TUMOR GROWTH AND METASTASIS IN TRIPLE NEGATIVE BREAST 
CANCER 
 
PART I: EXAMINING THE EFFECTS OF DIABETIC HYPERGLYCEMIA-
INDUCED MATRIX GLYCATION ON TUMOR PROGRESSION  
 
PART II: PARSING APART INTRATUMOR HETEROGENEITY IN THE 
CONTEXT OF THE MIGRATORY PHENOTYPE AND METASTATIC 
POTENTIAL 
 
 
 
A Dissertation 
Presented to the Faculty of the Graduate School 
of Cornell University 
In Partial Fulfillment of the Requirements for the Degree of 
Doctor of Philosophy 
 
 
 
by 
Lauren Ann Hapach 
May, 2021
 
 
 
 
 
 
 
 
 
 
 
 
 
© 2021 Lauren Ann Hapach
 
TUMOR GROWTH AND METASTASIS IN TRIPLE NEGATIVE BREAST 
CANCER 
 
Lauren Ann Hapach, Ph.D. 
Cornell University 2021 
 
Triple negative breast cancer remains a poorly understood clinical subtype of breast 
cancer associated with increased aggression, limited treatment options, and ultimately, 
worsened patient outcomes. Cancer progression can be divided conceptually into two 
main phases: primary tumorigenesis, the critical initial phase which requires numerous 
cellular aberrations, and secondary metastasis, the later phase where cancer cells gain 
additional mutations that enable dissemination from an established tumor and 
colonization of disparate tissue sites in the body. In Part I of this thesis, the role of 
diabetic hyperglycemia in triple negative breast cancer primary tumor growth in the 
context of non-enzymatic glycation is investigated. While it is well established that 
Type II diabetes mellitus is associated with increased incidence and severity of 
numerous comorbidities including triple negative breast cancer, there has been little 
work done to elucidate the nature of these disease interactions. This work delineates a 
novel role for diabetic hyperglycemia-mediated non-enzymatic glycation due in 
increased triple negative breast cancer tumor growth. In Part II of this thesis, the 
metastatic potentials of highly migratory and weakly migratory subpopulations of triple 
negative breast cancer cells are compared. Contrary to the common paradigm in cancer 
migration research that adoption of a migratory phenotype is associated with metastatic 
ability, we found that the weakly migratory subpopulation was highly metastatic in an 
E-cadherin dependent manner compared to weakly metastatic, highly migratory 
subpopulation. This work highlights the importance of parsing apart intratumor 
 
heterogeneity and indicates that migratory ability is not always correlated with 
metastatic potential.
 
BIOGRAPHICAL SKETCH 
Lauren Ann Hapach was born in Pittsburgh, Pennsylvania on May 8, 1992 to Gary and 
Barbara Hapach, and grew up in New Brighton, Pennsylvania along with her younger 
siblings, Stephen and Allison. She graduated from New Brighton High School with third 
highest honors in 2010. Lauren attended University of Pittsburgh and graduated summa 
cum laude with a B.S. in Bioengineering, a minor in Neuroscience, and a certificate in 
Conceptual Foundations of Medicine in 2014. In August 2014, Lauren entered the Ph.D. 
program in Biomedical Engineering at Cornell University. She joined the lab of Cynthia 
A. Reinhart-King and began her research by studying the role of intratumor 
heterogeneity in the context of cancer cell migration and metastasis. She was awarded 
a prestigious National Science Foundation Graduate Research Fellowship in 2016 and 
received her M.S. in Biomedical Engineering in 2016. During her time as a Cornell 
student, Lauren also won a Biomedical Engineering Society Extended Abstract Award. 
Lauren defended her thesis in December of 2020 and completed her Ph.D. in the spring 
of 2021. When not in lab, Lauren enjoys cooking and baking, traveling, yoga, and rock 
climbing.   
  
v 
 
 
 
 
 
 
 
 
 
 
 
 
For my father, Gary S. Hapach, 
What a long, strange trip it’s been
vi 
 
ACKNOWLEDGMENTS 
I would first and foremost like to thank Dr. Cynthia Reinhart-King for allowing me to 
complete my thesis work in her lab, first at Cornell, and now, at Vanderbilt University. 
Cindy, thank you for taking me on when I had no relevant experience to apply in cancer 
mechanobiology. You quickly equipped me with the tools, expertise, and support 
needed to excel. When my first big experiment produced completely counter-intuitive 
results, you patiently allowed me to validate the results and then supported me as I went 
on what often felt like a wild goose chase, keeping me on track as I probed numerous 
avenues, and never stifling my enthusiasm to solve the mystery. Under your mentorship, 
I gained the ability to critically think, solve complex problems, design experiments 
independently, and effectively communicate results, all while developing an extensive 
scientific toolkit. I will forever be grateful for the time spent in your lab.  
 
I would also like to extend my appreciation to my committee members Drs. Warren 
Zipfel and Richard Cerione for their support and supervision of this work. Their 
expertise and insight greatly enriched this project. 
 
The Reinhart-King lab was an optimal environment for my development as a scientist. 
This remarkable community of scientists both past and present have left an indelible 
mark on my professional and personal progression. To my predecessors, Drs. Shawn 
Carey, Marsha Lampi-Kowal, Julie Kohn, Joe Miller, Danielle LaValley, and Aniqua 
Rahman-Zaman, thank you all for the guidance, wisdom, and inspiration you have 
imparted. To Prof. Francois Bordeleau, thank you for the mentorship and support, 
especially when I was just starting in the lab, knew absolutely nothing, and was scared 
that I would break everything. Your patience and willingness to both help and challenge 
me were invaluable. I will greatly miss our coffee breaks/scientific discussions/rants 
vii 
 
with Dr. Miller and other fellow caffeine addicts. To the lab member who started this 
“climb” with me back in 2014, Dr. Jacob VanderBurgh, thanks for the years of 
friendship and support. Like a trusty belayer, you have always been there with calm 
encouragement and support any time I felt I was reaching the end of my rope. We made 
it, so “take! .. Ready to lower!” To my contemporaries, Matt Zanotelli, Adam Muñoz, 
Sam Schwager, Paul Taufalele, Wenjun Wang, Jenna Mosier, Andrew Johnson, Neil 
Chada, Kyra Smart, Ismael Ortiz, and Emily Chu, thank you for the years of 
collaboration, support, and comradery. To CRK M.Eng. alumnae, Adeline Chen and 
Stephanie Cheng, thank you for always being there with warm hugs and smiles as well 
as cute cupcakes. To the undergraduates I’ve had the pleasure of mentoring, Devika 
Pokhriyal, Harrison Thomas, Katy Ufford, Isaac Richardson, Mara Rao, Mara Yella, 
and Sarah Platkin, thank you for your contributions to our work and your enthusiasm to 
learn. I was extremely privileged to work with so many brilliant and amiable individuals 
during my time in the Reinhart-King lab, and I cannot wait to see all of the amazing 
things everyone accomplishes as our journeys continue. 
 
While our time was cut short, I am infinitely thankful for those I met while at Cornell. 
Katie Hoff, thank you for making upstate NY much less dreary. To my cohort, 
especially, Drs. Aaron Chiou, Terence Gee, Zhexun Sun, Monet Roberts, Alex Loiben, 
and Mandy Rooney, thanks for being my Cornell family. I’ll forever treasure all of our 
adventures in NYC, cooking/fine dining/exploring waterfalls in Ithaca, and of course, 
game nights. Drs. Hannah Watkins, Dan Cheung, and Jacob VanderBurgh, thank you 
for welcoming me into your BME climbing gang and for making physical fitness and 
experiencing nature actually enjoyable. To the amazing staff at Cornell, especially 
Belinda Floyd, thank you for helping to make my experience there pure magic. I have 
never experienced such a talented, passionate, and caring team of people and will always 
viii 
 
appreciate all of the assistance and support you provided. 
 
Moving mid-Ph.D. to Vanderbilt University allowed me to explore a city I never had on 
my radar, Nashville, TN. Thank you to the Vandy community who welcomed me and 
to my fellow Cornell expats for making the transition less bumpy. Especially, thank you 
to my boyfriend, Dr. Thong M. Cao. Even though you had just helped me move my 
mountains of furniture and other belongings to a new apartment when the move to 
Vanderbilt was announced, you nevertheless helped me to pack everything up and do it 
all again, this time, 845 miles away. Your cool and collected personality was an anchor 
to me during that tumultuous period and continues to be as we navigate new, unexpected 
hurdles such as the COVID-19 pandemic. I can’t wait to finally start my next chapter 
with you, Harley, and Tycho in Boston. 
 
I would be remiss if I did not extend my appreciation to those who inspired me to dream 
of possibilities I never initially envisioned for myself. I would like to sincerely thank 
the EXCEL program at University of Pittsburgh, who taught me the skills needed to 
succeed at the very start of my undergraduate journey and later encouraged me to apply 
to Cornell for graduate studies. I would also like to thank Dr. Nicole Ostrowski, a fellow 
member of the engineering sorority, Phi Sigma Rho, who graciously coached me 
through my first research presentation in Hannover, Germany, and later answered my 
incessant graduate school application questions and provided guidance on my first stab 
at the National Science Foundation Graduate Research Fellowship. Lastly, while my 
time in the Reinhart-King lab did not overlap with hers, Dr. Casey Kraning has provided 
a great deal of encouragement and inspiration needed during my final year. Thanks 
Casey, for helping me see the light at the end of the tunnel. 
 
ix 
 
To my friends, especially Briana Roberts, Luke Joyner, Nina Rudolph, Audrey Lapic, 
Sam Fabyanic, Karmen Maddox, Richard Karpinski, and Ryan and Mara Mooney, I am 
eternally grateful for all of the late night texts/phone calls/skype sessions that 
entertained me while I was working in lab. You all kept me from going crazy, so thank 
you!  
 
To my family who has always inspired and supported me to do my best and know that 
that will do, thank you for instilling self-confidence and resilience into my being from 
day one. Please know that I would have never have made it this far if it were not for all 
of you. I would especially like to thank my mother, Barbara, for always being my 
number one fan, encouraging every interest/hobby I wanted to pursue, buying me every 
book I ever wanted, and for modeling how to live a life with true perseverance, virtue, 
and grace. To my extended family, especially those that comprised the caravan that 
moved me to New York in August of 2014, my grandparents, Hope and Ed, and aunt 
and uncle, Jodi and Big Fun, thank you from the bottom of my heart for all of your 
constant love and support.  
 
 
 
 
 
 
 
 
 
 
x 
 
TABLE OF CONTENTS 
 
Biographical Sketch………………………………………………………...…………v 
Dedication……………………………………………………………………….……vi 
Acknowledgements………………………………………………………..…………vii 
Table of Contents…………………………………………………………..…………xi 
 
Chapter 1 Introduction…………………………………………………………….1 
 
1.1 Introduction to breast cancer…………………………………………..1 
1.2 Intratumor heterogeneity………..……………………………………..5 
1.3 Matrix-mediated link between breast cancer and other diseases……..15 
1.4 In vitro models of cancer metastasis…………………..…………...…20 
1.5 Dissertation organization.…………………………………...………..31 
Chapter 2 Diabetic Hyperglycemia Promotes Primary Tumor Progression  
Through Non-Enzymatic Glycation 
  
2.1 Abstract……………………………………………………………….34 
2.2 Introduction………..………………………………………………….34 
2.3 Materials and Methods………………………………………………..36 
2.4 Results……………………………………………………………...…43 
2.5 Discussion…………………………………………………...………..56 
Chapter 3 Phenotypic Sorting Reveals a Weakly Migratory, Highly Metastatic  
Cell Subpopulation Which Metastasizes in an E-cadherin Dependent  
Manner 
  
3.1 Abstract……………………………………………………………….59 
3.2 Introduction………..………………………………………………….60 
3.3 Materials and Methods………………………………………………..62 
3.4 Results……………………………………………………………...…78 
3.5 Discussion…………………………………………………...…..…..123 
Chapter 4 Phenotypically Sorted Migratory Breast Cancer Subpopulations Exhibit  
Migratory and Metastatic Commensalism 
  
4.1 Abstract……………………………………………………………...124 
4.2 Introduction………..………………………………………………...125 
4.3 Materials and Methods……………………………………………....126 
4.4 Results…………………………………………………………….....130 
4.5 Discussion…………………………………………………...………138 
Chapter 5 Conclusions and Future Directions………………………………….147 
 
5.1 Conclusions…..……………………………………………………...147 
5.2 Future Directions……..……………………………………………...156 
xi 
 
 
Appendix A: Hyperglycemia in MMTV-PyMT mice ………………...……180 
Appendix B: Immunofluorescence staining…………………………………181 
Appendix C: Immunohistochemistry staining………………………………182 
Appendix D: Mouse breast cancer metastasis model………………………..184 
Appendix E: Circulating tumor cell collection……………………………...191 
Appendix F: Lung decellularization……………………………………...…194 
Appendix G: Tumor spheroid model…………………………………...…...197 
 
References……...…………………………………………………………....………203
xii 
 
LIST OF FIGURES 
 
Chapter 2 
 
Figure 2.1. Induction of diabetes mellitus within MMTV-PyMT mice……………....45 
Figure 2.2. Hyperglycemia promotes glycation and increases ECM stiffness……….47 
Figure 2.3. Hyperglycemia increases tumor growth through promoting cell 
proliferation…………………………………………………………………………...49 
Figure 2.4. Diabetes promotes tumor cell EMT……………………………………....51 
Figure 2.5. Glycation inhibition decreases ECM stiffness……………………………53 
Figure 2.6. Glycation inhibition decreases primary tumor progression……...……….54 
 
Chapter 3 
 
Figure 3.1. Heterogeneity of MDA-MB-231 human cancer cell migratory capability is 
heritable and can be sorted based on migration behavior using an in vitro transwell 
migration assay……………………………………………………….……………....79 
Figure 3.2. Altering transwell ECM coating conditions does not alter the migratory 
behaviors of MDA-MB-231 subpopulations……...………………….……………....81 
Figure 3.3. Phenotypically sorted subpopulations show differential metastatic 
potentials in vivo……...………………….……………………………..…………....84 
Figure 3.4. MDA-MB-231 subpopulations have similar proliferation rates in vitro and 
in vivo………………………………………...……………………………………....86 
Figure 3.5. MCF10CA1a cells can be phenotypically sorted and display differential 
metastatic potentials……………………………………..……………………………88 
Figure 3.6. SUM159 cells can be phenotypically sorted and display differential 
metastatic potentials…………………………………..……………………...……….89 
Figure 3.7. Validation of circulating tumor cell collection, endothelial monolayer for 
trans-endothelial assay, and lung decellularization for ex vivo colonization assay….90 
Figure 3.8. RNA sequencing reveals differential EMT gene regulation in 
phenotypically MDA-MB-231 sorted subpopulations. ……...……….……………....92 
Figure 3.9. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted MDA-MB-231 subpopulations. ……...…….……………........94 
Figure 3.10. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted MCF10CA1a subpopulations. ……...………………...............96 
Figure 3.11. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted SUM159 subpopulations.…...……………….………..……....98 
Figure 3.12. RNA sequencing reveals differential cell-cell adhesion gene regulation in 
phenotypically sorted MDA-MB-231 subpopulations.. ….………….……………...100 
Figure 3.13. RNA sequencing reveals differential cell-cell adhesion gene regulation in 
phenotypically sorted MCF10CA1a subpopulations. ……...…………………….....101 
Figure 3.14. RNA sequencing reveals differential cell-cell adhesion gene regulation in 
phenotypically sorted SUM159 subpopulations.…...……………….………..……..102 
 
xiii 
 
Figure 3.15. E-cadherin expression is necessary for metastasis in phenotypically sorted 
subpopulations. .…...……………….………..……...................................................104 
Figure 3.16. E-cadherin expression in MDA subpopulations and additional validation 
of E-cadherin knockdown and addition experiments. .…...……………….…….......106 
Figure 3.17. Representative confocal fluorescence images shows phenotypically sorted 
subpopulations and E-cadherin variants have differential E-cadherin expression.. ...109 
Figure 3.18. E-cadherin expression tunes migration ability and mode in phenotypically 
sorted subpopulations. .…...……………….………..…….........................................112 
Figure 3.19. Subpopulations exhibit differential morphologies, cell-ECM signaling, 
and contractility. .…...……………….………..……..................................................115 
Figure 3.20. Subpopulations possess differential actin cytoskeletal structure and focal 
adhesion activation. .…...……………….………..…….............................................117 
Figure 3.21. Quantitative polarization microscopy shows differential cell contractility 
in migration strands of in vitro tumor spheroid model.. .…...……………….……...119 
Figure 3.22. Circulating tumor cell clusters trend with worsened patient outcome...121  
Chapter 4 
Figure 4.1. Phenotypic sorted cancer cells show differential invasion. ….………....136 
Figure 4.2. Phenotypic sorted cancer cells form in vitro tumor spheroids with 
differential compaction behavior. .…...……………….…………………………….138 
Figure 4.3. Phenotypic sorted cancer cells exhibit differential migration modes in 
tumor spheroid model. .…...……………….………………………………………..140 
Figure 4.4. Co-culture MDAMIX spheroids exhibit leader-follower behavior. …..….141 
Figure 4.5. Phenotypic sorted cancer cells show commensal interactions leading to 
enhanced metastasis in vivo. .…...………………………………………….……….143 
Figure 4.6. Commensal interaction between phenotypically sorted subpopulations 
hypothesized to be mediated by leader-follower behavior. .…...….……….……….146 
 
Chapter 5 
 
Figure 5.1. Regions of vascular infiltration in MDA- en bloc sections.…….………163 
Figure 5.2. Quantification of vascularization in MDA-MB-231 subpopulation primary 
tumors..…...………….………………………………………………………………164 
Figure 5.3. Differential Gene Expression MDA- vs MDA+ (logF2C 1.5)…..………165 
Figure 5.4. Chick chorioallantoic membrane (CAM) model. .…...………….……...167 
Figure 5.5. Anisotropy maps of quantitative polarization microscopy signal from 
Picrosirius red stained MDA-MB-231 subpopulation primary tumor sections……..168 
Figure 5.6. EMT marker staining reveals differential EMT phenotypes between 
MCF10CA1a parental cell line and subpopulations. .…...………….………………169 
Figure 5.7. Principal component analysis of RNA sequencing data for three triple 
negative breast cancer subpopulations. .…...………….…………………………….173 
Figure 5.8. MCF10CA1a subpopulation 1:1 co-culture tumor spheroid model…….178 
 
 
 
 
xiv 
 
LIST OF TABLES 
 
 
Table 3.1. Top differentially regulated GO terms from RNA sequencing of 
phenotypically sorted MDA-MB-231, MCF10CA1a, and SUM159 highly migratory 
and weakly migratory subpopulations.……………… …………….…………….......93 
 
Table 3.2. Patient demographic and clinical outcomes for metastatic cancer patient 
blood samples used in clustering analysis.……...………………….……………......122 
 
Table 3.3. Patient demographic and clinical details for metastatic breast cancer patient 
blood samples used in E-cadherin staining experiment….…………....………….....122 
xv 
 
CHAPTER 1 
INTRODUCTION 
 
Portions of this chapter were published as a review article titled “Engineered models 
to parse apart the metastatic cascade” in Nature Precision Oncology1 
 
1.1 Introduction to breast cancer  
For women, breast cancer is the leading cancer in both frequency of diagnosis and 
cancer-related death2. Additionally, breast cancer is the leading cause of disability-
adjusted life years (DALYS), a metric defined as the number of years lost due to 
disability or early death, in women3,4. Incidence and mortality of breast cancer are 
increasing globally; however, in developed countries, the burden of breast cancer is 
easing while in developing countries, it is accelerating3. These increases in incidence 
can be attributed to a multitude of factors including an expanding and aging population, 
improved treatment of other major illnesses such as cardiovascular disease, increased 
exposure to carcinogens including air pollution, pesticides, and endocrine disrupting 
chemicals, more effective screening and detection, earlier menarche, delayed 
menopause, and lifestyle factors including smoking, alcohol consumption, 
hyperglycemia, obesity, lack of exercise, exposure to light at night, increase in maternal 
age at first pregnancy, and possibly the use of hormonal contraception5–7. In addition to 
these non-genetic factors, approximately 10% of breast cancers can be attributed to 
inherited genetic factors such as the high penetrance BRCA1 and BRCA2 mutations5. 
1 
 
While newly diagnosed cases of breast cancer continue to rise, much has been done to 
ameliorate diagnosis and treatment, which has dramatically improved outcomes for 
breast cancer patients with adequate access to these medical advancements. 
 
1.1.1 Breast cancer subtypes 
Like most cancers, breast cancer is actually a collection of different subtypes united by 
the broadest defining feature of cancer where cells proliferate in an uncontrolled manner 
due to a loss of regulatory control mechanisms8. The differences between the subtypes 
can include the type of cell which is affected and the manner of dysregulation. These 
distinctions are critical as optimal patient treatments and expected outcomes often vary 
according to subtype. In the most recent edition of the World Health Organization 
(WHO) “blue book” on breast cancer classification published in 2019, there are 44 
different carcinomas of the breast currently recognized9. The increasing number of 
classifications from the 10 types of breast carcinoma described in the 1st edition of WHO 
classifications set forth in 1968 are evidence of the amount of progress that has been 
made in clinically relevant characterization of breast cancer9.  
 
Twenty years ago, Perou and Sorlie defined four intrinsic subtypes of breast cancer 
which were classified based on nascent microarrays and gene expression profiling 
techniques10,11. This study marked a conceptual shift in clinical treatment of tumors 
from a largely uniform regimen towards seeking to target specific molecular 
vulnerabilities inherent to the individual tumor. As next generation sequencing 
technologies continue to be developed and used to probe naïve, treated, and recurrent 
2 
 
primary tumor and metastatic samples, improved understanding of the genetic 
underpinnings of breast cancer progression, metastasis, and recurrence shows enormous 
potential to provide more insightful prognoses, personalized treatment plans, and novel, 
targeted therapies. 
 
The surrogate intrinsic subtypes of breast cancer implemented clinically in diagnosing 
breast cancer are classified as: luminal A-like, luminal B-like HER2-, luminal B-like 
HER2+, HER2-enriched (non-luminal), and triple negative breast cancer (TNBC)5. 
These classifications are made based on histological assessment along with 
immunohistochemistry of a biopsy sample for the following markers: oestrogen 
receptor (ER), progesterone receptor (PR), human epidermal growth factor receptor 2 
(HER2), and Ki-67, a marker of proliferation5. When organized along the spectrum of 
tumor aggressiveness as well as prevalence, at one end, Luminal A-like breast cancer is 
the most common (60-70% of cases) and least aggressive, with relatively low Ki-67 
index and a greater degree of normal tissue structure preserved, while at the opposite 
end of the spectrum, TNBC is less common (10-15% of cases) but is considered more 
highly aggressive with less organized histological features and a relatively higher Ki-67 
index5.  
 
TNBC lacks ER, PR, and HER2 expression, which inherently reduces potential 
treatment options since ER-, PR-, and HER2-targeted chemotherapies available to other 
subtypes will be ineffective in these cases5,12. Compounding further problematic 
features along with the relative paucity of viable treatment options, TNBC frequently 
3 
 
becomes resistant to administered chemotherapeutics13. Additionally, compared to other 
breast cancer subtypes, TNBC presents earlier, often in premenopausal women under 
40 years old, and shows a higher likelihood of recurrence and metastasis14,15. African 
American women are disproportionately affected by TNBC, but the reason underlying 
this remains unknown14,15. Regardless of stage or substage, TNBC has significantly 
worsened overall and cause-specific survival compared to other breast cancer 
subtypes16. TNBC is the most lethal subtype of breast cancer, and thus, the work 
outlined in this thesis was performed within the scope of this subtype.  
 
1.1.2 TNBC subtypes 
TNBC is itself extremely heterogeneous, and was initially classified into the following 
six subtypes based on patterns found in gene expression signatures by Lehmann et al.: 
basal-like 1 and 2, immunomodulatory, mesenchymal, mesenchymal-stem-like, and 
luminal androgen receptor subtypes17. With advances in single cell analysis, this 
classification system was later refined to four subtypes: basal-like 1 and 2, 
mesenchymal, and luminal androgen receptor subtypes18. While treatment options 
currently do not take these subtypes into consideration, these characterizations are 
clinically relevant, with significant differences between subtypes to similar neoadjuvant 
treatment regimens18. Additional classifications of TNBC subtypes have been made by 
other groups with the objective for all of these assessments to thoroughly characterize 
and increase our broader understanding of this subtype’s heterogeneity to improve 
patient outcomes19–21.  
 
4 
 
1.2 Intratumor heterogeneity 
In addition to the well-established intertumor heterogeneity of TNBC, this subtype also 
exhibits extensive intratumor heterogeneity22. Broadly, intratumor heterogeneity (ITH) 
is defined as the “uneven distribution, spatially or temporally, of genomic 
diversification in an individual tumor, fostered by accumulated genetic mutations23. 
This key feature of cancer development has been appreciated from morphological 
standpoint as early as the 1950s24,25. With the onset of cancer genomics research, our 
understanding of ITH has rapidly evolved to encompass both phenotypic and genetic 
variation within the primary tumor. To become cancer, cells must accumulate a number 
of genetic mutations that bypass regulatory circuits essential to multiple aspects of cell 
and tissue homeostasis8. The path to obtaining these necessary mutations is varied and 
not necessarily linear.  
 
1.2.1 Models of cancer evolution 
There are two primary competing models by which tumors evolve: 1) clonal evolution 
theory and 2) cancer stem cell (CSC) theory. In the clonal evolution theory, tumor 
development has been likened to the Darwinian model of evolution of species with 
branching of cancer subclones caused by new mutations and convergence of subclones 
caused by clonal competition8,25,26. Another hypothesis for tumor development put forth 
more recently is the cancer stem cell theory, which suggests that a small subset of 
undifferentiated CSCs are responsible for generating the tumor via unchecked 
proliferation and introducing heterogeneity via aberrant cell division27,28.  
 
5 
 
A third theory for the creation of intratumor heterogeneity indicates a role for the tumor 
microenvironment in affecting cellular genotypic and phenotypic behavior28. The tumor 
microenvironment consists of tumor cells, extracellular matrix (ECM) that the tumor 
cells are situated in, vasculature, stromal cell types including cancer associated 
fibroblasts (CAFs) and infiltrating immune cells, and the gradients of soluble 
components such as gases, nutrients, and metabolites29. The “dynamic reciprocity” of 
cells and their extracellular matrix is well established30. In the context of cancer 
progression, many studies including work by our group have contributed to elucidating 
the relationship between cancer cell contractility, matrix stiffness, and cancer 
progression/aggressiveness31–33. Matrix stiffness has been shown to shift cancer cell 
phenotype in vitro34–37. Further, there is a strong correlation between genomic mutation 
and matrix stiffness across multiple cancers38. While there is compelling evidence for 
each of these theories, it may be most realistic to assume that varying degrees of all 
three play a role in the diversification of breast tumors.  
 
1.2.2 Clonal cooperativity  
As our understanding of intratumor heterogeneity has vastly expanded with the 
development of next generation sequencing, our understanding of the complexity of 
subclonal interactions has also deepened. While the clonal evolution model implies that 
competitive interactions between cancer subclones are a feature contributing to tumor 
progression, clonal cooperation has been observed as early as the 1980s39,40. This 
increased complexity in subclonal interactions seems to support the maintenance of ITH 
and may offer an overall advantage as theoretically, cooperating cells could individually 
6 
 
only have to acquire several of the ‘Hallmarks of Cancer’ in order to collectively thrive 
as a tumor41. In fact, this could be likened to a complex society where individuals 
specialize in skills/occupations that contribute to the community broadly while 
receiving other necessary services from other individuals.  
 
Since being recognized, cooperative clonal interactions have been observed to affect 
numerous aspects of cancer progression including proliferation, resistance, invasion, 
and metastasis42–46. For instance, using the MMTV-Wnt1 transgenic mouse model, 
Cleary et al. showed that basal and luminal cancer cell subtypes required each other for 
tumor growth and maintenance42. Acquired resistance was imparted to sensitive 
colorectal cancer subclones by resistant subclones in a paracrine manner43. Using a 
zebrafish-melanoma xenograft model, invasive and non-invasive melanoma cell lines 
were coinjected and found to cooperatively invade such that the non-invasive cells could 
follow behind the invasive cells44. CSC-enriched and CSC-depleted subclones of the 
PC-3 prostate cancer cell line were shown to cooperate to enhance metastasis with both 
direct and paracrine mechanisms45,47. Using a mouse model of breast cancer metastasis, 
Wagenblast et al. assessed heterogeneity using a retroviral barcoding strategy for 
lineage-tracing and further probed the phenotypes of breast cancer subclones injected 
orthotopically both individually and as a mixture46. They found that while one subclone 
was the most proliferative and thus dominated the primary tumor, several other 
subclones were more adept at performing vascular mimicry which led to increased 
presence of those subclones in the blood as CTCs as well as at metastatic sites as 
lesions46. Altogether, these studies validate the clonal cooperativity and underscore the 
7 
 
need for elegant in vitro and in vivo experiments to examine subclonal interactions and 
identify critical phenotypes for tumor progression, resistance, and metastasis. 
 
1.2.3 Experimental methods to study intratumor heterogeneity 
As suggested by Marusyk et al., one of the reasons for the slow progress in 
understanding the relationship between genetic heterogeneity and tumor evolution is the 
lack of experimental models capable of recapitulating the complexity of ITH observed 
in human patient samples48. The very traits that make most genetically engineered 
mouse models (GEMMs) attractive to researchers for studying molecular mechanisms 
in vivo, reproducibility and convenient experimental time-frames, hamper the study of 
ITH 48. Next-generation GEMMs have shown promise in recapitulating many aspects 
of human tumor development including ITH48. To examine interactions between cancer 
cell subpopulations, Zhang et al. used a Trp53-null mouse model of breast cancer to 
generate ITH49. Recently, both the KPC model of pancreatic ductal carcinoma and 
Sleeping Beauty model of myeloid leukemia were evaluated in the context of ITH and 
found to be comparable to their respective human tumors48,50,51. GEMMs such as these 
are critical to allow findings from patient sample sequencing data to be validated and 
expanded upon in a research setting.   
 
One approach to studying the heterogeneity within a cancer cell population is by sorting 
or isolating specific genotypes or phenotypes of interest for further interrogation. This 
goal can be accomplished in a multitude of ways both in vitro and in vivo. Both modern 
studies as well as many early ones including the pioneering work of the Fidler group 
8 
 
began with expanding single cancer cell subclones to interrogate ITH45,46,52–54. Although 
these studies have done much to elucidate the roles of certain cancer genotypes and 
phenotypes, they are somewhat limited in the sense that they may be only reflective of 
singular cells out of their vast heterogeneous parental tumor population. Other 
techniques that collect entire subpopulations of cells based on certain traits may be 
considered more robust due to the comparatively comprehensive sampling from the 
parental population.  
 
Fluorescence activated cell sorting (FACS) is an efficient method commonly utilized to 
allow for sorting of cell lines and dissociated tumor cell suspensions for characteristics 
such as histological subtype, CSC markers, and EMT status42,49,55. While FACS is 
commonly used for selecting cells based on expression profiles of specific proteins 
using antibodies coupled with fluorescent tags, the use of photoactivatable or 
photoconvertible fluorescent reporter systems can allow for more complex phenotypes 
to be selected. The Marcus group have used an image-guided FACS-based selection 
technique called SaGA to isolate leader and follower cells in invasion strands of in vitro 
tumor spheroids embedded in 3D matrix56–58. Numerous studies have sorted cancer cell 
lines based on cellular properties such as morphology59,60, adhesion strength61,62, 
rigidity63,64, size65, invasion ability66, and metabolism67,68. Sorting studies such as these 
highlight the importance of considering ITH when drawing conclusions from population 
measurements. In this experimental approach, the potentially confounding issue of ITH 
is removed in order to more precisely elucidate the relationship between selected 
phenotypic characteristics and clinically relevant outcomes. 
9 
 
 
Cancer cell sorting to parse apart ITH can also be performed using in vivo systems. This 
approach enables researchers to bridge the gap between scientifically relevant but 
prohibitively complex in vivo phenotypic traits and clinically implementable genotypic 
profiling. The Condeelis group has performed extensive downstream analysis of 
subpopulations of cancer cells capable of invading out of the primary tumor in response 
to a chemotactic gradient69–72. Increasingly aggressive and/or metastatic variants of 
cancer cell lines can be generated by collecting the cells which metastasize to distant 
sites in vivo45,73,74. This process can be repeated multiple times for further phenotypic 
enrichment. The Massagué group has isolated and characterized subpopulations of 
breast cancer cells exhibiting metastatic organotropism to bone75,76, lung77, and brain78. 
Selecting for cell phenotypes and interrogating these subsets compared to their 
respective parental cell populations has led to numerous important discoveries regarding 
ITH and cancer progression as well as many novel biomarkers and potential targets for 
therapeutics. 
 
1.2.4 Clinical strategies to defeat intratumor heterogeneity 
Intratumor heterogeneity is known to complicate cancer diagnosis and treatment. In 
particular, much attention has been focused on measures to overcome treatment 
resistance which can occur either when during the treatment, a cancer cell acquires 
additional mutations that endow them with resistance or when a pre-existing subclone 
that may have been present as a small percentage of the primary tumor becomes more 
prevalent as sensitive subclones are killed. This is particularly important in TNBC as 
10 
 
patients who have a complete response to neoadjuvant treatment show markedly better 
outcomes compared to those that show an incomplete response79,80. A study by Kim et 
al. used complimentary single cell and bulk sequencing to determine that chemoresistant 
subclones in TNBC that persisted after completion of neoadjuvant chemotherapy were 
indeed present in the primary tumor originally while transcriptional profiles were 
shaped by the selective pressures of treatment19. 
 
The most straight-forward method to confront resistant subclones may be to inhibit the 
mechanism enabling resistance. Any compound that improves the reduces survival 
signaling, improves pro-apoptotic signaling, or otherwise improves the efficacy of a 
chemotherapeutic could be considered a chemosensitizer81.  One drug of interest as a 
chemosensitizer for breast cancer is metformin, which has been shown to improve in 
vivo treatment outcomes of both conventional chemotherapeutics and also pro-apoptotic 
TRAIL receptor agonists82,83. Unfortunately, the use of chemosensitizers have shown 
limited success clinically thus far84. 
 
One potential solution to better manage cancer treatment and avoid the emergence of 
resistant subclones is called adaptive therapy28. This method of treatment could be 
implemented in contrast to the ‘treat-to-eradicate’ approach outlined in the Norton-
Simon model where the maximum tolerated dose density of chemotherapy is 
administered85. Although the ‘treat-to-eradicate’ approach can lead to complete 
response in some patients, in others, some would argue that this extreme treatment does 
more harm than good by accelerating the growth and evolution of resistant subclones 
11 
 
that may ultimately lead to patient death84. The accelerated growth of resistant subclones 
in response to the decimation of susceptible subclones can be likened to the ecological 
phenomenon of ‘competitive release’84. In adaptive therapy, the goal is not tumor 
eradication but rather tumor containment ideally by a personalized regimen optimized 
to maintain a manageable tumor size comprised of cells that maintain sensitivity to the 
chemotherapeutic86.  Interestingly, the principle upon which this treatment approach can 
potentially work is the ‘cost of resistance’ where resistant subclones are less fit when 
pitted directly against subclones that are sensitive86. Adaptive therapy has been 
validated in vivo using both ovarian and breast cancer cell lines in mouse models84,87 
and multiple clinical trials are currently underway for several different cancers86. 
Another exciting aspect of adaptive therapy is that in vivo studies suggest that gradually 
smaller doses of chemotherapy are sufficient to maintain tumor burden, which bodes 
well for the patient’s quality of life87. 
  
The extent to which the tumor microenvironment influences cancer progression and 
response to treatment has become increasingly recognized. One potential avenue to 
prevent the emergence of resistant subclones that has not been fully explored is 
exploiting the relationship between tumor ECM stiffness, intratumor heterogeneity and 
cancer cell resistance. ECM gene signatures have been used to organize breast cancer 
tumors into groups that were predictive of patient outcome88,89. It has recently been 
shown that increased substrate stiffness reduces cancer cell susceptibility to 
chemotherapeutics at the genetic level90. Further, collagen density has also been shown 
to impede infiltration and killing capability of CD8+ T-cells in mammary tumors91. 
12 
 
Clinically, we could envision this knowledge being implemented in two potential 
avenues: 1) establishing a diagnostic metric to assess matrix deposition/ECM stiffness 
in the untreated tumor or 2) adding an additional neoadjuvant treatment that interferes 
with the mechanosensing pathways and/or an inhibitor of genes that are upregulated in 
stiff environments that contribute to resistance, such as Abcb9 and Abcc3, which are 
genes that encode for members of the ATP-binding cassette transporter family that 
contribute to multi-drug resistance90. 
 
Almost all novel cancer treatments are eventually rendered ineffective due to the 
development of resistance. Thus, it may ultimately be more effective to target the 
changes in stromal cell populations elicited by diverse cancer cell subclones as stromal 
cell types are inherently more stable than cancer cells28. The gene signatures of stromal 
cell populations have been shown to be predictive of patient outcome92. In particular, 
immunotherapy has recently shown great promise in treating many cancers including 
TNBC where patient outcomes can differ greatly based on tumor-associated immune 
cell activity. Based on the IMpassion 130 clinical trial, the Food and Drug 
Administration accelerated approval for atezolizumab, a monoclonal antibody targeting 
PD-L1, plus nab-paclitaxel, a conventional chemotherapeutic, for the treatment of 
locally advanced, unresectable, or metastatic TNBC93–95. Anti-PD-L1 immunotherapy 
shows great promise in treating TNBC, effectively bypassing the issue of resistance by 
working to activate immune cells. 
 
Another important aspect of cancer diagnosis and treatment where ITH needs to be 
13 
 
addressed is metastasis. There is conflicting evidence regarding whether subclones 
ultimately responsible for distal metastasis are present in the primary tumor or if 
metastatic cells undergo alterations after dissemination that render them distinct from 
their ancestors. Much of this controversy could be explained by intertumor 
heterogeneity as well as differences in methodology. The current prevailing theory of 
cancer metastasis is that a small fraction of subclones present in the primary tumor can 
possess the potential to complete all the steps that constitute the metastatic cascade96. 
Due to the extensive ITH in the primary tumor, these metastatic subclones may not be 
present in readily-detectable levels upon biopsy 28. Single cell sequencing studies have 
shown great potential in characterizing the genetic profiles of subclones in the primary 
tumor. However, the ability to compare these samples with paired metastatic lesions is 
hindered by a lack of paired samples collected at metastatic sites as it can be 
procedurally difficult and not offer significant benefit from a clinical standpoint.  
 
An attractive alternative to sample collection at the metastatic site is “liquid biopsies,” 
blood samples which can contain circulating tumor cells (CTCs), cell-free circulating 
tumor DNA, and tumor cell-derived exosomes97. These biopsies are advantageous as 
they are minimally invasive and can be collected along with other routine bloodwork. 
Further, liquid biopsies can be collected at multiple points over time to monitor tumor 
progression and response to treatment in real time. Although there is a less direct 
relationship between characteristics of CTCs and metastatic sites since only a small 
percentage of CTCs will survive and have the potential to complete the remaining steps 
of the metastatic cascade, both CTC number as well as biomarker expression have been 
14 
 
shown to be clinically relevant in multiple cancers. In breast cancer, CTC detection was 
shown to correlate with reduced overall survival and disease free survival in both early 
and metastatic patients98,99. Despite the strides made in characterizing CTC 
heterogeneity and metastatic potential, much work remains to be done; however, the 
number of clinical trials currently in progress which utilize CTCs to inform patient 
treatment emphasize the potential of this technique for clinical implementation to 
improve patient outcomes100. 
 
1.3 Matrix-mediated link between breast cancer and other diseases 
Aging is the leading risk factor for numerous chronic diseases including cancer101. 
While the reasons for the relationship between aging and cancer remain somewhat 
controversial, it stands that the median age of cancer patients in the United States is 70 
years old, and the death rates attributed to cancer increase with age102. While factors 
including telomere shortening, mutation accumulation, reactive oxygen species, and 
decreasing immune function all have been proposed to contribute to the increased 
incidence of disease with increasing age, another important factor implicated in this 
relationship is tissue stiffening101–105. Tissue stiffening occurs naturally with aging in 
almost every tissue or organ in the human body and has been shown to play key roles 
in numerous diseases including atherosclerosis, diabetes, glaucoma, and 
osteoarthritis103,106,107. While many diseases share this common biophysical trait, the 
potential for increased incidence of comorbidities due to synergistic tissue stiffening 
remains largely unstudied. 
 
15 
 
1.3.1 TNBC and diabetes 
A recent multicenter observational study of Austrian cancer patients found that 86% of 
patients possessed at least one comorbidity while a vast majority of the elderly cancer 
patient population reported more than three comorbidities108. While the overall clinical 
impact of comorbidities in cancer patients remains unresolved with numerous factors 
potentially contributing to discrepancies between studies, there is a documented 
negative effect on cancer survival109. This correlation may be due to changes in medical 
treatment since the presence of comorbidities is also associated with less aggressive 
cancer treatment, which is most likely connected to the potential for drug-drug 
interactions and increased chemotherapeutic toxicity109,110. However, potential 
synergistic effects between comorbid diseases and cancer progression is also a 
possibility that must be investigated. The most common comorbidities found in breast 
cancer patients are hypertension, arthritis, and diabetes111. Each of these comorbidities 
are associated with a significant negative effect on quality of life in the domain of 
physical functioning and bodily pain111. In TNBC, high comorbidity was associated 
with increased cancer-related mortality across all stages of disease112. Importantly, 
African-American/Black TNBC patients, who are disproportionately affected by TNBC 
compared to other breast cancer subtypes, had a significantly higher risk of mortality 
when comorbidities were present113. This effect was more drastic for African-
American/Black patients than for white patients with TNBC113.  
 
Type II diabetes mellitus (T2DM) is a complex metabolic disorder that is characterized 
by hyperglycemia caused by insulin resistance and insufficient insulin secretion114. If 
16 
 
unmanaged, chronic hyperglycemia from T2DM can lead to damage, disfunction, and 
eventual failure of organs including kidneys, eyes, nerves, heart, and vasculature114. 
T2DM is projected to increase in incidence, affecting upwards of 500 million people 
worldwide by 2035115. The sharpest increases in T2DM are expected to occur in low- 
and middle-income countries where 80% of the current T2DM population resides115. 
Numerous factors including poor diet, sedentary lifestyle, obesity, increased BMI, 
smoking, and genetics have all been implicated as risk factors for T2DM115.  
 
More than a century ago, early biostatisticians noted an association between cancer and 
diabetes107,116,117. Modern studies have worked to further clarify this association and 
have found that the relationship is cancer-type dependent with many cancers including 
colorectal, liver, pancreas, bladder, breast, and endometrial cancers showing a positive 
correlation while interestingly, prostate cancer shows an anti-correlation118. Women 
with T2DM have an increased risk of breast cancer119,120. Women who have T2DM 
generally present with more aggressive breast cancer as indicated by more advanced 
staging121,122. Further, T2DM is associated with poor overall survival and disease free 
survival in breast cancer patients123. While risk is increased and outcomes are worsened, 
T2DM does not predispose a patient to present with any specific clinical subtype of 
breast cancer compared to patients without diabetes124. The potential reasons for these 
correlations are numerous as both cancer and diabetes are complicated and multifaceted 
diseases. Hyperglycemia, hyperinsulinemia, dyslipidemia, pro-inflammatory signaling, 
and even the gut microbiome have been implicated in the interplay between these 
diseases with the effects of obesity, which commonly presents with T2DM, further 
17 
 
complicating matters125. Given the prevalence of both diabetes and breast cancer 
globally as well as the strong evidence of synergy between these comorbidities, it is 
important to determine the key interactions and develop therapeutic regimens to better 
improve outcomes for this population. 
 
1.3.2 Glycation: pathological matrix modifications in diabetes mellitus 
One of the results of chronic hyperglycemia and etiological sources of diabetes-
associated organ and tissue damage is the irreversible crosslinking of proteins and lipids 
in a process called glycation107. Also known as the Maillard or browning reaction, this 
non-enzymatic reaction was first recognized to occur with collagen, the most abundant 
protein in the human body, when discolored dura mater (the tough, collagenous sac 
covering the brain) was observed by Cerami and Koenig in tissues collected from 
diabetic and elderly patients126. Glycation is a multistep reaction which commences 
when the ring structure of glucose briefly opens allowing its nascent aldehyde group to 
react with the amino group on a protein forming a highly unstable intermediate called a 
Schiff base126. The Schiff base quickly reconfigures into a relatively more stable but 
still reversible Amadori product also known as an early glycation product107,126. The 
Amadori product can undergo degradation via dehydration, oxidation, or other reactions 
to form a dicarbonyl compound127. This intermediate then degrades to form an 
irreversible structure called an advanced glycation end product (AGE)126,127. AGEs can 
bind to and accumulate on long-lasting proteins, impairing their function and causing 
them undergo accelerated degradation, resulting in free AGEs127. AGEs also crosslink 
proteins, reducing flexibility and increasing stiffness126,127. In addition to structural and 
18 
 
functional modifications of proteins, AGEs also act as a signaling molecule, binding to 
cellular receptors for advanced glycation end products (RAGEs) which trigger pro-
inflammatory signaling, reactive oxygen species generation, and cellular stress and 
death128. AGEs have been shown to contribute to normal aging as well as numerous 
pathologies affecting virtually every tissue type in the human body126,129.   
 
Although glucose is the least reactive sugar present in the human body, the abundance 
of glucose relative to other sugars makes its effects on proteins far more important 
physiologically126. Further, while enzymatic means of ECM crosslinking occur with far 
greater speed than non-enzymatic glycation, the irreversibility of AGE formation 
coupled with the longevity of ECM proteins mean that the effects of glycation are long 
lasting and cumulative over many years129. For T2DM patients with chronic 
hyperglycemia, it becomes appreciable that AGE accumulation and associated 
stiffening has been shown to play critical roles in diabetic-associated cardiomyopathy, 
nephropathy, neuropathy, retinopathy, and osteopenia130–133. It also makes intuitive 
sense that glycation could play a role in the interplay between T2DM and breast cancer 
progression as recent data from Lega et al. found that only patients with an extended 
duration of T2DM had a significantly increased all-cause and cancer-specific 
mortality134. The extended time frame required to observe a correlation between T2DM 
and breast cancer progression suggests that long term effects of T2DM may be 
responsible. Since glycation reactions and AGE formation occurs on the scale of weeks 
to months135, it is important to determine whether glycation-mediated stiffening of 
mammary tissue could play a role in this clinical observation. 
19 
 
 
1.4 In vitro models of cancer metastasis 
Metastasis is one of the leading causes of death globally136. During tumor development, 
cancer cells acquire genetic mutations, co-opt their microenvironment, and induce 
angiogenic sprouting that can potentially lead to metastasis. Metastatic progression of 
solid tumors can be divided into five major steps: 1) invasion of the basement membrane 
and cell migration; 2) intravasation into the surrounding vasculature or lymphatic 
system; 3) survival in the circulation; 4) extravasation from vasculature to secondary 
tissue; and finally, 5) colonization at secondary tumor sites. Each stage of metastasis 
imposes different, often harsh conditions and energetically taxing challenges for the 
cancer cells to complete. As the cascade progresses, the number of viable cancer cells 
which survive and successfully complete each stage decreases precipitously; however, 
the underlying reason for this is not clear.   
 
Given the dynamic, multi-step nature of metastasis, and the well-documented presence 
of intratumor heterogeneity, certain cancer cell subpopulations may potentially perform 
some steps of metastasis more efficiently than others. Moreover, cooperative synergies 
may exist between cancer cell subpopulations such that it may not be necessary for a 
single subpopulation to complete the entire cascade alone41. Thus, success in one aspect 
of metastatic fitness is not necessarily predictive of success overall. This consideration 
highlights the importance of developing engineered in vitro models to assess cellular 
performance at various steps of the metastatic cascade in a controlled and reproducible 
manner. It is critical to leverage these models to bridge the existing gap between the 
20 
 
staggering level of genotypic characterization for metastatic subclones generated by 
next-generation sequencing studies and the paucity of corresponding phenotypic 
characterization of these critical subtypes. 
 
1.4.1 Tumor cell invasion and intravasation 
It is well established that metastasis is the leading cause of cancer-related death. Cancer 
cell invasion beyond the confines of the primary tumor site through the basement 
membrane and its surrounding layer of myoepithelial cells is considered the 
differentiating factor between pre-cancerous neoplasia and malignant carcinoma137. The 
basement membrane, which is composed of extracellular matrix components including 
laminin, type IV collagen, nidogen, and proteoglycans acts as a barrier to cancer cell 
migration138. Collagen I, one of the predominant stromal ECM components often 
undergoes significant changes including increased deposition, crosslinking, and 
alignment during tumor development that can aid in cancer cell escape139. ECM 
biophysical properties have been shown to be clinically relevant in tumor progression. 
Illustrating this fact, mammographic density is the greatest independent risk factor for 
the development of breast cancer, and specific collagen fiber architectures located at the 
tumor boundary have been shown to possess prognostic value, being associated with 
decreased patient survival139–142.  
 
Tumor spheroids 
The in vitro tumor spheroid model has been used extensively to observe cancer cell 
invasion and migration into 3D matrix. This model allows both cell-cell and cell-ECM 
21 
 
interactions during invasion to be observed with a high level of control and 
reproducibility in numerous cancer types143. Several methods exist to form cells into 
spheroids including the hanging drop method, round bottom non-adherent well plates, 
and more sophisticated custom techniques such as fabricated agarose molds144. As 
cancer cells have been shown to migrate predominantly in a collective fashion in vivo, 
characterization of cell-cell interactions are critical to capture in engineered in vitro 
models145. Leader-follower behavior is a pattern of migration where cells that are more 
adept at migrating through matrix can enable less migratory cells that may not be able 
to do so alone to migrate behind them. Our lab and others have used tumor spheroids to 
characterize leader cells, which have distinct genetic signatures and have been shown 
to require actomyosin contractility and proteolytic activity56,66,146. Recently, our group 
assessed the energetic regulation of leader-follower cell dynamics within cancer cell 
invasion strands using tumor spheroids embedded in type I collagen matrix147. As leader 
cells can enable invasion of less migratory follower cell types, they not only initiate 
metastatic invasion but can also introduce cellular heterogeneity into the metastatic 
cascade, increasing chances of successful metastasis and making these cells important 
targets to improve patient outcomes148,149. Leader-follower interactions highlight the 
importance of assays such as tumor spheroids where these and other important 
multicellular interactions can be studied.  
 
Further, stromal cell types can be integrated into either the spheroids themselves or the 
surrounding ECM which allows for heterotypic interactions between the invading 
cancer cells and stromal cells to be observed150. Endothelial cells are also amenable to 
22 
 
spheroid formation to assess angiogenic sprouting151, and when co-seeded with cancer 
cells, vascularized networks within the tumor spheroid can be established such that 
intravasation events can be observed152. Labernadie et al. co-cultured vulval cancer cells 
and cancer associated fibroblasts (CAFs) in tumor spheroids and found that CAFs 
exerted force on cancer cells to enable collective migration via heterotypic E-
cadherin/N-cadherin adhesion150. As primary cancer cells more accurately reflect the 
heterogeneity of patient tumors compared to cell lines, more studies are incorporating 
these precious cells into tumor spheroid assays. Tumor spheroids have been made with 
primary cells from numerous cancer types including glioma, hepatocellular carcinoma, 
and ovarian cancer for invasion and chemosensitivity studies153–155. While there are still 
significant impediments to implementing primary cell cultures as a part of the standard 
of care for cancer diagnosis and treatment, the in vitro tumor spheroid assay allows great 
control, reproducibility, and flexibility with the ability to incorporate stromal cell types 
in single or co-culture systems to design more physiologically relevant experiments to 
model tumor cell invasion. 
 
Organoids 
Organoids are another option for cancer cell invasion studies that incorporates more 
features of the tumor microenvironment including ECM and stromal cell types. 
Organoids are formed by enzymatically digesting either patient or animal model tumors 
into fragments that can be embedded in 3D matrix. The Ewald group created organoids 
using the MMTV-PyMT mouse model of breast cancer with and without genetic 
modifications to mechanistically assess the roles of cytokeratin-14, E-cadherin, and 
23 
 
myoepithelial cells in cancer invasion156–158. Organoids can also be created from 
samples of human patient tumors, which have shown great promise as invasion assays 
and drug screening platforms159,160. As organoids maintain some degree of the cellular 
and ECM organization that was present in the tumor sample, this method is thought to 
be more physiologically relevant but also inherently allows less control over aspects 
such as organoid size, shape, and composition. Efforts to engineer greater levels of 
experimental control and higher throughput along with further characterization of the 
cellular heterogeneity captured in organoid models will continue to bring this system 
closer to clinical implementation as tools for personalized medicine. 
 
1.4.2 Survival in the circulation and attachment to the endothelium 
While few cancer cells reach the circulation, even fewer survive the hemodynamic shear 
forces, immune stresses, and red blood cell collisions they encounter once there161. 
Circulating tumor cells (CTC) arrest in a vessel and extravasate through two primary 
mechanisms: physical occlusion and adhesion after rolling.  During physical occlusion, 
a CTC’s diameter surpasses that of the microvasculature, and the cell becomes lodged 
before attaching and extravasating.  During rolling-adhesion, CTCs collide with the 
endothelium, roll via E-selectin or P-selectin binding, and arrest via intercellular 
adhesion molecule-1 (ICAM-1) or vascular cell adhesion molecule-1 (VCAM-1) 
binding. In vitro models for this process require spatiotemporal control of shear forces, 
tunable substrate functionalization, and real-time imaging capability.   
 
 
24 
 
Circulating Tumor Cell Capture 
Microfluidic and microtubing systems enabling the collection of both single and 
clustered CTCs from patient blood have contributed greatly to understanding cancer 
metastasis162–165.   These platforms often employ surfaces functionalized with CTC-
specific adhesion proteins or antibodies to optimize adhesion dynamics for CTCs while 
minimizing that of leukocytes also present in whole blood162,163,165.  Physical entrapment 
under flow can also be utilized to isolate CTC clusters that have been suggested to have 
increased metastatic potential compared to single CTCs164.  Recent work has shown that 
single cell encapsulation of CTCs into microdroplets can be utilized to profile enzyme 
secretion166.  Additionally, single cell RNA sequencing of human patient CTCs has been 
optimized to assess inter- and intra-patient heterogeneity and identify potential 
therapeutic targets using a microfluidic platform and barcoding technique167.  As the 
methods to capture viable CTCs become more tractable, further probing of later stages 
of metastasis using isolated CTCs could provide insight into the properties of these rare 
but crucial cells.  Engineered platforms have the potential to elucidate the changes CTCs 
may undergo as they transition from solid tissue to the circulation, as well as determine 
the properties of CTCs best suited for extravasation and colonization.  
 
Shear Stress Models 
Cone and plate viscometers are often used to expose cancer cells to physiological shear 
forces in cell culture medium or whole blood168,169 and study the interactions of CTCs 
with neutrophils, platelets, and endothelial monolayers170,171,172. Efforts to increase 
throughput have led to the development of a cone viscometer platform that interfaces 
25 
 
with standard 96-well plates to enable more streamlined testing173. While cone and plate 
viscometers facilitate highly controlled, reproducible exposure to shear conditions, they 
lack relevant vessel-like architecture and do not allow for real-time imaging during 
shear exposure.  
 
Numerous commercially available, relatively inexpensive platforms are used to produce 
shear stresses in vitro.  Motorized expulsion through a needle has been used to assess 
cancer cell viability and conditioning after shear stress exposure174,175.  Parallel plate 
flow chambers can be used to assess rolling-adhesion interactions between circulating 
cells perfused over a substrate coated with ECM, ligands, or endothelial 
monolayers168,176,177. Others assess the rolling and adhesion of cancer cells to 
physiologically relevant proteins using controlled perfusion through functionalized 
microtubing178–180.  
 
Microfluidic systems provide an immense degree of customization, with the ability to 
incorporate complex structures and dynamic flow patterns in perfused channels that can 
be coated with ECM or endothelial monolayers181. More complete microfluidic models 
can incorporate spatially defined chemokine gradients and or organ-specific cells, such 
as primary lung endothelial cells or osteo-differentiated bone marrow derived stem 
cells182–184. While numerous microfluidic platforms mimic the vasculature, CTCs can 
also travel through the lymphatic system. Notably, it was observed that low shear 
stresses mimicking lymphatic flow induced cancer cell motility while high shear 
stresses mimicking arterial and venous flows inhibited cell motility in a YAP1-
26 
 
dependent manner, highlighting the importance of selecting physiologically relevant 
shear stresses since different ranges can elicit divergent cell behaviors185. 
 
Tumor cell arrest during extravasation can also occur through cancer cell occlusion in 
capillary networks. Serial deformation and transmigration chambers in microfluidic 
devices have been designed to mimic constrictions in capillaries and relevant 
endothelial/ECM barriers that cells must bypass to transmigrate after arresting186.  For 
example, a microfluidic device with capillary-sized channels was used to show that CTC 
clusters isolated from patient blood can traverse these constrictions while remaining 
intact187.  
 
Recently, self-assembled perfusable microvascular networks have been developed to 
investigate physical occlusion and rolling-adhesion events leading to extravasation188.  
To stabilize self-assembled networks, co-culturing fibroblasts segregated from 
endothelial cells provides the necessary paracrine signaling for network stabilization, 
while co-seeding with pericytes regulates vessel diameter and decreases vessel 
permeability188,189. While self-assembled microvascular networks do not necessarily 
require specialized equipment, control over network formation is limited.  
 
Three-dimensional printing of carbohydrate glass sacrificial fibers can create highly 
controlled, multiscale, perfusable vascular networks190. Though geared towards 
improving tissue engineering designs, this platform could be adapted to study 
extravasation in capillary networks. Live-cell lithography was developed to better 
27 
 
control cell placement for in vitro vascular network assembly191. In this system, multiple 
optical tweezers are used to manipulate placement of cells in three dimensions allowing 
the controlled addition of pericytes, smooth muscle cells, and fibroblasts outside of the 
vessel. Advances like these lay the foundation for systems that better recapitulate the 
complexity of the tumor microenvironment. As patient CTC capture platforms improve, 
further incorporation of these precious clinical samples into downstream assays will be 
crucial towards investigation of CTC performance during subsequent stages of 
metastasis. If assays can be streamlined and correlated with clinical data, theranostic 
platforms with CTCs isolated from patient blood have the potential to improve clinical 
outcomes.     
 
1.4.3 Extravasation and colonization 
Following arrest within the vessel, cancer cells must extravasate from the vessel to 
colonize new sites. This process differs from intravasation, where cancer cells navigate 
tumor-modified stroma via chemotactic and durotactic gradients toward leaky, nascent 
vasculature without experiencing hemodynamic stressors; rather, during extravasation, 
the vasculature that is breached by cancer cells is healthier, and cancer cells actively 
experience fluid shear stresses due to blood flow192. After extravasation, cancer cells 
have one final task to complete: colonization of secondary sites. This process is thought 
to be extremely inefficient with only a minute percentage of CTCs growing into 
lesions96. Metastatic niches possess cell types and ECM compatible for tumor cell 
survival and growth96, including perivascular niches, spaces around blood capillaries 
where cancer cells can seed. Extravasation and colonization models require tissue-
28 
 
specific cell types, microenvironmental cues, and vascularization. Leveraging tissue 
engineering advancements to model metastatic sites may be key in understanding 
factors driving colonization as it is possible to tailor the site to isolate roles of cells 
types, growth factors, and ECM architectures.  
 
Extravasation and secondary tissue microenvironment models 
As bone metastasis occurs frequently in breast and prostate cancers and correlates with 
shortened patient prognoses, many models of metastatic colonization in bone have been 
created193. Osteo-differentiated mesenchymal stem cells, mineralized hydroxyapatite-
incorporated ECM, and ex vivo bone scaffolds have all been shown to elicit relevant cell 
behavior in in vitro bone tissue models194–197. Incorporation of perfusable vascular 
networks in these models allows for cancer cells to be flowed though, recapitulating 
extravasation events at the metastatic site. Bioreactors can be used to create complex, 
mature tissue constructs for seeding as well as to expose seeded scaffolds to tunable, 
physiological compressive forces to observe colonization behavior198,199.  
 
Several different models have been exploited to assess colonization in various organ 
systems. Decellularization of tissues including mammary fat pad, lymph node, and 
lungs has been used to three-dimensionally map the spatial distribution of ECM 
components of these tissues in health and disease200. Moreover, decellularization 
provides a scaffold that can then be re-seeded with cancer cells to examine colonization 
in a simplified yet physiological setting201. Decellularization proves a powerful 
technique to enhance and inform tissue-engineered constructs of metastatic colonization 
29 
 
sites and assess cancer cell-ECM interactions. LiverChipÒ is a commercialized 
microfluidic model of the hepatic niche used to observe interactions between cancer 
cells, hepatocytes, and non-parenchymal cells202. Infiltration of the brain-blood barrier 
has been modeled by adding cancer cells to numerous permutations of co-cultures 
containing endothelial cells, pericytes, glial cells, astrocytes, and cancer-associated 
fibroblasts203–207.  As lung, liver, brain, and lymph node are all extremely common 
metastatic sites, further work should be directed towards developing more complex, 
physiologically relevant in vitro models for assessment of cancer cell metastatic 
colonization at these distinct locations.   
 
Currently, metastatic colonization assays are in their infancy relative to assays focusing 
on earlier stages. As colonization is the stage where metastasis gains its lethality and 
where confounding events like drug-resistance and dormancy often occur, it is 
promising as a key point of intervention. While much emphasis is placed on the 
personalized side of patient specific cancer cells, understanding patient-specific, non-
tumor cells in metastatic sites may help explain drug-resistance and dormancy. 
 
1.4.4 Full metastatic cascade models 
While in vivo models can be used to study the entirety of the cascade, the complexity 
and timescale of metastasis limits their utility. In vitro models successfully recapitulate 
individual steps in metastasis, yet few encompass more than one stage in the process.  
 
Recent work to develop a more complete metastatic platform has resulted in a 
30 
 
microfluidic metastasis-on-a-chip model where hydrogels embedded with host tissue 
cells are combined with microfluidics to represent the spread of metastatic cells from 
primary to secondary tissue208. Specifically, microfluidic chambers containing a gut 
tissue-like ‘source’ seeded with colon cancer cells and a liver tissue-like ‘sink’ are 
connected by a perfused flow channel. In this model, cancer cells can exit the gut 
chamber and spread to the liver chamber. This three-dimensional construct facilitates 
drug-screening and visualization of metastasis, though it still lacks important features 
like endothelial barrier function, intravasation, and extravasation. Despite limitations, it 
marks one of the first steps toward an in vitro model distilling the key components 
representing the diverse microenvironments cancer cells encounter during metastasis.  
 
While there are still limitations hindering in vitro recapitulation of the full metastatic 
cascade, approaches that stitch together multiple sequential steps into a single assay 
have fewer impediments. Reductionist models that incorporate a primary tumor site and 
a metastatic niche site separated by ECM serve as a simplified approach to assessing 
metastatic potential209. However, metastasis is a dynamic, multi-step process and by 
simplifying models to exclude parts of the cascade, we are only gaining insight as to 
how well cancer cells perform specific steps out of context. Thus, it is critical that more 
complete models be developed so that metastasis can be observed in the correct series 
of events.   
 
1.5. Dissertation organization 
The objective of Part I of this work is to assess the role of diabetic hyperglycemia and 
31 
 
non-enzymatic glycation in triple negative breast cancer progression. This work is 
highly relevant since there is a well-documented clinical correlation between diabetes 
mellitus and breast cancer.  
 
In Chapter 2, we characterize a novel mouse model of TNBC breast tumor development 
with pre-existing diabetic hyperglycemia. We show that tumors in mice with diabetic 
hyperglycemia are larger, stiffer, and more aggressive tumors than tumors of mice with 
normal blood glucose levels. Further, we use several pharmacological inhibitors of non-
enzymatic glycation to show that tumor development is enhanced by this reaction 
independent of blood glucose levels. This work establishes a novel role for the structural 
alterations in the microenvironment and advanced-glycation-end products (AGEs) in 
the synergy between cancer and diabetes mellitus. 
 
The objective of Part II of this work is to investigate the relationship between the 
migratory phenotype and the metastatic phenotype in triple negative breast cancer. 
Because adoption of a migratory phenotype is often considered essential for the initial 
steps of cancer cell metastasis, specifically, dissemination from the primary tumor and 
invasion into the stroma, migratory ability is often considered an indicator of metastatic 
potential. 
 
In Chapter 3, we assess the relationship between migratory ability and metastatic 
potential in the context of phenotypically sorted strongly and weakly migratory breast 
cancer cell subpopulations. We demonstrate that our novel transwell sorting assay 
32 
 
results in distinct, stable subpopulations that are representative of the most and least 
migratory cells within the parent population. Using an orthotopic mouse model, we find 
that migration ability is not correlated with metastasis. We then mechanistically show 
that the weakly migratory cells express E-cadherin, a cell-cell adhesion protein which 
is both necessary for weakly motile subpopulation metastatic success and sufficient for 
enabling metastasis of the highly migratory subpopulation.  
 
In Chapter 4, phenotypically sorted highly migratory and weakly migratory TNBC cell 
subpopulations are compared alone and combined in vitro and in vivo. Using an in vitro 
spheroid model, we observe that when seeded alone, our subpopulations migrate using 
distinct migration modes: highly migratory cells migrate largely as single cells while 
weakly migratory cells migrate collectively. In the 1:1 co-culture condition, the highly 
migratory subpopulation enables the weakly migratory subpopulation to displace further 
than they could when migrating alone. We show that the subpopulations exhibit leader-
follower behavior when in co-culture. Our in vivo findings suggest that this behavior 
translates into increased metastatic fitness where highly migratory, weakly metastatic 
cells assist weakly migratory, highly metastatic cells in disseminating from the primary 
tumor. 
 
In Chapter 5, conclusions and future directions for the work presented herein are 
discussed.
33 
  
 
CHAPTER 2 
 
DIABETIC HYPERGLYCEMIA PROMOTES PRIMARY TUMOR 
PROGRESSION THROUGH NON-ENZYMATIC GLYCATION  
 
Data in this chapter was collected/analyzed with Wenjun Wang. 
 
2.1 Abstract 
Diabetes mellitus is a complex metabolic disorder that is associated with increased 
risk of breast cancer. Despite this correlation, the interplay between tumor 
progression and diabetes is still mechanistically unclear. Here, we established a 
murine model where diabetic hyperglycemia was induced prior to mammary 
tumorigenesis. We find that pre-existing hyperglycemia causes increases in tumor s 
that are larger, stiffer, and more aggressive as evidenced by histological grading and 
epithelial-to-mesenchymal transition markers. We hypothesized that this phenotype 
was facilitated by increased levels of glycation, a non-enzymatic reaction which has 
been implicated in numerous diabetes-associated conditions. When mice with 
diabetic hyperglycemia are treated glycation inhibitors, we observed a reduction in 
size, stiffness, and EMT levels of diabetic tumors to levels comparable with non-
diabetic tumors. Altogether, our study describes a novel mechanism by which 
diabetic hyperglycemia promotes breast tumor progression. Additionally, our work 
also provides evidence that glycation inhibition holds promise as a potential 
adjuvant therapy for diabetic cancer patients. 
 
2.2 Introduction 
Type 2 diabetes mellitus (T2DM) is a chronic metabolic disorder characterized by 
 34 
 
hyperglycemia caused by defects in insulin secretion, insulin resistance, or both115. 
T2DM is considered the seventh leading cause of death in the United States with a 
majority of diabetes-associated deaths being attributed to cardiovascular disease 
triggered by chronic hyperglycemia115,210. Beyond cardiomyopathy, diabetes affects 
the patient systemically with retinopathy, nephropathy, osteopenia, and neuropathy 
presenting as common diabetes-associated complications127,131,211. Interestingly, 
there is an increased incidence of breast cancer in diabetic patients119,120. Further, 
diabetic breast cancer patients also present with more advanced staging and 
worsened prognosis compared to those without diabetes121–123. The relationship 
between these two highly prevalent diseases has been attributed to numerous factors 
including hyperinsulinemia, insulin-like growth factors, hyperglycemia, 
dyslipidemia, adipokines, inflammatory cytokines, and even factors from the gut 
microbiome125. However, given the critical role of matrix stiffness in cancer 
progression and diabetic complications, the potential role of diabetes-induced 
alterations to the tumor microenvironment remains understudied. 
 
One of the most prominent matrix modifications to be studied in the context of 
diabetes mellitus is non-enzymatic glycation126,127,130,132,133. Also known as the 
Maillard or browning reaction, glycation was first recognized to occur in the human 
body due to the discoloration of collagen-rich tissues observed in diabetic 
individuals126. Glycation occurs when the aldehyde group of a reducing sugar, most 
commonly glucose, reacts with the amine group of a protein126. The final result of 
glycation is a highly stable structure called an advanced glycation end product 
(AGE)212. Glycation is a complex, multistep reaction that occurs over the time span 
of weeks and accumulates on long lasting proteins including collagen I 
 35 
 
fibers126,127,130. Glycation reactions can result in ECM crosslinking, resulting in 
stiffening of tissues126. AGEs also act as signaling ligands with the receptor for 
advanced glycation end products (RAGE), which triggers apoptosis, reactive oxygen 
species production, and pro-inflammatory signaling213. Importantly, the AGE-
RAGE signaling axis has been shown to promote tumor malignancy107. As non-
enzymatic glycation results in impaired protein function, altered ECM structure, and 
activation of multiple key signaling pathways, it is critical to isolate and characterize 
the role of glycation-mediated ECM stiffening in the interplay between diabetes and 
cancer. 
 
Here, we utilize the MMTV-PyMT mouse spontaneous mammary tumor model with 
streptozocin (STZ) injections to establish diabetic hyperglycemia prior to tumor 
formation. Using this model, we assess the role of diabetic hyperglycemia on tumor 
progression and find that diabetic tumors are larger, stiffer, and have are further 
along the EMT spectrum compared to control tumors. We use glycation inhibitors 
aminoguanidine and alagebrium to assess the role of non-enzymatic glycation in the 
observed diabetic tumor phenotype and find that independent of blood glucose 
levels, glycation inhibition reduces tumor size, stiffness, and EMT state to non-
diabetic control levels. This study reveals a novel mechanism by which diabetes 
promotes breast cancer progression via non-enzymatic glycation. 
 
2.3 Materials and Methods 
Antibodies and reagents 
Antibodies used for staining in Western blot and immunohistochemistry 
experiments were as follows: anti-advanced glycation end product antibody (AGEs; 
 36 
 
ab23722, Abcam, Cambridge, United Kingdom), anti-fibronectin antibody (sc-9068 
and sc-6953, Santa Cruz Biotechnology, Dallas, TX, USA), anti-transforming 
growth factor-𝛽 antibody (TGF-𝛽; ab92486 Abcam), and anti–glyceraldehyde-3-
phosphate dehydrogenase antibody (GAPDH, Poly6314; BioLegend, San Diego, 
CA, USA). Primary and secondary antibodies used in immunofluorescence staining 
included anti-vimentin antibody (ab92547, Abcam), anti-polyoma virus medium T 
antigen antibody (PyMT; ab15085, Abcam), anti-E-cadherin antibody (7870, Santa-
Cruz), anti-Ki67 antibody (14-5698-82, Thermo Fisher Scientific, Waltham, MA, 
USA), anti-rat Alexa Fluor 647 goat antibody (A21247, Thermo Fisher Scientific), 
and anti-rabbit Alexa Fluor 568 donkey antibody (A10042, Thermo Fisher 
Scientific). All other reagents used here were purchased from Thermo Fisher 
Scientific. 
 
Mousework 
FVB/N-Tg(MMTV-PyVT)634Mul/J (MMTV- PyMT) mice were utilized in this 
study. All mice were maintained following a protocol approved by the Vanderbilt 
University Institutional Animal Care and Use Committee. Female MMTV-PyMT 
mice of the FVB strain background (Jackson Laboratory, Bar Harbor, ME, USA) 
were fed with D12492 high fat diet (NC0004611, research diet Inc, New Brunswick, 
NJ, USA) starting at 4 weeks of age. To induce diabetes, mice were given 5 
consecutive daily doses of streptozotocin (STZ, Millipore Sigma) through 
intraperitoneal injection at 70 mg/kg body weight. STZ is a small molecule that has 
been shown to selectively destroy pancreatic beta islet cells and induce diabetic 
hyperglycemia within mice214. The control group of mice were injected with citrate 
buffer (0.1 mmol/L, pH 4.5; Millipore Sigma) as vehicle control. Body mass, tumor 
 37 
 
volume, and glucose levels of mice were measured weekly starting prior to 
beginning STZ injections and continuing until the mice were sacrificed. The tumor 
volume of the MMTV-PyMT transgenic mice was calculated as follows: 𝜋 ×  length2 
× width / 12 based on caliper measurements215. Mice that had a blood glucose level 
above 400 mg dL-1 were considered diabetic. To further determine that diabetes was 
induced in the mice, presence of beta islet cells within the pancreas were quantified 
with immunohistochemistry at study endpoint. For STZ-injected mice receiving 
aminoguanidine, mice were treated with 3 mg/kg body weight aminoguanidine 
(Millipore Sigma) in drinking water until study endpoint. For STZ-injected mice 
receiving alagebrium (ALT-711) treatment, mice were treated with 1 mg/kg body 
weight alagebrium (MedChemExpress, Monmouth Junction, NJ, USA) injected 
daily intraperitoneally until study endpoint. Seven weeks after STZ injection, mice 
were humanely sacrificed by CO2 asphyxiation and necropsied. Mammary tumors 
were collected and snap frozen in 70% ethanol dry ice slurry or fixed with 4% (v/v) 
paraformaldehyde in PBS. Snap frozen samples were then sectioned at 5 𝜇𝑚 
thickness for immunofluorescence and immunohistochemical staining and at 20 µm 
for atomic force microscopy (AFM). Fixed tissues were sectioned at 5 µm thickness 
for collagen quantification via picrosirius staining. 
 
Glucose and insulin tolerance assays 
Glucose tolerance assay and insulin tolerance assay were also performed within one 
week post STZ injection. The mice were fasted 8 hours before both experiments. 
For glucose tolerance assay, mice were injected with sterile 10% glucose (Millipore 
Sigma) intraperitoneally at 1 g/kg body weight. For insulin tolerance assay, insulin 
(Eli lily, Indianapolis, IN, USA) was administered through intraperitoneal injection 
 38 
 
at 2 unit/kg body weight. Blood was collected immediately prior to the 
administration and at 15, 30, 60, 90, and 120 minutes after administration. 
 
Immunofluorescence 
Tumor tissue sections extracted from MMTV-PyMT mice were fixed with 4% (v/v) 
paraformaldehyde in PBS for 10 min at room temperature. After fixation, samples 
were washed with PBS, permeabilized with 1% (v/v) Triton X-100 (Thermo Fisher 
Scientific) in PBS. After permeabilization, tissue sections were then blocked with 
10% (v/v) fetal bovine serum (FBS), 5% (v/v) donkey serum, and 5% (v/v) goat 
serum for 2 hours at room temperature. Samples were then stained with primary 
antibodies, including anti-Vimentin antibody (ab92547, Abcam), anti-PyMT 
antigen antibody (ab15085, Abcam), anti-E-cadherin antibody (7870, Santa-Cruz), 
and anti-Ki67 antibody (14-5698-82, Thermo Fisher Scientific) at 1:50 diluted in 
PBS with 10% (v/v) FBS, 5% (v/v) donkey serum, and 5% (v/v) goat serum 
overnight at 4°C. After washing with PBS supplemented with 0.02% Tween, 
samples were then incubated with secondary antibody, including goat anti-rat Alexa 
Fluor 647 antibody (A21247; Thermo Fisher Scientific) and donkey anti-rabbit 
Alexa Fluor 568 antibody (A10042; Thermo Fisher Scientific) at 1:100 diluted in 
PBS with 10% (v/v) FBS, 5% donkey serum, and 5% goat serum for 1 h at room 
temperature in the dark. Immunofluorescent images were acquired with a Zeiss 
LSM800 confocal microscope using a 40x/1.1 NA water immersion objective and 
405, 568, and 640 excitation laser lines and z-stack imaging. Protein colocalization 
analysis was performed as previously described 216. 
 
Western blot analysis 
 39 
 
Snap frozen tumors extracted from mice were ground using a pre-cooled (-80 °C) 
mortar and pestle. The pulverized tumor was then lysed in Laemmli buffer and 
centrifuged at 14,000 x g at 4°C for 10 min. The supernatant was collected and used 
for Western Blotting. The protein concentration was measured with the DC Assay 
Kit (Bio-Rad), subjected to gel electrophoresis [8% (w/v) acrylamide gel], and 
transferred to PVDF membranes as described previously217. Membranes were then 
blocked with 5% bovine serum albumin (Millipore Sigma) in Tris-buffered saline 
(TBS). Membranes were then incubated with primary antibody overnight at 4 °C, 
washed with TBS supplemented with 0.1% Tween. After washing, membranes were 
stained with secondary antibody for 1 hour at room temperature. Primary antibodies 
were prepared at 1:1000 dilution in 5% bovine serum albumin. Secondary antibodies 
conjugated to horseradish peroxidase were prepared at 1:2000 dilution. Membranes 
were imaged with West Pico or West Dura (Thermo Fisher Scientific) per their 
respective protocols, using an ImageQuant LAS-4000 system. Quantification was 
performed with ImageJ (National Institutes of Health, NIH). Relevant protein 
expression level was expressed as the ratio of the protein of interest to GAPDH. 
 
Atomic force microscopy 
ECM stiffness of tumors extracted from PyMT mice was measured using contact 
mode atomic force microscopy (MFP-3D, Asylum Research). Snap frozen tissue 
samples were sectioned at 20 µm thickness and incubated in PBS supplemented with 
protease inhibitor cocktail (Thermo Fisher Scientific) during the measurement. A 
silicon nitride cantilever having a nominal spring constant of 0.06 N m−1 and 5 µm 
diameter spherical borosilicate glass bead (Novascan, Boone, IA, USA) was used. 
The probe spring constant was measured prior to each session. Force-displacement 
 40 
 
curves were obtained by indenting 2 or 3 force maps (10-by-10 grid spaced 10 µm 
apart) on each tissue section with approach and retract speeds of 2 µm s−1 until 
reaching the maximum set force of 3 nN. To calculate the Young’s modulus from 
each indentation, force-displacement curve was fitted to the Hertz model assuming 
a Poisson’s ratio of 0.5.  
 
Immunohistochemistry 
Tumor tissue sections extracted from MMTV-PyMT mice were fixed with 4% (v/v) 
paraformaldehyde in PBS for 10 min at room temperature. After incubating with 
primary antibodies overnight at 4°C, endogenous peroxidases in the samples were 
oxidized by exposure to 0.3% H2O2 in TBS for 15 min at room temperature. Samples 
were then incubated with secondary antibody, HRP-conjugated anti-rabbit goat 
antibody (611-103-122, Rockland Immunochemicals, Limerick, PA, USA) at 1:400 
diluted in TBS with 10% (v/v) FBS 10% goat serum for 1 h at room temperature. 
Samples was then incubated with DAB (Cell Signaling) for signal development and 
with Mayer’s Hematoxylin for counterstaining. After dehydration and mounting, 
tissue sections were then digitalized using Leica SCN400 Slide Scanner. Signal 
intensity and percentage of signal positive area were quantified with Leica Digital 
Image Hub. 
 
Collagen deposition assay 
Fixed tumor sections collected from PyMT mice were hydrated and stained using a 
Picrosirius red stain kit (24901-250, Polysciences, Inc.) according to manufacturer’s 
instructions. After dehydration and mounting, quantitative polarization microscopy 
was performed on the sections as previously described using an inverted Axiovert 
 41 
 
microscope equipped with a rotatable linear polarizer and a circular polarizer with a 
Zeiss Axiocam 506 color camera218. Quantification of optical retardance was 
performed using the pre-established method218. Briefly, average retardance of a 
randomly selected region of interest from each image with background subtraction 
was used to quantify collagen content in the tumor sections. 
 
Unconfined compression assay 
Snap frozen tumors from PyMT mice were thawed in PBS with protease inhibitor 
cocktail immediately before mechanical testing was performed as described 
previously151. Tumor mechanical properties were measured using a TA Electroforce 
Model 3100 (TA Instruments). Tissues were subjected to 15% strain with 3% 
stepwise displacements over 5 steps. Each step consisted of a 1 mm indentation with 
a 15 min relaxation time. Using a custom MATLAB script, the equilibrium modulus 
was calculated from the slope of the stress-strain curves generated using a 
poroviscoelastic model.  
 
Statistical analysis 
Statistical analysis was all performed with GraphPad Prism 8.0a (GraphPad 
Software, La Jolla, CA, USA). Data here are presented as mean ± standard error of 
mean (SEM). Parametric one- or two-tailed one-way analysis of variance (ANOVA) 
followed by a Mann-Whitney tests test or a post-hoc Tukey’s Honest Significant 
Difference test were used where appropriate. p < 0.05 was considered statistically 
significant. 
 42 
  
 
 
2.4 Results 
Establishment of diabetic hyperglycemia in PyMT mouse model pre-tumorigenesis 
To address the interplay between diabetes and the progression of other diseases, 
including breast cancer, previous studies have used diabetic mouse models generated 
by inducing genetic mutations, injecting mice with different dosages of streptozocin 
(STZ), or providing a high fat diet214,219,220. These methods have enabled researchers to 
induce hyperglycemia within mice of several genetic backgrounds. To create a model 
where we can observe the effects of diabetic hyperglycemia on mammary 
tumorigenesis, we have tested several permutations of one of these commonly used 
methods, STZ injections, on MMTV-PyMT mice which develop spontaneous mammary 
tumors. These regimens of STZ injections either fail to elevate blood glucose at all or 
elevated glucose at a delayed rate so that diabetic levels were not reached until week 8 
when tumor progression already initiated. Primary tumor progression within the 
mammary glands of MMTV-PyMT mice generally starts when mice are at 6 weeks of 
age221. This makes current existing methods not suitable for studying how pre-existing 
hyperglycemia affects tumor progression and tumor ECM properties.  
 
To induce hyperglycemia in PyMT mice prior to tumor formation, we fed the mice a 
high fat diet starting at 4 weeks of age and injected the mice with 5 daily consecutive 
70 mg kg-1 doses of STZ at the 5 weeks of age (Figure 2.1A). To verify the induction of 
diabetic hyperglycemia, blood glucose levels were measured weekly pre- and post-
fasting. Compared with non-diabetic PyMT control mice, blood glucose levels were 
 43 
 
significantly increased in the PyMT mice injected with STZ (Figure 2.1B,C). The blood 
glucose of STZ injected mice increased to above 400 mg dL-1 after STZ injection and 
remained stably elevated at diabetic levels over the entire course of the study. To 
evaluate health status of mice during experiments, body weight of mice was also 
measured weekly until the end of experiments. The body weight of both control and 
STZ injected mice increased steadily over the course of the study (Figure 2.1D).  
 
To further validate the successful induction of diabetic hyperglycemia, glucose 
tolerance and insulin tolerance of mice were tested at week 6. In both assays, glucose 
levels of STZ injected mice were increased compared with control group mice at each 
time point tested (Figure 2.1E,F). Noting that STZ induces diabetic hyperglycemia 
through destroying pancreatic beta islet cells, we next sought to compare beta cell 
concentration in diabetic and non-diabetic mice. Using immunohistochemical staining 
for insulin in pancreases collected from diabetic and non-diabetic mice at the study 
endpoint, we found that the percentage of beta islet cells in pancreases of the diabetic 
mice was significantly decreased (0.364 +/- 0.02064%) compared with that of non-
diabetic mice (1.506 +/- 0.1779%) (Figure 2.1G,H). Together, these results showed that 
diabetic hyperglycemia was induced in MMTV-PyMT mice successfully at week 6, 
before tumor formation. 
 44 
 
 
Figure 2.1. Induction of diabetes mellitus within MMTV-PyMT mice. (A) 
Schematic overview of diabetic MMTV-PyMT mouse model. (B) Pre-fasting and (C) 
post-fasting blood glucose level of MMTV-PyMT mice injected once daily with 5 
consecutive daily 70 mg/kg doses of either streptozotocin (STZ) or sodium citrate 
buffer vehicle control (Ctrl). (D) Body weight tracking of MMTV-PyMT mice 
injected once with 5 consecutive daily 70 mg/kg doses of either sodium citrate (Ctrl) 
or streptozotocin (STZ). (E) Glucose levels over time post-insulin injection for Ctrl 
group and STZ group. (F) Glucose levels over time post-glucose injection for Ctrl 
group and STZ group during glucose tolerance assay. (G) Representative images of 
immunohistochemical staining for insulin in the pancreases of MMTV-PyMT mice 
injected with either sodium citrate (Ctrl) or streptozotocin (STZ). (H) Quantification 
of beta islet cell percentage in the control group or mice injected with STZ (Ctrl: N = 
7, STZ: N = 5). Data are presented as mean ± SEM. *** p < 0.001, **** p < 0.0001. 
 
 
Diabetic hyperglycemia increases tumor matrix stiffness 
Noting that hyperglycemia results in increased glycation of ECM in numerous 
tissues127,222, we next verified whether hyperglycemia increased glycation within tumor 
ECM. Non-enzymatic glycation was assessed by quantifying the concentration of 
advanced glycation end products (AGEs) in diabetic and non-diabetic tumors using 
immunohistochemical staining and Western Blotting. While tumors collected from 
diabetic mice showed increased AGEs formations which indicates increased glycation 
(Figure 2.1A-D). Since sugars crosslink collagen molecules through a non-enzymatic 
 45 
 
reaction resulting in stiffened matrix126,223,224, we extended our investigation to assess 
collagen concentration, a major ECM component in mammary tumors, between diabetic 
and non-diabetic PyMT tumors. To assess collagen concentration in diabetic and non-
diabetic tumors, Picrosirius red staining was performed and tumor sections were imaged 
using quantitative polarization microscopy (Figure 2.2E). When the optical retardance 
of the collagen signal in the tumors was quantified, there was no significant difference 
observed between diabetic and non-diabetic tumors indicating that collagen 
concentrations are similar between our experimental groups. To assess if diabetic 
hyperglycemia stiffens tumors via non-enzymatic glycation, mechanical properties of 
diabetic and non-diabetic tumors were tested. To test macro-scale stiffness, unconfined 
compression testing was performed. We found that tumors collected from diabetic mice 
had higher equilibrium modulus compared with that from non-diabetic mice (Figure 
2.2F). Consistent with this result, AFM measurements showed that diabetic 
hyperglycemia increased tumor elastic modulus (Figure 2.2G,H) as well as the spatial 
heterogeneity of tumor stiffness as evidenced by representative force maps (Figure 
2.2I). These data suggest that diabetic hyperglycemia induced glycation stiffens breast 
tumors and increases heterogeneity of tumor stiffness. 
 46 
 
 
Figure 2.2. Hyperglycemia promotes glycation and increases ECM stiffness. (A) 
Representative images of tumor sections with IHC staining for AGEs collected from 
diabetic (STZ) and non-diabetic (Ctrl) mice. (B) Corresponding quantification of 
normalized AGEs positive area (Ctrl: N = 5, n = 14; STZ: N = 4, n = 17). (C) 
Representative Western blot protein bands for AGEs and GAPDH in tumors from 
diabetic (STZ) and non-diabetic (Ctrl) mice. (D) Corresponding quantification of 
AGEs normalized by GAPDH (N = 4, n = 6). (E) Quantification of collagen 
deposition of diabetic and non-diabetic tumors (N = 3, n = 3; 30 measurements per 
condition). (F) Unconfined compression assay showing equilibrium modulus of 
tumors extracted from diabetic or non-diabetic mice (Ctrl: N = 4, n = 6; STZ: N = 4, 
n = 5; 20 measurements per condition). (G) Elastic modulus of diabetic (STZ) or non-
diabetic (Ctrl) tumors measured by AFM (N = 3, n=3, 250~300 measurement per 
condition). (H) Corresponding histogram of AFM measurements for diabetic (STZ) 
and non-diabetic (Ctrl) tumors. (I) Representative force map of tumors extracted from 
diabetic (STZ) or non-diabetic (Ctrl) mice. Data are presented as mean ± SEM. * p < 
0.05, *** p < 0.001, **** p < 0.0001. 
 
 
Diabetic hyperglycemia promotes tumor progression in PyMT mice 
Since we determined that hyperglycemia induced glycation increases PyMT tumor 
stiffness, and ECM stiffness is known to promote tumor malignancy35,139,225, we next 
 47 
 
studied the progression of mammary tumors in our mouse model. Tumor volume of 
diabetic or non-diabetic mice was measured weekly using calipers until the study 
endpoint at week 12. Diabetic mice displayed accelerated tumor progression (Figure 
2.3A) and ultimately generated larger tumors (Figure 2.3B,C) compared with non-
diabetic group.  Because diabetic hyperglycemia promotes breast cancer growth in 
PyMT mice, tumor aggressiveness was further assessed with hematoxylin and eosin 
staining followed by tumor grading by a blinded veterinary pathologist (Figure 2.3D). 
To determine if increased tumor cell proliferation was responsible for the increased 
tumor growth observed in diabetic mice, we quantified the percentage of cells with 
Ki67+ nuclei using immunofluorescence staining (Figure 2.3E). The results indicated a 
significant increase in cell proliferation in the tumors collected from diabetic mice 
compared with control tumors (Figure 2.3F).  
 
Hyperglycemia causes PyMT tumor cells to undergo epithelial-mesenchymal transition 
Since previous studies have shown that matrix stiffening promotes tumor cell epithelial-
to-mesenchymal transition (EMT)226, which is also associated with increased tumor 
aggressiveness and worsened patient prognosis, we next investigated whether 
hyperglycemia induced glycation affects tumor progression through upregulating tumor 
cell EMT. We used immunofluorescence staining and confocal microscopy of diabetic 
and non-diabetic tumor sections stained for epithelial marker, E-cadherin, and 
mesenchymal marker, vimentin, to assess EMT status of the tumors (Figure 2.4A). The 
ratio of vimentin expression to E-cadherin expression was used as our metric for extent 
of tumor cell EMT (Figure 2.4B). Diabetic tumors showed an increase in EMT ratio  
 48 
 
 
Figure 2.3. Hyperglycemia increases tumor growth through promoting cell 
proliferation. (A) Weekly average tumor volume measurements for diabetic (STZ) 
and non-diabetic (Ctrl) mice from when tumors become palpable and large enough 
for caliper measurements (week 9) to study endpoint (Ctrl: N = 14, STZ: N = 14). (B) 
Average tumor volume for STZ injected (STZ) and control (Ctrl) mice at study 
endpoint (Ctrl: N = 7, n = 22; STZ: N = 9, n = 30). (C) Average tumor burden per 
mouse for STZ injected (STZ) and control (Ctrl) mice at study endpoint (Ctrl: N = 7, 
STZ: N = 9).  ( (D) Tumor differentiation grading of tumors extracted from control 
group and mice treated with STZ (N = 3, n = 3). (E) Representative images showing 
Ki67 and nucleus co-localization within diabetic (STZ) and non-diabetic (STZ) 
tumors. (F) Corresponding quantification of percentage of cells with Ki67+ nuclei (N 
= 3, n = 3, 21 imaging fields included per condition). Data are presented as mean 
± SEM. * p < 0.05, ** p < 0.01. 
 
 
compared to control tumors, indicating that diabetic hyperglycemia is causing a shift 
towards EMT. Since increased expression of fibronectin has been associated with the 
pro-metastatic breast cancer phenotype and induced changes in cell morphology and 
expression of EMT markers227, fibronectin expression of diabetic and non-diabetic 
 49 
 
tumors was compared with IHC staining. Quantification of fibronectin positive area 
showed that tumors of diabetic mice have higher fibronectin expression than the ones 
from non-diabetic mice. Besides fibronectin, TGF-𝛽 has also been shown to regulate 
EMT through multiple signaling pathways and transcriptional reprogramming, which 
further inactivates genes encoding epithelial proteins, such as E-cadherin, and activates 
genes encoding mesenchymal proteins, such as vimentin228,229. Thus, we also tested 
whether diabetic hyperglycemia regulates TGF-𝛽. Our results showed diabetic tumors 
have higher TGF-𝛽 expression compared with non-diabetic tumors (Figure 2.4E,F). 
These results were also confirmed with Western blotting showing that tumors extracted 
from diabetic mice expressed higher levels of TGF-𝛽 and fibronectin compared to 
control tumors (Figure 2.4G-I). Altogether, our results suggest that diabetic tumors are 
shifted further along the EMT spectrum towards an aggressive mesenchymal phenotype 
compared to non-diabetic tumors.  
 
Glycation inhibition results in reduced ECM stiffness 
Noting that glycation is a multistep, chemical reaction that increases ECM stiffness, we 
next tested whether glycation inhibition via aminoguanidine (AG) or AGE crosslink 
breaker, alagebrium (ALT-711), could ameliorate the observed phenotypes in mice with 
diabetic hyperglycemia. To determine whether glycation inhibition affected AGE 
concentration in STZ-injected mice, immunohistochemistry staining was performed 
(Figure 2.5A). When quantified, AGE concentration was increased in STZ-injected 
mice that received either inhibitor treatment compared to STZ only control mice (Figure 
2.5B). Next, collagen concentration was assessed as before using Picrosirius red  
 50 
 
 
Figure 2.4. Diabetes promotes tumor cell EMT. (A) Representative images of 
immunofluorescence staining showing E-cadherin and Vimentin of tumors extracted 
from diabetic (STZ) and non-diabetic (STZ) tumors. (B) Corresponding 
quantification of expression ratio of E-cadherin and vimentin (N = 3, n = 5, 37 fields 
per tumor section were imaged). (C) Representative images showing fibronectin (FN) 
expression within tumors extracted from diabetic (STZ) or non-diabetic (Ctrl) mice. 
(D) Corresponding quantification of fibronectin positive area within tumors (Ctrl: N 
= 4, n = 13; STZ: N = 5, n = 17). (E) Representative images of TGF-𝛽 stained tumor 
sections extracted from mice treated with (STZ) or without (Ctrl) STZ. (F) 
Corresponding quantification of TGF-𝛽 positive area within tumors (N = 3, n = 6). 
(G) Protein bands generated by western blotting showing fibronectin (FN) and TGF-
𝛽 (TGF-𝛽) expression within diabetic (STZ) and non-diabetic (Ctrl) tumors. GAPDH 
was utilized as loading control. (H & I) Corresponding quantification of fibronectin 
(H) and TGF-𝛽 (I) expression normalized over GAPDH expression (N = 3, n = 6). 
Data are presented as means ± SEM. * p < 0.05, ** p < 0.01, *** p < 0.001. 
 
 
staining of tumor sections (Figure 2.5C). When the optical retardance of the tumor 
sections was quantified, we observed no significant difference in mean optical 
retardance between STZ-injected mice that received inhibitor treatment and STZ-only 
 51 
 
control mice. After determining that collagen concentration was not impacted by 
glycation inhibition, bulk mechanical stiffness was assessed using unconfined 
compression testing. Here, we observed that both AG and ALT-711 treatments lowered 
STZ-injected mouse tumor stiffness to that of control non-diabetic mouse tumors 
(Figure 2.5D). To compare microscale stiffness, AFM measurements were performed 
on tumor samples with and without glycation inhibition. Average equilibrium modulus 
for tumors from diabetic mice treated with either inhibitor was decreased compared to 
those from diabetic control mice (STZ only) (Figure 2.5E). The distribution of 
equilibrium modulus values shows an overall shift towards lower values along with an 
absence of upper range values in AG or ALT-711 treated diabetic tumors compared to 
untreated diabetic controls (Figure 2.5F). These experiments suggest that the increased 
stiffness in tumors from mice with diabetic hyperglycemia is due to increased glycation, 
which is successfully abrogated to non-diabetic control tumor levels by either AG or 
ALT-711 AGE inhibiting treatments, resulting in lower tumor ECM stiffness despite 
mice still presenting with diabetic hyperglycemia. 
 
Glycation inhibition attenuates primary tumor progression 
As ECM stiffness is known to promote malignancy, and glycation inhibition via AG or 
ALT-711 treatment decreased ECM stiffness in STZ-injected mice, we next assessed 
whether glycation inhibition also altered tumor progression.  Blood glucose of mice was 
measured weekly to verify that either glycation inhibitor did not affect blood glucose 
throughout the experiment (Figure 2.6A). By measuring tumor volume weekly, we 
observed that tumor volume over time for the diabetic mice with either AG or ALT-711  
 52 
 
 
Figure 2.5. Glycation inhibition decreases ECM stiffness. (A) Representative 
images of tumor sections with IHC staining for AGEs collected from diabetic mice 
treated with AG or ALT-711. (B) Corresponding quantification of normalized AGEs 
positive area (STZ: N = 4, n = 8; STZ + AG: N = 3, n = 7; STZ + ALT-711: N = 4, n 
= 7). (C) Quantification of collagen deposition of diabetic and non-diabetic tumors 
(STZ: N = 5, n = 7; 349 measurements per condition; STZ + AG: N = 2, n = 5; 126 
measurements per condition; STZ + ALT711: N = 4, n = 8; 284 measurements per 
condition). (D) Unconfined compression assay showing equilibrium modulus of 
tumors extracted from diabetic mice treated with glycation inhibitors, AG or ALT-
711 (STZ: N = 6, n = 6, 40 measurements included; STZ + AG: N = 5, n = 7, 20 
measurements included; STZ + ALT-711: N = 3, n = 6, 20 measurements included). 
(E) Elastic modulus of diabetic tumors treated with or without glycation inhibitors 
measured by AFM (N = 3, n=3, 550-860 measurements per condition). (F) 
Corresponding histogram of AFM measurements of diabetic tumors treated with or 
without glycation inhibitors. Data are presented as mean ± SEM. * p < 0.05, **** p < 
0.0001. 
 
 
treatment was significantly lower than the volume of diabetic STZ only control tumors 
(Figure 2.6B). At study endpoint, average tumor volume was significantly lower for 
STZ-injected mice with either inhibitor treatment compared to STZ only control mice 
(Figure 2.6C). Additionally, the average total tumor burden per mouse was significantly  
 53 
 
 
 
Figure 2.6. Glycation inhibition decreases primary tumor progression. (A) Blood 
glucose level of diabetic mice treated with or without glycation inhibitors (STZ N = 
12, STZ + AG N = 10, STZ + ALT-711 N = 12). (B) Weekly average tumor volume 
measurements for diabetic mice treated with AG or ALT-711 from when tumors 
become palpable and large enough for caliper measurements (week 9) to study 
endpoint (STZ N = 12, STZ + AG N = 6, STZ + ALT-711 N = 9). (C) Average tumor 
volume for diabetic mice treated with glycation inhibitors at study endpoint (STZ: 
N = 12, n = 89; STZ + AG: N = 6, n = 18; STZ + ALT-711: N = 9, n = 54). (D) 
Average tumor burden per mouse for diabetic mice treated with glycation inhibitors 
at study endpoint (STZ N = 18, STZ + AG N = 9, STZ + ALT-711 N = 10). (E) 
Quantification of percentage of cells with Ki67+ nuclei (STZ: N = 5, n = 7, 56 imaging 
fields included; STZ + AG: N = 3, n = 3, 32 imaging fields included; STZ + ALT711: 
N = 4, n = 8, 64 imaging fields included). (F) Representative images showing Ki67 
and nucleus co-localization within diabetic tumors treated with or without glycation 
inhibitors.  (G) Representative images of immunofluorescence staining showing E-
cadherin and vimentin of tumors extracted from diabetic mice treated with AG or 
ALT-711. (H) Corresponding quantification of expression ratio of E-cadherin and 
vimentin (STZ: N = 5, n = 7, 52 imaging fields included; STZ + AG: N = 3, n = 4, 25 
 54 
 
imaging fields included; STZ + ALT711: N = 4, n = 8, 64 imaging fields included). 
(I) Representative images of TGF-𝛽 stained tumor sections extracted from diabetic 
mice treated with or without glycation inhibitors. (J) Corresponding quantification of 
TGF-𝛽 positive area within tumors (STZ: N = 4, n = 8; STZ + AG: N = 3, n = 7; STZ 
+ ALT711: N = 4, n = 8). (K) Corresponding quantification of fibronectin positive 
area within tumors (STZ: N = 4, n = 8; STZ + AG: N = 3, n = 7; STZ + ALT711: N 
= 4, n = 8). (L) Representative images showing fibronectin expression within tumors 
extracted from diabetic mice treated with or without glycation inhibitors. Data are 
presented as means ± SEM. * p < 0.05, ** p < 0.01, **** p < 0.0001. 
 
 
lower for glycation inhibitor-treated diabetic mice compared to diabetic only controls 
(Figure 2.6D). When stained with hematoxylin and eosin (H&E), tumor sections were 
graded by a pathologist (Figure 2.6E). Grading results indicated attenuated tumor 
progression for both inhibitors treated diabetic tumor conditions compared to the 
relatively more aggressive diabetic control tumors (Figure 2.6F). To assess tumor cell 
proliferation, Ki67 staining was performed, and the percentage of Ki67-positive cells 
was quantified (Figure 2.6G). We found that significantly less cells were Ki67 positive 
in glycation inhibited diabetic tumors compared to diabetic control tumors (Figure 2.6G, 
H). Additionally, EMT status was assessed by staining tumor sections for E-cadherin 
and vimentin and calculating the expression ratio of these EMT markers for AG and 
ALT-711 treated and untreated diabetic mouse tumors (Figure 2.6I,J). When quantified, 
we observed the EMT ratio was lower for glycation inhibitor treated STZ tumors 
compared to STZ only controls, indicating less EMT relative to STZ only tumors. 
Immunohistochemistry was used to assess fibronectin expression for diabetic tumors 
with and without glycation inhibitor treatments (Figure 2.6K). When the percentage of 
fibronectin-positive tumor area was quantified, inhibitor treated STZ-injected mouse 
tumors were observed to have reduced fibronectin positive area compared to STZ-only 
 55 
 
control mouse tumors (Figure 2.6L). TGF-beta was also stained for in glycation 
inhibitor treated and untreated diabetic mouse tumors using immunohistochemistry 
(Figure 2.6M,N). Quantification of TGF-beta-positive tumor area indicates that TGF-
beta expression is reduced in both AG- and ALT-711-treated diabetic tumors compared 
to untreated diabetic controls. Cumulatively, these results suggest that glycation 
inhibition by either AG or ALT-711 treatment reduces tumor aggressiveness in mice 
with diabetic hyperglycemia. 
 
2.5 Discussion 
In the present study, we have established a novel mouse model where diabetic 
hyperglycemia is established in female MMTV-PyMT mice via streptozocin injections 
prior to spontaneous mammary tumor generation. Using this model, we have elucidated 
a novel role for non-enzymatic glycation in promoting tumorigenesis in diabetic 
hyperglycemic conditions. When compared to non-diabetic tumors, we found that 
diabetic mouse tumors were stiffer, more proliferative, and further along the EMT 
spectrum, all of which correlate with worsened prognosis. Using AGE inhibitors, we 
mechanistically demonstrated that AGE mediated crosslinking uniquely contributes to 
the enhanced tumorigenesis present in hyperglycemic mice.  This model shows great 
potential in further probing the synergy between diabetes and cancer progression, and 
our findings highlight the importance in considering the interplay between clinically 
relevant comorbidities, especially as diabetes and cancer both rank in the top ten causes 
of death in the United States. 
 
 56 
 
Here, we have provided a mechanistic study detailing the role of AGE mediated 
crosslinking in tumor progression in the context of diabetic hyperglycemia. The effects 
of glycation and AGEs as a result of chronic hyperglycemia are far reaching and have 
been implicated in numerous diabetes-associated complications including retinopathy, 
neuropathy, nephropathy, and cardiomyopathy127. While other studies have focused on 
the chemical signaling role of AGEs and their receptors, RAGEs, in cancer progression, 
we focus on the mechanical role of non-enzymatic glycation, where AGEs are a 
bioactive byproduct of this reaction230,231. To parse apart the effect of the mechanical 
crosslinking and AGE signaling, we implemented treatment with two inhibitors, 
aminoguanidine and alagebrium, which affect different steps in the non-enzymatic 
glycation reaction. As aminoguanidine inhibits AGE formation, this inhibitor blocks 
both the chemical and mechanical effects of non-enzymatic glycation232. Alagebrium 
acts as an AGE crosslink breaker which will allow AGEs to form but removes the 
crosslinks, effectively allowing chemical signaling while negating the mechanical 
effects of non-enzymatic glycation233. As we found similar effects across all metrics for 
both aminoguanidine and alagebrium treated diabetic mice, our findings suggest that 
AGE mediated ECM glycation plays a critical role in the increased aggressiveness of 
mammary tumors in hyperglycemic mice that cannot be compensated for with AGE-
RAGE signaling alone. 
 
This study emphasizes the multi-faceted effects of AGEs which have been implicated 
in numerous pathologies including diabetes, cancer, atherosclerosis, rheumatoid 
arthritis, bone fragility fracture, kidney failure, Alzheimer’s disease, and premature 
 57 
 
aging211,212. While efforts have been made to develop and test a variety of AGE 
inhibiting drugs, clinical trials have produced mixed results211,234. As our study along 
with ample experimental and clinical evidence shows that outcomes are worsened for 
breast cancer patients with pre-existing diabetes, this subset of diabetic cancer patients 
may uniquely benefit from the addition of AGE inhibiting treatment to the traditional 
therapeutic strategy123. One commonly used drug, metformin, which lowers blood 
glucose and is a first line agent in treatment of type II diabetes, is also known to inhibit 
AGEs235. Interestingly, diabetic patients on metformin have a roughly one-third 
reduction in cancer incidence and mortality compared to patients on other anti-diabetic 
drugs236. Outside of diabetes, metformin has recently shown incredible promise as a 
complimentary anti-cancer therapeutic alongside standard of care treatment236. Given 
the outcome of our study and the benefits of metformin in both diabetes and cancer 
treatment, further work to delineate the synergy of breast cancer and diabetes and test 
the potential complimentary effects of other AGE inhibiting drugs should be pursued. 
 58 
  
 
CHAPTER 3 
 
PHENOTYPIC SORTING REVEALS A WEAKLY MIGRATORY, HIGHLY 
METASTATIC CELL SUBPOPULATION WHICH METASTASIZES IN AN E-
CADHERIN DEPENDENT MANNER 
 
Portions of this data were contributed by Shawn Carey, Samantha Schwager, Paul 
Taufalele, Wenjun Wang, Jenna Mosier, Nerymar Ortiz-Otero, Tanner McArdle, 
Zachary Goldblatt, Marsha Lampi, Francois Bordeleau, Jocelyn Marshall, Isaac 
Richardson, and Jiahe Li. 
 
3.1 Abstract 
While intratumor genomic heterogeneity can impede cancer research and treatment, less 
is known about the effects of phenotypic heterogeneities. To investigate the role of cell 
migration heterogeneities in metastasis, we phenotypically sorted metastatic breast 
cancer cells into two subpopulations based on migration ability. While migration is 
typically associated with metastasis, when injected orthotopically in vivo, the weakly 
migratory subpopulation metastasized significantly more than the highly migratory 
subpopulation. To investigate the mechanism by which weakly migratory cells 
metastasized while migratory cells did not, we utilized a combination of in vitro and in 
vivo engineered models to examine both subpopulations’ performance at each stage of 
the metastatic cascade, including dissemination from the primary tumor, survival in the 
circulation, extravasation, and colonization in the lung. While both subpopulations 
 59 
 
performed each step successfully, the poorly migratory cells presented as circulating 
tumor cell (CTC) clusters in the circulation, suggesting clustering as a potential 
mechanism behind the increased metastatic fitness of these cells. When E-cadherin 
expression was knocked down in weakly migratory cells, distal metastasis was 
abrogated. When E-cadherin expression was induced in highly migratory cells, 
metastasis increased. CTCs from blood collected from metastatic cancer patients also 
exhibit CTC clustering that correlates with worsened patient outcome. Our results 
demonstrate that the approach of deconvolving phenotypic heterogeneities can reveal 
fundamental insights into metastatic progression. More specifically, our results indicate 
that migration ability does not necessarily correlate with metastatic potential, and E-
cadherin promotes metastasis in phenotypically-sorted breast cancer cell subpopulations 
by enabling clustering of CTCs.  
 
3.2 Introduction 
Genetic and epigenetic subclones generated within the primary tumor generate 
significant phenotypic diversity that greatly complicates cancer diagnosis and 
treatment28. Intratumor heterogeneity has been identified as a prognostic marker 
associated with decreased patient survival237, and it can obscure the most aggressive 
drivers of cancer migration and metastasis1,238,239. Subclones that ultimately drive 
disease progression may be present in the primary tumor at the time of diagnosis in 
almost undetectable amounts28. If these important metastatic subclones could be 
identified and targeted early, patient outcomes would greatly improve. As metastasis is 
a complex, multi-step process, parsing apart the phenotypes which can complete the 
 60 
 
entire metastatic cascade is critical to improving therapeutic strategies. While much has 
been done to elucidate which cellular properties endow tumor cells with the ability to 
perform the various steps of the metastatic cascade, it remains unclear whether success 
in one part of the metastatic cascade is indicative of success in others1.   
 
As tumor cell infiltration into the surrounding stroma is an early step in cancer 
metastasis, tumor cell migration has been considered essential for cancer malignancy240. 
However, clinical evidence as well as in vitro and in vivo studies indicate that 
heterogeneity exists in cancer cell migration ability and mode, both between and within 
individual tumors241. This range of migratory phenotypes is thought to be partially due 
to the crosstalk between cancer cells and the tumor stroma, where extracellular matrix 
(ECM) architectural features can induce specific migratory behaviors242–245. While 
ECM features can account for part of the heterogeneity observed in local tumor 
dissemination, less is known regarding how tumor-intrinsic phenotypic heterogeneity 
impacts cancer cell migration and ultimately, metastasis.  
 
To address intratumor heterogeneity of cellular migration behaviors and isolate 
the role of migration in metastasis, we sorted metastatic cells based on their ability to 
migrate, one of the earliest proposed requirements for metastatic progression246,247. 
While migration is believed to be correlated with metastasis, when these subpopulations 
were used in an orthotopic metastasis model, the poorly migratory subpopulation readily 
colonized distal locations including lungs, liver, and bone, while the highly migratory 
cells showed poor metastatic potential in all tissues. Our data indicate that E-cadherin 
 61 
 
is essential to the metastatic fitness of the weakly migratory subpopulation, and the 
highly migratory subpopulation can be rescued with induction of E-cadherin expression. 
In patient samples, circulating tumor microemboli concentration and patient survival 
time are correlated, and we observed that circulating tumor cells in metastatic breast 
cancer samples were E-cadherin positive. Together, these findings indicate that in vitro 
migration ability does not necessarily correlate with metastatic capability, and E-
cadherin is a major driver of the metastatic phenotype.    
 
3.3 Materials and Methods 
Cell culture 
MDA-MB-231 cells (ATCC Cat# HTB-26) and HEK293T cells (ATCC Cat# CRL-
3216) were maintained in DMEM (Life Technologies) supplemented with 10% fetal 
bovine serum (FBS; Atlanta Biologicals), and 1% penicillin-streptomycin (Invitrogen). 
HUVECs (Lonza) were cultured in EBM medium (Lonza) supplemented with EGM 
SingleQuot Kit (Thermo Fisher Scientific). MCF10CA1a cells (Barbara Ann Karmanos 
Cancer Institute) were maintained in DMEM:F-12 (1:1) (Life Technologies) 
supplemented with 5% horse serum (Invitrogen), 20 ng/mL EGF (Invitrogen), 10 
mg/mL insulin (Sigma-Aldrich), 0.5 mg/mL hydrocortisone (Sigma-Aldrich), 100 
ng/mL cholera toxin (Sigma-Aldrich), and 1% penicillin-streptomycin (Invitrogen). 
SUM159PT cells (BioIVT) were maintained in Ham’s F-12 medium (Life 
Technologies) supplemented with 5% FBS, 10 mM HEPES (Life Technologies), 1 
µg/mL hydrocortisone, and 5 µg/mL insulin. All cells were free of mycoplasma. 
Mycoplasma tested was performed using Universal mycoplasma detection kit (30-
 62 
 
1012K, ATCC) after cells were sorted, when thawing new cells, and prior to use in vivo 
experiments. All cell culture and time-lapse imaging was performed at 37°C and 5% 
CO2.  
 
Materials and reagents 
The following antibodies and reagents were used for Western blot analysis and 
fluorescent imaging: monoclonal anti-GAPDH (Millipore Cat# MAB374); anti-rabbit 
and anti-mouse conjugated to horseradish peroxidase (Rockland Immunochemicals, 
Limerick, PA); monoclonal anti-E-cadherin (Cell Signaling Technology Cat# 3195); 
monoclonal anti-vimentin (Sigma-Aldrich Cat# V6630); polyclonal anti-FAK (Cell 
Signaling Technology Cat# 3285); poly-clonal anti-pFAKY397 (Cell Signaling 
Technology Cat# 3283); polyclonal anti-Ki67 (Santa Cruz Biotechnology Cat# sc-
7846); anti-polyclonal anti-GFP (Thermo Fisher Scientific Cat# A-6455); biotin-
conjugated anti-CD45 (Biolegend Cat# 304004); streptavidin-conjugated Alexa Fluor 
594 (Biolegend); FITC-conjugated pan-cytokeratin (BD Biosciences Cat# 347653); 
Alexa Fluor 568-conjugated anti-rabbit (Thermo Fisher Scientific Cat# A10042); Alexa 
Fluor 488-conjugated anti-mouse (Thermo Fisher Scientific Cat# A-11001); Alexa 
Fluor 488-conjugated phalloidin (Life Technologies); DAPI (Sigma-Aldrich). Human 
fibronectin (BD Biosciences) and Matrigel (growth factor-reduced, Corning Life 
Sciences) were used in transwell migration assays.  
 
Microscopy 
Phase contrast imaging was performed in a custom temperature, humidity, and CO2-
 63 
 
controlled stage of a Zeiss Axio Observer Z1m inverted phase contrast microscope 
equipped with a Hamamatsu ORCA-ER camera and operated by AxioVision software 
(v. 4.8.1.0). Confocal fluorescence and reflectance imaging was performed with a Zeiss 
LSM800 confocal microscope on a Zeiss Axio Observer Z1 inverted stand equipped 
with a long working distance water-immersion C-Apochromat 40x/1.1 NA Zeiss 
objective and a dry 10x/0.3 NA Zeiss objective or a Zeiss LSM700 confocal microscope 
on a Zeiss Axio Examiner Z1 upright stand equipped with a water-immersion Plan-
Apochromat 20x/1.0 NA Zeiss objective. Both confocal microscopes were operated by 
Zen software (v. 2010).  
 
For quantitative polarization microscopy of in vitro tumor spheroids, an in-house system 
was utilized218.  Briefly, on an inverted Axiovision microscope, the rotation of a circular 
polarizer is controlled by a motorized rotation stage (Thorlabs, max speed of 20 degrees 
sec-1) installed in the illumination plane above the microscope condenser with a 
polarizer analyzer positioned in the imaging plane. Images of migrating strands were 
acquired using a 20x 0.8 NA polarization objective, and image sequences were collected 
with the rotating polarizer moving at a 5° interval over a range of 0° to 180° using Zen 
software. The polarized image sequences were then processed with a custom Matlab 
code to obtain a pixel-by-pixel anisotropy map. The resulting anisotropy images were 
then background subtracted and set to the same colorimetric scale using ImageJ.  
 
Single cell migration studies 
For 3D migration assessment, MDA-MB-231 cells were seeded sparsely within 1.5 
 64 
 
mg/mL collagen matrices and 3 mg/mL collagen microtracks prepared from acid-
solubilized type I rat tail tendon collagen as previously described146,248. For all single 
cell migration studies in 3D type I collagen gels, following polymerization at 37°C, 
matrices were overlaid with culture media. Time-lapses of phase contrast images were 
acquired > 200 µm above the bottom surface of 3D matrices. All image analysis was 
performed using Fiji. To accurately represent heterogeneity in the 3D migration 
phenotype, all cells were included in motility analysis unless they divided or interacted 
with other cells during the indicated observation period. For determination of the 
fraction of motile cells, matrices were imaged at 10x for 20 min intervals for 12-24 h at 
least 12 h after polymerization. A cell was considered motile if the cell body displaced 
at least one cell diameter during a 2-h period, and motile fraction was defined as the 
ratio of motile cells to total cells (not including cells that undergoing cell division or 
touching another cell). For motile fraction, each data point represents a single field of 
view containing multiple cells averaged over time.  Following 12 h of spreading in 3D 
matrix and 4 h of spreading in microtracks, single cell migration speeds were quantified 
during 4-8 h observation periods as indicated. Migration traces were created by 
outlining individual cells and plotting the x and y coordinates of the cell centroid over 
8 h with all coordinates shifted so that the cell location at t=0 equals (0,0). 3 independent 
experiments were performed for each migration study.   
 
Cell proliferation assays 
To test the proliferative activity of the MDA-MB-231 subpopulations in vitro, 2,500 
cells/well were seeded into a 96 well plate daily for 5 days.  An MTT cell proliferation 
 65 
 
assay kit (Cayman Chemical) was performed according to manufacturer’s directions. A 
microplate reader (Biotek) was used to collect absorbance measurements. 4 wells per 
condition were processed and 3 independent experiments were performed.   
 
To test the proliferative activity of the MDA-MB-231 subpopulations in vivo, primary 
tumor sections were immunohistochemically stained for Ki-67. Whole slide imaging of 
immunohistochemical stained tissue sections was performed in the Digital Histology 
Shared Resource at Vanderbilt University Medical Center 
(www.mc.vanderbilt.edu/dhsr).  Quantification of the Ki-67 H score in MDA+ and 
MDA- primary tumor sections was performed using Leica Digital Image Hub Tissue 
Image Analyzer (v. 4.0.6).   
 
Contractility assay and traction force microscopy 
For the 3D collagen gel contraction assay, cells were seeded at 300,000 cells/mL in 
floating 1.5 mg/mL collagen matrices. Media was refreshed at day 2. To quantify 3D 
collagen gel contraction, gel areas at day 4 were compared to the initial gel area and the 
change in gel area was normalized to MDAPAR control. At least 3 individual experiments 
with 1-2 gels per condition were performed.   
 
To measure traction force generation, cells were seeded on 5 kPa polyacrylamide 
substrates coated with 100 µg/mL type I collagen (Becton Dickinson). For traction force 
microscopy quantification, traction fields were obtained from bead displacements using 
the LIBTRC analysis library developed by Dr. Micah Dembo249. Data are presented as 
 66 
 
total force magnitude |F|, which is equal to the integral of the traction field over the area 
of the cell. At least 3 individual experiments were performed with at least 15 cells per 
subpopulation. 
 
Spheroid outgrowth assays  
MDA-MB-231 spheroids were generated as previously described146. Briefly, cells were 
harvested and resuspended in spheroid compaction medium containing 0.25% 
methylcellulose (H4100; Stem Cell Technologies), 4.5% horse serum (Life 
Technologies), 18 ng/mL hEGF (Life Technologies), 0.45 µg/mL hydrocortisone 
(Sigma-Aldrich), 9 µg/mL insulin (Sigma-Aldrich), 90 ng/mL cholera toxin (Sigma-
Aldrich), 90 U/mL penicillin (Life Technologies), and 90 µg/mL streptomycin in 
DMEM/F12 (Life Technologies). The cell suspension was seeded into a 96-well round-
bottom microplate (Corning) with 5,000 cells/well, which was then centrifuged at 300 
× g for 5 min at room temperature. Spheroids formed in 2-3 days.  Compacted spheroids 
were embedded in 1.5 mg/mL collagen gels seeded on glass-bottom microplates 
(MatTek Corporation) and incubated at 37 °C for 30 min to polymerize. After 
polymerization, collagen gels were overlaid with culture media.  Confocal microscopy 
z-stacks were taken at 10 µm intervals at 48 h. Maximum z-projection was performed 
to collapse the z-stacks for quantification of the spheroid outgrowth and core area.  
Outgrowth area minus the core area was used for quantification at 48 h250. Only 
spheroids that were > 200 µm away from the bottom or sides of the dish were used. 
 
 
 67 
 
Phenotypic cell sorting  
To purify differentially migratory cells, parental MDA-MB-231 cells (MDAPAR) were 
seeded in a transwell migration assay. Briefly, a thin 1 mg/mL collagen gel (~10 µm 
thickness) was polymerized in an 8-µm pore transwell insert (Corning) for 20 min and 
the coated insert was equilibrated in serum-free DMEM. Cells were plated at 40,000 
 
cells/cm2 on the gel surface in DMEM + 0.5% FBS, and the insert was placed in a 6 
well plate containing DMEM + 10% FBS. On day 2 of culture, the media in the upper 
reservoir was refreshed. On day 4 of culture, media was removed from both the 
transwell insert and well plate and reserved. Transwell inserts were washed twice with 
PBS then 0.25% Trypsin-EDTA was added and the transwell plate was incubated on an 
orbital shaker at 37ºC for 5 min. The wellplate was then removed and tapped gently to 
help loosen cells then placed back on the orbital shaker for another 5 min. Equal volume 
of complete media was added above and below the transwell then cell suspensions were 
removed from both compartments and pooled with their respective reserved media. 
Highly migratory (MDA+) and weakly migratory (MDA-) cell subpopulations were 
recovered by centrifugation from the lower and upper compartment solutions, 
respectively, then counted using a hemocytometer. Purifying cell sorting was achieved 
 
by repeatedly seeding MDA+ and MDA- cells on separate freshly prepared transwell 
migration assays as described above. For purification of MDA+ cells, cells that migrated 
 
to the bottom of the transwell were recovered and reseeded; for purification of MDA-
cells, cells that did not migrate through the transwell were recovered and reseeded. 
Twenty rounds of purification were performed. Subpopulations were used in 
experiments for up to 20 passages following purification with no discernible changes in 
 68 
 
behavior.  
 
Transwell migration assay 
In the transwell migration assay, the fraction of cells migrated was calculated as the 
number of migrated cells in the lower compartment divided by the total number of cells 
in the upper and lower compartments on day 4. Three independent transwells were 
separately quantified for each condition. For transwells coated with fibronectin and 
collagen, fibronectin was mixed with neutralized 1 mg/mL collagen prior to transwell 
coating and allowed to polymerize before hydration and cell seeding. For transwells 
coated with Matrigel and collagen, Matrigel was kept on ice until mixing with 
neutralized 1 mg/mL collagen prior to transwell coating and allowed to polymerize 
before hydration and cell seeding.  
 
For trans-endothelial transwell migration assay, HUVECs were seeded upon 
polymerized collagen coated transwell inserts at 300,000 cells/well and cultured for 3 d 
to allow monolayer formation. To validate endothelial monolayer formation prior to 
cancer cell seeding, extra HUVEC-seeded transwells were subjected to a permeability 
assay where briefly, a FITC-dextran dye was added to the media at the top of the 
transwell and time lapse z-stack confocal imaging was performed to monitor 
concentration of FITC-dextran dye above and below the transwell. The relative 
concentrations above and below the transwell over time over 1 h were used to calculate 
relative permeability. At day 3, cancer cell subpopulations were seeded in the insert as 
before. For all permutations of the transwell migration assay, each experimental 
 69 
 
condition was performed in triplicate and three biological replicates were performed. 
 
Mousework 
Experiments were performed in accordance with AAALAC guidelines and were 
approved by the Vanderbilt University Institutional Animal Care and Use Committee 
(Protocol#: M1700029-00).  6-8 week old female NOD/SCID immunodeficient mice 
(The Jackson Laboratory) were injected with 1x106 MDA+ or MDA- cells 
subcutaneously at the mammary gland.  Bioluminescence imaging (BLI) was performed 
using an IVIS™ Spectrum system (PerkinElmer). 150 mg/kg D-luciferin (Gold 
Biotechnology) was injected intraperitoneally 10 min before imaging.  Mice were 
imaged weekly or biweekly for 4 weeks or until primary tumors approached 100-200 
mm3 in volume. Primary tumor removal surgery followed sterile surgical techniques.  
After tumor removal, mice were monitored for 4 weeks using BLI. All tissue samples 
collected were fixed using 4% paraformaldehyde and then processed for histological 
analysis.  
 
For en bloc tumor collection, mice were injected subcutaneously, and at 4 weeks, the 
tissue encompassing the primary tumor including the overlying skin and the underlying 
peritoneum was removed such that the surrounding stroma was undisturbed. For 
circulating tumor cell isolation, mice were injected subcutaneously as before, and at 4 
weeks, blood was obtained via cardiac puncture. The buffy coat was extracted and 
plated in serum-free DMEM before fixation with 4% paraformaldehyde. For lung 
decellularization, lungs from FVB mice were collected, processed, and seeded as 
 70 
 
previously described201.  
 
Primary tumor growth quantifications 
The short (width [W]) and long (length [L]) axes of tumors were measured as tumors 
became palpable using 6” digital calipers.  The equation used to calculate tumor volume 
was tumor volume = 0.5*(L*W2). The final tumor volume measurement was taken prior 
to surgical tumor removal.  
 
Bioluminescence imaging was also used to monitor primary tumor growth.  Using 
Living Image Software (Perkin Elmer), a circular region of interest (ROI) was generated 
at the subcutaneous cancer cell injection site for each mouse for photon quantification. 
From each ROI measurement, average radiance (photons/s/cm2/sr) was obtained for 
each mouse for each time point.  All subsequent measurements were normalized by the 
initial day 1 measurement for each individual mouse.  The normalized average radiance 
values were then log transformed and plotted. Only the MDA-MB-231 subpopulation 
was compared statistically using multiple Student’s t-tests since MCF10CA1a and 
SUM159 tumors had differing growth rates and were removed at different weeks based 
on caliper measurements. 
 
Quantification of lung and liver metastasis 
To quantify lung and liver metastasis in mice injected orthotopically with MDA-MB-
231 subpopulations, lungs and liver were fixed overnight in 4% PFA, washed, and 
embedded in paraffin. Random sectioning was performed and slides were 
 71 
 
immunohistochemically stained for GFP expression. Whole slide imaging of 
immunohistochemical stained tissue sections was performed in the Digital Histology 
Shared Resource at Vanderbilt University Medical Center 
(www.mc.vanderbilt.edu/dhsr).  Quantification of GFP-positive cells and total cells in 
lung and liver sections was performed using Leica Digital Image Hub Tissue Image 
Analyzer (v. 4.0.6). To obtain the percentage of GFP-positive cells, the number of GFP-
positive cells was divided by total cells in the tissue section and multiplied by 100. 1-2 
slides were used per mouse and 6-9 mice were used for each condition. Additionally, 
liver nodules were quantified by visually examining fixed whole livers under an Zeiss 
SteREO Discovery.V8 dissection microscope prior to embedding, and counting 
individual macroscopic metastatic nodules, which were detectable upon gross 
inspection without exogeneous contrast/staining by their white coloration251.  
 
Ex vivo MDA- single cell migration assay 
To assess whether MDA- metastasized cells retained their migration phenotype ex vivo, 
mice were injected subcutaneously as before, and at 4 weeks, lungs were collected and 
dissociated using a mouse lung dissociation kit (Miltenyi Biotec) with a gentleMACS 
dissociator (Miltenyi Biotec). GFP-positive cells were collected using a BD FACS Aria 
III (BD Biosciences) before migration studies were performed as described above.  
MDA+ cells were also collected via FACS, but the cell count retrieved was too low for 
downstream experiments. 
 
 
 72 
 
En bloc tumor local migration assay 
For en bloc tumor sections, 3 mice per condition were used and 1-2 tissue sections per 
mouse (> 200 µm apart) were used for quantification. Immunohistochemical staining 
for GFP expression (A6455, Life Technologies) was used to identify cancer cells that 
had migrated beyond the primary tumor into the stroma.  Outgrowth index was 
quantified as the number of GFP-positive cells in the stroma normalized by the length 
of the tumor periphery that was interfacing with stroma to account for any potential 
differences in tumor size of relative location of sectioning.  Cell counting and 
measurement of the tumor periphery was performed manually in FIJI using the Cell 
Counter plugin and measure function. 
 
Circulating tumor cell isolation from orthotopically-injected mice 
In circulating tumor cell (CTC) isolation experiments, blood was collected from at least 
3 mice per condition for processing and three independent replicates were performed. 
Cells were stained with DAPI, EpCAM (Abcam, ab71916), and CD45 (Biolegend, 30-
F11). Only EpCAM+ CD45- GFP+ cells were quantified as CTCs. Average CTC number 
per field of view was scaled based on blood volume and plating surface area. For cluster 
percentage quantification, the number of single cells and cell clusters were quantified 
across all fields of view for each sample. All samples were sparsely seeded (< 10 cells 
per field of view) such that there was no difficulty in identifying real cell clusters.   
 
Ex vivo decellularized lung colonization assay 
For lung decellularization, lungs were collected from FVB mice and were processed 
 73 
 
and seeded as previously described201. 0.2x106 MDA+ or MDA- + GFP/luc were seeded 
onto decellularized lung cubes in petri dishes treated with Rain-X (ITW Global Brands) 
to prevent cell attachment to plastic and allowed to attach for 2 h before washing away 
unattached cells. Cells on decellularized lung cubes were cultured in completed DMEM 
for 9 d before fixation, staining with DAPI, and imaging. Colonization index was 
calculated as the number of cancer cells identified by GFP-expression in the field of 
view normalized by the area of the lung ECM visualized with confocal reflectance.  At 
least 15 fields of view were collected for each sample, and 3 independent experiments 
were performed.  
 
RNA sequencing 
RNA was isolated from 2D culture with the RNeasy mini-kit [74104] and RNase-Free 
DNase Set [Qiagen 79254] according to manufacturer instructions. Sequencing was 
performed by the Vanderbilt Technologies for Advanced Genomics core (VANTAGE). 
In brief, RNASeq libraries were prepared using 500 ng of total RNA and the NEBNext® 
Ultra™ II RNA Library Prep (NEB, Cat: E7765S) per manufacturer’s instructions, with 
mRNA enriched via poly-A-selection using oligoDT beads. The RNA was then 
thermally fragmented and converted to cDNA, adenylated for adaptor ligation and PCR 
amplified.  The libraries were sequenced using the NovaSeq 6000 with 150 bp paired 
end reads.  RTA (version 2.4.11; Illumina) was used for base calling and analysis was 
completed using MultiQC v1.7. HISAT2 was used to map sequencing reads to the 
human genome (Grch38), then HTSeq was used to obtain gene level counts, and 
differential gene expression was analyzed via DESeq2. GO-term analysis was derived 
 74 
 
from significantly differentially expressed genes with a log2 fold change greater than 2 
and completed using the MSigDB (https://www.gsea-msigdb.org/gsea/msigdb). 
Epithelial and mesenchymal scores were computed from the average normalized 
expression of a manually curated list of epithelial and mesenchymal genes252,253.  
 
Kaplan-Meier Plots 
Kaplan Meier plots were created using the KM-Plotter tool254. The mRNA gene chip 
analysis for breast cancer was selected. CDH1 (201131_s_at) was selected for analysis. 
Trichrotomization of the patient data into upper and lower terciles was performed. For 
distal metastasis free survival, 1803 samples were included in the analysis. For overall 
survival, 1402 samples were included in the analysis. Redundant samples were removed 
and biased samples were not included in either analysis.  
 
Metastatic cancer patient sample criteria 
Cancer patient blood was obtained from patients with stage IV cancer from Guthrie 
Corning Hospital, after informed consent (IRB#1402-17). Samples were shipped 
overnight and processed as described previously255,256. Patient outcomes were 
monitored post-collection for up to 2 years. See Table 1 for patient demographic and 
clinical outcome information for samples used in circulating tumor cell clustering 
analysis. For this experiment, samples from any type of metastatic cancer were 
accepted.  See Table 2 for patient demographic and clinical outcome information for 
samples used in staining circulating tumor cells for E-cadherin.  Samples from any 
metastatic breast cancer patient was accepted regardless of subtype. For all experiments 
 75 
 
involving clinical samples, researchers were blinded to patient demographics and 
clinical outcomes. 
 
Metastatic cancer patient circulating tumor cell isolation 
For CTC isolation from human cancer patient blood samples, blood was separated using 
a Ficoll gradient with 15 min centrifugation at 2000 x g to isolate the buffy coat.  Buffy 
coat was washed twice with HBSS and incubated with CD45 magnetic MicroBeads 
(human, Miltenyi Biotec) for negative selection of leukocytes.  Unbound beads were 
removed by centrifugation at 300 x g for 5 min.  Cell suspensions were passed through 
a column in the presence of a magnetic field to remove leukocytes. After 3 washes, the 
cell suspension that passed through the column was fixed with 4% paraformaldehyde 
and cytospun onto glass slides. Isolated cells were immunostained for cytokeratin, 
CD45, and DAPI to confirm tumor cells and clusters were quantified after isolation and 
scaled by blood volume and plating surface area.  Only cells that were CD45-negative, 
cytokeratin-positive, and with an intact nucleus (observed with DAPI) were considered 
human cancer cells and quantified.  5 random fields of view were taken at 20x 
magnification for each sample.  Cell counts were normalized to sample blood volume 
and used slide surface area. For metastatic breast cancer patient samples, staining for E-
cadherin was also performed. 
 
Quantitative RT-PCR 
mRNA was isolated from target cells using PureZOL RNA isolation reagent (Bio-Rad). 
First strand cDNA was synthesized from RNA template using iScript Select cDNA 
 76 
 
Synthesis Kit (Bio-Rad). Quantitative reverse-transcriptase (RT)-PCR was done using 
SYBR green (Thermo Fisher Scientific) for quantification of double-stranded DNA. 
Relative expression was determined with the 2−ΔΔCT method, and the housekeeping gene 
transcript B-actin was used to normalize the results. The primers used for the CDH1 real 
time PCR were: GGAAGTCAGTTCAGACTCCAGCC (sense) and 
AGGCCTTTTGACTGTAATCACACC (antisense). 
 
Western blotting  
Protein was extracted from cells seeded on tissue culture plastic at 70-80% confluency 
using preheated (95°C) 2× Laemmli sample buffer and subjected to SDS-PAGE. 
Western blots were performed using standard methods257. Horseradish peroxidase 
signal was revealed using SuperSignal West Pico or West Femto kits (Thermo 
Scientific). For quantification of total protein expression, total protein signal was 
normalized to GAPDH loading control. Densitometry was performed using Fiji. 
 
Statistical analysis  
All statistical analysis for in vitro and in vivo studies was performed using GraphPad 
Prism Software (v. 7.0a). Unpaired, two-tailed Student’s t-tests, Mann-Whitney tests, 
ordinary, one-way ANOVAs, and Kruskal-Wallis H tests were performed as appropriate 
with p < 0.05 as the cutoff for statistical significance. All data are shown as mean ± SD 
except for bioluminescence measurements of primary tumor growth which are shown 
as mean ± SEM or box-and-whisker plots, where boxes represent medians and 25th/75th 
percentiles and bars indicate maximum and minimum values. 
 77 
 
3.4 Results 
Sorting metastatic cancer cells based on migration ability to produce distinct, stable 
subpopulations 
MDA-MB-231 human breast cancer cells are known to be genetically and 
phenotypically heterogeneous258,259. To characterize the heterogeneity of their migration 
phenotype, we seeded MDA-MB-231 cells in 3D collagen matrices and performed time-
lapse phase microscopy. The fraction of motile cells increased with time and plateaued 
at approximately 0.65, indicating that not all cells actively migrate (Figure 3.1A). Of 
those that migrate, some cells displace much more than others (Figure 3.1B). A wide 
distribution of net and total displacements exists (Figure 3.1C). To determine whether 
the migration behavior is heritable, and therefore has the potential to be the basis for 
phenotypic sorting, the speed of cells prior to and after cell division was measured. Our 
data indicate that the mother and daughter cell speeds are correlated (Figure 3.1D, 
R2=0.62), suggesting that migration behavior is heritable.   
To sort MDA-MB-231 cells based on migratory ability, MDA-MB-231 
(MDAPAR) cells were seeded in low-serum media on top of a collagen-coated transwell 
with complete media in the bottom chamber. After 4 days, the cells that migrated 
through the collagen and to the bottom of the transwell were collected separately from 
the cells that remained on the top. Both groups were subsequently separately reseeded 
in fresh transwells, and this sorting process was repeated for 20 rounds to produce stable 
strongly migratory (MDA+) and weakly migratory (MDA-) subpopulations (Figure 
3.1E). After 20 rounds of sorting, MDA+ cells were significantly more migratory than 
 78 
 
MDAPAR, which were significantly more migratory than MDA-. This trend persisted 
with no significant difference for either subpopulation between their respective freshly 
sorted cells, cells seeded after a freeze-thaw cycle, and cells that underwent a freeze-
thaw cycle followed by subsequent passaging. This indicates the establishment of two 
stable cell subpopulations that differ in their migration phenotype. 
 
Figure 3.1. Heterogeneity of MDA-MB-231 human cancer cell migratory 
capability is heritable and can be sorted based on migration behavior using an 
in vitro transwell migration assay. a) Fraction of motile MDA-MB-231 cells 
following seeding in 1.5 mg/mL collagen matrix. Dashed line indicates maximal 
 79 
 
motile fraction achieved by steady state. b) Single-cell migration paths and c) total 
and net cell migration speeds over 8 h. d) Correlation of total cell migration speed 
before and after mitosis. Each pair of daughter cells, D1 and D2, is connected by a 
vertical line, and the average daughter cell speed was used to determine correlation 
(R2 = 0.62). e) Schematic of migration cell sorting technique. f) Fraction of cells 
migrated through transwell assay for indicated populations after 4 days. g) Motile 
fraction h) single-cell migration paths and i) migration speeds of MDA-MB-231 
subpopulations and MDA-MB-231 parental cells after seeding in 1.5 mg/mL collagen 
matrix. Data in (a), (f), and (g) display mean ± SD. Statistical significance in (f) was 
calculated using one-way ANOVA (n = 3,4). Statistical significance in (g) was 
calculated using a Mann-Whitney test (n = 13,14). In (i), box and whisker plot show 
medians, 25th/75th, and minimum and maximum values. Statistical significance in (i) 
was calculated using a Kruskal-Wallis H test (n = 34,41,28). * p < 0.05, **** p < 
0.0001, n.s., non-significant. 
 
The sorting process requires cells to migrate through collagen. To assess the robustness 
of the migratory phenotypes and to address the possibility that the response is collagen-
specific, we tested the migration behavior of both MDA-MB-231 cell subpopulations in 
several different matrix compositions and densities. In transwells of increasing collagen 
density, more migration was observed in MDA+ compared to MDA- even as both 
subpopulations showed decreased fractions of cells migrated with increasing collagen 
density (Figure 3.2A). In transwells containing collagen doped with either fibronectin 
or matrigel, no discernable effect on fractions of cells migrated in either subpopulation 
was observed with increasing fibronectin or Matrigel concentrations (Figure 3.2B,C). 
These results demonstrate that MDA+ and MDA- subpopulations retain their migration 
phenotypes across various ECM compositions and densities. 
 80 
 
 
Figure 3.2. Altering transwell ECM coating conditions does not alter the 
migratory behaviors of MDA-MB-231 subpopulations. a) Motile fraction of MDA 
subpopulations in transwell assays with increasing collagen density (n = 3). b) Motile 
fraction of MDA subpopulations in transwell assays with the addition of fibronectin 
to 1 mg/mL collagen coating (n = 3). c) Motile fraction of MDA subpopulations in 
transwell assays with the addition of matrigel to 1 mg/mL collagen coating (n = 3). 
Data in (a),(b), and (c) represent mean ± SD. Light green bars: MDA+, Red bars: 
MDA-. Statistical significance was determined using ordinary one-way ANOVA. n.s., 
nonsignificant. 
 
When seeded in 3D collagen matrix, MDA+ motile fraction was significantly higher 
than both MDAPAR and MDA- motile fractions with MDA- being significantly lower 
than MDAPAR (Figure 3.1G). MDA+ cells displaced readily while MDA- remained 
largely stationary, toggling in place (Figure 3.1H). MDA+ cells exhibited 3D migration 
speeds comparable to MDAPAR, which were both significantly higher than MDA- 
migration speed (Figure 3.1I). These findings suggest that sorting is selecting the most 
migratory (MDA+) and least migratory (MDA-) cells from the parental MDA-MB-231 
(MDAPAR) population.   
Weakly migratory subpopulation exhibits increased metastatic potential and CTC 
clustering compared to highly migratory subpopulation  
 81 
 
After observing a robust difference in migratory abilities between MDA+ and MDA- 
cells, we next sought to assess their metastatic potentials. Since the acquisition of a 
migratory phenotype is associated with dissemination from the primary tumor and is 
often considered essential for metastasis260,261, we hypothesized that MDA+ cells would 
metastasize to a greater extent than MDA- cells. When cancer cell subpopulations were 
injected orthotopically into immunocompromised mice, both subpopulations formed 
primary tumors that grew steadily over the course of four weeks (Figure 3.3A). There 
was no significant difference in primary tumor size upon removal (Figure 3.3B). 
Additionally, subpopulation cell proliferation was compared both in vitro and in vivo 
and found to be similar (Figure 3.4A,B). After primary tumor removal, mice were 
monitored for bioluminescence signal indicating metastatic spread. Counter to our 
hypothesis, distal metastatic spread was observed in all MDA- mice while minimal distal 
signal was observed in MDA+ mice (Figure 3.3C). These findings were confirmed using 
anti-GFP immunohistochemistry (IHC) staining of lung, liver, and bone. Extensive 
metastatic spread was noted in lung, liver, and bone in MDA- -injected mice while only 
rarely were micrometastases observed in MDA+ injected mice (Figure 3.3D). Metastatic 
nodules were observed in MDA- livers while MDA+ livers appear healthy (Figure 3.3E). 
Quantification of IHC staining of tissue sections indicates the presence of significantly 
more GFP-positive cells in MDA- livers compared to MDA+ livers (Figure 3.3F). In 
addition, MDA- livers had significantly more metastatic nodules than MDA+ livers 
(Figure 3.3G). IHC staining of lung tissue sections were also quantified for GFP-
positive cells, and MDA- lungs had significantly more metastatic colonization than 
MDA+ lungs (Figure 3.3H). To determine whether the weakly migratory phenotype of 
 82 
 
MDA- cells that metastasized was preserved, lungs from mice injected orthotopically 
with MDA- cells were dissociated and GFP-tagged MDA- cells were collected via 
fluorescence activated cell sorting. When seeded in 1.5 mg/mL collagen matrix, 
metastasized MDA- cells retained their weakly motile behavior ex vivo (Figure 3.3I). 
Due to the very few metastases that formed in mice injected with MDA+ cells, we were 
unable to collect a sufficient number of GFP-positive cells from MDA+ lungs to make 
a comparison.   
Since this finding that MDA- are highly metastatic and MDA+ are poorly metastatic 
contradicted the prevailing view that robust migration leads to metastasis, we sought to 
determine whether this finding was specific to MDA-MB-231 cells or whether it also 
occurs in other cells. MCF10CA1a and SUM159 cells, two other metastatic breast 
cancer cell lines, were each sorted as before to obtain highly migratory (CA1a+, 
SUM159+)  and weakly migratory (CA1a-, SUM159-)  subpopulations (Figures 3.5A 
and 3.6A). Upon in vitro seeding in 3D collagen, motile fractions were similar to the 
behavior of the sorted MDA-MB-231 subpopulations (Figures 3.5B and 3.6B). 
Similarly, both CA1a-- and SUM159-- injected mice exhibited metastasis to lung and 
lymph nodes to a greater extent than CA1a+- and SUM159+- injected mice (Figures 
3.5C,D and 3.6C,D). These findings confirmed that the anti-correlation between 
migration and metastasis was not cell-type specific.  
 83 
 
 
 84 
 
Figure 3.3. Phenotypically sorted subpopulations show differential metastatic 
potentials in vivo. a) Primary tumor growth quantified by bioluminescence imaging 
using log(normalized average radiance) (n = 4). b) Primary tumor volume as 
measured by calipers (n = 3). c) Endpoint image showing metastasis via 
bioluminescence imaging. d) Representative images of anti-GFP 
immunohistochemical (IHC) staining of lung, liver, and bone samples counterstained 
with hematoxylin. e) Representative images of livers collected at study endpoint and 
f) quantification of percentage of GFP-positive cells in liver anti-GFP IHC sections 
(n = 9). g) Quantification of metastatic liver nodules at study endpoint (n = 5,6). h) 
Representative images of lungs at study endpoint. i) Quantification of GFP-positive 
cells measured via IHC of lung histological sections (n = 8,9). j) Representative 
migration traces of MDA- cells seeded in 1.5 mg/mL collagen after isolation from 
lungs of mice at 4 weeks post injection. Dark grey dashed line represents average 
distance migrated from MDA- cells in vitro with light grey dotted line representing 
standard deviation. k) Schematic of stages of the metastatic cascade assessed by each 
experiment. l) Anti-GFP IHC staining of en bloc tumor sections; white arrow 
indicates local spread at the primary tumor periphery; black arrows indicate cell 
migration into the stroma. m) Quantification of outgrowth index to assess local 
dissemination in en bloc histology (n = 6,5). n) Quantification of circulating tumor 
cells (CTCs) per mL blood in orthotopically injected mice after 4 weeks (n = 3).  o) 
Percentage of clustering of CTCs in mouse blood at 4 weeks (n = 3). p) Relative trans-
endothelial migration fraction of MDA-MB-231 subpopulations (n = 3). q) 
Representative images of ex vivo lung decellularization colonization assay for MDA-
MB-231 subpopulations imaged 9 d post-seeding using confocal reflectance and 
immunofluorescence; GFP-tagged cells: green, nuclei: blue, extracellular matrix: 
white; Scale bars: 50 µm. r) Quantification of colonization index in ex vivo lung tissue 
at 9 d post-seeding (n = 50). Data in (a) display mean ± SEM. Data in (b), (f), (g), (i), 
(m), (n), (p), and (r) display mean ± SD. Statistical significance in (a) was calculated 
using multiple t-tests. Statistical significance in (b), (f), (g), (m), (n), (p), and (r) were 
calculated using unpaired, two-tailed Student t-tests. Statistical significance in (i) was 
calculated using a Mann-Whitney test. * p < 0.05, ** p < 0.01., **** p < 0.0001, n.s., 
nonsignificant. 
 
Since MDA- successfully metastasize whereas MDA+ do not, we sought to determine at 
which stage of the metastatic cascade the phenotypically sorted subpopulations differ 
(Figure 3.3K). To assess each subpopulation’s ability to migrate locally from the 
primary tumor, en bloc sections were collected. Based on GFP signal, both 
subpopulations migrated into stroma adjacent to the primary tumor (Figure 3.3L). 
 85 
 
Quantification of each subpopulation’s outgrowth index revealed that both 
subpopulations migrate locally from the primary tumor; however, MDA+ cells migrate 
to a significantly greater extent compared to MDA- cells (Figure 3.3M), as expected 
based on their in vitro behavior.  
 
 
Figure 3.4. MDA-MB-231 subpopulations have similar proliferation rates in 
vitro and in vivo. a) MTT cell proliferation rate assay performed with MDA+ and 
MDA- cells over 5 days with initial seeding density of 2,500 cells/well (n = 3). b) Ki-
67 H-Score in MDA+ and MDA- primary tumors upon tumor removal at 4 weeks (n 
= 4,5). Data in (a) and (b) display mean ± SD. Statistical significance was determined 
using unpaired, two-tailed student’s t-test. n.s., nonsignificant. 
 
We tested whether both subpopulations could successfully intravasate and survive in 
the circulation. Blood was collected from mice at 4 weeks after tumor induction and 
circulating tumor cells (CTCs) were isolated and stained for EpCAM and CD45. 
GFP+EpCAM+CD45- cells were considered CTCs, which were present in blood at 
similar, high concentrations for both subpopulations (Figure 3.3N). However, MDA- 
CTCs were present in clusters as well as single cells while MDA+ CTCs presented as 
single cell CTCs exclusively (Figure 3.3O). To ensure CTCs were being correctly 
 86 
 
identified, blood from a mouse without cancer cell injections was collected and 
processed alongside experimental samples, and no CTC-like cells were detected (Figure 
3.7A,B,C).  Once CTCs either roll and then attach to the endothelium or lodge in 
capillary beds, they extravasate out of the circulation1. To assess extravasation ability, 
we added the subpopulations to a collagen-coated transwell seeded with an endothelial 
monolayer. The monolayer was validated using a permeability assay (Figure 3.7D). 
After 4 days, both subpopulations migrated into the bottom chamber, although the 
MDA+ cells did so significantly more (Figure 3.3P). To assess colonization ability, lung 
tissue was collected from mice, decellularized, and re-seeded with the subpopulations 
of MDA-MB-231 cells (Figure 3.7E,F). After 9 days of culture, both subpopulations 
successfully colonized the decellularized lung ECM (Figure 3.3Q) to similar extents 
(Figure 3.3R).   
 
Together, these data indicate that despite differences in their migration behaviors, 
both subpopulations can perform key steps of the metastatic cascade. To further 
investigate the mechanism by which MDA- successfully form metastases and MDA+ 
do not, we focused on the presence of CTC clusters in MDA- injected mouse blood, 
also known as circulating tumor microemboli (CTM), as a potential differentiating 
factor between the subpopulations which could explain the enhanced metastatic 
potential of the MDA- subpopulation.   
 
 87 
 
 
Figure 3.5. MCF10CA1a cells can be phenotypically sorted and display 
differential metastatic potentials. a) Fraction of cells migrated for parental and 
subpopulation MCF10CA1a cells post-sorting (n = 3) b) Motile fraction of CA1a 
subpopulations in 1.5 mg/mL collagen (n = 15) c) Primary tumor growth of 
MCF10CA1a subpopulation tumors as quantified by bioluminescence imaging (n = 
4) d) Representative endpoint bioluminescence image of MCF10CA1a 
subpopulation-injected mice. Data in (a) and (b) display mean ± SD. Data in (c) 
display mean ± SEM. Statistical significance was determined using one-way ANOVA 
for (a) and a Mann-Whitney test for (b). * p < 0.05, **** p <0.0001. 
 
 
 88 
 
 
Figure 3.6. SUM159 cells can be phenotypically sorted and display differential 
metastatic potentials. a) Fraction of cells migrated for parental and subpopulation 
SUM159 cells post-sorting (n = 3). b) Motile fraction of SUM159 subpopulations in 
1.5 mg/mL collagen (n = 12). c) Primary tumor growth of SUM159 subpopulation 
tumors as quantified by bioluminescence imaging (n = 4). d) Representative endpoint 
bioluminescence image of SUM159 subpopulation-injected mice. Data in (a) and (b) 
display mean ± SD. Data in (c) display mean ± SEM. Significance was determined 
using one-way ANOVA for (a) and an unpaired, two-tailed student’s t-test for (b).  * 
p < 0.05, ** p < 0.01, *** p < 0.001. 
 
 
 89 
 
 
Figure 3.7. Validation of circulating tumor cell collection, endothelial monolayer 
for trans-endothelial assay, and lung decellularization for ex vivo colonization 
assay. Representative CTC from processed blood from a) MDA+-injected and b) 
MDA--injected mice. c) Representative CD45+ immune cell from mouse that were 
not injected with cancer cells as a negative control for the circulating tumor cell 
collection assay (only CD45+ cells and no EpCAM+GFP+CD45- cells were found in 
control blood). d) Relative permeability of collagen coated transwell inserts with and 
without HUVEC monolayer 3 d post-seeding (n = 22-26). e) Representative image of 
lungs with heart and trachea attached prior to decellularization. d) Representative 
image of decellularized lungs prior to mincing and seeding for ex vivo colonization 
assay. Heart, trachea, and residual connective tissue are removed prior to seeding so 
that only lung tissue is used for the assay. Scale bars: 20µm. Data in (d) displays mean 
± SEM. Significance for (d) was determined using an unpaired, two-tailed student’s 
t-test. **** p < 0.0001. 
 
 
 90 
 
Phenotypically sorted subpopulations exhibit differential EMT gene regulation 
Given these initially counter-intuitive findings that more highly migratory cells are less 
metastatic and weakly migratory cells are highly metastatic, we sought to investigate 
the genotypic differences between our phenotypically sorted subpopulations using 
RNAseq. Initial analysis of sequencing data between MDA+ and MDA- subpopulations 
indicates a large degree of differential gene expression as indicated by the z score heat 
map (Figure 3.8A). As GO term analysis showed biological adhesion and cell motility 
categories were highly differentially regulated between the subpopulations (Table 3.1), 
we hypothesized that differential regulation of genes involved in epithelial-to-
mesenchymal transition (EMT) could explain the phenotypic differences between our 
subpopulations. Using a previously published list of genes involved in identification of 
EMT states253, a z score heat map comparing epithelial and mesenchymal gene 
regulation between MDA+ and MDA- subpopulations was generated (Figure 3.9A). We 
also compared the regulation of these genes between the subpopulations using log 2 fold 
change (Figure 3.8B). In both of these assessments, MDA+ cells showed greater 
upregulation of mesenchymal genes while MDA- cells showed greater upregulation of 
epithelial genes. MDA+ cells had both a significantly increased mesenchymal score and 
decreased epithelial score while MDA- cells showed the opposite relationship (Figure 
3.8C,D).  
 
 91 
 
 
Figure 3.8. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted MDA-MB-231 subpopulations. a) RNAseq z score heatmap 
showing differential gene regulation of MDA+ and MDA- subpopulations (n = 3). b) 
Log 2 fold change of MDA-/MDA+ epithelial-to-mesenchymal transition genes (n=3). 
c) Epithelial and d) mesenchymal scores for MDA+ and MDA- subpopulations (n = 
3). Data in (b) display mean ± SEM. Data in (c) and (d) display mean ± SD. Statistical 
significance in (c) and (d) was calculated using unpaired, two-tailed Student t-tests. 
**** p < 0.0001. 
 
 92 
 
MDA
Gene Set Name # Genes p-value FDR q-value
GO_POSITIVE_REGULATION_OF_MULTICELLULAR_ORGANISMAL_PROCESS 71 8.98E-25 9.22E-21
GO_NEUROGENESIS 66 1.11E-23 5.68E-20
GO_NEURON_DIFFERENTIATION 59 1.69E-22 5.77E-19
GO_POSITIVE_REGULATION_OF_DEVELOPMENTAL_PROCESS 60 2.45E-22 6.3E-19
GO_BIOLOGICAL_ADHESION 59 2.14E-21 4.39E-18
GO_CELL_CELL_SIGNALING 63 7.31E-20 1.22E-16
GO_TUBE_DEVELOPMENT 50 8.31E-20 1.22E-16
GO_NEURON_DEVELOPMENT 49 3.85E-19 4.94E-16
GO_REGULATION_OF_CELL_DIFFERENTIATION 64 6.16E-19 7.03E-16
GO_CELL_MOTILITY 61 1.89E-18 1.89E-15
CA1a
Gene Set Name # Genes p-value FDR q-value
GO_BIOLOGICAL_ADHESION 133 4.95E-49 5.09E-45
GO_EXTRACELLULAR_MATRIX 82 2.38E-45 1.1E-41
GO_CELL_MOTILITY 142 3.2E-45 1.1E-41
GO_CELL_CELL_SIGNALING 138 9.1E-43 2.34E-39
GO_COLLAGEN_CONTAINING_EXTRACELLULAR_MATRIX 67 1.76E-39 3.62E-36
GO_INTRINSIC_COMPONENT_OF_PLASMA_MEMBRANE 129 7.91E-39 1.35E-35
GO_POSITIVE_REGULATION_OF_MULTICELLULAR_ORGANISMAL_PROCESS 133 2.48E-38 3.64E-35
GO_REGULATION_OF_CELL_DIFFERENTIATION 136 5.3E-38 6.81E-35
GO_TUBE_DEVELOPMENT 102 5.21E-37 5.95E-34
GO_REGULATION_OF_CELLULAR_COMPONENT_MOVEMENT 101 1.79E-36 1.84E-33
SUM
Gene Set Name # Genes p-value FDR q-value
GO_BIOLOGICAL_ADHESION 64 2E-25 2.05E-21
GO_EXTRACELLULAR_MATRIX 40 1.42E-23 7.3E-20
GO_ION_TRANSPORT 60 1.5E-19 4.16E-16
GO_CELL_CELL_SIGNALING 62 1.62E-19 4.16E-16
GO_SIGNALING_RECEPTOR_BINDING 59 2.13E-19 4.37E-16
GO_POSITIVE_REGULATION_OF_SIGNALING 63 3.67E-19 6.29E-16
GO_REGULATION_OF_CELL_DIFFERENTIATION 63 1.29E-18 1.89E-15
GO_NEUROGENESIS 58 1.72E-18 2.2E-15
GO_POSITIVE_REGULATION_OF_MULTICELLULAR_ORGANISMAL_PROCESS 61 2.29E-18 2.61E-15
GO_COLLAGEN_CONTAINING_EXTRACELLULAR_MATRIX 30 6.62E-18 6.8E-15
 
Table 3.1. Top differentially regulated GO terms from RNA sequencing of 
phenotypically sorted MDA-MB-231, MCF10CA1a, and SUM159 highly migratory 
and weakly migratory subpopulations. 
 
 93 
 
 
Figure 3.9. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted MDA-MB-231 subpopulations. a) Heat map representing Z 
scores for EMT genes for MDA+ and MDA- subpopulations. 
 
Additionally, we performed RNAseq on the MCF10CA1a and SUM159 subpopulations 
and found the same trends with both subpopulations showing a high degree of 
differential regulation both overall and for EMT genes (Figures 3.10A,B,C, 3.11A,B,C). 
Both CA1a+ and SUM159+ showed significant upregulation of mesenchymal scores and 
downregulation of epithelial scores while both CA1a- and SUM159- subpopulations 
showed the opposite trends (Figures 3.10D,E, 3.11D,E). When GO term analysis on 
these subpopulations were performed, both showed biological adhesions as one of the 
 94 
 
most differentially regulated GO term categories (Table 3.1). Given the consistency of 
this finding across cell lines and the CTC clustering observed in MDA—injected mice, 
we hypothesized that cell-cell adhesion may be involved in the differential metastatic 
phenotype. To further characterize cell-cell adhesion behavior across all 
subpopulations, RNAseq data for genes encoding cell-cell adhesion proteins were 
compared (Figures 3.12A,B, 3.13A,B, 3.14A,B). Cell-cell adhesion scores were 
calculated and all weakly migratory subpopulations showed a significantly higher score 
than their respective highly migratory subpopulations, indicating greater upregulation 
of cell-cell adhesion genes (Figures 3.12C, 3.13C, 3.14C). Interestingly, E-cadherin, a 
cell-cell adhesion protein that has been recently implicated in CTC clustering and 
metastatic potential in other systems158,262, was the most highly differentially 
upregulated epithelial marker in MDA- cells compared to MDA+ cells. 
 95 
 
 
 
 
 
 96 
 
Figure 3.10. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted MCF10CA1a subpopulations. a) RNAseq z score heatmap 
showing differential gene regulation of CA1a+ and CA1a- subpopulations (n = 3). b) 
Top differentially regulated GO term categories for CA1a+ and CA1a- 
subpopulations. c) Log 2 fold change of CA1a-/CA1a+ epithelial-to-mesenchymal 
transition genes (n=3). d) Epithelial and e) mesenchymal scores for CA1a+ and CA1a- 
subpopulations (n = 3). f) GO categories found to be differentially regulated between 
CA1a+ and CA1a- subpopulations. Data in (c) display mean ± SEM. Data in (d) and 
(e) display mean ± SD. Statistical significance in (a) was calculated using multiple t-
tests. Statistical significance in (d) and (e) was calculated using unpaired, two-tailed 
Student t-tests. **** p < 0.0001. 
 
 97 
 
 
 
 
 
 
 
 98 
 
Figure 3.11. RNA sequencing reveals differential EMT gene regulation in 
phenotypically sorted SUM159 subpopulations. a) RNAseq z score heatmap 
showing differential gene regulation of SUM159+ and SUM159- subpopulations (n = 
3). b) Top differentially regulated GO term categories for SUM159+ and SUM159- 
subpopulations. c) Log 2 fold change of SUM159-/ SUM159+ epithelial-to-
mesenchymal transition genes (n=3). d) Epithelial and e) mesenchymal scores for 
SUM159+ and SUM159- subpopulations (n = 3). f) GO categories found to be 
differentially regulated between SUM159+ and SUM159- subpopulations. Data in (c) 
display mean ± SEM. Data in (d) and (e) display mean ± SD. Statistical significance 
in (a) was calculated using multiple t-tests. Statistical significance in (d) and (e) was 
calculated using unpaired, two-tailed Student t-tests. **** p < 0.0001. 
 
 99 
 
 
Figure 3.12. RNA sequencing reveals differential cell-cell adhesion gene 
regulation in phenotypically sorted MDA-MB-231 subpopulations. a) RNAseq 
z score heatmap showing differential gene regulation of MDA+ and MDA- 
subpopulations (n = 3). b) Log 2 fold change of MDA-/ MDA+ cell-cell adhesion 
genes (n=3).  
 
 
 100 
 
 
 
Figure 3.13. RNA sequencing reveals differential cell-cell adhesion gene 
regulation in phenotypically sorted MCF10CA1a subpopulations. a) RNAseq z 
score heatmap showing differential gene regulation of CA1a+ and CA1a- 
subpopulations (n = 3). b) Log 2 fold change of CA1a-/ CA1a+ cell-cell adhesion 
genes (n = 3).  
 101 
 
 
 
Figure 3.14. RNA sequencing reveals differential cell-cell adhesion gene 
regulation in phenotypically sorted SUM159 subpopulations. a) RNAseq z score 
heatmap showing differential gene regulation of SUM159 subpopulations (n = 3). b) 
Log 2 fold change of SUM159-/SUM159+ cell-cell adhesion genes (n = 3).  
 102 
 
E-cadherin expression in phenotypically sorted subpopulations is necessary for 
successful completion of the metastatic cascade 
Based on the formation of MDA- CTM in the circulation and our RNAseq data, we 
investigated the role of E-cadherin, a cell-cell adhesion protein implicated in CTM 
formation263–265. MDA- cells were found to possess significantly higher E-cadherin 
compared to MDA+  based on both qPCR (Figure 3.15A) and western blotting (Figures 
3.15B, 3.16A). Immunofluorescence staining for E-cadherin in MDA- cells indicated 
that it is localized to cell-cell junctions within cell clusters (Figure 3.15C). These assays 
confirm that MDA- cells express E-cadherin. 
 
To determine whether E-cadherin expression was involved in the increased metastatic 
potential of MDA- cells, an E-cadherin knockdown cell line was created by transducing 
MDA- cells with E-cadherin shRNA (Figure 3.16B). When E-cadherin knockdown 
MDA- (EcadKD) cells were injected orthotopically, BLI signal indicated slightly 
slowed growth of EcadKD primary tumors compared to MDA- scrambled shRNA 
control (Scrambled) primary tumors (Figure 3.15D). To compensate for this, EcadKD 
primary tumors were removed after they reached a comparable size to control tumors. 
After primary tumor removal, BLI showed reduced distal metastasis in EcadKD mice 
compared to control mice (Figure 3.15E). Anti-GFP IHC staining of lungs and liver 
confirmed a marked decrease in distal metastasis in EcadKD mice compared to 
scrambled control mice (Figure 3.15F). Additionally, macroscopically, there were 
significantly fewer liver nodules in EcadKD compared to controls (Figure 3.16C). When  
 103 
 
 
 104 
 
Figure 3.15. E-cadherin expression is necessary for metastasis in phenotypically 
sorted subpopulations. a) qPCR of E-cadherin in MDA-MB-231 subpopulations 
normalized to MDA+ cells (n = 3). b) Western blot of E-cadherin and GAPDH in 
subpopulations. c) Immunofluorescence staining of E-cadherin expression in MDA- 
cells. Scale bar: 25 µm. d) Primary tumor growth of MDA- E-cadherin knockdown 
and scrambled control tumors monitored via bioluminescence imaging (BLI) (n = 3). 
e) End point BLI of scrambled and E-cadherin knockdown mice at 4 weeks post tumor 
removal. f) Representative images of lung and liver histological sections stained with 
anti-GFP IHC. Quantification of percentage of GFP-positive cells in g) liver (n = 
10,12) and h) lung (n = 10,12) histological sections. i) Immunofluorescence staining 
of E-cadherin expression in MDA+ + E-cadherin cells. Scale bar: 50 µm. Inset scale 
bar: 20 µm. j) Primary tumor growth of MDA+ + E-cadherin tumors monitored via 
BLI (n = 6). k) End point BLI of MDA+ + E-cadherin mice at 4 weeks post tumor 
removal l) Representative image of lungs and livers histological sections stained with 
anti-GFP IHC. Quantification of percentage of GFP-positive cells in m) liver (n = 
6,9) and n) lung (n = 6,9) IHC-stained tissue sections. Data in (a), (g), (h), (m), and 
(n) display mean ± SD. Data in (d) and (j) display mean ± SEM. Statistical 
significance for (a), (g), and (h) was calculated using an unpaired, two-tailed 
Student’s t-test.  Statistical significance for (m) and (n) was calculated using a Mann-
Whitney test. *p < 0.05, **p < 0.01. 
 
quantified, a significantly greater percentage of GFP-positive cells were found in liver 
(Figure 3.15G) and lungs (Figure 3.15H) of scrambled control mice compared to 
EcadKD. These findings suggest that E-cadherin promotes distal metastasis of the 
MDA- subpopulation and may be required for metastasis of these cells.  
 
To determine whether E-cadherin expression can enable the MDA+ subpopulation to 
metastasize, we transduced MDA+ cells to express E-cadherin (Figure 3.15I, 3.16D). 
Primary tumor growth was monitored using BLI, and primary tumors were removed at 
4 weeks (Figure 3.15J). At the study endpoint, distal metastasis can be consistently 
observed in MDA+ + E-cadherin injected mice via BLI in locations indicative of lungs, 
liver, and bone (Figure 3.15K). Metastasis to lungs and liver was confirmed using anti- 
 105 
 
 
Figure 3.16. E-cadherin expression in MDA subpopulations and additional 
validation of E-cadherin knockdown and addition experiments. a) Normalized E-
cadherin expression Western blot quantification (n = 3). b) Western blot for E-
cadherin and GAPDH of MDA-  E-cadherin knockdown cells, scrambled control, and 
untransfected MDA- control. c) Representative images of livers at study endpoint and 
quantification of metastatic liver nodules. d) Western blot for E-cadherin and GAPDH 
in MDA+ + E-cadherin cells compared to both subpopulations. e) Representative 
images of livers at study endpoint and quantification of metastatic liver nodules. f) 
Average circulating tumor cell (CTC) count per mL blood across conditions. Note: 
MDA+ and MDA- data also graphed in Fig. 3.3n. g) Percentage of CTC clustering 
across conditions. Note: MDA+ and MDA- data also graphed in Fig. 2o. Data in (a), 
(c), and (e) display mean ± SEM. Data in (f) display mean ± SD. Statistical 
significance was determined using an unpaired, two-tailed student’s t-test for (a) and 
a Mann-Whitney test for (c) and (e).  * p < 0.05, **** p <0.0001. 
 106 
 
 
GFP IHC staining (Figure 3.15L). Additionally, quantification of metastatic liver 
nodules reveals a significantly higher nodule count for MDA+ + E-cadherin mice 
compared to MDA+ mice (Figure 3.16E). When quantified, the percentage of GFP-
positive cells was significantly higher in liver (Figure 3.15M) and lungs (Figure 3.15N) 
of MDA+ + E-cadherin mice compared to MDA+ mice. At 4 weeks post-injection, CTCs 
were collected from MDA-EcadKD- and MDA+ + E-cadherin-injected mice as before, 
and no significant difference in CTC concentration for these conditions compared to 
either MDA+- or MDA--injected mice was observed (Figure 3.16F). Further, CTC 
clusters were abrogated in MDA-EcadKD-injected mice while CTC clusters were 
present in MDA+ + E-cadherin-injected mice (Figure 3.16G). Together, these results 
establish that E-cadherin expression enables successful distal metastasis in both of our 
phenotypically sorted subpopulations and that E-cadherin plays a role in CTC 
clustering.  
 
E-cadherin expression tunes migration mode and extent in subpopulations 
Our data indicate that E-cadherin expression is related to the metastatic phenotype of 
highly migratory MDA+ and weakly migratory MDA- cell subpopulations. However, 
the molecular mechanism governing their difference in migration behavior in vitro is 
not clear. To determine whether E-cadherin plays a role in migration phenotype, we 
sought to further characterize these subpopulations and their variants to determine if 
modulation of E-cadherin expression would be sufficient to affect their distinct 
 107 
 
migratory phenotypes in vitro. 
 
First, we characterized the localization of E-cadherin across all MDA-MB-231 
conditions (Figure 3.17A). As seen in the representative immunofluorescence images, 
as expected, E-cadherin localizes to junctions as well as the cytoplasm in cells within 
MDA-, MDA+EcadLow (used for in vivo experiments described above), and 
MDA+EcadHigh (transduced with comparatively much higher E-cadherin expression) 
subpopulations while it is observed cytoplasmically in cells within the MDAPAR 
population and is absent from MDA+ and MDA-EcadKD cells. We revisited the 
transwell migration sorting platform to determine if modulation of E-cadherin 
expression affects either subpopulation’s ability to migrate through a collagen-coated 
transwell membrane. As expected, MDA+ cells migrated through the transwell 
significantly more than MDAPAR, which migrated significantly more than MDA- 
(Figure 3.18A). When MDA+ cells were induced to express a relatively modest amount 
of E-cadherin that was used previously in our in vivo experiments (MDA+EcadLow), the 
fraction of cells migrated significantly decreased, whereas when E-cadherin was 
knocked down in MDA- cells, the fraction of cells migrated significantly increased. 
However, even with a much higher level of E-cadherin expression, MDA+EcadHigh cells 
still migrated significantly more than MDA- cells. The migration fraction for 
MDA+EcadLow cells was significantly higher than MDA-EcadKD cells indicating that 
E-cadherin alone is not responsible for the differences in migration ability. We also 
compared the motile fraction of each condition seeded as single cells in 3D collagen 
 108 
 
matrix (Figure 3.18B). Here, we found similar trends where E-cadherin addition 
significantly attenuated the motile fractions of both MDA+EcadLow and MDA+EcadHigh 
conditions compared to MDA+, and knockdown significantly enhanced the motile 
fraction of MDA-EcadKD cells compared to MDA- cells. Here, even the much higher 
level of E-cadherin expression in MDA+EcadHigh was not sufficient to lower motile 
fraction to MDA- or MDA-EcadKD levels. These data show that E-cadherin affects the 
migration of the phenotypically sorted subpopulations but does not completely explain 
their distinct migratory abilities. 
 
Figure 3.17. Representative confocal fluorescence images show phenotypically 
sorted subpopulations and E-cadherin variants have differential E-cadherin 
expression. a) Representative images of MDA-MB-231 parental, MDA+, MDA-, 
MDA+EcadLow, MDA+EcadHigh,  and MDA-EcadKD cells seeded on collagen-coated 
glass. Scale bar: 20µm. 
 
Given that MDA- are relatively weak migrators, we sought to determine whether their 
migration defect was related to their ability to remodel matrix to facilitate movement. It 
 109 
 
has been previously shown that cancer cells can utilize channels found in native 
mammary stroma to migrate independent of proteolysis and matrix reorganization242. 
To test whether MDA- are capable of migrating in the more permissive environment of 
collagen channels, subpopulations were seeded in channels micromolded into collagen, 
and phase time-lapse microscopy was performed. MDA+ cells in channels displaced 
more readily than MDA- cells in channels, indicating that this permissive environment 
is not sufficient to improve MDA- migration ability to levels comparable to MDA+. Cell 
velocities in the channels were quantified and MDA+ cells migrated significantly faster 
than MDAPAR cells, which migrated significantly faster than MDA- cells (Figure 3.18C). 
In the channels, MDA+EcadLow and MDA+EcadHigh cell velocities were not significantly 
different from that of MDA-EcadKD cells. This suggests that while there are still 
additional factors responsible for part of the differential migration modes of these 
subpopulations, decreased ability to remodel matrix could also contribute to a portion 
of the residual differences that remain between the subpopulations after E-cadherin is 
altered.  
 
While single cell migration models are often used to study cancer cell migration, 
collective migration is thought to be the more prevalent mode of migration in vivo143. 
To assess the migration ability of these subpopulations and their E-cadherin variants in 
a collective context, in vitro tumor spheroids were formed. When embedded in collagen 
and monitored for 48 h, all conditions resulted in spheroid outgrowth into the 
surrounding matrix (Figure 3.18D). MDA- spheroids outgrew significantly less than 
 110 
 
MDAPAR, which outgrew significantly less than MDA+ spheroids (Figure 3.18E). 
Consistent with single cell measurements, outgrowth area of MDA+EcadLow and 
MDA+EcadHigh spheroids were significantly reduced compared to MDA+ spheroids, and 
outgrowth area significantly increased in MDA-EcadKD spheroids compared to MDA- 
control spheroids. Interestingly, MDA+ spheroids migrated out largely as single cells 
while MDA- spheroids migrated as collective strands while MDAPAR spheroids 
presented as a mixture of the two modes. This suggests that while MDA- cells do not 
migrate well as single cells, they are able to migrate collectively albeit less effectively 
than MDA+ cells.  Additionally, while MDA+EcadLow spheroids exhibit reduced 
outgrowth but no noticeable change in migration mode compared to MDA+ spheroids, 
MDA+EcadHigh spheroids shift from single to collective migration modes and obtain a 
reduced outgrowth area similar to that of MDA- spheroids. Conversely, MDA-EcadKD 
spheroids retain the collective mode of migration while exhibiting a higher outgrowth 
area compared to MDA- spheroids. Together, this suggests that while E-cadherin 
expression can modulate cell migration in both single and collective contexts, there are 
still other inherent differences between these subpopulations that affect their 
determining migration ability. 
 
 111 
 
 
 
 
 
 
 
 112 
 
Figure 3.18. E-cadherin expression tunes migration ability and mode in 
phenotypically sorted subpopulations. a) Fraction of cells migrated for MDA 
subpopulations with and without E-cadherin in transwell assays (n = 3). b) Motile 
fraction of MDA subpopulations with and without E-cadherin in 1.5mg/mL 3D 
collagen (n = 10-35). c) Microtrack migration speeds for MDA-MB-231 
subpopulations with and without E-cadherin (n = 30-36). d) Representative images of 
tumor spheroid outgrowth at 48 h post-embedding. Scale bar: 100 µm e) Outgrowth 
area to quantify cell migration from in vitro tumor spheroids made from MDA-MB-
23 subpopulations with and without E-cadherin embedded in 1.5 mg/mL 3D collagen 
matrix at 48 h. Data from (a) and (b) display mean ± SEM. The box and whisker plot 
in (c) and (e) show medians, 25th/75th percentiles, and minimum and maximum 
values. Statistical significance for (a) and (b) was calculated using an ordinary, one-
way ANOVA. Statistical significance for (c) and (e) was calculated using a Kruskal 
Wallis H test. Statistical significance for (f) was calculated using an unpaired, two-
tail Student’s t-test. * p < 0.05, ** p < 0.01,  *** p < 0.001, **** p < 0.0001, n.s., 
non-significant. 
 
Phenotypically sorted subpopulations display different morphologies and contractility 
After further characterizing the migration abilities of phenotypically sorted weakly and 
highly migratory cells and their E-cadherin variants, we sought to examine cytoskeletal 
architecture, cell-ECM signaling, and mechanotransduction, since these all can impact 
migration ability. To assess the morphologies of the phenotypically sorted cells with 
and without E-cadherin, we imaged them using phase-contrast microscopy (Figure 
3.19A). The MDA+ and MDA- display distinct phenotypes which are present in the 
heterogeneous parental cells. When comparing the morphologies of MDA+, 
MDA+EcadLow, and MDA+EcadHigh, more cell spreading and cell-cell contacts occur as 
E-cadherin expression increases. Conversely, when comparing the MDA- and MDA-
EcadKD cell morphologies, there is a slight reduction in cell-cell contacts and cell 
spreading. When cell area is quantified, MDA- cells have significantly greater cell area 
compared to MDAPAR cells which have greater cell area than  MDA+ cells (Figure 
 113 
 
3.19B). Interestingly, both the addition or knockdown of E-cadherin to MDA+ and 
MDA- cells, respectively, result in increased cell area compared to the baseline 
phenotypes.  
 
Since migration behavior is dependent on cytoskeletal architecture245, actin staining was 
performed on both subpopulations. MDA- cells exhibit robust stress fibers while MDA+ 
cells display both stress fibers as well as ruffling lamellipodia (Figure 3.20A). Again, 
MDAPAR cells display attributes of both subpopulations and E-cadherin addition or 
knockdown can attenuate these phenotypes in each respective subpopulation. 
Lamellipodia formation has been associated with both migratory and metastatic 
phenotypes266. Additionally, stress fibers in non-motile cells are expected to be more 
robust compared to stress fibers in migratory cells267. Thus, the prevalence of 
lamellipodia in MDA+ and stress fibers in MDA- are consistent with their migratory 
phenotypes. Together, these data show that the subpopulations display distinct 
morphologies that are present in the heterogeneous parental population and can be 
altered by changing E-cadherin expression. 
Since the actin cytoskeleton and cell-matrix adhesions are interdependent268, we 
examined adhesive structures and signaling across all cell conditions. Since focal 
adhesion kinase (FAK) has been correlated with an invasive phenotype and 
metastasis269, we investigated whether subpopulations had differential FAK expression 
and activation. While total FAK levels were similar between subpopulations, MDA- 
cells had significantly higher FAK activation compared to MDA+ as indicated by  
 114 
 
 
 
 115 
 
Figure 3.19. Subpopulations exhibit differential morphologies, cell-ECM 
signaling, and contractility. a) Representative two-dimensional morphologies with 
red arrows pointing to parental cells with weakly motile-like morphologies and the 
green arrows pointing to parental cells with strongly motile-like morphologies; Scale 
bar: 50 µm. b) Quantification of cell area for subpopulations with and without E-
cadherin. c) Western blot of pFAK, FAK, and GAPDH for MDAPAR, MDA+, and 
MDA-. d) Western blot of pFAK, FAK, and GAPDH for MDA- control and MDA-
EcadKD. e) Western blot of pFAK, FAK, and GAPDH for MDA+, MDA+EcadLow, 
and MDA+EcadHigh. f) Quantification of FAK expression and activation from Western 
blot in (c) (n = 3). g) Quantification of FAK expression and activation from Western 
blot in (d) (n = 3). h) Quantification of FAK expression and activation from Western 
blot in (e) (n = 3). i) Quantification of focal adhesion area for subpopulations with 
and without E-cadherin. j) Total traction force magnitude, |F|, of MDA-MB-231 
subpopulations with and without E-cadherin (n = 51-129).  k) Percentage of bulk 
collagen matrix contraction  for collagen gels seeded with MDA-MB-231 
subpopulations with and without E-cadherin after 4 days (n = 4-14). Data from (f), 
(g), (h) and (k) display mean ± SEM. The box and whisker plots in (b), (i), and (j) 
shows medians, 25th/75th and minimum and maximum values. Statistical significance 
for (b), (i), and (j) were calculated using a Kruskal-Wallis H test. Statistical 
significance for (f), (h), and (k) were calculated using an ordinary, one-way ANOVA. 
Statistical significance for (g) was calculated using an unpaired, two-tailed Student’s 
t-test. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, n.s. non-significant.  
 
pFAK/FAK levels (Figure 3.19C,F). Western blotting shows that loss of E-cadherin in 
MDA- reduces FAK activation (Figs. 6d,g) while addition of E-cadherin to MDA+ 
increases FAK activation (Figure 3.19E,H). To further characterize focal adhesions, we 
performed staining for phosphorylated FAK, which revealed small, nascent focal 
adhesions localized at the lamellipodium in MDA+ cells and large, elongated focal 
adhesions capping stress fibers in MDA- (Figure 3.20A). When quantified, MDA- cells 
had significantly larger focal adhesions than MDAPAR, which had larger focal adhesions 
than MDA+ (Figure 3.19I). When E-cadherin was knocked out in MDA-, focal adhesion 
area decreased while when E-cadherin was added to MDA+ cells, it was increased. 
Together, these results suggest that MDA+ and MDA- subpopulations have differential 
 116 
 
FAK signaling which is mediated in part by E-cadherin. 
 
Given the robust differences in morphologies and cytoskeletal architecture between the 
subpopulations, traction force microscopy (TFM) was performed to quantify contractile 
forces generated by MDA+ and MDA- cells with and without E-cadherin. Prior work 
indicates traction stresses are increased in malignant and more metastatic cells31,32. 
While both subpopulations were not significantly different from MDAPAR, traction 
forces of MDA+ were significantly increased compared to MDA- (Figure 3.19J). To 
determine whether this trend persists in 3D collagen matrix, a collagen contraction assay 
was performed. MDA+  seeded collagen gels contracted significantly more than  
 
Figure 3.20. Subpopulations possess differential actin cytoskeletal structure and 
focal adhesion activation. a) Representative images of MDA-MB-231 parental, 
MDA+, MDA-, MDA+EcadLow, MDA+EcadHigh,  and MDA-EcadKD cells stained for 
actin and pFAKY397. Scale bar: 10µm. Inset scale bar: 5µm. 
 117 
 
 
MDAPAR seeded collagen gels which contracted more than MDA- seeded collagen gels 
(Figure 3.19K). These data indicate that the highly motile, weakly metastatic 
subpopulation is more contractile than the weakly motile, highly metastatic 
subpopulation, which is counter to the hypothesis that more malignant and metastatic 
cells exert higher forces31,32. While MDA+EcadHigh exhibited relatively weak traction 
stresses in the TFM assay, this condition showed compaction comparable to those of 
MDA+ gels in the collagen gel contraction assay (Figure 3.19J,K). In addition, in vitro 
tumor spheroids were imaged at 24 h post-embedding using quantitative polarization 
microscopy to assess cell contractility in the context of multicellular cancer cell strand 
migration (Figure 3.21A). Here, optical anisotropy (which represents the contractility 
of cells within the migration strands) of each condition shows comparable trends to the 
TFM and gel contraction assay data. The optical anisotropy signal is highest in MDA+ 
migrating cells compared to MDAPAR cells which show higher optical anisotropy signal 
than cells in MDA- migration strands. Additionally, MDA+EcadLow migrating cells have 
lower optical anisotropy signal compared to MDA+ cells but greater signal than cells in 
MDA+EcadHigh migration strands. MDA-EcadKD migration strands showed higher 
optical anisotropy signal compared to MDA- migration strands. Together, this data 
demonstrates that phenotypically sorted subpopulations possess unique cytoskeletal 
features, cell-ECM signaling, and contractility behavior that are distinct even when 
modulated by E-cadherin expression. 
 118 
 
 
Figure 3.21. Quantitative polarization microscopy shows differential cell 
contractility in migration strands of in vitro tumor spheroid model. a) 
Representative images of MDA-MB-231 parental, MDA+, MDA-, MDA+EcadLow, 
MDA+EcadHigh,  and MDA-EcadKD tumor spheroid invasion strands at 24 h. Scale 
bar: 50µm. 
 
Clustering of circulating tumor cells in patient blood correlates with prognosis 
Based on our data, the MDA- subpopulation has a higher metastatic fitness that is 
associated with E-cadherin expression and clustering of CTCs. To determine if CTC 
clustering and E-cadherin expression are clinically relevant markers of metastatic 
 119 
 
potential, blood samples from patients with metastatic cancer were collected and 
processed (Table 3.2). After staining for cytokeratin, single CTCs and CTC clusters 
were quantified (Figure 3.22A). Our data indicate there is a significant trend between 
the concentration of CTC clusters per mL blood and patient survival,  where patients 
with survival time of less than 12 months had significantly increased CTC cluster 
concentrations compared to patients with 12-24 month survival and those alive with 
progression (Figure 3.22B). In a separate cohort of metastatic breast cancer patient 
samples (Table 3.3), when stained for E-cadherin, both single cell and clustered CTCs 
in metastatic breast cancer patient blood were positive for E-cadherin in all patient 
samples tested (Figure 3.22C). Additionally, we used a Kaplan-Meier Plotter to 
determine the relationship between E-cadherin and distal metastasis-free survival 
(DMFS) and overall survival (OS) in breast cancer patients254. The upper tercile of 
patients based on E-cadherin gene expression showed a significant reduction in both 
DMFS (Figure 3.22D) and OS (Figure 3.22E) compared to the lower tercile of patients. 
These results suggest E-cadherin expression and CTC clustering have prognostic value 
in breast cancer.  
 
 120 
 
 
Figure 3.22. Circulating tumor cell clusters trend with worsened patient 
outcome. a) Representative image of circulating tumor cells (CTCs) from human 
cancer patient blood stained for cytokeratin (green) and DAPI (blue); filled arrows 
denote CTC clusters while empty arrows denote single CTCs. Scale bar: 50 µm. b) 
Quantification of CTC cluster concentration from blood isolated from 11 metastatic 
cancer patients binned by patient survival time. c) Representative image of CTCs 
isolated from a separate cohort of metastatic breast cancer patient blood stained for 
E-cadherin (red) staining with cytokeratin (green), CD45 (magenta), and DAPI (blue) 
staining. Scale bar: 20 µm. Kaplan Meier curves displaying d) distal metastasis free 
survival and e) overall survival for breast cancer patients divided into top and bottom 
terciles based on E-cadherin gene expression. Data in (b) display mean ± SD. 
 
 
 
 
 121 
 
Table 3.2. Patient demographic and clinical outcomes for metastatic cancer patient 
blood samples used in clustering analysis. 
 
Gender Age Cancer origin Outcome  
Male 41 Bile duct Deceased after 18 months 
Male 69 Prostate Alive with disease 
Female 46 Breast Alive with disease 
Male 67 Kidney Deceased after 20 months 
Male 37 Skin Alive with disease 
Male 63 Prostate Deceased after 8 months 
Female 55 Breast Deceased after 18 months 
Female 86 Pancreas Deceased after 12 months 
Female 80 Pancreas Deceased after 3 months 
Female 51 Breast Alive with disease 
Female 75 Lung Deceased after 5 months 
Male 69 Prostate Deceased after 3 months 
 
 
 
 
Table 3.3. Patient demographic and clinical details for metastatic breast cancer patient 
blood samples used in E-cadherin staining experiment. 
 
Total, N 8 
Median age (range) 60 (49-73) 
Gender (%)  
     Female 8 (100%) 
     Male 0 (0%) 
Cancer subtype, N (%)  
     ER+, PR+, Her2- 5 (62.5%) 
     ER+, PR+, Her2+ 1 (12.5%) 
     ER+, PR-, Her2- 1 (12.5%) 
     ER+, PR-, Her2+ 1 (12.5%) 
Location of Known Metastases, N (%)  
     Bone 6 (75%) 
     Lung 2 (25%) 
     Peritoneum 1 (12.5%) 
     Brain 1 (12.5%) 
 
 
 
 
 122 
 
3.5 Discussion 
Here, we used the approach of sorting cells based on migratory phenotype to identify 
the phenotype associated with the most aggressively metastatic cells. MDA+ and MDA- 
subpopulations were sorted based on their ability to migrate through a transwell 
migration assay. Notably, MDA- cells were highly metastatic in vivo compared to 
MDA+ cells when injected orthotopically into mice. RNA sequencing revealed robust 
differences in gene regulation across many aspects of cellular function including cell 
adhesion and motility, which are associated with epithelial-to-mesenchymal transition 
(EMT). MDA- cells showed increased epithelial gene expression compared to MDA+ 
cells, which showed increased mesenchymal gene expression. The increased metastatic 
potential of MDA- was found to be dependent on E-cadherin, which enabled CTC 
clusters to form and disseminate in the circulation leading to enhanced colonization of 
distant tissues. Further, we established that induced expression of E-cadherin was 
sufficient to rescue MDA+ cells and allow them to metastasize to the same tissues as 
MDA-. Notably, in cancer patients, the number of CTC clusters in blood trended with 
worsened patient outcomes, CTCs from metastatic breast cancer patients were found to 
be positive for E-cadherin, and E-cadherin was associated with decreased distal 
metastasis free survival and overall survival. Together, these data indicate that migration 
ability in vitro is not positively correlated with metastasis, and E-cadherin plays a 
significant role in determining metastasis. 
Perhaps one of the most surprising results from this study is that migration 
ability does not correlate with metastasis. Our results generate numerous questions 
 123 
 
regarding intratumor heterogeneity in the context of migratory and metastatic 
phenotypes. Importantly, how much value should be placed on single cell migration 
studies and the most migratory cells in a population in cancer research? Even in highly 
permissive collagen microtracks, MDA- cells migrated poorly until placed in the 
collective context of in vitro tumor spheroids. Given that our MDA- subpopulation 
migrates poorly as single cells compared to in spheroids when they are permitted to 
migrate collectively, many researchers focused on single cell assays might inadvertently 
dismiss these cells, as motility is often correlated with cancer aggressiveness. Thus, 
these results highlight the need for physiologically relevant multicellular migration 
assays that can allow more complex cell behaviors, such as collective migration to be 
observed. Even with that consideration, our data indicate that migration ability, even 
when in a collective context, is not predictive of metastasis. 
Based on our findings, E-cadherin, an epithelial protein that is generally 
considered a tumor suppressor, enables breast cancer metastasis. Epithelial-to-
mesenchymal transition (EMT) has traditionally been considered necessary for 
metastasis, with cancer aggressiveness and metastatic potential correlating with the 
extent of EMT270,271. Increasing evidence calls into question the necessity of 
traditionally defined EMT and places increasing importance on cells that possess the 
capacity to exhibit epithelial properties required for metastatic colonization55,156,271. E-
cadherin along with other epithelial markers including keratin-14 have been shown to 
facilitate metastasis156,272. The enabling role of E-cadherin for metastasis and reciprocal 
effects on migration were recently delineated in several breast cancer models158,262. Our 
data may suggest that E-cadherin expression in the MDA- subpopulation endows these 
 124 
 
cells with a hybrid epithelial/mesenchymal (E/M) phenotype, which has been suggested 
to enhance metastatic potential273–276. The MDA+ subpopulation may represent an 
extreme mesenchymal phenotype, which lacks the epithelial traits necessary for 
metastatic colonization271,273,277,278. Probing the effects on both migration and metastasis 
of other epithelial and mesenchymal genes that we assessed using RNA sequencing 
could help determine which are associated with hybrid or extreme EMT states. As 
intermediate EMT states are being characterized, further work should be done towards 
probing heterogeneity of the EMT spectrum within primary tumors and metastatic sites. 
We have shown that MDA- are capable of migrating collectively in an in vitro 
tumor spheroid model but do not know what specific cues and machinery enable this 
migration. In Drosophila, E-cadherin has been shown to play a critical role as a 
mechanical signal integrator where it was necessary for directed migration of cellular 
clusters279. Using MDCK cells, it was recently shown that FAK-Src signaling is 
involved in relaxation of E-cadherin tension which facilitates migration via β-catenin 
signaling280. Given our findings that MDA- cells exhibited higher FAK activation 
compared to MDA+ cells, further work should be done to determine if FAK signaling 
plays a role in the collective migration abilities of MDA-. E-cadherin loss in MDA- cells 
increased migration, consistent with findings of others158,281, but was not sufficient to 
completely alter migration mode, indicating that there are other factors, most likely 
other epithelial cell-cell adhesion proteins, contributing to this behavior.  
One caveat of this study is that the implications of our phenotypic cell sorting 
process on the resulting subpopulations remain incompletely understood. One limitation 
 125 
 
of the transwell sorting platform is that it does not lend itself to real-time observation, 
which could provide more insight into the migration modes of the subpopulations when 
migrating through the transwell membrane. While none of the behaviors observed in 
either subpopulations appear to reside outside of the range of MDAPAR behaviors, it 
remains possible that the 20 rounds of sorting could condition or evolve the cells in a 
way that we have not yet identified. While this assay’s greatest perceived strength is 
that instead of selecting for protein expression or another well-defined metric, we are 
selecting for the multi-faceted behavior of cell migration, it is important to note that this 
limits our understanding of how these subpopulations relate to fully heterogeneous, 
unsorted parental cells. Further efforts to elucidate how representative these 
subpopulations are in the context of their parental cell lines should be considered as well 
as continued interrogation of the MCF10CA1a and SUM159 subpopulations as the 
notable transcriptional differences observed in the subpopulations across all three triple 
negative breast cancer cell lines tested may contribute novel insight into mechanisms of 
migration and metastasis. Additionally, comparison with subpopulations obtained from 
phenotypic sorting methods based on other cell behaviors such as adhesion and 
metastatic site preference could provide a more complete understanding of intratumor 
heterogeneity in the context of metastasis61,62,76.  
Our data supports the critical role of hybrid E/M phenotypes and the concept 
that certain subpopulations of cells, in this case MDA+ cells, can be too mesenchymal, 
lacking epithelial features required to colonize metastatic sites. We have demonstrated 
that E-cadherin facilitates cancer metastasis in weakly motile cells and can ameliorate 
the metastatic potential of highly migratory cells. Since colonization is the step in 
 126 
 
metastasis in which cancer gains its lethality, future therapeutic strategies could 
potentially take advantage of these findings by forcing metastatic cancer cells towards 
an extreme mesenchymal phenotype to prevent colonization and disease progression.   
 
 127 
  
 
CHAPTER 4 
 
PHENOTYPICALLY SORTED MIGRATORY BREAST CANCER 
SUBPOPULATIONS EXHIBIT MIGRATORY AND METASTATIC 
COMMENSALISM 
 
Portions of this data were contributed by Isaac Richardson. 
 
4.1 Abstract 
Intratumor heterogeneity is a well-established hallmark of cancer that impedes cancer 
research, diagnosis, and treatment. To assess the relationship between cancer cell 
migration and metastatic fitness, we phenotypically sorted highly migratory and weakly 
migratory MDA-MB-231 breast cancer cell subpopulations. Previously, we showed that 
counter to many prominent theories, the weakly migratory subpopulation metastasized 
far more than the highly migratory subpopulation in vivo. Our prior work characterized 
these subpopulations separately. However, they were originally isolated from a 
heterogeneous parental population so here, we compare the separate and combined 
behaviors of highly and weakly migratory cancer subpopulations in vitro and in vivo. 
Using an in vitro tumor spheroid model, we found that the weakly migratory 
subpopulation migrated a shorter distance collectively compared to the highly migratory 
subpopulation, which migrated further distances as both single cells and migration 
strands. When co-seeded with migratory cells, the non-migratory cells were able to 
migrate a further distance compared to cells in non-migratory only spheroids. In mixed 
spheroids, leader-follower behavior was observed with highly migratory cells leading 
 128 
 
the weakly migratory cells. Finally, we injected the highly migratory, weakly migratory, 
or a 1:1 mixture of both subpopulations orthotopically into mice and allowed tumors to 
grow. After primary tumor removal, significantly more distal metastasis was observed 
in mice injected with the 1:1 mixture compared to non-migratory only injected-mice, 
which also showed significant distal metastasis compared to migratory-injected mice. 
Together, our findings show that clonal cooperation is an important feature of 
intratumor heterogeneity and suggest that leader-follower behavior plays an important 
role in enhancing overall metastatic fitness. 
 
4.2 Introduction 
Metastasis, the spread of tumor cells from their tissue of origin to distant sites, is 
responsible for the vast majority of cancer-related deaths worldwide282. While 
metastasis is a dynamic, multistep process, many studies have focused on the initial 
steps of local dissemination where cancer cells adopt a motile phenotype, leave the 
primary tumor, and migrate through the stroma1. Despite the fact that in numerous 
cancer types, collective cell migration is the predominant migration mode observed in 
clinical samples and is associated with worsened patient prognosis, far less is known 
about this mode of migration as compared to single cell migration143,241,283. This may be 
partly due to the multicellular nature of collective migration, which imparts an inherent 
complexity that can impede both the design and interpretation of mechanistic studies. 
In collective migration, the external chemical and physical cues and intracellular 
signaling and mechanotransduction events that dictate single cell migration are 
integrated across cohesive sheets, strands, or streams of coordinated migrating 
 129 
 
cells145,284. While both single and collective cancer cell migration have been 
simultaneously observed in the same patients samples285, it remains challenging to parse 
apart the relative contributions of each cell’s spatiotemporally unique interactions with 
the microenvironment and their intrinsic genetic disposition to determine whether this 
observed spectrum of migration modes reflects cellular plasticity or phenotypic 
diversity. 
 
Intratumor heterogeneity is known to complicate cancer diagnosis and treatment and 
contribute to recurrence28. While the clinical impacts of intratumor heterogeneity are 
recognized, less is understood about how intratumor heterogeneity affects phenotypic 
behaviors such as migration and metastasis. During early tumor progression, negative 
clonal interactions such as competition are generally focused on as subclones acquire 
mutations and compete for resources following classic Darwinian evolutionary 
principles41,286. However, in later stages of tumor development, genetic instability can 
increase as mutations accumulate, causing the scales to tip from Darwinian selection 
towards genetic diversification41,282. The emergence of diverse cancer subclones allow 
for functional specialization and cooperative interactions including commensalism, 
synergism, and mutualism that can enhance tumor growth, metastatic potential, and 
chemoresistance41,282. In the context of migration, our group and others have studied 
leader-follower behavior, a pattern of collective migration where highly motile cells 
enable or guide directed migration of less motile cells56,66,146,147,287,288. Importantly, this 
form of collective migration can facilitate the metastasis of other less motile clonal 
subpopulations, introducing heterogeneity into metastatic sites and potentially 
 130 
 
imparting advantages such as enhanced survival and chemoresistance55,288–290. As 
leader-follower behavior can facilitate these late-stage effects which directly contribute 
to worsened patient outcomes, it is imperative that this cooperative clonal interaction be 
further studied.  
 
In this study, we use highly migratory (MDA+) and weakly migratory (MDA-) human 
breast cancer cells that have previously been phenotypically sorted based on their ability 
to move through a transwell migration assay. Using an in vitro tumor spheroid model, 
we found that MDA+ spheroids migrated out a farther distance as both single cells and 
in migration strands compared to MDA- spheroids, which migrated a shorter distance 
and in a predominantly collective fashion. When combined into 1:1 MDA+:MDA- 
mixed spheroids, leader-follower behavior is observed with MDA+ cells guiding MDA- 
cells resulting in increased migration distance for MDA- cells compared to MDA- only 
spheroids. Finally, we used an in vivo orthotopic metastasis model to show that MDA- 
cells metastasize more so than MDA+ cells and that MDA- metastasis is increased when 
co-injected with MDA+ cells. These data suggest that migratory and non-migratory 
cancer cell subpopulations cooperate in a commensal fashion to enhance metastatic 
fitness. 
 
4.3 Materials and Methods 
Cell Culture and Plasmids 
MDA-MB-231 breast adenocarcinoma cells (HTB-26, ATCC, Rockville, MD) were 
maintained in DMEM with high glucose (25 mM; Life Technologies, Grand Island, NY) 
 131 
 
supplemented with 10% fetal bovine serum (FBS; Atlanta Biologicals, Flowery Branch, 
GA), 100 µg mL-1 streptomycin (Life Technologies), and 100 U mL-1 penicillin (Life 
Technologies). All cell culture and time-lapse imaging was performed in a humidified 
environment at 37°C and 5% CO2. FUW-GFP-E2A-fluc and FUW-mCherry-E2A-rluc 
plasmids were created in-house, and all subpopulations were stably transduced prior to 
in vivo studies. Lentiviral particles were prepared and cells were transduced as 
described previously291. 
 
Phenotypic cell sorting  
To purify subpopulations based on migration ability, parental MDA-MB-231 cells 
(MDAPAR) were seeded in a transwell migration assay as described previously (Hapach 
et al., submitted to Cancer Res, in revision). Briefly, a coating of 1 mg mL-1 collagen 
gel (~10 µm thickness) was polymerized in a 6-well plate transwell insert with 8-µm 
pores (Corning) for 20 min. The coated insert was then hydrated in serum-free DMEM. 
Cells were seeded in the collagen-coated transwell insert at 40,000 cells cm-2 in low-
serum DMEM (DMEM + 0.5% FBS), and the insert was placed in a 6 well plate 
containing complete DMEM. On day 2, the low-serum DMEM was refreshed. On day 
4, highly migratory (MDA+) and weakly migratory (MDA-) cells were collected 
separately with 0.25% Trypsin-EDTA from the bottom and top compartments, 
respectively. Twenty additional rounds of sorting were performed to further purify 
subpopulations. Subpopulations were used in experiments for up to 20 passages 
following purification without discernible changes in behavior.  
 
 132 
 
Spheroid preparation and embedding 
MDA-MB-231 spheroids were generated as previously described147. Briefly, cells at 
approximately 80% confluency were harvested and resuspended in spheroid compaction 
medium containing 0.25% methylcellulose (H4100; Stem Cell Technologies, 
Cambridge, MA), 4.5% horse serum (Life Technologies), 18 ng mL-1 hEGF (Life 
Technologies), 0.45 µg mL-1 hydrocortisone (Sigma-Aldrich, St. Louis, MO), 9 µg mL-
1 insulin (Sigma-Aldrich), 90 ng mL-1 cholera toxin (Sigma-Aldrich), 90 U mL-1 
penicillin and 90 µg mL-1 streptomycin in DMEM/F12 (Life Technologies). The cell 
suspension was seeded into a 96-well round-bottom microplate with 5,000 cells in each 
well, which was then centrifuged at 300 × g for 5 min at room temperature.  
 
After 3 days of compaction, spheroids were embedded in 1.5 or 4.5 mg mL-1 type I 
collagen gels. Collagen gels were prepared as previously described146. Briefly, type I 
collagen was acid-extracted from rat tail tendons (BioIVT), purified via centrifugation 
and lyophilization, and reconstituted at 10 mg mL-1 in 0.1 % acetic acid. Stock collagen 
solution was diluted to either 1.5 or 4.5 mg mL-1 by gently mixing with ice-cold DMEM, 
and the solution was neutralized to pH 7.0 with 1 N NaOH. Spheroids were removed 
from culture plates and individually embedded within 500 µL collagen gels in glass-
bottom 24-well plates (MatTek, Ashland, MA). After 45 min of gel polymerization at 
37°C, gels were overlaid with 500 µL of complete DMEM complete media with or 
without inhibitors. 
 
Microscopy 
 133 
 
Static or time-lapse imaging were carried out with a Zeiss LSM800 confocal 
microscope, equipped with an environment control chamber. A 10X dry lens N.A. = 0.3 
was used to image embedded tumor spheroids with 20-µm-interval Z-stacks. A 40X 
water-immersion lens N.A. = 1.1 was used to image invasion strands with surrounding 
collagen architecture using confocal reflectance microscopy. 
 
Mousework 
Experiments were performed in accordance with AAALAC guidelines and were 
approved by the Vanderbilt University Institutional Animal Care and Use Committee 
(Protocol #: M1700029-01). 6-8 week old female NOD/SCID immunodeficient mice 
(The Jackson Laboratory) were injected with 1x106 MDA+, MDA-, or a 1:1 mix of 
MDA+:MDA- cells subcutaneously at the mammary gland. For MDA+ and MDA- only 
conditions, cells were tagged with a GFP plasmid. For the 1:1 mixed condition, half of 
the mice were injected with MDA+ + GFP and MDA- + mCherry and the other half were 
injected with MDA+ + mCherry and MDA- + GFP to control for any potential plasmid-
dependent differences. At 4 weeks or when primary tumors approached 100-200 mm3 
in volume, primary tumor removal surgery was performed following sterile surgical 
techniques. 4 weeks after tumor removal, mice were euthanized and tissue samples were 
collected and fixed using 4% paraformaldehyde or snapfrozen before processing for 
histological analysis.  
 
Data Quantification and Statistics 
All statistical analysis for in vitro and in vivo studies was performed using GraphPad 
 134 
 
Prism Software (v. 7.0a). Unpaired, two-tailed Student’s t-tests, Mann-Whitney tests, 
ordinary, one-way ANOVAs, and Kruskal-Wallis H tests were performed as appropriate 
with p < 0.05 as the cutoff for statistical significance. All data are shown as mean ± SD 
or box-and-whisker plots, where boxes represent medians and and bars indicate 10th/90th 
percentiles with outliers represented as dots. 
 
4.4 Results 
Previously, we have shown that phenotypic cell sorting of the human breast cancer cell 
line, MDA-MB-231 (MDAPAR), based on migration ability results in stable, distinct 
highly migratory (MDA+) and weakly migratory (MDA-) subpopulations (Hapach et al., 
submitted to Cancer Res). Sorting is performed by seeding cells in low-serum media on 
a collagen-coated transwell insert (Figure 4.1A). A serum gradient is established by 
filling the bottom reservoir of the transwell chamber with complete media. Over the 
course of 4 days, cells are able to migrate through the transwell into the bottom chamber. 
After 4 days, cells on the top and bottom of the transwell were separately collected and 
seeded into fresh transwells. This process was repeated 20 times to purify migratory and 
non-migratory subpopulations. Interestingly, when a 1:1 mixture of MDA+:MDA- cells 
(MDAMIX) were co-seeded into the same transwells, migration of MDA- cells in 
MDAMIX transwells was increased compared to MDA- only transwells while MDA+ cell 
migration in MDAMIX transwells was unaffected compared to MDA+ only transwells 
(Fig 1B). This suggested that cooperative interactions where MDA+ cells enhance 
MDA- cell migration may be occurring in MDAMIX transwells. 
 135 
 
 
Figure 4.1. Phenotypic sorted cancer cells show differential invasion. A) 
Schematic of transwell sorting assay that was performed on parental MDA-MB-231 
(MDAPAR) cells to create highly migratory MDA+ (green) and weakly migratory 
MDA- (red) subpopulations. B) Fraction of cells migrating through the transwell 
assay post-sorting. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p <0.0001 
To investigate the potential for cooperative interactions between the motile and non-
motile subpopulations, an in vitro tumor spheroid model was utilized. MDA+, MDA-, 
and MDAMIX spheroids were all formed over 3 days after seeding in a round-bottom 96-
well plate (Figure 4.2A). However, spheroids compacted to different extents with 
MDA+ being more compacted than MDAMIX which are more compacted than MDA- 
spheroids based on cross sectional area (Figure 4.2B). Additionally, when observing 
MDAMIX tumor spheroids at 72 h, MDA+ cells appear to preferentially localized to the 
 136 
 
exterior of the spheroid while the MDA- cells localize more towards the spheroid core 
(Figure 4.2C). When the relative intensity of the GFP- and mCherry-tagged cells across 
the diameter of the spheroid were compared at 0 (Figure 4.2D) and 72 h (Figure 4.2E), 
it is appreciable that there is initially a random distribution of cells immediately after 
seeding at 0 h and then after compaction at 72 h, there is an intensity shift towards the 
spheroid exterior for MDA+ cells and towards the spheroid interior for MDA- cells. 
 
After characterizing tumor spheroid formation, we next assessed cancer cell migration 
from the spheroids after embedding into 1.5 and 4.5 mg mL-1 3D type I collagen matrix. 
After 24 h post-embedding, both MDA+ and MDAMIX spheroids can be noticed 
migrating using both single and collective modes whereas MDA- spheroids appear to 
utilize a predominantly collective migration mode (Figure 4.3A). When comparing the 
use of single versus collective migration modes between the spheroid conditions, we 
observe that single cell migration is more prevalent in MDA+ spheroids and MDA+ cells 
in MDAMIX spheroids and compared to both MDA- spheroids and MDA- cells in 
MDAMIX spheroids (Figure 4.3B). Conversely, all spheroids appear to show collective 
or strand-like or streaming migration behavior with MDA+ cells still exhibiting 
significantly more migration in this mode compared to MDA- cells (Figure 4.3C). 
Further, we observed both MDA+ and MDAMIX spheroids migrated out into the 
surrounding collagen matrix a significantly larger area than MDA- spheroids (Figure 
4.3D). Interestingly, MDA- cells in the presence of MDA+ cells in the MDAMIX 
spheroids were able to migrate out significantly further that MDA- cells alone. This 
suggests that the presence of MDA+ cells in the co-culture spheroid is somehow 
 137 
 
enhancing the migration ability of the MDA- cells. These results led us to investigate 
the potential mechanism behind the enhanced migration distance of MDA- cells in 
MDAMIX  spheroids. 
 
Figure 4.2. Phenotypic sorted cancer cells form in vitro tumor spheroids with 
differential compaction behavior. A) Representative images of in vitro tumor 
spheroids immediately after seeding (0 h) and before embedding (72 h) with MDA+ 
(green), MDA- (red), and MDAMIX (1:1 MDA+:MDA- co-culture) conditions; scale 
bar: 100 µm. B) Compaction curve of MDA+, MDA-, and MDAMIX spheroids from 0 
h to 72 h. C) Representative image of fully compacted MDAMIX spheroid (left) with 
individual MDA+ (green) and MDA- (red) channels shown across spheroid diameter 
(right); scale bar: 50 µm. D) Relative intensity histogram across diameter of MDAMIX 
spheroid from (C) at 0 h. E) Relative intensity histogram across diameter (white 
dotted line) of fully compacted MDAMIX spheroid from (C) at 72 h. 
 
 138 
 
 
 
 
 
 
 139 
 
 
Figure 4.3. Phenotypic sorted cancer cells exhibit differential migration modes 
in tumor spheroid model. A) Representative images of MDA+, MDA-, and MDAMIX 
tumor spheroids in 1.5 mg mL-1 collagen at 24 h post-embedding. Black arrows inside 
inset images mark single cell migration while white arrows mark collective strand 
migration; scale bar: 100 µm. B) Average number of single cells migrating from 
spheroids at 24 h post-embedding. C) Average number of strands of migrating cells 
per spheroid at 24 h post-embedding.  D) Average invasion area of MDA+, MDA-, 
and MDAMIX spheroids in 1.5 mg mL-1 24 h post-embedding. * p < 0.05, ** p <0.01, 
*** p <0.001, **** p < 0.0001 
 
When observing the migration strands of MDAMIX spheroids in 1.5 and 4.5 mg mL-1 
collagen gels 24 h post-embedding, leader-follower behavior where MDA+ cells lead 
and MDA- cells follow was observed in many of the migration strands (Figure 4.4A). 
When quantified, we observed significantly more MDA+ leader cells than MDA- leader 
cells in both 1.5 and 4.5 mg mL-1 collagen matrices (Figure 4.4B). When comparing the 
average percentage of MDA+ leader cells in MDAMIX spheroids at 1.5 and 4.5 mg mL-
1, the  percentage is significantly higher for the 4.5 mg mL-1 condition (Figure 4.4C). 
Thus, these results suggest that leader-follower behavior where MDA+ leader cells 
facilitate MDA- follower cell migration may be responsible for the significantly further 
distances that MDA- cells in MDAMIX spheroids can displace compared to MDA- cells 
in MDA- only spheroids. 
 
 
 
 
 140 
 
 
 
 
 
Figure 4.4. Co-culture MDAMIX spheroids exhibit leader-follower behavior. A) 
Representative images of migration strands featuring leader-follower behavior in 1.5 
and 4.5 mg mL-1 collagen gels 24 h post-embedding. White arrows note migration 
strands where an MDA+ cell (green) is observed in the leader cell position with an 
MDA-cell (red) in the follower position.  B) Average number of leader cells per 
subpopulation per spheroid in 1.5 and 4.5 mg mL-1 collagen gels. C) Relative 
percentage of leader cells for MDA+ and MDA- cells in MDAMIX spheroids in 1.5 and 
4.5 mg mL-1 collagen gels; ** p < 0.01. 
 141 
 
 
We next wanted to test whether the commensal interactions between MDA+ and MDA- 
subpopulations we have observed in vitro affected metastatic potential in vivo. To test 
this, an orthotopic metastasis model was used where MDA+, MDA-, or MDAMIX (1:1 
mixture of MDA+:MDA-) cells were injected orthotopically at the mammary gland of 
6-8 week old female NSG mice (Figure 4.5A). At 4 weeks or when tumors reached 1 
cm3, primary tumors were surgically removed, and mice were monitored for an 
additional 4 weeks before sacrifice and tissue collection. MDA+, MDA-, or MDAMIX 
injections all resulted in tumors, and when tumors were resected, they were measured 
via calipers. At primary tumor removal, there was no statistically significant difference 
observed in tumor volume (Figure 4.5B). At study endpoint, tissue collection revealed 
macroscopically detectable distal metastasis to lung for both MDA- and MDAMIX 
injected mice and not MDA+ injected mice (Figure 4.5C). These results are consistent 
with our previous results where we observed MDA--injected mice showed significantly 
more distal metastasis compared to MDA+-injected mice (Hapach et al., Cancer Res). 
When GFP-positive and mCherry-positive cells were quantified in both lung and liver 
tissues and normalized by tissue area, we found that the percentage of metastatic 
colonization for MDAMIX tissues was significantly greater than those of MDA- which 
both far exceeded MDA+ in lungs (Figure 4.5D). This trend followed in the liver (Figure 
4.5E,F). Importantly, in the MDAMIX tissues, MDA- cells were still responsible for the 
vast majority of metastatic colonization. There was no significant difference in the 
percent area of metastatic colonization for MDA+ cells in lung or liver. In contrast, the 
percent area of metastatic colonization for MDA- cells was significantly increased in 
 142 
 
both lung and liver of MDAMIX-injected mice compared to MDA--injected mice. These 
results suggest that commensal interactions between migratory and non-migratory 
subpopulations occur in vivo and lead to increased metastatic fitness. 
 
 
 143 
 
Figure 4.5. Phenotypic sorted cancer cells show commensal interactions leading 
to enhanced metastasis in vivo. A) Schematic of orthotopic breast cancer 
spontaneous metastasis model where mice are injected with either MDA+, MDA-, or 
1:1 MDA+:MDA- (MDAMIX) cells, tumors are allowed to grow and then surgically 
removed, and metastasis is measured after collection of lung and liver tissues. B) 
Average primary tumor volume at surgical removal (4 wks post-injection). C) 
Representative images of lung sections from MDA+, MDA-, and MDAMIX-injected 
mice at study endpoint. D) Average relative area covered by metastasis in lungs as 
evidenced by GFP and/or mCherry expression in MDA+ and MDA- subpopulations. 
E) Representative images of liver sections from MDA+, MDA-, and MDAMIX-injected 
mice at study endpoint.  F) Average relative area covered by metastasis in liver as 
evidenced by GFP and/or mCherry expression in MDA+ and MDA- subpopulations. 
* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, n.s. non-significant. 
 
4.5 Discussion 
This study contributes additional nuance to the important discussion of clonal 
heterogeneity and cooperativity in the context of cancer migration and metastasis. While 
it is well established that intratumor heterogeneity is a clinically relevant metric 
associated with worsened patient prognosis and our ability to characterize and quantify 
genetic and phenotypic heterogeneity has vastly improved, much remains to be 
elucidated towards the functional roles of specific subclones in overall metastatic 
fitness.  
 
In this paper, we took phenotypically sorted subpopulations of MDA-MB-231 human 
breast cancer cells that migrate adeptly, MDA+, and poorly, MDA-, and compared their 
migratory and metastatic behaviors alone and together using an in vitro tumor spheroid 
model and an in vivo orthotopic mouse model. We found that the distance that MDA- 
cells were able to migrate out in the tumor spheroid model was enhanced by MDA+ cells 
in 1:1 co-culture MDAMIX tumor spheroids compared to MDA- only spheroids. 
 144 
 
Additionally, we observed that leader-follower behavior was occurring in the MDAMIX 
co-culture spheroids where MDA+ cells were leading MDA- cells in many of the 
migration strands in both 1.5 and 4.5 mg mL-1 collagen. We hypothesize that the MDA+ 
cells clear pathways in the collagen matrix called microtracks and lead MDA- cells out 
by physical and/or chemical means.  
 
To test the relevance of this finding in vivo, we performed orthotopic injections of 
MDA+, MDA-, and MDAMIX into NSG mice, surgically removed primary tumors, and 
allowed 4 weeks for metastasis to occur prior to tissue collection. We found that, 
consistent with our previous findings, MDA--injected mice showed significantly greater 
distal metastasis to lungs and liver compared to MDA+-injected mice. Further, when 
injected as a 1:1 MDAMIX, there was a significant increase in metastasis to lung and 
liver for MDA- cells compared to MDA- only injected mice. We have established that 
MDA+ cells are adept at migration but are poor at metastatic colonization while 
conversely, MDA- cells are poor at migration but are adept at metastatic colonization. 
We believe that the enhanced metastatic spread in MDAMIX mice is due to increased 
cancer cell dissemination in a leader-follower fashion, which ultimately results in more 
MDA- cells arriving at metastatic sites in MDAMIX injected mice compared to MDA- 
only injected mice (Figure 4.6). When we take into account the additional consideration 
that half of MDA- cells are initially injected into MDAMIX mice compared to MDA- 
mice, but we observe a roughly two-fold increase in metastatic colonization in MDAMIX 
mice compared to MDA- mice, this enhancement of metastatic efficacy is quite robust.  
 145 
 
 
Together, these results indicate that commensal interactions occur between these 
phenotypically sorted subpopulations to enhance overall cancer cell migratory and 
metastatic fitness. Future work should be performed to determine how prevalent leader-
follower behavior is in heterogeneous parental cancer cell populations and by what 
chemical or physical means this behavior is mediated. More broadly, this project 
highlights the need for identification and characterization of subclones that are essential 
to metastasis as well as supportive and competing subclones and assessment of how 
these clonal interactions affect metastatic fitness at critical points of the metastatic 
cascade such as intravasation, survival in the circulation, and colonization.  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 146 
 
Figure 4.6. Commensal interaction between phenotypically sorted 
subpopulations hypothesized to be mediated by leader-follower behavior. A) 
Schematic of hypothesized commensal interactions between MDA+ (green) and 
MDA- (red) subpopulations in vivo where dissemination from the primary tumor by 
weakly migratory, highly metastatic MDA- cells is enhanced by highly motile, weakly 
metastatic MDA+ cells in a leader-follower fashion resulting in more MDA- cells 
arriving at metastatic sites and increased colonization compared to MDA- cells alone. 
The proposed overall effect on metastatic fitness is depicted at the bottom of the 
schematic where MDA+ impart an early advantage during dissemination that is 
carried through to the final stage of metastasis, colonization, by MDA- cells. 
 
 147 
 
CHAPTER 5 
 
CONCLUSIONS AND FUTURE DIRECTIONS 
 
5.1 Conclusions 
In this dissertation, we had two primary goals:  
• In Part I, we sought to examine the role of diabetic hyperglycemia and non-
enzymatic glycation on breast tumor progression. 
• In Part II, we wanted to determine the relationship between cancer cell migration 
ability and metastatic potential in both isolated and pseudo-heterogeneous 
contexts.  
As comorbidities and intratumor heterogeneity both present distinct yet significant 
clinical hurdles, both of these goals are critical to improving breast cancer diagnosis and 
treatment, especially for TNBC patients who suffer from heterogeneous and aggressive 
disease along with a paucity of chemotherapeutic options.  
 
In Chapter 2, a novel mouse model was characterized to observe the effects of diabetic 
hyperglycemia established pre-tumorigenesis. Our mouse model provides improved 
physiological relevance compared to a similar recently published experimental system 
that induces hyperglycemia at week 8 since MMTV-PyMT mammary tumorigenesis 
has been reported to occur starting at 6 weeks of age221. As it has been shown clinically 
that increased incidence of breast cancer only occurs in patients who have had an 
extended history of T2DM134, we believe that our model which induces diabetic 
 148 
 
hyperglycemia at week 5 better mimics the pertinent sequence of events more closely. 
In addition to model creation and validation, we also tested the role of diabetic 
hyperglycemia in tumor progression and found striking results. Even in the absence of 
hyperinsulinemia which is often implicated in T2DM-cancer interactions, we still 
observed a striking shift in phenotype with larger, stiffer, and more advanced tumors in 
mice with diabetic hyperglycemia. Importantly, picrosirius staining indicated that 
matrix stiffening was not due to increased collagen deposition, suggesting architectural 
changes could be involved. We also found that diabetic hyperglycemia increased AGE, 
fibronectin, and TGF-𝛽 concentrations as well as EMT ratio of cancer cells in the 
primary tumor. 
 
By treating diabetic mice with either AGE inhibitor, aminoguanidine, or AGE crosslink 
breaker, alagebrium (ALT-711), we saw decreases in tumor size and stiffness 
independent of blood glucose levels. AGE inhibition also resulted in decreased AGE, 
fibronectin, and TGF-𝛽 concentrations as well as reduced cancer cell EMT ratio in the 
primary tumors. All relevant tumor metrics approached non-diabetic control tumor 
levels. This indicates that non-enzymatic glycation plays a role in the observed effects 
of diabetic hyperglycemia on mammary tumor progression. Importantly, as we did not 
observe a difference in AGE concentration between treatments, we were unable to parse 
apart the signaling and crosslinking effects of glycation and AGE formation as we had 
intended. This may potentially be due to the creation of soluble AGEs by alagebrium 
that would not be detected in tumor IHC staining. Comparison of AGE concentrations 
in blood across all conditions could aid in clarifying whether this explanation is accurate 
 149 
 
or not.  
 
Despite being unable to disentangle the relative roles of glycation-mediated matrix 
crosslinking and AGE-RAGE signaling, we still reveal a novel role for non-enzymatic 
glycation in the diabetic hyperglycemia-mediated promotion of tumor progression. This 
finding is somewhat surprising as it indicates that in addition to the relatively direct role 
of elevated blood glucose in fueling cancer cell glycolytic metabolism there is also a 
significant effect from the more indirect effects of glucose via increasing glycation to 
affect cancer cell behavior. Importantly, with these current experiments, we are unable 
to qualitatively or quantitatively assess the relative contributions of glucose metabolism 
and glycation to the in vivo tumor phenotype. This is an important limitation of this 
study that could be at partially remedied with further experimental controls and perhaps 
experiments in a more controllable complementary in vitro system. 
 
In Chapter 3, phenotypically sorted highly migratory (MDA+) and weakly migratory 
(MDA-) breast cancer cell subpopulations were characterized in vitro and in vivo. These 
cells were shown to be distinct and stable in culture with MDA+ cells migrating adeptly 
compared to MDA- which migrated poorly in 3D culture. As cancer cell motility is often 
associated with metastatic potential, we then injected the subpopulations orthotopically 
into NSG mice. We hypothesized that MDA+ cells would metastasize more efficiently 
than MDA- cells. After primary tumor removal, surprisingly, we found that MDA- 
metastasized to lungs, liver, and bone to a greater extent than MDA+, which 
metastasized poorly and were only overtly observed metastasizing to lymph nodes. To 
 150 
 
determine the robustness of these striking findings, in vivo experiments were repeated 
with two additional sets of TNBC phenotypically sorted subpopulations, MCF10CA1a 
and SUM159 cells. After primary tumor removal, we saw the same trends as before, 
where weakly migratory CA1a- and SUM159- subpopulations metastasized to distant 
organs while highly migratory CA1a+ and SUM159+ cells barely metastasized.  
 
To further assess the metastatic abilities of phenotypically sorted subpopulations, a 
series of experiments designed to parse apart the metastatic cascade were implemented. 
To assess initial cancer cell dissemination from the primary tumor, ex vivo en bloc 
tumor sections were used to compare local cell dissemination. Intravasation ability was 
characterized using a modified transwell invasion assay featuring an endothelial 
monolayer. Circulating tumor cells (CTCs) were isolated, stained, and quantified. Ex 
vivo lung colonization was assessed using decellularized lung matrix, which has been 
shown to predict colonization ability. Both subpopulations were able to perform in these 
assays. With steps involving migration, dissemination and intravasation, MDA- cells 
performed to a lesser extent then MDA+ cells but there were no outright failures at any 
step of the deconstructed metastatic cascade. A striking difference between the 
subpopulations was noted upon CTC collection where MDA- cells presented as clusters 
in addition to single cells while MDA+ cells were almost exclusively single cell CTCs. 
This suggested that cell-cell adhesion ability could potentially explain the difference in 
in vivo metastatic potential between MDA- and MDA+ subpopulations. 
 
As our in vivo findings contradicted our initial hypothesis, we performed RNA 
 151 
 
sequencing to determine what broad genetic differences the highly and weakly 
migratory subpopulations possessed. Using GO term analysis, we determined that 
categories involved in epithelial-to-mesenchymal transition (EMT) were highly 
differentially regulated and could explain the in vitro and in vivo phenotypic behaviors 
we had characterized. Across all cell lines, highly migratory cells showed a relatively 
increased mesenchymal index and decreased epithelial index compared to weakly 
migratory cells which showed the opposite trends. In particular, out of a previously 
published EMT gene list, E-cadherin was highly differentially regulated in the weakly 
migratory subpopulations across all three cell lines.  
 
Given our RNA sequencing findings, we first validated the differential expression of E-
cadherin in the phenotypically sorted subpopulations at both transcriptional and 
translational levels. We found that MDA- cells expressed E-cadherin while MDA+ cells 
did not. We then performed an shRNA knockdown of E-cadherin in the weakly 
migratory MDA- subpopulation and used lentivirus to induce E-cadherin expression in 
highly migratory MDA+ cells. When injecting these cells in vivo, we observed that the 
metastatic phenotypes had reversed with MDA-EcadKD cells failing to metastasize 
distally and MDA+ + E-cadherin cells metastasizing readily to lung, liver, and bone. 
These findings suggested that E-cadherin was both necessary for MDA- distal metastasis 
and sufficient to allow MDA+ metastasis to all of the tissues that MDA- cells were able 
to metastasize. Our findings support the recent literature that has found a pro-metastatic 
role for E-cadherin, which has typically been considered a tumor-suppressor protein158.  
 
 152 
 
To gain further understanding of the phenotypic differences between the subpopulations 
and the effect of E-cadherin on these phenotypes, in vitro characterization of migration, 
mechanotransduction, morphologies, and cytoskeletal architecture was performed. 
Here, we found several important distinctions. With several single cell migration assays 
including seeding in 3D collagen matrix, transwell migration assay, and seeding in 
collagen microtracks, we found that no matter the setting, MDA- cells always migrated 
more weakly compared to MDA+. Interestingly, when E-cadherin was knocked down 
or added to the MDA- and MDA+ cells, respectively, the migration behaviors 
approached each other becoming non-significantly different. However, when formed 
into in vitro tumor spheroids and embedded in 3D collagen, behaviors were more 
distinct where MDA+ spheroids migrating as single cells with a larger outgrowth area 
while MDA- spheroids migrating collectively with a shorter outgrowth area. The single 
cell and collective migration modes of MDA+ and MDA- cells were not altered when E-
cadherin was added and knocked down, respectively. However, the outgrowth area was 
enhanced in MDA-EcadKD compared to MDA-, and outgrowth area was attenuated in 
MDA++EcadLow compared to MDA+. Interestingly, in all of these assays, when a much 
higher expression level of E-cadherin was lentivirally induced in MDA+, the migration 
mode shifted to collective migration and outgrowth area was reduced to levels 
comparable with MDA-.  
 
Morphologically, there were distinct differences in the subpopulations with MDA+ cells 
presenting with more ruffling lamellipodia and less cell-cell interaction compared to 
MDA- cells which presented with more prominent stress fibers and more cell-cell 
 153 
 
clustering compared to MDA+. Importantly, both phenotypes could be observed within 
the parental MDA-MB-231 population. Tuning of E-cadherin made the subpopulations 
appear slightly more similar; however, neither the MDA+EcadLow or MDA+EcadHigh lost 
their lamellipodia and still look very distinct from MDA- and MDA-EcadKD. 
Conversely, MDA-EcadKD still exhibits some clustering behavior and doesn’t exhibit 
the ruffling lamellipodia seen in MDA+ with and without E-cadherin. Focal adhesion 
kinase (FAK) activation was found to be increased in MDA- cells compared to MDA+ 
cells. Tuning of E-cadherin lessened this difference with no significant difference in 
focal adhesion size between MDA+EcadLow and MDA-EcadKD, indicating that E-
cadherin expression affects FAK activity. Traction force microscopy and 3D collagen 
gel compaction assays showed that force transduction was greater in MDA+ compared 
to MDA- and again, E-cadherin tuning in each subpopulation brought the variants closer 
to moderate parental levels. The only exception to this was with MDA+EcadHigh collagen 
gels where interestingly, contraction was comparable to MDA+. This suggests that the 
relatively high expression of E-cadherin somehow increases the ability of these cells to 
mechanotransduce and remodel their collagen microenvironment. Interestingly, when 
observed using confocal microscopy, the morphology of MDA+EcadHigh is quite similar 
to MDA- with chainlike networks of interconnected cells that appear quite different 
compared to the largely single cell presentation of the other conditions independent of 
relative collagen gel contraction. Together, these results indicate that while numerous 
behaviors including motility and FAK signaling can be affected by E-cadherin 
expression, some others including morphology and cell contractility are less impacted. 
These findings reinforce that the phenotypically sorted highly and weakly migratory 
 154 
 
subpopulations have extensive differences at both the genotypic and phenotypic levels.   
 
In Chapter 4, in order to better understand how these subpopulations may interact in the 
heterogeneous context of the parental cell population, we examine the potential for 
cooperative interactions between the highly and weakly migratory phenotypically sorted 
subpopulations characterized separately in Chapter 3. Using the in vitro tumor spheroid 
model and the in vivo orthotopic mouse metastasis model also utilized in Chapter 3, we 
compare these subpopulations separately versus a 1:1 co-culture of both subpopulations. 
In both co-culture systems, we observed commensal interactions where the highly 
migratory subpopulation can aid the weakly migratory subpopulation to invade farther 
in vitro and metastasize more effectively in vivo. Interestingly, in the co-culture did not 
result in significantly increased metastasis of the highly migratory cells, which indicates 
that the presence of the highly metastatic, weakly migratory subpopulation was not 
sufficient for the highly migratory cells to improve their metastatic fitness.  
 
We assessed the behavior of the co-culture spheroid model and noticed that the two 
subpopulations had different styles of migration. The highly migratory cells displayed 
far greater single cell migration than the weakly migratory cells, which migrated 
predominantly as collective strands. Additionally, leader-follower behavior was 
observed at both concentrations of collagen matrix where the highly migratory cells 
were more likely to be at the head of a migration strand of cells than the weakly 
migratory cells which were often in follower position behind a cell at the front of a 
migration strand. We hypothesize that the leader-follower behavior observed in vitro is 
 155 
 
also occurring during the initial steps of tumor invasion and intravasation in vivo, 
resulting in enhanced metastasis at secondary sites. This work contributes to the 
literature highlighting the potential role of cooperative subclonal interactions in cancer 
metastasis. Future work should continue profiling the relative strengths and weaknesses 
of subclones within the same primary tumor and observing the dynamics of these 
subclones over the course of the metastatic cascade in order to better understand these 
interactions. 
 
Overall, this body of work has contributed to the already monumental strides that have 
been made in understanding triple negative breast cancer tumor progression and 
metastasis. As we increase our knowledge of this disease, these efforts will eventually 
translate into more effective diagnostic strategies, novel therapeutic approaches, and 
optimized patient treatment regimens. The implications of comorbidities such as 
diabetes mellitus on cancer progression need additional attention as these diseases are 
projected to increase in prevalence worldwide in the future. This work has demonstrated 
a novel role for non-enzymatic glycation in the enhancement of tumor progression due 
to hyperglycemia. Further, intratumor heterogeneity is a characteristic inherent to most 
cancers that imparts this disease with resilience that greatly impedes diagnosis and 
treatment. This work parsed apart the heterogeneity in TNBC cell migration and found 
that the weakly motile subpopulation was essential for metastasis. This finding would 
have been obscured if not for the phenotypic sorting technique employed to separate 
these cells by migratory phenotype. While the protein we implicated in our system, E-
cadherin is not an easy target as it is involved with normal epithelial function, further 
 156 
 
study into how cancer utilizes E-cadherin for metastasis could lead to improved targeted 
treatment approaches as it may have other involved proteins that are less ubiquitous.  
 
5.2 Future Directions 
In Chapter 2, we present a novel mouse model that was used to assess the role of diabetic 
hyperglycemia on mammary tumor progression. There are several major questions 
prompted by these findings that could be expanded upon with the following proposed 
experiments: 
1) How do these findings stand up in a more complex T2DM environment? 
2) Does diabetic hyperglycemia encourage or inhibit metastatic colonization? Does 
glycation inhibition have positive or negative effects on cancer metastasis? 
3) How does diabetic hyperglycemia affect tumor evolution? 
4) How does diabetic hyperglycemia affect tumor associated stromal cells – cancer 
associated fibroblasts, endothelial cells, immune cells?  
5) Is glycation inhibition sufficient to shift diabetic tumor metabolism independent of 
blood glucose levels? 
 
Question 1 addresses an important consideration in our model; our experimental system 
isolates the effects of diabetic hyperglycemia but does not reflect the entire metabolic 
endocrine disorder that is T2DM. Although the method of hyperglycemia induction 
utilized in our studies, a series of daily streptozocin (STZ) injections, is extremely 
common experimentally, the resulting system does not fully recapitulate the 
complexities of T2DM. In fact, as the mechanism of STZ is to chemically destroy the 
 157 
 
pancreatic beta islet cells that are responsible for insulin secretion, there is no 
hyperinsulinemia or insulin resistance developed in this model, and thus, it is actually 
more physiologically similar to Type 1 diabetes292. While the lack of hyperinsulinemia 
has helped reduce the complexity of our experimental system allowing for clearer 
results, for improved clinical relevance, supplementary model systems should be 
considered to further validate our findings.  
 
In order to assess the relative weight of non-enzymatic glycation in a more complete 
T2DM system, Leprdb/db mice which possess an autosomal recessive mutation in the 
leptin receptor could be utilized292. These mice present with hyperphagia, obesity, 
hyperglycemia, and hyperinsulinemia, making them more faithful models of T2DM292. 
These mice have a C57BLKS/J background so to allow for cancer studies to be 
incorporated, orthotopic injections of the syngeneic EO771 and metastatic derivative 
EO771.LMB cell lines could be implemented293. These cells complement the studies 
performed in the MMTV-PyMT mouse model as these cells are also triple negative and 
basal-like293. Orthotopic models such as this, while not as physiologically relevant as 
spontaneous models such as the MMTV-PyMT mouse, do allow for primary tumor 
removal surgery which can extend observation times so that metastatic spread can be 
compared.  
 
While our model benefits from earlier induction of hyperglycemia than others, one 
limitation is that the tumor size increase in STZ-injected mice is so robust that studies 
cannot continue past 12 weeks. At this point, there is no readily detectable metastatic 
 158 
 
spread which generally occurs starting at week 14 in the MMTV-PyMT model294. 
Another possible orthotopic model to observe metastasis in our system would be to use 
PyMT- FVB mice with and without STZ injections with orthotopic injection of 
syngeneic cells such as the weakly metastatic DB.7 and/or highly metastatic Met-1 cell 
lines295. Either of these proposed orthotopic models would allow us to address question 
2, allowing further assessment of the effects of diabetic hyperglycemia and non-
enzymatic glycation extending beyond primary tumor progression to metastasis. 
 
To answer questions 3 and 4, single cell RNA sequencing (scRNAseq) experiments of 
PyMT mammary tumors with and without diabetic hyperglycemia could be performed 
at several timepoints correlating with premalignant lesions, early carcinoma, and late 
carcinoma as delineated by Lin et al.294. Recently, scRNAseq was performed on kidney 
cortex of diabetic and non-diabetic patients to assess the single cell changes that occur 
during diabetic nephropathy296. Interestingly, they identified increased potassium 
secretion and angiogenic signaling as early kidney responses to diabetes296. As our 
model saw no difference in vascular density but a difference in vascular permeability, 
using scRNAseq to compare the regulation of diabetic and non-diabetic tumor 
endothelial cells at the aforementioned stages of tumor progression could elucidate 
critical changes and provide novel therapeutic targets to normalize vascular 
permeability for enhanced drug delivery. 
 
To address question 5 and assess the metabolic impacts of diabetic hyperglycemia with 
and without glycation on tumor metabolism, several in vitro and in vivo experiments 
 159 
 
could be implemented. To parse apart the impacts of glucose and glycation on cancer 
cell behavior, an in vitro experiment where cancer cells are seeded in glycated and non-
glycated collagen matrices should be implemented. As cancer cells cannot metabolize 
ribose but ribose can be involved in the glycation of collagen fibers, the use of this sugar 
can aid in testing the relative contributions of these parameters in a highly controlled, 
reproducible manner. Cells must be adjusted to either normal or hyperglycemic levels 
of glucose in media prior to seeding. The conditions tested should be as follows: 1) cells 
in non-glycated collagen with hyperglycemic glucose levels, 2) cells in non-glycated 
collagen with normal glucose levels, 3) cells in glycated collagen with hyperglycemic 
glucose levels, and 4) cells in glycated collagen with normal glucose levels. After 
seeding and allowing time to respond to their environments, cells will be fixed and 
stained for Ki-67, vimentin, E-cadherin, phospho-myosin light chain, and actin to assess 
proliferation, EMT status, and mechanotransduction. Metabolic activity could be 
broadly assessed using several probes including 2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-
yl)amino)-2-deoxyglucose (2-NBDG), tetramethyl rhodamine ethyl ester (TMRE), and 
PercevalHR (with accompanying pH probe), which can measure glucose uptake, 
mitochondrial membrane potential, and ATP:ADP ratio, respectively291,297. This will 
allow more exact comparison of the roles of non-enzymatic glycation and glucose 
metabolism than in the in vivo system. 
 
As a parallel experiment to the in vitro system outlined above, either ribose or L-glucose 
could be used to parse apart the metabolic and non-enzymatic glycation roles of glucose 
in the in vivo PyMT mouse model used for this work thus far. Both ribose and L-glucose 
 160 
 
cannot be metabolized like D-glucose but can still be involved in glycation so treating 
control mice with ribose or L-glucose should theoretically allow the effect of glycation 
alone to be observed. However, there are several major experimental design challenges 
with this setup. First, dosing of mice with either would require some sort of live 
detection method since the goal would be to approximate the level of glycation 
occurring in the STZ-injected hyperglycemic mice. This could be done by using blood 
samples collected via either tail snip or jaw bleed and performing a test for glycated 
serum protein (GSP) concentration to approximate systemic glycation activity. Once the 
serum levels of glycated proteins are measured in the STZ-injected mice, the ribose or 
L-glucose injections can be performed at several levels to find a dosage with comparable 
GSPs. Once these parameters are determined, the same stains and mechanical tests on 
ex vivo tumor samples could be performed to assess the isolated role of increased 
glycation independent of increased glucose metabolism. 
 
 18F-fluorodeoxyglucose (18F-FDG) is a glucose analog commonly utilized to assess 
metabolic activity of tumors with positron emission tomography/computed tomography 
(PET/CT)298. This method could be used to compare the relative levels of tumor 
metabolic activity between normal and hyperglycemic tumors with and without 
glycation inhibition to determine if glycation can affect tumor metabolism independent 
of blood glucose concentrations. As an important caveat, these experiments may not be 
feasible in our model since hyperglycemic conditions may impede appropriate data 
collection and/or interpretation and thus, quickly managing glucose levels via an insulin 
injection/pump prior to imaging may need to be considered. In addition, isolation of 
 161 
 
primary PyMT tumor cells from diabetic and non-diabetic mice and metabolic 
characterization using Agilent Seahorse XF Analyzer could provide insight into the 
metabolic phenotypes of these cells. This assay has been used to assess mitochondrial 
function in primary cells isolated from leptin-deficient and wildtype MMTV-PyMT 
mouse tumors299. This platform was also recently used to determine that mitochondrial 
metabolism, oxidative phosphorylation, and ATP production are inhibited in pancreatic 
beta islet cells in diabetic mice300. Given the wealth of new information implicating 
metabolic dysregulation as a hallmark of cancer progression301, this is an especially 
relevant topic as diabetes is thought of primarily as a metabolic disorder. 
 
In Chapter 3, we compared the relative metastatic potentials of highly and weakly 
migratory subpopulations of triple negative human breast cancer cell lines, MDA-MB-
231, MCF10CA1a, and SUM159. Here, we surprisingly found that for each cell line, 
the weakly migratory subpopulation was highly metastatic and the highly migratory 
subpopulation was weakly metastatic. Using RNA sequencing, we found that the 
subpopulations showed differential EMT and cell-cell adhesion gene regulation and 
focused in on E-cadherin as it was the most differentially expressed epithelial marker in 
the MDA- subpopulation. While this work showed that E-cadherin was necessary and 
sufficient for distal metastasis in either subpopulation, there are many unanswered 
questions that could be further interrogated to increase the impact of this work. These 
major questions include: 
1) How do the MDA- cells disseminate from the primary tumor? Does this 
mechanism differ from that of MDA+ cells? 
 162 
 
2) Is there a transcription factor(s) responsible for the difference in these observed 
phenotypes? Are there other cell-cell adhesion proteins that can act in the same 
manner as E-cadherin in this system? 
3) What role of E-cadherin is essential to metastasis: cell-cell adhesion or cell 
signaling? 
4) If we tune expression levels of E-cadherin in the subpopulations, do we observe 
a dose-dependent effect on metastasis?  
5) Does this finding hold true outside of triple negative breast cancer?  
 
Question 1 was prompted by several experimental observations from already collected 
data. In several regions of the en bloc tumor sections, vessel infiltration/occlusion in the 
tumor periphery were observed (Figure 5.1A). Additionally, when staining for CD31 
was performed on MDA-MB-231 subpopulation primary tumors, the percent 
vascularity as calculated by Leica Digital Image Analyzer software trended towards 
higher vascularity in the MDA- tumors as compared to MDA+ tumors (Figure 5.2A). 
This trend followed for the calculation of vessel number as calculated by the Leica 
Digital Image Analyzer software with increased vessel count in MDA- tumors compared 
to MDA+ tumors (Figure 5.2B). This data could suggest a differential ability between 
the subpopulations in accessing the vasculature within or nearby the primary tumor, 
which could have important albeit subtle implications in metastatic fitness.  
 
 163 
 
 A 
Figure 5.1. Regions of vascular infiltration in MDA- en bloc sections. A) Regions 
of interest in en bloc tumor sections with immunohistochemical staining for GFP 
(brown) and hematoxylin counter staining (blue). 
 
 
 
In addition, several of the most differentially upregulated genes in the RNA sequencing 
for the weakly migratory subpopulations across all three cell types were genes involved 
in EGFR signaling (EREG, ANO9, AREG) or genes associated with endothelial cells 
(TNFSF15, LIPG) (Figure 5.3A). As EGFR signaling is associated with regulation of 
angiogenesis302, it may be interesting to investigate whether there is a difference 
between highly and weakly motile subpopulations in inducing angiogenesis during 
primary tumor progression. Another implication of increased endothelial gene 
expression in the weakly migratory subpopulation that could be investigated is the 
potential for vascular mimicry303. Vascular mimicry is the process by which non-
endothelial cells can form their own patent vascular channels303. This process has been 
implicated in numerous cancer types including triple negative breast cancer and is 
 164 
 
associated with cancer cell survival, metabolism, and metastasis303,304. The method for 
identifying these vascular mimicry is to perform CD31 and periodic acid-Schiff (PAS) 
co-staining304. Tumor vessels can be identified by the presentation of a characteristic 
pattern of CD31- PAS+ cells often with erythrocytes in the center304. As vascular 
mimicry would put the cancer cells in direct contact with blood flow, it is associated 
with increased metastasis303. While the discovery of vascular mimicry emerged with a 
heavy amount of skepticism305, this phenomenon has gained acceptance in the field and 
could offer an explanation as to how weakly motile cells reach the circulation in 
comparable numbers to their highly motile counterparts. 
 
A B 
   
Figure 5.2. Quantification of vascularization in MDA-MB-231 subpopulation 
primary tumors. A) Percent vascularization and B) normalized number of vessels of 
MDA+ and MDA- primary tumors at 4 weeks post-injection based on CD31 
immunohistochemical staining as calculated using Leica Digital Image Analyzer (n 
= 3). 
 
 
 
 
 
 
 165 
 
 
A B  
Gene list 
ANO9 
TNFSF15 
PPP1R14C 
EREG 
INHBA 
AREG 
LIPG 
Figure 5.3. Differential Gene Expression MDA- vs MDA+ (logF2C 1.5).  A) 
Graphical representation of genes from RNA sequencing data that are differentially 
upregulated in weakly motile subpopulations in each cell line. B) Gene list of genes 
differentially upregulated in weakly migratory subpopulation of all 3 cell lines. 
Data provided by Emily Chu. 
 
 
To investigate cancer cell interactions with the vasculature and angiogenic activity, live 
cell imaging at the primary tumor by adapting the orthotopic mouse model to include 
an imaging window would be helpful to visualize cell behavior in real time in vivo. This 
model would be further enhanced with transgenic fluorescent tagging of endothelial 
cells so that they could be readily appreciated both within and outside of the primary 
tumor. However, unfortunately, this type of transgenic system is not readily available 
in immunocompromised mice. One potential avenue for observing these behaviors in 
real time would be to utilize the chick chorioallantoic membrane (CAM) model as used 
by our group and others151. Depending on the experimental goals, multiple setups can 
be employed including: 1) direct injection/onplant of cancer cells, 2) an avascular 
collagen matrix onplant seeded with cancer cells, or 3) an acellular collagen matrix 
onplant doped with concentrated conditioned media from either subpopulation. An 
 166 
 
example image of fluorescence microscopy of a CAM with a tumor spheroid onplant 
after fluorescent tracer injection to mark the vasculature (Figure 5.4A). This model 
benefits from relatively easy monitoring of angiogenic activity without extensive 
system manipulation. However, there is a considerable degree of intra- and inter-CAM 
heterogeneity in regards to vascular structure during CAM development so placement 
of onplants and study outputs must be standardized as much as possible to draw accurate 
conclusions. Additionally, the CAM model is not necessarily high-throughput and 
suffers from considerable sample attenuation due to contamination and failed tracer 
injections. Despite the limitations, the CAM model offers an intermediate level of 
complexity and accessibility between traditional in vitro and in vivo models of cancer-
vessel interactions that can greatly complement these studies. 
 
Another consideration in the answering of question 1 involves the possibility that the 
cancer cell subpopulations interact differentially with the primary tumor 
microenvironment or that they could exhibit differential interactions with stromal cell 
types that could assist in initiating metastasis. A preliminary observation of the collagen 
architecture of MDA-MB-231 primary tumor sections shows that MDA- tumors have 
more dense, bundled collagen fibers compared to MDA+ tumors (Figure 5.5A). This 
experimental observation warrants further investigation since more dense, stiffened 
extracellular matrix is associated with increased metastasis and worsened patient 
outcomes105. In particular, cancer cells on stiffer matrices are more invasive and exhibit 
a more malignant phenotype31,306. The more dense, bundled collagen observed in MDA- 
tumors could be due to the cancer cells themselves or stromal cells such as cancer-
 167 
 
associated fibroblasts (CAFs) interacting differentially with matrix and/or differentially 
depositing matrix. However, the overall state of the tumor ECM is attributed largely to 
fibroblast activity139. Thus, future directions should compare the ability of each 
subpopulation to recruit and activate fibroblasts/CAFs at the primary tumor.  
 
 A 
 
Figure 5.4. Chick chorioallantoic membrane (CAM) model. Representative image 
of periphery of tumor spheroid onplant interacting with CAM vasculature ex ovo. 
MDA-MB-231 cancer cells were tagged with GFP (green), and CAM vasculature was 
indicated by TRITC-dextran tracer injection prior to imaging (red). 
 
 
 168 
 
A  
Figure 5.5. Anisotropy maps of quantitative polarization microscopy signal from 
Picrosirius red stained MDA-MB-231 subpopulation primary tumor sections. A) 
Representative images of anisotropy maps of Picrosirius red stained tumor sections 
obtained using quantitative polarization microscopy (QPOL). Heat map scale is 
normalized from 0 (blue) to 1 (red).  
 
 
For question 2, the results of the RNA sequencing data analysis can be further delved 
into. For instance, PRRX1 was the most highly upregulated EMT-associated gene for 
MDA+ and SUM159+ subpopulations. PRRX1 is a EMT-associated transcription factor 
that has been implicated in cancer cell migration in metastasis in TNBC as well as other 
cancers307. Importantly, Ocaña et al. discovered that Prrx1 must be repressed for 
metastasis to occur and that Prxx1 decouples EMT and stemness, which are generally 
correlated277.  
 
Interestingly, in the MCF10CA1a subpopulations, the gene regulation signature looks 
very different with ZEB1, ZEB2, TWIST1, TWIST2,  PRRX1, and SNAI1 all highly 
upregulated in the CA1a+ subpopulation compared to their CA1a- counterparts. This 
suggests that the mechanism behind the differential EMT phenotypes may differ 
between the cell lines. Additionally, the MCF10CA1a subpopulations have the most 
 169 
 
strikingly different phenotypes as evidenced by their morphologies as well as 
immunofluorescence staining and Western blotting for E-cadherin and vimentin (Figure 
5.6A,B). 
A B 
 
Figure 5.6. EMT marker staining reveals differential EMT phenotypes between 
MCF10CA1a parental cell line and subpopulations. A) Representative images of 
immunofluorescence staining of MCF10CA1a parental and subpopulations for 
vimentin (green), E-cadherin (red), and DAPI (blue). B) Western blot of 
MCF10CA1a subpopulations for E-cadherin, vimentin, and GAPDH. 
 
This more pronounced difference in EMT phenotype led to the search for a more 
comprehensive way to compare the RNA sequencing data amongst the cell lines and 
their respective subpopulations. Principal component analysis was used for this purpose. 
When comparing the first two principal components, the results were based primarily 
on the cell line with the subpopulations clustering tightly with their respective parental 
cell lines (Figure 5.7A). It is interesting to note that for all three cell lines, the distance 
between the weakly migratory subpopulation and the parental population is closer than 
the distance between the highly migratory subpopulation and their respective parental 
population. This could potentially suggest that the majority of cells in the parental 
populations are more similar to the weakly migratory subpopulation than the highly 
 170 
 
migratory subpopulation. Further assessment of this by either flow 
cytometry/fluorescence activated cell sorting or single cell RNA sequencing could 
provide more insight into this finding. The third component stratified the cell lines and 
subpopulations by EMT status (Figure 5.7B). Here, we see that the MCF10CA1a 
subpopulations have the most distance between each other and their respective parental 
subpopulation compared to the MDA-MB-231 and SUM159 cell line. It is unclear what 
this means about the cell lines and the sorting process, but two options should be 
considered: 1) If the sorting process is somehow evolving cells, the MCF10CA1a 
subpopulation was the most susceptible to these environmental pressures and 2) if the 
difference reflects the inherent heterogeneity of the cell line, the MCF10CA1a cell line 
is the most heterogeneous of the three with both extreme mesenchymal and extreme 
epithelial phenotypes coexisting. One way to test this would be examining changes in 
somatic mutations over the course of phenotypic cell sorting as has been done for stem 
cell over serial passaging308. However, it is important to perform further validation of 
these findings as false positives are common in SNP studies. 
 
Questions 3 and 4 for Chapter 3 both relate to E-cadherin, an epithelial cell-cell adhesion 
protein implicated in cancer metastasis in our study. Question 3 arises from the dual 
roles of E-cadherin: cell-cell adhesion and cell proliferation signaling 
functions158,272,309–311. Several studies have implicated the role of E-cadherin in 
metastasis via the formation of circulating tumor cell clusters or circulating tumor cell 
microemboli (CTM) while others have pointed to the necessity of E-cadherin to induce 
proliferation towards colonization of metastasis sites158,272,312. While this study shows 
 171 
 
indirect evidence that could support either or both of these hypotheses, we have not 
actively probed this further in the context of our phenotypically sorted cancer cell 
subpopulations. To determine the exact role of E-cadherin in this system, E-cadherin 
adhesive function activating antibodies could be administered in vivo311 or 
subpopulations could be made to express E-cadherin mutations with deletions of 
domains with adhesive or proliferation signaling functions310.  
 
Question 4 delves further into the relationship between E-cadherin and metastasis. 
Specifically, is any amount of E-cadherin sufficient for metastasis, is the relationship 
related to the amount of E-cadherin expressed, or is there a certain threshold of E-
cadherin expression perhaps in relation to other mesenchymal markers that prevents cell 
migration to the extent that metastasis is inhibited? This question is especially relevant 
since clinically, evidence has been conflicting as to whether E-Cadherin’s role in breast 
cancer metastasis is suppressive or supportive270. Thus, understanding the context of 
when E-cadherin is suppressing or enhancing metastasis is critical. While the relative 
abundance of E-cadherin could potentially be a factor, the role of E-cadherin may likely 
also be determined by additional factors inherent to certain classifications of cancer. E-
cadherin loss is considered a hallmark of invasive lobular carcinoma312. However, in 
the most common type of breast cancer, invasive ductal carcinoma (IDC), as well as in 
inflammatory breast cancer, E-cadherin is found in primary tumors, circulating tumor 
microemboli, as well as metastases312. Even more compelling is that in IDC, E-cadherin 
has been shown to correlate with worsened patient survival and is enriched at metastatic 
sites compared to primary tumor levels281,313. While more needs to be done to determine 
 172 
 
in which instances E-cadherin expression can possess prognostic value, the presence of 
hybrid EMT states may also potentially serve as a prognostic indicator. Co-expression 
of E-cadherin and vimentin in invasive breast cancer tumors was recently shown to be 
associated with worsened patient outcome314.  Several recent efforts to characterize 
cellular EMT status with recognition of hybrid EMT states show potential for future 
clinical translation315,316.   
 
 
 
 
 
 
 
 
 
 173 
 
A  
 
 
CA1a+ 
MDA+  
SUM159+ 
CA1a- 
 MDA- 
SUM159- 
 CA1aPAR 
MDAPAR 
 SUM159PAR 
 
 
 
 
B 
 
 
CA1a+  
MDA+ 
SUM159+  
CA1a- 
MDA- 
SUM159- 
CA1aPAR 
MDAPAR 
SUM159PAR 
 
Figure 5.7. Principal component analysis of RNA sequencing data for three 
triple negative breast cancer subpopulations. A) First and second principal 
components cluster data by cell line. B) First and third principal components allow 
separation based on EMT status. Data contributed by Tanner McArdle. 
 
 
 
 174 
 
To answer question 4, which pertains to the potential effects of varying levels of E-
cadherin expression on migration and metastasis, two different approaches could be 
employed. Similar to this study, different amounts of lentivirus could be used or the 
promoter in the lentiviral plasmid could be changed to allow tunable levels of E-
cadherin expression for each subpopulation. In a more complicated but physiologically 
relevant approach, epithelial-to-mesenchymal transcription factor signaling could be 
tuned with subsequent changes in E-cadherin expression levels measured. While this 
approach could be more realistic, it is impeded by the complexity of EMT signaling, 
where E-cadherin expression can be affected by numerous transcription factors (TFs) 
including Snail, Slug, and Twist, as well as post-translational changes and epigenetic 
modifications and cannot be isolated from other changes resulting from the master 
regulator roles of these TFs317,318.  
 
Question 5 considers the scope of these findings as we focused this study on three 
human triple negative breast cancer cell lines. While recently, several studies have 
shown a direct relationship between E-cadherin expression and metastasis, traditionally, 
E-cadherin has been characterized as a tumor suppressor in breast cancer and more 
generally across cancer types319,320. In non-small cell lung cancer (NSCLC), bladder 
cancer, oropharyngeal squamous cell carcinoma, acute myeloid leukemia reduced E-
cadherin expression has been associated with worsened patient outcomes321–325. 
Interestingly, a study comparing metastatic and non-metastatic sublines of a 
hepatocellular carcinoma (HCC) cell line found E-cadherin necessary for intrahepatic 
metastasis326. Also in HCC, the localization of E-cadherin in the cytosol as opposed to 
 175 
 
the membrane was associated with worsened patient outcomes327. In gastric cancer, 
soluble E-cadherin measured in patient blood has been shown to be associated with poor 
prognosis and recurrence328. 
 
While this work generates far more questions than it answers, it contributes greatly to 
several rapidly evolving conversations in the field of cancer research. This project set 
out to parse apart intratumor heterogeneity, an aspect of tumor biology that we postulate 
obscures findings such as ours in many commonly performed population-based cellular 
assays. After obtaining initially counterintuitive results when weakly migratory 
subpopulations were found to be highly metastatic and highly migratory subpopulations 
were weakly metastatic, extensive in vitro and in vivo characterization has imparted a 
great deal of information about these weakly and highly migratory subpopulations. In 
particular, the characterization of their place in the EMT spectrum opens up numerous 
avenues of inquiry. The current heightened interest in E-cadherin in cancer progression 
by our group and others shows that there are important nuances regarding the role of 
EMT in cancer metastasis that need to be further elucidated.  
 
In Chapter 4, cooperative interactions between the phenotypically sorted cancer cell 
subpopulations were explored in vitro and in vivo. Here, we found commensal 
interactions where the highly migratory subpopulation assisted the weakly migratory 
subpopulation in migrating in a leader-follower fashion. We postulate this behavior 
ultimately led to enhanced metastasis of the weakly migratory cells in vivo. While this 
seems a logical conclusion, further work to better characterize this system and solidify 
 176 
 
our findings would include performing histology of en bloc tumor sections and 
circulating tumor cell collection to validate that leader-follower behavior is indeed 
occurring in vivo. Aside from further experimental validation, several questions that 
arise from this work include: 
1) What is the mechanism underlying leader-follower behavior between the 
subpopulations? 
2) Would the SUM159 and MCF10CA1a subpopulations also exhibit any 
cooperative behavior in vitro or in vivo? 
3) How much overlap exists between the phenotypes and genotypes of leader and 
follower cells within the parental MDA-MB-231 cell line and the phenotypically 
sorted highly and weakly migratory subpopulations? 
 
Question 1 asks for mechanistic insight into the nature of the leader-follower behavior 
between the subpopulations which would greatly enhance this work. Based on the data 
collected thus far, we hypothesize that the highly migratory cells have enhanced 
collagen remodeling ability compared to their weakly migratory counterpart. Thus, one 
potential mechanism is simply the creation of collagen microtracks within the collagen 
matrix enable MDA- cells to migrate within these permissive channels where matrix 
metalloprotease (MMP) activity is no longer necessary. Another could be via 
mechanical forces transduced via cell-cell connections such as the heterotypic E-
cadherin and N-cadherin mediated leader-follower behavior between cancer cells and 
cancer associated fibroblasts shown by Labernadie et al150. Another option would be 
that cell-cell signaling such as that recently delineated in breast cancer cells via AKT 
 177 
 
signaling could be responsible for the leader-follower behavior between these 
subpopulations329. To test these potential hypotheses, administration of compounds 
preventing MMP activity, cell binding, and cell secretion could be separately explored. 
Additionally, the secretome of each subpopulation could be assayed and compared to 
determine whether there are any known chemoattractants secreted by the MDA+ cells 
that the MDA- cells are responsive to. While it may be that one or a combination of 
these potential mechanisms are responsible for the leader-follower behavior observed 
between these subpopulations, it is important to determine as one could envision 
treatments aimed at preventing the enhanced dissemination of these weakly migratory 
but highly metastatic cancer cells by disrupting leader-follower behavior. 
 
Question 2 inquires about the possibility of cooperative subclonal interactions between 
the other two sets of phenotypically sorted cancer cell subpopulations that were not 
tested here. In preliminary experiments with the MCF10CA1a subpopulations using the 
in vitro tumor spheroid model, we observed spatial separation between the 
subpopulations within the 1:1 co-culture system (Figure 5.8A). The reason for this 
rearrangement of the subpopulations into essentially two opposing hemispheres may be 
due to differential cell-cell or cell-ECM interactions or mechanical forces. However, 
regardless of the cause, this behavior is not immediately helpful for direct comparison 
with the MDA-MB-231 subpopulations. It would be interesting to test the co-culture 
model with the SUM159 subpopulations since their EMT status appear more similar to 
MDA-MB-231 subpopulations and less divergent than the MCF10CA1a 
subpopulations, but this has not yet been done. Given these cell line dependent 
 178 
 
differences observed in the in vitro tumor spheroid model where it does not appear that 
the MCF10CA1a subpopulations interact heavily with each other unlike the MDA-MB-
231 subpopulations, performing the in vivo orthotopic mouse metastasis model with co-
injections of the MCF10CA1a and SUM159 subpopulations and comparing these to 
their respective baselines could provide information about how universal leader-
follower behavior is in TNBC.  
 
 
A 
 
Figure 5.8. MCF10CA1a subpopulation 1:1 co-culture tumor spheroid model. 
A) Representative images of 1:1 co-culture tumor spheroid with highly migratory 
(green) and weakly migratory (red) MCF10CA1a subpopulations embedded in 1.5 
mg/mL collagen gel at 2 (left) and 4 (right) days post-embedding. Arrows indicate 
protrusion or migration behavior with color corresponding to subpopulation. 
 
 
 
 179 
 
Lastly, question 3 asks how the phenotypically sorted highly and weakly migratory 
subpopulations would compare to leader and follower cells selected out of the parental 
MDA-MB-231 cell line as has been done by Konen et al56. Given that the 
subpopulations tested by Konen et al. were selected for by the behavior that our 
subpopulations elicit, it would be a highly relevant comparison for our phenotypic 
sorting method. This also helps to answer whether leader-follower behavior 
synonymous with highly and weakly migratory behavior or if leader-follower cells are 
specialized subsets of the broader highly and weakly migratory cell populations. This 
could be parsed apart by adopting the methods of Konen et al. in our already sorted 
subpopulations and comparing sequencing data from the leader and follower MDA+ 
cells, the leader and follower MDA- cells, and their respective “parental” MDA+ and 
MDA- populations. This analysis could be expanded beyond the MDA-MB-231 cell line 
if other subsequently sorted cell lines also exhibited leader-follower behavior for more 
robust findings. 
  
 
 
 
 
 
 
 
 
 180 
 
APPENDIX A 
 
Induction of Diabetic Hyperglycemia in MMTV-PyMT Mice Pre-Tumorigenesis 
Protocol developed by Dr. Lauren A. Griggs 
 
Materials: 
• Streptozocin (S0130 Sigma) 
• Sodium citrate buffer 
• 26 G 1/2” syringe and needle 
• Glucose meter and glucose test strips 
• Isoflurane and isoflurane vaporizer 
 
1) At 4 weeks of age, switch mice to high fat diet 
2) On the morning of the first injection on week 5, fast mice in a clean cage for 5 hrs 
(use deprivation alert card ) 
3) Measure the blood glucose level of each mouse and record 
4) Measure ~7 mg of streptozocin into an Eppendorf tube, keep tube on ice and 
covered from light 
5) Anesthetize mice to be injected using IACUC guidelines 
6) Weigh mice and measure blood glucose, record, perform calculations to inject 
mice at 70 mg/kg 
7) Prior to injection, dissolve streptozocin with 1mL citrate buffer 
8) Inject each mouse intraperitoneally with appropriate volume of STZ solution 
(citrate buffer only for control mice) 
9) Allow mice to recover in clean cage and monitor until ambulating normally 
10) Replenish food in cages 
11) Repeat steps 2-10 daily for 5 consecutive days  
12) Perform weekly glucose and weight measurements until study endpoint 
 181 
 
APPENDIX B 
 
Immunofluorescence Staining of Frozen Mouse Tissues 
Protocol developed by Wenjun Wang 
 
1) Remove slides from -80 ºC freezer and immediately fix samples with 4% 
paraformaldehyde in PBS for 10 min at RT 
2) Wash 3x 5 min with PBS + 0.02% Tween 20 
3) Permeabilize tissue with PBS + 1% Triton for 5 min 
4) Wash 3x 5 min with PBS + 0.02% Tween 20 
5) Use hydrophobic pen to draw circle around each tissue on slide to minimize 
amount of blocking agent and antibody needed for staining 
6) Block samples with PBS + 10% FBS 5% donkey serum 5% goat serum for 2 h at 
RT 
7) Incubate tissues in primary antibody diluted in blocking solution overnight at 4ºC  
8) Wash 3x 5 min with PBS + 0.02% Tween 20 
9) Incubate tissues in secondary antibody diluted in blocking solution 2 h at RT in the 
dark 
10) Wash 3x 5 min with PBS + 0.02% Tween 20 in the dark 
11) Stain 1:500 with DAPI for 15 min at RT in the dark 
12) Wash 3x 5 min with PBS + 0.02% Tween 20 in the dark 
13) Mount slides with Vectashield with No. 1.5 coverslip then seal edges with nail 
polish 
14) Image slides on laser scanning confocal microscope 
 
 
 
 
 182 
 
APPENDIX C 
 
Immunohistochemistry Staining of Frozen Mouse Tissues 
Protocol developed by Wenjun Wang 
 
1) Remove slides from -80 ºC freezer and immediately fix samples with 4% 
paraformaldehyde in PBS for 10 min at RT 
2) Wash 3x 5 min with TBS + 0.02% Tween 20 
3) Permeabilize tissue with TBS + 1% Triton for 5 min 
4) Wash 3x 5 min with TBS + 0.02% Tween 20 
5) Use hydrophobic pen to draw circle around each tissue on slide to minimize 
amount of blocking agent and antibody needed for staining 
6) Block samples with TBS + 10% FBS 5% donkey serum 5% goat serum for 2 h 
at RT 
7) Incubate tissues in primary antibody diluted in blocking solution overnight at 
4ºC  
8) Wash 3x 5 min with TBS + 0.02% Tween 20 
9) Incubate slides with TBS + 0.3% H2O2 for 15 min at RT 
10) Incubate tissues in secondary antibody diluted in blocking solution 1 h at RT  
11) Wash 3x 5 min with TBS + 0.02% Tween 20 
12) Develop with DAB 1.5-3 min at RT (be consistent between samples) 
13) Rinse samples with running tap water for 5 min 
14) Counterstain with Mayer’s Hematoxylin for 3 min 
15) Rinse in tap water until nuclei are blue (5 min) 
16) Dehydrate and clear 
a. 50% ethanol in tap water 
b. 70% ethanol in tap water 
c. 95% ethanol in tap water 
d. 100% ethanol in tap water x2 
 183 
 
e. 2:1 ethanol:xylene 
f. 1:1 ethanol:xylene 
g. 1:2 ethanol:xylene 
h. xylene x2 
17) Remove excess xylene from edges of slide and mount with Vectamount 
medium (less is more!), cover with No. 1.5 coverslip  
18) Optional: to speed up drying time, can bake slides for 2-3 h daily until dry 
19) Submit slides for digitalization 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 184 
 
APPENDIX D 
 
Breast Cancer Metastasis Model in NOD-SCID Mice 
This protocol was co-authored with Samantha Schwager 
 
 
Purpose: To observe dynamics of metastasis in vivo. 
 
I. Injection of Cell Suspension into female NOD-SCID mice 
 
Materials: 
● Cells resuspended in Serum Free (SF) DMEM in Eppendorf tube on ice 
● P1000 pipet and sterile tips 
● 1 mL syringe with detachable 26G ½” needle (need fresh needle each mouse) 
● Scale (to record mass) 
● Tweezers  
 
**Note: Remove hair using depilatory cream and ear punch mice ~1-3 days before 
injections 
 
Procedure: 
1. Collect and resuspend cancer cells transduced with GFP/luciferase in SF 
DMEM (performed in cell culture room biosafety cabinet) 
a. Each mouse gets 1 million cancer cells (for MDA-MB-231 cell line; 
other cell lines may need more or less depending on 
proliferation/survival) 
b. For 4 mice, have ~10 million cells (try to make roughly double the 
amount of solution you need based on number of mice since there is 
loss due to dead volume in each needle/syringe) 
 185 
 
i. 2 80-90% confluent T150 flasks per cell line are sufficient 
ii. After trypsinization, spin cells down at 1000 rpm for 5 min 
iii. resuspend in 1 mL SF DMEM 
iv. Dilute as required and count using hemocytometer 
v. Dilute to desired concentration for injections 
c. Want final concentration of cells to be 1 million cells/100 uL SF 
DMEM (or the desired cell concentration).  i.e. 10 million cancer cells 
in 1 mL of SF DMEM 
d. Place final concentration of cells in 1.7 mL eppendorf tube on ice (want 
to minimize time until injections are performed to optimize cell 
viability; immediately begin next steps after cells are prepared) 
2. Set up biosafety cabinet in mouse facility  
a. Prepare isoflurane set up (see isoflurane unit instructions on laminated 
cards attached to unit in ESB012A procedure room) 
b. Get clean cages 
c. Set up paper towel bed for mouse and nosecone 
3. Anesthetize mouse in induction chamber with 1-4% isoflurane 
a. Ensure mice are properly anesthetized with toe pinch and observation 
of ~1 Hz breathing rate 
b. Weigh mice with scale if desired 
4. Prepare first syringe of cells 
a. Mix cells with P1000 (cells sink to the bottom of the Eppendorf tube 
and must be resuspended prior to each injection) 
b. With syringe (needle not attached – minimizes shear on cells), pull up 
to ~0.2 mL of cell suspension 
c. Attach needle, remove bubbles from syringe and adjust volume to 0.1 
mL 
d. Expel extra volume back into Eppendorf tube  
5. Inject mouse 
 186 
 
a. Pull skin by mammary gland taut with tweezers such that injection can 
be performed parallel to the surface of the mouse body into the tissue 
within the tented skin (helps to avoid injecting too deeply which can 
lead to tumor invasion into the peritoneum) 
b. While keeping skin taut with tweezers, enter needle into the skin below 
mammary gland (visualize with mammary papilla – “nipple”) (needle 
bevel up!) and position tip of needle directly under the mammary gland  
c. Lift needle up, expel cell suspension slowly. Bolus should form; pause 
5-10 seconds with needle in place 
d. Remove tweezers, press down slightly with syringe while slowly 
pulling out to avoid leakage 
e. Record time of injection 
6. Place mouse in clean cage to recover 
7. Repeat steps 3-6 until all mice are injected 
a. If injecting different conditions (i.e. subpopulations or 
knockdown/control), it is best practice to alternate which group is 
injected to ensure that there is not an artifact due to difference in time 
in suspension prior to injection 
b. If injecting multiple cages, avoid housing mice based on their 
conditions (best to have both conditions in each cage as opposed to one 
cage for each condition) to avoid artifacts specific to that particular 
batch/cage 
8. Image on IVIS next day to get baseline BLI reading 
 
II. Bioluminescence imaging (BLI) 
 
Preparation of Luciferin (purchased from GoldBio, LUCK-100): 
1. Retrieve 100 mg desiccated powder from -20°C freezer. 
2. Dissolve 1g D-luciferin in 33.33mL Dulbecco’s Phosphate-Buffered Saline 
(DPBS) to obtain concentration of 30 mg/mL. 
3. Aliquot 1 mL volumes into Eppendorf tubes and store in -20°C freezer. 
 187 
 
 
 
1. Place mice into induction chamber and induce anesthesia in mice using 1-3% 
isoflurane gas. 
2. Once mice are adequately anesthetized, administer 100-125 µL by 
intraperitoneal injection using a 26G 1/2” insulin syringe (~5 µL/g mouse body 
weight). 
3. Allow mice to recover and wait 10 minutes for luciferin to react before 
imaging. Place mice back into induction chamber at ~2 min from imaging 
time. 
4. Switch anesthesia to imaging chamber and arrange anesthetized mice in 
chamber up to 3 mice can be imaged at a time) such that they are breathing 
into the nose cones and are in desired position for imaging. 
 
Bioluminescence imaging of mice using the Xenogen Ivis200 system: 
1. Open Living Image program on computer desktop. 
2. Log in using initials. 
3. To check the field of view (options range from A-E), check “Alignment Grid” 
box and green laser grid will appear in chamber.  Make sure the grid 
encompasses the area of interest.  
4. Initialize system by pressing “initialize” button.  You will see a red light on the 
door showing that it is now locked. Camera temperature needs to decrease 
before imaging can commence (box will turn green). 
5. Arrange mice ensuring that pertinent body areas are all adequately exposed and 
are “flat” – limbs may block luminescence signal so gentle adjustments so that 
body signal is not covered is critical to ensuring accurate imaging. If mouse 
body is leaning at an angle and is not “flat” in the prone position, time point 
comparisons will not be accurate. 
 188 
 
6. Press “Acquire” to obtain single image. Adjust exposure time to avoid over-
exposure. Collection of “auto” exposure as well as a range of exposure times 
helps to ensure comparable images at study endpoint. 
 
Note: Exposure time, binning, f-stop, and subject height can be adjusted to obtain 
the optimal image.   
 
III. Tumor Removal Surgery 
 
1. Prepare sterile kit with scalpel, scissors, forceps, tweezers, Q-tips, and sterile 
pad.   
2. Autoclave sterile kit.   
3. Induce anesthesia in mouse using 1-4% isoflurane gas. 
4. Administer 2-5 mg/kg ketoprofen subcutaneously. 
5. Use Nair to remove hair, wipe away Nair with a paper towel damp with 70% 
ethanol. 
6. Sterilize surgical site with chloroprep swabs. 
7. Make “C” shaped incision around periphery of tumor on the medial side by 
tenting skin and cutting with scissors. 
8. Once incision is made, use forceps/tweezers and scalpel to separate tumor from 
surrounding tissue; to avoid local recurrence, skin immediately above tumor 
should also be removed. 
9. Once tumor is isolated from surrounding tissue, use forceps to grasp the tumor 
and remove it using scalpel/scissors to detach any remaining tissue. 
10. Collect tumor for histology by either fixing in 4% paraformaldehyde or 
snapfreezing. 
11. Observe exposed tissue to ensure that all of tumor has successfully been 
removed. If desired, radical mastectomy can be performed by removing entire 
mammary gland in addition to tumor. 
12. Use wound clip applicator to seal wound. 
 189 
 
13. Monitor mice until they regain consciousness. 
14. Every 24 hours for 48 hours, administer 2-5 mg/kg ketoprofen subcutaneously.   
15. Optional: Change water source from nozzle to a water bottle and administer 
antibiotic in water bottle  (sulfamethoxazole-trimethoprim suspension has been 
used with success, 6mL added to 250mL water and shaken to resuspend each 
day). 
16. Remove wound clips 1 week post-operation. 
 
IV. Endpoint tissue collection and processing 
 
1. Prepare 4% paraformaldehyde (PFA) and/or liquid nitrogen/dry ice slurry as 
needed depending on desired downstream application. 
2. Use CO2  to humanely euthanize mice. After CO2 is finished, perform cervical 
dislocation to confirm death. 
3. Mount specimen onto necropsy board and open abdominal cavity. 
4. Prepare 26G ½” needle and syringe with either PFA or PBS to inflate lungs. 
5. Carefully access trachea (ribbed tube structure). Ligate distal end with 
tweezers and insert needle into trachea. Inflate lungs with either PFA or PBS 
then carefully remove lungs. Add lungs to either PFA in 50 mL conical tube or 
to O.C.T. mounting medium or conical tube for snapfreezing. 
6. Remove liver and place in either PFA in a 50 mL conical tube or O.C.T. 
mounting medium or conical tube for snapfreezing. If metastatic liver nodules 
are observed, quantify visually using dissection microscope. 
7. Using large, less expensive scissors (bone will damage good necropsy 
scissors), collect hindlimb, remove tissue including muscle to expose bone. 
Can remove paw to keep both major long bones.  Fix bone in PFA in 50 mL 
conical tube. 
8. Collect any additional tissues. Place snapfrozen samples in -80 freezer and 
fixed samples in refrigerator overnight.   
 190 
 
9. Wash fixed samples in PBS x3 1 hr each on rocker before storing in 70% 
ethanol. 
10. Send samples to Translational Pathology Shared Resource for processing 
including embedding, sectioning, and some IHC stainings. (GFP stain from 
TPSR can be used to quantify metastases)
 191 
  
 
APPENDIX E 
 
 
Circulating tumor cell collection from orthotopic mouse model 
This protocol was adapted from the protocol written by Jocelyn Marshall 
 
 
Purpose: The purpose of this protocol is to isolate circulating tumor cells (CTCs) found 
in the blood stream of mice that have previously been orthotopically injected with 
MDA-MB-231 human breast cancer cells. This protocol is a terminal procedure 
generally done at the endpoint of primary tumor growth (~4wks). This isolation allows 
the cells to be cultured prior to fixation, IF staining, and imaging to assess CTC count 
per mL blood (associated with poor patient prognosis/cancer aggressiveness) and the 
presence of biomarkers. 
 
Materials: 
● 50 mg/mL heparin in DI water 
● 26G ½” needles  
● 1 or 3 mL syringes (personal preference) 
● 2 mL BD vacutainers with sodium heparin 33USP (Cat no. 367671) 
● HBSS (without Mg++/Ca++) 
● Histopaque 1077 (MP Biomedicals, 0219083780) 
● Red Blood Cell Lysis buffer (eBioscience, 00-4300-54) 
● Serum-free DMEM + 5% Penn/Strep 
● Sterile Mattek dishes 
 
Part I. Harvesting Tissues/Blood from Mice  
1. Prepare heparin solution fresh prior to blood collection, obtain dry ice to create 
a slurry for flash-freezing tissues, aliquot sterile PBS into 50 mL conical tubes 
for lungs and keep on ice  
192 
 
2. Work with one mouse at a time to ensure that blood does not clot, connect cage 
with mouse to CO2 at ACUP recommended flow rate and duration 
3. While mouse is being euthanized, prime syringe with 50 uL heparin solution 2-
3x, each time adding to vacutainer, rotate/invert vacutainer to coat glass with 
heparin solution, leave 50 uL in syringe  
4. Once CO2 is shut off, perform cervical dislocation on mouse to ensure it is 
deceased 
5. Immediately bring mouse into biological safety cabinet and perform cardiac 
puncture by coming in with the needle at a 30º angle right below the xiphoid 
process (bottom of rib cage at the midline), maintain slight negative pressure 
with syringe throughout collection, a good stick will see a spike of blood in the 
hub of the needle 
6. Massage rib cage around location of needle to help circulation 
7. Slowly withdraw needle while maintaining slight negative pressure, repeat 
until no more blood can be obtained 
8. If you want to collect the primary tumor and any other tissues, do so at this 
time (do not euthanize the next mouse until you are completely ready; mouse 
blood clots very quickly and yield will be significantly impaired with extended 
wait time prior to blood collection) 
Note: A good collection is ~1-1.5 mL blood/mouse. Downstream processing to 
obtain circulating tumor cells needs a minimum volume of 3mL of blood pooled. 
Blood can be stored at RT if downstream processing is done within 1 hr of 
collection; if not, keep on ice. 
 
Part II. Blood Processing 
1.  Bring blood in vacutainers upstairs to biosafety cabinet and transfer to a 15 
mL conical tube (record total blood volume for each condition) 
2. Dilute blood 1:1 with HBSS (if less than 2.5 mL is obtained, make volume to 5 
mL with HBSS) 
3. In a fresh 15 mL conical tube, add 5 mL histopaque 1077 
193 
 
4. Add blood to histopaque 1077 very carefully down the edge of a 15 mL 
conical tube 
5. Centrifuge at 2000 x g 15 min at RT with acceleration set to 6 and deceleration 
set to 1 (slow!!) 
6. While spinning, prepare RBC lysis buffer 1:100 in DI water (make 10 mL) 
7. After centrifugation, blood should have separated into plasma, buffy coat, 
histopaque, and hematocrit (from top to bottom order) 
8. Remove buffy coat (wispy, transluscent layer on top of histopaque (clear, 
colorless) and below plasma (clear, sometimes pinkish) into a new 15 mL 
conical tube using a Pasteur pipette and bulb, make sure to get all of the buffy 
coat even if excess solution is obtained (take ~1.5-2 mL to be sure) 
9. Dilute with 2x volume of HBSS and centrifuge at 1100 rpm (~200-300 x g) for 
10 min at RT with full acceleration and deceleration 
10. Carefully aspirate supernatant and add 10 mL HBSS and spin down at the 
same conditions as step 8 
11. Carefully aspirate supernatant and then add 1 mL of diluted RBC lysis buffer 
to pellet, mix for 10 “Mississippi-less” seconds then immediately add 10 mL 
HBSS 
12. Resuspend each pellet in ~2 mL serum free DMEM + 5% Penn/Strep; Plate in 
sterile Mattek dishes (can divide up sample into multiple dishes for different 
stainings) 
13. Allow attachment overnight in incubator at 37C, 5% CO2 
14. Fix and stain  
15. Can take images at random locations in the dish to calculate CTC number per 
mL blood and scale using surface area of plate 
 
 
 
 
 
 
 
 
194 
  
 
APPENDIX F 
  
 
Mouse lung decellularization protocol 
Based on protocol from Xiong et al201 
 
 
Objective: This protocol prepares mouse lungs such that cellular material is removed, 
leaving the extracellular matrix components (ECM) in place. The decellularized lungs 
can then be prepared as an ECM scaffold for ex vivo cell culture experiments with 
cancer cells. Xiong et al. showed that metastatic cells can ‘colonize’ the decellularized 
lung scaffolds more readily than non-metastatic cell types.  
 
Materials: 
• FVB or other strain of mouse 
• Surgical scissors 
• 6 mL syringe 
• ½” needle 20G or lower 
• Rinse solution: DI water + 5x penicillin/streptomycin (P/S) 
• 0.1% Triton X-100 + 5x P/S 
• 2% sodium deoxycholate + 1x P/S  
• DNAse solution: 30 mg/mL DNAse in 1.3 mM MgSO4 and 2 mM CaCl2 
• 1 M NaCl + 5x P/S 
• PBS 
• Rain-X 
• Untreated well plate or petri dish 
• Cell culture medium 
 
1. Euthanize mice using CO2 followed by cervical dislocation 
195 
 
2. Affix mouse to dissection board using dissection pins, spray down with ethanol 
3. Using surgical scissors, cut along midline from hindlegs to jaw with slight 
deviance from center past the rib cage to avoid accidentally puncturing the trachea 
4. Once initial cut is made, access lungs from beneath zyphoid process (bottom of 
sternum) by moving GI organs and liver aside, carefully cut diaphragm to visualize 
lungs 
5. From above rib cage, carefully clear away tissue at the neck until trachea is 
visualized (small whitish tube with bandings) 
6. Firmly hold trachea as proximal to skull as possible with tweezers and insert 
needle with syringe filled with rinse solution next to the tweezers proximal to the 
lungs 
7. Inflate lungs by injecting ~3mL of rinse solution via the trachea 
8. Excise lungs with trachea and heart attached and place in sterile petri dish filled 
with rinse solution 
9. Incubate lungs in rinse solution for 1 hr at 4°C 
10. Inject via the trachea 3-5 times with rinse solution to wash 
11. Inject via the trachea with 0.1% Triton X-100 + 1x P/S solution and incubate for 8 
hr at room temperature (RT) to lyse cells 
12. Inject via the trachea 3-5 times with rinse solution to wash 
13. Inject via the trachea with 2% sodium deoxycholate + 1x P/S solution and incubate 
overnight at 4°C  
14. At some point prior to seeding, take 6 well plate or petri dish and coat with Rain-X 
using Q-tip 
15. Let dry 15 min and then thoroughly buff off remaining Rain-X with Kimwipe/Q-
tip 
16. Sterilize dish with UV for 1 hr; set aside until lung scaffolds are ready for seeding 
17. Inject via the trachea 3-5 times with rinse solution to wash 
18. Inject via the trachea with 1 M NaCl + 5x P/S and incubate for 1 hr at RT to lyse 
residual nuclei 
19. Inject via the trachea 3-5 times with rinse solution to wash 
196 
 
20. Inject via the trachea with DNAse solution and incubate for 1 hr at RT to remove 
DNA 
21. Inject via the trachea 3-5 times with PBS + 5x P/S to wash 
22. Cut decellularized lungs with sterile surgical scalpel into 5 mm3 cubes and place in 
sterile 6 well plate or petri dish 
23. Seed desired cells (suggested seeding density for MDA-MB-231 or MCF7 cells is 
0.2x106 cells/lung/well) and culture at 37°C 5% CO2 for 2 hr to allow cell 
attachment 
24. Gently aspirate media and add fresh media to remove unattached cells 
25. Incubate at 37°C 5% CO2 for up to 9 days, feeding cells every 2 days by refreshing 
media 
26. Scaffolds and cells can be imaged with confocal reflectance microscopy on 
LSM700/800 
197 
 
APPENDIX G 
 
 
Tumor Spheroid formation and embedding in 3D collagen matrix 
This protocol was adapted from the ones written by Drs. Shawn Carey and Jian Zhang 
 
 
General tips: 
• Coordinate growth of cells 
• Can label cells with CellTracker dye (maintain unlabeled flasks for further 
propagation) or use cells transduced with fluorescent marker 
• Methocult is only necessary for cells that do not readily form spheroids  
 
Optional: Live cell labeling 
CellTracker Green CMFDA (Thermo Fisher) 
Prepare 10 mM stock (500 µg lyophilized powder in 90 µL DMSO); store at -
20°C for ~1 month 
CellTracker Orange CMRA (Thermo Fisher)Prepare 10 mM stock (50 µg lyophilized 
powder in 9 µL DMSO); store at -20°C for ~1 month 
1. Remove media from T-25 flask by aspiration 
2. Rinse 1X with sterile PBS 
3. Add 1 mL working solution (10 µM CMFDA or CMRA in PBS) 
4. Incubate 10 min in incubator on orbital shaker 
5. Add 10 mL of complete media and spin down cells 
6. Resuspend pellet and continue with protocol 
 
Optional: Addition of Matrigel to improve spheroid formation 
  Matrigel (“MG”) [BD 356234; Lot 15900]  
1. Thaw MG at 4°C overnight on ice 
2. Freeze 20-200µL sterile pipette tips and centrifuge tubes at -20°C 
198 
 
3. On ice, aliquot into appropriate volumes 
    For 1% MG: 144 µL MG per 14.4 mL complete media (enough for 150 µL 
x 96 wells) 
4. Quickly transfer aliquots to cold -20°C freezer box; store at -20°C 
5. Immediately before use, remove from -20°C, thaw on ice, dilute in cold media 
on ice 
Note: Avoid multiple freeze-thaws 
 
Methocult preparation: 
  Methocult (“MC”) [H4100] – ordered from Stem Cell 
1. Thaw MC overnight at 4°C 
2. Add 98 mL of complete culture media to MC (2.6%) and shake vigorously. 
Solution is now 0.75%. Allow several days for bubbles to settle out of solution. 
Store at 4°C. 
 
I. Spheroid Formation 
 
1. Remove media from T-25 flask of cultured cells 
2. Rinse 1X with sterile PBS 
3. Trypsinize cells at 37°C (do not over-trypsinize) 
4. Neutralize trypsin with 3-5 mL media; transfer to 15 mL tube 
5. Centrifuge cells at 1100 rpm for 5 min 
6. Aspirate supernatant and resuspend cells in 1 mL culture media; count cells 
7. Mix MC by swirling bottle and pipetting up and down (do not introduce 
bubbles) 
8. Make media / 0.25% MC by mixing 1 part MC with 2 parts media: 
 
per 96-well plate (200 µL/well): 
0.75% MC  7.33 mL 
Complete media 14.66 mL 
199 
 
      22 mL media / 0.25% MC 
 
9. Add desired number of cells to media/0.25% MC; cell concentration should be 
desired number of cells per 200 µL (i.e. for 10,000-cell spheroids, 10,000 
cells/200 µL) 
Note: If volume of cell suspension to add is more than 500 µL, repellet and 
suspend in media / 0.25%MC 
10. Pour cell suspension into sterile reagent reservoir 
11. Use multichannel pipette to add 200 µL suspension to each well of a non-
treated, round bottom, 96-well plate 
12. Centrifuge at 1100rpm, 5 min at room temp 
13. Optional (only needed if adding Matrigel): Incubate at 37°C, 5% CO2 for 2 
hours on orbital shaker (~60rpm) 
14. Remove from shaker and incubate at 37°C, 5% CO2 
15. Optional (if adding Matrigel): immediately replace culture media with 
complete MCF10A media / 0.25% MC / 1% MG  
16. Optional: Can feed spheroids every 2-3 days by removing and replacing 150-
175 µL media. 
 
II. Spheroid embedding 
 
1. Prepare spheroids– allow spheroids to compact over 2-3 days 
2. Clean and UV glass bottom 24 well plates any time before embedding (can 
also activate if gel detachment is an issue) 
3. On the last day of spheroid compaction, select “good” spheroids (well 
compacted, mostly round, no satellite mini-spheroids, no debris) using tissue 
culture microscope – mark/circle wells with lab marker to allow easy 
collection later 
4. Spray down biosafety cabinet with 70% EtOH and bring in the following 
items: 
200 
 
• ice bucket with 10mg/mL collagen in 0.1% acetic acid stock solution 
and 1N NaOH (4C) 
• DMEM complete medium (RT) 
• empty 15 mL conical tubes for each condition plus some extra for 
backfill and just in case 
• wide bore 1000uL and 200uL pipette tips 
• UVed glass bottom well plates 
• P1000, P200, and P20 pipettors 
5. Using P200 pipettor set to ~70uL with wide bore tip, collect 10 good spheroids 
and deposit droplet containing spheroid on lid of 96 well plate in the respective 
position – spheroid should be visible as tiny white “snowflake” dot in pipette 
tip upon aspiration 
6. After 10 spheroids are collected, prepare collagen gel mixture: 
• Add using either P1000 or P200 with wide bore tip, add 150uL 
collagen stock to 15 mL conical tube 
• Using P1000, collect 850uL DMEM complete then release ~100-150uL 
back into aliquot 
• Using same pipettor with DMEM, collect each spheroid (do not need 
all the media in the droplet just the spheroid) 
•  If pipettor reaches 850 before all spheroids are collected, expel and 
collect spheroids on lid until volume is correct and all spheroids are in 
the pipette tip 
• If all spheroids are collected and pipettor still hasn’t reached 850uL, 
collect remaining volume needed from complete medium aliquot 
• Add 850uL DMEM + 10 spheroids to 150uL collagen in the conical 
tube 
• Mix gently but thoroughly, mixture should change from pink + clear to 
a uniform orange-ish yellow  
• Using P20, add 3uL 1N NaOH to neutralize mixture and mix until 
color returns to pinkish 
201 
 
7. Using P200 set to 40uL with a wide bore tip, collect collagen mixture + 1-2 
spheroids and pipette into center of glass bottom well plate 
8. Can quickly check to see if spheroid is positioned at edge of droplet – if so can 
quickly manipulate position with pipette tip or give up and aspirate out droplet 
(droplet should also be manipulated/aspirated if 2 or more spheroids within 
close proximity are observed) 
9. If this assessment is very fast (less than ~30 sec) can add an additional droplet 
with 1-2 spheroids in a neighboring well 
10. Immediately put well plate into incubator and start stopwatch 
11. For polymerization, the following flipping times can be used and modified if 
initial runs show spheroids on glass or top of droplet: 
• 30 sec - flip to invert well plate (at 30 sec) 
• 30 sec – flip well plate back to upright (at 60 sec) 
• 30 sec – flip to invert well plate (at 1:30) 
• 30 sec – flip back to upright (at 2:00) 
• 45 sec – flip to invert (at 2:45) 
• 45 sec – flip back to upright (at 3:30) 
• 45 sec – flip to invert (at 4:15) 
• 45 sec – flip back to upright (at 5:00) 
• 60 sec – flip to invert (at 6:00) 
• 60 sec – flip back to upright (at 7:00) 
• 60 sec – flip to invert (at 8:00) 
• After 60 sec, flip back to upright and check spheroid position(s) under 
microscope 
12. Once up to 8 inside wells of 24 well plate is full, back fill with 500uL of 
1.5mg/mL collagen per well and put in incubator to polymerize for 45 min – 1 
h 
13. Hydrate gels with 500uL DMEM complete 
202 
 
14. Image using LSM800 at 0h, 24h, 48h, or accordingly to your experimental 
goals 
 
 
 
 
203 
 
REFERENCES 
1. Hapach, L. A., Mosier, J. A., Wang, W. & Reinhart-King, C. A. Engineered 
models to parse apart the metastatic cascade. npj Precis. Oncol. 3, (2019). 
2. Bray, F. et al. Global cancer statistics 2018: GLOBOCAN estimates of 
incidence and mortality worldwide for 36 cancers in 185 countries. CA. Cancer 
J. Clin. 68, 394–424 (2018). 
3. Fitzmaurice, C. et al. The Global Burden of Cancer 2013. JAMA Oncol. 1, 505–
527 (2015). 
4. Murray, C. J. L., Lopez, A. D., Organization, W. H., Bank, W. & of Public 
Health, H. S. The Global burden of disease : a comprehensive assessment of 
mortality and disability from diseases, injuries, and risk factors in 1990 and 
projected to 2020 : summary / edited by Christopher J. L. Murray, Alan D. 
Lopez. Published by the Harvard School of Public Health o (1996). 
5. Harbeck, N. et al. Breast cancer. Nature Reviews Disease Primers 5, (2019). 
6. Hiatt, R. A. & Brody, J. G. Environmental Determinants of Breast Cancer. 
Annu. Rev. Public Health 39, 113–133 (2018). 
7. Li, N. et al. Global burden of breast cancer and attributable risk factors in 195 
countries and territories, from 1990 to 2017: Results from the Global Burden of 
Disease Study 2017. J. Hematol. Oncol. 12, 1–12 (2019). 
8. Hanahan, D. & Weinberg, R. A. The hallmarks of cancer. Cell 100, 57–70 
(2000). 
9. Cserni, G. Histological type and typing of breast carcinomas and the WHO 
classification changes over time. Pathologica 112, 25–41 (2020). 
10. Perou, C. M. et al. Molecular portraits of human breast tumours. letters to 
nature 748. Nature 533, 747–752 (2000). 
11. Ades, F., Tryfonidis, K. & Zardavas, D. The past and future of breast cancer 
treatment - From the papyrus to individualised treatment approaches. 
Ecancermedicalscience 11, 1–11 (2017). 
12. Mehanna, J., Haddad, F. G. H., Eid, R., Lambertini, M. & Kourie, H. R. Triple-
negative breast cancer: Current perspective on the evolving therapeutic 
landscape. Int. J. Womens. Health 11, 431–437 (2019). 
13. Nedeljkovic, M. & Damjanovic, A. Mechanisms of Chemotherapy Resistance 
in Triple-Negative Breast Cancer - How We Can Rise to the Challenge. Cells 8, 
204 
 
(2019). 
14. Morris, G. J. et al. Differences in breast carcinoma characteristics in newly 
diagnosed African-American and Caucasian patients: A single-institution 
compilation compared with the national cancer institute’s surveillance, 
epidemiology, and end results database. Cancer 110, 876–884 (2007). 
15. Yin, L., Duan, J. J., Bian, X. W. & Yu, S. C. Triple-negative breast cancer 
molecular subtyping and treatment progress. Breast Cancer Res. 22, 1–13 
(2020). 
16. Li, X. et al. Triple-negative breast cancer has worse overall survival and cause-
specific survival than non-triple-negative breast cancer. Breast Cancer Res. 
Treat. 161, 279–287 (2017). 
17. Lehmann, B. D. et al. Identification of human triple-negative breast cancer 
subtypes and preclinical models for selection of targeted therapies. J. Clin. 
Invest. 121, 2750–2767 (2011). 
18. Lehmann, B. D. et al. Refinement of triple-negative breast cancer molecular 
subtypes: Implications for neoadjuvant chemotherapy selection. PLoS One 11, 
1–22 (2016). 
19. Kim, J. et al. Genomic characteristics of triple-negative breast cancer nominate 
molecular subtypes that predict chemotherapy response. Mol. Cancer Res. 18, 
253–263 (2020). 
20. Ding, Y. C. et al. Molecular subtypes of triple-negative breast cancer in women 
of different race and ethnicity. Oncotarget 10, 198–208 (2019). 
21. Liu, Y. R. et al. Comprehensive transcriptome analysis identifies novel 
molecular subtypes and subtype-specific RNAs of triple-negative breast cancer. 
Breast Cancer Res. 18, 1–10 (2016). 
22. Kim, C. et al. Chemoresistance Evolution in Triple-Negative Breast Cancer 
Delineated by Single-Cell Sequencing. Cell 173, 879-893.e13 (2018). 
23. Li, X. & Hou, J. A richer understanding of intratumoral heterogeneity: Single-
cell genomics put it within reach. J. Thorac. Dis. 10, 1178–1182 (2018). 
24. Huxley, J. The Biological Aspects of Cancer. (Harcourt, Brace, 1958). 
25. McGranahan, N. & Swanton, C. Clonal Heterogeneity and Tumor Evolution: 
Past, Present, and the Future. Cell 168, 613–628 (2017). 
26. Nowell, P. C. The clonal evolution of tumor cell populations. Science (80-. ). 
194, 23–28 (1976). 
205 
 
27. Chaffer, C. L. et al. Normal and neoplastic nonstem cells can spontaneously 
convert to a stem-like state. Proc. Natl. Acad. Sci. U. S. A. 108, 7950–7955 
(2011). 
28. Zardavas, D., Irrthum, A., Swanton, C. & Piccart, M. Clinical management of 
breast cancer heterogeneity. Nat. Rev. Clin. Oncol. 12, 381–394 (2015). 
29. Nicolson, G. L. Generation of phenotypic diversity and progression in 
metastatic tumor cells. Cancer Metastasis Rev. 3, 25–42 (1984). 
30. Bissell, M. J., Hall, H. G. & Parry, G. How does the extracellular matrix direct 
gene expression? J. Theor. Biol. 99, 31–68 (1982). 
31. Paszek, M. J. et al. Tensional homeostasis and the malignant phenotype. 
Cancer Cell 8, 241–254 (2005). 
32. Kraning-Rush, C. M., Califano, J. P. & Reinhart-King, C. A. Cellular traction 
stresses increase with increasing metastatic potential. PLoS One 7, (2012). 
33. Zaman, M. H. et al. Migration of tumor cells in 3D matrices is governed by 
matrix stiffness along with cell-matrix adhesion and proteolysis. Proc. Natl. 
Acad. Sci. U. S. A. 103, 10889–10894 (2006). 
34. Tilghman, R. W. et al. Matrix rigidity regulates cancer cell growth and cellular 
phenotype. PLoS One 5, 1–13 (2010). 
35. Levental, K. R. et al. Matrix Crosslinking Forces Tumor Progression by 
Enhancing Integrin Signaling. Cell 139, 891–906 (2009). 
36. Provenzano, P. P., Inman, D. R., Eliceiri, K. W. & Patricia, J. Matrix density-
induced mechanoregulation of breast cell phenotype, signaling, and gene 
expression through a FAK-ERK linkage. Matrix 28, 4326–4343 (2010). 
37. Berger, A. J. et al. Scaffold stiffness influences breast cancer cell invasion via 
EGFR-linked Mena upregulation and matrix remodeling. Matrix Biol. 85–86, 
80–93 (2020). 
38. Pfeifer, C. R., Alvey, C. M., Irianto, J. & Discher, D. E. Genome variation 
across cancers scales with tissue stiffness – an invasion-mutation mechanism 
and implications for immune cell infiltration. Curr Opin Syst Biol. 176, 103–
114 (2017). 
39. Miller, B. E., Miller, F. R. & Heppner, G. H. Interactions between Tumor 
Subpopulations Affecting Their Sensitivity to the Antineoplastic Agents 
Cyclophosphamide and Methotrexate. Cancer Res. 41, 4378–4381 (1981). 
40. Miller, B. E., Machemer, T., Lehotan, M. & Heppner, G. H. Tumor 
206 
 
Subpopulation Interactions Affecting Melphalan Sensitivity in Palpable Mouse 
Mammary Tumors. Cancer Res. 51, 4378–4387 (1991). 
41. Tabassum, D. P. & Polyak, K. Tumorigenesis: It takes a village. Nat. Rev. 
Cancer 15, 473–483 (2015). 
42. Cleary, A. S., Leonard, T. L., Gestl, S. A. & Gunther, E. J. Tumour cell 
heterogeneity maintained by cooperating subclones in Wnt-driven mammary 
cancers. Nature 508, 113–117 (2014). 
43. Hobor, S. et al. TGFα and amphiregulin paracrine network promotes resistance 
to EGFR blockade in colorectal cancer cells. Clin. Cancer Res. 20, 6429–6438 
(2014). 
44. Chapman, A. et al. Heterogeneous tumor subpopulations cooperate to drive 
invasion. Cell Rep. 8, 688–695 (2014). 
45. Celià-Terrassa, T. et al. Epithelial-mesenchymal transition can suppress major 
attributes of human epithelial tumor-initiating cells. J. Clin. Invest. 122, 1849–
1868 (2012). 
46. Wagenblast, E. et al. A model of breast cancer heterogeneity reveals vascular 
mimicry as a driver of metastasis. Nature 520, 358–362 (2015). 
47. Mateo, F. et al. SPARC mediates metastatic cooperation between CSC and non-
CSC prostate cancer cell subpopulations. Mol. Cancer 13, 1–17 (2014). 
48. Marusyk, A., Janiszewska, M. & Polyak, K. Intratumor Heterogeneity: The 
Rosetta Stone of Therapy Resistance. Cancer Cell 37, 471–484 (2020). 
49. Zhang, M. et al. Intratumoral heterogeneity in a Trp53-null mouse model of 
human breast cancer. Cancer Discov. 5, 521–533 (2015). 
50. Mann, K. M. et al. Analyzing tumor heterogeneity and driver genes in single 
myeloid leukemia cells with SBCapSeq. 34, 962–972 (2016). 
51. Niknafs, N. et al. Characterization of genetic subclonal evolution in pancreatic 
cancer mouse models. Nat. Commun. 10, 1–10 (2019). 
52. El Hilali, N., Rubio, N., Martinez-Villacampa, M. & Blanco, J. Combined 
noninvasive imaging and luminometric quantification of luciferase-labeled 
human prostate tumors and metastases. Lab. Investig. 82, 1563–1571 (2002). 
53. Fidler, I. J. & Kripke, M. L. Metastasis results from preexisting variant cells 
within a malignant tumor. Science (80-. ). 197, 893–895 (1977). 
54. Kripke, M. L., Gruys, E. & Fidler, I. J. Metastatic Heterogeneity of Cells from 
207 
 
an Ultraviolet Light-induced Murine Fibrosarcoma of Recent Origin. Cancer 
Res. 38, 2962–2967 (1978). 
55. Fischer, K. R. et al. Epithelial-to-mesenchymal transition is not required for 
lung metastasis but contributes to chemoresistance. Nature 527, 472–476 
(2015). 
56. Konen, J. et al. Image-guided genomics of phenotypically heterogeneous 
populations reveals vascular signalling during symbiotic collective cancer 
invasion. Nat. Commun. 8, 1–15 (2017). 
57. Commander, R. et al. Subpopulation targeting of pyruvate dehydrogenase and 
GLUT1 decouples metabolic heterogeneity during collective cancer cell 
invasion. Nat. Commun. 11, 1–17 (2020). 
58. Summerbell, E. R. et al. Epigenetically heterogeneous tumor cells direct 
collective invasion through filopodia-driven fibronectin micropatterning. Sci. 
Adv. 6, 1–15 (2020). 
59. Tamura, M. et al. Morphology-based optical separation of subpopulations from 
a heterogeneous murine breast cancer cell line. PLoS One 12, 1–13 (2017). 
60. Hasle, N. et al. High-throughput, microscope-based sorting to dissect cellular 
heterogeneity. Mol. Syst. Biol. 16, (2020). 
61. Fuhrmann, A., Banisadr, A., Beri, P., Tlsty, T. D. & Engler, A. J. Metastatic 
State of Cancer Cells May Be Indicated by Adhesion Strength. Biophys. J. 112, 
736–745 (2017). 
62. Beri, P. et al. Cell adhesiveness serves as a biophysical marker for metastatic 
potential. Cancer Res. 80, 901–911 (2020). 
63. Islam, M. et al. Microfluidic cell sorting by stiffness to examine heterogenic 
responses of cancer cells to chemotherapy. Cell Death Dis. 9, (2018). 
64. Masaeli, M. et al. Multiparameter mechanical and morphometric screening of 
cells. Sci. Rep. 6, 1–11 (2016). 
65. Mu, L. et al. Small-sized colorectal cancer cells harbor metastatic 
tumorinitiating cells. Oncotarget 8, 107907–107919 (2017). 
66. Jill M. Westcott et al. An epigenetically distinct breast cancer cell 
subpopulation promotes collective invasion. 125, 1927–1943 (2015). 
67. Shah, A. T., Cannon, T. M., Higginbotham, J. N., Coffey, R. J. & Skala, M. C. 
Autofluorescence flow sorting of breast cancer cell metabolism. J. Biophotonics 
10, 1026–1033 (2017). 
208 
 
68. Fiorillo, M., Sotgia, F. & Lisanti, M. P. ‘Energetic’ cancer stem cells (e-CSCs): 
A new hyper-metabolic and proliferative tumor cell phenotype, driven by 
mitochondrial energy. Front. Oncol. 9, 1–15 (2019). 
69. Wyckoff, J. B., Segall, J. E. & Condeelis, J. S. The collection of the motile 
population of cells from a living tumor. Cancer Res. 60, 5401–5404 (2000). 
70. Wang, W. et al. Gene expression analysis on small numbers of invasive cells 
collected by chemotaxis from primary mammary tumors of the mouse. BMC 
Biotechnol. 3, 1–12 (2003). 
71. Hernandez, L., Smirnova, T., Wyckoff, J., Condeelis, J. & Segall, J. E. In Vivo 
Assay for Tumor Cell Invasion. Methods Mol Biol. 571, 227–238 (2009). 
72. Patsialou, A. et al. Selective gene-expression profiling of migratory tumor cells 
in vivo predicts clinical outcome in breast cancer patients. Breast Cancer Res. 
14, R139 (2012). 
73. Kozlowski, J. M. et al. Metastatic Behavior off Human Tumor Cell Lines 
Grown in the Nude Mouse. Cancer Res. 44, 3522–3529 (1984). 
74. Santner, S. J. et al. Malignant MCF10CA1 cell lines derived from premalignant 
human breast epithelial MCF10AT cells. Breast Cancer Res. Treat. 65, 101–
110 (2001). 
75. Kang, Y. et al. A multigenic program mediating breast cancer metastasis to 
bone. 3, 537–549 (2003). 
76. Minn, A. J. et al. Distinct organ-specific metastatic potential of individual 
breast cancer cells and primary tumors. J.Clin.Invest. 115, 44–55 (2005). 
77. Minn, A. J. et al. Genes that mediate breast cancer metastasis to lung. 436, 518–
524 (2005). 
78. Bos, P. D. et al. Genes that mediate breast cancer metastasis to the brain. 
Nature 459, 1005–1009 (2009). 
79. Cortazar, P. et al. Pathological complete response and long-term clinical benefit 
in breast cancer: The CTNeoBC pooled analysis. Lancet 384, 164–172 (2014). 
80. Waks, A. G. & Winer, E. P. Breast Cancer Treatment: A Review. JAMA - J. 
Am. Med. Assoc. 321, 288–300 (2019). 
81. Guestini, F., McNamara, K. M. & Sasano, H. The use of chemosensitizers to 
enhance the response to conventional therapy in triple-negative breast cancer 
patients. Breast Cancer Manag. 6, 127–131 (2017). 
209 
 
82. Wang, J. C. et al. Metformin inhibits metastatic breast cancer progression and 
improves chemosensitivity by inducing vessel normalization via PDGF-B 
downregulation. J. Exp. Clin. Cancer Res. 38, 1–17 (2019). 
83. Strekalova, E., Malin, D., Rajanala, H. & Cryns, V. L. Metformin sensitizes 
triple-negative breast cancer to proapoptotic TRAIL receptor agonists by 
suppressing XIAP expression. Breast Cancer Res. Treat. 163, 435–447 (2017). 
84. Kam, Y., Das, T., Minton, S. & Gatenby, R. A. Evolutionary strategy for 
systemic therapy of metastatic breast cancer: Balancing response with 
suppression of resistance. Women’s Heal. 10, 423–430 (2014). 
85. Norton, L. & Simon, R. Tumor size, sensitivity to therapy, and design of 
treatment schedules. Cancer Treat. Rep. 61, 1307—1317 (1977). 
86. West, J. et al. Towards multidrug adaptive therapy. Cancer Res. 80, 1578–1589 
(2020). 
87. Gatenby, R. A., Silva, A. S., Gillies, R. J. & Frieden, B. R. Adaptive Therapy. 
Cancer Res. 69, 4894–4903 (2009). 
88. West, R. B. et al. Determination of stromal signatures in breast carcinoma. 
PLoS Biol. 3, 1101–1110 (2005). 
89. Bergamaschi, A. et al. Extracellular matrix signature identifies breast cancer 
subgroups with different clinical outcome. J Pathol. 214, 357–367 (2008). 
90. Medina, S. H. et al. Identification of a mechanogenetic link between substrate 
stiffness and chemotherapeutic response in breast cancer. Biomaterials 202, 1–
11 (2019). 
91. Kuczek, D. E. et al. Collagen density regulates the activity of tumor-infiltrating 
T cells. J. Immunother. Cancer 7, 1–15 (2019). 
92. Finak, G. et al. Stromal gene expression predicts clinical outcome in breast 
cancer. Nat. Med. 14, 518–527 (2008). 
93. Schmid, P. et al. Atezolizumab and Nab-Paclitaxel in Advanced Triple-
Negative Breast Cancer. N. Engl. J. Med. 379, 2108–2121 (2018). 
94. Schmid, P. et al. Atezolizumab plus nab-paclitaxel as first-line treatment for 
unresectable, locally advanced or metastatic triple-negative breast cancer 
(IMpassion130): updated efficacy results from a randomised, double-blind, 
placebo-controlled, phase 3 trial. Lancet Oncol. 21, 44–59 (2020). 
95. Cyprian, F. S., Akhtar, S., Gatalica, Z. & Vranic, S. Targeted immunotherapy 
with a checkpoint inhibitor in combination with chemotherapy: A new clinical 
210 
 
paradigm in the treatment of triple-negative breast cancer. Bosn. J. Basic Med. 
Sci. 19, 227–233 (2019). 
96. Massagué, J. & Obenauf, A. C. Metastatic colonization by circulating tumour 
cells. Nature (2016). doi:10.1038/nature17038 
97. Bulfoni, M. et al. Dissecting the heterogeneity of circulating tumor cells in 
metastatic breast cancer: Going far beyond the needle in the Haystack. Int. J. 
Mol. Sci. 17, 1–25 (2016). 
98. Riethdorf, S. et al. Detection of circulating tumor cells in peripheral blood of 
patients with metastatic breast cancer: A validation study of the cell search 
system. Clin. Cancer Res. 13, 920–928 (2007). 
99. Zhang, L. et al. Meta-analysis of the prognostic value of circulating tumor cells 
in breast cancer. Clin. Cancer Res. 18, 5701–5710 (2012). 
100. Menyailo, M. E., Tretyakova, M. S. & Denisov, E. V. Heterogeneity of 
circulating tumor cells in breast cancer: Identifying metastatic seeds. Int. J. Mol. 
Sci. 21, (2020). 
101. Zhang, X., Meng, X., Leng, S. X. & Zhang, H. The Biology of Aging and 
Cancer The Biology of Aging. Cancer J. 23, 201–205 (2017). 
102. Ershler, W. B. & Longo, D. L. Aging and cancer: Issues of basic and clinical 
science. J. Natl. Cancer Inst. 89, 1489–1497 (1997). 
103. Kohn, J. C., Lampi, M. C. & Reinhart-King, C. A. Age-related vascular 
stiffening: Causes and consequences. Front. Genet. 6, 1–17 (2015). 
104. Davalli, P., Mitic, T., Caporali, A., Lauriola, A. & D’Arca, D. ROS, Cell 
Senescence, and Novel Molecular Mechanisms in Aging and Age-Related 
Diseases. Oxid. Med. Cell. Longev. 2016, (2016). 
105. Lampi, M. C. & Reinhart-King, C. A. Targeting extracellular matrix stiffness to 
attenuate disease: From molecular mechanisms to clinical trials. Sci. Transl. 
Med. 10, 1–15 (2018). 
106. Liu, B., McNally, S., Kilpatrick, J. I., Jarvis, S. P. & O’Brien, C. J. Aging and 
ocular tissue stiffness in glaucoma. Surv. Ophthalmol. 63, 56–74 (2018). 
107. Rojas, A., Añazco, C., González, I. & Araya, P. Extracellular matrix glycation 
and receptor for advanced glycation end-products activation: A missing piece in 
the puzzle of the association between diabetes and cancer. Carcinogenesis 39, 
515–521 (2018). 
108. Balic, M. et al. Prevalence of comorbidity in cancer patients scheduled for 
211 
 
systemic anticancer treatment in Austria: A cross-sectional multicentre 
observational study. Memo - Mag. Eur. Med. Oncol. 12, 290–296 (2019). 
109. Sarfati, D., Koczwara, B. & Jackson, C. The impact of comorbidity on cancer 
and its treatment. CA. Cancer J. Clin. 66, 337–350 (2016). 
110. Van De Poll-Franse, L. V. et al. Less aggressive treatment and worse overall 
survival in cancer patients with diabetes: A large population based analysis. Int. 
J. Cancer 120, 1986–1992 (2007). 
111. Fu, M. R. et al. Comorbidities and quality of life among breast cancer 
survivors: A prospective study. J. Pers. Med. 5, 229–242 (2015). 
112. Parise, C. A. & Caggiano, V. The influence of comorbidity on treatment and 
survival of triple-negative breast cancer. Breast J. 26, 1729–1735 (2020). 
113. Swede, H. et al. Mortality Risk from Co-Morbidities independent of Triple- 
Negative Breast Cancer Status: NCI SEER-based Cohort Analysis. Cancer 
Causes Control 27, 627–636 (2016). 
114. Diagnosis and classification of diabetes mellitus. Diabetes Care 36, 67–74 
(2013). 
115. DeFronzo, R. A. et al. Type 2 diabetes mellitus. Nat. Rev. Dis. Prim. 1, 1–23 
(2015). 
116. Maynard, G. D. A statistical study in cancer death-rates. Biometrika 7, 276–304 
(1910). 
117. Pearson, K., Lee, A. & Elderton, E. M. On the Correlation of Death-Rates. 73, 
534–539 (1910). 
118. Shikata, K., Ninomiya, T. & Kiyohara, Y. Diabetes mellitus and cancer risk: 
Review of the epidemiological evidence. Cancer Sci. 104, 9–14 (2013). 
119. Larsson, S. C., Mantzoros, C. S. & Wolk, A. Diabetes mellitus and risk of 
breast cancer: A meta-analysis. Int. J. Cancer 121, 856–862 (2007). 
120. Boyle, P. et al. Diabetes and breast cancer risk: A meta-analysis. Br. J. Cancer 
107, 1608–1617 (2012). 
121. Fleming, S. T., Pursley, H. G., Newman, B., Pavlov, D. & Chen, K. 
Comorbidity as a predictor of stage of illness for patients with breast cancer. 
Med. Care 43, 132–140 (2005). 
122. Srokowski, T. P., Fang, S., Hortobagyi, G. N. & Giordano, S. H. Impact of 
diabetes mellitus on complications and outcomes of adjuvant chemotherapy in 
212 
 
older patients with breast cancer. J. Clin. Oncol. 27, 2170–2176 (2009). 
123. Zhao, X. B. & Ren, G. S. Diabetes mellitus and prognosis in women with breast 
cancer: A systematic review and meta-analysis. Med. (United States) 95, e5602 
(2016). 
124. Bronsveld, H. K. et al. Diabetes and Breast Cancer Subtypes. PLoS One 12, 1–
16 (2017). 
125. Kang, C., LeRoith, D. & Gallagher, E. J. Diabetes, obesity, and breast cancer. 
Endocrinology 159, 3801–3812 (2018). 
126. Cerami, A., Vlassara, H. & Brownlee, M. Glucose and Aging. Sci. Am. 256, 
90–96 (1987). 
127. Singh, V. P., Bali, A., Singh, N. & Jaggi, A. S. Advanced glycation end 
products and diabetic complications. Korean J. Physiol. Pharmacol. 18, 1–14 
(2014). 
128. Volpe, C. M. O., Villar-Delfino, P. H., Dos Anjos, P. M. F. & Nogueira-
Machado, J. A. Cellular death, reactive oxygen species (ROS) and diabetic 
complications review-Article. Cell Death Dis. 9, (2018). 
129. Cox, T. R. & Erler, J. T. Remodeling and homeostasis of the extracellular 
matrix: Implications for fibrotic diseases and cancer. DMM Dis. Model. Mech. 
4, 165–178 (2011). 
130. Duran-Jimenez, B. et al. Advanced glycation end products in extracellular 
matrix proteins contribute to the failure of sensory nerve regeneration in 
diabetes. Diabetes 58, 2893–2903 (2009). 
131. Katayama, Y., Akatsu, T., Yamamoto, M., Kugai, N. & Nagata, N. Role of 
nonenzymatic glycosylation of type I collagen in diabetic osteopenia. J. Bone 
Miner. Res. 11, 931–937 (1996). 
132. Kaji, Y. et al. Advanced glycation end products in diabetic corneas. Investig. 
Ophthalmol. Vis. Sci. 41, 362–368 (2000). 
133. Makita, Z. et al. Advanced Glycosylation End Products in Patients with 
Diabetic Nephropathy. N. Engl. J. Med. 325, 836–842 (1991). 
134. Lega, I. C. et al. The impact of diabetes on breast cancer treatments and 
outcomes: A population-based study. Diabetes Care 41, 755–761 (2018). 
135. Booth, A. A., Khalifah, R. G., Todd, P. & Hudson, B. G. In vitro kinetic studies 
of formation of antigenic advanced glycation end products (AGEs). Novel 
inhibition of post-Amadori glycation pathways. J. Biol. Chem. 272, 5430–5437 
213 
 
(1997). 
136. Christofori, G. New signals from the invasive front. Nature 441, 444–450 
(2006). 
137. C.F., C. et al. Progression from ductal carcinoma in situ to invasive breast 
cancer: Revisited. Mol. Oncol. 7, 859–869 (2013). 
138. Sherwood, D. R. Cell invasion through basement membranes: an anchor of 
understanding. Trends Cell Biol. 16, 250–256 (2006). 
139. Winkler, J., Abisoye-Ogunniyan, A., Metcalf, K. J. & Werb, Z. Concepts of 
extracellular matrix remodelling in tumour progression and metastasis. Nat. 
Commun. 11, 1–19 (2020). 
140. McCormack, V. A. & Dos Santos Silva, I. Breast density and parenchymal 
patterns as markers of breast cancer risk: A meta-analysis. Cancer Epidemiol. 
Biomarkers Prev. 15, 1159–1169 (2006). 
141. Kerlikowske, K. et al. Automated and Clinical Breast Imaging Reporting and 
Data System Density Measures Predict Risk of Screen-Detected and Interval 
Cancers. Ann Intern Med. 168, 757–765 (2018). 
142. Conklin, M. W. et al. Aligned collagen is a prognostic signature for survival in 
human breast carcinoma. Am. J. Pathol. 178, 1221–1232 (2011). 
143. Lintz, M., Muñoz, A. & Reinhart-King, C. A. The Mechanics of Single Cell and 
Collective Migration of Tumor Cells. J. Biomech. Eng. 139, 1–9 (2017). 
144. Liao, W. et al. High-throughput three-dimensional spheroid tumor model using 
a novel stamp-like tool. J. Tissue Eng. 10, (2019). 
145. Haeger, A., Wolf, K., Zegers, M. M. & Friedl, P. Collective cell migration: 
Guidance principles and hierarchies. Trends Cell Biol. 25, 556–566 (2015). 
146. Carey, S. P., Starchenko, A., McGregor, A. L. & Reinhart-King, C. A. Leading 
malignant cells initiate collective epithelial cell invasion in a three-dimensional 
heterotypic tumor spheroid model. Clin. Exp. Metastasis 30, 615–630 (2013). 
147. Zhang, J. et al. Energetic regulation of coordinated leader–follower dynamics 
during collective invasion of breast cancer cells. Proc. Natl. Acad. Sci. 116, 
7867–7872 (2019). 
148. Cheung, K. J. & Ewald, A. J. Invasive leader cells: Metastatic oncotarget. 
Oncotarget 5, 1390–1391 (2014). 
149. Leonard, F. & Godin, B. Unraveling bio-social hierarchy in cancer collective 
214 
 
invasion: symbiosis between leaders and followers. Transl. Cancer Res. 6, 
S1084–S1087 (2017). 
150. Labernadie, A. et al. A mechanically active heterotypic E-cadherin/N-cadherin 
adhesion enables fibroblasts to drive cancer cell invasion. Nat. Cell Biol. 19, 
224–237 (2017). 
151. Bordeleau, F. et al. Matrix stiffening promotes a tumor vasculature phenotype. 
Proc. Natl. Acad. Sci. U. S. A. 114, 492–497 (2017). 
152. Ehsan, S. M., Welch-Reardon, K. M., Waterman, M. L., Hughes, C. C. W. & 
George, S. C. A three-dimensional in vitro model of tumor cell intravasation. 
Integr. Biol. 6, 603–610 (2014). 
153. Aaberg-Jessen, C. et al. Invasion of primary glioma- and cell line-derived 
spheroids implanted into corticostriatal slice cultures. Int. J. Clin. Exp. Pathol. 
6, 546–560 (2013). 
154. Raghavan, S. et al. Personalized Medicine Based Approach to Model Patterns 
of Chemoresistance, and Tumor Recurrence Using Ovarian Cancer Stem Cell 
Spheroids. Clin. Cancer Res. 23, 6934–6945 (2017). 
155. Song, Y. et al. Patient-derived multicellular tumor spheroids towards optimized 
treatment for patients with hepatocellular carcinoma. J. Exp. Clin. Cancer Res. 
37, 1–13 (2018). 
156. Cheung, K. J., Gabrielson, E., Werb, Z. & Ewald, A. J. Collective invasion in 
breast cancer requires a conserved basal epithelial program. Cell 155, 1639–
1651 (2013). 
157. Sirka, O. K., Shamir, E. R. & Ewald, A. J. Myoepithelial cells are a dynamic 
barrier to epithelial dissemination. J. Cell Biol. 217, 3368–3381 (2018). 
158. Padmanaban, V. et al. E-cadherin is required for metastasis in multiple models 
of breast cancer. Nature 324, (2019). 
159. Jabs, J. et al. Screening drug effects in patient-derived cancer cells links 
organoid responses to genome alterations. Mol. Syst. Biol. 13, 955 (2017). 
160. Padmanaban, V. et al. Organotypic culture assays for murine and human 
primary and metastatic-site tumors. Nat. Protoc. 15, 2413–2442 (2020). 
161. Wirtz, D., Konstantopoulos, K. & Searson, P. C. The physics of cancer: the role 
of physical interactions and mechanical forces in metastasis. Nat. Rev. Cancer 
11, 512–22 (2011). 
162. Stott, S. L. et al. [SI] Isolation of circulating tumor cells using a microvortex-
215 
 
generating herringbone-chip. Proc. Natl. Acad. Sci. (2010). 
doi:10.1073/pnas.1012539107 
163. Hughes, A. D., Powderly, J. D., Mattison, J., Greene, B. T. & King, M. R. 
Rapid Isolation of Viable Circulating Tumor Cells from Patient Blood Samples. 
J. Vis. Exp. (2012). doi:10.3791/4248 
164. Sarioglu, A. F. et al. A microfluidic device for label-free, physical capture of 
circulating tumor cell clusters. Nat. Methods (2015). doi:10.1038/nmeth.3404 
165. Mitchell, M. J., Castellanos, C. A. & King, M. R. Surfactant functionalization 
induces robust, differential adhesion of tumor cells and blood cells to charged 
nanotube-coated biomaterials under flow. Biomaterials (2015). 
doi:10.1016/j.biomaterials.2015.03.045 
166. Dhar, M. et al. Functional profiling of circulating tumor cells with an integrated 
vortex capture and single-cell protease activity assay. Proc. Natl. Acad. Sci. 
(2018). doi:10.1073/PNAS.1803884115 
167. Cheng, Y.-H. et al. Hydro-Seq enables contamination-free high-throughput 
single-cell RNA-sequencing for circulating tumor cells. Nat. Commun. 10, 2163 
(2019). 
168. Lawler, K., O’Sullivan, G., Long, A. & Kenny, D. Shear stress induces 
internalization of E-cadherin and invasiveness in metastatic oesophageal cancer 
cells by a Src-dependent pathway. Cancer Sci. 100, 1082–1087 (2009). 
169. Mitchell, M. & King, M. Fluid Shear Stress Sensitizes Cancer Cells to 
Receptor-Mediated Apoptosis via Trimeric Death Receptors. New J Phys 15, 
10.1088/1367-2630/15/1/015008 (2013). 
170. Liang, S., Slattery, M. J., Wagner, D., Simon, S. I. & Dong, C. Hydrodynamic 
Shear Rate Regulates Melanoma-Leukocyte Aggregation, Melanoma Adhesion 
to the Endothelium, and Subsequent Extravasation. Ann Biomed Eng 36, 661–
671 (2008). 
171. Egan, K., Cooke, N. & Kenny, D. Living in shear: Platelets protect cancer cells 
from shear induced damage. Clin. Exp. Metastasis 31, 697–704 (2014). 
172. Franzoni, M. et al. Design of a cone-and-plate device for controlled realistic 
shear stress stimulation on endothelial cell monolayers. Cytotechnology 68, 
1885–1896 (2016). 
173. Spruell, C. & Baker, A. B. Analysis of a high-throughput cone-and-plate 
apparatus for the application of defined spatiotemporal flow to cultured cells. 
Biotechnol. Bioeng. 110, 1782–1793 (2013). 
216 
 
174. Barnes, J. M., Nauseef, J. T. & Henry, M. D. Resistance to Fluid Shear Stress Is 
a Conserved Biophysical Property of Malignant Cells. PLoS One 7, (2012). 
175. Chivukula, V. K., Krog, B. L., Nauseef, J. T., Henry, M. D. & Vigmostad, S. C. 
Alterations in cancer cell mechanical properties after fluid shear stress 
exposure: A micropipette aspiration study. Cell Health Cytoskelet. 7, 25–35 
(2015). 
176. Tremblay, P. L., Huot, J. & Auger, F. A. Mechanisms by which E-selectin 
regulates diapedesis of colon cancer cells under flow conditions. Cancer Res. 
68, 5167–5176 (2008). 
177. Haddad, O., Chotard-Ghodsnia, R., Verdier, C. & Duperray, A. Tumor 
cell/endothelial cell tight contact upregulates endothelial adhesion molecule 
expression mediated by NFκB: Differential role of the shear stress. Exp. Cell 
Res. 316, 615–626 (2009). 
178. Rana, K., Liesveld, J. L. & King, M. R. Delivery of apoptotic signal to rolling 
cancer cells: A novel biomimetic technique using immobilized TRAIL and E-
Selectin. Biotechnol. Bioeng. 102, 1692–1702 (2009). 
179. Hsu, J. W. et al. Suppression of prostate cancer cell rolling and adhesion to 
endothelium by 1α,25-dihydroxyvitamin D3. Am. J. Pathol. 178, 872–880 
(2011). 
180. Rana, K., Reinhart-King, C. A. & King, M. R. Inducing apoptosis in rolling 
cancer cells: A combined therapy with aspirin and immobilized TRAIL and E-
selectin. Mol. Pharm. 9, 2219–2227 (2012). 
181. Fan, R. et al. Circulatory shear flow alters the viability and proliferation of 
circulating colon cancer cells. Sci. Rep. 6, 1–8 (2016). 
182. Kühlbach, C., da Luz, S., Baganz, F., Hass, V. & Mueller, M. A Microfluidic 
System for the Investigation of Tumor Cell Extravasation. Bioengineering 5, 40 
(2018). 
183. Bersini, S. et al. A combined microfluidic-transcriptomic approach to 
characterize the extravasation potential of cancer cells. Oncotarget 9, 36110–
36125 (2018). 
184. Song, J. W. et al. Microfluidic endothelium for studying the intravascular 
adhesion of metastatic breast cancer cells. PLoS One 4, (2009). 
185. Lee, H. J. et al. Fluid shear stress activates YAP1 to promote cancer cell 
motility. Nat. Commun. 8, (2017). 
186. Chaw, K. C., Manimaran, M., Tay, E. H. & Swaminathan, S. Multi-step 
217 
 
microfluidic device for studying cancer metastasis. Lab Chip 7, 1041–1047 
(2007). 
187. Au, S. H. et al. Clusters of circulating tumor cells traverse capillary-sized 
vessels. Proc. Natl. Acad. Sci. 113, 4947–4952 (2016). 
188. Chen, M. B. et al. On-chip human microvasculature assay for visualization and 
quantitation of tumor cell extravasation dynamics HHS Public Access Author 
manuscript. Nat Protoc 12, 865–880 (2017). 
189. Kim, S., Lee, H., Chung, M. & Jeon, N. L. Engineering of functional, 
perfusable 3D microvascular networks on a chip. Lab Chip 13, 1489–1500 
(2013). 
190. Miller, J. S. et al. Rapid casting of patterned vascular networks for perfusable 
engineered three-dimensional tissues. Nat. Mater. 11, 768–774 (2012). 
191. Sarveswaran, K. et al. Synthetic Capillaries to Control Microscopic Blood 
Flow. Sci. Rep. 6, 1–18 (2016). 
192. Rankin, E. B., Erler, J. & Giaccia, A. J. 3 - The Cellular Microenvironment and 
Metastases. in (eds. Niederhuber, J. E., Armitage, J. O., Doroshow, J. H., 
Kastan, M. B. & Tepper, J. E. B. T.-A. C. O. (Fifth E.) 40-51.e4 (Content 
Repository Only!, 2014). doi:https://doi.org/10.1016/B978-1-4557-2865-
7.00003-5 
193. Macedo, F. et al. Bone metastases: An overview. Oncol. Rev. (2017). 
doi:10.4081/oncol.2017.321 
194. Jeon, J. S. et al. Human 3D vascularized organotypic microfluidic assays to 
study breast cancer cell extravasation. Proc. Natl. Acad. Sci. 112, 214–219 
(2015). 
195. Marturano-Kruik, A. et al. Human bone perivascular niche-on-a-chip for 
studying metastatic colonization. Proc. Natl. Acad. Sci. 115, 1256–1261 (2018). 
196. Sano, E. et al. Engineering of vascularized 3D cell constructs to model cellular 
interactions through a vascular network. Biomicrofluidics 12, (2018). 
197. Jusoh, N., Oh, S., Kim, S., Kim, J. & Jeon, N. L. Microfluidic vascularized 
bone tissue model with hydroxyapatite-incorporated extracellular matrix. Lab 
Chip (2015). doi:10.1039/c5lc00698h 
198. Lynch, M. E. et al. In vivo tibial compression decreases osteolysis and tumor 
formation in a human metastatic breast cancer model. J. Bone Miner. Res. 
(2013). doi:10.1002/jbmr.1966 
218 
 
199. Mastro, A. M. & Vogler, E. A. A three-dimensional osteogenic tissue model for 
the study of metastatic tumor cell interactions with bone. Cancer Res. 69, 4097–
4100 (2009). 
200. Mayorca-Guiliani, A. E. et al. ISDoT: In situ decellularization of tissues for 
high-resolution imaging and proteomic analysis of native extracellular matrix. 
Nat. Med. (2017). doi:10.1038/nm.4352 
201. Xiong, G., Flynn, T., Chen, J., Trinkle, C. & Xu, R. Development of an ex vivo 
breast cancer lung colonization model utilizing decellularized lung matrix. 
Integr. Biol. 7, 1518–1525 (2015). 
202. Wheeler, S. E. et al. Spontaneous dormancy of metastatic breast cancer cells in 
an all human liver microphysiologic system. Br. J. Cancer (2014). 
doi:10.1038/bjc.2014.533 
203. Bos, P. D. et al. Genes that mediate breast cancer metastasis to the brain. 
Nature 459, 1005–1009 (2009). 
204. 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). 
205. Tominaga, N. et al. Brain metastatic cancer cells release microRNA- 181c-
containing extracellular vesicles capable of destructing blood–brain barrier. 
Nat. Commun. 6, doi: 10.1038/ncomms7716 (2015). 
206. Drolez, A. et al. ST6GALNAC5 expression decreases the interactions between 
breast cancer cells and the human blood-brain barrier. Int. J. Mol. Sci. 17, 
(2016). 
207. Xu, H. et al. A dynamic in vivo-like organotypic blood-brain barrier model to 
probe metastatic brain tumors. Sci. Rep. 6, 1–12 (2016). 
208. Skardal, A., Devarasetty, M., Forsythe, S., Atala, A. & Soker, S. A reductionist 
metastasis-on-a-chip platform for in vitro tumor progression modeling and drug 
screening. Biotechnol. Bioeng. (2016). doi:10.1002/bit.25950 
209. Parikh, M. R., Minser, K. E., Rank, L. M., Glackin, C. A. & Kirshner, J. A 
reconstructed metastasis model to recapitulate the metastatic spread in vitro. 
Biotechnol. J. (2014). doi:10.1002/biot.201400121 
210. Heron, M. Deaths: Leading causes for 2017. Natl. Vital Stat. Reports 68, 
(2019). 
211. Younus, H. & Anwar, S. Prevention of Non-Enzymatic Glycosylation 
(Glycation) : Implication in the Treatment of Diabetic Complication. Int. J. 
219 
 
Health Sci. (Qassim). 10, 247–263 (2016). 
212. Prasad, C., Davis, K. E., Imrhan, V., Juma, S. & Vijayagopal, P. Advanced 
Glycation End Products and Risks for Chronic Diseases: Intervening Through 
Lifestyle Modification. Am. J. Lifestyle Med. 13, 384–404 (2019). 
213. Ramasamy, R., Yan, S. F. & Schmidt, A. M. Receptor for AGE (RAGE): 
Signaling mechanisms in the pathogenesis of diabetes and its complications. 
Ann. N. Y. Acad. Sci. 1243, 88–102 (2011). 
214. Li, M. et al. Generation and Biological Characteristics of a Mouse Model of 
Breast Cancer that Copresents with Diabetes Mellitus. Anat. Rec. 302, 269–277 
(2019). 
215. Faustino-Rocha, A. et al. Estimation of rat mammary tumor volume using 
caliper and ultrasonography measurements. Lab Anim. (NY). 42, 217–224 
(2013). 
216. Huynh, J. et al. Age-related intimal stiffening enhances endothelial 
permeability and leukocyte transmigration. Sci. Transl. Med. 3, (2011). 
217. Wang, W., Lollis, E. M., Bordeleau, F. & Reinhart-King, C. A. Matrix stiffness 
regulates vascular integrity through focal adhesion kinase activity. FASEB J. 
33, 1199–1208 (2019). 
218. Wang, W., Miller, J. P., Pannullo, S. C., Reinhart-King, C. A. & Bordeleau, F. 
Quantitative Assessment of Cell Contractility Using Polarized Light 
Microscopy. J Biophotonics 11, (2018). 
219. Nonogaki, K., Strack, A. M., Dallman, M. F. & Tecott, L. H. Leptin-
independent hyperphagia and type 2 diabetes in mice with a mutated serotonin 
5-HT(2C) receptor gene. Nat. Med. 4, 1152–1156 (1998). 
220. Furman, B. L. Streptozotocin-Induced Diabetic Models in Mice and Rats. Curr. 
Protoc. Pharmacol. 70, 5.47.1-5.47.20 (2015). 
221. Nielsen, B. S. et al. Matrix metalloproteinase 13 is induced in fibroblasts in 
polyomavirus middle T antigen-driven mammary carcinoma without 
influencing tumor progression. PLoS One 3, (2008). 
222. Brownlee, M., Cerami, A. & Vlassara, H. Advanced Glycosylation End 
Products in Tissue and the Biochemical Basis of Diabetic Complications. N 
Engl J Med. 318, 1315–1321 (1988). 
223. Mason, B. N., Starchenko, A., Williams, R. M., Bonassar, L. J. & Reinhart-
King, C. A. Tuning three-dimensional collagen matrix stiffness independently 
of collagen concentration modulates endothelial cell behavior. Acta Biomater. 
220 
 
9, 4635–4644 (2013). 
224. Mason, B. N. & Reinhart-King, C. A. Controlling the mechanical properties of 
three-dimensional matrices via non-enzymatic collagen glycation. 
Organogenesis 9, 70–75 (2013). 
225. Wei, S. C. & Yang, J. Forcing through Tumor Metastasis: The Interplay 
between Tissue Rigidity and Epithelial-Mesenchymal Transition. Trends Cell 
Biol. 26, 111–120 (2016). 
226. Wei, S. C. et al. Matrix stiffness drives epithelial-mesenchymal transition and 
tumour metastasis through a TWIST1-G3BP2 mechanotransduction pathway. 
Nat. Cell Biol. 17, 678–688 (2015). 
227. Park, J. & Schwarzbauer, J. E. Mammary epithelial cell interactions with 
fibronectin stimulate epithelial-mesenchymal transition. Oncogene 33, 1649–
1657 (2014). 
228. Xu, J., Lamouille, S. & Derynck, R. TGF-Β-induced epithelial to mesenchymal 
transition. Cell Res. 19, 156–172 (2009). 
229. Vincent, T. et al. A SNAIL1-SMAD3/4 transcriptional repressor complex 
promotes TGF-B mediated epithelial-mesenchymal transition. Nat. Cell Biol. 
11, 943–950 (2009). 
230. Kwak, T. et al. Targeting of RAGE-ligand signaling impairs breast cancer cell 
invasion and metastasis. Oncogene 36, 1559–1572 (2017). 
231. Nasser, M. W. et al. RAGE mediates S100A7-induced breast cancer growth and 
metastasis by modulating the tumor microenvironment. Cancer Res. 75, 974–
985 (2015). 
232. Sell, D. R., Nelson, J. F. & Monnier, V. M. Effect of chronic Aminoguanidine 
treatment on age-related glycation, glycoxidation, and collagen cross-linking in 
the fischer 344 rat. Journals Gerontol. - Ser. A Biol. Sci. Med. Sci. 56, B405–
B411 (2001). 
233. Coughlan, M. T. et al. Combination therapy with the advanced glycation end 
product cross-link breaker, alagebrium, and angiotensin converting enzyme 
inhibitors in diabetes: Synergy or redundancy? Endocrinology 148, 886–895 
(2007). 
234. Nenna, A. et al. Pharmacologic approaches against Advanced Glycation End 
Products (AGEs) in diabetic cardiovascular disease. Res. Cardiovasc. Med. 4, 5 
(2015). 
235. Lahiri, S. W. Management of type 2 diabetes: What is the next step after 
221 
 
metformin? Clin. Diabetes 30, 72–75 (2012). 
236. Saraei, P., Asadi, I., Kakar, M. A. & Moradi-Kor, N. The beneficial effects of 
metformin on cancer prevention and therapy: A comprehensive review of recent 
advances. Cancer Manag. Res. 11, 3295–3313 (2019). 
237. Yang, F. et al. Intratumor heterogeneity predicts metastasis of triplenegative 
breast cancer. Carcinogenesis 38, 900–909 (2017). 
238. Turajlic, S. & Swanton, C. Metastasis as an evolutionary process. Science (80-. 
). 352, 169–175 (2016). 
239. Lin, J. M. G. et al. Linking invasive motility to protein expression in single 
tumor cells. Lab Chip 18, 371–384 (2018). 
240. Wu, Y., Siadaty, M. S., Berens, M. E., Hampton, G. M. & Theodorescu, D. 
Overlapping gene expression profiles of cell migration and tumor invasion in 
human bladder cancer identify metallothionein 1E and nicotinamide N-
methyltransferase as novel regulators of cell migration. Oncogene 27, 6679–
6689 (2008). 
241. Clark, A. G. & Vignjevic, D. M. Modes of cancer cell invasion and the role of 
the microenvironment. Curr. Opin. Cell Biol. 36, 13–22 (2015). 
242. Carey, S. P. et al. Comparative mechanisms of cancer cell migration through 
3D matrix and physiological microtracks. Am. J. Physiol. - Cell Physiol. 308, 
C436–C447 (2015). 
243. Riching, K. M. et al. 3D Collagen Alignment Limits Protrusions to Enhance 
Breast Cancer Cell Persistence. Biophysj 107, 2546–2558 (2014). 
244. Carey, S. P. et al. Local extracellular matrix alignment directs cellular 
protrusion dynamics and migration through Rac1 and FAK. Integr. Biol. 8, 
821–835 (2016). 
245. Bordeleau, F., Alcoser, T. A. & Reinhart-King, C. A. Physical biology in 
cancer. 5. The rocky road of metastasis: The role of cytoskeletal mechanics in 
cell migratory response to 3D matrix topography. Am. J. Physiol. - Cell Physiol. 
306, 110–120 (2014). 
246. Schor, S. L. Cell proliferation and migration on collagen substrata in vitro. J. 
Cell Sci. 41, 159–175 (1980). 
247. Aznavoorian, S., Murphy, A. N., Stetler, S. W. & Liotta, L. A. Molecular 
aspects of tumor cell invasion and metastasis. Cancer 71, 1368–1383 (1993). 
248. Kraning-Rush, C. M., Carey, S. P., Lampi, M. C. & Reinhart-King, C. A. 
222 
 
Microfabricated collagen tracks facilitate single cell metastatic invasion in 3D. 
Integr. Biol. 5, 606–616 (2013). 
249. Kraning-Rush, C. M., Carey, S. P., Califano, J. P. & Reinhart-King, C. A. 
Quantifying Traction Stresses in Adherent Cells. Methods in Cell Biology 110, 
(Elsevier Inc., 2012). 
250. Berens, E. B., Holy, J. M., Riegel, A. T. & Wellstein, A. A cancer cell spheroid 
assay to assess invasion in a 3D setting. J. Vis. Exp. 2015, 1–6 (2015). 
251. Wang, H. H. et al. B16 melanoma cell arrest in the mouse liver induces nitric 
oxide release and sinusoidal cytotoxicity: A natural hepatic defense against 
metastasis. Cancer Res. 60, 5862–5869 (2000). 
252. Puram, S. V. et al. Single-Cell Transcriptomic Analysis of Primary and 
Metastatic Tumor Ecosystems in Head and Neck Cancer. Cell 171, 1611-
1624.e24 (2017). 
253. Pastushenko, I. et al. Identification of the tumour transition states occurring 
during EMT. Nature (2018). doi:10.1038/s41586-018-0040-3 
254. Györffy, B. et al. An online survival analysis tool to rapidly assess the effect of 
22,277 genes on breast cancer prognosis using microarray data of 1,809 
patients. Breast Cancer Res. Treat. 123, 725–731 (2010). 
255. Hughes, A. D. et al. Differential drug responses of circulating tumor cells 
within patient blood. Cancer Lett 28, 28–35 (2014). 
256. Meye, A. et al. Isolation and enrichment of urologic tumor cells in blood 
samples by a semi-automated CD45 depletion autoMACS protocol. Int J Oncol. 
21, 521–530 (2002). 
257. Sambrook J, Fritsch, E., Maniatis, T. Molecular Cloning: A Laboratory Manual. 
Cold Springs Harb. Lab. Press (1989). 
258. Ali, R. & Wendt, M. K. The paradoxical functions of EGFR during breast 
cancer progression. Signal Transduct. Target. Ther. 2, 1–7 (2017). 
259. Martín-Pardillos, A. et al. The role of clonal communication and heterogeneity 
in breast cancer. BMC Cancer 19, 1–26 (2019). 
260. Koedoot, E. et al. Uncovering the signaling landscape controlling breast cancer 
cell migration identifies novel metastasis driver genes. Nat. Commun. 10, 1–16 
(2019). 
261. Jiang, W. G. et al. Tissue invasion and metastasis: Molecular, biological and 
clinical perspectives. Semin. Cancer Biol. 35, S244–S275 (2015). 
223 
 
262. Elisha, Y., Kalchenko, V., Kuznetsov, Y. & Geiger, B. Dual role of E-cadherin 
in the regulation of invasive collective migration of mammary carcinoma cells. 
Sci. Rep. 8, 1–15 (2018). 
263. Aceto, N. et al. Circulating tumor cell clusters are oligoclonal precursors of 
breast cancer metastasis. Cell 158, 1110–1122 (2014). 
264. Giuliano, M. et al. Perspective on circulating tumor cell clusters: Why it takes a 
village to metastasize. Cancer Res. 78, 845–852 (2018). 
265. Mutlu, B. R. et al. In-flow measurement of cell–cell adhesion using oscillatory 
inertial microfluidics. Lab Chip 1612–1620 (2020). doi:10.1039/d0lc00089b 
266. Baugher, P. J., Krishnamoorthy, L., Price, J. E. & Dharmawardhane, S. F. Rac1 
and Rac3 isoform activation is involved in the invasive and metastatic 
phenotype of human breast cancer cells. Breast Cancer Res. 7, R965–R974 
(2005). 
267. Tojkander, S., Gateva, G. & Lappalainen, P. Actin stress fibers - Assembly, 
dynamics and biological roles. J. Cell Sci. 125, 1855–1864 (2012). 
268. Huveneers, S. & Danen, E. H. J. Adhesion signaling - Crosstalk between 
integrins, Src and Rho. J. Cell Sci. 122, 1059–1069 (2009). 
269. Luo, M. & Guan, J.-L. Focal Adhesion Kinase: a Prominent Determinant in 
Breast Cancer Initiation, Progression, and Metastasis. Cancer Lett 289, 127–
139 (2010). 
270. Venhuizen, J.-H., Jacobs, F. J. C., Span, P. N. & Zegers, M. M. P120 and E-
cadherin: double-edged swords in tumor metastasis. Semin. Cancer Biol. 1–14 
(2019). doi:10.1016/j.semcancer.2019.07.020 
271. Wells, A., Chao, Y. L., Grahovac, J., Wu, Q. & Lauffenburger, D. A. Epithelial 
and mesenchymal phenotypic switchings modulate cell motility in metastasis. 
Front. Biosci. (Landmark Ed. 16, 815–37 (2011). 
272. Russo, G. C. et al. E-cadherin promotes cell hyper- proliferation in breast 
cancer. BioRxiv (2020). 
273. Liu, X. et al. Epithelial-type systemic breast carcinoma cells with a restricted 
mesenchymal transition are a major source of metastasis. Sci. Adv. 5, eaav4275 
(2019). 
274. Jolly, M. K. et al. Stability of the hybrid epithelial/mesenchymal phenotype. 
Oncotarget 7, 27067–27084 (2016). 
275. Lu, M., Jolly, M. K., Levine, H., Onuchic, J. N. & Ben-Jacob, E. MicroRNA-
224 
 
based regulation of epithelial-hybrid-mesenchymal fate determination. Proc. 
Natl. Acad. Sci. U. S. A. 110, 18144–18149 (2013). 
276. Saitoh, M. Involvement of partial EMT in cancer progression. J. Biochem. 164, 
257–264 (2018). 
277. Ocaña, O. H. et al. Metastatic Colonization Requires the Repression of the 
Epithelial-Mesenchymal Transition Inducer Prrx1. Cancer Cell 22, 709–724 
(2012). 
278. Tsai, J. H., Donaher, J. L., Murphy, D. A., Chau, S. & Yang, J. Spatiotemporal 
Regulation of Epithelial-Mesenchymal Transition Is Essential for Squamous 
Cell Carcinoma Metastasis. Cancer Cell 22, 725–736 (2012). 
279. Cai, D. et al. Mechanical feedback through E-cadherin promotes direction 
sensing during collective cell migration. 157, 1146–1159 (2015). 
280. Gayrard, C., Bernaudin, C., Déjardin, T., Seiler, C. & Borghi, N. Src- and 
confinement-dependent FAK activation causes E-cadherin relaxation and β-
catenin activity. J. Cell Biol. 217, 1063–1077 (2018). 
281. Chao, Y. L., Shepard, C. R. & Wells, A. Breast carcinoma cells re-express E-
cadherin during mesenchymal to epithelial reverting transition. Mol. Cancer 9, 
1–18 (2010). 
282. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: The next generation. Cell 
144, 646–674 (2011). 
283. VanderVorst, K. et al. Wnt/PCP signaling contribution to carcinoma collective 
cell migration and metastasis. Cancer Res. 79, 1719–1729 (2019). 
284. Thüroff, F., Goychuk, A., Reiter, M. & Frey, E. Bridging the gap between 
single-cell migration and collective dynamics. Elife 8, (2019). 
285. Palmer, T. D., Ashby, W. J., Lewis, J. D. & Zijlstra, A. Targeting tumor cell 
motility to prevent metastasis. Adv. Drug Deliv. Rev. 8, 568–581 (2011). 
286. Greaves, M. & Maley, C. C. Clonal Evolution in Cancer. Nature 481, 306–313 
(2012). 
287. Trepat, X. et al. Physical forces during collective cell migration. Nat. Phys. 5, 
426–430 (2009). 
288. Cheung, K. J., Padmanaban, V., Silvestri, V., Schipper, K. & Cohen, J. D. 
Polyclonal breast cancer metastases arise from collective dissemination of 
keratin 14-expressing tumor cell clusters. (2016). doi:10.1073/pnas.1508541113 
225 
 
289. Cheung, K. J. & Ewald, A. J. A collective route to metastasis: Seeding by tumor 
cell clusters. Science (80-. ). 352, 167–169 (2016). 
290. Deisboeck, T. S. & Couzin, I. D. Collective behavior in cancer cell populations. 
BioEssays 31, 190–197 (2009). 
291. Zanotelli, M. R. et al. Regulation of ATP utilization during metastatic cell 
migration by collagen architecture. Mol. Biol. Cell mbc.E17-01-0041 (2017). 
doi:10.1091/mbc.E17-01-0041 
292. King, A. J. F. The use of animal models in diabetes research. Br. J. Pharmacol. 
166, 877–894 (2012). 
293. Johnstone, C. N. et al. Functional and molecular characterisation of 
EO771.LMB tumours, a new C57BL/6-mouse-derived model of spontaneously 
metastatic mammary cancer. DMM Dis. Model. Mech. 8, 237–251 (2015). 
294. Lin, E. Y. et al. Progression to Malignancy in the Polyoma Middle T 
Oncoprotein Mouse Breast Cancer Model Provides a Reliable Model for 
Human Diseases. Am. J. Pathol. 163, 2113–2126 (2003). 
295. Borowsky, A. D. et al. Syngeneic mouse mammary carcinoma cell lines: Two 
closely related cell lines with divergent metastatic behavior. Clin. Exp. 
Metastasis 22, 47–59 (2005). 
296. Wilson, P. C. et al. The single-cell transcriptomic landscape of early human 
diabetic nephropathy. Proc. Natl. Acad. Sci. U. S. A. 116, 19619–19625 (2019). 
297. Madonna, M. C. et al. Optical imaging of glucose uptake and mitochondrial 
membrane potential to characterize HER2 breast tumor metabolic phenotypes. 
Mol. Cancer Res. 17, 1545–1555 (2019). 
298. Croteau, E. et al. PET Metabolic Biomarkers for Cancer. Biomark. Cancer 8s2, 
BIC.S27483 (2016). 
299. Park, J., Kusminski, C. M., Chua, S. C. & Scherer, P. E. Leptin receptor 
signaling supports cancer cell metabolism through suppression of mitochondrial 
respiration in vivo. Am. J. Pathol. 177, 3133–3144 (2010). 
300. Haythorne, E. et al. Diabetes causes marked inhibition of mitochondrial 
metabolism in pancreatic β-cells. Nat. Commun. 10, (2019). 
301. Pavlova, N. N. & Thompson, C. B. Emerging metabolic hallmarks of cancer. 
Cell Metab 12, 27–47 (2016). 
302. De Luca, A. et al. The role of the EGFR signaling in tumor microenvironment. 
J. Cell. Physiol. 214, 559–567 (2008). 
226 
 
303. Wechman, S. L. et al. Vascular mimicry: Triggers, molecular interactions and 
in vivo models. 1–34 (2020). doi:10.1016/bs.acr.2020.06.001.Vascular 
304. Andonegui-Elguera, M. A. et al. An Overview of Vasculogenic Mimicry in 
Breast Cancer. Front. Oncol. 10, 1–8 (2020). 
305. Seftor, R. E. B. et al. Tumor cell vasculogenic mimicry: From controversy to 
therapeutic promise. Am. J. Pathol. 181, 1115–1125 (2012). 
306. Parekh, A. & Weaver, A. M. Regulation of cancer invasiveness by the physical 
extracellular matrix environment. Cell Adhes. Migr. 3, 288–292 (2009). 
307. Lv, Z. D. et al. miR-655 suppresses epithelial-to-mesenchymal transition by 
targeting Prrx1 in triple-negative breast cancer. J. Cell. Mol. Med. 20, 864–873 
(2016). 
308. Cai, J. et al. Whole-genome sequencing identifies genetic variances in culture-
expanded human mesenchymal stem cells. Stem Cell Reports 3, 227–233 
(2014). 
309. Petrova, Y. I., Schecterson, L. & Gumbiner, B. M. Roles for E-cadherin cell 
surface regulation in cancer. Mol. Biol. Cell 27, 3233–3244 (2016). 
310. Stockinger, A., Eger, A., Wolf, J., Beug, H. & Foisner, R. E-cadherin regulates 
cell growth by modulating proliferation-dependent β-catenin transcriptional 
activity. J. Cell Biol. 154, 1185–1196 (2001). 
311. Na, T. Y., Schecterson, L., Mendonsa, A. M. & Gumbiner, B. M. The 
functional activity of E-cadherin controls tumor cell metastasis at multiple 
steps. Proc. Natl. Acad. Sci. U. S. A. 117, 5931–5937 (2020). 
312. Shamir, E. R. & Ewald, A. J. Adhesion in mammary development: Novel roles 
for E-cadherin in individual and collective cell migration. Current Topics in 
Developmental Biology 112, (Elsevier Inc., 2015). 
313. Tan, D. S. P. et al. The biological and prognostic significance of cell polarity 
and E- cadherin in grade I infiltrating ductal carcinoma of the breast. J. Pathol. 
189, 20–27 (1999). 
314. Yamashita, N. et al. Epithelial Paradox: Clinical Significance of Coexpression 
of E-cadherin and Vimentin With Regard to Invasion and Metastasis of Breast 
Cancer. Clin. Breast Cancer 18, e1003–e1009 (2018). 
315. George, J. T., Jolly, M. K., Xu, S., Somarelli, J. A. & Levine, H. Survival 
outcomes in cancer patients predicted by a partial EMT gene expression scoring 
metric. Cancer Res. 77, 6415–6428 (2017). 
227 
 
316. Polioudaki, H. et al. Variable expression levels of keratin and vimentin reveal 
differential EMT status of circulating tumor cells and correlation with clinical 
characteristics and outcome of patients with metastatic breast cancer. BMC 
Cancer 15, 399 (2015). 
317. Serrano-Gomez, S. J., Maziveyi, M. & Alahari, S. K. Regulation of epithelial-
mesenchymal transition through epigenetic and post-translational modifications. 
Mol. Cancer 15, 1–14 (2016). 
318. Lu, W. & Kang, Y. Epithelial-mesenchymal plasticity in cancer progression and 
metastasis Wei. Dev Cell 49, 361–374 (2019). 
319. Onder, T. T. et al. Loss of E-cadherin promotes metastasis via multiple 
downstream transcriptional pathways. Cancer Res. 68, 3645–3654 (2008). 
320. Li, Z., Yin, S., Zhang, L., Liu, W. & Chen, B. Prognostic value of reduced E-
cadherin expression in breast cancer: A meta-analysis. Oncotarget 8, 16445–
16455 (2017). 
321. Al-Saad, S. et al. The prognostic impact of NF-κB p105, vimentin, E-cadherin 
and Par6 expression in epithelial and stromal compartment in non-small-cell 
lung cancer. Br. J. Cancer 99, 1476–1483 (2008). 
322. Liu, D. et al. E-cadherin expression associated with differentiation and 
prognosis in patients with non-small cell lung cancer. Ann. Thorac. Surg. 71, 
949–954 (2001). 
323. Baumgart, E. et al. Identification and prognostic significance of an epithelial-
mesenchymal transition expression profile in human bladder tumors. Clin. 
Cancer Res. 13, 1685–1694 (2007). 
324. Lefevre, M. et al. Epithelial to mesenchymal transition and HPV infection in 
squamous cell oropharyngeal carcinomas: The papillophar study. Br. J. Cancer 
116, 362–369 (2017). 
325. Zhang, T. et al. CDH1 (E-cadherin) expression independently affects clinical 
outcome in acute myeloid leukemia with normal cytogenetics. Clin. Chem. Lab. 
Med. 55, 123–131 (2017). 
326. Osada, T. et al. E-cadherin is involved in the intrahepatic metastasis of 
hepatocellular carcinoma. Hepatology 24, 1460–1467 (1996). 
327. Schneider, M. R. et al. Evidence for a role of E-cadherin in suppressing liver 
carcinogenesis in mice and men. Carcinogenesis 35, 1855–1862 (2014). 
328. Liu, X. & Chu, K. M. E-cadherin and gastric cancer: Cause, consequence, and 
applications. Biomed Res. Int. 2014, 1–9 (2014). 
228 
 
329. Khalil, A. A. et al. Collective invasion induced by an autocrine purinergic loop 
through connexin-43 hemichannels. J. Cell Biol. 219, (2020). 
 
229