Metabolic and Microenvironmental Determinants of Breast Metastasis: Effects of Utilization on Metastatic Phenotypes and a Predictive Metastasis Microfluidic Device

by

Megan A. Altemus

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Cancer Biology) in the University of Michigan 2019

Doctoral Committee:

Professor Sofia D. Merajver, Chair Assistant Professor Carlos A. Aguilar Professor Maria G. Castro Professor Eric R. Fearon Assistant Professor Costas A. Lyssiotis

Megan A. Altemus

[email protected]

ORCID ID: 0000-0002-8191-8838

© Megan A. Altemus 2019

DEDICATION

For my mother, Barbara Hull, a breast cancer survivor and role model. I’m thankful every day for your never-ending support and love. I began my academic journey wanting to contribute to breast cancer research so other little girls would be able to grow up with their moms just like I got to with you. Everything that I have accomplished has been for and because of you. I love you.

ii ACKNOWLEDGEMENTS

To Dr. Sofia Merajver, thank you for your support throughout the years and for the opportunity to work and study as a member of your lab. As your student I have grown both as an independent scientist and as a person. You have taught me many things that will aide in my success in my future endeavors and career. Thank you. In addition, thank you to my committee members Eric Fearon, Maria Castro, Costas

Lyssiotis, Carlos Aguilar, and past member Shuichi Takayama for all your guidance during the initiation, execution, and completion of this dissertation work.

To past and current member of the Merajver lab, especially Ryan Oliver, thank you for sharing your experience and teaching me new things every day. Thank you to

Zhi Fen Wu whose dedication to the lab and the success of its members has been invaluable to the completion of this work. Thank you to Koh Meng Aw Yong, Sara

Caceres, Laura Goo, and Hannah Cheriyan for always being there when I needed advice or a coffee break.

To the Cancer Biology Program faculty, staff, and students, for providing a well- rounded graduate experience and help with completing this degree. Thank you especially to Beth Lawlor and Zarinah Aquil whose dedication to my success and the success of all Cancer Biology students went above and beyond.

iii Thank you to my constant companion and friend for the last 14 years, Missy, my

cat, whose unconditional love and affection brightened my every day. Whenever things

seem bleak, she is always there with a fluffy hug and a purr.

And lastly, thank you to my husband, Eric Altemus. You are my one and only. I

would not have been able to do this without you. I’m excited to begin this next chapter of

my life with you by my side.

FINANCIAL ACKNOWLEDGEMENTS

I would like to acknowledge all the sources of funding that made this work

possible. First, the Cancer Biology Program training grant, an NIH T-32 Training

Fellowship (T32CA009676). Also, funds provided by the Rogel Cancer Center and

Rogel Cancer Center Nancy Newton Loeb Fund which supported my second year in the

Cancer Biology Program. Additional funding for this work and the Merajver lab was

provided by National Cancer Institute of the National Institutes of Health under award

number P30CA046592, 5T32CA009676-23, CA196018, AI116482, and the METAvivor

and Breast Cancer Research Foundation organizations.

iv TABLE OF CONTENTS

DEDICATION ii

ACKNOWLEDGEMENTS iii

LIST OF TABLES ix

LIST OF FIGURES x

ABSTRACT xii

CHAPTER

1. Introduction: Aspects of Breast Cancer Metastasis; Effects of Hypoxic

Microenvironment on Initiation and , and the Brain as the Deadliest

Secondary Site 1

Hypoxia and metastasis 2

Metabolic reprogramming under hypoxia 4

Glycogen metabolism 5

Hypoxic regulation of glycogen in cancer 6

Distant site-specific metabolic alterations in cancer 7

Breast cancer brain metastasis 9

Current models of brain metastasis 10

v Blood-brain barrier microfluidic devices 12

CONCLUSION 14

FIGURES 15

REFERENCES 16

2. Breast Utilize Hypoxic Glycogen Stores via PYGB, the Brain

Isoform of , to Promote Metastatic Phenotypes 22

SUMMARY 22

INTRODUCTION 23

MATERIALS AND METHODS 27

Cell culture and media 27

Glycogen assay 28

Periodic acid-Schiff staining 29

RT-qPCR 30

shRNA knockdown 30

Western blotting 30

Proliferation assay 31

Wound-healing assay 31

Transwell invasion assay 32

RESULTS 33

DISCUSSION AND CONCLUSION 38

vi FIGURES 42

TABLES 51

REFERENCES 53

3. A Platform for Artificial Intelligence Based Identification of the

Extravasation Potential of Cancer Cells into the Brain Metastatic Niche 56

SUMMARY 56

INTRODUCTION 57

RESULTS AND DISCUSSION 59

Brain seeking breast cancer cell line reveals a distinct μBBN phenotypic pattern 59

PDX-derived brain metastatic and primary tumor cells display differential

phenotypic behaviors 62

Brain metastatic cancer cells degrade the endothelial barrier 65

Comprehensive differential cancer cell behavior in vitro leads to an index of

brain metastatic potential 66

MATERIALS AND METHODS 69

Study design 69

μm-Blood Brain Niche design and validation 70

Small molecule membrane transfer measurements 71

Cell culture and reagents 72

Patient-derived xenografts 73

vii Live subject statement 73

Seeding microfluidic device 74

Measurement of the cell attributes using confocal tomography 74

Statistical analysis 75

Artificial intelligence machine learning algorithm 75

Breast (cancer) cell lines 76

CONCLUSION 76

FIGURES 77

TABLES 84

REFERNCES 88

4. Conclusions and Future Directions 91

REFERENCES 98

viii LIST OF TABLES

TABLE

2.1 Primers for qPCR of glycogen . 51

2.2 shRNA oligonucleotides for shPYGL and shPYGB. 51

2.3 Glycogen phosphorylase knockdown decreases doubling time of MCF-7

and MCF-10A cells 52

3.1 Summary of metrics measured for each cell line. 84

3.2 Cell counts and volumes by location. 84

3.3 Summary of metrics measure for each PDX type. 85

3.4 PDX type counts and volumes by location. 85

3.5 Comparison of methods to classify cancer cells by brain met potential. 86

3.6 Confusion matrix for random forest. 86

3.7 Comparison of methods to classify breast PDX cancer cells by brain met

potential. 87

3.8 Confusion matrix for random forest using PDX cancer cells. 87

ix LIST OF FIGURES

FIGURE

1.1 Graphical depiction of the steps in the metastatic cascade 15

2.1 Glycogen accumulates in breast cancer cells under hypoxic conditions across subtypes. 42

2.2 Glycogen accumulates in additional breast cancer cells under hypoxic conditions. 43

2.3 Glycogen pathway expression changes in breast cancer cells exposed to hypoxia. 44

2.4 Additional glycogen pathway gene expression changes in breast cancer cells exposed to hypoxia. 45

2.5 Glycogen phosphorylase brain isoform knockdown inhibits glycogen utilization in breast cancer cells. 46

2.6 shPYGL and shPYGB reduces glycogen phosphorylase mRNA expression. 47

2.7 Glycogen phosphorylase knockdown in SUM-149 cells. 48

2.8 Loss of glycogen phosphorylase inhibits proliferation in MCF-7 and normal-like MCF-10A cells but not MDA-MB-231. 49

2.9 Loss of glycogen phosphorylase brain isoform inhibits wound-closure in MCF-7 and invasion in MDA-MB-231 cells. 50

3.1 Overview of method. 77

3.2 Microfluidic BBNiche device design to study brain metastatic process. 78

3.3 Differences in extravasation and morphology of brain-seeking cells compared to non-brain-seeking cell in the µBBN device analyzed using confocal tomography. 79

3.4 Profiling of patient derived xenografts in µBBN device. 80

3.5 Cancer cell interaction with the μBBN endothelium 81

x 3.6 Accurate identification of brain metastatic potential in μBBN device. 82

3.7 Mask of four-channel device 83

3.8 Optimizing media composition for co-culture 83

xi ABSTRACT

Breast cancer has the highest incidence rates of all cancer types among women in the United States, and the second highest mortality. While survival is high for localized disease, breast cancer that has metastasized has a five-year survival rate of roughly 30%. This highlights the need for a better understanding of the breast cancer metastatic process, both at the primary and distant sites such as the brain, in order to develop new treatments and preventative measures. The work presented here focuses on two different avenues for combatting breast cancer metastasis. The first of these is exploiting metabolic vulnerabilities of the cancer cells at the primary site in order to prevent invasion and metastasis. The second fills the pressing need for quick and biologically accurate models of distant sites in order study secondary site-specific metastatic processes of cancer cells and patient samples. Therefore, in this dissertation work we aim to 1) determine the effects hypoxic glycogen utilization has on metastatic phenotypes of breast cancer and 2) develop a microfluidic device that accurately mimics the interactions of cancer cells and the blood-brain niche.

We found that many different breast cancer cell lines increased stores of glycogen, the main storage molecule in the body, in response to hypoxia, such as a growing solid tumor might encounter. This extent of glycogen accumulation in response to hypoxia did not seem to correlate with breast cancer receptor status and

xii differences in glycogen pathway gene expression under hypoxic exposure were not conserved across cell lines. Additionally, using shRNA knockdowns of both the

(PYGL) and brain (PYGB) isoforms of the glycogen degradation glycogen phosphorylase, we were able to determine that the ability to utilize glycogen was mainly controlled by PYGB rather than PYGL in breast cancer cell lines and directly influences the rate of migration and invasion.

For the second aim, we created and validated a brain niche mimetic microfluidic device that in conjunction with machine learning algorithms can accurately predict the brain metastatic potential. The device consists of an upper vasculature-like chamber lined by endothelial cells and a lower brain stroma-like chamber filled with astrocytes suspended in a collagen gel separated by a porous membrane. Utilizing both breast cancer cell lines and samples from patient-derived xenografts in the device, we measured four specific phenotypes of the cancer cells and their interaction with the endothelium; percent volume extravasated, distance extravasated, sphericity, and overall volume. These measurements were then utilized to train a machine learning algorithm to accurately predict brain metastatic potential of the breast cell lines and patient-derived cells.

Overall, this body of work adds to our understanding of how altered glycogen metabolism may influence breast cancer metastasis and presents a new tool for studying and predicting metastasis to the brain.

xiii CHAPTER 1

Introduction: Aspects of Breast Cancer Metastasis; Effects of Hypoxic

Microenvironment on Initiation and Metabolism, and the Brain as the Deadliest

Secondary Site

Currently in the US, breast cancer has the highest incidence rates of all cancer

types among women, and the second leading cause of cancer-related mortality behind

lung cancer1. Breast cancers are divided into distinct subtypes based on receptor status and HER-2 expression that significantly impact treatment options and

prognosis. Hormone receptor positive breast cancer (expressing estrogen (ER) and/or

progesterone (PR) receptors) account for approximately 80% of all breast cancers and

typically have the best prognosis2,3. Around 20% of breast cancers are HER-2 positive,

over-expressing human epidermal growth factor 2, (which can be ER/PR+ or ER/PR-),

and have had historically poor prognoses; however, outlook for these patients has

improved greatly due to the development of anti-HER-2 therapies2–4. In contrast, triple-

negative breast cancers (TNBC) account for 10-20% of breast cancer diagnoses, are

ER/PR- and HER-2-, with few targeted therapies currently approved to treat these

patients, even within narrow subgroups2–4. Recently, poly(-

ribose) polymerase (PARP) inhibitors have been approved for metastatic TNBC in

patients with the BRCA gene mutation, and anti-programmed cell death-ligand 1 (PD-

L1) immunotherapies are likely to be approved for TNBC in which the tumor cells

1 express PD-L15,6. Around 34% of patients with TNBC present with disease recurrence at a distant site with median time to recurrence of 2.6 years, compared to a 20%

recurrence rate and 5 year median time to recurrence for patients with other subtypes,

making TNBC one of the most metastasis-prone subtypes of breast cancer, second only to inflammatory breast cancer7–9.

For cancer cells to form macro-metastases at a site distant from the primary

tumor, certain steps must occur known as the ‘metastatic cascade’ (Figure 1.1)10. First,

tumor cells must invade outwards from the primary tumor into the surrounding tissue;

these cells then must traverse the basement membrane of the tissue and intravasate

through the endothelium via gaps or imperfections in the vasculature, and into the

surrounding blood vessels. Once in the bloodstream, the tumor cells must survive until

they adhere at a distant site, where they again traverse the vascular endothelium in the

opposite direction into the stroma of the distant organ. The tumor cell must then adapt

to its new microenvironment and proliferate to form the metastatic lesion. Each of these

steps serves as a kind of sieve whereby many tumor cells attempt the process, but few

survive. Yet, despite these slim odds, the rate of metastasis in certain cancers, such as

TNBC, remains dangerously high. In order to combat these highly metastatic diseases,

we must better understand the adaptations of tumor cells that allow for survival through

each of these stages.

Hypoxia and metastasis

It has been observed that more than half of solid tumors, present with hypoxic or

anoxic areas compared to the local normal tissue11. Tumor hypoxia has been

2 associated with metastasis and poor prognosis in many cancers, including breast11,12.

Hypoxia develops in primary tumors due to multiple factors including increased distance from the blood supply and an imbalance between oxygen delivered to the tumor site and consumption by the cancer cells and tumor associated cells11. Tumor cells respond

to this hypoxic stimulus by activating signaling pathways leading to distinct gene

expression changes, the most notable being stabilization of the hypoxia-inducible

factors, HIF-1α and HIF-2α13. Under normoxic conditions, HIF-α subunits are

hydroxylated by the prolyl hydroxylase (PHD 1-3) that utilize oxygen as a

substrate14,15. While hydroxylated, the von Hippel-Lindau tumor suppressor (VHL) binds

HIF-α proteins and targets them for degradation by the E3 ubiquitin ligase complex16–18.

Under hypoxia, PHD enzymes lack oxygen as a and thus the hydroxylation event does not occur, preventing HIF-α degradation and promoting its stabilization, enabling it to translocate to the nucleus. HIF-α subunits dimerize and bind hypoxia- responsive elements (HREs) in the genome to activate transcription of multiple genes19–

21.

Hypoxia and HIF-1α activate transcriptional programs that induce the epithelial to

mesenchymal transition (EMT) and support metastasis in many cancers22–25. Hypoxia-

induced upregulated transcription factors and signaling pathways include Twist, Snail,

ZEB1, Notch, TGF-B, and Hedgehog signaling to name a few24,26–32. HIF-1α and

hypoxia have been shown to increase the metastatic phenotype in multiple cancer cell

types, including breast cancer. Some notable studies have shown that blocking HIF-1α

and/or HIF-2α in MDA-MB-231 breast cancer cells decreased formation and total

metastatic burden of lung metastases in mice after orthotopic xenograft, and the size of

3 tumors and blood vessel density in bone metastasis after intra-cardiac injection25,33.

Lung metastases were also decreased in the MMTV-PyMT breast cancer genetic

mouse model by a mammary epithelial knockout of HIF-1α. Increased HIF-1α levels

have been linked with increased risk of metastasis and mortality in many breast cancer

patient cohorts34.

Metabolic reprogramming under hypoxia

Hypoxia and subsequent stabilization of HIF-1α also induces many metabolic

changes in cancer cells, in part due to the ubiquity of oxygen as an electron acceptor in

many biosynthetic processes35. These include increased expression of glucose

transporters and other genes involved in glycolysis36–38, increased pyruvate

dehydrogenase kinase activity thus decreasing the amount of pyruvate that enters the

TCA cycle and decreasing oxidative phosphorylation39–41, and altered fatty-acid and lipid metabolism42,43. These changes in metabolic factors induced by hypoxia and HIF-1α

stabilization shift the cancer cells from utilization of the TCA cycle and oxidative

for their energy needs towards increased , very similar to the

“Warburg effect”44. In addition to these well-studied changes, glycogen metabolism has also been found to be altered under hypoxic conditions45, however the exact effects this

altered glycogen metabolism may have on the vulnerabilities of cancer cells has yet to

be determined.

4 Glycogen metabolism

Glycogen is a large branching polymer of glucose molecules and is the main glucose storage molecule in the body46. It is primarily stored in the liver where it is

utilized to maintain blood-glucose levels during periods of fasting (such as sleep) and in

the muscles, where it may be quickly mobilized for energy during exercise. Glycogen is

synthesized around a glycogenin core by addition of UDP-glucose onto growing glycogen chains. Glucose-1-phosphate available in the cell from either glucose transported into the cell or through glucose production via gluconeogenic substrates is catalyzed to UDP-glucose by UDP-glucose pyrophosphorylase-2 (UGP2). UDP-glucose is added onto glycogen via an α-1,4 linkage by glycogen synthase, the rate limiting enzyme in glycogen synthesis46. There are two isoforms of glycogen synthase, GYS1

which is mainly expressed in the liver, and the muscle isoform GYS2. After the chain

has reached at least 11 residues, glycogen branching enzyme (GBE) can move

approximately 6 residues from the growing chain and add to an interior glucose residue

via an α-1,6 linkage to form a branch46. During degradation, glucose-1-phosphate

molecules are removed from the non-reducing end of the glycogen molecule by

glycogen phosphorylase46. Glycogen phosphorylase has three different isoforms that

are typically expressed in different tissues, liver (PYGL), isoform (PYGM), and brain

isoforms (PYGB). Glycogen de-branching enzyme, also key to glycogen degradation,

has two functions, moving residues immediately before the α-1,6 linkage to another branch of glycogen, exposing the α-1,6 bond and cleaving it46. Free glucose-1-

phosphate molecules are then catalyzed to glucose-6-phosphate, the first intermediate

step in glycolysis, by phosphoglucomutase-1 (PGM1).

5 There are also enzymes that regulate glycogen synthase and glycogen

phosphorylase, the key rate-limiting enzymes of glycogen metabolism. Glycogen

synthase activity is regulated by allosteric effects and by phosphorylation events.

Phosphorylation of glycogen synthase by glycogen synthase kinase 3α and 3β

(GSK3α/β) at multiple serine residues inhibits its activity47. In turn, GSK3α/β is regulated

itself by protein kinase-A (PKA) and Akt, a major link between metabolism and cell

signaling.47 Glycogen phosphorylase is activated by phosphorylation at Ser-14 by

phosphorylation kinase46. Both glycogen synthase and glycogen phosphorylase are

regulated by protein phosphatase-1 (PP1). PP1 dephosphorylation activates glycogen

synthase, yet inhibits glycogen phosphorylase, leading to reciprocal regulation of

glycogen synthesis and degradation that results in a tight regulation of glucose

homeostasis46.

Hypoxic regulation of glycogen in cancer

High levels of glycogen have been found in many different cancer cell lines

including breast cancers. Glycogen levels inversely correlated with proliferation rate,

suggesting that glycogen may be utilized as a glucose source to sustain proliferation

under certain conditions, such as hypoxia45. It has been found that hypoxia induces glycogen metabolism in cancer in a similar manner as non-cancerous cells48–51. High

levels of glycogen have also been found in the hypoxic cores of some tumors and in

tumors treated with anti-angiogenic therapy52, again supporting the potential dynamic

regulation of glycogen metabolism as a way to store and access glucose. Hypoxia and

stabilization of HIF-1α have been shown to increase levels of many glycogen enzymes

6 and regulatory proteins that may contribute to the observed increase in glycogen,

including UGP2, GYS1, GBE, and PPP1R3C, the glycogen-associated regulatory

subunit 3C of protein phosphatase 151,53–55.

Through work described in later chapters, we investigated the link between

hypoxic glycogen storage and metastatic phenotypes in breast cancer. In order to fuel

migration and invasion, a necessary part of the metastatic cascade that is induced by

hypoxia, cancer cells require a source of glucose and energy. We aim to answer the

question; could this source of energy be glycogen, the accumulation of which is also

induced by hypoxia? If so, regulation of glycogen metabolism could prove a novel

strategy to prevent the initiation or control the progression of metastasis in highly

aggressive breast cancers, such as TNBC.

Distant site-specific metabolic alterations in cancer

In addition to potentially playing a role in metastasis initiation, metabolic alterations can support growth at a distant sites which present different metabolic challenges than the tumor site of origin56. The common locations of breast cancer

metastasis; lung, liver, bone, and brain, all have different levels of available nutrients,

oxygen, and oxidative stress that the cancer cells must adapt to. The lung is a unique

site that while having very high levels of oxygenation compared to other locations in the

body, also has high levels of reactive oxygen species due to the processes by which the

lungs remove toxic chemicals from the body57,58. Cancer cells, particularly breast, have

developed certain mechanisms to deal with this oxidative stress including upregulation

of PGC-1α, a transcriptional co-activator which promotes metabolic flexibility and

7 expression of some antioxidant genes such as glutathione peroxidase59–61, and upregulation of pyruvate carboxylase, similar to what is seen in non-small cell lung cancers to support the replenishment of TCA cycle intermediates62–65.

The liver presents with a very different environment. As discussed previously, the

liver is responsible for maintaining blood-glucose levels in the body as well as many

other important metabolic functions. As such, the liver has areas of differing oxygen

availability, presenting cancer cells with a hypoxic microenvironment66. Liver metastatic

breast cancer cells have been shown to increase glycolysis and reduce oxidative-

phosphorylation in part due to HIF-1α mediated expression of pyruvate dehydrogenase

kinase-1 (PDK-1)67. Additionally, this expression of PDK-1 was found to be essential to

breast cancer liver metastasis formation in mouse models67.

Breast cancer cells also alter their metabolism to support growth in the bone.

Breast cancer bone metastases are mostly osteolytic and thus have adapted to support

osteoclasts, cells involved in bone resorption. These adaptation included L-serine

synthesis, a necessary factor for bone marrow precursor differentiation to osteoclasts,

and increased lactate production, where lactate is a main fuel source for osteoclast

function68,69. This mobilization of osteoclasts and thus osteolysis, frees up space and nutrients for the colonization of bone microenvironment by metastatic breast cancer cells.

And lastly, the brain presents a unique metabolic challenge to colonizing breast cancer cells. The brain consumes more glucose than any other organ in the body and thus is a very glucose-scarce secondary site. Breast cancer cells have therefore developed adaptations to rely on non-glucose sources of energy, such as upregulating

8 acetyl-CoA synthetase enzyme 2 to convert acetate to acetyl-CoA to fuel the TCA

cycle70. Additionally breast cancer brain metastases have been shown to utilize

branched-chain amino acids and glutamate as glucose alternatives as well as

supporting gluconeogenesis via upregulation of fructose-1,6-bisphostase 271. These

differences in cancer cell metabolism based on secondary site location reveal

interesting potential therapeutic targets that may function to prevent seeding and

colonization by breast cancer metastases.

Breast cancer brain metastasis

Different subtypes of breast cancer have differing propensities to metastasize to

the various common secondary sites, including the brain. ER+ breast cancers have a 5-

10% rate of metastasis to the brain whereas TNBC and HER-2+ breast cancers have brain metastases rates upwards of 20%7,72. Survival time after diagnosis of a metastatic

lesion to the brain is extremely short, with a median overall survival of less than 11.5

months for HER-2+ and a meager 4.9 months for TNBC, making the diagnosis of brain

metastases the most devastating complication, from the survival standpoint, of breast

cancer73. The high rate of brain metastasis and longer median survival time in HER-2

positive breast cancers is hypothesized to be a result of increased use of the targeted

therapy trastuzumab74,75. Trastuzumab has minimal ability to cross the blood-brain

barrier but it prevents metastases at other distant sites, therefore effectively increasing

the rate at which we see the brain as the first site of distant metastases in these

patients.

9 The brain as a site of metastasis is especially unique. Within the brain, endothelial cells, astrocytes, and pericytes form a selective barrier to protect brain stroma known as the blood-brain barrier76. This barrier is stronger and more selective than the endothelial barrier present in other locations of the body due to tight junctions between the endothelial cells within the brain that are supported by astrocytes and pericytes. After the tumor cells have extravasated through the blood-brain barrier, these tight junctions serve to protect them from immune surveillance and drugs, such as chemotherapy77. While in the perivascular space, tumor cells remain in close proximity to the vasculature where they elongate and spread along the existing blood vessels, a process known as vessel cooption78–80.

Once a patient has developed brain metastasis, treatment involves surgical resection and/or radiation and is often only palliative81. Given the poor prognosis despite these therapies to treat macrometastases, the best avenue for future treatment of breast cancer brain metastases is to develop tests for early detection and, better still, preventative therapies. In order to develop more effective tests and treatments, we need accurate and reproducible model systems to study brain metastases.

Current models of brain metastasis

In order to study the process of cancer metastasis to the brain, models that mimic the human blood-brain barrier must be used. Models currently utilized include both murine in vivo models and various in vitro models.

In vivo brain metastasis models involve tumor cells that migrate and colonize the brain niche within the mouse, successfully developing solid tumors. These models can

10 be divided into three groups, syngeneic mouse models, genetically engineered mouse models, and human-mouse xenograft models.

Syngeneic mouse models are injected orthotopically or ectopically (into the same organ as a tumor would originate such as the mammary fat pad for breast cancer, or directly into the blood stream, respectively) with murine cancer cell lines. Genetically engineered mouse models have been developed that have inactivated tumor suppressor genes or express oncogenes and spontaneously develop metastases.

Genetically engineered mouse models of brain metastasis have been developed for certain lung cancers and melanoma but not for breast cancers82,83. Human-mouse xenograft models involve immune-compromised mice injected orthotopically or ectopically with human cancer cell lines or implanted with cells from the tumor of a patient, known as patient-derived xenografts (PDXs). These are the mouse models most commonly used to study breast cancer brain metastasis. Several breast cancer cell lines, including MDA-MB-231, MDA-MB-435, 4T1, and MCF-7, have been shown to extravasate through the blood-brain barrier in these mouse models80. Also, the brain- seeking cell line, MDA-MB-231BR, was developed through repeated re-injection of the

MDA-MB-231 TNBC cells that have metastasized to the brain through intracardiac injection in mice, and has been used in many brain metastases studies since84.

However, mouse models, while similar to the human brain, cannot capture the same complexity as the human tissues especially in reference to the cells present in the perivascular space of the brain, such as astrocytes which vary greatly from their murine counterparts85. Moreover, it is not possible with present technologies to investigate the

11 trajectory and biology of individual cancer cells as they traverse the blood-brain barrier

in murine models.

The shortcomings of murine models have spurred the development of in vitro

models of brain metastasis. These models have the benefit of being faster and more

cost-effective than mouse models and may use human-derived cells to mimic blood- brain barrier, however the models currently used do not have the same micro- environment complexity as murine models. The in vitro models used today vary greatly

in their components. The simplest consists of a monolayer of endothelial cells cultured

on a dish or on the apical side of a transwell insert86. Various layers of complexity have

been added to this basic structure including using astrocyte-conditioned medium, co-

culturing astrocytes and endothelial cells on either side of a membrane, and a layer of

endothelial cells on top of a collagen gel through which cancer cells can in vade and

grow87,88.

Both in vivo and in vitro models currently in use have major drawbacks. Mouse

models are costly, difficult and slow to reproduce, and the tumor microenvironment

cannot be monitored in real-time, whereas traditional in vitro models lack the complexity

to accurately represent the blood-brain niche microenvironment. The emergence of ever

more advanced microfluidic devices attempts to bridge this gap between complexity and

ease-of-use.

Blood-brain barrier microfluidic devices

Many microfluidic devices have been designed to mimic the blood-brain barrier.

These devices have mostly been developed for drug permeability studies and

12 incorporate various aspects of blood-brain niche biology. Some of these devices consist of endothelium exposed to a constant media flow to mimic the shear stress that physiologically occurs within brain capillaries. This also has the effect of increasing tight-junction and barrier formation that cannot be achieved in static in vitro models89,90.

Increased barrier function in microfluidic devices can also be achieved by co-culturing

endothelial cells with astrocytes or exposing the endothelium to astrocyte-conditioned

medium91,92. Some of the most advanced devices have been designed to screen for the effects of drugs on neurodegenerative diseases and neurobiology and consist of multiple brain cell types in a 3D structure including endothelium, astrocytes, pericytes, microglia, and neurons93,94. In terms microfluidic devices designed to study brain metastases, most have been designed to study drug permeability through the endothelium and the subsequent effect of the drugs on tumor cells95,96. These devices

have already established tumor cells on the basolateral side of the endothelium and

thus cannot be used to study the initiation events of brain metastases in order to

understand the changes the cancer cells may experience as they negotiate the

endothelial barrier

As described in later chapters, we have developed and validated a novel

microfluidic device to specifically study the initiation of brain metastases. This device

consists of two chambers; a long upper chamber lined with endothelium to mimic the brain vasculature and provide ample area for cancer cells to adhere, and a lower chamber filled with an astrocyte-containing collagen gel to mimic the brain stroma, support tight-junction and barrier formation, and allow room for the cancer cells to extravasate and grow in a 3D environment. This device fills the current need for an

13 easily visualized, reproducible, and biologically relevant model to study the formation of

brain metastases. In our work presented in detail in Chapter 3, this device was coupled

with high resolution confocal imaging and machine learning to enable quantitative

studies to be carried out to begin to predict the brain metastatic potential of individual

cells.

Conclusion

In many cancers, including breast, metastasis to a distant site results in a dramatically poorer prognosis compared to locoregional disease. Aggressive breast cancers such as TNBC have a high rate of recurrence at a distant site after a period of being disease-free following treatment for an early stage (I or II) tumor, highlighting a need for the development of metastasis-preventative therapies. In order to develop these therapies, the underlying biology of what drives and supports the initiation of metastases as well as what allows the cancer cells to infiltrate and survive at the distant site must be understood. In this dissertation work we discovered that hypoxic glycogen accumulation may support invasion and migration in breast cancer cells, phenotypes required for the initial steps of the metastatic cascade. We also developed a microfluidic device that mimics the blood-brain niche and can accurately distinguish between cancer cells that have the propensity for brain metastasis and those that do not. The results of these studies may lead to the future development of brain metastasis predictive clinical tests and targeted therapies to prevent metastases in aggressive breast cancer.

14 Figures

Figure 1.1 Graphical depiction of the steps in the metastatic cascade The process of progression from the initial tumor at the primary site to the development of metastases at a distant site is known as the metastatic cascade. In this work, Chapter II will focus on the contributions of hypoxia and glycogen metabolism to the initial steps of the metastatic cascade; invasion and migration. The ending steps of adherence, extravasation, and colonization will be explored in Chapter III where we describe the development of a microfluidic device to study the initiation of brain metastases.

15 References

1. American Cancer Society. Cancer Facts & Figures 2018. Am. Cancer Soc. (2018). doi:10.1182/blood-2015-12-687814 2. Howlader, N. et al. US incidence of breast cancer subtypes defined by joint hormone receptor and HER2 status. J. Natl. Cancer Inst. (2014). doi:10.1093/jnci/dju055 3. Press, D. J., Miller, M. E., Liederbach, E., Yao, K. & Huo, D. De novo metastasis in breast cancer: occurrence and overall survival stratified by molecular subtype. Clin. Exp. Metastasis (2017). doi:10.1007/s10585-017-9871-9 4. Goldhirsch, A., Wood, W. C., Coates, A. S., Gelber, R. D. & Thu, B. Strategies for subtypes — dealing with the diversity of breast cancer : highlights of the St Gallen International Expert Consensus on the Primary Therapy of Early Breast Cancer 2011. Ann. Oncol. (2011). doi:10.1093/annonc/mdr304 5. Robson, M. et al. Olaparib for Metastatic Breast Cancer in Patients with a Germline BRCA Mutation . N. Engl. J. Med. (2017). doi:10.1056/nejmoa1706450 6. Schmid, P. et al. Atezolizumab and Nab-Paclitaxel in Advanced Triple-Negative Breast Cancer. N. Engl. J. Med. 379, 2108–2121 (2018). 7. Kennecke, H. et al. Metastatic behavior of breast cancer subtypes. J. Clin. Oncol. 28, 3271–3277 (2010). 8. Dent, R. et al. Triple-negative breast cancer: Clinical features and patterns of recurrence. Clin. Cancer Res. (2007). doi:10.1158/1078-0432.CCR-06-3045 9. Yang, W. T. et al. Inflammatory breast cancer: PET/CT, MRI, mammography, and sonography findings. Breast Cancer Research and Treatment (2008). doi:10.1007/s10549-007-9671-z 10. Talmadge, J. E. & Fidler, I. J. The biology of cancer metastasis: historical perspective. Cancer Res. (2010). doi:10.1158/0008-5472.CAN-10-1040 11. Vaupel, P. & Mayer, A. Hypoxia in cancer: Significance and impact on clinical outcome. Cancer and Metastasis Reviews (2007). doi:10.1007/s10555-007-9055- 1 12. Vaupel, P., Mayer, A. & Höckel, M. Tumor Hypoxia and Malignant Progression. Methods Enzymol. (2004). doi:10.1016/S0076-6879(04)81023-1 13. Semenza, G. L. Hypoxia-inducible factors in physiology and medicine. Cell (2012). doi:10.1016/j.cell.2012.01.021 14. Bruick, R. K. & McKnight, S. L. A conserved family of prolyl-4-hydroxylases that modify HIF. Science (80-. ). (2001). doi:10.1126/science.1066373 15. Epstein, A. C. R. et al. C. elegans EGL-9 and mammalian homologs define a family of dioxygenases that regulate HIF by prolyl hydroxylation. Cell (2001). doi:10.1016/S0092-8674(01)00507-4 16. Jaakkola, P. et al. Targeting of HIF-α to the von Hippel-Lindau ubiquitylation complex by O2-regulated prolyl hydroxylation. Science (80-. ). (2001). doi:10.1126/science.1059796 17. Maxwell, P. H. et al. The tumour suppressor protein VHL targets hypoxia- inducible factors for oxygen-dependent proteolysis. Nature (1999).

16 doi:10.1038/20459 18. Ivan, M. et al. HIFα targeted for VHL-mediated destruction by proline hydroxylation: Implications for O2sensing. Science (80-. ). (2001). doi:10.1126/science.1059817 19. Wang, G. L., Jiang, B. H., Rue, E. A. & Semenza, G. L. Hypoxia-inducible factor 1 is a basic-helix-loop-helix-PAS heterodimer regulated by cellular O2 tension. Proc. Natl. Acad. Sci. (1995). doi:10.1073/pnas.92.12.5510 20. Wang, G. L. & Semenza, G. L. General involvement of hypoxia-inducible factor 1 in transcriptional response to hypoxia. Genetics (1993). doi:10.1073/pnas.90.9.4304 21. Schödel, J. et al. High-resolution genome-wide mapping of HIF-binding sites by ChIP-seq. Blood (2011). doi:10.1182/blood-2010-10-314427 22. Chan, D. A. & Giaccia, A. J. Hypoxia, gene expression, and metastasis. Cancer and Metastasis Reviews (2007). doi:10.1007/s10555-007-9063-1 23. Mak, P. et al. ERβ Impedes Prostate Cancer EMT by Destabilizing HIF-1α and Inhibiting VEGF-Mediated Snail Nuclear Localization: Implications for Gleason Grading. Cancer Cell (2010). doi:10.1016/j.ccr.2010.02.030 24. Yang, M.-H. et al. Direct regulation of TWIST by HIF-1alpha promotes metastasis. Nat. Cell Biol. (2008). doi:10.1038/ncb1691 25. Hiraga, T., Kizaka-Kondoh, S., Hirota, K., Hiraoka, M. & Yoneda, T. Hypoxia and hypoxia-inducible factor-1 expression enhance osteolytic bone metastases of breast cancer. Cancer Res. (2007). doi:10.1158/0008-5472.CAN-06-2355 26. Luo, D., Wang, J., Li, J. & Post, M. Mouse Snail Is a Target Gene for HIF. Mol. Cancer Res. (2011). doi:10.1158/1541-7786.MCR-10-0214 27. Zhang, W. et al. HIF-1α promotes epithelial-mesenchymal transition and metastasis through direct regulation of ZEB1 in . PLoS One (2015). doi:10.1371/journal.pone.0129603 28. Copple, B. L. Hypoxia stimulates hepatocyte epithelial to mesenchymal transition by hypoxia-inducible factor and transforming growth factor-beta-dependent mechanisms. Liver Int. (2010). doi:10.1111/j.1478-3231.2010.02205.x 29. Chen, J., Imanaka, N., Chen, J. & Griffin, J. D. Hypoxia potentiates Notch signaling in breast cancer leading to decreased E-cadherin expression and increased cell migration and invasion. Br. J. Cancer (2010). doi:10.1038/sj.bjc.6605486 30. Sahlgren, C., Gustafsson, M. V., Jin, S., Poellinger, L. & Lendahl, U. Notch signaling mediates hypoxia-induced tumor cell migration and invasion. Proc. Natl. Acad. Sci. (2008). doi:10.1073/pnas.0802047105 31. Spivak-Kroizman, T. R. et al. Hypoxia triggers hedgehog-mediated tumor-stromal interactions in pancreatic cancer. Cancer Res. (2013). doi:10.1158/0008- 5472.CAN-11-1433 32. Lei, J. et al. Hedgehog signaling regulates hypoxia induced epithelial to mesenchymal transition and invasion in pancreatic cancer cells via a ligand- independent manner. Mol. Cancer (2013). doi:10.1186/1476-4598-12-66 33. Zhang, H. et al. HIF-1-dependent expression of angiopoietin-like 4 and L1CAM mediates vascular metastasis of hypoxic breast cancer cells to the lungs. Oncogene (2012). doi:10.1038/onc.2011.365

17 34. Semenza, G. L. Defining the role of hypoxia-inducible factor 1 in cancer biology and therapeutics. Oncogene (2010). doi:10.1038/onc.2009.441 35. Pavlova, N. N. & Thompson, C. B. The Emerging Hallmarks of Cancer Metabolism. Cell Metabolism (2016). doi:10.1016/j.cmet.2015.12.006 36. Semenza, G. L. HIF-1: upstream and downstream of cancer metabolism. Current Opinion in Genetics and Development (2010). doi:10.1016/j.gde.2009.10.009 37. Ebert, B. L., Firth, J. D. & Ratcliffe, P. J. Hypoxia and mitochondrial inhibitors regulate expression of glucose transporter-1 via distinct cis-acting sequences. J. Biol. Chem. (1995). doi:10.1074/jbc.270.49.29083 38. Denko, N. C. Hypoxia, HIF1 and glucose metabolism in the solid tumour. Nature Reviews Cancer (2008). doi:10.1038/nrc2468 39. Papandreou, I., Cairns, R. A., Fontana, L., Lim, A. L. & Denko, N. C. HIF-1 mediates adaptation to hypoxia by actively downregulating mitochondrial oxygen consumption. Cell Metab. (2006). doi:10.1016/j.cmet.2006.01.012 40. Kim, J. W., Tchernyshyov, I., Semenza, G. L. & Dang, C. V. HIF-1-mediated expression of pyruvate dehydrogenase kinase: A metabolic switch required for cellular adaptation to hypoxia. Cell Metab. (2006). doi:10.1016/j.cmet.2006.02.002 41. Lu, C. W., Lin, S. C., Chen, K. F., Lai, Y. Y. & Tsai, S. J. Induction of pyruvate dehydrogenase kinase-3 by hypoxia-inducible factor-1 promotes metabolic switch and drug resistance. J. Biol. Chem. (2008). doi:10.1074/jbc.M803508200 42. Metallo, C. M. et al. Reductive glutamine metabolism by IDH1 mediates lipogenesis under hypoxia. Nature (2012). doi:10.1038/nature10602 43. Wise, D. R. et al. Hypoxia promotes isocitrate dehydrogenase-dependent carboxylation of -ketoglutarate to citrate to support cell growth and viability. Proc. Natl. Acad. Sci. (2011). doi:10.1073/pnas.1117773108 44. Heiden, M. G. Vander, Cantley, L. C. & Thompson, C. B. Understanding the warburg effect: The metabolic requirements of cell proliferation. Science (2009). doi:10.1126/science.1160809 45. Rousset, M., Fogh, J. & Zweibaum, A. Presence of glycogen and growth–related variations in 58 cultured huma tumor cell lines of various tissue origins. Cancer Res. (1981). 46. Berg, Jeremy, M., Tymoczko, J. L., Gatto, G. J. & Stryer, L. Biochemistry 8th Edition. Biochemistry (2015). doi:10.1007/978-3-8274-2989-6 47. Beurel, E., Grieco, S. F. & Jope, R. S. Glycogen synthase kinase-3 (GSK3): Regulation, actions, and diseases. Pharmacology and Therapeutics (2015). doi:10.1016/j.pharmthera.2014.11.016 48. Saez, I. et al. Neurons have an active glycogen metabolism that contributes to tolerance to hypoxia. J. Cereb. Blood Flow Metab. (2014). doi:10.1038/jcbfm.2014.33 49. Vigoda, A., Mamedova, L. K., Shneyvays, V., Katz, A. & Shainberg, A. Glycogen metabolism in rat heart muscle cultures after hypoxia. Mol. Cell. Biochem. (2003). doi:10.1023/A:1027302021428 50. Zois, C. E., Favaro, E. & Harris, A. L. Glycogen metabolism in cancer. Biochemical Pharmacology (2014). doi:10.1016/j.bcp.2014.09.001 51. Pelletier, J. et al. Glycogen Synthesis is Induced in Hypoxia by the Hypoxia- Inducible Factor and Promotes Cancer Cell Survival. Front Oncol (2012).

18 doi:10.3389/fonc.2012.00018 52. Favaro, E. et al. Glucose utilization via glycogen phosphorylase sustains proliferation and prevents premature senescence in cancer cells. Cell Metab. (2012). doi:10.1016/j.cmet.2012.10.017 53. Mole, D. R. et al. Genome-wide association of hypoxia-inducible factor (HIF)- 1alpha and HIF-2alpha DNA binding with expression profiling of hypoxia-inducible transcripts. J. Biol. Chem. (2009). doi:10.1074/jbc.M901790200 54. Shen, G. M., Zhang, F. L., Liu, X. L. & Zhang, J. W. Hypoxia-inducible factor 1- mediated regulation of PPP1R3C promotes glycogen accumulation in human MCF-7 cells under hypoxia. FEBS Lett. (2010). doi:10.1016/j.febslet.2010.09.040 55. Pescador, N. et al. Hypoxia promotes glycogen accumulation through hypoxia inducible factor (HIF)-mediated induction of glycogen synthase 1. PLoS One (2010). doi:10.1371/journal.pone.0009644 56. Schild, T., Low, V., Blenis, J. & Gomes, A. P. Unique Metabolic Adaptations Dictate Distal Organ-Specific Metastatic Colonization. Cancer Cell (2018). doi:10.1016/j.ccell.2018.02.001 57. Alvarado, A. & Arce, I. Metabolic Functions of the Lung, Disorders and Associated Pathologies. J. Clin. Med. Res. (2016). doi:10.14740/jocmr2668w 58. Valavanidis, A., Vlachogianni, T., Fiotakis, K. & Loridas, S. Pulmonary oxidative stress, inflammation and cancer: Respirable particulate matter, fibrous dusts and ozone as major causes of lung carcinogenesis through reactive oxygen species mechanisms. International Journal of Environmental Research and Public Health (2013). doi:10.3390/ijerph10093886 59. Lebleu, V. S. et al. PGC-1α mediates mitochondrial biogenesis and oxidative phosphorylation in cancer cells to promote metastasis. Nat. Cell Biol. (2014). doi:10.1038/ncb3039 60. Andrzejewski, S. et al. PGC-1α Promotes Breast Cancer Metastasis and Confers Bioenergetic Flexibility against Metabolic Drugs. Cell Metab. (2017). doi:10.1016/j.cmet.2017.09.006 61. St-Pierre, J. et al. Suppression of Reactive Oxygen Species and Neurodegeneration by the PGC-1 Transcriptional Coactivators. Cell (2006). doi:10.1016/j.cell.2006.09.024 62. Christen, S. et al. Breast Cancer-Derived Lung Metastases Show Increased Pyruvate Carboxylase-Dependent Anaplerosis. Cell Rep. (2016). doi:10.1016/j.celrep.2016.09.042 63. Sellers, K. et al. Pyruvate carboxylase is critical for non-small-cell lung cancer proliferation. J. Clin. Invest. (2015). doi:10.1172/JCI72873 64. Yin, G. et al. Predicting brain metastases for non-small cell lung cancer based on magnetic resonance imaging. Clin. Exp. Metastasis (2017). doi:10.1007/s10585- 016-9833-7 65. Shinde, A., Wilmanski, T., Chen, H., Teegarden, D. & Wendt, M. K. Pyruvate carboxylase supports the pulmonary tropism of metastatic breast cancer. Breast Cancer Res. (2018). doi:10.1186/s13058-018-1008-9 66. Kietzmann, T. Metabolic zonation of the liver: The oxygen gradient revisited. Redox Biology (2017). doi:10.1016/j.redox.2017.01.012 67. Dupuy, F. et al. PDK1-dependent metabolic reprogramming dictates metastatic

19 potential in breast cancer. Cell Metab. (2015). doi:10.1016/j.cmet.2015.08.007 68. Pollari, S. et al. Enhanced serine production by bone metastatic breast cancer cells stimulates osteoclastogenesis. Breast Cancer Res. Treat. (2011). doi:10.1007/s10549-010-0848-5 69. Lemma, S. et al. MDA-MB-231 breast cancer cells fuel osteoclast metabolism and activity: A new rationale for the pathogenesis of osteolytic bone metastases. Biochim. Biophys. Acta - Mol. Basis Dis. (2017). doi:10.1016/j.bbadis.2017.08.030 70. Mashimo, T. et al. Acetate is a bioenergetic substrate for human glioblastoma and brain metastases. Cell (2014). doi:10.1016/j.cell.2014.11.025 71. Chen, J. et al. Gain of glucose-independent growth upon metastasis of breast cancer cells to the brain. Cancer Res. (2015). doi:10.1158/0008-5472.CAN-14- 2268 72. Aversa, C. et al. Metastatic breast cancer subtypes and central nervous system metastases. Breast 23, 623–628 (2014). 73. Niikura, N., Saji, S., Tokuda, Y. & Iwata, H. Brain metastases in breast cancer. Jpn. J. Clin. Oncol. 44, 1133–1140 (2014). 74. E.M., O. et al. Incidence and risk of central nervous system metastases as site of first recurrence in patients with HER2-positive breast cancer treated with adjuvant trastuzumab. Ann. Oncol. (2013). doi:10.1093/annonc/mdt036 75. Lin, N. U. & Winer, E. P. Brain metastases: The HER2 paradigm. Clinical Cancer Research 13, 1648–1655 (2007). 76. Ballabh, P., Braun, A. & Nedergaard, M. The blood-brain barrier: An overview: Structure, regulation, and clinical implications. Neurobiology of Disease (2004). doi:10.1016/j.nbd.2003.12.016 77. Sagar, D. et al. Mechanisms of dendritic cell trafficking across the blood-brain barrier. Journal of Neuroimmune Pharmacology (2012). doi:10.1007/s11481-011- 9302-7 78. Carbonell, W. S., Ansorga, O., Sibson, N. & Muschel, R. The vascular basement membrane as ‘soil’ in brain metastasis. PLoS One (2009). doi:10.1371/journal.pone.0005857 79. Kienast, Y. et al. Real-time imaging reveals the single steps of brain metastasis formation. Nat. Med. 16, 116–22 (2010). 80. Lorger, M. & Felding-Habermann, B. Capturing changes in the brain microenvironment during initial steps of breast cancer brain metastasis. Am. J. Pathol. (2010). doi:10.2353/ajpath.2010.090838 81. Eichler, A. F. et al. The biology of brain metastases-translation to new therapies. Nature Reviews Clinical Oncology (2011). doi:10.1038/nrclinonc.2011.58 82. Meuwissen, R. et al. Induction of small cell lung cancer by somatic inactivation of both Trp53 and Rb1 in a conditional mouse model. Cancer Cell 4, 181–189 (2003). 83. Kato, M. et al. The herbal medicine Sho-saiko-to inhibits growth and metastasis of malignant melanoma primarily developed in ret-transgenic mice. J. Invest. Dermatol. 111, 640–644 (1998). 84. Yoneda, T., Williams, P. J., Hiraga, T., Niewolna, M. & Nishimura, R. A bone- seeking clone exhibits different biological properties from the MDA-MB-231 parental human breast cancer cells and a brain-seeking clone in vivo and in vitro.

20 J. Bone Miner. Res. (2001). doi:10.1359/jbmr.2001.16.8.1486 85. Oberheim, N. A. et al. Uniquely Hominid Features of Adult Human Astrocytes. J. Neurosci. 29, 3276–3287 (2009). 86. Martins-Green, M., Petreaca, M. & Yao, M. Chapter 8 An Assay System for In Vitro Detection of Permeability in Human ‘Endothelium’. Methods Enzymol. 443, 137–153 (2008). 87. Czupalla, C. J., Liebner, S. & Devraj, K. In vitro models of the blood-brain barrier. Methods Mol. Biol. (2014). doi:10.1007/978-1-4939-320-7_34 88. Jeon, J. S., Zervantonakis, I. K., Chung, S., Kamm, R. D. & Charest, J. L. In Vitro Model of Tumor Cell Extravasation. PLoS One 8, (2013). 89. Griep, L. M. et al. BBB on CHIP: Microfluidic platform to mechanically and biochemically modulate blood-brain barrier function. Biomed. Microdevices 15, 145–150 (2013). 90. Shao, X. et al. Development of a blood-brain barrier model in a membrane-based microchip for characterization of drug permeability and cytotoxicity for drug screening. Anal. Chim. Acta (2016). doi:10.1016/j.aca.2016.06.028 91. Prabhakarpandian, B. et al. SyM-BBB: a microfluidic blood brain barrier model. Lab Chip 13, 1093–1101 (2013). 92. Booth, R. & Kim, H. Characterization of a microfluidic in vitro model of the blood- brain barrier (μBBB). Lab Chip 12, 1784 (2012). 93. Brown, J. A. et al. Recreating blood-brain barrier physiology and structure on chip: A novel neurovascular microfluidic bioreactor. Biomicrofluidics (2015). doi:10.1063/1.4934713 94. Bang, S. et al. A Low Permeability Microfluidic Blood-Brain Barrier Platform with Direct Contact between Perfusable Vascular Network and Astrocytes. Sci. Rep. (2017). doi:10.1038/s41598-017-07416-0 95. Phan, D. T. T. et al. A vascularized and perfused organ-on-a-chip platform for large-scale drug screening applications. Lab Chip 17, 511–520 (2017). 96. Terrell-Hall, T. B., Ammer, A. G., Griffith, J. I. G. & Lockman, P. R. Permeability across a novel microfluidic blood-tumor barrier model. Fluids Barriers CNS (2017). doi:10.1186/s12987-017-0050-9

21 CHAPTER 2

Breast Cancers Utilize Hypoxic Glycogen Stores via PYGB, the Brain Isoform of

Glycogen Phosphorylase, to Promote Metastatic Phenotypes 1

Summary

In breast cancer, tumor hypoxia has been linked to poor prognosis and increased metastasis. Hypoxia activates transcriptional programs in cancer cells that lead to increased motility and invasion, as well as various metabolic changes. One of these metabolic changes, an increase in glycogen metabolism, has been further associated with protection from reactive oxygen species damage that may lead to premature senescence. Here we report that breast cancer cells significantly increase glycogen stores in response to hypoxia. We found that knockdown of the brain isoform of an enzyme that catalyzes glycogen breakdown, glycogen phosphorylase B (PYGB), but not the liver isoform, PYGL, inhibited glycogen utilization in estrogen receptor negative and positive breast cancer cells; whereas both independently inhibited glycogen utilization in the normal-like breast epithelial cell line MCF-10A. Functionally, PYGB knockdown and the resulting inhibition of glycogen utilization resulted in significantly decreased wound- healing capability in MCF-7 cells and a decrease in invasive potential of MDA-MB-231

1 Megan A. Altemus1,2, Laura E. Goo1, Andrew C. Little1, Joel A. Yates1, Hannah G. Cheriyan1,2, Zhi Fen Wu1, Sofia D. Merajver1. Breast cancers utilize hypoxic glycogen stores via PYGB, the brain isoform of glycogen phosphorylase, to promote metastatic phenotypes. (2019) Manuscript submitted for publication 1Department of Internal Medicine, University of Michigan, Ann Arbor, MI 48109, USA. 2Graduate Program in Cancer Biology, University of Michigan, Ann Arbor, MI 48109, USA.

22 cells. Thus, we identify PYGB as a novel metabolic target with potential applications in

the management and/or prevention of metastasis in breast cancer.

Introduction

More than half of solid tumors present with locally hypoxic or anoxic areas relative to the surrounding normal tissue1. Hypoxia has been associated with metastasis

and poor prognosis in many cancers, including breast1,2. Tumor hypoxia results from an

imbalance between oxygen delivered to the tumor niche and its consumption by cancer

cells and tumor associated cells. Hypoxia develops in primary solid tumors due to

multiple factors, including increased distance from blood supply, weakened vessel

integrity, and competition for oxygen and nutrients from neighboring tumor and tumor-

associated cells1. Both tumor and normal cells respond to a hypoxic environment by

activating specific signaling pathways that lead to distinct gene expression changes,

amongst the most immediate and salient common hubs being stabilization of the

hypoxia-inducible factors, HIF-1α and HIF-2α3.

In cancer, stabilized HIF-1α activates transcriptional programs that have been

recognized to induce the epithelial to mesenchymal transition (EMT) and support

metastasis in various cancer types4–7. Hypoxia-induced transcription factors and

signaling pathways include Twist, Snail, ZEB1, Notch, TGF-β, and Hedgehog, among

others6,8–14. Moreover, HIF-1α and hypoxia have been shown to increase the metastatic

phenotype in multiple cancer cell types, including breast cancer in in vivo experiments,

and have been linked to increased risk of metastasis and mortality in breast cancer

patient cohorts7,15–22. Investigation of how the hypoxic tumor microenvironment

23 contributes to increased cancer aggressiveness and metastatic potential may provide

novel therapeutic avenues.

Hypoxia and subsequent stabilization of HIF-1α induce downstream metabolic

changes in cancer cells. These include increased expression of glucose transporters

and genes involved in glycolysis, altered fatty-acid and lipid metabolism, and increased pyruvate dehydrogenase kinase activity thereby decreasing the amount of pyruvate that enters the TCA cycle and decreasing oxidative phosphorylation23–31. One additional metabolic change is glycogen accumulation, which has been previously described in both cancerous and non-cancerous cells under hypoxic conditions relative to their normal state32–35. Exploiting the potential vulnerabilities that arise from tumors’

adaptations to hypoxia has been shown to contribute to tumor control using a of

methods36. In this study, we focus on a deeper understanding of the potential

vulnerabilities exhibited by the hypoxic modulation of glycogen , carried out

by a delicate balance between synthesizers and degraders of glycogen.

Glycogen is a high molecular weight branched polysaccharide of glucose and is

the main glucose storage macromolecule in animals37. It is primarily stored in the liver

where it is utilized to maintain blood-glucose levels and in the muscles where it can be

mobilized quickly for energy production during exercise37. Glycogen is synthesized

around a glycogenin core by addition of UDP-glucose onto growing glycogen chains.

Glucose-1-phosphate available in the cell from either glucose transported into the cell or

gluconeogenic substrates is catalyzed to UDP-glucose by UDP-glucose

pyrophosphorylase-2 (UGP2). UDP-glucose is added onto glycogen via an α-1,4 linkage

by glycogen synthase, the rate limiting enzyme in glycogen synthesis37. There are two

24 isoforms of glycogen synthase: GYS1 which is mainly expressed in the liver and the

muscle isoform GYS2. During degradation, glucose-1-phosphate molecules are

removed from the non-reducing end of the glycogen molecule by glycogen

phosphorylase (PYG), the rate limiting enzyme of glycogen degradation37. PYG has

three different isoforms in humans that are typically expressed in different tissues: liver

(PYGL), muscle (PYGM), and brain isoforms (PYGB). Free glucose-1-phosphate

molecules are then catalyzed to glucose-6-phosphate, the first intermediate in

glycolysis, by phosphoglucomutase-1 (PGM1).

To maintain glycogen-free glucose balance, glycogen synthase and glycogen

phosphorylase are tightly regulated. Glycogen synthase activity is regulated

allosterically and via posttranslational modification e.g. phosphorylation.

Phosphorylation of glycogen synthase by glycogen synthase kinase 3α and 3β

(GSK3α/β) at multiple serine residues inhibits its activity38. Glycogen phosphorylase is

activated by phosphorylation at Ser-14 by phosphorylase kinase37 and by the allosteric

stimulator glucose-6-phosphate (G6P). Importantly, both glycogen synthase and

glycogen phosphorylase are regulated in synchrony by protein phosphatase-1 (PP1).

PP1 dephosphorylation activates glycogen synthase, but inhibits glycogen phosphorylase, leading to reciprocal regulation of glycogen synthesis and degradation37.

High levels of glycogen have been found in diverse cancer cell types including

breast cancers39. Recently, glycogen levels were found to be inversely correlated with

proliferation rate, indicating that glycogen was utilized as an energy source to sustain

proliferation39. High levels of glycogen have also been found in the hypoxic tumor cores

25 and in tumors treated with anti-angiogenic therapies40. Hypoxia and stabilization of HIF-

1α have been shown to increase levels of many glycogen enzymes and regulatory proteins including UGP2, GYS1, GBE, and PPP1R3C, the glycogen-associated regulatory subunit 3C of PP135,41–43.

In this study, we sought to understand the link between the fuel provided by

hypoxia-induced glycogen storage in aggressive breast cancers and the promotion of

invasion and migration. We found that six different breast cancer cell lines and a

normal-like breast epithelial cell line all increased their glycogen stores under hypoxia.

Glycogen gene expression changes under hypoxia were also evaluated, finding no

consensus change in expression that would account for this increase, indicating other

means of regulation of glycogen stores are in place, such as post-translational

modification or of the rate-limiting enzymes. In order to investigate

how proliferation, migration, and invasion are affected by glycogen storage and

utilization, we created glycogen phosphorylase knockdowns for both the liver and brain

isoforms of PYG. In the two breast cancer cell lines, MDA-MB-231 and MCF-7, loss of

the brain isoform PYGB inhibited hypoxic glycogen usage whereas the loss of both

PYGL and PYGB in the normal-like MCF-10A cell line exhibited this effect. Prohibition of

glycogen utilization resulted in a marked decrease of proliferation in MCF-10A cells and

a slight decrease in MCF-7 cells. Wound-healing was strikingly decreased in shPYGB

MCF-7 cells under both normoxic and hypoxic conditions. While loss of PYGB did not

affect the proliferation or wound-healing of triple-negative breast cancer (TNBC) MDA-

MB-231 cells, it did significantly decrease the invasiveness of these cells. These

findings indicate that attacking the cancer vulnerabilities derived from dysregulation of

26 glycogen metabolism could be a therapeutic strategy to inhibit development of distant

metastases in breast cancers such as TNBC, for which few targeted therapies currently

exist.

Materials and Methods

Cell culture and media

MDA-MB-231 cells (ATCC HTB-26) were maintained in RPMI-1640 (+) L-

glutamine (ThermoFisher 11875093) supplemented with 10% fetal bovine serum (FBS)

(Corning 35-010-CV). SUM-14944 cells (gift from Dr. Steve Ethier, Medical University of

South Carolina) were maintained in Ham’s F-12 (+) L-glutamine (ThermoFisher

11765054) supplemented with 5% FBS (GE Healthcare Life Sciences SH30071.03), 5

μg/mL insulin, and 1 μg/mL hydrocortisone. MCF-7 and MDA-468 (ATCC HTB-132) cells were maintained in DMEM (Corning 10-013-CV) supplemented with 10% FBS

(Corning 35-010-CV). MCF-10A cells (ATCC CRL-10317) were maintained in 50:50

DMEM/F-12 (Corning 10-090-CV) supplemented with 5% horse serum (ATCC 30-

2040), 10 μg/mL insulin, 0.5 μg/mL hydrocortisone, 0.02 μg/mL epidermal growth factor, and 0.1 μg/mL cholera toxin. BT-549 cells (ATCC HTB-122) were maintained in RPMI-

1640 (+) L-glutamine (ThermoFisher 11875093) supplemented with 10% FBS (Corning

35-010-CV) and 0.8 μg/mL insulin. SUM-19045 cells (gift from Dr. Steve Ethier, Medical

University of South Carolina) were maintained in Ham’s F-12 (+) L-glutamine

(ThermoFisher 11765054) supplemented with 1mg/mL bovine serum albumin (BSA)

(Sigma-Aldrich A8806), 1X ITS-X (ThermoFisher 51500056), 10 nM T3, 10 mM HEPES,

and 1 μg/mL hydrocortisone. All cell lines were supplemented with 1X Anti-Anti

27 (ThermoFisher 15240062) and 5 μg/mL gentamicin (ThermoFisher 15750060), except

SUM-149 and SUM-190 which were supplemented with 2.5 μg/mL amphotericin B

(ThermoFisher 15290018), 50 U/mL penicillin-streptomycin, and 5 μg/mL gentamicin

(ThermoFisher 15750060). Metabolism media (MM) was made from base DMEM

(Sigma-Aldrich D5030) which was supplemented with 44 mM sodium bicarbonate, 11

mM glucose, 2.5 mM glutamine, and 1 mM sodium pyruvate as well as all cell-line

specific supplements listed above. All cells were grown in 5% CO2, except SUM-149

and SUM-190 which were grown at 10% CO2. All hypoxia experiments were performed

at 1% O2 with stated CO2 concentration in an incubator system with O2 and CO2 sensors to maintain accurate gas concentration.

Glycogen assay

For each experiment four 10 cm plates of cells were plated in normal growth media and grown in CO2 controlled normal atmosphere (normoxia) for 24 hr. Plates were washed with Dulbecco’s phosphate-buffered saline (DPBS) (ThermoFisher

14190250) and changed to MM for one hour. Media was changed again with MM and two plates were placed at 1% O2 and two were kept in normoxic conditions. After 24 hr

and 48 hr, cells from one plate from each condition were harvested with 0.05% trypsin

(ThermoFisher 25300054) and pelleted via centrifugation. Cell pellets were washed with

DPBS and centrifuged a second time. Cell pellet was flash frozen in liquid nitrogen and

stored at -80°C for up to three days. Glycogen assay was conducted using a glycogen assay kit (Cayman Chemical 700480). Cell pellets were re-suspended in 400 mL Assay

Buffer and homogenized using 1 mL pestle tissue grinder. Glycogen concentration was

normalized to protein concentration in lysate. Each glycogen assay was performed in

28 biological triplicate. One-way ANOVA multiple comparison statistics were calculated

using GraphPad Prism 7.04 software. Glycogen utilization assays for knockdown cell

lines were conducted as described above with the following alterations: two plates of

each cell line were placed in 1% O2 conditions for 24 hr, following which one plate was

harvested and the second plate was moved to normoxic conditions for 24 hr before

harvesting.

Periodic acid-Schiff staining

Cells were grown in 6-well tissue culture plate on acid-washed coverslips.

Coverslips were prepared by heating overnight in 1 M HCl at 55°C. Solution was cooled

to room temperature and coverslips were rinsed in ddH2O. Coverslips were then rinsed

in a sonicating water bath in fresh ddH2O for 15min 3X, followed by 50%, 70%, then

95% ethanol for 15 min each. Coverslips were stored in 95% ethanol. Once cells

reached 50% confluent media was changed to MM and cells were incubated in

normoxia or 1% O2 conditions for 48hr. Cells were fixed in Carnoy’s solution of 60%

ethanol, 30% chloroform, and 10% glacial acetic acid, permeabilized in 0.5% Triton X-

100 at 4°C and quenched in PBS-glycine. Amylase controls were digested in 0.05

mg/mL α-amylase (Sigma-Aldrich A3176) at room temperature. All coverslips were

incubated in Periodic Acid Solution (Sigma-Aldrich 395B) and then Schiff’s Reagent

(Sigma-Aldrich 395B). Nuclei were counterstained in Hematoxylin Gill’s Solution No. 3

(Sigma-Aldrich 395B) and blued in Scott’s Tap Water Substitute (3.5 g sodium

bicarbonate, 20 g magnesium sulfate, 1 L distilled H2O). Coverslips were air dried and mounted on slides for imaging.

29 RT-qPCR

RNA was isolated for all cells using RNeasy mini kit (Qiagen) and reverse

transcription for cDNA preparation was performed using Reverse Transcription System

(Promega) according to manufacturer’s instructions. qPCR was performed using

QuantiTect SYBR Green PCR kit (Qiagen) on a Step-One Plus real-time PCR system

(Applied Biosystems). Biological triplicates were performed for each qPCR analysis. Ct values were normalized to RPL22 and RPL30. Statistical analysis for each triplicate was performed on dCT values using multiple comparisons one-way ANOVA in Graphpad

Prism 7.04. qPCR primers utilized in study are reported in Table 2.1. shRNA knockdown

Annealed shRNA oligonucleotides were cloned into pLentilox 3.7 GFP lentiviral expression vector obtained from University of Michigan Vector Core using restriction enzymes XhoI and HpaI. shRNA expression is driven by mU6 promoter with selection marker GFP expression driven by CMV promoter. An empty vector (EV) pLentilox 3.7

GFP plasmid was used as a control. Lentivirus was produced by University of Michigan

Vector Core. After viral transduction all cell lines were sorted for GFP expression and viability using the University of Michigan Flow Cytometry Core. shRNA oligonucleotides utilized are detailed in Table 2.2. Knockdowns were confirmed via western blot and RT- qPCR.

Western blotting

Protein samples were harvested from 6 cm tissue culture dish using RIPA buffer with protease inhibitors (cOmplete Mini Protease Inhibitor Cocktail, Sigma

11836153001). Protein concentration in lysate determined using BCA protein assay

30 (ThermoFisher 23225). 40 μg protein per sample were loaded onto pre-cast 4-12% Tris-

HCL gel (Bio Rad 3450027). Gel was transferred onto nitrocellulose membrane

(ThermoFisher IB23001) using iBlot2 system (ThermoFisher IB21001) and blocked in

5% blotting-grade non-fat dry milk (Bio Rad 1706404). Primary antibodies utilized were

PYGL rabbit polyclonal antibody (ThermoFisher PA5-51492), GPBB (PYGB) rabbit

polyclonal antibody (ThermoFisher PA5-28022), and Actin mouse monoclonal antibody

(Sigma A3854).

Proliferation assay

Cells were incubated for 24 hr at normoxic and 1% O2 conditions. After 24 hr cells were plated in black clear-bottom 96-well plates in technical quintuplet at 2,000 cells/well for MDA-MB-231 and 3,000 cells/well for MCF-7 and MCF-10A and incubated

overnight at normoxia or 1% O2 conditions. Media was changed and bright-field images

were taken every four hours for five days using Biotek Cytation5 and oxygen controlled

Biotek Biospa8 system. Experiments were performed in biological triplicate. Cell counts

were calculated at each timepoint using Biotek Gen5 software. Cell count data for each

technical replicate over time were fit to exponential curves using GraphPad Prism 7.04.

Geometric mean and geometric standard deviation of technical replicate rate constants

were calculated for each biological replicate. Multiple comparisons one-way ANOVA were conducted on rate constants of three biological replicates to determine significance.

Wound-healing assay

Cells were incubated in MM (see cell culture and media) for 24 hr at normoxia

and 1% O2 conditions. After 24hr cells were plated in Ibidi 2-well culture inserts (Ibidi

31 80209) affixed to 12-well plates at 50,000 cells/well in technical triplicate and incubated

overnight at normoxia and 1% O2 conditions. After overnight incubation, inserts were

removed and washed 2X with PBS before adding fresh MM. Bright-field images were

taken every four hours until wounds were fully closed using Biotek Cytation5 and

oxygen controlled Biospa8 system. Wound area for each image in pixels was calculated

using the ImageJ macro MRI Wound Healing Tool. These values were used to calculate

the percent closure of the wound at each time point. Reported values are the average

percent closure and standard error of biological triplicates at each timepoint. p-values

were determined using two-way ANOVA in GraphPad Prism 7.04

Transwell invasion assay

Cells were incubated in MM for 24 hr at normoxia and 1% O2 conditions. After 24

hr cells were plated in Corning BioCoat Matrigel Invasion Chambers (Corning 354480)

at 100,000 cells/chamber in serum-free MM with serum-containing MM in bottom chamber in technical duplicate. Cells were allowed to invade for 24 hr in normoxia or

1% O2 conditions, respectively. After invasion, Matrigel and top chamber were scrubbed

to remove non-invading cells. Inserts were fixed in 70% ethanol and stained in 0.2%

Crystal Violet. Bottom of insert was removed and mounted on slides for imaging. 5 images per insert were taken at 10X magnification. Each image was processed using

ImageJ. Images were converted to binary in ImageJ and converted to percent coverage based on total pixels per image. Statistical analyses were conducted on technical triplicates using multiple comparisons one-way ANOVA in GraphPad Prism 7.04.

32 Results

Glycogen levels were quantified in six different breast cancer cell lines and a normal-like breast cell line at 24 and 48h exposure to normoxia and 1% O2 hypoxia

(Figure 2.1 A-D and 2.2). Glycogen levels were significantly increased in the TNBC cell line MDA-MB-231, inflammatory TNBC SUM-149, ER+ MCF-7, and normal-like MCF-

10A (Figure 2.1). However, there was great variation in the extent to which glycogen accumulation increased under hypoxia compared to normoxia. MDA-MB-231 and MCF-

7 cells’ glycogen levels increased 1.5-3X under hypoxia, while SUM-149 and MCF-10A increased from 10-75X. In addition, we find variable basal glycogen levels in normoxia, with MCF-7 having the highest normoxic glycogen levels and normal-like MCF-10A having essentially no normoxic glycogen stores. Periodic-acid Schiff (PAS) staining for glycogen was conducted on MDA-MB-231 and SUM-149 cell lines to visually confirm glycogen accumulation (Figure 2.1 E-F). Glycogen accumulation was observed via PAS staining with no detectable change in staining intensity between normoxic and hypoxic

MDA-MB-231. SUM-149, with higher hypoxic glycogen increases, as determined via the glycogen assay (Figure 2.1 B), exhibited a diffuse cytoplasmic pink staining in hypoxic, but not normoxic conditions. These results, paired with data from additional TNBC cell lines MDA-MB-468, BT-549, and the inflammatory HER-2+ breast cancer cell line SUM-

190 (Figure 2.2), show a wide range of normoxic glycogen storage and glycogen accumulation in response to hypoxia in all the subtypes.

We then investigated whether hypoxia-induced transcriptional changes in glycogen metabolism genes contribute to the observed glycogen accumulation under hypoxia. Glycogen synthesis and degradation pathways are outlined in Figure 2.3 A.

33 Glycogen synthase expression increased significantly under hypoxia in MCF-7 cells only (Figure 2.3 B). We hypothesized that decreases in glycogen phosphorylase

(PYGL/B) would also account for increased glycogen accumulation; however, no significant changes in expression were observed between normoxic and hypoxic conditions (Figure 2.3 C-D) for PYGL/B. Expression of GYS2, GSK3α, GSK3β, PYGM,

GYG1, and GYG2 were also evaluated showing no significant changes between normoxia and hypoxia (Figure 2.4). However, there was a significant increase in the

PP1 complex member PPP1R3C observed in SUM-149 and MCF-10A cells under hypoxic conditions (Figure 2.3 E). These findings indicate that glycogen accumulation is likely not due to transcriptional regulation of the synthesis/degradation rate-limiting enzymes, but to other processes, such as allosteric regulation or post-translational modifications.

We hypothesized that the ability to mobilize glucose from glycogen is an important vulnerability of aggressive breast cancers in that the degradation of glycogen is needed for invasion and motility. In order to observe whether the ability to store and

utilize glycogen under hypoxia affects the ability of breast cancer cells to proliferate,

migrate, and invade, cells that are unable to use glycogen stores are a necessary

negative control. Since there were no consensus transcriptional changes observed that

may account for glycogen accumulation under hypoxia, we knocked down glycogen

phosphorylase, the rate-limiting enzyme of glycogen degradation in order to inhibit the

breast cancer cells from utilizing glycogen. Breast cells express both the liver (PYGL)

and brain (PYGB) isoforms of glycogen phosphorylase; therefore, we designed shRNA

vectors against both PYGL and PYGB. We confirmed PYGL and PYGB knockdown in

34 our cell models by western blot (Figure 2.5 A) and qPCR (Figure 2.6) compared to wild- type (WT) and empty-vector (EV) controls. To determine if glycogen utilization is regulated by PYGL or PYGB or both, we allowed the shPYGL and shPYGB cells to accumulate glycogen for 24 hr in hypoxia (hypoxic control), then transferred the cells to

normoxia for 24 hr to deplete their glycogen stores (normoxic exposure), and subsequently tested their levels of glycogen (Figure 2.5 B-D).

Loss of PYGB but not PYGL in MDA-MB-231 (Figure 2.5 B) and MCF-7 (Figure

2.5 C) resulted in significantly more glycogen retention (p<0.05) under normoxia

compared to wild-type and empty vector control cells. This indicates that PYGB, but not

PYGL, is responsible for degradation of hypoxic glycogen stores in these cells. In

normal-like breast epithelial MCF-10A cells, loss of PYGL resulted in significantly more

glycogen retention under normoxia (p<0.01) and loss of either PYGL and PYGB

maintained significantly more glycogen under hypoxia (p<0.05) compared to wild-type

and empty vector controls. This indicates that both PYGL and PYGB are responsible for

glycogen degradation in MCF-10A cells. The increase in glycogen stores under hypoxia

observed in these cell lines is likely due to inhibition of degradation, rather than an

increase in glycogen synthesis, given that we are reducing levels of glycogen

phosphorylase, the canonical rate-limiting enzyme of glycogen degradation.

In addition, we generated stably expressing shPYGB and shPYGL vectors in

SUM-149 cells. We confirmed the efficacy of PYGB or PYGL knockdown by RNA and

protein expression (Figure 2.7 A-B). In SUM-149, loss of PYGL, rather than PYGB,

abrogated glycogen utilization (Figure 2.7 C). Loss of PYGL in SUM-149 cells resulted in cessation of cellular propagation and, eventually, cell death. Future studies will focus

35 on alternate methods of impeding glycogen utilization in SUM-149 cells in order to

determine its effects on invasion and metastasis. Overall, these data show that the

glycogen phosphorylase isoform responsible for the majority of glycogen degradation

varies depending on breast cell line, with PYGB being the primary isoform responsible

in MDA-MB-231 and MCF-7, and both PYGL and PYGB contributing in normal-like

MCF-10A.

Proliferative effects due to the loss of either PYGL or PYGB were observed

under normoxic and hypoxic conditions by live-cell bright-field imaging and cell counting

every four hours for five days (Figure 2.8). Cell counts from the exponential growth

phase and corresponding growth rate were determined. Doubling times for each cell line

are reported in Table 1. Proliferation of shPYGL and shPYGB MDA-MB-231 cell lines did not differ from wild-type or empty vector controls in normoxic or hypoxic conditions

(Figure 2.8 A, Table 2.3). MCF-7 shPYGB cells had a significantly decreased proliferation rate (p<0.05) compared to both wild type and empty vector controls under normoxia; however, there were no significant differences between the shPYG cell lines and controls under hypoxic conditions (Figure 2.8 B, Table 2.3). Loss of PYGB and

PYGL significantly decreased the proliferation rate of MCF-10A cells compared to the empty vector control (p<0.01) in normoxic and hypoxic conditions (Figure 2.8 C, Table

2.3). These data correlate with Figure 2.5 C-D in that the loss of PYGB but not PYGL in

MCF-7 inhibited glycogen utilization and slowed proliferation whereas loss of either enzyme in MCF-10A inhibited glycogen utilization and slowed proliferation compared to controls.

36 Next, we determined whether ability to utilize glycogen stores would also affect

migration of breast cancer cells in both normoxic and hypoxic conditions using a wound-

healing assay. Cells were plated in a 2-well insert which was removed after cells

adhered to the bottom of the plate to create a uniform gap (“wound”) between the two

groups of cells. Bright-field images were taken every four hours until the wound was

fully closed, and the percentage of wound closure was calculated based on the area of

the wound at each time point. Loss of PYGL and PYGB had no effect on wound-healing

in MDA-MB-231 cells in normoxia or hypoxia (Figure 2.9 A). In MCF-7 cells, loss of

PYGB significantly inhibited wound-healing compared to wild-type (p<0.0001), empty vector control (p<0.0001), and to loss of PYGL (p<0.01), under normoxia (Figure 2.9 B).

Under hypoxic conditions, loss of PYGB again significantly inhibited wound closure compared to wild-type (p<0.0001), empty vector (p<0.0001), and loss of PYGL

(p<0.0001) (Fig 5B). Loss of PYGL significantly inhibited wound closure of MCF-7 cells under hypoxic conditions compared to wild-type (p<0.0001) and empty vector (p<0.05) controls, but the effect was smaller from that seen for the shPYGB MCF-7 (Figure 2.9

B). Even though knockdowns of PYGL and PYGB significantly reduced proliferation in

MCF-10A, they had no effect on wound-healing of these normal-like epithelial cells in either normoxic or hypoxic conditions (Figure 2.9 C).

Cellular invasion, as well as migration and proliferation, is an important phenotypic determinant of aggressive breast cancer progression and metastasis. TNBC cells MDA-MB-231 are highly invasive, unlike ER+ MCF-7 or normal-like MCF-10A.

Using Matrigel-coated transwell inserts and serum as a chemoattractant, we determined that loss of PYGB, but not PYGL, significantly reduced the invasiveness (p<0.05) of

37 MDA-MB-231 cells compared to wild-type and empty vector controls in normoxic

conditions (Figure 2.9 D). Under hypoxic conditions, loss of PYGB significantly reduced invasiveness compared to loss of PYGL (p<0.05) and trended towards reducing invasion compared to wild-type and empty vector controls, though not reaching

statistical significance (Figure 2.9 E). Representative images of invaded cells can be

seen in Figure 2.9 F. Importantly, these data are consistent with Figure 2.5 B in that loss of PYGB, but not PYGL, inhibits glycogen utilization in MDA-MB-231.

Discussion and Conclusion

It is well known that hypoxia increases migration, invasion, and metastasis in a variety of cancers, including breast cancers. Hypoxia also induces glycogen accumulation in cancer cells, promoting proliferation, protecting cells from reactive oxygen species, and preventing senescence34,39,40. Here we show that different types of

breast cancer cells exhibit hypoxic glycogen accumulation and utilization of these

glycogen stores contributes to proliferation, migration, and invasion.

All breast cancer cells tested increased glycogen stores in response to hypoxia.

However, baseline normoxic glycogen levels and the amount of glycogen increase

under hypoxic conditions varied widely between cell lines, with no discernible pattern

based on receptor status or sub-type. Inflammatory breast cancer cells increased their

glycogen stores in response to hypoxia by over 10-fold higher than other breast cancer

cell types, suggesting that interventions based on inhibiting glycogen utilization may be

more damaging to this aggressive breast cancer subtype. Overall, these data indicate

that glycogen metabolism phenotypes in breast cancer may vary widely depending on

38 each individual tumor. Even though there is no pattern to glycogen metabolism that is

distinct for the commonly used breast cancer biomarkers, relative glycogen levels in

patient biopsies can be determined by a simple and reliable histological test (PAS

staining), thus potentially facilitating patient selection for interventions based on

glycogen metabolism in the future.

We also found no single consensus glycogen gene expression signature that

would account for the hypoxic glycogen accumulation observed in our breast cancer

cells, indicating that there will be high degree of heterogeneity in the regulation of the

common event we describe of glycogen storage under hypoxia. Previous studies have

proposed that glycogen accumulation in breast cancer cells is due to HIF-1α mediated

increase in GYS1 and/or PPP1R3C expression in hypoxia42,43. In agreement with those

prior results, we found an increase in GYS1 expression in MCF-7 cells and increased

PPP1R3C expression in SUM-149 and normal-like MCF-10A cells; however, importantly, we find that there is no consensus glycogen-related gene expression among all breast cancer cells that leads to the observed hypoxic increase in glycogen.

This result is important because it suggests that modulation of the rate limiting reactions of glycogen synthesis or degradation, rather than interventions on upstream targets, would have more general utility in breast cancer. This accumulation of glycogen could also be caused by allosteric regulation or phosphorylation/dephosphorylation of the rate-limiting enzymes of glycogen metabolism. Future studies will need to determine the exact mechanism of hypoxic glycogen accumulation based on glycogen synthase and glycogen phosphorylase regulation and the possible relation to HIF-1α stabilization under hypoxia in breast cancer, in a context dependent manner.

39 Regardless of the mechanism of hypoxic glycogen accumulation, we successfully

inhibited glycogen utilization in breast cancer using shRNA knockdown of the glycogen

phosphorylase isoforms PYGL and PYGB. Previous work in the field focused solely on

the liver isoform of glycogen phosphorylase40. However, we determined that the brain

isoform of glycogen phosphorylase is primarily responsible for glycogen degradation in

MDA-MB-231 and MCF-7 breast cancer cells and both isoforms contribute in normal- like MCF-10A cells. This novel finding should inform future glycogen metabolism studies in breast and other cancers to include all isoforms of glycogen phosphorylase in addition to the well-studied liver isoform.

Inhibition of glycogen utilization also led to drastic phenotypic changes in breast cancer cells in both hypoxic and normoxic conditions. Proliferation was reduced in shPYGB MCF-7 cells and both shPYGL and shPYGB normal-like MCF-10A cells, which matches with the inhibition of glycogen utilization seen in these cells. In the ER+ MCF-7 breast cancer cell line, wound-healing was also inhibited in the shPYGB cells. Wound- healing assays measure the ability of cells to move and grow outwards from an area of dense cell population. Without the ability to utilize glycogen, MCF-7 cells were unable to close the wound as efficiently as the control or the shPYGL cells. This effect was not seen in the non-cancerous, normal-like breast epithelial MCF-10A cells, indicating that glycogen usage to promote migration is a cancer-specific phenotype and thus a possible vulnerability. Additionally, in the TNBC MDA-MB-231 cells, inhibition of glycogen utilization by PYGB knockdown led to a significant decrease in invasive potential, reaffirming the importance of the brain isoform, PYGB, in advantaging cancer cells to more aggressive phenotypes based on enhanced glycogen availability.

40 While all breast cells tested increased glycogen storage under hypoxia, glycogen

utilization as promoted by the brain isoform of glycogen phosphorylase, PYGB, affects

migration and invasion phenotypes only in cancer cells and not in normal-like epithelial cells. These findings suggest PYGB as a potential novel target to reduce invasiveness and metastasis of breast cancers. Future work will focus on recapitulating these in vitro

results in in vivo tumor xenograft and metastasis studies, as well as investigating the

anti-metastatic effects of treatments with glycogen phosphorylases inhibitors, such as

ingliforib 46.

41 Figures

Figure 2.1 Glycogen accumulates in breast cancer cells under hypoxic conditions across subtypes. (A-D) Glycogen levels normalized to total protein lysate in (A) MDA-MB-231, (B) SUM-149, (C) MCF-7, and (D) MCF-10A cell lines at 24 and 48 hr after media change and exposure to normoxia or 1% O2 (hypoxia) conditions. Error bars equal SEM of n=3. *p<0.05, **p<0.01, ***p<0.001 by multiple comparisons one-way ANOVA. (E and F) PAS stain of (E) MDA-MB-231 and (F) SUM-149 cells after 48 hr exposure to normoxia or 1% O2 conditions. Images taken at 10X magnification. Inset is amylase digestion control. Scale bar = 50μm.

42 Figure 2.2 Glycogen accumulates in additional breast cancer cells under hypoxic conditions. (A-D) Glycogen levels normalized to total protein lysate in MDA-MB-468, BT-549, and SUM-190 cell lines at 24 and 48 hr after media change and exposure to normoxia or 1% O2 (hypoxia) conditions. Error bars equal SEM of n=3. *p<0.05, **p<0.01 by multiple comparisons one-way ANOVA.

43 Figure 2.3 Glycogen pathway gene expression changes in breast cancer cells exposed to hypoxia. (A) Schematic of the glycogen synthesis and degradation pathways. (B-E) qPCR of (B) GYS1, (C) PYGL, (D) PYGB, and (E) PPP1R3C gene expression in MDA-MB-231, SUM-149, MCF-7, and MCF-10A cells. Points represent fold-change gene expression of cells exposed to hypoxia for 24 hr compared to normoxia for 24 hr. Normoxia values normalized to 1 are also displayed with associated error. Error bars represent SEM, n=3. *p<0.05 ****p<0.0001

44 Figure 2.4 Additional glycogen pathway gene expression changes in breast cancer cells exposed to hypoxia. (A-F) qPCR of GYS2, GYG1, GYG2, PYGM, GSK3α, and GSK3β gene expression in MDA-MB-231, SUM-149, MCF-7, and MCF-10A cells. Points represent fold-change gene expression of cells exposed to hypoxia for 24 hr compared to normoxia for 24 hr. Normoxia values normalized to 1 are displayed with associated error. No points mean the gene was not expressed in this cell line. Error bars represent SEM, n=3.

45

Figure 2.5 Glycogen phosphorylase brain isoform knockdown inhibits glycogen utilization in breast cancer cells. (A) Western blot of liver and brain glycogen phosphorylase (PYGL and PYGB) in wild-type, either empty vector, shPYGL and shPYGB transduced MDA-MB-231, MCF-7, and MCF-10A cells. Actin is shown as a loading control. (B-D) Glycogen levels normalized to protein in lysate after 24h in hypoxia (Hypoxic Control) or 24h in hypoxia with subsequent 24 hr in normoxia (Normoxic Exposure) for wild-type, empty vector, shPYGL, and shPYGB in (B) MDA-MB-231, (C) MCF-7, and (D) MCF-10A cells, respectively. Error bars represent SEM of n=3. *p<0.05, **p<0.01 compared to EV by multiple comparison one-way ANOVA.

46

Figure 2.6 shPYGL and shPYGB reduces glycogen phosphorylase mRNA expression. (A-C) qPCR confirmation of PYGL and PYGB knockdown in MDA-MB-231, MCF-7, and MCF-10A cells, respectively. Error bars represent SEM of n=3. *p<0.05, **p<0.01, ***p<0.001, ****p<0.001 compared to EV by multiple comparisons one-way ANOVA.

47

Figure 2.7 Glycogen phosphorylase knockdown in SUM-149 cells. (A) Western blot of liver (PYGL) and brain glycogen phosphorylase (PYGB) in wild-type, empty vector transduced, shPYGL and shPYG with actin loading control in SUM-149 cells. (B) qPCR confirmation of PYGL and PYGB knockdown in SUM-149, n=1. (C) Glycogen levels normalized to protein in lysate after 24h in hypoxia (Hypoxic Control) or 24 hr in hypoxia with subsequent 24 hr in normoxia (Normoxic Exposure) for wild-type, empty vector, shPYGL, and shPYGB in SUM-149, n=1.

48

Figure 2.8 Loss of glycogen phosphorylase inhibits proliferation in MCF-7 and normal-like MCF- 10A cells but not MDA-MB-231. Exponential curve fit of (A) MDA-MB-231, (B) MCF-7, and (C) MCF-10A cell proliferation in normoxic and hypoxic conditions, respectively. Dotted line indicates +/- standard error for n=3. *p<0.05 **p<0.01 compared to EV by multiple-comparison one-way ANOVA on doubling times.

49 Figure 2.9 Loss of glycogen phosphorylase brain isoform inhibits wound-closure in MCF-7 and invasion in MDA-MB-231 cells. (A-C) Wound-healing assays for wild-type, empty vector, shPYGL, and shPYGB in (A) MDA-MB-231, (B) MCF-7, and (C) MCF-10A cells in normoxia and hypoxia, respectively. Data represented as percentage of wound closed based on 0 hr time point image compared to images taken every 4 hr. Dotted lines represent SEM at each timepoint of n=3. *p<0.05, ****p<0.0001 compared to EV by two-way ANOVA. (D and E) Quantification of invasion assay of wild-type, empty vector, shPYGL, and shPYGB MDA-MB-231 cells in normoxia and hypoxia, respectively. Data represented as percent coverage of bottom side of transwell membrane after 24 hr invasion from SFM to serum-containing media. Error bars represent SEM of n=3 biological replicates. #p<0.05 compared to wild-type, empty vector, and shPYGL and *p<0.05 compared to shPYGL by multiple-comparison one-way ANOVA. (F) Representative 10X images of wild- type, empty vector, shPYGL, and shPYGB MDA-MB-231 invasion membranes in normoxia and hypoxia, scale bar = 200 μm.

50 Tables

Table 2.1 Primers for qPCR of glycogen genes.

Table 2.2 shRNA oligonucleotides for shPYGL and shPYGB.

51 Normoxia Mean p-value p-value Hypoxia Mean p-value p-value Cell Line Vector Doubling Time compare compare Doubling Time compare compare days (+/- SD) d to WT d to EV days (+/- SD) d to WT d to EV

WT 1.62 (1.04-1.56) - 0.28 1.54 (1.48-1.60) - 0.97 MDA-MB- EV 1.50 (1.48-1.53) 0.28 - 1.50 (1.41-1.59) 0.97 - 231 shPYGL 1.60 (1.58-1.61) 0.97 0.47 1.58 (1.49-1.68) 0.98 0.83 shPYGB 1.66 (1.52-1.81) 0.94 0.14 1.65 (1.44-1.88) 0.76 0.51 WT 1.88 (1.72-2.06) - 0.997 1.86 (1.72-2.02) - 0.99 EV 1.84 (1.65-2.05) 0.997 - 1.82 (1.73-1.92) 0.99 - MCF-7 shPYGL 2.19 (1.88-2.56) 0.52 0.42 1.84 (1.73-1.96) 0.999 0.999 shPYGB 2.87 (2.45-3.35) *0.02 *0.01 2.05 (1.69-2.49) 0.73 0.60 WT 1.01 (0.92-1.10) - 0.93 1.64 (1.50-1.80) - 0.75 EV 0.96 (0.93-0.99) 0.93 - 1.26 (1.12-1.41) 0.75 - MCF-10A shPYGL 1.51 (1.34-1.70) **0.004 **0.002 4.09 (3.21-5.21) *0.04 **0.009 shPYGB 1.41 (1.26-1.58) *0.01 **0.005 5.72 (3.18-10.3) **0.007 **0.002

Table 2.3 Glycogen phosphorylase knockdown decreases doubling time of MCF-7 and MCF-10A cells *p-values were calculated using multiple comparison one-way ANOVA on ln-transformed rate constants of the exponential curves

52 References

1. Vaupel, P. & Mayer, A. Hypoxia in cancer: Significance and impact on clinical outcome. Cancer and Metastasis Reviews (2007). doi:10.1007/s10555-007-9055- 1 2. Vaupel, P., Mayer, A. & Höckel, M. Tumor Hypoxia and Malignant Progression. Methods Enzymol. (2004). doi:10.1016/S0076-6879(04)81023-1 3. Semenza, G. L. Hypoxia-inducible factors in physiology and medicine. Cell (2012). doi:10.1016/j.cell.2012.01.021 4. Chan, D. A. & Giaccia, A. J. Hypoxia, gene expression, and metastasis. Cancer and Metastasis Reviews (2007). doi:10.1007/s10555-007-9063-1 5. Mak, P. et al. ERβ Impedes Prostate Cancer EMT by Destabilizing HIF-1α and Inhibiting VEGF-Mediated Snail Nuclear Localization: Implications for Gleason Grading. Cancer Cell (2010). doi:10.1016/j.ccr.2010.02.030 6. Yang, M.-H. et al. Direct regulation of TWIST by HIF-1alpha promotes metastasis. Nat. Cell Biol. (2008). doi:10.1038/ncb1691 7. Hiraga, T., Kizaka-Kondoh, S., Hirota, K., Hiraoka, M. & Yoneda, T. Hypoxia and hypoxia-inducible factor-1 expression enhance osteolytic bone metastases of breast cancer. Cancer Res. (2007). doi:10.1158/0008-5472.CAN-06-2355 8. Luo, D., Wang, J., Li, J. & Post, M. Mouse Snail Is a Target Gene for HIF. Mol. Cancer Res. (2011). doi:10.1158/1541-7786.MCR-10-0214 9. Zhang, W. et al. HIF-1α promotes epithelial-mesenchymal transition and metastasis through direct regulation of ZEB1 in colorectal cancer. PLoS One (2015). doi:10.1371/journal.pone.0129603 10. Copple, B. L. Hypoxia stimulates hepatocyte epithelial to mesenchymal transition by hypoxia-inducible factor and transforming growth factor-beta-dependent mechanisms. Liver Int. (2010). doi:10.1111/j.1478-3231.2010.02205.x 11. Chen, J., Imanaka, N., Chen, J. & Griffin, J. D. Hypoxia potentiates Notch signaling in breast cancer leading to decreased E-cadherin expression and increased cell migration and invasion. Br. J. Cancer (2010). doi:10.1038/sj.bjc.6605486 12. Sahlgren, C., Gustafsson, M. V., Jin, S., Poellinger, L. & Lendahl, U. Notch signaling mediates hypoxia-induced tumor cell migration and invasion. Proc. Natl. Acad. Sci. (2008). doi:10.1073/pnas.0802047105 13. Spivak-Kroizman, T. R. et al. Hypoxia triggers hedgehog-mediated tumor-stromal interactions in pancreatic cancer. Cancer Res. (2013). doi:10.1158/0008- 5472.CAN-11-1433 14. Lei, J. et al. Hedgehog signaling regulates hypoxia induced epithelial to mesenchymal transition and invasion in pancreatic cancer cells via a ligand- independent manner. Mol. Cancer (2013). doi:10.1186/1476-4598-12-66 15. Zhang, H. et al. HIF-1-dependent expression of angiopoietin-like 4 and L1CAM mediates vascular metastasis of hypoxic breast cancer cells to the lungs. Oncogene (2012). doi:10.1038/onc.2011.365 16. Semenza, G. L. Defining the role of hypoxia-inducible factor 1 in cancer biology

53 and therapeutics. Oncogene (2010). doi:10.1038/onc.2009.441 17. Schindl, M. et al. Overexpression of hypoxia-inducible factor 1α is associated with an unfavorable prognosis in lymph node-positive breast cancer. Clin. Cancer Res. (2002). 18. Kronblad, Å., Jirström, K., Rydén, L., Nordenskjöld, B. & Landberg, G. Hypoxia inducible factor-1α is a prognostic marker in premenopausal patients with intermediate to highly differentiated breast cancer but not a predictive marker for tamoxifen response. Int. J. Cancer (2006). doi:10.1002/ijc.21676 19. Giatromanolaki, A. et al. c-erbB-2 related aggressiveness in breast cancer is hypoxia inducible factor-1α dependent. Clin. Cancer Res. (2004). doi:10.1158/1078-0432.CCR-04-1068 20. Generali, D. et al. Hypoxia-inducible factor-1α expression predicts a poor response to primary chemoendocrine therapy and disease-free survival in primary human breast cancer. Clin. Cancer Res. (2006). doi:10.1158/1078-0432.CCR-05- 2690 21. Bos, R. et al. Levels of hypoxia-inducible factor-1α independently predict prognosis in patients with lymph node negative breast . Cancer (2003). doi:10.1002/cncr.11246 22. Dales, J. P. et al. Overexpression of hypoxia-inducible factor HIF-1α predicts early relapse in breast cancer: Retrospective study in a series of 745 patients. Int. J. Cancer (2005). doi:10.1002/ijc.20984 23. Metallo, C. M. et al. Reductive glutamine metabolism by IDH1 mediates lipogenesis under hypoxia. Nature (2012). doi:10.1038/nature10602 24. Wise, D. R. et al. Hypoxia promotes isocitrate dehydrogenase-dependent carboxylation of -ketoglutarate to citrate to support cell growth and viability. Proc. Natl. Acad. Sci. (2011). doi:10.1073/pnas.1117773108 25. Pavlova, N. N. & Thompson, C. B. The Emerging Hallmarks of Cancer Metabolism. Cell Metabolism (2016). doi:10.1016/j.cmet.2015.12.006 26. Semenza, G. L. HIF-1: upstream and downstream of cancer metabolism. Current Opinion in Genetics and Development (2010). doi:10.1016/j.gde.2009.10.009 27. Ebert, B. L., Firth, J. D. & Ratcliffe, P. J. Hypoxia and mitochondrial inhibitors regulate expression of glucose transporter-1 via distinct cis-acting sequences. J. Biol. Chem. (1995). doi:10.1074/jbc.270.49.29083 28. Denko, N. C. Hypoxia, HIF1 and glucose metabolism in the solid tumour. Nature Reviews Cancer (2008). doi:10.1038/nrc2468 29. Kim, J. W., Tchernyshyov, I., Semenza, G. L. & Dang, C. V. HIF-1-mediated expression of pyruvate dehydrogenase kinase: A metabolic switch required for cellular adaptation to hypoxia. Cell Metab. (2006). doi:10.1016/j.cmet.2006.02.002 30. Papandreou, I., Cairns, R. A., Fontana, L., Lim, A. L. & Denko, N. C. HIF-1 mediates adaptation to hypoxia by actively downregulating mitochondrial oxygen consumption. Cell Metab. (2006). doi:10.1016/j.cmet.2006.01.012 31. Lu, C. W., Lin, S. C., Chen, K. F., Lai, Y. Y. & Tsai, S. J. Induction of pyruvate dehydrogenase kinase-3 by hypoxia-inducible factor-1 promotes metabolic switch and drug resistance. J. Biol. Chem. (2008). doi:10.1074/jbc.M803508200 32. Saez, I. et al. Neurons have an active glycogen metabolism that contributes to tolerance to hypoxia. J. Cereb. Blood Flow Metab. (2014).

54 doi:10.1038/jcbfm.2014.33 33. Vigoda, A., Mamedova, L. K., Shneyvays, V., Katz, A. & Shainberg, A. Glycogen metabolism in rat heart muscle cultures after hypoxia. Mol. Cell. Biochem. (2003). doi:10.1023/A:1027302021428 34. Zois, C. E., Favaro, E. & Harris, A. L. Glycogen metabolism in cancer. Biochemical Pharmacology (2014). doi:10.1016/j.bcp.2014.09.001 35. Pelletier, J. et al. Glycogen Synthesis is Induced in Hypoxia by the Hypoxia- Inducible Factor and Promotes Cancer Cell Survival. Front Oncol (2012). doi:10.3389/fonc.2012.00018 36. Brown, J. M. & Wilson, W. R. Exploiting tumour hypoxia in cancer treatment. Nature Reviews Cancer (2004). doi:10.1038/nrc1367 37. Berg, Jeremy, M., Tymoczko, J. L., Gatto, G. J. & Stryer, L. Biochemistry 8th Edition. Biochemistry (2015). doi:10.1007/978-3-8274-2989-6 38. Beurel, E., Grieco, S. F. & Jope, R. S. Glycogen synthase kinase-3 (GSK3): Regulation, actions, and diseases. Pharmacology and Therapeutics (2015). doi:10.1016/j.pharmthera.2014.11.016 39. Rousset, M., Fogh, J. & Zweibaum, A. Presence of glycogen and growth–related variations in 58 cultured huma tumor cell lines of various tissue origins. Cancer Res. (1981). 40. Favaro, E. et al. Glucose utilization via glycogen phosphorylase sustains proliferation and prevents premature senescence in cancer cells. Cell Metab. (2012). doi:10.1016/j.cmet.2012.10.017 41. Mole, D. R. et al. Genome-wide association of hypoxia-inducible factor (HIF)- 1alpha and HIF-2alpha DNA binding with expression profiling of hypoxia-inducible transcripts. J. Biol. Chem. (2009). doi:10.1074/jbc.M901790200 42. Shen, G. M., Zhang, F. L., Liu, X. L. & Zhang, J. W. Hypoxia-inducible factor 1- mediated regulation of PPP1R3C promotes glycogen accumulation in human MCF-7 cells under hypoxia. FEBS Lett. (2010). doi:10.1016/j.febslet.2010.09.040 43. Pescador, N. et al. Hypoxia promotes glycogen accumulation through hypoxia inducible factor (HIF)-mediated induction of glycogen synthase 1. PLoS One (2010). doi:10.1371/journal.pone.0009644 44. Rao, G. S., Murray, S. & Ethier, S. P. Radiosensitization of human breast cancer cells by a novel ErbB family receptor tyrosine kinase inhibitor. Int. J. Radiat. Oncol. Biol. Phys. (2000). doi:10.1016/S0360-3016(00)01358-4 45. Woods Ignatoski, K. M. & Ethier, S. P. Constitutive activation of pp125(fak) in newly isolated human breast cancer cell lines. Breast Cancer Res. Treat. (1999). doi:10.1023/A:1006135331912 46. Tracey, W. R. et al. Cardioprotective effects of ingliforib, a novel glycogen phosphorylase inhibitor. Am. J. Physiol. Heart Circ. Physiol. (2004). doi:10.1152/ajpheart.00652.2003

55 CHAPTER 3

A Platform for Artificial Intelligence Based Identification of the Extravasation

Potential of Cancer Cells into the Brain Metastatic Niche 2

Summary

Brain metastases are the most lethal complication of advanced cancer; therefore, it is critical to identify when a tumor has the potential to metastasize to the brain. There are currently no interventions that shed light on the potential of primary tumors to metastasize to the brain. We constructed and tested a platform to quantitatively profile the dynamic phenotypes of cancer cells from aggressive triple negative breast cancer cell lines and patient derived xenografts (PDXs), generated from a primary tumor and brain metastases from tumors of diverse organs of origin. Combining an advanced live cell imaging algorithm and artificial intelligence, we profile cancer cell extravasation within a microfluidic blood-brain niche (µBBN) chip, to detect the minute differences

2 Megan A. Altemus1*, C. Ryan Oliver1,2*, Trisha M. Westerhof1, Hannah Cheriyan1, Xu Cheng1, Michelle Dziubinski1, Zhifen Wu1, Joel Yates1, Aki Morikawa1, Jason Heth3, Maria G. Castro3,4, Brendan M. Leung1,2,5, Shuichi 2,6 1 Takayama , Sofia D. Merajver . A platform for Artificial Intelligence based identification of the extravasation potential of cancer cells into the brain metastatic niche. Lab on a Chip 19(7), (2019)

1Department of Internal Medicine, University of Michigan, Ann Arbor, MI 48109, USA. 2Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI 48109, USA. 3Department of Neurosurgery, University of Michigan, Ann Arbor, MI 48109, USA. 4Department of Cell and Developmental Biology, University of Michigan, Ann Arbor, MI 48109, USA. 5Current affiliation: Department of Applied Oral Sciences and School of Biomedical Engineering, Dalhousie University, Nova Scotia, Canada 6Current affiliation: The Wallace H Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory School of Medicine, Atlanta GA 30332 USA. *These authors have contributed equally to this work.

56 between cells with brain metastatic potential and those without with a PPV of 0.91 in the

context of this study. The results show remarkably sharp and reproducible distinction

between cells that do and those which do not metastasize inside of the device

Introduction

Brain metastatic spread of cancer is the most lethal event in cancer progression.

Approximately 15% of all breast cancer patients develop a brain metastatic lesion, making it the most frequent tissue of origin of brain metastases in women. Brain metastases as a result of breast cancer are increasing in incidence due to improved imaging technologies leading to increased detection and better primary tumor management which allows more time for metastases to develop1–5. While there have

been significant advances in the development of targeted therapies for some metastatic

breast cancers (e.g. anti-estrogen and anti-HER2 drugs), systemic therapy currently has a limited role in the treatment of brain metastasis6. Moreover, there is a lack of

predictive tools with clinically relevant metrics to predict if subpopulations of the

patient’s primary tumor cells will metastasize to the brain. Because of these challenges,

we propose a platform which could be used in a precision medicine approach to identify

the likelihood of brain metastases arising from primary lesions. We posed that artificial

intelligence could identify cancer cells which exhibited a brain metastatic phenotype

using accurate 3D measurement of their behavior in an ex vivo BBB model (Figure

1.1)7.

Three-dimensional measurement of each cancer cell in a live patient’s tumor

micro-environment would be ideal. However, current technology such as MRI is unable

57 to meet this need because it is both expensive and lacks single cell fidelity (0.2 mm x

0.2 mm x 1.2 mm resolutions for 7 Tesla MRI from Siemens specification sheet).

Therefore, the current practice is to biopsy the suspected tumor and a pathologist scans individual slices from the sample, each layer only a few microns thick8. An experienced pathologist can identify cancers and even cancer cells with reasonable accuracy9.

However, it is tedious and there is a large variation among pathologist based on experience10. Moreover, this approach is focused on the question of identifying a tumor or metastasis already grown and present at the biopsy location. There is no method to identify the probability of a cell to migrate across the patient’s blood brain barrier in the future. It is unknown how many cells in a tumor have this capacity, but it is thought to only be a small percentage, thus the importance of identifying them. It then follows that it is important to sample a large number of cells from the patient’s tumor with high fidelity and reproducibility to detect minute differences that relate to the probability and potential to metastasize the brain11. Such a technical challenge indicates a need for methods to capture measurements of the morphologic phenotype of live cancer cells in

3D from an ex vivo micro-environment representing tissue to which they metastasize, such as the BBB. This approach differs from murine models which are largely slow to metastasize and whose brain micro-environments differ significantly from humans12–14.

We solve this challenge by the use of confocal imaging combined with mesh-based tomography of cancer cell phenotypes in a published BBB organ on a chip model15–18.

Finally, the visual differences between cancer cells that can metastasize to the brain and those that cannot are subtle. Trained professionals may have difficulty telling them apart in many cases resulting in delayed treatment9. It is known that treatment

58 early in disease progression is critical to positive outcomes highlighting an opportunity

for improvement19. Artificial intelligence has already been shown to be effective in 2D

pathology and we pose that if combined with 3D confocal tomography of an ex vivo

blood brain barrier it could be trained to reliably identify the minute differences between

cells with metastatic potential and those without.

Thus, in our approach (Figure 1.1) we combine a BBB on a chip with advanced

imaging software (confocal tomography) to improve the quantitation and reliability of the

measurements of cellular dynamic phenotypes and features as the cells traverse the

BBB (Figure 3.2 C-D). Using this platform, we characterize the migratory and

proliferative phenotypes of cancer cells with varying degrees of brain metastatic

potential as well as cells from cancer patient samples with known metastatic potential.

These results when combined with artificial intelligence (AI), lead to a model which can

be used to predict the metastatic potential of cancer cells.

Results and Discussion

Brain seeking breast cancer cell line reveals a distinct μBBN phenotypic pattern

Having confirmed the endothelium’s barrier function by small molecule exclusion

and comparing it to previously published models (Figure 3.2 F-G), we profiled breast

cancer cell behavior in the μBBN device to identify important phenotypic features. After

a confluent monolayer of hCMEC/D3-DsRed cells was established, GFP-expressing

MDA-MB-231 cells (triple-negative breast cancer), MDA-MB-231-BR cells (brain- seeking subclone of MDA-MB-231), or MCF10A cells (normal-like breast epithelium) were seeded onto the top chamber. After 24 or 48 hr, the entire μBBN device was

59 imaged via confocal microscopy (See methods) (Figure 3.2 C-E) to measure the final stages in the metastatic cascade.

Figure 3.3 A-B show examples of diverse morphologies of representative cells in the device, by cell line, at 24 hr and 48 hr, respectively20. The four parametric variables chosen for this study were used to characterize the cancer cells behavior. Percent volume extravasated is the percentage (0-100%) of the cell’s volume that has passed through the barrier plane. For this metric, MCF10A and MDA-MB-231 cells had a substantial proportion of cells that were less than 50% extravasated (59% and 57%, respectively). In contrast, the MDA-MB-231-BR had 32-35% greater proportion of cells >

50% extravasated (Figure 3.3 C, Table 3.1). After 48 hr, unlike the MCF10A cells both cancer cell lines bulk populations had extravasated more than 50% (87% and 95%, respectively). However, only the MDA-MB-231-BR cells had a large sub-population

100% extravasated (14% compared to < 1% for all other cell lines). The distribution of percent volume extravasated was statistically significantly different across all cell lines at both time points according to a Kolmogorov-Smirnov test (all p-values < 8e-5) (Table

3.1).

Distance extravasated (Figure 3.3 C, Table 3.1) is the distance in μm between the center of a cell in the stromal chamber and the endothelial cell layer. One way to analyze the cellular behavior is by binning the distances the cells extravasated and studying the percentage of cells near or in the endothelium (Zone 1: < 50 μm, Green), migratory cells (Zone 2: > 50 - 100 μm, Yellow), and cells attracted to the astrocytes

(Zone 3: > 100 μm, Red). At 24 hr after seeding, all cancer cell lines tended to remain close to the plane in Zone 1 (MCF10A: 36%, MDA-MB-231: 26%, MDA-MB-231-BR:

60 74%). The cancer cell lines had 3-fold as many cells in Zone 2. However, at 48 hr, a

subpopulation of MDA-MB-231-BR cells (Zone 2: 11%, Zone 3: 12%) moved ~100 μm

into the bottom chamber (max = 190.6 μm). Extravasated subpopulations are not

observed in the parental MDA-MB-231 (1.6% in Zones 2-3) or the normal-like MCF10A

(0.3% in Zones 2-3). The distribution of distances cancer cells traversed differed

significantly across all cell lines at both time points, by the Kolmogorov-Smirnov test (all

p-values < 8e-06) (Table 3.1).

The morphologies of the cells in the device were quantified by the sphericity of cells extravasated, with 1.0 being the sphericity of a perfect sphere (Figure 3.3 E, Table

3.1) (see Methods). Sphericity is thought to be related to RhoA and RhoC expression and is a result of their role in migratory phenotypes. At 24 hr, the MCF10A cells were least spherical (0.44), and the MDA-MB-231-BR cells were most spherical (0.73). Over

48 hr post-seeding in the μBBN chip, the distribution of sphericity of MCF10A (0.47) and

MDA-MB-231 cells remained approximately the same, while the MDA-MB-231-BR cells became less spherical (0.52).

The cell populations detected within the device were separated into two subsets to assess if the extravasated subpopulation of cells (> 90% extravasated through the barrier) differed in size from those that interact with but did not traverse through the

BBB. Illustrating this point, the MDA-MB-231-BR cells that extravasated were 58% smaller than those that did not (Figure 3.3, Table 3.2).

Cancer and normal cell lines characterized by the system revealed a significant dynamic range of behaviors when encountering the BBB. Morphological differences of the brain metastatic cells measured over time suggest that cytoskeletal plasticity may

61 contribute to successful extravasation. Given that MDA-MB-231-BR cells extravasated

significantly between 24 and 48 hr (Figure 3.3 C-D) and also exhibited a distinct shift from high to low sphericity (Figure 3.3 E), these data strongly support that the cytoskeletal plasticity necessary to adopt a spherical shape during extravasation enables the brain-seeking cells to traverse the endothelial layer more easily, and once within the brain niche, the cells become elongated to initiate colonization. This method of spherical extravasation has been reported previously by Allen et al. 2017 in contrast to the elongated extravasation of leukocytes21. It is possible that smaller cells are better

able to traverse the endothelial layer, as the cytoskeleton would have less distortion in

shape and fewer tight junctions would need to be disrupted. The changes in sphericity

observed in our system are consistent with a report by Sanz-Moreno et al. in which they

note the role of actomyosin contractility as a means used by neoplastic cells to squeeze

through voids in a 3D matrix22. We find the propensity of the brain-seeking tumor cells to transition from spherical to non-spherical cell shapes to be a distinctive feature of the population that successfully extravasates through the membrane.

PDX-derived brain metastatic and primary tumor cells display differential phenotypic behaviors

It was unknown if patient-derived xenografts (PDXs) would survive and colonize the BBB niche. To demonstrate applicability of the system to patient’s cells, we measured the differential behavior of various first-generation patient-derived-xenografts

(PDXs). We profiled triple negative breast cancer (TNBC) from a primary tumor site

(PDX9040C1), as well as triple-negative breast (PDXbrC1), lung (PDXLuC1), ovarian

(PDXOvC1), and tongue PDX’s (PDXTonC1) developed from brain metastatic sites of

62 those diverse primary tumor types. Figure 3.4 A-B shows examples of morphologies

(low and high sphericity) of representative cells in the device for each cancer cell type.

The primary breast cancer PDX (PDX9040C1) had 87% and 79% of cells that were less than 50% extravasated after seeding at 24 and 48 hr, respectively. However, in contrast, a much higher proportions of extravasated cells were observed from the

PDXs derived from brain metastatic sites: 66% for breast PDXbrC1, 75% for lung

PDXLuC1, 58% for ovarian PDXOvC1, and 92% for tongue PDXTonC1. Of the brain metastatic breast cancer PDX cells, more than 50% had extravasated at 24 hr and ~4/5

(82%) had extravasated by 48 hr (Figure 3.4 C). The mean, median, and standard deviation of percent volume extravasated were calculated for each PDX and are reported in Table 3.3. The distribution the % extravasated for each PDX sample differed significantly (p-values < 0.05e-7) (Table 3.3).

The distance the PDX cancer cells extravasated into the device (Figure 4.4 D) also differed between PDX samples, with PDXbrC1and PDXTonC1 samples extravasating deepest into the device at 129 μm and 145 μm, respectively, at 24 hr.

This is 2.8-fold deeper than the primary (PDX9040C1) sample. Using the zone measurements defined above, PDX9040C1 had 34% of cells in Zone 1 at 24 hr. Only

0.3% of these cells traveled beyond Zone 1. In sharp contrast, the brain metastatic

PDXbrC1 and PDXTonC1 had a 4.6-fold and 59.0-fold larger proportion of cells beyond

Zone 1 and had 0.7% and 1.4% of cells that travelled into Zone 3. No PDX9040C1 cells travelled into Zone 3 at 24 or 48 hr. Moreover, in the PDXbrC1 samples, 87% of cells extravasated across the membrane in contrast to 35% in the PDX9040C1. PDXOvC1 did not migrate into the bottom chamber (> 99%), remaining instead clustered near the

63 endothelial layer. The distribution of cancer cell positions was significantly different

between the brain metastatic and primary tumor PDX samples, with p < 0.05.

The morphologies of the PDX cells in the device were also measured by calculating the sphericity index of cells in the device over time (Figure 3.4 E). Similar to the previous results, the primary breast cells (PDX9040C1) maintained their sphericity level between 24 and 48 hr (0.52 and 0.51). The brain metastatic breast cells

(PDXbrC1) shifted from more spherical to less spherical between 24 and 48 hr (0.64 to

0.53). Of the other sites of origin, lung (PDXLuC1) was the least spherical (0.53) while ovarian (PDXOvC1) was the most (0.61). Like the breast cell lines, the distribution of sphericity of PDXbrC1 cells showed two distinct populations, unlike the cells of

PDX9040C1. Moreover, cells that have traversed into the brain stromal like space

(those which have a percent volume extravasated > 90%) showed decreased sphericity, except for the primary tumor cells. The volumes of cells that traversed the barrier had smaller volumes (Figure 3.4 F, Table 3.4). For example, PDX9040C1 and PDXLuC1 that traversed the barrier were on average 59% and 80% smaller than the rest of the population.

First, we observed that the PDX cells from various primary locations survive and thrive in the in vitro human blood brain niche system which has not been shown previously. Primary human cells are known to be more sensitive to their environment and it has not been verified that an in vitro system can produce viable cultures.

Moreover, the fact that the patient primary breast and breast brain metastasis mimic the behavior of the cancer cell line in vitro is remarkable. Finally, it is important to observe that the brain metastatic PDX cells show differential phenotypes that suggest data will

64 be needed from many primary tumors to fully train an AI system to work across cancer

types.

Brain metastatic cancer cells degrade the endothelial barrier

We measured each cell line in the μBBN device for nine days to observe if the

cancer cells developed into pre-colonization clusters and how they interacted with the

cellular components of the μBBN device. Figure 3.5 shows that unlike normal-like breast

cells which allow the endothelium to continue to proliferate and become more dense,

cancer cells drastically degrade the endothelial barrier over time. Moreover, brain-

seeking MDA-231BR cells reduced the coverage of the endothelium to a greater degree

than the parental MDA-231 cells (Figure 3.5 B). Barrier degradation by cancer cells was

observed concomitant with a marked increase in the number and organization of cancer

cells in the stromal space (Figure 3.5 A and C).

The majority of cells consistently clustered near the endothelium after

extravasation instead of migrating far into the brain microenvironment and this is

consistent with in vivo reports23. From our data, we pose that metastatic cells may

preferentially remain near the barrier to 1) de-regulate the barrier tight junctions and amplify the number of barrier traversing cancer cells, and/or 2) promote angiogenesis and redirection of the endothelial cells to support tumor colonization of the metastatic site. These hypotheses are supported by the observed drastic degradation of the endothelium (Figure 3.5) by the colonizing brain seeking cells. While the data indicate that aggressive cells may remain near the barrier, it does not account for potential repair mechanisms of the barrier that may occur after tumor colonization. This means that the

65 window for therapeutic efficacy, especially for large molecule drugs, may be small, if the barrier will undergo repair after extravasation of the cancer cells.

Comprehensive differential cancer cell behavior in vitro leads to an index of brain metastatic potential

An important goal of this study was to confirm if the phenotypic expression of the cancer cells in the platform would enable the identification of cancer cells that had shown in vivo brain metastatic potential. From the results obtained above we realized that simply crossing the barrier (yes/no) is not the only measurement needed to identify the cells with brain metastatic potential. An analysis, of the data using simple linear models produced poor results, therefore to evaluate the potential of the proposed method to identify cancer cells capable of colonizing the brain based on their behavior in an in vitro BBB, we applied a type of AI termed machine learning to generate an index corresponding to the probability a cell was derived from a brain metastatic clone.

This model was developed in stages. First, we characterize the physical characteristics of cells described above in the process of traversing the BBB and discern the features of cells that traverse the BBB. The predictive power of the algorithm then, depended on its ability to predict if cancer cell subclones not previously encountered would traverse the BBB in a specific way. The traversal behavior of these new cells was confirmed in vivo for cancer cell lines. Ten machine learning algorithms (Table 3.5) were investigated which produce a probabilistic value (0%-1%) of high metastatic potential. The models were trained and cross-validated using the phenotypic behavior of the cells measured previously according to the workflow shown in Figure 3.6 A. The machine learning methods we investigated include Naive Bayes, Random forest, Tree, Logistic

66 regression, k-Nearest-Neighbor (kNN), Stochastic Gradient descent, Neural network and Adaboost (Random forest) (Table 3.5). To compare the methods, we scored them using three statistics: the area under the curve (AUC), accuracy (CA), and the weighted average of precision and recall (F1). When used in tandem these statistics provide insights into the performance and types of errors that the models may make when measuring a cells metastatic potential24. The top three performing methods according to their AUC were the neural network (AUC = 0.95), AdaBoost (Random forest) (AUC =

0.95), and the Random forest (AUC = 0.95) (Figure 3.6 B).

Important translational metrics are the positive predictive value (PPV) and negative predictive value (NPV). Both the PPV and NPV are 0.87 (Table 3.6), which is generally considered excellent in predictive models of a clinical behavior, and of metastatic behavior in particular.

The same models were tested on patient derived xenografts (PDX) taken from brain metastasis to determine if brain metastatic cells could be differentiated from primary tumor cells, under the expectation that the performance would be likely to degrade due to their heterogeneity (Table 3.7). The brain metastatic PDX cells were defined as the metastatic cells and the primary breast cancer PDX cells were defined as a non-brain metastatic control. The data measured in the chip was used to test the ability of the system to predict if cells in the chip belonged to the brain metastatic or non-brain metastatic cell type using the metastatic potential index (Table 3.7). This was done using the training performed on the cancer cell lines and a fresh training set taken from the PDX data. Moreover, this technique could identify between the different tumors of origin. The top three performing methods were the Neural Network (AUC = 0.97)

67 followed by the Random Forest (AUC = 0.96) and AdaBoost (Random forest) (AUC =

0.96) (Figure 3.6 C and Table 3.7). The resulting positive probability value (PPV) and

negative probability value (NPV) are 0.91 and 0.85, respectively (Table 3.8), indeed

defying the a priori prediction that these parameters were likely to decrease, as

heterogeneity increased.

There is a need for robust diagnostics that predict the future occurrence of brain metastasis from breast or other cancers at the time of primary diagnosis. In addition to the study, imaging tools are being tested for this application. Yin et al. has shown that

MRI of brain composition can predict the number of days until brain metastases for non- small cell lung cancer, with accuracy, sensitivity, and specificity of 70%, 75%, and 66%, respectively25. Graesslin et al. reported on a model to predict brain metastasis for

patients with metastatic breast cancer with an area under the curve (AUC) of 0.7426. A

limitation of these methods is that the patient’s brain must have undergone a change

prior to diagnosis and is therefore, already susceptible to metastasis. Thus, a more

direct approach which measures the behavior of a patient’s primary cells in a brain

environment, with increased accuracy and at a lower cost would be attractive. In this

context, the results, support the measurement power of the system to predict the

behavior of the cancer cells themselves. The low cost and quick turnaround time (24-48

hr) of the µBBN device along with the demonstrated ability to identify cancer subclones

that then metastasized to the brain in murine models, make the device an excellent

candidate to meet this technological gap. We acknowledge more study is necessary to

optimize the brain micro-environment in the device which is a known area of active

study in the field27–30. Devices are being developed to include flow, additional cell types

68 and improved barrier function overcoming the limitations of hCMEC endothelial cells.

Regardless, we chose a realistic, simple and well understood model BBB employing the widely used hCMEC that was available at the start of the study to guide development of the imaging algorithm and AI components of the platform and the results are a positive contribution to the field. Moreover, the system identified patient derived cells based on their site of origin and previously known brain metastatic state, even though we tested a limited number of diverse PDXs. Future work expanding the library of PDX cells matched with knowledge of the patient outcomes would be the basis of a clinically usable training set. Taken together, this device could be further developed to identify cancer patients that may require additional screening and, as available, personalized treatment strategies to minimize the probability of brain metastasis.

Materials and Methods

Study design

The primary hypothesis investigated in this study were to (i) verify that cancer cells cultured in vitro mimic known in vivo behavior and (ii) establish that the behavior was distinct to brain metastatic cells and could be exploited as a diagnostic. The experiments presented were designed to compare first established lines with known brain seeking subclones and then to follow up with patient cells from biopsied tumors.

All data presented are the result of three independent biological replicates with three technical replicates performed for each, but each measurement includes hundreds of cells.

69 μm-Blood Brain Niche design and validation

The complex series of processes by which a cancer cell moves from the primary

tumor site to distant sites is known as the metastatic cascade17,31–33. Initially, cancer

cells invade into the tissue around the primary site. The cells then intravasate into the

bloodstream where they may survive in the circulatory system until they adhere to the

endothelium at a distant site. In the case of brain metastasis, the cells extravasate

through the BBB endothelium where they then colonize and grow within the brain

stroma. The μBBN device (Figure 3.2 A-B) was designed to study the late and most definitively clinically impactful steps of the brain metastatic cascade: adherence, extravasation, and colonization. It is based on previously published and accepted models such as the one by Wang et al., Booth et al. and, Marino et al34–37. The µBBN

we designed is composed of two chambers separated by a 5 μm microporous

membrane (Figure 3.2 B and Figure 3.7). The device is fabricated from

Polydimethylsiloxane (PDMS) because of its inertness during cell culture. The upper

chamber mimics the brain vasculature with a lining of human cerebral microvascular

endothelial cells, hCMEC/D3. The lower chamber is filled with hTERT immortalized

normal human astrocytes (NHA) suspended in a type I collagen matrix. The porous

membrane is sized to separate the upper and lower chambers without obstructing the

inlets and outlets. After various pilot trials, we chose 5 µm pore membranes to optimize

three functions: support for the endothelial layer, ease of extravasation, and clarity of

imaging. Micropipette tips are placed in each inlet/outlet port to serve as medium

reservoirs and facilitate easy loading of liquids and cells. The assembled chip is bonded

to a 50 mm x 75 mm glass slide and positioned using an alignment fixture (Figure 3.2

70 C). This chip has four independent sets of chambers imprinted on it for ease of

replication of each experiment.

To use the engineered brain niche to study cancer cell extravasation, we first

validated the barrier function of the human cerebral microvascular endothelial cell

(hCMEC/D3) monolayer in the device. The hCMEC/D3 cells formed a monolayer in the

upper channel over a 3-day period post-seeding. Co-culture of the astrocytes and

endothelial cells required that their respective media be mixed at a 50:50 ratio to

promote simultaneous healthy growth of both cell lines and barrier formation (Figure

3.8). The endothelial barrier permeability was characterized using FITC-dextran exclusion.

Small molecule membrane transfer measurements

We conducted a FITC-Dextran exclusion assay in which 10 kDa FITC-Dextran was introduced to the upper chamber. Diffusion of the dye was measured over 24 hr using the relative fluorescence intensity (RFI) approach (Figure 3.2 F-G). The intensity of the region measured was plotted as a percentage of the normalized intensity of a calibration solution of FITC-DEX38. There was an increase in dye content over the first

hour in the lower chamber, in line with other models in the literature32, with a maximal

RFI of 4% followed by a stable diffusion (Figure 3.2 G). In contrast, the diffusion was

significantly higher, with a maximal RFI of 74% (p < 0.05) when no endothelial layer was

present.

FITC-dextran conjugated dye (Sigma-Aldrich, CAS: 60842-46-8, 5 mg/ml, 10

KDa) was used to assess total integration fluorescence of small molecules across the endothelial layer in the device. Images were taken using an inverted fluorescent

71 microscope with a 10X objective. The fluorescent signal at the interface between the

upper and lower chamber was measured at 1, 5, 10, 20, 60 and 1440 minutes. Three

images were taken at the interface between the upper and lower chamber at three pre-

determined locations along the device (center, 5 mm from top and 5 mm from bottom)

for three channels and n=3 biological replicates.

Cell culture and reagents

MCF10A (ATCC CRL-10317) cells were maintained in 50:50 DMEM:F12 medium

(Corning 10-090-CV) supplemented with 5% horse serum, 10 μg/ml insulin, 0.5 μg/ml hydrocortisone, 0.02 μg/ml epidermal growth factor, 0.1 μg/ml cholera toxin, and 5 mL antibiotic-antimycotic (Gibco 15240062). MCF10A cells were stained using Cell Tracker

Green according to the manufacturer’s protocol. MDA-MB-231 and MDA-MB-231-BR-

GFP cells were obtained from Patricia Steeg, PhD and were maintained in DMEM

(Corning 10-013-CV), supplemented with 10% FBS and 5mL antibiotic-antimycotic

(Gibco 15240062). hCMEC/D3 (EMD Millipore SCC066) cells were maintained in EGM-

2 medium (Lonza CC-3162). MDA-MB-231-GFP fluorescent cells were created by transfecting MDA-MB-231 cell with empty vector pLLEV-GFP lentivirus. hCMEC/D3-

DsRed fluorescent cells were created by transfecting hCMEC/D3 cells with empty vector pLL3.7-dsRed lentivirus. Normal human astrocytes (NHA) were obtained from

Lonza (CC-2565) and were immortalized using lentiviral induced hTERT expression and were maintained in AGM media (Lonza CC-3186). Lentivirus was created using pLOX-

TERT-iresTK lentiviral vector obtained from Addgene (12245) and packaging vectors psPAX2 and pMD2.G also obtained from Addgene (12260 and 12259) in HEK-293T cells. Cells were grown at 37°C in 5% CO2.

72 Patient-derived xenografts

Human tumor tissue was collected at the University of Michigan under approved

IRB protocols at University of Michigan. Animal studies were performed under approved

University of Michigan institutional animal care and use committee (IACUC) protocols.

Freshly resected human tumor tissue was immediately and directly implanted into eight-week-old NSG mice (Jackson lab). Breast tumor tissue was implanted into the mammary fat pad and non- breast tumor tissue was implanted into both flanks. Tumor growth was monitored once a week and were harvested when the tumor size reached

0.6-0.7 cm.

Excised tumors were cut into 2-4 mm pieces and dissociated to single cell suspensions using a gentle MACS dissociation platform (Miltenyi Biotec) according to the manufacturer’s protocol. The single cell suspensions were counted using a hematocytometer, then resuspended in 80µL of 1X PBS, pH 7.4 supplemented with

0.5% BSA and 20µL of mouse cell depletion cocktail (cat# 130-104-694) containing magnetically labeled antibodies per 107 cells. Samples were incubated at 4oC for 15 minutes then applied to a LS column to deplete the magnetized mouse cells from the purified human tumor cells. The purified human tumor cells were then stained with Cell

Tracker Green according to the manufacturer’s protocol, counted using a hematocytometer before use.

Live subject statement

All human tissues were collected from patients treated at Michigan Medicine, who provided written, in-person informed consent under a protocol approved by the

University of Michigan Institutional Review Board (IRB).

73 Seeding microfluidic device

The bottom chamber of the devices was seeded with 1x106 NHA suspended in a solution of 1mL of 3mg/mL PureCol type I bovine collagen with 128 μL 0.8M NaHCO3 and 40 μL 10X high-glucose (250 mM) DMEM and incubated at 37°C for one hour. Top chamber was coated with 2% growth-factor reduced Matrigel in AGM and incubated at

37°C for one hour. hCMEC cells were seeded in the top chamber by pipetting 30 μL of cells at 1 million cells per mL in 50:50 AGM/EGM-2 in both inlet and outlet twice, with 15 min between each inoculation. Devices were then incubated at 37°C and 5% CO2, changing media in both chambers every 12 hrs. Cancer and normal-like cells were seeded into the top chamber inlet of a confluent device at a density of 3x104 cells.

Measurement of the cell attributes using confocal tomography

Extracting meaningful data from images of the organ-on-a-chip device poses a major challenge. Thus, we developed software based on the Visualization tool kit (VTK) library, termed confocal tomography, to measure the extravasation behavior of cancer cells relative to the endothelial layer20. In brief, the algorithm fits a 2D surface to the endothelial layer using the centers of the cells (centroids) as points on the plane. Then, each cancer cell is converted from the confocal z-stack into a 3D object. Each 3D cancer cell object is then compared to the endothelial layer plane to measure the cell’s morphological and functional metrics. To calculate the percent volume of the cell that has extravasated across the endothelial plane the cells 3D mesh is cut using a Boolean operation and the resulting hole closed to divide the cell into two bodies. A representative result of these measurements is shown in Figure 3.2 E.

74 Sphericity was measured by comparing the Volume and Area of the cell, the ratio of

which may deviate from a perfect sphere according to the equation below:

(6 ) = 1 2 3 3 𝜋𝜋 𝑉𝑉 𝑆𝑆 Statistical analysis 𝐴𝐴

Analysis was performed using Python, R/R Studio. Comparison of populations of

cells were made using a Smirnov-Kolmogorov test and Kruskal-Wallis Rank Sum Test

with a p-value test at 0.05. The double-sided t-test was used to compare the means of

the RFI results for µBBN chips with an endothelial and no-endothelial layer. The means

of the cell area coverage in Figures 3.4 and 3.5 were tested using a Kruskal-Wallis test

and if significant also by a pairwise comparison Tukey and Kramer (Nemenyi) test.

Artificial intelligence machine learning algorithm

Binary classification was performed in Orange (Figure 3.6 A). The data was

filtered to remove bad measurements, defined as those that failed a Boolean operation

or giving parametric variable values outside of known bounds (-100-200, 0-1 and 0-

2000). Using the cell lines as an example the MDA-MB-231-BR-GFP cells were labeled as brain metastatic and the remaining cell lines were labelled as non-brain metastatic.

The features used to classify cells included all parametric variables. The data was sampled into a training (80%) and test set (20%). The training set was stratified and cross-validated using 10 folds against each model/classifier. The models/classifiers studied included Neural Network, Naïve Bayes, Random forest, Tree, Logistic regression, kNN, Stochastic Gradient Descent and AdaBoost latched to Random forest.

After training the data the test data was used to score the performance of the model by classifying the cells in the chip according to the probability, they were in the brain

75 metastatic cell line from 0 to 1. The model performance was ranked according to the area under the curve (AUC) of the ROC, the accuracy and the F1 score. When used in tandem, these statistics provide insights into the performance and types of errors that the models may make when measuring a cells metastatic potential24.

Breast (cancer) cell lines

All laboratory Breast cell lines (MCF10A, MDA-MB-231) used in this research have been authenticated via ATCC’s STR profiling service prior to fluorescent labelling.

Conclusion

In conclusion, we presented the development and potential of a platform designed to identify the subtle phenotypic differences between cancer cells that show brain metastatic behavior and those that do not, based on their behavior in a brain like tumor micro-environment with potential for translation to the clinic as a brain metastatic predictive diagnostic given additional study. The method was validated by measuring the extravasation and metastatic events of breast cancer cells and PDX cancer cells in the device. Future work will expand the library of patient samples used to train the system to improve its clinical applicability. Additional work will use the device to evaluate the molecular determinants of the migration, survival of metastatic cancer cells and to test the efficacy of potential new treatments on metastatic cancer within the brain niche.

76 Figures

Figure 3.1 Overview of method. The concept we demonstrate is to culture cells from a cell line or patient in an in vitro BBB device allowing the cancer cells to undergo late stage metastatic processes. The result is then imaged via confocal tomography after 24 and 48 hrs. The confocal z-stack is converted to a 3D mesh and single cell phenotypic measurements are calculated such as the distance from the endothelial layer and shape. The feature measurements are evaluated by a trained artificial intelligence (AI) model to determine if the cells have a high, medium, or low brain metastatic potential index.

77 Figure 3.2 Microfluidic BBNiche device design to study brain metastatic process. (A) Schematic of µBBN device. (B) Image of the µBBN device indicating top channel, bottom chamber, and porous membrane. (C) Confocal images of µBBN device are analyzed using 3D rendered objects. hCMEC/D3 endothelial layer in μBBN device forms a barrier between the top and bottom chambers. Image of the µBBN from top of device showing the area being imaged (black rectangle). (D) 3D-rendered µBBN device 24 hrs post-seeding with MDA-MB-231-BR cells from top-down. Scale bar = 1000 μm. (E) Side view of the channel with insets showing cancer cells at various stages of traversing endothelium. Dashed line shows plane fit to the endothelium. Circles highlight representative cells. Scale bar = 200 μm. (F) μBBN device at 0, 1, and 24 hrs after addition of 10 kDa FITC-Dextran for device with hCMEC/D3 cells and device without. Black box indicates area of chip shown in images. (G) Quantification of permeability after addition of FITC-Dextran in devices with and without hCMEC/D3 layer. Values are the average of three areas per time point with n=3 biological replicates. Error bars indicate standard deviation. *p<0.05

78 Figure 3.3 Differences in extravasation and morphology of brain-seeking cells compared to non- brain-seeking cell in the µBBN device analyzed using confocal tomography. (A) Representative images of morphology of cells in device. Scale bar = ~25 μm. (B) Representative images of morphology of cells in device at 48 hr. Scale bar = ~25 μm. (C) Violin plot of percent total volume of cells extravasated through endothelial plane at 24 (orange) and 48 hrs (light blue). Dashed lines represent quartiles, longer dashed line represents the mean. (D) Strip plot of distance in μm of cancer cell centroids from plane at 24 and 48 hr. (E) Violin plot of sphericity of cancer cells in µBBN device at 24 and 48 hr. Sphericity ranges from 1:Spherical to 0:Not Spherical. (F) Plot of volume of each cell line in µBBN device in voxels for cells <90% and >90% extravasated. *** p<0.0001 for 24hr timepoint, *** p<0.0001 for 48hr timepoint.

79 Figure 3.4 Profiling of patient derived xenografts in µBBN device. (A) Representative images of morphology of PDX cells in device with low and high sphericity. Scale bar = 25 μm. (B) Violin plot of percent total volume of cells extravasated through plane for PDXs. (C) Strip plot of distance in μm of PDX cell center from the endothelial layer. (D) Violin plot of sphericity of PDX cells in µBBN device. (E) Box and whisker plot of volume PDX cells in µBBN device in voxels for cells <90% and >90% extravasated *** p-value < 0.05.

80 Figure 3.5 Cancer cell interaction with the μBBN endothelium (A) Comparison of cell line degradation of the endothelium over 7 d (168 hrs) after adding the breast cells. The left and right panel shows cancer cells (green) and endothelial cells (red) 24 hrs and 7 d after seeding the cancer cells respectively. (B) Comparison of the endothelial coverage at 24 hrs and 168 hrs by cell line. (C) Comparison of cancer cell coverage as a percentage of area in the channel at 24 hrs and 168 hr. Scale bar = 400 µm.

81

Figure 3.6 Accurate identification of brain metastatic potential in μBBN device. (A) Schematic of model development process. This process is divided into a training (grey) and testing (light blue) phase that outputs a probability a cell has a high metastatic potential (green). (B) Receiver operating characteristic (ROC) curve for the AdaBoost classification methods for cell lines. Upper left corner indicates good classification accuracy. (C) ROC curve for the AdaBoost classification methods for PDX samples.

82

Figure 3.7 Mask of four-channel device

Figure 3.8 Optimizing media composition for co-culture

83

Tables

24 hr

Distance Cell % Extravasated Cell line p-value extravasated p-value Sphericity p-value count by volume (μm) MCF10A 671 36.45±34.83 - -14.74±34.23 - 0.45±0.14 - MDA-MB- 3780 43.44±30.90 8e-05 -0.77±33.87 6e-09 0.60±0.14 2e-16 231 MDA-MB- 6459 56.91±20.12 2e-16 6.91±14.03 8e-06 0.74±0.13 2e-16 231-BR

48 hr Distance Cell % Extravasated Cell line p-value extravasated p-value Sphericity p-value count by volume (μm) MCF10A 2951 41.55±22.30 - -2.18±20.62 - 0.47±0.14 - MDA-MB- 11125 42.25±28.47 2e-13 0.89±25.73 2e-16 0.57±0.17 0.068 231 MDA-MB- 5290 51.35±31.60 2e-16 18.34±6.33 2e-16 0.52±0.11 3e-06 231-BR

Table 3.1 Summary of metrics measured for each cell line.

Cell count Volume < Cell count Volume > Cell line p-value p-value < 90% 90% (voxels) > 90% 90% (voxels) MCF10A 472 503.6±502.5 - 43 370.2±361.9 - MDA-MB-231 3068 429.0±435.6 2e-16 387 285.9±284.5 1e-09 MDA-MB-231-BR 5986 167.7±201.8 2e-15 440 89.2±137.0 0.011

Table 3.2 Cell counts and volumes by location.

84 24 hr

% Distance Cell p- p- PDX Extravasated extravasated Sphericity count value value by volume (μm) Primary breast 389 34.63±15.41 - -4.22±14.36 - 0.57±0.11 - (PDX9040C1) TNBC 517 53.12±21.61 2e-16 10.90±16.39 2e-16 0.65±0.14 2e-16 (PDXbrC1) Tongue 2e-16 6e-08 1423 78.30±21.26 2e-16 30.79±24.27 0.59±0.14 (PDXTonC1) Lung 2e-16 1e-07 677 70.90±31.37 2e-16 -16.68±14.57 0.53±0.11 (PDXLuC1) Ovarian 2e-16 7e-03 239 51.75±16.86 2e-16 -29.50±6.05 0.61±0.13 (PDXOvC1) 48 hr

% Distance Cell p- p- PDX Extravasated extravasated Sphericity count value value by volume (μm) Primary breast 1e-14 520 39.18±22.62 8e-06 4.30±17.69 0.55±0.10 0.02 (PDX9040C1) TNBC 2e-16 322 72.34±33.17 2e-16 17.59±18.59 0.58±0.15 2e-08 (PDXbrC1)

Table 3.3 Summary of metrics measure for each PDX type.

24 hr

Cell count Volume < Cell count Volume > PDX p-value p-value < 90% 90% (voxels) > 90% 90% (voxels) Primary breast 324 519.2±479.7 - 2 474.1±684.1 - (PDX9040C1) TNBC 417 589.0±472.8 1e-04 19 304.2±328.6 0.63 (PDXbrC1) Tongue 455 687.8±500.1 6e-15 460 538.5±493.3 0.70 Lung 377 485.2±322.1 9e-11 275 420.3±299.6 0.50 Ovarian 199 613.5±419.0 6e-08 7 233.4±93.6 0.75

48 hr Cell count Volume < Cell count Volume > PDX p-value p-value < 90% 90% (voxels) > 90% 90% (voxels) Primary breast 448 460.2±433.8 4e-05 29 110±78.9 0.24 (PDX9040C1) TNBC 173 440.1±414.3 3e-06 127 297.6±311.2 0.67 (PDXbrC1)

Table 3.4 PDX type counts and volumes by location.

85

Method AUC CA F1 Precision Recall Neural Network 0.951 0.871 0.871 0.867 0.876 AdaBoost 0.950 0.876 0.876 0.874 0.877 Random Forest 0.946 0.874 0.874 0.873 0.875 Tree 0.917 0.843 0.839 0.857 0.823 kNN 0.868 0.787 0.776 0.817 0.739 Logistic Regression 0.848 0.779 0.783 0.769 0.796 Naïve Bayes 0.833 0.751 0.757 0.740 0.774 SGD 0.774 0.774 0.778 0.763 0.795

Table 3.5 Comparison of methods to classify cancer cells by brain met potential.

Predicted 0 1

0 698 102 800∑

Actual 1 100 700 800

798 802 1600 ∑ Table 3.6 Confusion matrix for random forest.

86

Method AUC CA F1 Precision Recall Neural Network 0.972 0.881 0.878 0.910 0.847 Random Forest 0.964 0.888 0.887 0.900 0.875 AdaBoost 0.957 0.881 0.879 0.899 0.861 Tree 0.954 0.867 0.865 0.884 0.847 Logistic Regression 0.897 0.832 0.831 0.843 0.819 Naïve Bayes 0.896 0.846 0.849 0.838 0.861 kNN 0.882 0.818 0.814 0.838 0.792 SGD 0.861 0.860 0.853 0.906 0.806

Table 3.7 Comparison of methods to classify breast PDX cancer cells by brain met potential.

Predicted 0 1

0 93 7 100 ∑

Actual 1 8 92 100

101 99 200 ∑

Table 3.8 Confusion matrix for random forest using PDX cancer cells.

87 References

1. Aversa, C. et al. Metastatic breast cancer subtypes and central nervous system metastases. Breast 23, 623–628 (2014). 2. Kennecke, H. et al. Metastatic behavior of breast cancer subtypes. J. Clin. Oncol. 28, 3271–3277 (2010). 3. Niikura, N., Saji, S., Tokuda, Y. & Iwata, H. Brain metastases in breast cancer. Jpn. J. Clin. Oncol. 44, 1133–1140 (2014). 4. Nayak, L., Lee, E. Q. & Wen, P. Y. Epidemiology of brain metastases. Curr. Oncol. Rep. 14, 48–54 (2012). 5. Lin, N. U. & Winer, E. P. Brain metastases: The HER2 paradigm. Clinical Cancer Research 13, 1648–1655 (2007). 6. Lockman, P. R. et al. Heterogeneous blood-tumor barrier permeability determines drug efficacy in experimental brain metastases of breast cancer. Clin. Cancer Res. 16, 5664–5678 (2010). 7. Olson, E. M. et al. Clinical outcomes and treatment practice patterns of patients with HER2-positive metastatic breast cancer in the post-trastuzumab era. Breast (2013). doi:10.1016/j.breast.2012.12.006 8. Binnig, G., Huss, R. & Schmidt, G. Tissue phenomics: Profiling cancer patients for treatment decisions. Tissue Phenomics: Profiling Cancer Patients for Treatment Decisions (2018). doi:10.1201/9781351134279 9. Williams, B. J. et al. Digital pathology for the primary diagnosis of breast histopathological specimens: an innovative validation and concordance study on digital pathology validation and training. Histopathology (2018). doi:10.1111/his.13403 10. Jha, S. & Topol, E. J. Adapting to artificial intelligence: Radiologists and pathologists as information specialists. JAMA - Journal of the American Medical Association (2016). doi:10.1001/jama.2016.17438 11. Schmidt, R. L. et al. Risk-benefit analysis of sampling methods for fine-needle aspiration cytology: A mathematical modeling approach. Am. J. Clin. Pathol. (2013). doi:10.1309/AJCPEAKR4MO2GQBO 12. Meuwissen, R. et al. Induction of small cell lung cancer by somatic inactivation of both Trp53 and Rb1 in a conditional mouse model. Cancer Cell 4, 181–189 (2003). 13. Kato, M. et al. The herbal medicine Sho-saiko-to inhibits growth and metastasis of malignant melanoma primarily developed in ret-transgenic mice. J. Invest. Dermatol. 111, 640–644 (1998). 14. Oberheim, N. A. et al. Uniquely Hominid Features of Adult Human Astrocytes. J. Neurosci. 29, 3276–3287 (2009). 15. Shah, K. K., Yang, L. & Abbruscato, T. J. In vitro models of the blood-brain barrier. Methods Mol. Biol. (2012). doi:10.1007/978-1-61779-452-0_29 16. Prabhakarpandian, B. et al. SyM-BBB: a microfluidic blood brain barrier model. Lab Chip 13, 1093–1101 (2013). 17. Booth, R. & Kim, H. Characterization of a microfluidic in vitro model of the blood-

88 brain barrier (μBBB). Lab Chip 12, 1784 (2012). 18. Ostrom, Q. T. et al. CBTRUS statistical report: Primary brain and other central nervous system tumors diagnosed in the United States in 2009-2013. Neuro. Oncol. (2016). doi:10.1093/neuonc/now207 19. Niikura, N. et al. Treatment outcomes and prognostic factors for patients with brain metastases from breast cancer of each subtype: A multicenter retrospective analysis. Breast Cancer Res. Treat. (2014). doi:10.1007/s10549-014-3090-8 20. Schroeder, W. J., Avila, L. S. & Hoffman, W. Visualizing with VTK: A tutorial. IEEE Comput. Graph. Appl. (2000). doi:10.1109/38.865875 21. Allen, T. A. et al. Angiopellosis as an Alternative Mechanism of Cell Extravasation. Stem Cells (2017). doi:10.1002/stem.2451 22. Sanz-Moreno, V. et al. Rac Activation and Inactivation Control Plasticity of Tumor Cell Movement. Cell (2008). doi:10.1016/j.cell.2008.09.043 23. Eichler, A. F. et al. The biology of brain metastases-translation to new therapies. Nature Reviews Clinical Oncology (2011). doi:10.1038/nrclinonc.2011.58 24. Sokolova, M. & Lapalme, G. A systematic analysis of performance measures for classification tasks. Inf. Process. Manag. (2009). doi:10.1016/j.ipm.2009.03.002 25. Yin, G. et al. Predicting brain metastases for non-small cell lung cancer based on magnetic resonance imaging. Clin. Exp. Metastasis (2017). doi:10.1007/s10585- 016-9833-7 26. Graesslin, O. et al. Nomogram to predict subsequent brain metastasis in patients with metastatic breast cancer. J. Clin. Oncol. (2010). doi:10.1200/JCO.2009.24.6314 27. Biemans, E. A. L. M., Jäkel, L., de Waal, R. M. W., Kuiperij, H. B. & Verbeek, M. M. Limitations of the hCMEC/D3 cell line as a model for Aβ clearance by the human blood-brain barrier. J. Neurosci. Res. (2017). doi:10.1002/jnr.23964 28. Choi, Y. P. et al. Cancer-associated fibroblast promote transmigration through endothelial brain cells in three-dimensional in vitro models. Int. J. Cancer 135, 2024–2033 (2014). 29. Van Horssen, R., Galjart, N., Rens, J. A. P., Eggermont, A. M. M. & Ten Hagen, T. L. M. Differential effects of matrix and growth factors on endothelial and fibroblast motility: Application of a modified cell migration assay. J. Cell. Biochem. (2006). doi:10.1002/jcb.20994 30. Song, J. W. et al. Computer-controlled microcirculatory support system for endothelial cell culture and shearing. Anal. Chem. (2005). doi:10.1021/ac050131o 31. Wirtz, D., Konstantopoulos, K. & Searson, P. C. The physics of cancer: The role of physical interactions and mechanical forces in metastasis. Nature Reviews Cancer (2011). doi:10.1038/nrc3080 32. Brown, J. A. et al. Recreating blood-brain barrier physiology and structure on chip: A novel neurovascular microfluidic bioreactor. Biomicrofluidics (2015). doi:10.1063/1.4934713 33. Takeshita, Y. et al. An in vitro blood-brain barrier model combining shear stress and endothelial cell/astrocyte co-culture. J. Neurosci. Methods 232, 165–172 (2014). 34. Wang, Y. I., Abaci, H. E. & Shuler, M. L. Microfluidic blood–brain barrier model provides in vivo-like barrier properties for drug permeability screening. Biotechnol.

89 Bioeng. (2017). doi:10.1002/bit.26045 35. Marino, A. et al. A 3D Real-Scale, Biomimetic, and Biohybrid Model of the Blood- Brain Barrier Fabricated through Two-Photon Lithography. Small (2018). doi:10.1002/smll.201702959 36. Esch, M. B., Post, D. J., Shuler, M. L. & Stokol, T. Characterization of In Vitro Endothelial Linings Grown Within Microfluidic Channels . Tissue Eng. Part A (2011). doi:10.1089/ten.tea.2010.0371 37. Chen, M. B., Whisler, J. a, Jeon, J. S. & Kamm, R. D. Mechanisms of tumor cell extravasation in an in vitro microvascular network platform. Integr. Biol. (Camb). 5, 1262–71 (2013). 38. Shi, L., Zeng, M., Sun, Y. & Fu, B. M. Quantification of Blood-Brain Barrier Solute Permeability and Brain Transport by Multiphoton Microscopy. J. Biomech. Eng. (2013). doi:10.1115/1.4025892

90 CHAPTER 4

Conclusions and Future Directions

Breast cancer has long had the highest incidence rate of any cancer type among women, and has the second highest mortality, despite significant advancements in breast cancer early detection and treatment. These advancements have mainly been in the treatment of ER/PR+ breast cancers with hormone-based therapies and HER-2+ breast cancers with anti-HER-2 therapies like trastuzumab. Additionally, with the widespread use of screening using mammography, breast cancer has been significantly down-staged in North American and Western Europe, regions with the highest incidence worldwide. Due to these advances, four-year breast cancer-specific survival rates in these subtypes have risen to 90-93% for ER/PR+ and 83-90% for HER-2+ cancers1. However, overall survival for cancers negative for ER/PR and with normal

HER-2, termed triple-negative breast cancer (TNBC) is at a relatively low 77%1. In addition to differences in treatment, this disparity in survival is also due in part to increased rates of metastasis at a relatively short term of 1-5 years post-diagnosis2,3.

According to the SEER breast cancer statistics, five-year relative survival rates drop from 85-99% for local/regional disease to 27% for disease present at a distant site4.

This highlights the need for new treatments and therapies aimed at preventing metastases, and the need to identify those patients most at risk for developing them. In this thesis work, we investigated hypoxic glycogen stores as a potential fuel to drive the

91 initiation of metastases and developed a microfluidic device as a tool to investigate and

identify the propensity of cells to form brain metastases, the most lethal event in cancer

progression.

Glycogen, the main glucose storage molecule of the body, has been previously

shown to accumulate under hypoxic conditions5,6. This includes the hypoxic cores of solid tumors7,8. Additionally, hypoxia and HIF-1α stabilization activate transcriptional

programs that initiate EMT and other metastatic phenotypes9–12. We hypothesized that

this accumulation of glycogen can be used to fuel the invasion and migration induced by

hypoxia. And indeed, consistent with our hypothesis, we found that the loss of glycogen

phosphorylase, and thus the ability to utilize glycogen stores, correlated with inhibited

migration in MCF-7 breast cancer cells and invasion in MDA-MB-231 TNBC cells.

Interestingly, it was the loss of the brain isoform of glycogen phosphorylase, PYGB, but

not the liver isoform, that was responsible for blocking glycogen utilization and the

decrease in migration and invasion. This is a novel finding in that previous studies have

focused on the effect of PYGL loss on proliferation and premature senescence in

cancer cells13.

In future research, we will attempt to recapitulate these in vitro findings in vivo.

One approach we intend to follow is to implant our shPYG cells in orthotopic xenograft

models and monitor the size of the tumor and the development of metastases in the

bone, brain, liver, and lung. Once harvested, we will also evaluate the glycogen

accumulation in the primary tumor and any metastases via PAS staining. If the results of

these experiments hold true to the in vitro predictions reported here, we will next test the

efficacy of glycogen phosphorylase inhibitors such as ingliforib on reducing metastasis

92 in cell line and patient-derived xenograft models of various aggressive breast cancers14.

Additionally, in order to confirm it is the prevention of glycogen utilization and not another, off-target effect such as glycogen macromolecules spatially hindering cell motility, we will knockdown glycogen synthase and confirm that similar phenotypes occur. Glycogen phosphorylase inhibitors, in combination with standard-of-care treatment, could prove a promising new therapy to prevent metastasis in patients with high levels of high levels of tumor glycogen. While the expression of common biomarkers such as hormone receptors and HER-2 did not correlate with hypoxic glycogen accumulation in our studies, PAS staining of tumor biopsies could be easily used to determine which patients have a tumor with increased glycogen accumulation and therefore may benefit from anti-glycogen phosphorylase treatment. Currently there are few targeted therapies against TNBC, which, as stated above, has a lower survival rate compared to other subtypes of breast cancer. Survival in breast cancer is highly correlated to the formation of metastases at distant sites. If anti-glycogen therapies prove to be efficacious at preventing or reducing metastases, this could be a valuable treatment option for women with high risk of developing metastases, such as those with

TNBC. Importantly, inhibition of the brain isoform may be less toxic than inhibition of the liver or muscle isoforms which are important for everyday energy needs. The brain isoform of glycogen phosphorylase is mainly important in fetal development and during extreme to provide a minimum of glucose to brain cells15–17.

As well as investigating glycogen’s role in cancer cell phenotypes that drive the first steps of the metastatic cascade, we developed a tool to study cancer cell adherence, extravasation, and colonization of the brain as the secondary site. Around

93 20% of TNBC and HER-2+ breast cancers metastasize to the brain2,18. The brain is the deadliest secondary site of breast cancer metastases with median survival of around 5-

12 months depending on HER-2 status19. Current treatments are merely palliative, highlighting a need to better understand the brain metastatic process in order to develop new treatments aimed at preventing brain metastases altogether, since poor drug penetration in the brain will likely be difficult to overcome. Moreover, we aim to be able to predict which patients are at increased risk of developing brain metastases so more aggressive treatment options can be pursued before they occur. In order to do this, we need an accurate, quick, and reproducible model. In vivo mouse models and existing in vitro blood-brain barrier models do not meet this need. Mouse models are expensive, slow, and the development of metastases at a cellular level cannot be monitored in real-

time. Additionally, while similar to human, the mouse brain microenvironment differs

especially when it comes to the phenotypes of brain cells present, like astrocytes20. In

vitro models can utilize human brain cells, but current models are too simple in their

composition and dynamics, and do not provide an accurate dynamic 3D

microenvironment21–23. To combat these issues, we developed a microfluidic device to mimic the human blood-brain niche and study the mechanisms of brain metastasis formation.

Using confocal z-stack images of the blood-brain niche device, we measured

different aspects of the cancer cells in relation to the endothelium such as distance

extravasated, the percent volume of the cancer cell extravasated, and the sphericity of

the cancer cells. We validated the device using a brain seeking TNBC cell line

compared to the parental line and a normal-like epithelial cell line. The brain-seeking

94 TNBC cell line had a subset of cells that were able to extravasate and colonize the

bottom chamber of the brain-niche device, whereas the parental and normal-like cell

lines had few cells that were able to colonize the niche. Once the device was validated,

we were able to utilize it to measure the different aspects of brain metastatic and

primary site PDX cells as they interacted with the device. Using this information, we

trained an artificial intelligence (AI) algorithm to accurately predict whether a subset of

cells are brain metastatic or non-metastatic. With further validation and ease-of-use

modifications this device could be used with direct from patient samples in order to

determine whether the patient’s tumor has the potential to become brain metastatic.

This could inform treatment options and lead to prevention of brain metastatic lesions

and possible increase in survival.

Besides potential clinical applications, future work will focus on increasing the

complexity and accuracy of the device, investigating how the brain cells within the niche

affect the behavior of the cancer cells and vice-versa, and isolating the colonizing subset of the cancer cells in the device for further analysis. In order to increase the accuracy and reproducibility of the device, we are working on implementing microglia and pericyte cells within the device in addition to the astrocytes, as well as finding an epithelial cell line that more accurately represents the blood-brain barrier. While

hCMEC/D3 cells have been traditionally used in such applications, there have been

documented problems with their accuracy in reproducing the tight junctions found within

the brain microvasculature24. We will also add constant flow to the top channel of the

device to stimulate tight-junction formation and increase the accuracy of the

microenvironment. Additionally, we will investigate how the cancer cells and brain niche

95 cells interact within the device. During the course of the work detailed earlier in this

dissertation, we found the devices with brain-seeking cells exhibited a marked degradation of the endothelial layer that was not evident with the parental or normal-like cells. We will investigate the cell-cell interaction differences as well as any secretions exclusive to the brain-seeking cells that may account for this observation. Also, we are investigating the secretions of the microglia and astrocytes within the niche and how they may contribute to the selective migration of the brain-seeking cells through the device. And lastly, we will isolate the cancer cells that were able to colonize the niche from the other cells within the device and perform further downstream analysis, such as

NanoString mRNA panels, to identify what transcriptional changes may separate these cells from the bulk of the cells that were unable to colonize the device.

While both of these projects seem separate in nature, the natural next step is a convergence of the two. As detailed in the first chapter of this work, different secondary sites of breast cancer metastases all present with very different metabolic environments. While the ability to utilize glycogen contributes to invasiveness and motility, could glycogen storage and utilization also be an important metabolic adaptation for colonizing certain distant sites? Two of the most likely distant sites for which glycogen storage may serve useful to breast cancer would be the liver, which naturally stores and utilizes glycogen and is hypoxic and nature, and the brain, a glucose-poor site where the ability to store and use glucose at a later point may be invaluable to cancer cell survival. It would be very interesting to observe how breast cancer cells with inhibited glycogen metabolism behave in the blood-brain niche device, does the ability to use glycogen confer a survival or invasive advantage within this

96 microenvironment? Additionally, the microfluidic device could be cultured with hepatic endothelium and hepatocytes suspended in a collagen gel to form a liver niche device.

This device could be cultured in a normoxic or hypoxic microenvironment to accurately mimic in vivo liver oxygen levels and utilized to determine the possible advantage that glycogen utilization might provide to breast cancer cells within the liver microenvironment.

In conclusion, the work presented in this dissertation explores different aspects of the breast cancer metastatic cascade. We determined that the ability to utilize glycogen stores, mobilized by PYGB, contribute to the mobility and invasiveness of breast cancer cells, pointing towards a potential new target for breast cancer metastasis prevention.

We also developed a microfluidic device and analysis that accurately predicts the brain metastatic potential of known cell lines and PDX cells. These contributions could pave the way for new metastasis prevention treatments in breast cancer and a clinical test to determine patients at risk for brain metastasis, thus combatting the deadliest complication of breast cancer and potentially increasing patient survival.

97 References

1. Howlader, N., Cronin, K. A., Kurian, A. W. & Andridge, R. Differences in breast cancer survival by molecular subtypes in the United States. Cancer Epidemiol. Biomarkers Prev. (2018). doi:10.1158/1055-9965.EPI-17-0627 2. Kennecke, H. et al. Metastatic behavior of breast cancer subtypes. J. Clin. Oncol. 28, 3271–3277 (2010). 3. Dent, R. et al. Triple-negative breast cancer: Clinical features and patterns of recurrence. Clin. Cancer Res. (2007). doi:10.1158/1078-0432.CCR-06-3045 4. Howlader, N. et al. SEER Cancer Statistics Review, 1975-2014, National Cancer Institute. Bethesda, MD, http://seer.cancer.gov/csr/1975_2012/, based on November 2016 SEER data submission, posted to the SEER web site, April 2017. Http://Seer.Cancer.Gov/Csr/1975_2012/ (2016). 5. Saez, I. et al. Neurons have an active glycogen metabolism that contributes to tolerance to hypoxia. J. Cereb. Blood Flow Metab. (2014). doi:10.1038/jcbfm.2014.33 6. Vigoda, A., Mamedova, L. K., Shneyvays, V., Katz, A. & Shainberg, A. Glycogen metabolism in rat heart muscle cultures after hypoxia. Mol. Cell. Biochem. (2003). doi:10.1023/A:1027302021428 7. Zois, C. E., Favaro, E. & Harris, A. L. Glycogen metabolism in cancer. Biochemical Pharmacology (2014). doi:10.1016/j.bcp.2014.09.001 8. Pelletier, J. et al. Glycogen Synthesis is Induced in Hypoxia by the Hypoxia- Inducible Factor and Promotes Cancer Cell Survival. Front Oncol (2012). doi:10.3389/fonc.2012.00018 9. Chan, D. A. & Giaccia, A. J. Hypoxia, gene expression, and metastasis. Cancer and Metastasis Reviews (2007). doi:10.1007/s10555-007-9063-1 10. Mak, P. et al. ERβ Impedes Prostate Cancer EMT by Destabilizing HIF-1α and Inhibiting VEGF-Mediated Snail Nuclear Localization: Implications for Gleason Grading. Cancer Cell (2010). doi:10.1016/j.ccr.2010.02.030 11. Yang, M.-H. et al. Direct regulation of TWIST by HIF-1alpha promotes metastasis. Nat. Cell Biol. (2008). doi:10.1038/ncb1691 12. Hiraga, T., Kizaka-Kondoh, S., Hirota, K., Hiraoka, M. & Yoneda, T. Hypoxia and hypoxia-inducible factor-1 expression enhance osteolytic bone metastases of breast cancer. Cancer Res. (2007). doi:10.1158/0008-5472.CAN-06-2355 13. Favaro, E. et al. Glucose utilization via glycogen phosphorylase sustains proliferation and prevents premature senescence in cancer cells. Cell Metab. (2012). doi:10.1016/j.cmet.2012.10.017 14. Tracey, W. R. et al. Cardioprotective effects of ingliforib, a novel glycogen phosphorylase inhibitor. Am. J. Physiol. Heart Circ. Physiol. (2004). doi:10.1152/ajpheart.00652.2003 15. Richter, F., Böhme, H. J. & Hofmann, E. Developmental changes of glycogen phosphorylase b in rat tissues. Biomed. Biochim. Acta 42, 1229–35 (1983). 16. Sato, K., Satoh, K., Sato, T., Imai, F. & Morris, H. P. patterns of glycogen

98 phosphorylase in rat tissues and transplantable hepatomas. Cancer Res. 36, 487–95 (1976). 17. Suh, S. W. et al. Astrocyte glycogen sustains neuronal activity during hypoglycemia: studies with the glycogen phosphorylase inhibitor CP-316,819 ([R- R*,S*]-5-chloro-N-[2-hydroxy-3-(methoxymethylamino)-3-oxo-1- (phenylmethyl)propyl]-1H-indole-2-carboxamide). J. Pharmacol. Exp. Ther. 321, 45–50 (2007). 18. Aversa, C. et al. Metastatic breast cancer subtypes and central nervous system metastases. Breast 23, 623–628 (2014). 19. Niikura, N., Saji, S., Tokuda, Y. & Iwata, H. Brain metastases in breast cancer. Jpn. J. Clin. Oncol. 44, 1133–1140 (2014). 20. Oberheim, N. A. et al. Uniquely Hominid Features of Adult Human Astrocytes. J. Neurosci. 29, 3276–3287 (2009). 21. Czupalla, C. J., Liebner, S. & Devraj, K. In vitro models of the blood-brain barrier. Methods Mol. Biol. (2014). doi:10.1007/978-1-4939-320-7_34 22. Jeon, J. S., Zervantonakis, I. K., Chung, S., Kamm, R. D. & Charest, J. L. In Vitro Model of Tumor Cell Extravasation. PLoS One 8, (2013). 23. Martins-Green, M., Petreaca, M. & Yao, M. Chapter 8 An Assay System for In Vitro Detection of Permeability in Human ‘Endothelium’. Methods Enzymol. 443, 137–153 (2008). 24. Biemans, E. A. L. M., Jäkel, L., de Waal, R. M. W., Kuiperij, H. B. & Verbeek, M. M. Limitations of the hCMEC/D3 cell line as a model for Aβ clearance by the human blood-brain barrier. J. Neurosci. Res. (2017). doi:10.1002/jnr.23964

99