buffer (PBS with 4% FBS). Cells were treated with anti-CD24 antibody conjugated with PE (BD Biosciences), anti-CD44 conjugated with APC (BD Biosciences), and anti-GD2 conjugated with PE (Biolegend 357304) and were incubated for 30 min on ice. The cells were washed with PBS and resuspended in 200

l FACS buffer. The cells were analyzed using the BD Accuri.

3.2.16 Western blot

Cell pellets were obtained by trypsinization. Cell pellets were lyzed with RIPA buffer

(Sigma R0278) containing phosphatase inhibitor (Roche 4906837001) and protease inhibitor

(Roche 11697498001) and vortexed vigorously then incubated on ice for 30 minutes, spun down, 42 and cell debris removed. After quantification by Bradford assay, 50 g of protein was denatured in -mercaptoethanol at 100 C for 10 minutes. Samples were loaded onto a 10% SDS-agarose gel and run at 110 V. For blots, antibodies were used at 1:1000 in 5% milk in TBST. Antibodies used were anti-Snail (Cell Signalling 3879S; 1:1000 in BSA), anti-FOXC2 (Bethyl A302-383A;

1:1000 in 5% milk), anti-ZEB1 (Santa Cruz sc-25388; 1:1000 in 5% milk), anti-actin (Santa Cruz sc-1616-R; 1:1000 in 5% milk), anti-tubulin (Cell Signaling 2144s; 1:1000 in BSA), and anti-GLS2

(generously provided by Dr. Jose Mates, University of Malaga, Malaga, Spain; 1:200 in BSA).

3.2.17 RNA extraction and qRT-PCR

Total RNA was isolated using the RNeasy Plus kit (Qiagen) according to the manufacturer’s instructions. All quantitative reverse transcription PCR (qRT-PCR) experiments were run in triplicate and a mean value was used for the determination of mRNA levels. Relative quantification of the mRNA levels was performed using the comparative Ct method with human

RPLPO as the reference gene and with the formula 2− ΔΔCt.

3.2.18 Evaluation of TGFβ induction

Cells were seeded at 0.1X106 cells per 10-cm plate and treated with 10 µg/µl of TGFβ for

6 days. Samples for RNA isolation taken after 12, 24, 48, 96, and 144 hours.

3.2.19 Immunofluorescence

Cells were plated at 50% confluency on glass coverslips in 6-well plates. After 24 hours, the medium was removed aspiration, and cells were washed with PBS. Cells were fixed in 4% paraformaldehyde at room temperature for 20 minutes and then washed with PBS. Cells were quenched with 4% glycine for 15 minutes at room temperature and washed with PBS. Cells were blocked with 5% BSA, 0.1% Triton-X100 in PBS for 1 hour at room temperature. After 3 washes with PBS, cells were incubated with primary antibody in 1% BSA for 2 hours at room temperature.

Cells were washed three times with PBS and then incubated with a secondary antibody at 1:1000 in 1% BSA for 1 hour at room temperature in the dark. Cells were washed with PBS and nuclear

43 stain DAPI was added to the cells at 1:1000. After 5 minutes in the dark. Cells were washed three times with PBS, and coverslips were mounted on glass slides with DAKO mounting solution and incubated at 4 °C overnight to dry. Fluorescent images of the slides were taken with the same exposure time for each slide.

3.2.20 Animal studies

All animal procedures were approved by the Institutional Animal Care and Use Committee

(IACUC). Luciferase labeled 4T1 cells (1X104 cells in 100 µl of DMEM/F-12, 10% FBS, with penicillin/streptomycin) were injected into the mammary fat pad of 4-6 week old Balb/c female mice. Injection efficiency was verified by injecting 1.5 mg of D-Luciferin subcutaneously and imaging with an IVIS Spectrum. The cells were injected at a single site per mouse, and each group had ten mice. Tumors were allowed to grow for 7 days, at which time tumors were palpable.

Treatment began on day 7. Group one received the 0.9% saline vehicle by intraperitoneal (IP) injection. Group two received 3 cycles of 100 mg/kg of carboplatin, every seventh day by IP injection. Group 3 received daily subcutaneous injections of 3.8 mg/kg of SB203580 (in 0.9% saline). Group four received a combination of carboplatin (3 cycles, 100 mg/kg, every 7th day) and SB203580 (3.8 mg/kg, 5 consecutive days, 3 cycles). Mice were imaged every week, and monitored daily for health status. At the endpoint, mice were euthanized by CO2 and cervical dislocation. Circulating tumor cells were harvested by cardiac puncture, the lungs and tumor were harvested and imaged immediately after dissection. Dissected tissue was held in PBS overnight and then fixed in formalin.

3.2.21 Drugs

SB203580 (Selleckchem S1076), carboplatin (MD Anderson Pharmacy), paclitaxel

(Sigma T7402), and 5-Aza-2’-deoxycytidine (Sigma 3656) were used in this study.

3.2.22 Plasmids and viral transduction

HMLER-derived cell lines, HMLER-SNAIL and HMLER-FOXC2, were generated as described previously (Hollier et al., 2013b). The S367E and S367A FOXC2 mutants were generated as described previously (Werden et al., 2016). The HMLE-ZEB1, HMLER-ZEB1,

HMLE-Snail-shFOXC2pZEB1, HMLER-Snail-shFOXC2pZEB1, and SUM159shFOXC2pZEB1 cell lines were made with a lentiviral FUW-2A-mStrawberry ZEB1 expression construct (a gift from Li Ma, MD Anderson) and selected with 4 µg/mL puromycin. For stable GLS2 expression, a GLS2 cDNA expression vector (purchased from MD Anderson core) was sub-cloned into the pHAGE-EF1alpha-PURO vector (a gift from Kenneth Scott, Baylor College of Medicine, Houston,

TX) using the Gateway system (Invitrogen) and selection with 4 µg/mL puromycin. pGIPZ-based shRNAs designed to deplete cells of GLS2 and ZEB1 were purchased from MD Anderson shRNA core.

3.2.23 Proliferation assay

Cells were plated in triplicate at 1X104 cells per well in 6 well-plates and counted manually every 24 hours for 72 hours after staining an aliquot with Trypan Blue.

3.2.24 Mammosphere assay

Mammosphere culture medium was made with 1% methylcellulose and MEGM prepared as described above without bovine pituitary extract. In a 96-well low attachment plate, 500 cells per well in six replicates were cultured for 14 days. Every 3 days, 50 µl of fresh media was added.

Mammospheres were photographed, and spheres with a diameter greater than 80 µm were counted.

3.2.25 Lactate assay

In a 96 well plate, 1X104 cells were plated per well in triplicate. After 24 hours, spent medium was diluted 1:50, and lactate was measured with EnzyChrom L-Lactate Assay Kit

(BioAssay Systems).

3.2.26 ROS detection

Cells were stained with 2 µM of ROS Indicator CM-H2DCFDA (Invitrogen) and incubated for

1 hour, trypsinized, washed with PBS, and resuspended in 300 µl FACS buffer (2% FBS in PBS).

Samples were analyzed on BD Accuri in FL2, and FlowJo was used to plot cell populations.

3.3 Tables of Reagents

3.3.1 Table of primers

Gene Forward Primer 5’-3’ Reverse Primer 5’-3’

RPLPO GCGACCTGGAAGTCCAACTA ATCTGCTTGGAGCCCACAT

ACA CCC TCAGCC TCA TGC ATG GCT CCT GATACA GCT GLS2 AT GAC TT

CAC TGC CCT CCCATT ACC GAA GCT CAA GCA TGGGAA GLS TAG CAG

FOXC2 GCCTAAGGACCTGGTGAAGC TTGACGAAGCACTCGTTGAG

ZEB1 GCACAACCAAGTGCAGAAGA CATTTGCAGATTGAGGCTGA

Chapter 4: Metabolic alterations associated with EMT predict poor clinical outcome

Contents of this chapter are based on Bhowmik, S. K., E. Ramirez-Pena, J. M. Arnold, V. Putluri,

N. Sphyris, G. Michailidis, N. Putluri, S. Ambs, A. Sreekumar and S. A. Mani (2015). EMT- induced metabolite signature identifies poor clinical outcome. Oncotarget 6(40): 42651-42660.

Copyright permission was not required as Oncotarget journal states “authors retain ownership of the copyright for their article”.

4.1 Introduction and rationale

Cancer cells require specific metabolic programs that provide energy and building blocks to maintain their highly proliferative phenotype (Ward and Thompson, 2012). Furthermore, dependencies of these metabolic programs reveal the possibility of exploiting the vulnerability of tumors to specific nutrients for therapeutics (Jang et al., 2013). The EMT is a latent embryonic program that is aberrantly induced in breast cancer cells conferring migratory and invasive capabilities that leads to metastasis, chemotherapeutic resistance, and relapse (May et al.,

2011). This process is re-enforced by the enhanced expression of EMT-TFs Snail, Twist, and

Goosecoid (Mani et al., 2008; Taube et al., 2010b). Although EMT marker transcript levels serve as indicators of poor metastasis-free survival in some cancers, in breast cancer they are not always good predictors of survival (Tan et al., 2014; Taube et al., 2010b). Furthermore, the changes in cellular metabolism that occur during EMT are not well understood.

Gene expression profiling using methods such as microarrays and RNA-seq have been employed to generate gene signatures that are clinically relevant and are capable of predicting clinical outcomes in breast cancer patients (Lehmann et al., 2011; Prat et al., 2010). This methodology has led to the discovery of molecular differences between the breast cancer subtypes and an enhanced understanding of breast cancer heterogeneity (Beca and Polyak,

2016; Lehmann et al., 2011). Furthermore, this technology has also led to the development of clinically actionable tests to evaluate risk factors for disease progression (Lehmann and 48

Pietenpol, 2014; Taube et al., 2010b). These analyses offer a limited view of the physiologic changes taking place within the tumors, however.

By improving our understanding of the differences in metabolic phenotypes that result from genetic differences, integration of metabolomics with genomics and proteomics has the potential to improve disease diagnostics. Traditional gene expression profiling evaluates changes in transcription that may or may not exert functionally relevant phenotypes.

Metabolomics measures metabolites that arise from metabolic reactions and are a direct readout of biochemical activity and cellular phenotype (Fiehn, 2002). We reasoned that studying altered metabolic pathways and metabolites associated with EMT will uncover the global phenotypic changes that occur during this physiological program. It is possible that alterations in certain metabolic pathways may facilitate EMT during tumor progression.

The reason for developing metabolite-based screening assays for clinical risk assessment in breast cancer is to improve diagnostics. In this study, multiple cell line models of immortalized human mammary epithelial cells were induced to undergo EMT by stably expressing EMT- inducing transcription factors. Cell lines that express Snail, Twist, or Goosecoid (HMLESNAIL,

HMLETWIST, HMLEGOOSECOID, respectively) were used to study the metabolic alterations associated with EMT. This study sought to identify phenotypic metabolic signature associated with EMT and determine if it could serve as a prognostic tool to assess patient outcome.

4.2 Results

4.2.1 EMT induces significant metabolic reprogramming

A previously described microarray (Taube et al., 2010b) comparing the transcriptomes of the three cell lines HMLESNAIL, HMLETWIST, and HMLEGOOSECOID to the control HMLEGFP transcriptome revealed that 13% of the common differentially expressed transcripts were metabolic enzymes and transporters (Figure 4). The differentially expressed genes include those encoding factors associated with glycolysis and with glutathione, aminosugars, and nucleotide metabolism. These

49 findings prompted our investigation of the metabolic alterations induced during EMT. An in-depth analysis of the genes associated with metabolism during EMT is described in Chapter 5.

Figure 4: Analysis of metabolism-associated genes differentially expressed upon EMT. Of genes (p<0.001) differentially expressed between HMLESNAIL, HMLETWIST, and HMLEGOOSECOID cells and the epithelial control HMLEGFP, 13% are associated with major metabolic pathways.

4.2.2 Individual models exhibit unique metabolite abundances

To investigate the metabolic alterations associated with EMT, targeted SRM-based mass spectrometry was used to measure the differential levels of metabolites between each of the models in which EMT is induced by expression of an EMT-TF (HMLESNAIL, HMLETWIST,

HMLEGOOSECOID) relative to the epithelial control HMLEGFP. In total, 97 metabolites were measured across all cell lines. Each EMT transcription factor generated a unique metabolic signature

(Figure 5). HMLESNAIL cells had a 15-fold increase in N-acetylaspartylglutamate, a peptide neurotransmitter (Neale et al., 2000), and a 3.5-fold increase in lactate, a product of glycolysis

(Sun et al., 2017) relative to the control cells. The metabolites that were significantly reduced in

HMLESNAIL cells included pyroglutamate (3 fold), which is a product of glutamate metabolism; hippurate (3.8 fold), an acylated glycine product; methylnicotinamide (4.1 fold), a product of nicotinamide metabolism; N-acetylmethionine (9.4 fold), an acetylated form of methionine; ornithine (10 fold), a component of the urea cycle; and xanthine (140 fold) a purine base formed by the degradation of adenosine monophosphate (Figure 5A).

The metabolites with significantly higher abundance in the HMLETWIST cells than control cells included UDP-glucose (3 fold), glucose-6-phosphate/fructose-6-phosphate (4.2 fold), and

UDP-glucuronate (5.2 fold). These metabolites are associated with the nucleotide sugar pathway. Other abundant metabolites such as cystathionine (11.2 fold), a precursor to homocysteine; S-adenosylmethionine (5.2 fold), critical for transmethylation reactions; and citrate (3.7 fold), a critical TCA cycle intermediate, were also specifically increased HMLETWIST cells (Figure 5B).

Unique to HMLEGOOSECOID were increases in abundance of glucose/fructose isomer (3.2 fold), which contributes to glycolysis; methionine sulfoxide (3.6 fold), a marker for oxidative stress; glycylproline (3.6 fold) a product of collagen degradation; glucosamine-6-phosphate (3.6 fold), which contributes to de novo glucosamine synthesis; spermidine (4.1 fold), a polyamine;

52 lysine (6.32 fold), an essential amino acid; and 2-aminoadipate (36 fold), a product of lysine degradation. HMLEGOOSECOID cells had the most elevated abundances of metabolites of the three cell lines analyzed. The metabolites decreased in HMLEGOOSECOID cells relative to the control were reduced glutathione (3.4 fold), an important antioxidant; S-adenosylhomocysteine (4.2 fold), a product of methylation reactions; and cysteine (4.4 fold), an amino acid that contributes to redox reactions (Figure 5C).

Some metabolites were increased in abundance (at least 1.5-fold increase) in two of the cell lines induced to undergo EMT. In HMLESNAIL and HMLETWIST cells, both uridine and succinate were present at elevated levels. HMLETWIST and HMLEGOOSECOID shared elevated levels of spermidine and malate. Tyrosine and isoleucine were more abundant in HMLESNAIL and

HMLEGOOSECOID cells than in the control cells.

To define an EMT-metabolite signature (EMS), metabolites present at higher abundances

(at least 1.5-fold increase compared to HMLEGFP cells) in each of the three EMT-induced cell lines were chosen. Four metabolites met these criteria: beta-alanine, glutamine, glutamate, and glycylleucine (Figure 6A-B). Beta-alanine is a product of pyrimidine degradation. Glutamine is a substrate that contributes to the product glutamate during glutaminolysis. Glutamate is a neurotransmitter. Glycyleucine is a dipeptide whose role is not clearly understood. Three metabolites were decreased in the three EMT-induced cells: phosphoenolpyruvate, a glycolytic intermediate; urate, a purine degradation product; and deoxycarnitine, a precursor carnitine.

These were not included in the EMS. These findings demonstrate that EMT-induced cells exhibit significant changes in abundance of metabolites involved in metabolic pathways such as glycolysis, TCA cycle, redox homeostasis, aminosugar metabolism, and nucleotide metabolism in vitro.

Figure 5: Individual heatmaps of significantly differential metabolites. Significantly different metabolites in (A) HMLESNAIL, (B) HMLETWIST, and (C) HMLEGOOSECOID cells relative to the control

HMLEGFP cells (HMLE-VC).

Figure 6: Metabolites with significantly elevated abundances form the EMS. (A) Venn diagram of overlap of significantly elevated metabolites in HMLE cells that express indicated

EMT-inducing TFs (fold change >1.5) compared to control. (B) Table of significantly elevated metabolites.

4.2.3 Breast cancer patients can be stratified by the EMS

Cancer cells that exhibit features of EMT are highly aggressive and metastatic (Hollier et al., 2013a; Mani et al., 2008). To test if the EMS has prognostic value, a principal component analysis (PCA) was used to stratify human breast tumor samples on the basis of EMS abundance. The PCA was applied to a previously published metabolomics dataset of 67 human breast tumor samples (Terunuma et al., 2014). When data from the patient samples were plotted in the two-dimensional space of the PCA, the stratification of the groups was mostly driven by glutamate and beta-alanine. Glutamine and glycylleucine did not contribute significantly to the stratification. Patients were divided into two groups based on this analysis; Group 1 included all patient with principle component score ≥ 0 (Figure 7).

Next we sought to analyze clinical parameters such as survival, subtype, and ER status within the groups. Notably, we determined that 82% of Group 1 patients had been diagnosed with ER-negative tumors and only 44% of Group 2 patient tumors were ER-negative (p=0.001)

(Figure 8A). No significant association between lymph node status and the EMS was found

(Figure 8B). We further determined the distribution of PAM50 subtype and determined that 82% of Group 1 tumors were basal-like and HER2-over-expressing subtypes, whereas only 22% of

Group 2 were (Figure 8C). Additionally, we found a significant difference in survival between the two groups (hazard ratio (HR) = 2.3; 95% confidence interval (CI): 1.31-4.20, p=0.03; Figure 8D).

Interestingly, the single HMLEGOOSECOID metabolic profile also yielded a significant stratification

(HR=5.43; CI: 2.29-12.88; p=0.00002); however, the HMLESNAIL and HMLETWIST metabolic profiles did not (for HMLESNAIL, HR = 1.52; CI: 0.85-2.73; for HMLETWIST, HR = 0.93, CI: 0.49-

1.76; Figure 9A-C).

To test for differences in expression of genes associated with EMT between Group 1 and

Group 2 patients, we analyzed tumor gene expression profiles. Amongst the EMS stratified

56 groups, we found no significant differences in abundances of genes known to be involved in

EMT: CDH2, GOOSECOID, TWIST1, ZEB2, SNAI1, and VIM (Figure 9 D).

Collectively, our data provide evidence of metabolic changes that accompany EMT.

Furthermore, these metabolic changes are linked with characteristics of aggressiveness and poor overall survival in patients. The stratification based on EMS was not associated with differences in expression of genes that are up-regulated during EMT.

Figure 7: Principal component analysis of EMS in patient metabolomic profiles. A previously described metabolomics dataset (Terunuma et al., 2014) was used and patient groups were created on the basis of PC1 and PC2, which accounted for 94% of the variation in the data.

Group 1 includes all patient samples with PC1 and PC2 scores >1.

Figure 8: The EMS is associated with aggressive breast cancer molecular subtypes. (A)

ER status distribution in Group 1 and Group 2. (B) Lymph node status in Group 1 and Group 2.

(C) EMS is associated with aggressive breast cancer molecular subtypes. Patients in Group 1 had high frequency of basal-like and HER2-over-expressing molecular subtypes. Patient tumors in Group 2 were predominantly of luminal subtype. (D) EMS metabolites are predictive of worse overall survival as shown for patients stratified based on EMS.

Figure 9: EMT expression within tumor samples from Group 1 and Group 2 samples.

Associations of (A) HMLESNAIL, (B) HMLETWIST, and (C) HMLEGOOSECOID metabolic profiles with patient survival in groups defined by EMS. (D) EMT marker expression in groups defined by

EMS.

4.3 Discussion

In this chapter I described how we arrived at an EMS using three different cell lines in which EMT was induced by expression of an EMT-TF. We conducted targeted metabolomic profiling to identify alterations in biochemical processes in these cells relative to a control HMLE line. We profiled cell line models in which EMT had been induced by mass spectrometry and applied the observed signature to disease prognostication. The metabolites that make up the

EMS are amino acids glutamine, glutamate, and beta-alanine and a di-peptide glycylleucine.

With the exception of glycylleucine, the three amino acids of the EMS are at the center of multiple pathways important in cancer such as pyrimidine metabolism, TCA, and glutaminolysis. This analysis identified metabolites that are likely critical for the bioenergetics of EMT rather than merely consequences of the transition. Future work remains to be done to thoroughly characterize how these metabolites become abundant and to conclusively demonstrate that they offer bioenergetic support during EMT.

Chapter 5: The role of FOXC2 in EMT-associated metabolic reprogramming

5.1 Introduction and rationale

For cancer cells to adapt, survive and grow in an environment of fluctuating nutrient and/or oxygen availability, like that of the tumor microenvironment, metabolic reprogramming must occur. Metabolic reprogramming is the re-wiring of bio-energetic pathways to give cancer cells an advantage during tumorigenesis. Reprogrammed bio-energetic pathways provide the highly proliferative cancer cells with support for the production of biomass, ensure redox balance, and enable growth under stresses like nutrient and oxygen deprivation (DeBerardinis and

Chandel, 2016).

There is increasing evidence that the acquisition of EMT properties requires metabolic adaptations that enable survival in environments in which nutrient and oxygen fluctuate. Several reports provide evidence that not only do EMT-TFs regulate classical EMT processes such as invasion, migration, chemotherapeutic resistance, and resistance to apoptosis, but they also regulate metabolic reprogramming. In support of this, studies show that individual EMT-TFs regulate specific metabolic genes. For example, Snail represses transcription of FBP1, a bisphosphatase that diverts carbons toward the pentose phosphate pathway, favoring glucose uptake and inhibiting oxidative phosphorylation (Dong et al., 2013). Furthermore, Wnt-Snail signaling inhibits production of COX, the cytochrome C oxidase, in the mitochondrial respiratory chain inducing metabolic reprogramming toward glycolysis (Lee et al., 2012). The gene encoding

EMT-TF Twist is a direct target of HIF-1α and over-expression of Twist results in the up-regulation of critical metabolism genes that encode the lactate dehydrogenase LDHA, pyruvate kinase

PKM2, hexokinase HK2, and the phosphate dehydrogenase G6PD involved in NADPH production (Yang and Wu, 2008).

A key mediator of metastatic potential, EMT-TF FOXC2 regulates adipocyte metabolism.

A few reports have linked FOXC2-driven EMT with metabolic changes. One study employed murine models of claudin-low and basal-like breast cancer to determine if diet-induced obesity

(DIO) and calorie restriction (CR) affected cancer progression in these aggressive breast cancer 63 subtypes. Mice orthotopically implanted with either mesenchymal (M)-Wnt or epithelial (E)-Wnt cells derived from MMTV-Wnt-1 transgenic mouse mammary tumors were fed DIO or CR diets.

Interestingly, the DIO diet enhanced tumor progression, increased levels of intratumoral adipocytes, and enhanced expression of FOXC2 among other EMT-TFs in M-Wnt cells. In contrast, a CR diet significantly reduced tumor progression and the expression of EMT- associated markers (Dunlap et al., 2012). These findings suggest that (M)-Wnt cells respond to dietary energy changes and that a DIO diet supports an EMT phenotype, which is associated with high FOXC2 levels and poor prognosis in breast cancer patients. In nasopharyngeal carcinoma, FOXC2 expression promotes expression of the glycolytic enzyme HK2 through its interaction with the transcriptional regulator YAP. Glycolysis is inhibited in FOXC2-expressing cells in vitro when HK2 is targeted with the chemical inhibitor 3-BrPA. The tumorigenic potential of FOXC2-expressing cells is significantly inhibited in vivo with 3-BrPA treatment (Song et al.,

2017). These data suggest that FOXC2 expression promotes the expression of metabolism- associated genes and that targeting the FOXC2-induced metabolic phenotype will significantly attenuate tumorigenic potential. I hypothesized that FOXC2 dictates metabolic functions during

EMT and metastasis.

In this chapter, I describe how the expression of metabolic genes altered in EMT are regulated by FOXC2 and present evidence from a metabolomics analysis that metabolic reprogramming occurs in cells that have undergone EMT in parallel with changes in expression of metabolism-associated genes. Determining if the metabolism-associated genes with altered expression in cells post EMT are directly linked with the metabolite abundance changes will improve will guide our understanding of the bioenergetics needs that are required for the EMT- associated functions that promote cancer stem cell properties, invasion, and metastasis.

5.2 Results

5.2.1 Metabolic genes altered in EMT

To determine identify genes associated with metabolism changes that occur upon induction of EMT, we analyzed a previously reported microarray containing samples of immortalized HMLE cells induced to undergo EMT by expression of EMT-TF Snail, Twist, or

Goosecoid (HMLESNAIL, HMLETWIST, HMLEGOOSECOID, respectively) compared to vector control

(HMLEGFP) described in Chapter 4. We identified genes that were differentially expressed in the three models of induced EMT compared to the control cells (Figure 10). Many of the genes downregulated upon EMT encode enzymes that operate in nucleotide metabolism such as adenyl succinate lyase (ADSL), purine nucleoside phosphorylase (PNP), hypoxanthine phosphoribosyltransferase (HPRT1), uridine monophosphate synthetase (UMPS), uridine phosphorylase (UPP1), adenine phosphoribosyltransferase (APRT), and xanthine dehydrogenase (XDH). Other genes encode factors involved in spermidine biosynthesis (SRM), amino acid metabolism (GOT1), and the oxidation of L-2-hydroxyglutarate to α-ketoglutarate

(L2HGDH). Upregulated metabolic genes are involved in the generation of nitric oxide (DDAH1/2) and the synthesis of triglycerides and glycerophospholipids (GK). Interestingly, we also identified two genes that encode enzymes that convert glutamine to glutamate: GLS and GLS2. Based on the microarray data, GLS expression is upregulated during EMT and GLS2 expression is downregulated. Previous reports have also shown that GLS expression is enhanced in EMT- enriched non-small cell lung cancer cells (Ulanet et al., 2014) and breast cancer cells (Lee et al.,

2016b). Remarkably, it was also reported that EMT-enriched non-small cell lung cancer cells exhibit a reduced capacity for oxidative phosphorylation despite GLS expression.

Figure 10: Metabolic genes that are altered in EMT. Differentially expressed genes associated with metabolism in HMLEGoosecoid (GSC), HMLESNAIL, and HMLETWIST cells compared to HMLEGFP control. Upregulated genes are red. Downregulated genes are green.

5.2.2 XDH expression is enhanced and nucleotides are increased in abundance in

FOXC2-deficient cells

A majority of the downregulated metabolic genes in the three models of induced-EMT encode enzymes involved in purine nucleotide metabolism. We have previously demonstrated that depletion of FOXC2 in vitro inhibits EMT properties such as migration, invasion, and metastasis and reduces expression of cancer stem cell markers in vivo (Hollier et al., 2013b)

(Mani et al., 2007b). Further, FOXC2 plays a critical role in regulating metabolism in adipocytes

(Cederberg et al., 2001). I used a short hairpin RNA (shRNA) designed to deplete cells of FOXC2 to test the hypothesis that FOXC2 regulates metabolism-associated genes in EMT. To determine if inhibiting FOXC2 restores expression and results in biochemical consequences, I quantified the XDH, UPP1, APRT, ADSL, HPRT1, PNP, and UMPS mRNAs in HMLESNAIL cells with and without expression of FOXC2 shRNA. Of the seven genes, three (XDH, UPP1, and APRT) were upregulated by FOXC2 depletion (Figure 11A); there were no significant differences in levels of the other mRNAs in treated and control cells (Figure 11B). XDH exhibited the most striking upregulation upon FOXC2 depletion (100-fold upregulation), therefore I hypothesized that its expression and function could be functionally relevant to EMT.

Figure 11: FOXC2 regulates the expression of XDH, UPP1, and APRT in EMT (A) qRT-PCR analysis of metabolism genes XDH, UPP1, and APRT in vector control (V) and shFOXC2-treated cells. (B) qRT-PCR analysis of ADSL, HPRT1, PNP, and UMPS in control and shFOXC2-treated cells.

XDH is an enzyme in the de novo purine synthesis pathway that catalyzes the oxidation of hypoxanthine to xanthine and then xanthine to uric acid. Hypoxanthine and xanthine contribute to the production and salvage of adenine and guanine purines, which are components of DNA and RNA. In hepatocellular carcinoma, low XDH expression is correlated with higher tumor grades and poorer prognosis (Chen et al., 2017). The same study evaluated effects of inhibition and enhancement of XDH activity. Reduction of XDH activity through shRNA treatment or chemical inhibition by oxypurinol enhances EMT properties such as migration, invasion, and

EMT-TF expression in vitro. Over-expression of XDH reduces TGFβ signaling, cell migration, invasion in vitro, and lung colonization by MHCC97H cells after tail vein injection. This evidence suggests that inhibition of XDH supports EMT. The study did not, however, propose a regulatory mechanism upstream of XDH that could explain its loss of expression during EMT. Furthermore, the study did not provide metabolomics data to determine if products of XDH activity are decreased or if the entire nucleotide pathway is attenuated.

To determine if there are differential global nucleotide abundance differences before and after EMT, we used mass spectrometry to measure abundance of intracellular nucleotides in

HMLERSNAIL and HMLERFOXC2 cells compared to epithelial control cells (HMLERGFP). The cells that had undergone EMT had significantly higher levels of hypoxanthine and lower levels of xanthine than control cells (Figure 12A). I hypothesized that the observed accumulation of hypoxanthine and decrease in xanthine after EMT resulted from reduced XDH expression. This hypothesis is supported by the observation that XDH expression was higher in cells treated with shFOXC2 than in control cells (Figure 11A). Moreover, the cells treated with shFOXC2 had higher levels of xanthine than HMLERFOXC2 cells (Figure 12B). The levels of hypoxanthine remained relatively high in the shFOXC2-treated cells, therefore the enhanced xanthine abundance is a result of enhanced XDH expression.

I hypothesized that reduced XDH activity and xanthine abundance used for de novo purine synthesis during EMT affects DNA synthesis and is reflected by cell cycle differences. To test this hypothesis, I performed a cell cycle analysis in control and HMLERSNAIL cells (Figure 69

12C). The HMLERSNAIL cells exhibited a longer DNA synthesis phase (S-phase) than the

HMLERGFP cells. EMT-enriched CSCs are generally regarded as slow-cycling cells (Moore and

Lyle, 2011). The quiescent or dormant state of CSCs enables these cells to be resistant to chemotherapeutics and consequently results in tumor recurrence (Borst, 2012). We have previously reported that FOXC2 regulates the G2/M transition of CSCs and that FOXC2 protein levels fluctuate in a cell cycle-dependent manner (Pietila et al., 2016). We propose that the regulation of G2/M transition is highly regulated in CSCs due to their slow cycling nature allowing them to have a higher capacity for DNA repair. I reasoned that there could be a link between low purine biosynthesis and longer S-phase. To determine if supplementing cells post EMT with nucleotides decreased the length of time to complete the S-phase, I supplemented HMLERSNAIL cells with a cocktail of ten nucleotides (Table 1). In support of my hypothesis of a link between efficiency of purine biosynthesis and cell cycle, the nucleotide cocktail-treated HMLERSNAIL cells remained predominantly in growth phase (G1-phase) for 14 hours (Figure 12D), whereas untreated HMLERSNAIL cells were predominantly in S-phase by 8 hours (Figure 12C).

To determine if the observed impact of XDH in vitro is relevant in breast cancer patients,

I analyzed breast cancer patient data from an online database “KMplotter.com”. All breast cancer patients were divided into two groups based on XDH expression. Interestingly, higher expression of XDH is associated with improved relapse-free survival (RFS) (HR=0.79, p=0.0028; Figure

12E). Strikingly, higher XDH expression is also associated with improved RFS in the basal breast cancer subtype, grade 3 breast cancers, and the mesenchymal-like subtype (Figure 12E). This observation is indicative that more aggressive tumors enriched with EMT/CSC properties exhibit lower XDH expression. These data collectively suggest that regulation of XDH expression during

EMT by FOXC2 plays a critical functional role. Understanding how products of purine de novo synthesis impact EMT properties warrants further investigation.

Figure 12: XDH and nucleotide metabolism influence EMT. (A) Heatmap of nucleotide metabolite abundances in HMLERFOXC2 and HMLERSNAIL cells compared to HMLER-vector cells.

Data from each of four replicates are shown. (B) Heatmap of metabolite abundances in

HMLERSNAIL cells that express shFOXC2 (shFOXC2) compared to HMLERSNAIL cells that express vector (V). Data from each of four replicates are shown. (C) Cell cycle analysis in HMLERSNAIL cells and HMLERGFP control cells. (D) Cell cycle analysis of untreated HMLERSNAIL cells and nucleotide cocktail-treated HMLERSNAIL cells. (E) RFS is shown for patients divided into groups with high (red) or low (black) expression of XDH.

5.3 Discussion

In this chapter, I provide evidence that FOXC2 regulates expression of genes associated with nucleotide metabolism. In particular, the metabolic enzyme XDH is significantly downregulated after EMT induction. XDH catalyzes the oxidation of hypoxanthine to xanthine and subsequently oxidizes xanthine to uric acid. Both hypoxanthine and xanthine contribute to the production and salvage of purines, which are critical components of DNA and RNA. To gather insight into its functional role during EMT, we employed a mass spectrometry-based approach to examine the abundance of nucleotides in cells that had undergone EMT compared to epithelial control cells. We found that the induction of EMT by EMT-TFs SNAIL and FOXC2 elicited an accumulation of hypoxanthine and significant reduction in xanthine. We attribute the accumulation of hypoxanthine to the decreased expression of XDH. Furthermore, we observed that inhibition of EMT by reduction of FOXC2 expression enhanced expression of XDH and the conversion of hypoxanthine to xanthine. The functional roles of the enzyme XDH and its metabolite product xanthine during EMT remains to be discovered, but our analysis of gene expression data from breast cancer patient samples indicates that higher levels of XDH expression are correlated with better disease outcomes.

Chapter 6: Characterizing the functional relevance of altered glutamine metabolism in

EMT

Chapter 6: Characterizing the functional relevance of altered glutamine metabolism in

EMT

6.1 Introduction and Rationale

Although many studies have characterized metabolism within primary tumors and therapeutics have been developed to target primary tumor metabolism, the bioenergetic programs that support EMT poorly understood. EMT has been shown to contribute to intra-tumor heterogeneity resulting in the aggressive attributes of metastatic breast cancer (Scheel and

Weinberg, 2012). The properties that EMT promotes elicit migratory and invasive features and expression of stem cell surface markers (CD44high/CD24low), mesenchymal markers (vimentin, fibronectin, and N-cadherin), and transcription factors such as ZEB1 and FOXC2 (Hollier et al.,

2013a). We have previously demonstrated that EMT promotes and maintains p38MAPK-FOXC2 signaling in breast and prostate cancer (Paranjape et al., 2016b; Werden et al., 2016). We demonstrated that chemical inhibition of p38MAPK destabilizes FOXC2 protein reversing EMT in vitro and resulting in reduced metastatic potential in vivo. Interestingly, FOXC2 is a forkhead mesenchyme transcription factor that is only expressed in adult adipocytes where it directs a metabolic network of energy turnover and mitochondrial biogenesis promoting expression of mitochondrial and metabolic genes such as CPT1 and PPARγ (Cederberg et al., 2001;

Hakansson et al., 2011). I hypothesized that in addition to conferring metastatic potential, FOXC2 controls a metabolic program that supports EMT. As EMT contributes to tumor heterogeneity, it is possible that this process also contributes to metabolic heterogeneity within the tumor warranting further investigation of the metabolic changes induced by EMT.

6.2 Results

6.2.1 The epithelial to mesenchymal transition operates with reduced mitochondrial

metabolism

Our previous work determined that induction of EMT results in metabolic alterations and identified an EMT-metabolite signature (Bhowmik et al., 2015). Two of the amino acids in the

EMS are glutamine and glutamate; therefore, I hypothesized that they contribute to the bioenergetics necessary for the onset of EMT. Because metabolite abundance alone cannot discriminate whether there are alterations in conversion or accumulation I sought to determine if cells that have undergone EMT exhibit differential utilization of amino acids by performing a

Biolog unbiased phenotypic microarray assay. I obtained kinetic data over 24 hours that measured the utilization of individual amino acids in HMLERSNAIL and HMLERGFP cells. I compared the amino acid levels in the two cell lines at 8 hours (Figure 13A). This time point was chosen as signal had not yet reached saturation at this time. Negative values were designated as loss of utilization and positive values were designated as gain in utilization. I chose differences less than -1.5 and higher than 1.5 as the cut off for significant loss or gain, respectively. One of the top four metabolites with significant loss of utilization in HMLERSNAIL cells was L-glutamine.

That cells that have undergone EMT did not utilize glutamine was a surprising finding because glutamine was one of metabolites of the EMS. I reasoned that the accumulation could be due to reduced glutaminolysis.

Cells that have transitioned through EMT employ glycolysis, and therefore I sought to determine if cells were susceptible to decreasing glucose. To determine the concentration of

GFP SNAIL glucose at which viability decreases by 50% (IC50), I cultured HMLER , HMLER , and

HMLERFOXC2 for 24 hours in a range of concentrations of glucose and measured viability by MTS

SNAIL FOXC2 assay. The HMLER and HMLER cells were viable at lower concentrations (IC50 = 1.5

GFP and IC50 = 1.07 respectively) of glucose than the HMLER cells (IC50 = 3.47) (Figure 12B). This finding was surprising because I expected the cells that had undergone EMT to be more sensitive to reductions in glucose than epithelial control cells. The data indicate that after EMT cells have a higher degree of metabolic flexibility because they are able to survive despite low glucose concentrations even in the absence of glutamine.

I next measured the oxygen consumption rate (OCR) and extracellular acidification rate

(ECAR) over time with a Seahorse MitoStress assay in cells in post-EMT cells and control epithelial cells. Compared to HMLERGFP cells, HMLERSNAIL and HMLERFOXC2 cells exhibited lower spare respiratory capacity in response mitochondrial complex inhibitors oligomycin (O), FCCP

(F), and rotenone with antimycin (R+A) over time (Figure 13C), had significantly lower basal respiration (Figure 13D), and were less energetic as determined by lower OCR/ECAR ratios

(Figure 13E). The cells that had undergone EMT also produced less ATP than control cells

(Figure 13F), which is in line with the observation that mitochondrial activity is reduced.

To determine if there are differences in conversion of glutamine to glutamate as a result of reduced mitochondrial activity, I measured the biochemical conversion of intracellular glutamine to glutamate in HMLERGFP, HMLERSNAIL, and HMLERFOXC2 cells. The cells expressing the EMT-TFs had a lower intracellular concentration of glutamate than control cells (Figure 13G).

I hypothesized that post EMT there would also be a lower production of GSH. Indeed, EMT- induced cells have lower GSH compared to vector cells (Figure 13 H) Moreover, HMLERSNAIL and HMLERFOXC2 cells had lower mitochondrial membrane potential than control cells as evaluated using MitoTracker Red staining (Figure 13I-J).

The metastatic breast cancer cell lines SUM159 and MDA-MB-231 are enriched in EMT properties compared to the non-metastatic cell line MCF7. I measured the mitochondrial membrane potential, conversion of glutamine to glutamate, and intracellular GSH in these three cell lines to determine if they behave similarly to the HMLER model cells post EMT. Indeed, the

EMT-enriched cancer cell lines exhibited lower conversion of glutamine to glutamate (Figure

14A), lower GSH (Figure 14B), and lower mitochondrial membrane potential (Figure 14C-D).

Similar to my observations, previous studies have determined that triple negative breast cancer cell lines exhibit lower mitochondrial respiration compared to ER-positive cells (Pelicano et al.,

2014).

It is possible that the reduced conversion of glutamine to glutamate and the enrichment of glutamine could be due to an accumulation in the cells that have undergone EMT. To interrogate this, we performed a steady-state mass spectrometry analysis of HMLERGFP cells and

HMLERFOXC2 cells. From this analysis, we determined that glycolytic metabolites (3PG/2PG,

FBP/GBP, G6P/F6P, and PEP) are significantly more abundant in FOXC2-expressing cells

(Figure 15A). Furthermore, we performed a mass spectrometry flux analysis of HMLERGFP and

HMLERFOXC2 cells with labeled glutamine (13C-glutamine) after 0 and 6 hours of incorporation

(Figure 15B). This analysis allows detection of incorporation of labeled carbons into downstream metabolites. There was significantly less incorporation of the 13C at position M+5 of GSH and glutamine in HMLERFOXC2 cells than in control cells at 6 hours. This provides further evidence that the metabolism of cells that have undergone EMT affects how glutamine is incorporated into downstream metabolites. Taken together, these data suggest that post EMT cells operate with a reduced mitochondrial metabolism and glutamine independence.

Figure 13: Induction of EMT results in repressed mitochondrial metabolism. Induction of

EMT results in repressed mitochondrial metabolism. (A) Bar graph of the difference in amino acid consumption between HMLERSNAIL cells and HMLERGFP cells. Green bars indicate <1.5 fold metabolic loss. Red bars indicate >1.5 fold metabolic gain. (B) Glucose IC50 of indicated HMLER cells. (C) Seahorse analysis of OCR. (D) Basal respiration. (E) OCR vs. ECAR. (F) ATP production relative to HMLERGFP cells. (G) Conversion of glutamine to glutamate relative to

HMLERGFP cells. (H) Intracellular GSH. (I) Representative images of MitoTracker Red staining of

(left to right) HMLERGFP cells, HMLERSNAIL cells, and HMLERFOXC2 cells. (J) Relative fluorescence intensity (RFI) of MitoTracker Red staining calculated with ImageJ software. 78

Figure 14: EMT-enriched breast cancer cell lines exhibit reduced mitochondrial metabolism. (A) Relative conversion of glutamine to glutamate in MCF7 and SUM159 cells. (B)

Intracellular GSH. (C) Representative images of MitoTracker Red staining. (D) Quantification of

MitoTracker Red staining measured as relative fluorescence intensity (RFI).

Figure 15: Levels of glycolytic intermediates are higher and glutamine flux is decreased after EMT. (A) Mass spectrometry analysis of glycolytic metabolites quantified as the difference between HMLERFOXC2 and HMLERGFP cells. (F) Mass spectrometry based flux analysis of

HMLERFOXC2 and HMLERGFP cells grown in the presence of 13C-glutamine.

6.2.2 GLS2 is inversely correlated with EMT in breast cancer

To further explore metabolism in cells that have undergone EMT, I analyzed a previously reported EMT-microarray for metabolic gene expression (Taube et al., 2010a) and identified altered expression of genes associated with nucleotide and aminosugar metabolism (Chapter 5).

Furthermore, I found an inverse expression pattern of two glutaminases (GLS2 and GLS) in cells post EMT. GLS was previously reported to be induced during EMT (Kung et al., 2011; Lee et al.,

2016a). Both enzymes catalyze the conversion of glutamine to glutamate; however, their expression patterns and regulation are distinct. GLS appears to contribute to glutamine addiction in primary tumors, but the role of GLS2 in tumor progression and metastasis is less well- understood.

We sought to profile the expression of GLS2 and GLS in human breast epithelial (E- cadherin-positive) and mesenchymal (vimentin-positive) tumors from publicly available data. The breast cancer patient data demonstrate significantly induced expression of GLS in mesenchymal breast tumors (p=1.526e-005) compared to epithelial tumors (Figure 16A). The expression of

GLS2 is significantly reduced (p=3.77e-011) in mesenchymal tumors compared to epithelial tumors (Figure 16B). The epithelial tumors have high levels of CDH1, which encodes E-cadherin, and reduced levels of CDH2 and VIM, which encode N-cadherin and vimentin, respectively, compared to the mesenchymal tumors (Figure 16C-E). That mesenchymal markers (VIM and

CDH2) have a negative correlation with GLS2 expression and positive correlation with GLS expression in breast tumor samples is similar to the pattern I observed in the HMLER cell models

(Chapter 5). Furthermore, we observed a negative correlation between GLS2 and GLS expression in breast cancer patients (Figure 16F).

I validated the observation that GLS2 is suppressed and GLS expression is increased relative to controls in the HMLER and non-transformed HMLE cells that over-express EMT-TFs and in metastatic triple negative breast cancer cell lines SUM159 and MDA-MB-231 by RT-PCR

(Figure 17A-F). Thus, the suppression of GLS2 is observed in multiple models where EMT is induced. The expression of RAS in HMLE cells did not induce the expression of FOXC2 or alter the expression of GLS2, therefore I concluded that GLS2 suppression is not transformation dependent; it is EMT dependent (Figure 17I-J). It has been previously reported that GLS expression is enhanced following TGFβ induction of MCF10A cells, and I demonstrated that this occurs in parallel with GLS2 repression (Figure 17 G-H). Collectively, these data demonstrate that gene expression changes occur during EMT in genes associated with metabolism and that

GLS2 expression is inversely correlated with EMT.

I next performed immunofluorescece staining of HMLERGFP, HMLERSNAIL, and

HMLERFOXC2 cellsfor GLS2. There was a significantly lower intensity of staining in the cells that had undergone EMT compared to the epithelial control (Figure 18A). The relative levels of protein expression pattern were similar to the relative levels of mRNA expression in these cells. The triple negative breast cancer cells SUM159 and MDA-MB-231 WHAT??? (Figure 18A). The breast cancer cells that had undergone EMT also expressed higher levels of FOXC2 compared to the MCF7 control (Figure 18B). The HMLER cells that had undergone EMT expressed higher levels of FOXC2, SNAIL, and ZEB1 than the control HMLERGFP cells (Figure 19). The expression of this EMT-inducing transcription factor shows that the EMT-program is active in these cells.

Figure 16: GLS2 is downregulated in mesenchymal breast cancers. Gene expression analysis of (A) GLS, (B) GLS2, (C) CDH1, (D) CDH2, and (E) VIM in breast cancer patients with epithelial (E) and mesenchymal (M) tumors. (F) Correlation analysis of GLS2 versus GLS expression in breast cancer patients.

Figure 17: GLS2 is downregulated in cells that have undergone EMT and in metastatic breast cancer cell lines. (A-B) qRT-PCR analysis of A) GLS2 and B) GLS in indicated HMLER cells. Data are relative to levels in epithelial control HMLERGFP. (C-D) qRT-PCR analysis of C)

GLS2 and D) GLS in metastatic breast cancer cells relative to non-metastatic MCF7 cells. (E-F) qRT-PCR analysis of E) GLS2 and F) GLS in indicated HMLE cells. Data are relative to levels in epithelial control HMLEGFP. (G-H) qRT-PCR analysis of G) GLS2 and H) GLS in MCF10A cells treated with TGF for indicated times. Data are relative to levels in untreated cells. (I-J) qRT-

PCR analysis of I) FOXC2 and J) GLS in indicated cells.

Figure 18: GLS2 protein is expressed in cells that have undergone EMT. (A)

Immunofluorescence staining of indicated cells with human GLS2 antibody (green); nuclei were stained with DAPI (blue). (B) Co-immunofluorescence staining with GLS2 and FOXC2 antibodies in MCF7 and Sum159 cells.

Figure 19: EMT marker expression in HMLER cells. (A) ZEB1, FOXC2, and actin protein expression in indicated cells as demonstrated by western blot. (B) Snail and tubulin protein expression in indicated cells as demonstrated by western blot.

6.2.3 Suppression of GLS2 is dependent on FOXC2

I next sought to determine if inhibiting EMT results in a reversion of the glutamine independence that we observed in the HMLER cells that express EMT-TFs. We previously reported that transcription factor FOXC2 is required and necessary for EMT (Mani et al., 2007a)

(Hollier et al., 2013a). FOXC2 also has a fundamental metabolic role in adipocyte metabolism as it directly promotes the expression of mitochondrial genes (Cederberg et al., 2001; Lidell et al.,

2011) so I hypothesized that FOXC2 regulates metabolic gene expression during EMT. We have demonstrated that inhibition of FOXC2 activity in EMT-TF-expressing cells with a chemical inhibitor or an shRNA blocks metastatic potential.

To understand the relationship between FOXC2 and GLS2, we analyzed expression of these genes in human breast cancer patients using a publicly available dataset. Levels of FOXC2 mRNA were negatively correlated with GLS2 mRNA levels, whereas FOXC2 and GLS are positively correlated (Figure 20 A-B). For further validation of EMT status and GLS2 expression in the patient samples, we performed a correlation analysis of VIM versus GLS2 and of CDH1 versus FOXC2. As expected, the mesenchymal marker VIM and GLS2 are negatively correlated as are mesenchymal marker CDH1 and FOXC2 in the patient samples (Figure 20C-D).

To test if the expression of GLS2 is restored by inhibiting EMT, I measured the relative mRNA transcript levels of GLS2 in cells that express shFOXC20. In HMLER and HMLE lines that express the shRNA, GLS2 expression was significantly enhanced reative to the cells that do not express the shRNA; however, GLS expression did not differ (Figure 21A-D). To determine if

GLS2 protein and mRNA have similar expression patterns, I performed immunofluorescent staining of HMLERSNAIL cells that do or do not express shFOXC2 with human GLS2 antibody. I observed that cells depleted of FOXC2 have higher intensity of GLS2 than the cells that express

FOXC2 (Figure 21G).

I also observed that GLS2 is suppressed in HMLER cells that express ZEB1 (Figure 17A and 22B). The HMLERZEB1 cells have high levels of endogenous FOXC2 (Figure 22E). We previously reported that FOXC2 regulates the expression of ZEB1 by directly binding to its promoter (Werden et al., 2016). ZEB1 is generally considered to be a transcriptional repressor.

Therefore I hypothesized that ZEB1 acts downstream of FOXC2 to suppress GLS2 expression.

To test this hypothesis, I generated HMLESNAIL and SUM159 cells that express shRNA targeting

ZEB1. After confirming by RT-PCR that ZEB1 levels are reduced in cells that express shZEB1

(Figure 22H and J), I measured mRNA transcript levels of GLS2, GLS, and FOXC2. The inhibition of ZEB1 expression resulted in a modest enhancement of GLS2 expression and no change in

GLS or FOXC2 levels (Figure 22???). It is unclear if this means that ZEB1 represses GLS2 as expression of shZEB1 only resulted in 2.5 fold increase in GLS2 in HMLESNAIL cells, whereas expression of shFOXC2 enhanced GLS2 expression nearly 25 fold. Cells that express shZEB1 express FOXC2, and it is possible that FOXC2 may exert an effect on GLS2 expression not mediated by ZEB1.

To obtain a clearer understanding of a possible suppressive role of ZEB1 on GLS2, I over-expressed ZEB1 in HMLESNAIL, HMLERSNAIL, and SUM159 cells that express shFOXC2.

Surprisingly, ZEB1 expression in shFOXC2-expressing cells did not repress GLS2 in any of the models that I generated (Figure 23). Despite ZEB1 expression, GLS2 expression did not change in the absence of FOXC2. Taking these data collectively, I conclude that FOXC2 is the transcription factor that exerts the predominant effect on GLS2 repression during EMT.

I sought to determine if restoring GLS2 expression by inhibiting FOXC2 expression would enhance mitochondrial metabolism. To determine if expression of shFOXC2 restores mitochondrial activity, I performed MitoTracker Red staining of HMLERSNAIL cells with and without expression of shFOXC2. Depletion of FOXC2 resulted in an increase in mitochondrial membrane potential (Figure 24A-B). Furthermore, to determine if oxygen consumption was enhanced upon

FOXC2 repression, I performed a Seahorse MitoStress analysis after oligomycin, FCCP, 88 rotenone, and antimycin treatments over time. The HMLERSNAIL cells that express shFOXC2 have higher respiratory capacity than the HMLERSNAIL control cells (Figure 24C-D).

I wanted to determine if enhanced respiratory capacity in cells with shFOXC2 expression also led to an enhanced conversion of glutamine to glutamate. Therefore I measured the biochemical conversion of glutamine to glutamate in HMLERSNAIL cells that do and do not express shFOXC2. Indeed, enhanced mitochondrial activity was detected in cells that express shFOXC2

(Figure 24E). I measured intracellular GSH to determine if enhanced GLS2 expression and conversion of glutamine to glutamate had further downstream metabolic consequences. In accordance with enhanced respiratory capacity, the cells that expressed shFOXC2 had higher levels of GSH than control cells (Figure 24F). These data suggest that blocking EMT through inhibition of FOXC2 expression results in enhanced GLS2 expression and a metabolic shift toward mitochondrial metabolism.

To determine if loss of FOXC2 resulted in differential utilization of amino acids, I performed a Biolog unbiased phenotypic microarray assay on HMLERSNAIL cells that do and do not express shFOXC2. I measured the difference in metabolite concentration at 8 hours. Five metabolites, including L-glutamine, with levels decreased in cells that express EMT-TFs had higher levels in HMLERSNAIL cells that express shFOXC2 than in HMLERSNAIL cells indicating that inhibiting FOXC2 expression results in restoration of amino acid utilization typical of epithelial cells.

Figure 20: FOXC2 is negatively correlated with GLS2 in breast cancer patient samples.

Correlation analyses of (A) FOXC2 and GLS2 expression, (B) FOXC2 and GLS expression, (C)

CDH1 and FOXC2 expression, and (D) VIM and GLS2 expression in breast cancer samples.

Figure 21: Cells deficient in FOXC2 have higher GLS2 expression. (A-B) qRT-PCR analysis of A) GLS2 and B) GLS mRNA expression in HMLERSNAIL cells that do and do not express shFOXC2. (C-D) qRT-PCR analysis of C) GLS2 and D) GLS mRNA expression in HMLESNAIL cells that do and do not express shFOXC2. (E-F) qRT-PCR analysis of F) GLS2 and G) GLS mRNA expression in SUM159 cells that do and do not express shFOXC2. (G)

Immunofluorescent staining of GLS2 protein in HMLERSNAIL cells that do and do not express shFOXC2.

Figure 22: ZEB1 induces FOXC2-mediated repression of GLS2. qRT-PCR analysis of ZEB1

(A), GLS2 (B), and GLS (C) mRNA expression in HMLE-vector (V) and HMLE-ZEB1 (Z) cells. qRT-PCR analysis of ZEB1 (D), FOXC2 (E), and GLS (F) mRNA expression in HMLER-vector

(V) and HMLER-ZEB1 (Z) cells. (G) Phase images of HMLER-vector (V) and HMLER-ZEB1 cells. qRT-PCR analysis of ZEB1 (H), GLS2 (I), mRNA expression in HMLE-Snail-vector (V) and

HMLE-Snail-shZEB1 cells. qRT-PCR analysis of ZEB1 (J), FOXC2 (K), GLS (L), and GLS2 (M), mRNA expression in SUM159-vector (V) and SUM159-shZEB1 cells.

Figure 23: In the absence of FOXC2, ZEB1 does not repress GLS2. (A-C) qRT-PCR analysis of A) ZEB1, B) GLS, and C) GLS2 mRNA in SUM159 cells that express shFOXC2 and that do or do not over-express ZEB1. (D-E) qRT-PCR analysis of D) ZEB1, E) GLS, and F) GLS2 mRNA in HMLESNAIL cells that express shFOXC2 and that do or do not over-express ZEB1. (G) Phase image of HMLESNAIL cells that express shFOXC2 and that do not (left) and do (right) over-express

ZEB1. (H-J) qRT-PCR analysis of H) ZEB1, I) GLS, and J) GLS2 mRNA in HMLERSNAIL cells that express shFOXC2 and that do or do not over-express ZEB1.

Figure 24: Inhibition of EMT by expression of shFOXC2 enhances mitochondrial metabolism. (A) MitoTracker Red staining of HMLERSNAIL cells that do and do not express shFOXC2. (B) Quantification of MitoTracker Red staining of HMLERSNAIL cells that do and do not express shFOXC2 measured by relative fluorescence intensity (RFI) with ImageJ software. (C)

Seahorse analysis of OCR of HMLERSNAIL cells that do and do not express shFOXC2 over time with treatments with oligomycin (O), FCCP (F), and rotenone plus antimycin (R+A) indicated. (D-

G) Measurements of D) basal respiration, E) conversion of glutamine to glutamate, F) GSH, and

G) % relative viability with decreasing glucose of HMLERSNAIL cells that do and do not express shFOXC2. (H) Biolog phenotypic microarray measuring the difference in amino acid utilization between HMLERSNAIL cells that do and do not express shFOXC2.

6.2.4 p38MAPK-FOXC2 signaling suppresses GLS2 expression

Understanding how suppression of GLS2 is regulated during the EMT program may reveal therapeutic targets, Therefore I sought to characterize this mechanism. We have previously shown that EMT and metastasis occur through p38MAPK-FOXC2 signaling

(Paranjape et al., 2016a; Werden et al., 2016) in breast and prostate cancers. Direct targeting of

FOXC2 is not yet a viable option; however, we have shown that chemically targeting p38MAPK results in decreased FOXC2 protein stability, which abrogates EMT-inducing signaling and thus decreases metastatic potential. To test if GLS2 expression is regulated through the p38MAPK-

FOXC2 signaling axis, I treated HMLERFOXC2 cells with p38MAPK inhibitor SB203580 (20 µM) or with vehicle for 24 hours. Using RT-PCR analysis, I showed that GLS2 mRNA expression was significantly enhanced in the cells treated with SB203580 (Figure 25A). GLS mRNA expression was not changed by SB203580 treatment (Figure 25B) much like the expression pattern observed upon expression of shFOXC2 in these cells (Figure 21).

p38MAPK signaling induces phosphorylation of FOXC2 at serine 367. We have previously generated FOXC2 mutants that mimic constitutive phosphorylation (FOXC2-S367E) and that are not phosphorylatable (FOXC2-S367A). To test if GLS2 expression is dependent on the phosphorylation status of FOXC2, I quantified GLS2 mRNA in cells that express these mutants. GLS2 expression was suppressed in the S367E mutant cells but was expressed at similar levels in the FOXC2-S367E mutant cells as in the control cells (Figure 25C). GLS expression was significantly induced in the FOXC2-S367E mutant cells (12 fold) and in the

FOXC2-S367A mutant cells (5 fold) (Figure 25D). Interestingly, ZEB1 expression was 70-fold higher in the FOXC2-S367A mutant cells than in the control cells, which could be the reason why

GLS was expressed (Figure 25E). This suggests that a S367-phosphoylation independent mechanism may be operating.

To explore an alternative regulatory mechanism, I tested whether GLS2 expression could be suppressed by DNA methylation in our model systems. Previous reports have observed reduced GLS2 mRNA expression in liver and colon cancer due to promoter hypermethylation

(Zhang et al., 2013a). To test this in breast cancer and EMT-TF-expressing cells, I treated

SUM159 and HMLERFOXC2 cells with a chemical that results in DNA demethylation, 5-aza-

2’deoxycytidine (5d-AzaC). GLS2 mRNA expression was upregulated in a manner dependent on

5d-AzaC dose (Figure 25F-G). Interestingly, FOXC2 mRNA also became downregulated after

5d-AzaC treatment compared to vehicle control. Thus, the inverse relationship between FOXC2 and GLS2 that we observed in other models was seen in these experiments.

To determine if the FOXC2 mutants have metabolically different phenotypes reflective of p38MAPK-FOXC2 signaling, I measured mitochondrial activity in cells that expressed the mutants. Similar to cells that express wild-type FOXC2, the HMLERFOXC2 cells, those that express the S367E mutant have reduced mitochondrial membrane potential and reduced intracellular

GSH relative to HMLERGFP cells indicating that the p38MAPK-FOXC2 signaling maintained by

S367 phosphorylation directs this metabolic program (Figure 26A-B). I treated of HMLERSNAIL and HMLERFOXC2 cells with 20 µM SB203580 or vehicle for 24 hours and stained with MitoTracker

Red. Similar to the cells expressing the FOXC2-S367A mutant, chemical inhibition of p38MAPK-

FOXC2 signaling resulted in enhanced mitochondrial membrane potential (Figure 26D-F). These data demonstrated that GLS2 repression is mediated by a critical EMT-inducing signaling pathway.

Figure 25: Inhibition of p38MAPK-FOXC2 signaling enhances GLS2 expression. qRT-PCR of GLS2 (A) and GLS (B) mRNA expression in vehicle and SB203580 treated HMLER-FOXC2 cells. qRT-PCR of GLS2 (C) GLS (D), a ZEB1 (E) mRNA expression in HMLER-vector (V),

HMLER-FOXC2 (S367E), and HMLER-FOXC2 (S367A) mutants. (F) qRT-PCR of GLS2 and

FOXC2 (G) mRNA expression in SUM159 and HMLER-FOXC2 cells treated with 5 deoxy azacytidine (5d-AzaC) and vehicle control. 97

Figure 26: Inhibition of p38MAPK-FOXC2 signaling enhances GLS2 expression. (A)

MitoTracker Red and nuclear DAPI staining of HMLERGFP, HMLERFOXC2(S367E), and

HMLERFOXC2(S367A) cells. (B) Quantification of MitoTracker Red by relative fluorescence intensity

(RFI). (C) GSH quantification HMLERGFP, HMLERFOXC2(S367E), and HMLERFOXC2(S367A) cells.(D)

MitoTracker red staining of HMLERSNAIL and HMLERFOXC2 cells treated with SB203580. (E-F)

Quantification of MitoTracker Red by relative fluorescence intensity (RFI) of E) HMLERSNAIL and

F) HMLERFOXC2 cells cells treated with SB203580.

6.2.5 GLS2 re-expression in GLS2-negative breast cancer cell line restores mitochondrial metabolism

To understand how GLS2 suppression metabolically supports EMT I over-expressed

GLS2 in the metastatic breast cancer cell line SUM159 and in HMLESNAIL, HMLERSNAIL, and

HMLERFOXC2 cells. These cell lines do not express GLS2 endogenously. Over-expressing GLS2 in these cell lines did not change the expression of GLS, FOXC2, or ZEB1 in any of the models

(Figure 27 B, C, E, G, H, and J). Immunofluorescent staining for human GLS2 in SUM159 cells over-expressing GLS2 confirmed enhanced protein expression (Figure 27D).

To test if loss of GLS2 in epithelial cells was sufficient to promote EMT, I generated

HMLE, HMLER, and MCF7 cells that express shGLS2 (Figure 28A). I performed a mammosphere assay to test CSC potential these cells. Reduced GLS2 expression did not promote mammosphere formation in these cell models (Figure 28C) and did not result in a CSC- like surface marker expression (Figure 28D).

To determine if GLS2 over-expression in cell lines with reduced mitochondrial metabolism induced a gain in metabolic effects, I measured spare respiratory capacity (Figure 29 A), basal respiration (Figure 29 B) and ATP production (Figure 29 C), mitochondrial membrane potential

(Figure 29 E), conversion of glutamine to glutamate (Figure 29 F) and GSH (Figure 29 G). Indeed

GLS2 over-expression enhanced mitochondrial metabolism. I also tested if GLS2 over- expression could impact viability in decreasing concentrations of glucose but only observed a modest reduction (Figure 29 D). I further hypothesized that this shift towards mitochondrial metabolism could decrease the glycolytic metabolism described previously. To test this I measured lactate from the culture media. The GLS2 over-expressing SUM159 cells had significantly lower levels of secreted lactate which implies a metabolic shift away from glycolytic metabolism (Figure 29 H). Functionally, GLS2 has been shown to promote the production of intermediates that neutralize reactive oxygen species (ROS). I measured intracellular ROS by

99 flow cytometry to conclude if over-expression of GLS2 also resulted in neutralization of ROS. I observed that GLS2 over-expression resulted in a decreased formation of ROS (Figure 29 I).

Taken together, this data demonstrates that expressing GLS2 in EMT-induced cells enhances mitochondrial metabolism.

My findings suggest that there is a shift in mitochondrial metabolism toward enhanced glutaminolysis. Therefore I hypothesized that this could lead to changes in amino acid utilization.

To determine if GLS2 over-expressing cells exhibit differences in amino acid utilization, I performed a phenotypic microarray (Biolog) in HMLER-FOXC2-vector and HMLER-FOXC2-

GLS2 cells (Figure 30). I measured the difference of initial rates in each well at 8 hours as previously described in figure 1. Only two of the 5 metabolites (designated by black arrows) that previously exhibited metabolic loss (L-glutamine and L-histidine) had restored utilization in GLS2 over-expressing cells. This means that while FOXC2 knockdown exerts a significant metabolic gain, GLS2 over-expression alone can only rescue two of the amino acids. This implies that its function is specific to glutamine metabolism.

Triple negative breast cancer cells are enriched with cancer stem cells that have properties that allow them to metastasize. The mammosphere assay allows the quantification of cancer stem cell potential as single cells grow in a low-attachment matrix which will give rise to mammospheres. To test if metastatic potential is altered with GLS2 expression, I performed a mammosphere assay and quantified spheres greater than 80 micrometers after 14 days in low attachment conditions. The cells with GLS2 over-expression had significantly decreased the number of mammospheres indicating a decrease in stemness properties (Figure 31A).

Furthermore, I confirmed that the decrease in sphere formation is not due to proliferation differences because I did not observe differences in two-dimensional growth between the vector and GLS2 expressing cells in vitro. Moreover, I provide evidence that a shift in metabolic program toward mitochondrial metabolism reduced metastatic potential despite no changes in FOXC2 expression and ZEB1. 100

I analyzed breast cancer patient data from a publicly available database in KMplotter.com and found that improved relapse-free survival (RFS) is associated with higher GLS2 expression with a hazard ratio (HR) of 0.69 in all breast cancers (Figure 31) and the highly aggressive mesenchymal breast cancer subtype (Figure 31 C). Moreover, higher GLS2 expression is associated with better overall survival in the ER-negative subtype (Figure 31 D). GLS expression does not stratify patients (Figure 31 E). This data is in line with our model that demonstrates that breast tumors with high EMT features and low GLS2 expression are characteristically more aggressive than those tumors that are more differentiated and have higher expression of GLS2.

101

Figure 27: Over-expression of GLS2 does not alter FOXC2 expression. qRT-PCR analysis of GLS2 (A), GLS (B), FOXC2 (C), and ZEB1 (E) mRNA in SUM159-vector (V) and SUM159-

GLS2 cells. (D) Immunofluorescence staining of GLS2 (green) and nuclear DAPI (blue) in

SUM159-vector (V) and SUM159-GLS2 cells. qRT-PCR analysis of GLS2 (F), FOXC2 (G), GLS

(H) mRNA expression in HMLE-Snail-vector (V) and HMLE-Snail-GLS2 cells. qRT-PCR analysis of GLS2 (I) and GLS (J) mRNA expression in HMLER-Snail-vector (V) and HMLER-Snail-GLS2 cells.

102

Figure 28: Loss of GLS2 does not result in a CSC phenotype. (A-B) qRT-PCR analysis of A)

GLS2 and B) GLS in HMLEGFP cells and HMLEGFP cells that express shGLS2. (C) Quantification of mammospheres (>80 µm) after 12 days in culture. (D) FACS analysis of CD24, CD44, GD2, and CD140b cell surface marker expression in HMLEGFP cells and HMLEGFP cells that express shGLS2.

103

Figure 29: Over-expression of GLS2 enhances mitochondrial metabolism. (A) Seahorse analysis of OCR, (B) basal respiration, (C) ATP production, (D) glucose IC50, (E) MitoTracker

Red and nuclear DAPI staining, (F) conversion of glutamine to glutamate, (G) GSH, (H) L-lactate, and (I) ROS in SUM159 cells and SUM159 cells that overexpress GLS2.

104

Figure 30: Amino Acid utilization. (A) Biolog phenotypic microarray measuring the difference of amino acid consumption in GLS2-expressing HMLER-FOXC2 cells compared to HMLER-

FOXC2 vector cells.

105

Figure 31: Higher GLS2 expression is correlates with improved patient outcomes. (A)

Quantification of mammospheres >80m from in SUM159-vector (V) and SUM159-GLS2 cells.

RFS of GLS2 expression in all breast cancer patients (B), RFS of GLS2 expression in the mesenchymal breast cancer subtype (C), overall survival of GLS2 expression in ER-negative breast cancer patients. (E) RFS of GLS expression in all breast cancer patients.

106

6.3 Discussion

In this chapter, I have employed genetic and metabolomics approaches to characterize the bioenergetics of EMT-induced cells. I report that EMT-induced cells exhibit glutamine independence even in low-glucose conditions. Induction of EMT results in decreased mitochondrial metabolism resulting in decreased glutamine utilization, oxygen consumption, and reduced glutathione. I sought to understand if changes in metabolism-related genes upon EMT induction could contribute to glutamine independence. I observed an inverse correlation with

EMT and Glutaminase 2 (GLS2). GLS2 catalyzes the conversion of glutamine to glutamate therefore I hypothesized that its expression could be regulated in EMT to maintain glutamine independence. I further demonstrate that transcriptional suppression of GLS2 occurs through p38MAPK-FOXC2 signaling. Targeting p38MAPK-FOXC2 signaling with a chemical inhibitor of p38MAPK enhances GLS2 expression and results in increased glutamine utilization, glutaminolysis, and oxygen consumption. Moreover, we demonstrate that GLS2 over-expression shifts the glycolytic metabolic program and reduces cancer stem cell properties. This is in line with breast cancer patient data indicating higher GLS2 is associated with better overall survival.

107

Chapter 7: Inhibition of p38MAPK-FOXC2 signaling sensitizes breast cancer stem cells

to standard of care chemotherapeutics

108

Chapter 7: Inhibition of p38MAPK-FOXC2 signaling sensitizes breast cancer stem cells

to standard of care chemotherapeutics

7.1 Introduction and Rationale

It is of urgent clinical need to develop therapeutic strategies that not only decrease the tumor bulk but also inhibit cancer stem cells to prevent metastasis. Metastasis is the leading cause for cancer mortality as is it incurable and survival outcomes have not significantly improved. In the

United States, 5-year survival rates for metastatic breast cancer patients did not significantly improve from 1992-1999 (23.3%) to 2005-2011(25.9%) (Cardoso et al., 2016). Currently, the 5- year survival rate for women diagnosed with localized breast cancer is 98.9% compared to 26.9% in women diagnosed with metastatic disease (Figure 32). To develop therapeutic strategies that have the potential to prevent or target metastatic cells, it is necessary to provide refined preclinical models that can provide evidence of improved outcomes.

Several compounds that target EMT-related pathways are currently in Phase 1, 2, and 3 clinical trials in combination with chemotherapeutic drugs. The use of these compounds in combination with chemotherapeutics is to reduce tumor burden and elicit depletion and/or inhibit the generation of CSC or reversal of drug resistance. However, none of these compounds currently in clinical trials target the p38MAPK-FOXC2 signaling axis that directs EMT (Marcucci,

Stassi et al. 2016). We identified that p38MAPK stabilizes FOXC2 protein enabling it to exert its functions that lead to the EMT/CSC phenotype. In efforts to target cells with EMT/CSC properties, we propose that targeting those cells that express high levels of FOXC2 would be ideal because the cells with highly aggressive features could be eliminated. However, therapeutically targeting a transcription factor like FOXC2 is currently not feasible so we reasoned that by targeting p38MAPK (MAPK14) signaling we can destabilize FOXC2 protein and attenuate its activity.

Although many signaling pathways converge on p38MAPK we observed that inhibiting the p38MAPK-FOXC2 axis was sufficient to inhibit CSC properties and metastasis in mouse models

109 of metastatic breast cancer. This is particularly important in CSC-enriched triple negative breast cancer that have no targeted therapeutic strategies available (Figure 33).

Treating patients with chemotherapy to reduce the primary tumor is not always sufficient to eliminate the CSC population. 30% of breast cancer patients treated with chemotherapy will succumb to tumor relapse and chemotherapy resistance offering them little therapeutic alternatives. Targeting both the CSC population and the tumor bulk is key to eliminating the risk of relapse, chemo-resistance, and metastasis. Therefore, we hypothesize that employing a p38MAPK inhibitor in combination with a standard of care chemotherapeutic like carboplatin will eliminate the CSC population and the tumor bulk. The proposed combination strategy is a novel approach as there are no approved therapies that directly target p38MAPK-FOXC2 signaling.

To model the proposed combination, we chose the metastatic murine breast cancer cell line

4T1 to be implanted into the mammary gland of immune competent Balb/c mice. A model where the immune system is intact is more physiologically relevant than using human cells in immune- compromised mice. Emerging studies demonstrate how the immune system influences tumor progression and metastasis. The 4T1 cell line was derived from a subpopulation of cells arising from the spontaneous mammary tumor of a BALB/cfC3H female mouse (Dexter et al., 1978).

4T1 is reported to be thioguanine-resistant with the ability to spontaneously metastasize to the lung and liver (Aslakson and Miller, 1992). Notably, gene expression profiling of the four murine mammary carcinoma cell lines with differing metastatic abilities identified FOXC2 to be significantly up-regulated in 4T1, the only cell line out of the four analyzed that is capable of forming primary tumors, release tumor cells into circulation, and generate micro and macrometastasis (Mani et al., 2007b). Furthermore, this study demonstrated that FOXC2 is required for metastatic ability in 4T1. For these aforementioned lines of evidence we chose 4T1 cells as the cell lines in which we will test our combination strategy.

110

To target p38MAPK in vivo, we will use SB203580, a pyridinyl imidazole that competitively binds to the ATP pocket of p38MAPK. This compound is currently not approved for use in human patients but has shown efficacy in mice (Mader et al., 2008). To target the tumor bulk, we will use carboplatin in combination with SB203580. Carboplatin enters the cell mainly by passive diffusion. The platinum atom of carboplatin forms covalent bonds to the N7 positions of purine bases in DNA forming 1,2 or 1,3 intrastrand or interstrand crosslinks. The linkage to the DNA occurs where the cyclobutane-1,1-dicarboxylate (CBDCA) are in the compound. Binding of the compound with DNA distorts the DNA structure by bending the duplex. This distortion causes replication and transcription inhibition, cell cycle arrest and ultimately results in cell death by apoptosis (Wang and Lippard, 2005).

Clinical trials have shown that the best patient complete response (pCR) occurs in TNBC patients who are chemotherapy naïve, and treated in neoadjuvant setting with carboplatin (von

Minckwitz and Martin, 2012). However, carboplatin has not shown significant efficacy when used as a monotherapy. Currently, carboplatin is used in combination with other chemotherapeutics including taxanes, and anthracyclines (Martin, 2001) (von Minckwitz and Martin, 2012).

Increasing interest in the use of carboplatin in patients with TNBC is emerging because a significant portion of these patients harbor BRCA1 mutations. This result in cells with faulty DNA repair mechanisms and are most responsive to carboplatin. The effects of carboplatin in mice injected with 4T1 in the posterior flank have been previously tested (de Souza et al., 2014; Souza et al., 2013). In these studies, carboplatin treated mice exhibited significantly decreased tumor volume, metastasis, and histological analysis determined decrease in proliferation, blood vessel formation and increased apoptosis compared to vehicle control.

In this chapter, I describe the effects of SB203580 and carboplatin on 4T1 cells, measuring specific criteria such as determining effective concentrations as single agents and in combination that will target p38MAPK-FOXC2 signaling and CSC properties in vitro, and tumor burden, and metastasis in vivo. 111

Figure 32: Survival outcomes in women diagnosed with distant metastasis compared to localized or regional breast cancer. (A) 5-year relative survival rates from 2007-2013 in women

(all races) diagnosed with localized, regional, or distant breast cancer at time of diagnosis. (B)

Percent surviving women by time since diagnosis in women diagnosed with localized, regional, or distant breast cancer from 2000-2013. Graphs adapted from seer.cancer.gov

112

Figure 33: Graphical model of the proposed advantage of a combination treatment that targets CSCs and bulk tumor. Conventional chemotherapy targets the tumor bulk but CSCs are not eliminated and these residual CSCs result in relapse and metastasis. Using only CSC targeted therapy would only target a subpopulation of cells leaving behind the tumor bulk without effectively diminishing it. The ideal strategy is to target both CSCs and differentiated tumor cells to prevent relapse by eliminating the cells that can self-renew and decrease the tumor bulk simultaneously.

113

7.2 Results

7.2.1 p38MAPK inhibitor SB203580 inhibits CSC properties in metastatic 4T1 murine

breast cancer cell line

We have used SB203580 in human cells lines extensively in vitro and in vivo and have determined an effective dose that decreases FOXC2 protein and EMT/CSC properties. To determine if FOXC2 expressing 4T1 cells respond to SB203580-mediated FOXC2 inhibition, I treated 4T1 cells with vehicle, 20 uM, 40 uM, and 80 uM of SB203580 for 24 hours. I performed a western blot analysis to analyze the effects of FOXC2 protein with SB203580 treatment and the status of p-ATF/total ATF which are directly regulated by p38MAPK (Figure 34 A). The non- metastatic murine cell line 67NR does not express FOXC2, or significant levels of ATF. During the treatment, total ATF levels remained the same while p-ATF decreased in a dose-dependent manner. FOXC2 protein levels also decreased similarly in a dose-dependent manner suggesting that the inhibitory effect of SB203580 on 4T1 affects the signaling mediated by p38MAPK in parallel to FOXC2 inhibition (Figure 34 A). The morphology of 4T1 cells is also affected with increasing concentrations of SB203580 (Figure 34 B). The 4T1 cells appear to have tighter cell islands than vehicle control cells. To further characterize the effects of SB203580 on 4T1 cells, I determined the IC50 values after 72 hours of treatment. The IC50 of 4T1 cells treated with

SB203580 is 62.42 uM therefore it will serve as a guide to understand which concentrations affect CSC/EMT properties (Figure 34 C). To determine if pre-treatment with SB203580 in 2D cultures inhibits CSC properties, I treated 4T1 cells with 20 uM SB203580 (a concentration below

IC50 that does not affect proliferation) for 72 hours prior to plating them for sphere formation assay. I hypothesized that pre-treatment would eliminate the CSC potential in the cells and consequently inhibit sphere-formation. I divided the pre-treated cells into two groups; one group received continuous SB203580 throughout the course of the sphere assay and the other group did not continue to receive drug treatment. After 14 days, the spheres greater than 80 um were counted (Figure 34 D). Indeed, pre-treated 4T1 cells had a reduced capacity to form spheres in

114 a dose-dependent manner suggesting that SB203580 effectively eliminates CSC potential in highly CSC-enriched 4T1 cells. To determine if SB203580 can reduce sphere formation in treatment naïve cells, I performed a sphere assay with increasing concentrations of SB203580 for 14 days. Sphere formation was significantly decreased at a higher concentration (80 uM) than the pre-treated cells (20-40 uM) however it is still the concentration at which I determined p38MAPK-FOXC2 is attenuated (Figure 34 E-F). Collectively these data demonstrate that

FOXC2- expressing 4T1 murine cells respond to p38MAPK inhibitor SB203580 and decreasing

FOXC2 levels correspond with decreasing CSC-potential.

115

Figure 34: Inhibiting p38MAPK with SB203580 reduces FOXC2 and sphere formation in

4T1 cells. (A) Western blot analysis of FOXC2, p-ATF, total ATF, and Actin protein expression in 67NR cells and 4T1-RFP cells treated with vehicle, 20, 40, and 80 m of SB203580. (B)

Phase images of 4T1-RFP cells treated with vehicle, 20, 40, and 80 m of SB203580. (C) IC50 of SB203580 in 4T1-RFP cells. (D) Quantification of mammospheres >80 m at vehicle, 20, 40, and 160 m of SB203580. (E) Phase images of mammospheres treated with vehicle, 20, 40,

80, 120, and 160 m of SB203580. (F) Quantification of E

116

7.2.2 The combination of SB203580 with carboplatin decreases the IC50 of 4T1 cells compared to using a single agent

We sought to use a chemotherapeutic drug that is clinically relevant for highly metastatic triple negative breast cancer so we tested carboplatin and paclitaxel in 4T1 cells (Fountzilas et al., 1998). I determined IC50 values against 4T1 cells. The IC50 of carboplatin is 50.93 uM (Figure

35 A) and paclitaxel is 9.61 uM (Figure 35 B). These concentrations follow a similar pattern to the IC50 concentrations determined in highly metastatic breast cancer MDA-MB-231 to the parent platinum compound cisplatin (21.8 uM) and paclitaxel (0.13 uM). Interestingly, non-metastatic epithelial breast cancer cell line MCF7 is highly resistant to cisplatin (>250 uM) and paclitaxel

(>100 uM) (Fountzilas et al., 1998). This offers evidence that using paclitaxel or a platinum compound like carboplatin is an appropriate compound in metastatic TNBC. The goal is to identify an appropriate combination strategy that can decrease the viability of 4T1 cells more effectively than a monotherapy therefore I performed combination treatments with SB203580 and carboplatin or paclitaxel. Strikingly, the combination of carboplatin and SB203580 decreased the

IC50 of carboplatin by approximately 100-fold (IC50=0.065 uM) 9Figure 35 C) at the SB203580 concentration in which FOXC2 protein is decreased (80 µM) (Figure 34 A). The IC50 of 4T1 was also decreased nearly 100-fold (IC50=0.016) using paclitaxel in combination with SB203580 at

80 uM (Figure 35 D). Moving forward, we chose to use carboplatin in our combination strategy model.

117

Figure 35: Employing Chemotherapeutics in combination with SB203580 in vitro (A) IC50 of carboplatin in 4T1-RFP cells (B) IC50 of paclitaxel in 4T1-RFP (C) IC50 of 4T1-RFP cells treated with carboplatin and SB23580 (D) IC50 of 4T1-RFP cells treated with paclitaxel and SB23580

118

7.2.3 Testing the combination of SB203580 and carboplatin is a novel combination

strategy to inhibit CSCs and reduce bulk tumor

The goal of this study is to determine if combination strategies that target differentiated cells and cancer stem cells decrease tumor burden and lung metastasis. To determine the efficacy of a novel combination strategy using SB203580 and carboplatin, we used four treatment groups with ten female mice per group. We injected 40 female mice with RFP-luciferase labeled

4T1 cells in the mammary gland above the 5th nipple on the left side. We waited 7 days for the injected 4T1-RFP cells to form small palpable tumors before beginning treatments. After 7 days we confirmed that the mice had detectable tumors by luciferin based imaging and we randomized ten mice into each group. Group 1; vehicle control group, Group 2; receiving carboplatin (100 mg/kg) by IP injection once a week for four cycles, Group 3; receiving SB203580 (3.8 mg/kg) subcutaneously for 5 consecutive days for 4 cycles, and Group 4; the combination group receiving carboplatin (100 mg/kg) by IP once weekly for four cycles, and SB203580 (3.8 mg/kg) subcutaneously for 5 consecutive days for 4 cycles (Figure 36). We imaged the mice weekly to monitor tumor burden.

After 4 weeks of treatment cycles the experiment was terminated. Tumor reduction by nearly 50% was evident in the group receiving the combination and carboplatin alone compared to SB203580 alone and vehicle control as we predicted (Figure 37 A). In the combination group,

100% of the mice survived whereas only 90 % survived carboplatin treatment, and 80% survived in SB203580, and vehicle control groups (Figure 37 B). The mice that died succumbed to tumor burden the week prior to termination. This suggests that the combination treatment contributed to improved survival outcomes. Furthermore, the tumor volume was significantly decreased in the two groups receiving carboplatin treatment. There was a modest reduction of tumor volume in the group receiving SB203580 treatment alone as predicted. These results indicate that carboplatin effectively contributed to reducing tumor bulk (Figure 37 C). To determine if the treatments inhibited the ability of the highly metastatic 4T1 cells to complete the metastatic 119 cascade and successfully colonize the lungs, I counted the tumor nodules in the dissected lungs.

Large tumor nodules covered the lungs of the vehicle control mice as expected. The group with carboplatin treatment reduced visible lung metastasis by 60% compared to vehicle control.

Although the group that received SB203580 treatment only decreased the lung metastasis by

50% (although not statistically significantly), the metastatic nodules were much smaller in size upon visual examination (data not shown). The group that received the combination treatment also had a significant reduction in lung metastatic nodules (80% reduction) (Figure 37 D).

Compared to carboplatin or SB203580 treatment alone, the combination treatment reduced lung metastasis by 66% and 41%, respectively. This data collectively suggests that the combination treatment is more potent in reducing not only tumor burden but also lung metastasis and results in improved survival.

120

Figure 36: Schematic of in vivo treatment groups.

121

Figure 37: The combination of SB203580 with carboplatin reduces tumor volume and lung metastasis (A) Representative images of treatment groups (B) Quantification of % survival per group (C) Quantification of tumor volume (D) Quantification of lung metastasis nodules

122

7.1 Discussion

In this chapter I have described a novel combination treatment to target cancer stem cells within a tumor and also reduce tumor bulk. It is necessary to target both types of cell population to efficiently reduce the risk of metastasis and tumor recurrence. In recent years, the characterization of tumors has brought to light that the degree of heterogeneity within a tumor can impact outcomes in patients. Those tumors that are highly heterogeneous have a higher risk of drug resistance and relapse. A subpopulation of cells that greatly contribute to heterogeneity are CSCs because they have the potential to self-renew and also give rise to differentiated cancer cells maintaining the tumor cell population. Before proposing a novel treatment in patients, a preclinical study in mice is an effective method to test the efficacy of the combination. This way it is possible to set defined criteria that can provide evidence of the effectiveness and specificity of the proposed treatment.

123

Ch. 8 Discussion and Future Directions

124

Ch. 8 Discussion and Future Directions

8.1 Metabolic Alterations associated with EMT

In chapter 4, I described a metabolite signature driven by the induction of three different EMT-

TFs, Snail, Twist, and Goosecoid. Each EMT-TF generated its own distinct profile of altered metabolites however, for the EMS we chose abundant metabolites common across the three models compared to the vector control. The EMS is composed of amino acids glutamine, glutamate, beta-alanine, and a di-peptide glycylleucine (Figure 6). With the exception of glycylleucine, the three amino acids of the EMS are at the center of multiple pathways important in cancer such as pyrimidine metabolism, TCA, and glutaminolysis. We wanted to determine if this EMS derived from a cell line model could have prognostic value for breast cancer patient samples. Notably, when we analyzed clinical parameters such as ER status, breast cancer subtype, and survival, we found significant differences between patient samples that were grouped with a high EMS score (Group 1) and a low EMS score (Group 2) (Figure 7). We determined that 82% of Group 1 consists of ER-negative tumors and only 44% of Group 2 consists of ER-negative tumors (p=0.001) (Figure 8). We further determined the distribution of the PAM50 subtype and 82% of Group 1 consisted of basal-like and HER2-over-expressing subtypes while only making up 22% of Group 2. Additionally, we found a significant difference in survival yielding a hazard ratio (HR) of 2.3, (95% confidence interval (CI): 1.31-4.20, p=0.03

9Figure 8). Collectively, these data demonstrate that indeed there are metabolic alterations occurring in parallel to the onset of EMT and that these alterations can be used for clinical prognosis. This was the first report of an analysis that employed EMT-induced cell line models profiled by mass spectrometry and applied the observed signature to disease prognostication.

This report raises substantial questions about the nature of the metabolic alterations we observed and their functional relevance for the metabolic requirements of EMT. While other studies have employed mRNA transcripts levels for prognostic indicators, we chose to employ

125 metabolite abundances as a prognostic indicators. Measuring metabolites with metabolomics approaches provides a direct readout of biochemical activity and phenotype (Fiehn, 2002).

Metabolites are connected to large networks of biochemical reactions and using metabolomics can potentially identify alterations in entire networks. We identified both glutamine and glutamate in the EMS and both belong to a network that feed into the TCA cycle (Altman et al., 2016b). We explored the functional relevance of these alterations and this study served as the foundation for the work elaborated in depth in Chapter 6.

Future work must be done to extensively validate the EMS with a larger patient sample data set. A larger sample set would offer more statistical significance and it could include larger groups of the breast cancer subtypes with varying degrees of EMT status. Notably, the breast cancers of the luminal androgen receptor (LAR) subtype based on the Pietenpol subtyping are heavily enriched in metabolic pathways including TCA cycle, glutathione, tyrosine, steroid, and fatty acid metabolism (Lehmann et al., 2011). It would be interesting to understand how the EMS contributes to this subtype heavily enriched by metabolic pathways or alternatively, identify its own unique LAR metabolite signature. Interestingly, a previous study using metabolomics in breast cancer patient samples and cell lines identified onco-metabolite 2-hydroxyglutarate (2HG) to be highly accumulated in basal-like tumors (Terunuma et al., 2014). Of clinical importance, a subgroup with the highest 2HG levels in this study consisted of ER-negative tumors predominantly from African American patients was identified from this analysis. Based on this and multiple other studies, circulating 2HG is considered a potential metabolite-based biomarker

(Borger et al., 2014). Furthermore, it is of urgent clinical need to address the racial disparities that affect African America women diagnosed with metastatic disease. The 5-year survival rate for African American women diagnosed with metastatic breast cancer is 17% compared to 26.9% in European American women (Cardoso et al., 2016). Studies like Borger et al that analyze data sets inclusive of different ethnic groups for the development of biomarkers are necessary not

126 only to improve risk assessment and monitoring but also understand underlying biological differences that may contribute to the racial disparities in outcomes.

Furthermore, it would be valuable to measure the EMS against breast cancer patient samples that are treatment naïve, during treatment, and samples of tumors post treatment that are both sensitive and resistant to the treatment. One could hypothesize that as a tumor size decreases or increases during chemotherapy, the population of cells with EMT/CSC features also changes and may be reflected by their EMS status. For further validation, it would be valuable to compare the EMS with other cancers generally associated with EMT features and aggressiveness such as glio/neuro-blastomas, osteosarcomas, and malignant melanomas and cancers with more epithelial features, such as colorectal or gastric cancer (Barretina et al., 2012;

Garnett et al., 2012; Tan et al., 2014). This analysis would determine if our EMS, which we established from EMT-induced human mammary epithelial cells (Mani et al., 2008), can also offer prognostic value for other cancer types, particularly in highly aggressive cancers enriched with

EMT features. Collectively, these data could provide substantial evidence for the development of metabolite based biomarkers. To this end, the EMS would need to be validated in serum and/or urine samples of pre and post treatment breast cancer patients compared with healthy females with no malignancies. The development of an EMT-specific metabolite based biomarker that is non-invasive could be an effective means to monitor patients during treatments for therapy resistance (Hinestrosa et al., 2007). Non-invasive serum and/or blood biomarker surveillance strategies that reflect a patient’s response to treatment could also serve as cost effective alternatives to imaging. Metastatic breast cancer incurs a significant economic burden not only to patients and their caregivers, but also to the healthcare systems in both private and public systems (Cardoso et al., 2016). This is particularly important in low income countries where nearly 50-80% patients are initially diagnosed with advanced stage breast cancer compared to women from high-income countries where only 8% are initially diagnosed with advanced breast cancer (Dowsett et al., 2015; O'Shaughnessy, 2005).

127

8.2 Metabolic genes and metabolic reprogramming regulated by FOXC2

In chapter 5, I report that genes associated with metabolism, particularly genes that encode enzymes involved in nucleotide and amino metabolism, spermidine biosynthesis, triglyceride synthesis, and the generation of nitric oxide are de-regulated during EMT (Figure 10).

The most striking downregulated gene, Xanthine dehydrogenase (XDH), operates in the purine nucleotide de novo synthesis pathway. We sought to characterize the functional role it may play in EMT. We identified the metabolite upstream of XDH, hypoxanthine to be significantly abundant and the product of its oxidation, xanthine, significantly lower. This accumulation is reversed when we inhibited EMT by knockdown of FOXC2. In FOXC2 knockdown cells, XDH mRNA expression is upregulated nearly 100-fold (Figure 11 and 12). We reasoned that XDH expression and metabolite products of this pathway could affect the cell cycle of EMT induced cells so we exposed EMT-induced cells to a cocktail of nucleotides. The nucleotide supplements prolonged the G1-phase in the EMT-induced cells. How this enhancement of the G1-phase affects EMT remains to be discovered. It is interesting to speculate that there is an intentional regulatory mechanism in EMT where FOXC2 directs XDH expression and thus downstream metabolites are affected. Based on breast cancer patient data from KMplotter.com, I observed that high XDH expression is generally associated with a longer relapse-free survival (RFS) (Figure 12).

Interestingly, the most striking effect of an increased RFS with high XDH expression is in the mesenchymal-stem like subtype, which is associated with features of EMT/CSCs. This finding raises substantial questions of the general regulation of nucleotide metabolism in EMT. While we determined that inhibition of EMT re-expresses XDH and enhances the oxidation of hypoxanthine to xanthine, it has yet to determine what functional role this enzyme’s activity contributes to the maintenance of EMT/CSC properties. Future work must be done to determine if XDH over- expression suppresses EMT/CSC properties or if xanthine metabolite supplementation is sufficient. This would provide additional information about how critical the metabolic product is for the bioenergetics of EMT or if XDH exerts an alternative function. Validation of our findings

128 from the metabolomics data in breast cancer patient samples including all subtypes would be useful in understanding if hypoxanthine or xanthine abundance can serve as a prognostic biomarker for CSC features and survival outcomes.

The role XDH plays in tumorigenesis and metastasis is not well understood but a few reports provide insight into its function. Similar to our observations, one study described that lower XDH expression is correlated with higher tumor grades and poor prognosis in hepatocellular carcinoma (Chen et al., 2017). Over-expression of XDH resulted in an attenuation of EMT properties in vitro and in vivo. It is plausible that XDH over-expression may have similar effects in breast cancer cell lines that are highly metastatic. Interestingly, XDH was determined to perform retinol oxidation in human mammary epithelial cells (HMEC) (Taibi et al., 2008).

Retinoic acid is an active metabolite that controls differentiation, proliferation, and maintenance of epithelia during embryonic development. Our study did not look into metabolites involved in retinoic acid biosynthesis but it is interesting to speculate that these metabolites, in addition to nucleotides, could be altered during EMT and may provide insight into the metabolic pathways that support a dedifferentiated phenotype. A follow-up study demonstrated that indeed the breast cancer cell lines MCF7 and MDA-MB-231 have reduced ability to produce retinoic acid compared to HMECs (Taibi et al., 2009). Furthermore, this study demonstrated that the enzymatic activity is further decreased due to exposure to estrogenic hormone estradiol (E2). The most significant

E2 mediated inhibition of XDH protein and activity occurred in MCF7 cells, most likely due to the presence of ER. The interaction of E2 with the estrogen receptor (ER) results in DNA binding and transcriptional activation of multiple signaling pathways during mammary gland development and breast cancer progression. It would be valuable to elucidate the mechanism in normal mammary gland development that connects FOXC2 and XDH activity with ER. Questions remain about the regulatory role that directs XDH expression upstream of FOXC2 and it would be interesting to delineate how receptor status and response to E2 signaling could modulate XDH expression during EMT.

129

8.3 Suppression of GLS2 in EMT directed by FOXC2

In chapter 4 we identified a prognostic metabolite signature that included glutamine and glutamate. In chapter 5 and 6 I further explored the role of FOXC2 as a regulator of metabolism in EMT. Furthermore, I identified that FOXC2 regulates the expression of GLS2, a glutaminase that converts glutamine to glutamate. I identified a novel mechanism by which p38MAPK-FOXC2 signaling represses GLS2 (Figure 25). This signaling pathway results in the de-regulation of glutamine metabolism during EMT which in turn supports the pre-dominant glycolytic metabolism. In breast cancer patient samples, high GLS2 expression is associated with better relapse free survival. I reasoned that during EMT GLS2 expression is modulated intentionally to support the bioenergetics of EMT progression. I characterized multiple biochemical parameters downstream of GLS2 activity that are involved in mitochondrial metabolism such as ATP, oxygen consumption, reduced glutathione and mitochondrial membrane potential. During EMT, these parameters decrease and over-expression of GLS2 reverses this phenotype. Enhancement of mitochondrial metabolism not only led to a decrease in glycolytic activity but also a decrease in the production of ROS (Figure 28). I attribute this reduction to the enhanced GSH levels that can neutralize ROS. These effects led to the significant decrease in sphere formation, without decreasing EMT markers or FOXC2 levels. This provocative implication suggests that altering the metabolic program alone can significantly impact EMT/CSC properties.

One of the biggest challenges in understanding the distinct roles between GLS and GLS2 in cancer has been elucidating a potentially distinct function in addition to glutaminase activity. Evidence that alludes this idea comes from the distinct primary structure of the GLS and

GLS2 proteins. Both proteins contain a glutaminase domain that is highly conserved, mitochondrial import pre-sequences, ankyrin repeats, and even nuclear receptor interaction motifs. However, in contrast to GLS, GLS2 contains a PDZ protein interaction motif (ESMV) at the C-terminus (Aledo et al., 2001; Olalla et al., 2001). Two GLS2 PDZ domain interacting proteins have been identified in the human brain; α1-syntropin and glutaminase-interacting 130 protein (GIP). It is possible that GLS2 may interact with distinct proteins in human mammary cells and these interactions could provide insight into how GLS2 is regulated during EMT.

Alternatively, Both GLS and GLS2 contain a nuclear receptor interacting motif (LXXLL) and nuclear localization of GLS2 in rat and monkey brain (Olalla et al., 2002). Interestingly there is a predicted ER-α recognition site on the promoter of GLS2 (Martin-Rufian et al., 2012). ER-α belongs to a family of nuclear hormone receptors and plays a critical role in breast cancer progression and metastasis (Fuqua, 2001). Substantial work characterizing mutations of ER-α in breast cancer patients has shown their contribution in metastasis and therapy resistance (Li et al., 2013; Skliris et al., 2010; Zhang et al., 1997). Future work to explore if indeed ER-α directly regulates GLS2 would elucidate a new regulator mechanism for GLS2 and would also offer new insights into the role of metabolism associated gene regulation during EMT by ER-α and its associated mutations.

During embryonic development, metabolic status is linked with fundamental stem cell function (Folmes et al., 2012) and metabolic plasticity favors the evolving demands of self- renewal and differentiation in stem cells. A limitation of the model system we used is that the cells are “locked” in either extreme of EMT. Questions remain about modeling the metabolic phenotype of the more physiologically relevant hybrid epithelial/mesenchymal (hybrid E/M) state.

While cells undergo EMT, a hybrid E/M state is intermediary and cells have mixed epithelial and mesenchymal traits (Jolly et al., 2016). In the context of metastasis, this hybrid E/M state has been observed in circulating tumor cells (CTCs) of patients with advanced stage lung, prostate, and breast cancers (Armstrong et al., 2011; Lecharpentier et al., 2011). This state allows cells to undergo collective migration, efficiently entering and exiting the bloodstream, resist anoikis, and seed new tumors compared to CTCs that move individually and have only mesenchymal traits

(Aceto et al., 2015; Joosse et al., 2015). One could hypothesize that a high degree of metabolic plasticity enables the hybrid E/M state. To this end, a potential future research endeavor could characterize the metabolic program that accompanies the hybrid E/M state and determine how

131 this program is regulated and maintained flexible to support the completion of the metastatic cascade.

In development and cancer, ROS influence cell fate by activating ERKs, JNKs, and p38MAPKs (Son et al., 2013b). A future challenge is to determine if ROS activates p38MAPK in the context of EMT, similar to the activation that occurs in development. It is likely that in addition to fluctuating nutrient levels, CSCs are exposed to fluctuating levels of oxidative stress therefore it is necessary to understand how this activation could offer CSCs an advantage in circulation and upon arriving in the foreign microenvironment of a distant organ. In hematopoietic stem cells, the activation of p38MAPK by ROS induced stem cell exhaustion and inhibition of self-renewal

(Ito et al., 2006). While we observe higher ROS in EMT-induced cells with intact p38MAPK-

FOXC2 signaling, one could hypothesize that modulation of this signaling could impact the hybrid

E/M state. Furthermore, the role of ROS in tumor progression and metastasis is controversial. In normal cells, excessive ROS causes apoptosis however, it can also be pro-tumorigenic and cause DNA damage and genomic instability (Ames et al., 1993; Behrend et al., 2003). However, cancer cells also have the capacity to maintain high levels of ROS by increasing their antioxidant capacity (Chandel and Tuveson, 2014; Sullivan and Chandel, 2014). In contrast, in our models we observe that induction of EMT results in a metabolic pathway that reduces the levels of antioxidant GSH and maintains high ROS. Another report determined that neutralization of ROS with antioxidant N-acetylcysteine (NAC) inhibited the onset of EMT in the TGFβ-inducible system in renal tubular epithelial cells. Interestingly, NAC treatment inhibited p38MAPK activation (Rhyu et al., 2005). This raises questions about how ROS is modulated during EMT to maintain its signaling.

Although others have demonstrated intact oxidative phosphorylation and glutamine addiction to be important for metastasis (Commisso et al., 2013; LeBleu et al., 2014), we demonstrate EMT-induced cells through our proposed mechanism exert glutamine independence. Glutamine independence has been previously proposed to arise from adaptations 132 of glutamine addicted cells to survive in low-glutamine conditions by employing pyruvate carboxylase (PC) to use glucose derived pyruvate for survival (Cheng et al., 2011). I further propose that the lack of reliance on glutamine provides flexibility to employ glycolysis in an efficient manner. Others have identified signaling that promotes mitochondrial metabolism in breast cancer (LeBleu et al., 2014; Sotgia et al., 2012). In more recent studies comparing metabolic gene signatures of multiple metastatic sites arising from the murine metastatic cells line 4T1 primary tumors distinct differences in metabolic programs at the different organ metastatic sites have been identified (Dupuy et al., 2015). It is unclear if these differences in metabolic programs are adaptations to the organ microenvironment upon seeding or due to metabolic heterogeneity within the primary tumor that favors dissemination to a specific site.

There is an increased appreciation for the idea that a high degree of metabolic plasticity favors the evolving needs of a CSC (Lehuede et al., 2016; Luo and Wicha, 2015). In a need to identify therapeutic targets to inhibit CSCs, the metabolic dependencies of the hybrid E/M state could be explored.

My thesis work has presented FOXC2 as a regulator of metabolism in EMT but future work remains on characterizing its role as a direct transcriptional repressor of metabolic genes.

EMT-TF FOXC2 has been generally described as a transcriptional activator however; a recent report presented evidence that FOXC2 transcriptionally represses p120-catenin enhancing a migratory phenotype in breast cancer (Pham et al., 2017). This is an important point to explore because both GLS2 and XDH were significantly down-regulated in EMT-induced cells and re- expressed upon FOXC2 inhibition. Alternatively, it is possible that there is cooperation between

ZEB1 and FOXC2 for repressive activity as I also observed that ZEB1 did not repress GLS2 in the absence of FOXC2.

Further, there is substantial evidence in glioblastoma, liver, and colon cancer demonstrating that GLS2 expression is lost as a result of DNA hyper-methylation (Szeliga et al.,

2016; Zhang et al., 2013a). In these studies using liver and colon cancer cell lines, treatment with 133

DNA demethylating agent 5-aza-2’-deoxycytidine (Aza) increased GLS2 mRNA levels. Further, restoring GLS2 in these cancer cell lines resulted in inhibition of cell proliferation, and cell cycle arrest in the G2/M phase. Similarly, I observed that treatment of Aza in HMLER-FOXC2 and

SUM159 cells significantly enhances GLS2 mRNA expression (Chapter 6) and downregulates

FOXC2. This is the first evidence that DNA methylation may be playing in a role in GLS2 suppression in EMT-induced cells and cells enriched with EMT properties. A future challenge is to determine if the induction of EMT results in hyper-methylation of the GLS2 promoter. This would provide insights about the epigenetic modifications that affect metabolism during EMT.

Epigenetic regulation of cellular metabolism has evolved in eukaryotes to have the ability to respond to fluctuating environmental signals (Xu et al., 2016). Dysregulation of DNA methylation significantly contributes to metabolic disorders like diabetes by regulating gene expression and silencing (Barres and Zierath, 2011; de Mello et al., 2014). It is interesting to speculate that this mechanism could also play a role in dysregulating metabolism associated genes like GLS2 during EMT.

8.4 Therapeutic potential of targeting p38MAPK-FOXC2 in a novel strategy

In chapter 7, I described a novel dual targeting therapeutic strategy where I employed a p38MAPK inhibitor to target CSCs and a chemotherapeutic agent carboplatin to simultaneously reduce the bulk tumor. I hypothesized that targeting p38MAPK-FOXC2 signaling would efficiently eliminate CSCs as we have evidence that this signaling not only drives EMT and metastatic potential but also maintains a distinct metabolic program (Chapter 6). To test the effectiveness of this strategy, I used the highly aggressive mouse metastatic breast cancer cell line, 4T1. This cell line has high endogenous levels of FOXC2 and when implanted into the mouse mammary gland, has the ability to complete the metastatic cascade resulting in lung metastasis (Mani et al., 2007a). We also reasoned that in using a mouse cell line, we would be modeling a more physiologically relevant scenario since we could inject these cells into a mouse with an intact immune system (Balb/c). We hypothesized that a dual targeting strategy would work in synergy 134 to maximize the therapeutic benefit (Figure 33). Our endpoint criteria were reduction in tumor volume and lung metastasis. The group that received p38MAPK inhibitor exhibited a modest reduction in tumor volume however, the dual treatment had a comparable decrease in tumor volume as the group receiving carboplatin alone. Notably, the group that received the dual treatment also had less lung metastasis than the groups receiving carboplatin or p38MAPK inhibitor alone (Figure 37). Furthermore, this group also had the highest survival rate which can be attributed to the reduction of tumor burden and metastasis. This is the first evidence of a dual targeting strategy modeled in vivo that targets p38MAPK-FOXC2 signaling specifically.

Questions remain about how specific the dual treatment was to reducing the CSC population in the bulk tumor. To address this, immunohistochemical analysis of the tumors for p38MAPK, FOXC2, E-cadherin, and vimentin protein would determine if the dual treatment reduced the levels of EMT/CSC associated markers. Furthermore, including Ki67 to determine differences in proliferation and caspase-3 for apoptosis would determine the efficacy of combination treatment-induced cell death.

One of the challenges facing treatment options in metastatic breast cancer is chemotherapy resistance and tumor relapse (Zardavas et al., 2013). Ideally, the goal for a treatment strategy is to prevent metastasis however 30% of breast cancer patients progress after treatment (O'Shaughnessy, 2005). Once the cancer has metastasized, treatment options are limited. The CSC population contributes to this propensity for resistance and relapse and high degree of heterogeneity (Cancer Genome Atlas, 2012; Shah et al., 2012). A future challenge of this project is to determine how effective this strategy is for reducing tumor relapse. One possible way to model this is to remove the tumor bulk by performing survival surgeries on mice that receive single and dual treatment and monitor metastasis. This data could provide evidence to determine if p38MAPK-FOXC2 signaling is effective to eliminating the CSC population that leads to tumor relapse. This analysis would be informative and clinically relevant because FOXC2 expression has been shown to be elevated in residual surviving cells in patients treated with 135 conventional chemotherapy (Creighton et al., 2009). A potential future experimental adjustment of this model could include the p38MAPK inhibitor, LY2228820, which is currently being used in patients in clinical trials for advanced and metastatic cancers (Goetz et al., 2012), including glioblastoma, and ovarian cancer.

136

Bibliography

1. Aceto, N., Toner, M., Maheswaran, S., and Haber, D.A. (2015). En Route to Metastasis:

Circulating Tumor Cell Clusters and Epithelial-to-Mesenchymal Transition. Trends

Cancer 1, 44-52.

2. Al-Hajj, M., Wicha, M.S., Benito-Hernandez, A., Morrison, S.J., and Clarke, M.F. (2003).

Prospective identification of tumorigenic breast cancer cells. Proc Natl Acad Sci U S A

100, 3983-3988.

3. Aledo, J.C., Rosado, A., Olalla, L., Campos, J.A., and Marquez, J. (2001).

Overexpression, purification, and characterization of glutaminase-interacting protein, a

PDZ-domain protein from human brain. Protein Expr Purif 23, 411-418.

4. Altenberg, B., and Greulich, K.O. (2004). Genes of glycolysis are ubiquitously

overexpressed in 24 cancer classes. Genomics 84, 1014-1020.

5. Altman, B.J., Stine, Z.E., and Dang, C.V. (2016a). From Krebs to clinic: glutamine

metabolism to cancer therapy. Nature reviews. Cancer 16, 749.

6. Altman, B.J., Stine, Z.E., and Dang, C.V. (2016b). From Krebs to clinic: glutamine

metabolism to cancer therapy. Nature reviews. Cancer 16, 619-634.

7. Ames, B.N., Shigenaga, M.K., and Hagen, T.M. (1993). Oxidants, antioxidants, and the

degenerative diseases of aging. Proc Natl Acad Sci U S A 90, 7915-7922.

8. Armstrong, A.J., Marengo, M.S., Oltean, S., Kemeny, G., Bitting, R.L., Turnbull, J.D.,

Herold, C.I., Marcom, P.K., George, D.J., and Garcia-Blanco, M.A. (2011). Circulating

tumor cells from patients with advanced prostate and breast cancer display both

epithelial and mesenchymal markers. Mol Cancer Res 9, 997-1007.

9. Aslakson, C.J., and Miller, F.R. (1992). Selective events in the metastatic process

defined by analysis of the sequential dissemination of subpopulations of a mouse

mammary tumor. Cancer Res 52, 1399-1405.

137

10. Barres, R., and Zierath, J.R. (2011). DNA methylation in metabolic disorders. Am J Clin

Nutr 93, 897S-900.

11. Barretina, J., Caponigro, G., Stransky, N., Venkatesan, K., Margolin, A.A., Kim, S.,

Wilson, C.J., Lehar, J., Kryukov, G.V., Sonkin, D., et al. (2012). The Cancer Cell Line

Encyclopedia enables predictive modelling of anticancer drug sensitivity. Nature 483,

603-607.

12. Beca, F., and Polyak, K. (2016). Intratumor Heterogeneity in Breast Cancer. Adv Exp

Med Biol 882, 169-189.

13. Behrend, L., Henderson, G., and Zwacka, R.M. (2003). Reactive oxygen species in

oncogenic transformation. Biochem Soc Trans 31, 1441-1444.

14. Beloribi-Djefaflia, S., Vasseur, S., and Guillaumond, F. (2016). Lipid metabolic

reprogramming in cancer cells. Oncogenesis 5, e189.

15. Bergstrom, J., Furst, P., Noree, L.O., and Vinnars, E. (1974). Intracellular free amino

acid concentration in human muscle tissue. J Appl Physiol 36, 693-697.

16. Bhowmik, S.K., Ramirez-Pena, E., Arnold, J.M., Putluri, V., Sphyris, N., Michailidis, G.,

Putluri, N., Ambs, S., Sreekumar, A., and Mani, S.A. (2015). EMT-induced metabolite

signature identifies poor clinical outcome. Oncotarget 6, 42651-42660.

17. Bid, H.K., Roberts, R.D., Manchanda, P.K., and Houghton, P.J. (2013). RAC1: an

emerging therapeutic option for targeting cancer angiogenesis and metastasis. Mol

Cancer Ther 12, 1925-1934.

18. Bonnet, D., and Dick, J.E. (1997). Human acute myeloid leukemia is organized as a

hierarchy that originates from a primitive hematopoietic cell. Nat Med 3, 730-737.

19. Borger, D.R., Goyal, L., Yau, T., Poon, R.T., Ancukiewicz, M., Deshpande, V.,

Christiani, D.C., Liebman, H.M., Yang, H., Kim, H., et al. (2014). Circulating

oncometabolite 2-hydroxyglutarate is a potential surrogate biomarker in patients with

isocitrate dehydrogenase-mutant intrahepatic cholangiocarcinoma. Clin Cancer Res 20,

1884-1890. 138

20. Borst, P. (2012). Cancer drug pan-resistance: pumps, cancer stem cells, quiescence,

epithelial to mesenchymal transition, blocked cell death pathways, persisters or what?

Open Biol 2, 120066.

21. Brooks, M.D., Burness, M.L., and Wicha, M.S. (2015). Therapeutic Implications of

Cellular Heterogeneity and Plasticity in Breast Cancer. Cell Stem Cell 17, 260-271.

22. Campbell, R.M., Anderson, B.D., Brooks, N.A., Brooks, H.B., Chan, E.M., De Dios, A.,

Gilmour, R., Graff, J.R., Jambrina, E., Mader, M., et al. (2014). Characterization of

LY2228820 dimesylate, a potent and selective inhibitor of p38 MAPK with antitumor

activity. Mol Cancer Ther 13, 364-374.

23. Cancer Genome Atlas, N. (2012). Comprehensive molecular portraits of human breast

tumours. Nature 490, 61-70.

24. Cano, A., Perez-Moreno, M.A., Rodrigo, I., Locascio, A., Blanco, M.J., del Barrio, M.G.,

Portillo, F., and Nieto, M.A. (2000). The transcription factor snail controls epithelial-

mesenchymal transitions by repressing E-cadherin expression. Nat Cell Biol 2, 76-83.

25. Cardoso, F., Beishon, M., Cardoso, M.J., Corneliussen-James, D., Gralow, J., Mertz,

S., Papadopoulos, E., Paonessa, D., Peccatori, F., Sabelko, K., et al. (2016). Global

status of advanced/metastatic breast cancer (ABC/mBC): A Decade Report 2005-2015.

European Journal of Cancer 57, S5-S6.

26. Cederberg, A., Gronning, L.M., Ahren, B., Tasken, K., Carlsson, P., and Enerback, S.

(2001). FOXC2 is a winged helix gene that counteracts obesity, hypertriglyceridemia,

and diet-induced insulin resistance. Cell 106, 563-573.

27. Chandel, N.S., and Tuveson, D.A. (2014). The promise and perils of antioxidants for

cancer patients. N Engl J Med 371, 177-178.

28. Chen, G.L., Ye, T., Chen, H.L., Zhao, Z.Y., Tang, W.Q., Wang, L.S., and Xia, J.L.

(2017). Xanthine dehydrogenase downregulation promotes TGFbeta signaling and

cancer stem cell-related gene expression in hepatocellular carcinoma. Oncogenesis 6,

e382. 139

29. Cheng, T., Sudderth, J., Yang, C., Mullen, A.R., Jin, E.S., Mates, J.M., and

DeBerardinis, R.J. (2011). Pyruvate carboxylase is required for glutamine-independent

growth of tumor cells. Proc Natl Acad Sci U S A 108, 8674-8679.

30. Commisso, C., Davidson, S.M., Soydaner-Azeloglu, R.G., Parker, S.J., Kamphorst, J.J.,

Hackett, S., Grabocka, E., Nofal, M., Drebin, J.A., Thompson, C.B., et al. (2013).

Macropinocytosis of protein is an amino acid supply route in Ras-transformed cells.

Nature 497, 633-637.

31. Creighton, C.J., Li, X., Landis, M., Dixon, J.M., Neumeister, V.M., Sjolund, A., Rimm,

D.L., Wong, H., Rodriguez, A., Herschkowitz, J.I., et al. (2009). Residual breast cancers

after conventional therapy display mesenchymal as well as tumor-initiating features.

Proceedings of the National Academy of Sciences of the United States of America 106,

13820-13825.

32. de Mello, V.D., Pulkkinen, L., Lalli, M., Kolehmainen, M., Pihlajamaki, J., and Uusitupa,

M. (2014). DNA methylation in obesity and type 2 diabetes. Ann Med 46, 103-113.

33. de Souza, C.M., Araujo e Silva, A.C., de Jesus Ferraciolli, C., Moreira, G.V., Campos,

L.C., dos Reis, D.C., Lopes, M.T., Ferreira, M.A., Andrade, S.P., and Cassali, G.D.

(2014). Combination therapy with carboplatin and thalidomide suppresses tumor growth

and metastasis in 4T1 murine breast cancer model. Biomed Pharmacother 68, 51-57.

34. DeBerardinis, R.J., and Chandel, N.S. (2016). Fundamentals of cancer metabolism.

Science advances 2, e1600200.

35. DeSantis, C.E., Bray, F., Ferlay, J., Lortet-Tieulent, J., Anderson, B.O., and Jemal, A.

(2015). International Variation in Female Breast Cancer Incidence and Mortality Rates.

Cancer Epidemiol Biomarkers Prev 24, 1495-1506.

36. Dexter, D.L., Kowalski, H.M., Blazar, B.A., Fligiel, Z., Vogel, R., and Heppner, G.H.

(1978). Heterogeneity of tumor cells from a single mouse mammary tumor. Cancer Res

38, 3174-3181.

140

37. Dirat, B., Bochet, L., Dabek, M., Daviaud, D., Dauvillier, S., Majed, B., Wang, Y.Y.,

Meulle, A., Salles, B., Le Gonidec, S., et al. (2011). Cancer-associated adipocytes

exhibit an activated phenotype and contribute to breast cancer invasion. Cancer Res

71, 2455-2465.

38. Dong, C., Yuan, T., Wu, Y., Wang, Y., Fan, T.W., Miriyala, S., Lin, Y., Yao, J., Shi, J.,

Kang, T., et al. (2013). Loss of FBP1 by Snail-mediated repression provides metabolic

advantages in basal-like breast cancer. Cancer Cell 23, 316-331.

39. Dowsett, M., Forbes, J.F., Bradley, R., Ingle, J., Aihara, T., Bliss, J., Boccardo, F.,

Coates, A., Coombes, R.C., Cuzick, J., et al. (2015). Aromatase inhibitors versus

tamoxifen in early breast cancer: patient-level meta-analysis of the randomised trials.

Lancet 386, 1341-1352.

40. Dunlap, S.M., Chiao, L.J., Nogueira, L., Usary, J., Perou, C.M., Varticovski, L., and

Hursting, S.D. (2012). Dietary energy balance modulates epithelial-to-mesenchymal

transition and tumor progression in murine claudin-low and basal-like mammary tumor

models. Cancer Prev Res (Phila) 5, 930-942.

41. Dupuy, F., Tabaries, S., Andrzejewski, S., Dong, Z., Blagih, J., Annis, M.G., Omeroglu,

A., Gao, D., Leung, S., Amir, E., et al. (2015). PDK1-Dependent Metabolic

Reprogramming Dictates Metastatic Potential in Breast Cancer. Cell Metab 22, 577-

589.

42. Dykxhoorn, D.M., Wu, Y., Xie, H., Yu, F., Lal, A., Petrocca, F., Martinvalet, D., Song, E.,

Lim, B., and Lieberman, J. (2009). miR-200 enhances mouse breast cancer cell

colonization to form distant metastases. PLoS One 4, e7181.

43. Elenbaas, B., Spirio, L., Koerner, F., Fleming, M.D., Zimonjic, D.B., Donaher, J.L.,

Popescu, N.C., Hahn, W.C., and Weinberg, R.A. (2001). Human breast cancer cells

generated by oncogenic transformation of primary mammary epithelial cells. Genes &

development 15, 50-65.

141

44. Fiehn, O. (2002). Metabolomics--the link between genotypes and phenotypes. Plant Mol

Biol 48, 155-171.

45. Fillmore, C.M., and Kuperwasser, C. (2008). Human breast cancer cell lines contain

stem-like cells that self-renew, give rise to phenotypically diverse progeny and survive

chemotherapy. Breast Cancer Res 10, R25.

46. Folmes, C.D., Dzeja, P.P., Nelson, T.J., and Terzic, A. (2012). Metabolic plasticity in

stem cell homeostasis and differentiation. Cell Stem Cell 11, 596-606.

47. Fountzilas, G., Athanassiades, A., Papadimitriou, V., Dimopoulos, M.A., Bafaloukos, D.,

Aravantinos, G., Nicolaides, C., Kalofonos, H., Papakostas, P., Xiros, N., and Razi, E.

(1998). Paclitaxel and carboplatin as first-line chemotherapy for advanced breast

cancer. Oncology (Williston Park) 12, 45-48.

48. Fristrom, D. (1988). The cellular basis of epithelial morphogenesis. A review. Tissue

Cell 20, 645-690.

49. Fuqua, S.A. (2001). The role of estrogen receptors in breast cancer metastasis. J

Mammary Gland Biol Neoplasia 6, 407-417.

50. Galluzzi, L., Kepp, O., Vander Heiden, M.G., and Kroemer, G. (2013). Metabolic targets

for cancer therapy. Nature reviews. Drug discovery 12, 829-846.

51. Gan, L., Liu, Z., Feng, F., Wu, T., Luo, D., Hu, C., and Sun, C. (2018). Foxc2

coordinates inflammation and browning of white adipose by leptin-STAT3-PRDM16

signal in mice. Int J Obes (Lond) 42, 252-259.

52. Gao, P., Tchernyshyov, I., Chang, T.C., Lee, Y.S., Kita, K., Ochi, T., Zeller, K.I., De

Marzo, A.M., Van Eyk, J.E., Mendell, J.T., and Dang, C.V. (2009). c-Myc suppression of

miR-23a/b enhances mitochondrial glutaminase expression and glutamine metabolism.

Nature 458, 762-765.

53. Garnett, M.J., Edelman, E.J., Heidorn, S.J., Greenman, C.D., Dastur, A., Lau, K.W.,

Greninger, P., Thompson, I.R., Luo, X., Soares, J., et al. (2012). Systematic

142

identification of genomic markers of drug sensitivity in cancer cells. Nature 483, 570-

575.

54. Giacobbe, A., Bongiorno-Borbone, L., Bernassola, F., Terrinoni, A., Markert, E.K.,

Levine, A.J., Feng, Z., Agostini, M., Zolla, L., Agro, A.F., et al. (2013). p63 regulates

glutaminase 2 expression. Cell Cycle 12, 1395-1405.

55. Goetz, M.P., Tolcher, A.W., Haluska, P., Papadopoulos, K.P., Erlichman, C., Beeram,

M., Lensing, J., Rasco, D.W., Molina, J.R., Arcos, R., et al. (2012). A first-in-human

phase I study of the oral p38 MAPK inhibitor LY2228820 dimesylate in patients with

advanced cancer. Journal of Clinical Oncology 30.

56. Gum, R.J., McLaughlin, M.M., Kumar, S., Wang, Z., Bower, M.J., Lee, J.C., Adams,

J.L., Livi, G.P., Goldsmith, E.J., and Young, P.R. (1998). Acquisition of sensitivity of

stress-activated protein kinases to the p38 inhibitor, SB 203580, by alteration of one or

more amino acids within the ATP binding pocket. The Journal of biological chemistry

273, 15605-15610.

57. Hakansson, J., Eliasson, B., Smith, U., and Enerback, S. (2011). Adipocyte

mitochondrial genes and the forkhead factor FOXC2 are decreased in type 2 diabetes

patients and normalized in response to rosiglitazone. Diabetology & metabolic

syndrome 3, 32.

58. Hanahan, D., and Weinberg, R.A. (2011). Hallmarks of cancer: the next generation. Cell

144, 646-674.

59. Hay, N. (2016). Reprogramming glucose metabolism in cancer: can it be exploited for

cancer therapy? Nature reviews. Cancer 16, 635-649.

60. Hinestrosa, M.C., Dickersin, K., Klein, P., Mayer, M., Noss, K., Slamon, D., Sledge, G.,

and Visco, F.M. (2007). Shaping the future of biomarker research in breast cancer to

ensure clinical relevance. Nature reviews. Cancer 7, 309-315.

61. Hollier, B.G., Tinnirello, A.A., Werden, S.J., Evans, K.W., Taube, J.H., Sarkar, T.R.,

Sphyris, N., Shariati, M., Kumar, S.V., Battula, V.L., et al. (2013a). FOXC2 expression 143

links epithelial-mesenchymal transition and stem cell properties in breast cancer.

Cancer Research 73, 1981-1992.

62. Hollier, B.G., Tinnirello, A.A., Werden, S.J., Evans, K.W., Taube, J.H., Sarkar, T.R.,

Sphyris, N., Shariati, M., Kumar, S.V., Battula, V.L., et al. (2013b). FOXC2 expression

links epithelial-mesenchymal transition and stem cell properties in breast cancer.

Cancer Res 73, 1981-1992.

63. Hong, S.S., Lee, H., and Kim, K.W. (2004). HIF-1alpha: a valid therapeutic target for

tumor therapy. Cancer Res Treat 36, 343-353.

64. Howlader, N., Altekruse, S.F., Li, C.I., Chen, V.W., Clarke, C.A., Ries, L.A.G., and

Cronin, K.A. (2014). US Incidence of Breast Cancer Subtypes Defined by Joint

Hormone Receptor and HER2 Status. Jnci-J Natl Cancer I 106.

65. Hu, W., Zhang, C., Wu, R., Sun, Y., Levine, A., and Feng, Z. (2010). Glutaminase 2, a

novel p53 target gene regulating energy metabolism and antioxidant function. Proc Natl

Acad Sci U S A 107, 7455-7460.

66. Ishitsuka, K., Hideshima, T., Neri, P., Vallet, S., Shiraishi, N., Okawa, Y., Shen, Z.,

Raje, N., Kiziltepe, T., Ocio, E.M., et al. (2008). p38 mitogen-activated protein kinase

inhibitor LY2228820 enhances bortezomib-induced cytotoxicity and inhibits

osteoclastogenesis in multiple myeloma; therapeutic implications. British journal of

haematology 141, 598-606.

67. Ito, K., Hirao, A., Arai, F., Takubo, K., Matsuoka, S., Miyamoto, K., Ohmura, M., Naka,

K., Hosokawa, K., Ikeda, Y., and Suda, T. (2006). Reactive oxygen species act through

p38 MAPK to limit the lifespan of hematopoietic stem cells. Nat Med 12, 446-451.

68. Jang, M., Kim, S.S., and Lee, J. (2013). Cancer cell metabolism: implications for

therapeutic targets. Exp Mol Med 45, e45.

69. Jolly, M.K., Tripathi, S.C., Jia, D., Mooney, S.M., Celiktas, M., Hanash, S.M., Mani,

S.A., Pienta, K.J., Ben-Jacob, E., and Levine, H. (2016). Stability of the hybrid

epithelial/mesenchymal phenotype. Oncotarget 7, 27067-27084. 144

70. Joosse, S.A., Gorges, T.M., and Pantel, K. (2015). Biology, detection, and clinical

implications of circulating tumor cells. EMBO Mol Med 7, 1-11.

71. Kaelin, W.G., Jr. (2009). SDH5 mutations and familial paraganglioma: somewhere

Warburg is smiling. Cancer Cell 16, 180-182.

72. Katt, W.P., Lukey, M.J., and Cerione, R.A. (2017). A tale of two glutaminases:

homologous enzymes with distinct roles in tumorigenesis. Future Med Chem 9, 223-

243.

73. Kung, H.N., Marks, J.R., and Chi, J.T. (2011). Glutamine synthetase is a genetic

determinant of cell type-specific glutamine independence in breast epithelia. PLoS

Genet 7, e1002229.

74. Lacey, J.M., and Wilmore, D.W. (1990). Is glutamine a conditionally essential amino

acid? Nutr Rev 48, 297-309.

75. Lagadec, C., Vlashi, E., Della Donna, L., Meng, Y., Dekmezian, C., Kim, K., and Pajonk,

F. (2010). Survival and self-renewing capacity of breast cancer initiating cells during

fractionated radiation treatment. Breast Cancer Res 12, R13.

76. LeBleu, V.S., O'Connell, J.T., Gonzalez Herrera, K.N., Wikman, H., Pantel, K., Haigis,

M.C., de Carvalho, F.M., Damascena, A., Domingos Chinen, L.T., Rocha, R.M., et al.

(2014). PGC-1alpha mediates mitochondrial biogenesis and oxidative phosphorylation

in cancer cells to promote metastasis. Nat Cell Biol 16, 992-1003, 1001-1015.

77. Lecharpentier, A., Vielh, P., Perez-Moreno, P., Planchard, D., Soria, J.C., and Farace,

F. (2011). Detection of circulating tumour cells with a hybrid (epithelial/mesenchymal)

phenotype in patients with metastatic non-small cell lung cancer. Br J Cancer 105,

1338-1341.

78. Lee, H.J., and Zheng, J.J. (2010). PDZ domains and their binding partners: structure,

specificity, and modification. Cell Commun Signal 8, 8.

79. Lee, H.W., Osis, G., Handlogten, M.E., Lamers, W.H., Chaudhry, F.A., Verlander, J.W.,

and Weiner, I.D. (2016a). Proximal tubule-specific glutamine synthetase deletion alters 145

basal and acidosis-stimulated ammonia metabolism. Am J Physiol Renal Physiol 310,

F1229-1242.

80. Lee, S.Y., Jeon, H.M., Ju, M.K., Jeong, E.K., Kim, C.H., Park, H.G., Han, S.I., and

Kang, H.S. (2016b). Dlx-2 and glutaminase upregulate epithelial-mesenchymal

transition and glycolytic switch. Oncotarget 7, 7925-7939.

81. Lee, S.Y., Jeon, H.M., Ju, M.K., Kim, C.H., Yoon, G., Han, S.I., Park, H.G., and Kang,

H.S. (2012). Wnt/Snail signaling regulates cytochrome C oxidase and glucose

metabolism. Cancer Res 72, 3607-3617.

82. Lehmann, B.D., Bauer, J.A., Chen, X., Sanders, M.E., Chakravarthy, A.B., Shyr, Y., and

Pietenpol, J.A. (2011). Identification of human triple-negative breast cancer subtypes

and preclinical models for selection of targeted therapies. J Clin Invest 121, 2750-2767.

83. Lehmann, B.D., and Pietenpol, J.A. (2014). Identification and use of biomarkers in

treatment strategies for triple-negative breast cancer subtypes. J Pathol 232, 142-150.

84. Lehuede, C., Dupuy, F., Rabinovitch, R., Jones, R.G., and Siegel, P.M. (2016).

Metabolic Plasticity as a Determinant of Tumor Growth and Metastasis. Cancer Res 76,

5201-5208.

85. Lengyel, E., Makowski, L., DiGiovanni, J., and Kolonin, M.G. (2018). Cancer as a Matter

of Fat: The Crosstalk between Adipose Tissue and Tumors. Trends Cancer 4, 374-384.

86. Li, S., Shen, D., Shao, J., Crowder, R., Liu, W., Prat, A., He, X., Liu, S., Hoog, J., Lu, C.,

et al. (2013). Endocrine-therapy-resistant ESR1 variants revealed by genomic

characterization of breast-cancer-derived xenografts. Cell Rep 4, 1116-1130.

87. Li, X., Lewis, M.T., Huang, J., Gutierrez, C., Osborne, C.K., Wu, M.F., Hilsenbeck, S.G.,

Pavlick, A., Zhang, X., Chamness, G.C., et al. (2008). Intrinsic resistance of tumorigenic

breast cancer cells to chemotherapy. J Natl Cancer Inst 100, 672-679.

88. Lidell, M.E., Seifert, E.L., Westergren, R., Heglind, M., Gowing, A., Sukonina, V., Arani,

Z., Itkonen, P., Wallin, S., Westberg, F., et al. (2011). The adipocyte-expressed

146

forkhead transcription factor Foxc2 regulates metabolism through altered mitochondrial

function. Diabetes 60, 427-435.

89. Luo, M., and Wicha, M.S. (2015). Metabolic plasticity of cancer stem cells. Oncotarget

6, 35141-35142.

90. Mader, M., de Dios, A., Shih, C., Bonjouklian, R., Li, T., White, W., Lopez de Uralde, B.,

Sanchez-Martinez, C., del Prado, M., Jaramillo, C., et al. (2008). Imidazolyl

benzimidazoles and imidazo[4,5-b]pyridines as potent p38alpha MAP kinase inhibitors

with excellent in vivo antiinflammatory properties. Bioorganic & medicinal chemistry

letters 18, 179-183.

91. Malorni, L., Shetty, P.B., De Angelis, C., Hilsenbeck, S., Rimawi, M.F., Elledge, R.,

Osborne, C.K., De Placido, S., and Arpino, G. (2012). Clinical and biologic features of

triple-negative breast cancers in a large cohort of patients with long-term follow-up.

Breast Cancer Research and Treatment 136, 795-804.

92. Mani, S.A., Guo, W., Liao, M.J., Eaton, E.N., Ayyanan, A., Zhou, A.Y., Brooks, M.,

Reinhard, F., Zhang, C.C., Shipitsin, M., et al. (2008). The epithelial-mesenchymal

transition generates cells with properties of stem cells. Cell 133, 704-715.

93. Mani, S.A., Yang, J., Brooks, M., Schwaninger, G., Zhou, A., Miura, N., Kutok, J.L.,

Hartwell, K., Richardson, A.L., and Weinberg, R.A. (2007a). Mesenchyme Forkhead 1

(FOXC2) plays a key role in metastasis and is associated with aggressive basal-like

breast cancers. Proceedings of the National Academy of Sciences of the United States

of America 104, 10069-10074.

94. Mani, S.A., Yang, J., Brooks, M., Schwaninger, G., Zhou, A., Miura, N., Kutok, J.L.,

Hartwell, K., Richardson, A.L., and Weinberg, R.A. (2007b). Mesenchyme Forkhead 1

(FOXC2) plays a key role in metastasis and is associated with aggressive basal-like

breast cancers. Proc Natl Acad Sci U S A 104, 10069-10074.

95. Marquez, J., Mates, J.M., and Campos-Sandoval, J.A. (2016). Glutaminases. Adv

Neurobiol 13, 133-171. 147

96. Martin-Rufian, M., Tosina, M., Campos-Sandoval, J.A., Manzanares, E., Lobo, C.,

Segura, J.A., Alonso, F.J., Mates, J.M., and Marquez, J. (2012). Mammalian

glutaminase Gls2 gene encodes two functional alternative transcripts by a surrogate

promoter usage mechanism. PLoS One 7, e38380.

97. Martin, M. (2001). Platinum compounds in the treatment of advanced breast cancer.

Clinical breast cancer 2, 190-208; discussion 209

98. Mates, J.M., Segura, J.A., Campos-Sandoval, J.A., Lobo, C., Alonso, L., Alonso, F.J.,

and Marquez, J. (2009). Glutamine homeostasis and mitochondrial dynamics. The

international journal of biochemistry & cell biology 41, 2051-2061.

99. May, C.D., Sphyris, N., Evans, K.W., Werden, S.J., Guo, W., and Mani, S.A. (2011).

Epithelial-mesenchymal transition and cancer stem cells: a dangerously dynamic duo in

breast cancer progression. Breast Cancer Res 13, 202.

100. Mayer, I.A., Abramson, V.G., Lehmann, B.D., and Pietenpol, J.A. (2014). New

strategies for triple-negative breast cancer--deciphering the heterogeneity. Clin Cancer

Res 20, 782-790.

101. Moore, N., and Lyle, S. (2011). Quiescent, slow-cycling stem cell populations in

cancer: a review of the evidence and discussion of significance. J Oncol 2011.

102. Morel, A.P., Lievre, M., Thomas, C., Hinkal, G., Ansieau, S., and Puisieux, A.

(2008). Generation of breast cancer stem cells through epithelial-mesenchymal

transition. PLoS One 3, e2888.

103. Naor, D., Sionov, R.V., and Ish-Shalom, D. (1997). CD44: structure, function,

and association with the malignant process. Adv Cancer Res 71, 241-319.

104. Neale, J.H., Bzdega, T., and Wroblewska, B. (2000). N-Acetylaspartylglutamate:

the most abundant peptide neurotransmitter in the mammalian central nervous system.

J Neurochem 75, 443-452.

148

105. Neve, R.M., Chin, K., Fridlyand, J., Yeh, J., Baehner, F.L., Fevr, T., Clark, L.,

Bayani, N., Coppe, J.P., Tong, F., et al. (2006). A collection of breast cancer cell lines

for the study of functionally distinct cancer subtypes. Cancer Cell 10, 515-527.

106. O'Shaughnessy, J. (2005). Extending survival with chemotherapy in metastatic

breast cancer. Oncologist 10 Suppl 3, 20-29.

107. Olalla, L., Aledo, J.C., Bannenberg, G., and Marquez, J. (2001). The C-terminus

of human glutaminase L mediates association with PDZ domain-containing proteins.

FEBS Lett 488, 116-122.

108. Olalla, L., Gutierrez, A., Campos, J.A., Khan, Z.U., Alonso, F.J., Segura, J.A.,

Marquez, J., and Aledo, J.C. (2002). Nuclear localization of L-type glutaminase in

mammalian brain. The Journal of biological chemistry 277, 38939-38944.

109. Palm, W., and Thompson, C.B. (2017). Nutrient acquisition strategies of

mammalian cells. Nature 546, 234-242.

110. Paranjape, A.N., Soundararajan, R., Werden, S.J., Joseph, R., Taube, J.H., Liu,

H., Rodriguez-Canales, J., Sphyris, N., Wistuba, I., Miura, N., et al. (2016a). Inhibition of

FOXC2 restores epithelial phenotype and drug sensitivity in prostate cancer cells with

stem-cell properties. Oncogene 35, 5963-5976.

111. Paranjape, A.N., Soundararajan, R., Werden, S.J., Joseph, R., Taube, J.H., Liu,

H., Rodriguez-Canales, J., Sphyris, N., Wistuba, I., Miura, N., et al. (2016b). Inhibition of

FOXC2 restores epithelial phenotype and drug sensitivity in prostate cancer cells with

stem-cell properties. Oncogene.

112. Park, S.M., Gaur, A.B., Lengyel, E., and Peter, M.E. (2008). The miR-200 family

determines the epithelial phenotype of cancer cells by targeting the E-cadherin

repressors ZEB1 and ZEB2. Genes Dev 22, 894-907.

113. Pelicano, H., Zhang, W., Liu, J., Hammoudi, N., Dai, J., Xu, R.H., Pusztai, L.,

and Huang, P. (2014). Mitochondrial dysfunction in some triple-negative breast cancer

cell lines: role of mTOR pathway and therapeutic potential. Breast Cancer Res 16, 434. 149

114. Pham, T.N.D., Perez White, B.E., Zhao, H., Mortazavi, F., and Tonetti, D.A.

(2017). Protein kinase C alpha enhances migration of breast cancer cells through

FOXC2-mediated repression of p120-catenin. BMC Cancer 17, 832.

115. Pietila, M., Vijay, G.V., Soundararajan, R., Yu, X., Symmans, W.F., Sphyris, N.,

and Mani, S.A. (2016). FOXC2 regulates the G2/M transition of stem cell-rich breast

cancer cells and sensitizes them to PLK1 inhibition. Sci Rep 6, 23070.

116. Possemato, R., Marks, K.M., Shaul, Y.D., Pacold, M.E., Kim, D., Birsoy, K.,

Sethumadhavan, S., Woo, H.K., Jang, H.G., Jha, A.K., et al. (2011). Functional

genomics reveal that the serine synthesis pathway is essential in breast cancer. Nature

476, 346-350.

117. Prat, A., Adamo, B., Cheang, M.C.U., Anders, C.K., Carey, L.A., and Perou,

C.M. (2013). Molecular Characterization of Basal-Like and Non-Basal-Like Triple-

Negative Breast Cancer. Oncologist 18, 123-133.

118. Prat, A., Parker, J.S., Karginova, O., Fan, C., Livasy, C., Herschkowitz, J.I., He,

X.P., and Perou, C.M. (2010). Phenotypic and molecular characterization of the claudin-

low intrinsic subtype of breast cancer. Breast Cancer Research 12.

119. Rankin, E.B., and Giaccia, A.J. (2016). Hypoxic control of metastasis. Science

352, 175-180.

120. Recouvreux, M.V., and Commisso, C. (2017). Macropinocytosis: A Metabolic

Adaptation to Nutrient Stress in Cancer. Front Endocrinol (Lausanne) 8, 261.

121. Rhyu, D.Y., Yang, Y., Ha, H., Lee, G.T., Song, J.S., Uh, S.T., and Lee, H.B.

(2005). Role of reactive oxygen species in TGF-beta1-induced mitogen-activated

protein kinase activation and epithelial-mesenchymal transition in renal tubular epithelial

cells. J Am Soc Nephrol 16, 667-675.

122. Rohrig, F., and Schulze, A. (2016). The multifaceted roles of fatty acid synthesis

in cancer. Nature reviews. Cancer 16, 732-749.

150

123. Scheel, C., and Weinberg, R.A. (2012). Cancer stem cells and epithelial-

mesenchymal transition: Concepts and molecular links. Seminars in Cancer Biology

22396-403.

124. Shah, S.P., Roth, A., Goya, R., Oloumi, A., Ha, G., Zhao, Y., Turashvili, G.,

Ding, J., Tse, K., Haffari, G., et al. (2012). The clonal and mutational evolution spectrum

of primary triple-negative breast cancers. Nature 486, 395-399.

125. Shaul, Y.D., Freinkman, E., Comb, W.C., Cantor, J.R., Tam, W.L., Thiru, P.,

Kim, D., Kanarek, N., Pacold, M.E., Chen, W.W., et al. (2014). Dihydropyrimidine

accumulation is required for the epithelial-mesenchymal transition. Cell 158, 1094-1109.

126. Shimono, Y., Zabala, M., Cho, R.W., Lobo, N., Dalerba, P., Qian, D., Diehn, M.,

Liu, H., Panula, S.P., Chiao, E., et al. (2009). Downregulation of miRNA-200c links

breast cancer stem cells with normal stem cells. Cell 138, 592-603.

127. Skliris, G.P., Nugent, Z., Watson, P.H., and Murphy, L.C. (2010). Estrogen

receptor alpha phosphorylated at tyrosine 537 is associated with poor clinical outcome

in breast cancer patients treated with tamoxifen. Horm Cancer 1, 215-221.

128. Son, J., Lyssiotis, C.A., Ying, H., Wang, X., Hua, S., Ligorio, M., Perera, R.M.,

Ferrone, C.R., Mullarky, E., Shyh-Chang, N., et al. (2013a). Glutamine supports

pancreatic cancer growth through a KRAS-regulated metabolic pathway. Nature 496,

101-105.

129. Son, Y., Kim, S., Chung, H.T., and Pae, H.O. (2013b). Reactive oxygen species

in the activation of MAP kinases. Methods Enzymol 528, 27-48.

130. Song, L., Tang, H., Liao, W., Luo, X., Li, Y., Chen, T., and Zhang, X. (2017).

FOXC2 positively regulates YAP signaling and promotes the glycolysis of

nasopharyngeal carcinoma. Exp Cell Res 357, 17-24.

131. Sotgia, F., Whitaker-Menezes, D., Martinez-Outschoorn, U.E., Salem, A.F.,

Tsirigos, A., Lamb, R., Sneddon, S., Hulit, J., Howell, A., and Lisanti, M.P. (2012).

Mitochondria "fuel" breast cancer metabolism: fifteen markers of mitochondrial 151

biogenesis label epithelial cancer cells, but are excluded from adjacent stromal cells.

Cell Cycle 11, 4390-4401.

132. Soundararajan, R., Paranjape, A.N., Barsan, V., Chang, J.T., and Mani, S.A.

(2015). A novel embryonic plasticity gene signature that predicts metastatic

competence and clinical outcome. Sci Rep 5, 11766.

133. Souza, C.M., Gamba Cde, O., Campos, C.B., Lopes, M.T., Ferreira, M.A.,

Andrade, S.P., and Cassali, G.D. (2013). Carboplatin delays mammary cancer 4T1

growth in mice. Pathol Res Pract 209, 24-29.

134. Sullivan, L.B., and Chandel, N.S. (2014). Mitochondrial reactive oxygen species

and cancer. Cancer Metab 2, 17.

135. Sun, S., Li, H., Chen, J., and Qian, Q. (2017). Lactic Acid: No Longer an Inert

and End-Product of Glycolysis. Physiology (Bethesda) 32, 453-463.

136. Suzuki, S., Tanaka, T., Poyurovsky, M.V., Nagano, H., Mayama, T., Ohkubo, S.,

Lokshin, M., Hosokawa, H., Nakayama, T., Suzuki, Y., et al. (2010). Phosphate-

activated glutaminase (GLS2), a p53-inducible regulator of glutamine metabolism and

reactive oxygen species. Proc Natl Acad Sci U S A 107, 7461-7466.

137. Szeliga, M., Bogacinska-Karas, M., Kuzmicz, K., Rola, R., and Albrecht, J.

(2016). Downregulation of GLS2 in glioblastoma cells is related to DNA

hypermethylation but not to the p53 status. Mol Carcinog 55, 1309-1316.

138. Taibi, G., Carruba, G., Miceli, V., Cocciadiferro, L., and Nicotra, C.M. (2009).

Estradiol decreases xanthine dehydrogenase enzyme activity and protein expression in

non-tumorigenic and malignant human mammary epithelial cells. J Cell Biochem 108,

688-692.

139. Taibi, G., Di Gaudio, F., and Nicotra, C.M. (2008). Xanthine dehydrogenase

processes retinol to retinoic acid in human mammary epithelial cells. J Enzyme Inhib

Med Chem 23, 317-327.

152

140. Tan, T.Z., Miow, Q.H., Miki, Y., Noda, T., Mori, S., Huang, R.Y., and Thiery, J.P.

(2014). Epithelial-mesenchymal transition spectrum quantification and its efficacy in

deciphering survival and drug responses of cancer patients. EMBO Mol Med 6, 1279-

1293.

141. Taube, J.H., Herschkowitz, J.I., Komurov, K., Zhou, A.Y., Gupta, S., Yang, J.,

Hartwell, K., Onder, T.T., Gupta, P.B., Evans, K.W., et al. (2010a). Core epithelial-to-

mesenchymal transition interactome gene-expression signature is associated with

claudin-low and metaplastic breast cancer subtypes. Proceedings of the National

Academy of Sciences of the United States of America 107, 15449-15454.

142. Taube, J.H., Herschkowitz, J.I., Komurov, K., Zhou, A.Y., Gupta, S., Yang, J.,

Hartwell, K., Onder, T.T., Gupta, P.B., Evans, K.W., et al. (2010b). Core epithelial-to-

mesenchymal transition interactome gene-expression signature is associated with

claudin-low and metaplastic breast cancer subtypes. Proc Natl Acad Sci U S A 107,

15449-15454.

143. Terunuma, A., Putluri, N., Mishra, P., Mathe, E.A., Dorsey, T.H., Yi, M., Wallace,

T.A., Issaq, H.J., Zhou, M., Killian, J.K., et al. (2014). MYC-driven accumulation of 2-

hydroxyglutarate is associated with breast cancer prognosis. The Journal of clinical

investigation 124, 398-412.

144. Ulanet, D.B., Couto, K., Jha, A., Choe, S., Wang, A., Woo, H.K., Steadman, M.,

DeLaBarre, B., Gross, S., Driggers, E., et al. (2014). Mesenchymal phenotype

predisposes lung cancer cells to impaired proliferation and redox stress in response to

glutaminase inhibition. PLoS One 9, e115144.

145. van der Pluijm, G. (2011). Epithelial plasticity, cancer stem cells and bone

metastasis formation. Bone 48, 37-43.

146. Vander Heiden, M.G., Cantley, L.C., and Thompson, C.B. (2009).

Understanding the Warburg effect: the metabolic requirements of cell proliferation.

Science 324, 1029-1033. 153

147. Vaupel, P., and Mayer, A. (2007). Hypoxia in cancer: significance and impact on

clinical outcome. Cancer Metastasis Rev 26, 225-239.

148. Velletri, T., Romeo, F., Tucci, P., Peschiaroli, A., Annicchiarico-Petruzzelli, M.,

Niklison-Chirou, M.V., Amelio, I., Knight, R.A., Mak, T.W., Melino, G., and Agostini, M.

(2013). GLS2 is transcriptionally regulated by p73 and contributes to neuronal

differentiation. Cell Cycle 12, 3564-3573.

149. Venneti, S., Dunphy, M.P., Zhang, H., Pitter, K.L., Zanzonico, P., Campos, C.,

Carlin, S.D., La Rocca, G., Lyashchenko, S., Ploessl, K., et al. (2015). Glutamine-based

PET imaging facilitates enhanced metabolic evaluation of gliomas in vivo. Sci Transl

Med 7, 274ra217.

150. Vesuna, F., Lisok, A., Kimble, B., and Raman, V. (2009). Twist modulates breast

cancer stem cells by transcriptional regulation of CD24 expression. Neoplasia 11, 1318-

1328.

151. von Minckwitz, G., and Martin, M. (2012). Neoadjuvant treatments for triple-

negative breast cancer (TNBC). Annals of oncology : official journal of the European

Society for Medical Oncology / ESMO 23 Suppl 6, vi35-39.

152. Wagner, E.F., and Nebreda, A.R. (2009). Signal integration by JNK and p38

MAPK pathways in cancer development. Nature reviews. Cancer 9, 537-549.

153. Wang, D., and Lippard, S.J. (2005). Cellular processing of platinum anticancer

drugs. Nature reviews. Drug discovery 4, 307-320.

154. Wang, Y.Y., Attane, C., Milhas, D., Dirat, B., Dauvillier, S., Guerard, A.,

Gilhodes, J., Lazar, I., Alet, N., Laurent, V., et al. (2017). Mammary adipocytes

stimulate breast cancer invasion through metabolic remodeling of tumor cells. JCI

insight 2, e87489.

155. Wang, Y.Y., Lehuede, C., Laurent, V., Dirat, B., Dauvillier, S., Bochet, L., Le

Gonidec, S., Escourrou, G., Valet, P., and Muller, C. (2012). Adipose tissue and breast

epithelial cells: a dangerous dynamic duo in breast cancer. Cancer Lett 324, 142-151. 154

156. Ward, P.S., and Thompson, C.B. (2012). Metabolic reprogramming: a cancer

hallmark even warburg did not anticipate. Cancer Cell 21, 297-308.

157. Weigelt, B., Peterse, J.L., and van 't Veer, L.J. (2005). Breast cancer metastasis:

markers and models. Nature reviews. Cancer 5, 591-602.

158. Weigelt, B., and Reis, J.S. (2009). Histological and molecular types of breast

cancer: is there a unifying taxonomy? Nature Reviews Clinical Oncology 6, 718-730.

159. Wellner, U., Schubert, J., Burk, U.C., Schmalhofer, O., Zhu, F., Sonntag, A.,

Waldvogel, B., Vannier, C., Darling, D., zur Hausen, A., et al. (2009). The EMT-activator

ZEB1 promotes tumorigenicity by repressing stemness-inhibiting microRNAs. Nat Cell

Biol 11, 1487-1495.

160. Werden, S.J., Sphyris, N., Sarkar, T.R., Paranjape, A.N., LaBaff, A.M., Taube,

J.H., Hollier, B.G., Ramirez-Pena, E.Q., Soundararajan, R., den Hollander, P., et al.

(2016). Phosphorylation of serine 367 of FOXC2 by p38 regulates ZEB1 and breast

cancer metastasis, without impacting primary tumor growth. Oncogene 35, 5977-5988.

161. Wise, D.R., and Thompson, C.B. (2010). Glutamine addiction: a new therapeutic

target in cancer. Trends Biochem Sci 35, 427-433.

162. Xu, W., Wang, F., Yu, Z., and Xin, F. (2016). Epigenetics and Cellular

Metabolism. Genet Epigenet 8, 43-51.

163. Yang, M.H., Hsu, D.S., Wang, H.W., Wang, H.J., Lan, H.Y., Yang, W.H., Huang,

C.H., Kao, S.Y., Tzeng, C.H., Tai, S.K., et al. (2010). Bmi1 is essential in Twist1-

induced epithelial-mesenchymal transition. Nat Cell Biol 12, 982-992.

164. Yang, M.H., and Wu, K.J. (2008). TWIST activation by hypoxia inducible factor-1

(HIF-1): Implications in metastasis and development. Cell Cycle 7, 2090-2096.

165. Yeung, S.J., Pan, J., and Lee, M.H. (2008). Roles of p53, MYC and HIF-1 in

regulating glycolysis - the seventh hallmark of cancer. Cell Mol Life Sci 65, 3981-3999.

155

166. Yuneva, M., Zamboni, N., Oefner, P., Sachidanandam, R., and Lazebnik, Y.

(2007). Deficiency in glutamine but not glucose induces MYC-dependent apoptosis in

human cells. J Cell Biol 178, 93-105.

167. Zardavas, D., Baselga, J., and Piccart, M. (2013). Emerging targeted agents in

metastatic breast cancer. Nat Rev Clin Oncol 10, 191-210.

168. Zhang, C., Liu, J., Zhao, Y., Yue, X., Zhu, Y., Wang, X., Wu, H., Blanco, F., Li,

S., Bhanot, G., et al. (2016). Glutaminase 2 is a novel negative regulator of small

GTPase Rac1 and mediates p53 function in suppressing metastasis. Elife 5, e10727.

169. Zhang, J., Wang, C., Chen, M., Cao, J., Zhong, Y., Chen, L., Shen, H.M., and

Xia, D. (2013a). Epigenetic silencing of glutaminase 2 in human liver and colon cancers.

BMC Cancer 13, 601.

170. Zhang, L., Huang, G., Li, X., Zhang, Y., Jiang, Y., Shen, J., Liu, J., Wang, Q.,

Zhu, J., Feng, X., et al. (2013b). Hypoxia induces epithelial-mesenchymal transition via

activation of SNAI1 by hypoxia-inducible factor -1alpha in hepatocellular carcinoma.

BMC Cancer 13, 108.

171. Zhang, Q.X., Borg, A., Wolf, D.M., Oesterreich, S., and Fuqua, S.A. (1997). An

estrogen receptor mutant with strong hormone-independent activity from a metastatic

breast cancer. Cancer Res 57, 1244-1249.

156

Vita

Esmeralda Quetzalcitlali Ramirez-Peña was born in Mexico City, Mexico on August 25th, 1985, the daughter of Mrs. Esperanza Peña-Rosales and Mr. Arturo Ramirez-Cuellar. She graduated from Stephen F. Austin High School in Sugar Land, Texas in May 2003. The following August of

2003, she enrolled at the University of Houston pursuing a major in Biochemical and Biophysical

Sciences and a minor in Biology. She presented a senior honor thesis under the mentorship of

Dr. Masaya Fujita from the Department of Biology and Biochemistry and graduated with

University Honors in May 2008. For the next three years, she worked as a laboratory technician at Houston’s Methodist Hospital Research Institute under Dr. Paul Sumby. In August 2011, she began her doctoral studies at The University of Texas MD Anderson Cancer Center UTHealth

Graduate School of Biomedical Sciences, joined the Genes and Development graduate program and worked under the mentorship of Dr. Sendurai A. Mani.

157