ABSTRACT
Title of dissertation: STRUCTURE AND DYNAMICS
OF MICROTENTACLES:
FROM A PHYSICAL PERSPECTIVE
Eleanor Claire-higgins Ory,
Doctor of Philosophy, 2016
Dissertation directed by: Professor Wolfgang Losert
Department of Biophysics
While modern cancer diagnostics and treatments are often interpreted through
a biomolecular perspective, cancer abounds with many mechanically interesting
characteristics and questions. Metastasis, the process by which a primary tumor
spreads and forms a second tumor in a distant site is currently responsible for 90%
of cancer fatalities [1–3]. One of the key limiting steps in metastasis is extravasa-
tion; the process by which a circulating tumor cell (CTC) moves from the blood-
stream into surrounding tissue. So far, most in vitro studies in metastasis focus
on cell migration and invasiveness with few focused on reattachment of cells to a
blood vessel wall, and extravasation. One possible attachment mechanism involves
tubulin-based structures called microtentacles, which have been observed to poke
into crevices between cells that line blood vessels. Based on biomolecular assays,
the current hypothesis is that microtentacles are formed as the result of unbalanced,
mechanical interactions between microtubules and actin, allowing microtubules to
push the plasma membrane beyond the cell body. The focus of my thesis is to gain
insights into the dynamics and mechanical properties of microtentacles and evaluate
how microtentacles may be altered by cytoskeletal drugs.
In this thesis, I will measure changes in microtubule dynamics using cytoskele-
tal drugs to the actomyosin cortex and microtubules. The first study presented
examines how drug treatments targeting the actomyosin cortex impact microtubule
dynamics for attached cells. The results of the first study demonstrate that weaken-
ing the actomyosin cortex allows microtubule end-binding-protein-1 (EB1) to move
beyond the cell body boundary. Weakening the actomyosin cortex also results in
changes to the speed and straightness of microtubule growth. In the second study, an
image analysis framework is presented to quantify microtentacles as well as an eval-
uation of the dynamics of microtubules in suspended cells. The study demonstrates
a successful image analysis technique that can evaluate microtentacle phenotype for
both free-floating and tethered cells as well as dynamics for tethered cells. This
second study shows that while microtubule stabilizing drug treatment with Taxol
increases total microtentacle phenotype, it also reduces microtentacle dynamics. On
the other hand, while microtubule destabilizing drug treatment Colchicine decreases
total microtentacle phenotype, Colchicine also reduces microtentacle dynamics. As
a summary and outlook, I present a mechanical framework and present hypotheses
for 4 different genetic modifications spanning a spectrum of different cytoskeletal
states. I also show preliminary, qualitative results for 3 out of the 4 different cell
lines. Critical to evaluating microtentacles within this physical framework is a direct
mechanical assay; here, I show preliminary work taken at the University of Leipzig
on an optical stretcher.
Given that microtentacles have demonstrated to be a sufficient prerequisite for
reattachment, better understanding of what circumstances lead to microtentacles is
a critical basic research question. My work applies a physical perspective to the bal-
ance between the actomyosin cortex and microtubules and demonstrates changes in
microtubule dynamics. Such work contributes towards the possibility of identifying
morphological and dynamics signatures of CTCs with higher metastatic potential.
STRUCTURE AND DYNAMICS OF MICROTENTACLES FROM A
PHYSICAL PERSPECTIVE
by
EChO
Eleanor Claire-higgins Ory
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2016
Advisory Committee:
Wolfgang Losert, Chair/Advisor
Stuart Martin, Co-Advisor
Professor Arpita Uphadhyaya
Professor Marco Colombini
Professor Kimberly Stroka
c© Copyright by
Eleanor Claire-higgins Ory
2016
Preface
Nature uses only the longest threads to weave her patterns, so that
each small piece of her fabric reveals the organization of the entire
tapestry. -Richard Feynman
Simplicity is the ultimate sophistication -Leonardo DiVinci
ii
Dedication
Dedication: I dedicate my research in memory to the family members who
have profoundly impacted my life and succumbed to their battles with cancer most
especially to my late grandparents Grandma, Gigi, and Gramps.
To my busybody Grandma (Rita Higgins), who was a teacher, nurse, en-
trepreneur of a craft store, president of the Chicago women′s club, a mother of 9
children, and a very dedicated catholic who instilled in me a sense of civic duty and
work ethic.
To Gigi (Norma Ory), a free-spirited artist who filled my childhood with all
the imagination and art supplies a kid could ever wish for. She was passionate
about art and education; she was the dean for many years and founded many of the
outreach programs that continue 30+ years later at the Glassel School of Arts; and
inspired me to always follow my heart and dreams.
To Gramps (Edwin M Ory), an MD who pioneered research in infectious dis-
eases and internal medicine. Who never ceased to be amazed and interested in the
latest technologies and their future roles in medicine and impressed upon me the
impact that basic and clinical research can have on the lives of Americans.
iii
Acknowledgments
This work would not have been possible without the help of so many people.
Here are a few notable ones:
First of all, I cannot thank my parents enough for all their love and support
throughout the entirety of my education and for always being there for me no matter
what. I would like to thank them for listening to all my technical rants and for never
giving up on trying to understand what it is I do.
I want to thank several people from my undergraduate education at Smith
College. I am grateful to all my professors for encouraging me to pursue interdisci-
plinary projects early on. A special thanks to my philosophy mentor Nalini Bhushan
for continuing to give me advice on finishing my thesis and writing techniques in
the final stretch of my graduate education. I must thank all my wonderful Smithie
engineering girls for putting up with my doing work even on vacations; for supplying
a steady stream of hilarious youtube videos and messages, especially the cat ones,
when I was most stressed out; and for persistently insisting that I can do this!
I want to thank my wonderful labmates in both the Losert and Martin lab for
all their help and camaraderie: to Keyata, for her encouragement, great sense of
humor, and being an all-around awesome officemate; to Michele, for teaching me so
much about the phenotypes of many of the cancer cell models; to Becky, for showing
me the ropes in lab and feedback with so many biological techniques; to Kristi, for
teaching me and helping me with tethering techniques; to Lek, for teaching me
immunohistochemistry; to Josh, for all the carpooling and help with exponential
iv
fits; to Desu, for so much assistance with image analysis; to Chenlu and Rachel for
sharing ideas, articles, resources, and an office; and to anyone I haven′t listed.
I would like to say thank you to my committee for their time and patience and
making this defense possible on such a short time scale. I would like to extend a
thank you to Kimberly Stroka for joining my committee this summer. I would also
like to thank Marco Colombini for helping me out when I first got started on this
journey at College Park and for teaching me everything I know about membranes. A
sincere thanks to Arpita Uphadhaya for her patience with me on the NBT2 project,
for co-advising in the beginning, and for teaching me so much about EMT, image
analysis, and traction force microscopy.
I would like to express my gratitude to co-advisor Stuart Martin for all his
support, guidance, and direction throughout this process. For all the times I walked
into his office overwhelmed, I never managed to leave without an improved research
strategy and renewed enthusiasm for science. I greatly appreciate his sharing of so
much wisdom and encouragement.
Lastly, I want to give a special thanks to my advisor Wolfgang Losert for this
opportunity and having the courage to pursue high-risk research projects. It has
been a pleasure working with someone who has the collaborative vision to couple
together the right synergy between questions, technologies, and models in this fast-
growing interdisciplinary world. Every time that I hit a research roadblock, he
always had a new technique to try or question to pursue.
v
Table of Contents
List of Figures viii
List of Abbreviations x
1 Introduction 1
1.1 Cancer and Physics: a Historical Perspective . . . . . . . . . . 1
1.2 Metastasis Causes Cancer Mortality . . . . . . . . . . . . . . . . 5
1.3 Tumor Cell Circulation and Reattachment During Metastasis 8
1.3.1 Leukocyte extravasation . . . . . . . . . . . . . . . . . . . . . 8
1.3.2 Platelet and CTC interactions . . . . . . . . . . . . . . . . . . 9
1.3.3 Vascular permeability . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.4 Capillary Entrapment . . . . . . . . . . . . . . . . . . . . . . 11
1.3.5 Microtentacles . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Microtentacles Promote Tumor Cell Reattachment . . . . . . 12
1.5 Microtentacles from a Biophysical Perspective . . . . . . . . . 17
1.5.1 Tubulin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.5.2 Actomyosin Cortex and Model . . . . . . . . . . . . . . . . . . 19
1.5.3 Differences in adherent cells and suspended cells . . . . . . . . 28
1.6 Goals and Hypothesis . . . . . . . . . . . . . . . . . . . . . . . . . 31
2 Microtubule growth dynamics of breast tumor cells is
altered by changes in actin network structure and con-
tractility 32
2.1 Summary (Abstract) . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.3.1 Actin cytoskeleton integrity and contractility affects the struc-
ture of the microtubule network . . . . . . . . . . . . . . . . . 36
2.3.2 Quantifying microtubule growth dynamics via live cell imaging 40
2.3.3 Particle tracking microtubule end dynamics . . . . . . . . . . 44
2.3.4 Localizing microtubule tip relative to the cell body boundaries . 45
vi
2.3.5 Measuring the fraction of microtubule tips that are part of
localized protrusions . . . . . . . . . . . . . . . . . . . . . . . 48
2.3.6 Measuring microtubule tip speed . . . . . . . . . . . . . . . . 50
2.3.7 Measuring microtubule tip trajectory straightness . . . . . . . 53
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
2.5 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . 61
2.5.1 Breast cancer cells and culture conditions . . . . . . . . . . . . 61
2.5.2 Indirect Immunofluorescence . . . . . . . . . . . . . . . . . . . 63
2.5.3 Transfection and confocal microscopy . . . . . . . . . . . . . . 63
2.5.4 Image analysis, particle tracking and computation . . . . . . . 64
2.5.5 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.6 Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3 Extracting Microtentacle Dynamics in Non-adherent Cells 67
3.1 Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.3.1 Anisotropic Filter allows us to capture outline of microtentacles 70
3.3.2 Tethering prevents cells from drifting and improves visualiza-
tion of microtentacles . . . . . . . . . . . . . . . . . . . . . . 73
3.3.3 Image Analysis captures microtentacles qualitatively and quan-
titatively of drug treatments . . . . . . . . . . . . . . . . . . . 79
3.3.4 Dynamic analysis of morphology measures stability of drug
treatments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.5 Materials and Methods . . . . . . . . . . . . . . . . . . . . . . . . 87
3.5.1 Cell Culture . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.5.2 Free Floating Cells . . . . . . . . . . . . . . . . . . . . . . . . 88
3.5.3 Tethered Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.5.4 Confocal Microscopy . . . . . . . . . . . . . . . . . . . . . . . 89
3.5.5 Image Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.5.6 Metrics and Dynamics . . . . . . . . . . . . . . . . . . . . . . 99
3.5.7 Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4 Summary and Future directions 100
4.1 Summary and Discussion of Results . . . . . . . . . . . . . . . . 100
4.1.1 EB1 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.1.2 Suspended Cells . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.2 Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.2.1 Effects of Cancer Relevant Mutations . . . . . . . . . . . . . . 104
4.2.2 Direct Mechanical Measurements and Preliminary Optical Stretcher
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.3 Outlook and Final Thoughts . . . . . . . . . . . . . . . . . . . . . 114
Bibliography 117
vii
List of Figures
1.1 Cancer at Tissue and Cellular level. . . . . . . . . . . . . . . . . . . . 4
1.2 Metastatic cascade. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Initial cell attachment is dependent on McTN formation as affected
by tau. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.4 Microtentacles facilitate HMLE-endothelial attachment. . . . . . . . . 16
1.5 Models for force-induced modulation of cytoskeletal stiffness . . . . . 18
1.6 Cortex tension and cell shape. . . . . . . . . . . . . . . . . . . . . . . 22
1.7 Typical Cell has no microtentacles . . . . . . . . . . . . . . . . . . . . 23
1.8 Microtentacle model for loose actomyosin cortex . . . . . . . . . . . . 24
1.9 Microtentacle model for deforming actomyosin cortex . . . . . . . . . 25
1.10 Microtentacle model for breaking actomyosin cortex . . . . . . . . . . 26
1.11 Microtentacles form from unbalanced forces . . . . . . . . . . . . . . 27
1.12 Force-mediated regulation of integrin adhesions . . . . . . . . . . . . 30
2.1 Counterbalanced forces of microtubule expansion and actin cortex
contraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.2 Actin disruption promotes microtubule extension beyond the cell-
body boundary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3 Live-cell confocal microscopy of EB1-GFP . . . . . . . . . . . . . . . 42
2.4 Tracking algorithm identifies dynamic EB1-GFP tips and trajectories. 43
2.5 Analysis of EB1 Localization . . . . . . . . . . . . . . . . . . . . . . . 47
2.6 Average Velocity of EB1 Trajectories . . . . . . . . . . . . . . . . . . 52
2.7 Orientation Autocorrelation as a function of time for EB1 Trajectories 54
2.8 Distribution of Temporal orientation Autocorrelation as a function of
cumulative distance for EB1 Trajectories . . . . . . . . . . . . . . . . 55
2.9 Table of Orientation Autocorrelation Coefficients . . . . . . . . . . . 56
3.1 Lipid tethering and Image Analysis Techniques allows us to obtain
morphological attributes quantitatively . . . . . . . . . . . . . . . . . 72
3.2 Measurements of lateral cell drifting compare free floating cells to
tethered cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.3 Measurements of cell body attributes for Free-floating and tethered
cells. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
viii
3.4 Statistics of Free floating verses Tethered Cells metrics suggest that
tethered cells allow better visualization of microtentacles. . . . . . . . 78
3.5 Drug Panel shows all the Image analysis attributes for all 3 drug
treatments. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.6 Measurements of microtentacle attributes for different drug treatments. 82
3.7 Dynamic behavior is assessed by analyzing cumulative tip distance
and the ratio of full cell perimeter to cell body perimeter. . . . . . . . 84
3.8 Max Intensity of z-projections. . . . . . . . . . . . . . . . . . . . . . . 93
3.9 Full cell outline is the composite of analyses optimized for 3 distinct
cellular regions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.10 Results of processing steps optimizing for fine featured tentacles. . . . 95
3.11 Computing rotating Anisotropic filter. . . . . . . . . . . . . . . . . . 96
3.12 Rotating Anisotropic filter. . . . . . . . . . . . . . . . . . . . . . . . . 97
3.13 Composite rotating Anisotropic filter results. . . . . . . . . . . . . . . 98
4.1 Cytoskeleton States of 4 different cell lines . . . . . . . . . . . . . . . 109
4.2 Preliminary EB1 results Different Cell lines . . . . . . . . . . . . . . . 110
4.3 Optical Stretcher Results GFP vehicle vs Twist . . . . . . . . . . . . 113
ix
List of Abbreviations
α alpha
β beta
AMP Adenosine monophospate
AMPK AMP-activated protine kinase
CSC Cancer Stem Cells
CTC Circulating Tumor Cells
DMSO Dimethyl sulfoxide
ECM Extracellular Matrix
EMT Epithelial to Mesenchymal Transition
ENT Etinostat
EB1 End-Binding Protein-1
FITC Fluorescein isothiocyanate
GFP Green Fluorescent Protein
HMLE Human mammary epithelial cells
IREAP Institute for Research in Electronics and Applied Physics
McTN Microtentacles
MLCK Myosin Light Chain Kinase
MTOC Microtubule-Organizing Center
NF-κB Nuclear factor-kappaB
NMII Non-Muscle Myosin-II
NIH National Institute of Health
shRNA Small hairpin RNA
PTEN Phosphatase and tensin homolog
ROCK Rho-associated protein kinase
x
Chapter 1: Introduction
1.1 Cancer and Physics: a Historical Perspective
Humanity has known about cancer for thousands of years. The first docu-
mentation of cancer is found in the Edwin Smith Papyrus dating back to 3,000 BC
and describes breast cancer [4]. The word ‘cancer’ itself dates back to the ancient
Greeks’ father of medicine Hippocrates due to the way cancer moves about and has
crab-like projections [4, 5] (See Figure 1.1A). Since antiquity overall knowledge of
cancer has sophisticated greatly from a large macroscopic tissue perspective down
to the biomolecular level of signaling pathways. No timeframe has seen as extensive
changes and strides in the epistemology of cancer as the last 150 years. Only in the
this late modern period of the 1940s did cancer treatments begin to have less sur-
gical approaches and usher in radically new advances such as chemotherapy, novel
imaging techniques, and anticancer drugs [6].
As far back as antiquity and continuing into the more contemporary age of
research, cancer has been noted for its physical properties. In fact, it has been ar-
gued that the first involvement of physics in cancer also dates back to the ancient
Edwin Smith Papyrus papers which describe a cauterization technique using a fire
drill [7]. Cancer is universally recognized by patients to have mechanically stiff and
1
lump-like properties. The term ‘biophysics’, however, was not coined until Karl
Pearson′s publication of “Grammar of Science” book in 1892 [8]. Only 5 years later
in 1897, Emil Grubbe treated breast cancer with radiation for the first time [9, 10].
Eventually, many of those developments in radiation research formed the founda-
tions of what is now called medical physics and is usually applied to fields such as
radiation oncology, nuclear medicine, and radiology [11]. The field of ‘biophysics’
has grown to encompass a much broader scope of applications including migration,
environmental sensing, mechanosensing, gene regulatory networks, protein interac-
tions, electrophysiology, membrane channels, dynamics of self-assembly, soft matter,
molecular machines, etc.
Today, cancer continues to pose many interesting questions for physicists. Pos-
sibly no cancer process is richer in biomechanically interesting problems than the
metastatic cascade. Currently, most metastatic studies examine causal relations
between metastasis and biomolecular dominant mechanisms such as cell signal-
ing pathways, growth hormones, and genetics [12, 13]. Research in this inherently
motile phenomenon paints a picture of a plethora of mechanically driven systems
including motility, cell material properties (i.e. viscoelasticity), attachment, and
mechanosensing. Clinically, changes in mechanical properties in situ are associated
with higher metastatic risk; for example, increased collagen stiffness in human lu-
minal breast tumor tissue is correlated with a poor prognosis for survival [14]. Cells
with higher metastatic potential exhibit changes in mechanical phenotype both in
terms of mechanosensitivity [15] and material properties [16]. Perhaps cancer re-
search has spiraled full circle since antiquity, from tissue down to the cellular level,
2
as contemporary cancer research has recently discovered morphologically fingerlike
structures on the cellular level sufficient for cancer reattachment in breast cancer
cell lines called microtentacles (See Figure 1.1B). Microtentacles are hypothesized
to form as the result of unbalanced forces between microtubules and actin result-
ing in the extension of the plasma membrane beyond the actin cortex. A pubmed
search for cancer and cytoskeleton yields 16,037 publications; most of these studies
focus on migration though. A search for circulating tumor cells and cytoskeleton, on
the other hand, yields only 20 results, 10 of which examine microtentacles. Given
that microtentacles dominate the cytoskeletal literature for CTCs, and understand-
ing the mechanics of CTCs is arguably as important as understanding migration
in cancer research, I propose that understanding the dynamics and mechanics of
microtentacles is a critical biophysical question.
In today′s world of modern technology, the tools to probe cancer have grown
far beyond ancient fire drills and gross anatomy. With the advances of transfection
and fluorescent techniques, virtually any protein can be visualized; this, alongside
strides in microscopy that allow high-speed imaging at super-resolutions and com-
prehensive computational power for analysis, open up new scientific opportunities
never before explored. The goal of my thesis is better understand the mechanical
interplay between the actomyosin cortex and microtubules, and how these interac-
tions contribute to Microtentacle formation. I hypothesize that using drugs to alter
the actomyosin cortex or microtubule stability will result in differences in the pro-
trusiveness and dynamics of microtubule growth in both attached and suspended
cells.
3
A	
   B	
  
Figure 1.1: Cancer at Tissue and Cellular level. A. Tissue level illustration of Crab shaped
tumor shows finger-like protrusions Permission granted from Journal of Pioneering Medical Sci-
ences Blogs, 2014 [17]. B. Cellular level suspended Tumor cells show long spindly protrusions
called microtentacles. Early cellmask image taken by Eric Balzer.
4
1.2 Metastasis Causes Cancer Mortality
Cancer continues to be an epidemiologically significant disease responsible for
killing approximately 590,000 people every year [18]. Also, there are currently an
estimated 20 million people in the US who have been diagnosed with cancer [19].
This high number makes cancer the 2nd leading cause of death in the United States
[18]. Within that cancer mortality figure, metastasis is currently responsible for
90% of all those cancer fatalities [1–3]. Despite the fact that metastasis is the major
mechanism of death for the overwhelming majority of cancer, only 5% of cancer
research funding goes towards studies specifically investigating metastasis for both
basic and clinical research [20].
The metastatic cascade begins with a primary tumor in the host site where
tumor cells are pathological cells of the host tissue (See Figure 1.2). Next, tumor
cells begin to push out of the primary tumor and migrate, a process known as in-
vasion. Following, cells squeeze out of the primary tissue and into the blood stream
or lymphatic system in a process called intravasation. Once cells are circulating in
the bloodstream, they are known as circulating tumor cells or CTCs. Reattachment
happens when CTCs embed themselves on the endothelial cells. Finally, the patho-
logical cells will undergo extravasation or exiting the bloodstream and invading into
a distant tissue site. Once out of the bloodstream, the cancer cells colonize the
second organ forming a secondary or metastatic tumor.
While primary and secondary tumors tend to have similar genetic profiles,
there are often distinct differences in the genetic signatures and more mutations
5
in the primary than secondary tumors [21–23]. Predictably, secondary sites often
respond differently to drug therapies than the primary site behaves; the result is a
dearth in the efficacy of metastatic treatments [24]. In order for cancer therapies
to be more effective in preventing metastasis, more research studies that focus on
mechanisms specific to metastasis need to be conducted.
It is worth noting that the vast majority of cancers, in fact 90%, are carci-
noma types or comprised of epithelial cells; this is probably due to the fact that
epithelial cells frequently divide [25]. Interestingly, most epithelial cells die in the
bloodstream by a process called anoikis; anoikis is a form of cell suicide triggered
by loss of attachment to the extracellular matrix (ECM). Cells that don′t die by
anoikis are often ripped apart by the high magnitude of shear forces especially in
the small capillary tubes [26]. Overall, only an estimated .01% of CTCs are thought
to survive the bloodstream [27]. Even within research on the metastatic cascade
though, most studies focus on invasion and migration rather than the behavior of
CTCs or reattachment. With so many natural obstacles between when a cancer cell
enters the bloodstream until when it extravasates, this phase would seem a critical
rate-limiting stage within the metastatic cascade and an ideal place to target with
interventive therapies.
6
Primary Tumor
Invasion
Intravasation
CTC
Reattachment
Extravasation
MET/ Colonization
Micrometastasis
Cell Death
Figure 1.2: Metastatic cascade. Cancer begins as a primary tumor. Next, tumor cells begin
to push out of the primary tumor and migrate, a process known as Invasion. Following,
cells squeeze out of the primary tissue and into the blood stream or lymphatic system in a
process called Intravasation. Once cells are circulating in the bloodstream, they are known
as circulating tumor cells or a CTC. Most CTCs undergo Cell Death due to being ripped
apart from fluidic shear forces or an epithelial apoptotic process called anoikis. Reattachment
happens when CTCs embed themselves on the endothelial cells. Afterwards, the pathological
cells will undergo extravasation or exiting the bloodstream and invading into a distant site in a
process called MET/Colonization. Once out of the bloodstream, the cancer cells colonize the
second organ forming a Micrometastasis or secondary tumor.
7
1.3 Tumor Cell Circulation and Reattachment During Metas-
tasis
Currently, two of the most poorly understood steps in the metastatic cascade
are reattachment and extravasation, and there exist very few therapeutics target-
ing these steps. The extravasation cascade as a whole, can be broken down into 3
steps: loose attachment, tight attachment, and squeezing between the very tightly
connected epithelial cells lining the blood vessels called endothelial cells [28]. Thus,
for the purposes of this section, reattachment will be treated as a sub-step of ex-
travasation. For cancer specific extravasation, not a lot of basic research has honed
in on the exact mechanism by which a CTC manages to reattach itself to the blood
vessel walls, break through, and wedge itself in between those tightly connected
endothelial cells. Here, I review and summarize some of the existing research and
theories positing how CTCs manage to reattach and exit the blood stream as well
as why the discovery of microtentacles is significant.
1.3.1 Leukocyte extravasation
Perhaps one of the most studied forms of extravasation is the extravasation
of non-pathological leukocytes or white blood cells. Much of cancer extravasation
research studies leukocyte extravasation in the hopes of finding similarities between
the leukocyte extravasation and CTC extravasation. In healthy individuals, leuko-
8
cytes are responsible for dealing with inflammation and infections; to accomplish
this task, leukocytes travel through the bloodstream and must extravasate through
the endothelial cells to the site of injury or infection [28]. Leukocytes initiate contact
with endothelial cell adhesion molecules called selectins to form a transient connec-
tion with the endothelial cells [28]. Because this adhesion is weak, the bloodstream
continues to push and ultimately causes the leukocyte to ‘roll’. It is proposed that
this rolling behavior may be how CTCs also reattach [29]. It is also possible that
leukocytes act directly in cancer extravasation by forming a linker between tumor
cells and endothelial cells as shown in a study with neutrophil granuloctyes between
MDA-MB-468 and pulmonary epithelial cells [30]. More direct evidence of CTC
rolling behavior is demonstrated in one study by flowing MCF-7s across surfaces
functionalized with E-selectin [31]. Unlike leukocytes where L-selectin is established
as responsible for binding to the endothelial cells′ E-selectins, on the MCF-7 cells,
CD24 was implicated as the ligand responsible for rolling [31]. However, there
maybe be a considerable gap between E-selectin coated surfaces versus endothelial
E-selectin expression. Clinically, tumors with high levels of activated leukocytes are
associated with a more aggressive disease [27]. Furthermore, a rolling mechanism of
CTCs has not been observed in in vivo studies [26].
1.3.2 Platelet and CTC interactions
Interactions between platelet and CTCs interactions are also thought to play
a critical role in extravasation. It has been demonstrated that activation of platelets
9
increases metastatic potential [26, 32]. Platelets may play a role helping CTCs sur-
vive the bloodstream by attaching and preventing cell death mediated by natural
killer cell lymphocytes [27, 33]. There is also evidence that platelets protect CTCs
from shear forces directly [33]. Additionally, platelets may be involved in leukocyte
assisted reattachment. Evidence suggests that in places where there are high shear
forces that would make a rolling mechanism unlikely, platelets are involved in assist-
ing leukocytes [28]. In addition to helping CTCs survive the blood stream, studies
show that platelets can activate EMT pathways in circulating tumor cells [23]. Fur-
thermore, genetically modifying the platelet pathway or depleting platelets reduces
metastasis [23, 26, 33]. Clinically, an increase in platelet count is associated with
poor patient prognosis [23]. Also, for melanoma cancer, platelet behavior increases
metastasis in the lung [33].
1.3.3 Vascular permeability
Another important factor that may contribute to CTCs′ ability to exit the
bloodstream is vascular permeability. Endothelial cells use connections called Tight
junctions and Adherens junctions to control paracellular transport [34]. Indeed, the
endothelial junctions are so tight that it prevents transit of macromolecules larger
than 3 nm [34]. It has been shown that damaging the endothelial cells in mice can
increase metastasis [34]. Furthermore, studies show examples of cancer cells with
higher metastatic potential excreting factors known to disrupt endothelial cell-to-cell
junctions [23, 34]. On the other hand, only 20% of leukocytes use transcellular mi-
10
gration. Also it is unclear that an increase in vascular permeability is even necessary
for leukocytes to migrate through endothelial cells. [34].
1.3.4 Capillary Entrapment
Perhaps the simplest mechanism of reattachment proposed in the literature,
suggests that CTCs simply get trapped in narrow capillary tubes before extravasat-
ing [29]. This mechanism makes sense for places with narrow capillary tubes like
the lungs. Considering that lungs are the most frequent site of metastasis for breast
cancer [2, 24], this is conceivably an important factor for some metastases but not
all.
1.3.5 Microtentacles
While the above research shows many potential signaling pathways, none of
this research suggests a particular cell morphology identifying which CTCs are most
likely to metastasize. Additionally, research shows that under the shear forces of the
bloodstream, integrin mediated binding of tumor cells with endothelial cell selectins
is insufficient and suggests that the actomysin cortex or microtubule involvement
may be necessary [35]. It has been demonstrated that more malignant cancer cells
are more resistant to fluid shear forces [36]. More recently, morphological protrusions
called microtentacles (McTNs) have been discovered. Previous studies demonstrate
McTN protrusions poking in between endothelial cells indicating that a McTN-
mediated reattachment would most likely would assist CTCs between a loose and
11
tight attachment phases [37, 38].
1.4 Microtentacles Promote Tumor Cell Reattachment
One particularly interesting mechanism on how circulating tumor cells reattach
posits that tubulin-based structures called microtentacles or McTNs increase the
retention of circulating tumor cells in the capillaries of distant tissues (Figure 1.3).
Based on reviewing the McTN literature, the presence of McTNs appears to be a
sufficient condition for in vitro attachment [38–46]. Evidence also supports that
McTN-positive cells attach to endothelial cells [37, 38] (See Figure 1.4). Injection
of cells with McTNs into mice show that they get trapped and retained in the
lungs [47].
Previous research has implicated detyrosinated α-tubulin for the unique McTN
structure [38,48]. Detyrosinated α-tubulin is also called glu-tubulin due to the loss
of the c-terminal tyrosine resulting in the exposure of the glutamic acid residue.
Immunofluorescence of suspended cells stained for α-tubulin demonstrate that the
protrusions are made up primarily of microtubules [41]. Furthermore, Electron
Microscopy results showMcTNs comprised of several microtubules [49]. On a clinical
level this is relevant because an increase in the levels of glu-tubulin is an indicator of
poor prognosis [50]. Cell culture experiments demonstrate that glu-tubulin turnover
has a much longer persistence time on the scale of hours compared to α-tubulin which
only has a persistence on the scale of minutes [38,51]. Thus glu-tubulin is potentially
an important indicator of the possibility of McTN formation. Furthermore, it has
12
been demonstrated that McTNs are not the result of actin-based invadapodia [47].
From a biochemical and signaling perspective, pathways that stabilize tubu-
lin also stabilize McTNs. For example, vimentin is an intermediate filament that
colocalizes and aligns along microtubules which may help stabilize the tubulin
struts. [42]. Removing or inhibiting vimentin reduced McTNs and reattachment [42].
In a less direct technique, entinostat (ENT), a histone deacetylase inhibitor demon-
strated vimentin disassembly, weakening microtubules and causing a decrease in
McTN phenotype [45]. On a clinical level, vimentin has been identified as a marker
for breast cancer cells more likely to metastasize [42]. Anti-inflammatory agents
Parthenolide and costunolide destabilize tubulin, decrease detyrosinated tubulin,
and reduce McTNs [40]. Lastly, kinesin motor proteins are thought to cross-link
vimentin and glu-tubulin; inhibiting kinesin motors with tetracaine also inhibited
McTN formation [44]. Basically, McTNs can be increased by stabilizing micro-
tubules and decreased by depolymerizing microtubules.
In addition to microtubule manipulations, it appears that McTN formation
can also be modulated by manipulating the actin cortex. Unlike actin-based in-
vadapodia, McTNs are predominantly composed of microtubules, and do not require
actin support. Studies that remove actin polymerization by using Latrunculin-A
still have McTNs, [42, 44, 52]. Inhibiting actin cortex contractility indirectly via
Rho-associated kinase (ROCK) inhibitor Y27632 increases McTNs, while using Rho
Activator II decreases McTNs by making the actin cortex more contractile [39]. An-
other study used genetic modifications of c-Src to make the actomyosin cortex more
contractile or by making c-Src less active and weakening the actomyosin cortex [47]
13
The more contractile actomyosin cortex from overactive c-Src had less McTNs, and
weaker actomyosin cortex from inactive c-Src had more McTNs [47]. Additionally,
inhibiting AMP-activated protein kinase AMPK, activates actin severing protein
cofilin, weakening the actomyosin cortex causing an increase in McTNs and reat-
tachment [43]. Yet another example used shRNA to downregulate obscurins, de-
crease RhoA phosphorylation, ultimately decrease actomyosin contractility through
the myosin light chain causing an increase in McTNs [46].
Cells thought to have higher metastatic potential also have more McTNs. For
example, breast cancer cell lines scored positive for McTNs also positively correlated
with Cancer Stem Cell (CSC) markers [53]. Cells undergoing EMT have long been
of interest in metastatic research due to their association with invasion, and EMT
cells have a similar phenotype to CSCs [54]. Cells induced for EMT phenotype by
overexpressing transcripton factors twist and snail also increase McTNs [37]. Loss
of well-known tumor suppressor gene PTEN is also with associated high metastatic
potential and induces EMT [55–58]. Research shows that knocking down PTEN in
mammary epithelial cells induces McTNs [41].
14
Figure 1.3: Initial cell attachment is dependent on McTN formation as affected by tau. (a)
Over 1 h, MCF7 cells stably overexpressing GFP-tau (MCF-7tau1 and MCF-7tau2) attach
more efficiently (200 and 180%, respectively) compared with MCF-7 controls. Data (n=4)
represent the mean percentage of attachment±s.d. *P<0.01. (b) Over 1 h, siTau1 and siTau2
attach less efficiently (67 and 46%, respectively) compared with ZR-75-1 controls. Data (n=3)
are represented as the mean percentage of attachment±s.d.*P<0.01. (c) Glowscale of luciferase-
expressing cells injected into the tail vein of nude mice show equal initial signal retention in
regions of interest (ROI) emcompassing the lungs. Over a period of 8 days post-injection, MCF-
7tau1-luc cells are more retained than control MCF-7GFP-luc cells. (d). On average over all
injections, 61.8 and 44.4% of MCF-7tau1-luc cells (n=6) were retained in lung capillaries 1
day and 8 days post-injection. In comparison, 15.9 and 12.9% of control MCF-7GFP-luc cells
(n=5) were retained at the same time points. This difference was significantly different for
both 1 day (*P<0.02) and 8 days (**P<0.05) post-injection by a non-parametric t-test. Figure
reproduced with permission from Nature Publishing Group [59].
15
Figure 1.4: Microtentacles facilitate HMLE-endothelial attachment. Confocal imaging of GFP-
Membrane transfected HMLE cells suspended for 20 min over a confluent layer of mCherry-
labeled HBME. Top, angle, and side views of HMLE cells at the early stages of attachment
show HMLE-GFP vector control cells rounded without observable microtentacles. HMLE-
Twist displays a microtentacle anchoring to the top of the HBME (white arrow). HMLE-Snail
exhibits microtentacles extending under (white arrow) or bending toward (white arrowheads)
the HBME layer. Figure reproduced with permission from American Association for Cancer
Research [40].
16
1.5 Microtentacles from a Biophysical Perspective
McTNs are a particularly interesting phenomenon, not just for their potential
role in metastasis, but also as a physical phenomena that most likely stems from me-
chanical interactions within the cytoskeleton. Based on the biochemical alterations
outlined in previous studies reviewed above, it has been proposed that McTNs are
the result of unbalanced forces between the actin cortex and microtubules following
a tensegrity model [38]. The mechanical tensegrity model consists of stiff struts
that act as continuous compression members while cables connected to these struts
are in continuous tension and maintain the shape [60] (Figure 1.5A). Applied to a
biological cell, actin filaments act as the tension members, microtubules as com-
pression members, and apply these forces to anchor proteins (i.e. integrin) where
the cell adheres to the ECM [60] (Figure 1.5A). Thus, the tensegrity model is really
only applicable to attached cells and not fully applicable to CTCs. Furthermore,
the tensegrity model does not account for dynamically growing and shrinking mi-
crotubules, constantly changing actin cortex, nor the viscoelastic properties of the
cell [61]. Here, I discuss some the relevant cytoskeletal literature and present possi-
ble scenarios for how the tubulin and the actomyosin network interact mechanically.
I propose that ultimately, even though there are many possible scenarios, a simpli-
fied assumption that microtentacles form as the result of unbalanced forces remains
valid.
17
Figure 1.5: Models for force-induced modulation of cytoskeletal stiffness. (A) Tensegrity model:
Top left- A simplified version with compression struts and tensed cables exemplifying that stress
levels regulate cytoskeletal rigidity. Top right- In the cellular context, microtubules (gold rod)
apply compression on cell-matrix adhesions (represented by actin linking modules in pink and
integrin dimers) while the actin filaments (red) experience the cellular tension and hence stiffen
accordingly. (B) Semiflexible chain model is represented by the flexible actin cables (red) that
locally rigidify at points of stress application i.e. myosin (blue bundle) contraction. (C) Dipole
polarization model: Formation of contractile actomyosin dipoles is symbolically represented
by the arrow pairs. According to this model, they freely orient in response to applied stress
as experienced at a particular point. Adapted from Stamenovic D. Ingber DE. Tensegrity-
guided self assembly: from molecules to living cells. Soft Matter 2009; 5:1137-45. [DOI:
10.1039/B903916N] and [20816234]. Used by permission from MBInfo: www.mechanobio.info;
Mechanobiology Institute, National University of Singapore. [62]
18
1.5.1 Tubulin
Detyrosinated tubulin or glu-tubulin forms the basis of the microtubules com-
prising McTNs [37]. The outer diameter of a microtubule filament is 25 nm; tubulin
varies in strength and stability, and the contributions of tubulin detyrosination to
stabilized McTNs is unclear. In in vitro, for example glu-tubulin and α-tubulin have
similar polymerization rates [63]. In in vivo experiments, though, there is evidence
to suggest that detyrosinated tubulin depolymerizes more slowly [64]. However, glu-
tubulin alone does not increase in vivo microtubule stability [65]. Also, the cell can
quickly adjust detyrosination; for example, if glu-tubulin is microinjected into the
cell, it is quickly tyrosinated [66]. More likely, glu-tubulin gets its stability from
binding to intermediate filaments (IFs) like vimentin. It has been shown that vi-
mentin colocalizes to glu-tubulin as well as binds to glu-tubulin with 3 times the
affinity as α-tubulin [67,68]. Tau stabilizes microtubules by bundling them together
and can work competitively with taxol [69, 70]. It is also noteworthy that kinesin,
another tubulin crosslinker, prefers to bind to glu-tubulin [68]. Mechanically, buck-
ling experiments support the idea that microtubules bear significant compressive
cellular load [61,71].
1.5.2 Actomyosin Cortex and Model
The actomyosin cortex is a contractile meshwork near the cell membrane that
ranges in thickness from 50nm - 1 µm [72] (Figure 1.6). The actomyosin cortex is
a meshwork of actin filaments near the plasma membrane; mesh size can vary from
19
20 nm [73] to as large as 200 nm [74]. The actomyosin cortex is a dynamic system
where actin filaments are being continuously polymerized and breaking. Addition-
ally important, the actomyosin cortex is contractile where myosin walks along the
actin filaments and pulls the fibers towards each other (See Figure 1.5B and 1.6).
On a cellular level, one can think of the net effect of these moving actin filaments as
creating localized anisotropic contraction dipoles (See Figure 1.5C) and as having a
net pulling force toward the center of the cell (See Figure 1.6).
Thus, there is an actomyosin cortex that pulls inward and generates tension,
and compressive load-bearing microtubules pushing outward [15, 75, 76]. In a nor-
mal cell, the microtubule pushes against the actin meshwork, the actomyosin cortex
contracts along the membrane causing the membrane to pull inward, and no McTNs
form (See Figure 1.7). For McTN formation, I propose 3 possible ways in which mi-
crotubules can penetrate the actomyosin cortex: loose network, deforming network,
and a breaking network. In a loose actomyosin network, variations in meshwork size
are sufficient perhaps for a 25 nm microtubule to push through a 200nm hole and
beyond the actin cortex for cases of larger mesh size (See Figure 1.8) 200 nm [74].
Furthermore, in at least one immunofluorescent study with blebbistatin treated cells,
microtubule protrusion extension beyond the cell body boundary shows minimal or
sparse colocalization with F-actin [77]. Another possibility, is that the actomyosin
network exerts a weaker force, but gets stretched or deforms along the microtubule
(See Figure 1.9). Experiments showing colocalization of actin and McTNs might sug-
gest this mode [41]. At least one immunofluorescent study in attached cells shows
a thick, consistent layer of actin alongside microtubule protrusions [78]. Lastly, it
20
is possible that the actomyosin cortex would simply fail, either due to the strength
of the microtubules, brittleness of the actin cortex, or signaling that locally depoly-
merizes the actin filaments (See Figure 1.10). In at least one immunofluorescent
study of bacterial toxin on mammalian cells, microtubules protrude beyond the cell
body boundary and there are very distinct gaps of the actomyosin cortex where the
protrusions stick out suggesting breaks in the actomyosin cortex [79]. Another study
supporting this model shows that the actomyosin cortex breaks under tension from
a aspirating micropipette and creates a ‘window′ in the actomyosin cortex [80]. Re-
gardless of how the microtubules interact with the actomyosin cortex directly, this
process can be simplified to the idea that the cortex acts as a mechanical barrier
(See Figure 1.11).
21
Figure 1.6: Cortex tension and cell shape. (a) Actin filaments assemble into a thin network
connected to the cell membrane and undergoing continuous turnover. Myosin motors exert
forces in the network, giving rise to a cortical tension, T. Motors are assembled into mini-
filaments, which connect pairs of actin filaments and slide them with respect to each other.
This can result in contractile or expansile stresses, depending on the position of the motors
on the actin filaments (top right). Why contractile stresses dominate, giving rise to a positive
tension in the network, is not fully understood. Crosslinks in the network could contribute to
the generation of tension; crosslink turnover is a major determinant of cortex viscoelasticity.
(b) Because of the cell curvature, cortical tension gives rise to a hydrostatic pressure in the
cytoplasm. Gradients of motor-generated contractility within the cortex can drive tangential
flows of cortex in the plane of the membrane (left), whereas normal forces can drive cell
deformations with net displacement of the cytoplasm (right). Figure reproduced with permission
from Elsevier [72]
22
Typical	
  cell	
  without	
  microtentacles	
  
Figure 1.7: In a normal cell, forces between microtubules and actin cortex are balanced. Red
arrows represent pull of actin cortex toward center of cell; green arrow, microtubule pushing
outward; black arrow, membrane pulling back against microtubule (red-actomyosin cortex,
green-microtubules, black membrane).
23
Loose	
  actomyosin	
  cortex	
  
Figure 1.8: One hypothesis is that Microtentacles form when the actomyosin cortex is loose,
pushing through the meshwork. Red arrows represent pull of actin cortex toward center of
cell; green arrow, microtubule pushing outward; black arrow, membrane pulling back against
microtubule (red-actomyosin cortex, green-microtubules, black membrane).
24
Actomyosin	
  cortex	
  deforms	
  around	
  microtubule	
  
Figure 1.9: One hypothesis is that Microtentacles form when Actomyosin cortex deforms and
stretches around the microtubule. Red arrows represent pull of actin cortex toward center of
cell; green arrow, microtubule pushing outward; black arrow, membrane pulling back against
microtubule (red-actomyosin cortex, green-microtubules, black membrane).
25
Stabilized	
  microtubule	
  breaks	
  actomyosin	
  cortex	
  
Figure 1.10: One hypothesis is that Microtentacles form when Actomyosin cortex breaks and
the microtubule is allowed to move through. Red arrows represent pull of actin cortex toward
center of cell; green arrow, microtubule pushing outward; black arrow, membrane pulling back
against microtubule (red-actomyosin cortex, green-microtubules, black membrane).
26
Normal	
  Cell	
   Weak	
  
Actomyosin	
  
Stabilized	
  
Microtubules	
  
Ac#n	
   MT	
   Ac#n	
   MT	
   Ac#n	
   MT	
  
Figure 1.11: Microtentacles for when Actomyosin cortex is weak or tubulin is stabilized where
actomyosin cortex is in tension and microtubules are in compression.
27
1.5.3 Differences in adherent cells and suspended cells
The models outlined above are intended primarily for suspended cells, however
it is worth discussing some of the differences in mechanics between suspended and
attached cells. The tensegrity model is also mostly intended for attached cells but
makes the claim that cytoskeletal prestress exists for suspended cells [61]. Addition-
ally, the tensegrity model cites a study where increasing contact density to ECM
or substrate attachment causes an increase in cell stiffness [81] and argues that the
tensegrity model predicts cell softening for cells in suspension [61]. This is in direct
conflict with the studies in the literature that compare attached and fully detached
cells where measurements by both AFM and optical stretching show that detached
cells increase in stiffness [82,83]. Also in contrast with the tensegrity model are the
prevailing migration models where protrusions are actin based, form due to actin
polymerization, and push rather than pull the edge of the cell [84–86]. In the mi-
gration literature, the pushing of actin polymerization causes a decrease in local
membrane tension allowing the cell to move forward [86]. In one of the more con-
temporary mechanotransduction models, integrin acts as a clutch and stiffer ECM
substrates are able to provide a stronger reaction force and increase the push of the
plasma membrane outward (See Figure 1.12 ) [85].
In general, McTNs are not normally observed in attached cells, which is most
likely due to integrin anchoring unlike free-floating cells. While not outlined in my
model above, based on the migration′s actin polymerization model, I would sus-
pect that outward push of polymerization would lessen the net inward contractility.
28
This is consistent with some of the more recent schematics of actin protrusions,
where without adhesions, the net retrograde flow of actin would be higher than the
polymerizing flow [87]. As a result, there would be less opportunity for the plasma
membrane to deform around the microtubules.
29
Figure 1.12: Force-mediated regulation of integrin adhesions. Schematic of the ‘focal adhesion
clutch′ on stiff (a) versus soft (b) extracellular matrix (ECM). In all cases, integrins are coupled
to F-actin via linker proteins (for example, talin and vinculin). The linker proteins move
backwards (as indicated by the small arrows) as F-actin also moves backwards, under pushing
forces from actin polymerization and/or pulling forces from myosin II activity. This mechanism
transfers force from actin to integrins, which pull on the ECM. A stiff ECM (a) resists this
force so that the bound integrins remain immobile. A compliant matrix (b) deforms under this
force (as indicated by the compressed ECM labelled as deformed matrix) so that the bound
integrins can also move backwards. Their movement reduces the net loading rate on all the
force-bearing elements, which results in altered cellular responses. Reprinted by permission
from Macmillan Publishers Ltd: Nature Reviews Molecular Cell Biology, copyright 2014 [85]
30
1.6 Goals and Hypothesis
I show in my thesis that changes to the actomyosin network, change behav-
ior in microtubules in a manner consistent with the cortex acting as a barrier. I
hypothesize that by either weakening the actomyosin cortex or stabilizing micro-
tubules, there will be an increase in the distance of microtubule protrusions beyond
the cell body boundary in both attached and suspended cells, but also to a possibly
lesser extent in attached cells. I aim to develop quantitative assays to analyze the
dynamics of microtubule growth and morphology in both attached and suspended
cells in order to work towards determining distinct morphological signatures for
different types of interactions between the actomyosin cortex and microtubules.
I demonstrate these goals, first, by showing how drug treatments targeting
the actomyosin cortex impact microtubule dynamics for attached cells. Second, I
demonstrate how targeting the microtubules directly in suspended cells and causes
morphological changes to microtentacles. Throughout this process, I synthesize im-
age analysis tools optimized for microtubule growth in attached cells and cell outlines
in suspended cells. Additionally, I determine relevant metrics for morphology and
dynamics
31
Chapter 2: Microtubule growth dynamics of breast tu-
mor cells is altered by changes in actin net-
work structure and contractility
This chapter was adapted from Ory, E.C-h., et al, "Microtubule growth dynamics of
breast tumor cells is altered by changes in actin network structure and contractility",
submitted to Physical Biology, 2016. Ory proposed drug treatments Blebbistatin
and Latrunculin-A and Immunofluorescent experiments. Immunofluorescent exper-
iments were conducted by Martin. EB1 experiments and imaging were conducted
by both Martin and Ory. All analyses were implemented by Ory.
2.1 Summary (Abstract)
The periphery of epithelial cells is shaped by the balance of cytoskeletal physi-
cal forces generated by two dynamic force generating systems - growing microtubule
ends push against the boundary from the cell center, and the actin cortex contracts
the attached plasma membrane. Here we investigate how changes to the struc-
ture and dynamics of the actin cortex alter the dynamics of microtubules. Using
human MCF-7 breast tumor cells expressing GFP-tagged microtubule end-binding-
protein-1 (EB1) and co-expressing cytoplasmic mCherry, we analyze the trajecto-
ries of growing microtubule ends, and follow their position relative to the cell body
boundary. Actin depolymerization with Latrunculin-A reduces the growth speed
of microtubule ends, and increases the fraction of microtubule ends that protrude
beyond the cell body boundary, forming localized tubulin rich protrusions. Reduc-
32
ing contractility of the actin cytoskeleton upstream of myosin II via the Rho-kinase
inhibitor Y-27632 also enhances localized microtubule rich protrusions, but does
not slow down microtubule growth. Direct inhibition of myosin II with Blebbistatin
reduces microtubule growth speed and bends trajectories, but does not enhance lo-
calized protrusions. These results demonstrate that microtubule dynamics depend
on both actin network structure and dynamics.
2.2 Introduction
The delicate balance between the physical forces of microtubule growth and
actin network contraction that determines the shape of non-migratory cells is often
disrupted in diseases, in particular when epithelial cells transform into carcinomas
[38, 88]. Since more than 90% of human solid tumors arise as carcinomas from
epithelial cells [89], understanding this balance of microtubule growth and actin
network contractions in the context of epithelial cytoskeletal architecture has broad
potential implications.
Microtubule polymerization initiates from the centrosome, also known as the
microtubule-organizing center (MTOC), which is found in the center of eukaryotic
cells (See Figure 2.1). Extension of microtubules toward the cell surface is counter-
acted in several ways by the cortex of actin filaments that lies beneath, and is linked
to the plasma membrane [90, 91]. The mesh size of the actin cortex ranges from
20 to 250 nm [72], potentially providing a physical barrier for microtubule bundles,
since individual microtubules are 25nm in diameter. Furthermore, the end-binding
33
protein-1 (EB1), which is attached to the growing ends of microtubules, can also
bind to the actin cortex through association with the adenomatous polyposis pro-
tein (APC) [91, 92]. Thus microtubule ends couple to the actin cortex in multiple
ways, thus allowing for a balance of forces between pushing and contractility [38,90].
Unlike tensegrity networks with their balance of stationary tensile and load bearing
elements, the actin cortex is dynamic with actin polymerization and depolymer-
ization, and contraction mainly due to non-muscle myosin-II (NMII). The actin
cytoskeleton can also stabilize microtubules and allow them to bear larger com-
pressive forces within cells than is possible with purified microtubules in vitro [93].
Mechanotransduction in epithelial cells is known to be regulated by this balance
between microtubules and actin filaments, and alterations of cytoskeletal structure
and dynamics in epithelial tumor cells disrupt this balance [94, 95]. One important
manifestation of this altered balance is microtentacles (McTNs), discovered by one
of us as localized microtubule based protrusions that are found to protrude out from
the otherwise spherical body of circulating tumor cells [38,42,96].
Given the role of actin filaments in cell division and motility, actin has also
become a target for the development of cancer therapies [97,98]. Inhibition of actin
assembly and contractility are effective at reducing the growth and invasion of tu-
mor cells [99–101]. However, there is growing evidence that targeting the actin
cytoskeleton has consequences for other aspects of cancer progression. Toxins that
alter either microtubules or actin can disrupt the ability of epithelial cells to re-
spond to mechanical stimuli [102]. Compounds or genetic alterations that stabilize
microtubules [59, 96] or reduce actin integrity [47, 96] promote McTN formation
34
and increase the retention of circulating tumor cells in the lung capillaries of living
mice [47, 59]. Recently, it has also become clear that reducing actin contractility
can actually increase stem cell characteristics in epithelial tumor cells [103–105],
resulting in greater tumor-initiating capability.
Drugs altering the contractility of the actin cytoskeleton have recently demon-
strated efficacy in enhancing the ability of patient-derived tumor cells to grow in
vitro. Very recent studies show that even relatively short-term treatment with com-
pounds which reduce actin contractility by targeting ROCK (Y-27632) or myosin-II
(Blebbistatin) [106] can directly induce stem cell characteristics in epithelial tumor
cells [103] that promote long-term growth of patient-isolated cancer cells [104, 105]
and tumor formation in mice [103]. These results suggest that altering the mechan-
ical tension of epithelial tumor cells can regulate their ability to proliferate in vitro
and in vivo.
To improve our understanding of how altering actin integrity and contractil-
ity affects the cytoskeletal balance with microtubules, we use time-lapse confocal
microscopy to track the movement of of the microtubule end binding protein EB1
in the epithelial breast cancer cell line MCF-7. Since EB1 is tightly coupled to the
growing microtubule plus end [91, 107, 108], we infer microtubule growth from the
motion of EB1. The schematic shown in Figure 2.1 shows the three manipulations
of the actin cortex we use to investigate how actin integrity and contractility al-
ters microtubule growth dynamics. Normally, the actin cortex enhances membrane
tension [90] and resists growing microtubules from extending beyond the cell mem-
brane [91]. Complete disruption of actin polymerization with Latrunculin-A should
35
make it easier for microtubules to grow beyond the cell body boundary. Targeting
the Rho-kinase (ROCK) with the small molecule, Y-27632, inhibits activation of
myosin-II and thereby reduces contractility of actin filaments within the cortex [109].
Likewise, Blebbistatin binds directly to myosin-II and disrupts actomyosin contrac-
tility [90, 106]. We selected Y-27632 and Blebbistatin also since both were shown
to increase tumor stem cell characteristics [103–105] and aid proliferation of patient
derived tumor cells in vitro.
2.3 Results
2.3.1 Actin cytoskeleton integrity and contractility affects the structure
of the microtubule network
The microtubule network exhibits qualitative changes upon perturbations of
the actin network integrity and contractility using three different drugs: Latrunculin-
A [96], Blebbistatin [41, 106], and Y-27632 [52]. Network morphology and co-
localization of actin and microtubules were analyzed in fixed cells stained for DNA
with Hoechst 33342, F-actin with Alexa594-phalloidin, and α-tubulin monoclonal
antibody (DM1A). The three drugs did not significantly alter nuclear structure, but
observations of at least 25 cells per condition revealed qualitative changes in the
microtubule network. Figure 2.2 shows representative cells for control and the three
treatments.
The morphology of the MCF-7 cells under control conditions was consistent
36
with the relatively flat shapes associated with cultured epithelial breast tumor
cells [96]. The membrane edges appeared tightly anchored to the surface and the
cytoskeletal fibers were distributed throughout the cytoplasm. Polymerized actin
was enhanced in diffuse spots within the cytoplasm and near the nucleus, as well
as membrane ruffles at the cell periphery. Microtubule filaments were visible as a
network throughout the cytoplasm with some accumulation of α-tubulin filaments
near the nucleus.
In cells treated with 5µM of the actin depolymerizing compound, Latrunculin-
A [96], only minimal residual polymerized F-actin was observed with phalloidin
staining, as expected. Filaments of α-tubulin were distributed throughout the cell
body, as in control, but exhibited straight protrusions that do not contain F-actin
and extend beyond the cell body boundary. For cells treated with 25µMBlebbistatin
[41], F-actin clumped both within the cytoplasm and near the cell body boundary. α-
tubulin in these cells was denser near the nucleus. Small, straight local protrusions
containing both actin and α-tubulin were visible outside the cell body boundary.
Finally, in cells treated with 10µM of Y-27632 [52] the overall morphology of the cells
had fewer localized protrusions, but also had several regions with actin membrane
‘ruffles’ along the cell periphery, consistent with reports that Y-27632 can result in
Rac activation which is associated with laemellipodial membrane ruffles [110].
37
Figure 2.1: Counterbalanced forces of microtubule expansion and actin
cortex contraction. In a normal, adherent epithelial cell, there is a cortex
of cross-linked actin filaments beneath the plasma membrane that is un-
der contraction mediated by the non-muscle myosin-II motor protein (red
arrows). Microtubules are nucleated at the microtubule-organizing cen-
ter (MTOC) and grow outward from the cell center toward the plasma
membrane (green arrows). Microtubule end-binding protein-1 (EB1)
binds preferentially to the growing GTP-capped microtubule plus end
and forms comets projecting from the cell center. Microtubules reach-
ing the network of cortical actin filaments, which have an estimated
mesh size of 100nm. Interaction of EB1 with cortical proteins, such as
APC, also mediate capture of microtubules. Latrunculin-A depolymer-
izes actin, reducing the barrier to microtubule expansion. Phosphoryla-
tion of myosin-II by the Rho-kinase (ROCK) increases contractility, but
can be inhibited by the small molecule ROCK inhibitor, Y-27632. Sim-
ilarly, Blebbistatin binds directly to myosin-II and reduces contractility.
38
Ory	
  et	
  al.,	
  Figure	
  2	
  
DNA actin α-tubulin overlay
Vehicle
Latruculin-A
Blebbistatin
Y-27632
Latrun -­‐ 	
  
Figure 2.2: Actin disruption promotes microtubule extension beyond the cell-body bound-
ary. Human MCF-7 breast cancer cells were treated with vehicle control (0.5% DMSO), 5µM
Latrunculin-A, 25µM Blebbistatin or 10µM of Y-27632 for 30 minutes, fixed with formalde-
hyde and fluorescently stained for actin localization (red), microtubules (green) or DNA (blue).
Confocal microscopy images are shown for each channel and condition, along with an overlay
of all channels (right column)(scalebar = 10µm).
39
2.3.2 Quantifying microtubule growth dynamics via live cell imaging
In order to track microtubule tip growth, images of GFP-labeled EB1 proteins
were collected via fluorescence time-lapse microscopy [107, 108]. Taking one image
every 2 seconds allowed us to follow the growing trajectories of the labeled micro-
tubule plus ends. In control movies, EB1 clusters appeared uniformly distributed
across the inside of the cell (Figure 2.3-top). A maximum intensity time projection
written in Matlab of images across all 50 time points illustrated the microtubule tip
trajectories throughout the cell (Figure 2.3-bottom).
Blebbistatin treatment caused EB1 tips to move more slowly (Figure 2.3-top).
Trajectories were fainter and shorter than for other drug treatments or control (Fig-
ure 2.3-bottom), but quantitative tracking (Figure 2.4) was still possible for an av-
erage of more than 93% of EB1 tips with an automated cluster-finding and tracking
algorithm. The snapshots of Latrunculin-A treated cells showed little background
fluorescence, and EB1 clusters were scattered throughout the cytoplasm inside the
cell as well as outside the cell body boundary (Figure 2.3-top). The time-lapse pro-
jection images revealed long, straight trajectories, many extending beyond the cell
body boundary (Figure 2.3-bottom). Microtubule tip trajectories did not encircle
the nucleus, even though significant α-tubulin was seen in the immunofluorescence,
indicating that the α-tubulin around the nucleus may be more stable than mi-
crotubules in other regions of the cell (Figure 2.3-bottom). In cells treated with
40
Y-27632 [52], microtubule tips were scattered throughout the cytoplasm and near
the boundary, but scarce near the nucleus (Figure 2.3-top). The trajectories were
straight, and again scarce near the nuclei (Figure 2.3-bottom).
41
50 100 150 200 250
50
100
150
200
250
50 100 150 200 250
50
100
150
200
250
!"#$%&"'
50 100 150 200 250
50
100
150
200
250
50 100 150 200 250
50
100
150
200
250
50 100 150 200 250
50
100
150
200
250
()*+,-%,&$-'.'/0'
50 100 150 200 250
50
100
150
200
250
1&"22$3*)4-'5.'/0'
50 100 150 200 250
50
100
150
200
250
50 100 150 200 250
50
100
150
200
250
67589:5';<'/0'
''''
=$
->
&"
'$?
)>
"'
*7
@+
AB
"%
4A
-'
Ory	
  et	
  al.,	
  Figure	
  3	
  
Figure 2.3: Live-cell confocal microscopy of EB1-GFP allows imaging of microtubules dynam-
ics. (Top) Single image frame of EB1-GFP transfected cells for all 4 conditions. (Bottom)
Mean image across all 50 frames (scalebar = 10µm).
42
Ory	
  et	
  al.,	
  Figure	
  4	
  
Figure 2.4: Tracking algorithm identifies dynamic EB1-GFP tips and trajectories. A) Example
of EB1 tips found. B) Example of tips tracked. C) Individual trajectories: i. Normal trajectory
in cell bulk. ii. Trajectory in cell bulk treated with 25µM blebbistatin. iii. Trajectory in cell
bulk treated with 5µM latrunculin. iv. Trajectory in cell bulk treated with Y27632. v. Normal
trajectory near cell edge. vi. Trajectory in near cell edge treated with 25µM blebbistatin. vii.
Trajectory near cell edge treated with 5µM latrunculin. viii. Trajectory near cell edge treated
with Y27632. D) Trajectory overlays for normal cells where blue represents earlier time frames
and red represents later time frames (scalebar = 10µm). E) Trajectory overlays for cell treated
with 25µM blebbistatin where blue represents earlier time frames and red represents later time
frames. F) Trajectory overlays for cell treated with 5µM Latrunculin-A where blue represents
earlier time frames and red represents later time frames. G) Trajectory overlays for cell treated
with Y-27632 (10µM) where blue represents earlier time frames and red represents later time
frames.
43
2.3.3 Particle tracking microtubule end dynamics
The tip finding algorithm successfully picked out microtubule tips from images
(Figure 2.4A). We analyzed image sequences if the tip tracking algorithm tracked
at least 85% of visible tips for a minimum of 4 frames (Figure 2.4B), and at least
85% of tips moved more than 1 pixel. For control and drug treated cells, typical
trajectories are shown both in the cell bulk and near the cell body boundary (Figure
2.4C). For all 4 conditions, an overlay of all trajectories are shown in Figures 2.4D-G.
As expected, trajectories are generally in concordance with time-projected images
shown in Fig. 2.3, indicating that the microtubule tip finding and tracking algorithm
output was reasonable. The trajectories are color-coded to reveal the overall outward
motion of microtubules - blue indicated the starting-point of each trajectory and red
the endpoint (Figure 2.4D-G).
Control cells displayed microtubule growth throughout the cytoplasm, with
very few microtubule tip trajectories extending beyond the cell body boundary (Fig-
ure 2.4D). Likewise, Blebbistatin treated cells had short trajectories of microtubule
tips throughout the cytoplasm, consistent with the time-projected images (Figure
2.4E). For cells treated with Blebbistatin, the tracking algorithm found few EB1 tra-
jectories beyond the cell body boundary even though the fixed images show localized
tubulin containing protrusions extending beyond the cell body boundary. The lack
of actively growing microtubule tips in these Blebbistatin-induced protrusions sug-
44
gests that the microtubules extending the cytoplasmic edge were more stable in this
case.
In Latrunculin-A treated cells, the distribution of microtubule tips was similar
in fixed cells (Figure 2.2) and in live cell imaging (Figure 2.4F), indicating that
localized tubulin containing protrusions in this case did not contain significantly
more stable microtubules than other regions. Similar to Blebbistatin treated cells,
Y-27632 treated cells had short-lived tubulin growth near the edge of the cell body
boundary (Figure 2.4G). Beyond the cell body boundary, Y-27632 treated cells
exhibited localized α-tubulin protrusions. Though we found fewer protrusions com-
pared to blebbistatin treatment in both fixed imaging (Figure 2.2), we found more
protrusions than in blebbistatin treatment for live cell imaging (Figure 2.4G). This
suggests that in Blebbistatin-treated cells, localized tubulin containing protrusions
were stable, while Y-27632 and Latrunculin-A lead to more dynamic protrusions
with active microtubule growth.
2.3.4 Localizing microtubule tip relative to the cell body boundaries
Next, location of microtubule tips in relation to the cell body boundary was
measured. Analysis for control (0.1% DMSO) was comprised of 2 experiments from
19 independent cells at 950 frames (50 frames each) or 48,497 tip distance mea-
surements. The average distance of microtubule tips from the cell body boundary
measured 1.7 ±0.2 µm from the cell body boundary(Figure 2.5C). Similar analy-
45
sis of Latrunculin-A treated cells (2 experiments, 18 cells, 900 frames, 45,148 tips)
showed that the average distance of microtubule tips from the cell body boundary
decreased significantly to only 0. 16±0.2 µm compared to control (anova p<.0001
and ks-test p=1.0178e-6). Blebbistatin treated cells (2 experiments, 16 cells, 800
frames, 41,161 tips) had microtubule tip locations at an average distance compa-
rable to controls at 1.5±0.2 µm with no significant difference (anova p=.9397 and
ks-test p=.0528). Cells treated with Y-27632 (2 experiments, 19 cells, 950 frames,
44,821 tips) also exhibited a significantly smaller microtubule tip distance from the
boundary at 0.8±0.3 µm compared to control (anova p=.0263 and ks-test p=.0181).
Additionally, microtubule tip distance from the boundary for Latrunculin-A treated
cells was significantly less than Blebbistatin treated cells (anova p=.0004 and ks-test
p=6.4616e-6).
46
outside boundary bulk0
20
40
60
80
Pe
rc
en
t P
ar
tic
le
s
in
 R
eg
io
n
outside boundary bulk0
20
40
60
80
Pe
rc
en
t P
ar
tic
le
s
in
 R
eg
io
n
outside boundary bulk0
20
40
60
80
Pe
rc
en
t P
ar
tic
le
s
in
 R
eg
io
n
outside boundary bulk0
20
40
60
80
Pe
rc
en
t P
ar
tic
le
s
in
 R
eg
io
n
Control
Latrunculin
Blebbistatin
Yï27632
ï0.5
0
0.5
1
1.5
2
Di
st
an
ce
 fr
om
 B
ou
nd
ar
y 
(µ
m
)
Control Latrunculin Blebbistatin Yï27632
A	
  
C	
  
B	
  
Figure 2.5: Analysis of localization and distribution of dynamic microtubule ends. A) Cell
body boundary from mean mCherry image taken across 50 frames (left). B) Mean distance
from the boundary for all 4 conditions (bottom left). Control treated cells (.1% DMSO)
show an average distance from cell body boundary of measured 1.7 ±0.2 µm. Cells treated
with 5µM Latrunculin-A show an average distance of 0. 16±0.2 µm which is a significant
decrease when compared to control (anova p<.0001 and ks-test p=1.0178e-6). Cells treated
with 25µMBlebbistatin have an average distance of 1.5±0.2 µm comparable to controls with no
significant difference (anova p=.9397 and ks-test p=.0528). Cells treated with 10µM Y-27632
have an average tip distance of 0.8±0.3 µm which is significantly smaller than control (anova
p=.0263 and ks-test p=.0181). Additionally, microtubule tip distance from the boundary for
latrunculin treated cells was significantly less than blebbistatin treated cells (anova p=.0004
and ks-test p=6.4616e-6). C) Percentage of particles outside the boundary (left), near the
cell body boundary (within 10% of mCherry-defined cell body boundary) and in the cell bulk
(right).
47
2.3.5 Measuring the fraction of microtubule tips that are part of local-
ized protrusions
Next we quantified the fraction of growing microtubule tips located near the
cell body boundary, tips that were part of localized protrusions, and tips that were in
the cell′s interior (Figure 2.5C). Tips near the boundary were defined as those within
10% of the cell′s cytoplasmic area that is closest to the boundary. For control treated
cells the percentages of tips in the regions outside the cell body boundary, near the
cell body boundary, and in the cytoplasmic bulk were 3±2%, 19±2%, and 78±2%
where percentages for all 3 regions were significantly different (anova p<.0001). For
Latrunculin-A treated cells the percentages of tips in the regions outside the cell
body boundary, near the cell body boundary, and in the cytoplasmic bulk were
28±2%, 20±2%, and 52±2%. For cells treated with Latrunculin-A outside the cell
body boundary, the distribution had a skew of 12; thus, an additional ks-test was
conducted on all comparisons including the population of Latrunculin-A treated
EB1 tips outside the cell body boundary to check for significance. For Latrunculin-
A treated cells outside the cell body boundary, there was no significant difference
when compared to EB1 tips near the cell body boundary (anova p=.2663 and ks-test
p=.4255). For Latrunculin-A treated cells, there were significantly fewer EB1 tips
outside the cell body boundary and near the boundary compared to inside the cell
body bulk with anova computed p-values less than .0001 for both (ks-test p=2.9003e-
48
07 and p=3.8098e-08 respectively). For Blebbistatin treated cells the percentages
of tips in the regions outside the cell body boundary, near the cell body boundary,
and in the cytoplasmic bulk were 4±2%, 23±2%, and 73±2% where percentages
for all 3 regions were significantly different (anova p<.0001). For Y-27632 treated
cells the percentages of tips in the regions outside the cell body boundary, near the
cell body boundary, and in the cytoplasmic bulk were 12±2%, 33±2%, and 55±2%
where percentages for all 3 regions were significantly different (anova p<.0001).
For the region outside the cell body boundary, Latrunculin-A had a signifi-
cantly higher percentage of EB1 tips compared to all other regions control, Blebbis-
tatin, and Y-27632 (anova p<.0001 for all 3 and ks-test p=2.6693e-09, 1.3309e-08,
and 1.6487e-04 respectively). Additionally, while Y-27632 treated cells had signifi-
cantly less EB1 tips outside the cell body boundary, there were significantly more
EB1 tips outside the cell body boundary compared to control treated cells (anova
p=.0334 ks-test p=4.6625e-04). In the region closest to the boundary, only Y27632
had significantly more EB1 tip localization compared to control, Latrunculin-A, and
Blebbistatin (anova p=0.0011, 0.0031, 0.0391 respectively). Both the control and
Blebbistatin had over 70% of the EB1 tips in the cytoplasmic bulk and significantly
more EB1 tips in the bulk region compared to Latrunculin-A (anova p<.0001 for
both) and Y27632 (anova p<.0001 and 0.0008 respectively).
49
2.3.6 Measuring microtubule tip speed
In addition to quantifying EB1 tip localization, speeds of EB1 tips were mea-
sured to see whether actomyosin cortex integrity changed the speed of the EB1
tips for different drug treatments. For speed and dynamics experiments, videos
without mcherry staining were used. Analysis for control was comprised of 5 ex-
periments from 24 independent cells and a total of 4,979 individual EB1 tip tra-
jectories measured over time. The average speed of microtubule tips in untreated
cells was 0.075±0.002µm/s (Figure 2.6A). When broken down by regions outside,
boundary, and bulk, the average speeds were 0.057±0.003µm/s, 0.065±0.003µm/s,
and 0.084±0.003µm/s respectively. For cells treated with 5µM Latrunculin-A there
was a total 3 experiments, 22 cells and 3,703 trajectories where the average speed
was 0.065±0.004µm/s; when broken down by regions outside, boundary and bulk:
0.048±0.003µm/s, 0.058±0.004µm/s, and 0.071±0.005µm/s respectively. For cells
treated with 25µM Blebbistatin there was a total of 5 experiments, 18 cells, and
4,240 trajectories moving an average speed of 0.053±0.002µm/s; when the trajecto-
ries were broken down into their regions of outside, boundary and bulk, the speeds
were 0.037±0.003µm/s, 0.046±0.002µm/s, and 0.056±0.003µm/s respectively. For
cells treated with 10µM Y-27632, there was a total of 3 experiments, 25 cells, and
5,117 trajectories moving an average speed of 0.077±0.003µm/s; when the trajecto-
ries were broken down into their regions of outside, boundary and bulk, the speeds
were 0.059±0.004µm/s, 0.062±0.002µm/s, and 0.088±0.003µm/s respectively. For
cells treated with 25µM Blebbistatin, the average speed of EB1 tips moved sig-
50
nificantly slower than Control, 5µM Latrunculin-A, and 10µM Y-27632 (p<.0001,
p=0.0228, and p<.0001 respectively). Additionally, cells treated with Latrunculin-A
had an average EB1 tip speed moving significantly slower than trajectories in cells
treated with Y-27632 (p=.0130).
Within each cell treatment, the speed of trajectories was compared between
all 3 regions: outside the cell body boundary, near the cell body boundary, and
inside the cytoplasmic bulk. For the control and all drug treatments: Latrunculin,
Blebbistatin, and Y-27632 EB1 trajectories outside the cell body boundary moved
significantly slower than trajectories inside the cytoplasmic bulk (p<.0001, p=.0002,
p=.0001, and p<.0001 respectively). It is evident from this change in speed that
the membrane can exert a force that counteracts microtubule growth, even without
intact actin or actomyosin contraction. For the control and all drug treatments:
Latrunculin, Blebbistatin, and Y-27632 boundary region trajectories moved signif-
icantly slower than bulk region trajectories(p<0.0001, p=0.0475, p=0.0394, and
p<.0001 respectively), but no significant difference in speed between trajectories
near the boundary and those outside the cell body boundary. However the dif-
ferences in speed between the EB1 trajectories in the bulk region of the cell verses
boundary region of the cell were more pronounced in control conditions and Y-27632
compared to Latrunculin-A treated cells and Blebbistatin treated cells suggesting
that when contractility of the actomyosin cortex is compromised, there is less barrier
to slow down the speed of microtubule growth.
51
00.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Control Latrunculin Blebbistatin Yï27632
Sp
ee
d 
(µ
m
/s
)
outside boundary bulk0
0.05
0.1
Sp
ee
d 
(µ
m
/s
)
in
 R
eg
io
n
outside boundary bulk0
0.05
0.1
Sp
ee
d 
(µ
m
/s
)
in
 R
eg
io
n
outside boundary bulk0
0.05
0.1
Sp
ee
d 
(µ
m
/s
)
in
 R
eg
io
n
outside boundary bulk0
0.05
0.1
Sp
ee
d 
(µ
m
/s
)
in
 R
eg
io
n
Control
Latrunculin
Blebbistatin
Yï27632A	
  
B	
  
Figure 2.6: Average speed (µm/s) for all 4 conditions: control (black) at 0.075±0.002µm/s, 5µM
Latrunculin-A (purple), at 0.065±0.004µm/s, 25µM Blebbistatin (red) at 0.053±0.002µm/s,
and 10µM Y-27632 (yellow) at 0.077±0.003µm/s. B. Break down of speed for all 4 conditions
broken into 3 regions from left to right: outside the cell body boundary, near the cell body
boundary (the 10% of points inside the cell body closest to the boundary) and within the cell
bulk (remaining area inside the cell body): speeds for control (black) were 0.057±0.003µm/s,
0.065±0.003µm/s, and 0.084±0.003µm/s respectively; speeds for 5µM Latrunculin-A (pur-
ple), were 0.048±0.003µm/s, 0.058±0.004µm/s, and 0.071±0.005µm/s respectively; speeds for
25µM Blebbistatin (red) were 0.037±0.003µm/s, 0.046±0.002µm/s, and 0.056±0.003µm/s re-
spectively; and speeds for 10µM Y-27632 (yellow) were 0.059±0.004µm/s, 0.062±0.002µm/s,
and 0.088±0.003µm/s respectively. Error bars represent SEM.
52
2.3.7 Measuring microtubule tip trajectory straightness
To determine the straightness of trajectories of growing microtubules, the ori-
entation autocorrelation function was measured separately for each microtubule tip
trajectory. We measured both the change in direction of motion over time, which
yielded a characteristic turning time for growing tips, and the distribution of the
temporal orientation autocorrelation as a function of position, which yielded a char-
acteristic lengthscale over which trajectories bent. We analyzed all trajectories in 24
control cells, 22 cells treated with Latrunculin-A, 18 cells treated with Blebbistatin,
and 25 cells treated with Y-27632. The orientation correlations, averaged over all
trajectories for each condition (50 frames per cell) are shown as a function of time
delay (Figure 2.7) and as a function of cumulative distance (Figure 2.8) . We found
that microtubules in cells treated with Y-27632 maintained their direction similar to
control cells. In contrast, microtubules in Blebbistatin and Latrunculin-A treated
cells changed direction more quickly, and over shorter distances than control cells.
Note that Blebbistatin and Latrunculin-A treated cells also had microtubule tips
that move notably slower (Figure 2.6). An exponential fit of the time delay autocor-
relation function yielded a characteristic time and distance over which microtubule
tips change direction, as shown in Table 2.9. The higher τ and lower κ decay con-
stants in Control and Y-27632 indicate less change in direction and curving over
time and space.
53
0 2 4 6 8 10 12 14 16 18 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Lag, 6t (seconds) 
Co
rre
la
tio
n,
 C
(6
t)
 
 
Control
Latrunculin 5 µM
Blebbistatin 25 µM
Yï27632 10 µM
Ory	
  et	
  al.,	
  Figure	
  6	
  
Figure 2.7: Orientation Autocorrelation as a function of time for control (black squares),
Blebbistatin (red squares), Latrunculin-A (purple squares), and Y-27632 (yellow squares).
54
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Cumulative distance of EB1 tip (µm) 
Co
rre
la
tio
n 
di
st
rib
ut
io
n,
 l(
C(
6
t))
 
 
Control
Latrunculin 5 µM
Blebbistatin 25 µM
Yï27632 10 µM
Ory	
  et	
  al.,	
  Figure	
  7	
  
Figure 2.8: Distribution of Temporal orientation Autocorrelation as a function of cumulative
distance for control (black squares), Blebbistatin (red squares), Latrunculin-A (purple squares),
and Y-27632 (yellow squares).
55
Ory	
  et	
  al.,	
  Table	
  1	
  
Figure 2.9: Orientation autocorrelation coefficients as a function of time (top) and spatial
distribution of temporal orientation autocorrelation (bottom) for control, Latrunculin-A, Bleb-
bistatin, and Y-27632.
56
2.4 Discussion
Previous research suggests that microtentacles form in suspended cells when
the actomyosin cortex is compromised supporting a model where microtentacle for-
mation is the result of unbalanced forces between the actomyosin cortex and mi-
crotubules [39, 43, 46]. Here we present results that show, changes in microtubule
growth for attached cells as the result of changes in the actomyosin cortex. We
focused on the drugs listed above that target the actin cytoskeleton yet induce
changes in cell behavior that are strongly microtubule dependent (e.g. prolifera-
tion). The drugs are small molecule inhibitors of the actin cytoskeleton, both per-
turbing its integrity (Latrunculin-A) and contractility (Blebbistatin, Y-27632) [106].
Latrunculin-A binds directly to β-actin monomers, sequestering them and prevent-
ing their assembly into F-actin filaments [111, 112]. Blebbistatin and Y-27632 do
not alter actin polymerization, but reduce actomyosin contractility through inde-
pendent mechanisms. Blebbistatin binds to non-muscle myosin-II (NMII) and slows
down the release of phosphate that primes myosin-II for reattachment to actin fil-
aments [106]. As a result, NMII is blocked in an intermediate state with low actin
affinity, effectively releasing the actin-myosin cross-bridges that promote contrac-
tility [106]. The small molecule drug Y-27632 suppresses Rho-associated kinase
(ROCK) which in turn stimulates actomyosin contractility. Y-27632 competes with
ATP in the ROCK active site, reducing the kinase activity of ROCK and therefore
57
reducing NMII-induced actomyosin contractility [90]. Additionally, Y-27632′s inhi-
bition of ROCK results in the activation of cofilin an actin severing protein [101].
Therefore, while each of the three compounds target actin function, Latrunculin-A
broadly disrupts actin polymerization, Blebbistatin targets actomyosin contractility,
and Y-27632 targets both the actin filaments and actomyosin contractility.
Our analysis of microtubule structure and dynamics for the controls and three
drugs is carried out on MCF-7 breast tumor cells expressing the fluorescently tagged
microtubule end-binding-protein-1 (EB1) [107, 108]. Differences in localization and
amounts of all microtubules (Figure 2.2) and the actively growing microtubules mea-
sured via EB1 are used to infer microtubule stability. We find that microtubules
overall are more stable under Latrunculin A treatment, with few growing tips overall,
and a low concentration of tips near the cell body bulk region. Our results further
show that most active microtubule tips with EB1 are concentrated a few microns
from the cell body boundary (Figure 2.5A,C). Control cells showed less than 10% of
microtubule tips beyond the cell body boundary (Figure 2.5B), while Latrunculin-A
and Y-27632 increase localized protrusions, indicating that both the structural in-
tegrity and the dynamics of the cortex prevent localized protrusions of microtubules
through the actin cortex. Since microtubules will buckle and break when encoun-
tering a barrier [113], the relative softening of the actin cortex by Blebbistatin and
Y-27632 could produce a more flexible actin cortex that improves cortical capture
of growing microtubules, rather than breakage.
EB1 trajectories treated with Latrunculin-A move significantly slower than
control cells or cells treated with Y-27632. This suggest that the membrane ten-
58
sion beyond the cell body boundary is the largest inhibitor of speed and is further
supported by the fact that EB1 trajectory speeds are slower outside the cell body
boundary across all conditions where Latrunculin-A has significantly more EB1 tips
beyond the cell body boundary than all other conditions. Interestingly, Blebbistatin
moves significantly more slowly than any other conditions despite not having more
EB1 tips beyond the cell body boundary. It is worth considering that while other re-
sults provide evidence supporting the notion of growing microtubules contained by a
contractile actomyosin cortex, all the above experiments were conducted on attached
cells. One of the key, widely accepted models for attached cells posits that actin-
based protrusions form from actin polymerization pushing the plasma membrane
outward [114, 115]. Also, studies show that Blebbistatin weakens the protrusive
force of lamellipodia [116]. Furthermore, the speed of actin in the lamellipodia is
around speed 25nm/s and 2nm/s for retrograde flow [117]. The speed of the EB1 tips
in control cells was around 0.075±0.002µm/s and blebbistatin, 0.053±0.002µm/s or
a speed difference of about 22 nm/s and comparable to actin in the lamellipodia.
Straightness of microtubule tip trajectories is decreased by either actin de-
polymerization (Latrunculin-A) or the inhibition of actomyosin contractility with
Blebbistatin (Figure 2.7) In support of this model, previous studies have shown
that the cytoplasm provides a barrier to microtubule persistence length, compared
to in vitro microtubule growth [118]. Depolymerization of actin with Latrunculin-A
also decreased the straightness of the EB1 trajectories overall, but trajectories were
deflected in localized protrusions outside the cell body boundary.
Cells sense their mechanical microenvironment and the presence of wounds
59
through both actin contraction and microtubule extension [15, 38, 119]. Persistent
pulling of actomyosin contraction on both cell-matrix and cell-cell adhesion points
serve as signals for appropriate cell attachment [120]. Imbalances between actin cor-
tical contraction and microtubule extension from the cell center can also influence
metastatic potential through the formation of microtentacles on the surface of free-
floating tumor cells [38]. Reducing actin contractility with Blebbistatin enhances
microtentacles that promote tumor cell reattachment and aggregation [41]. Further
supporting a role for cytoskeletal balancing, microtentacles are induced by either
inhibition of actin polymerization with Latrunculin-A or hyperstabilization of mi-
crotubules with Paclitaxel [59, 96]. The importance of this microtubule-actin force
balancing is also demonstrated by several studies in the past year that have shown
that compounds which reduce actomyosin contractility, including Blebbistatin and
Y-27632, are capable of allowing long-term in vitro culture growth of patient-derived
tumor cell from many cancers [104, 105] using a new method termed ‘conditional
reprogramming′, because cells treated with cytoskeletal inhibitors divert from their
differentiation program and become capable of growing efficiently and for many gen-
erations [104,105]. These results suggest that some stem cell characteristics may be
regulated by mechanical changes in the cytoskeleton. In support of this model, both
Blebbistatin and Y-27632 induce stem cell characteristics in primary colon cancer
cells, including spheroid formation, CD44v expression and increased tumorgenicity
in mice [103]. If the balance of microtubule expansion against actin contraction is
partly responsible for these effects, it would be predicted that microtubule stabiliza-
tion would also be influenced by cell detachment and produce increased stem cell
60
characteristics. In fact, both increased microtubule stabilization upon cell detach-
ment [48] and the induction of stem cell characteristics by the common microtubule
stabilizing compound, Paclitaxel, have already been observed [121].
Currently, many clinical chemotherapeutic drugs target the cytoskeleton such
as Vinblastine, Paclitaxel, and Demecolcine. Cytoskeletal compounds which pro-
mote actin disruption or microtubule stabilization could inadvertently induce sur-
viving cancer cells to elevate microtentacles in circulating tumor cells, or localized
protrusions in attached cells, and increase stem cell characteristics. Overall our re-
sults show that it will be important to more clearly understand how proposed cancer
drugs alter the balance between microtubule extension and actin cortical contrac-
tion. We conclude that it will be important to understand this balance for a broad
range of cancer treatments.
2.5 Materials and Methods
2.5.1 Breast cancer cells and culture conditions
Human MCF-7 breast cancer cells, which retain many epithelial characteris-
tics [122, 123], were used to model the effects of actin disruption on cytoskeletal
balancing in epithelial carcinoma cells. MCF-7 cells were cultured in DMEM con-
taining 10%FBS and 1%Pen/Strep [59]. At 80% confluency, cells were detached
by trypsinization and replated onto dishes with glass coverslip bottoms to enable
imaging with confocal microscopy [59].
61
Drug reagents were obtained from Sigma Aldrich and concentrations were
based on previous studies: 5µM Latrunculin-A [37, 44], 25µM Blebbistatin [41, 52],
10µM Y-27632 [39,46]. Cell viability for the above concentrations were determined
by previous work in the Martin Lab employing a variety of different assays at 24-
hour time points. To determine whether drug treatment kills the cells by necrosis,
an XTT or CellTiter assay is frequently employed [37, 39, 44, 46, 53]. The XTT
assay works by staining for metabolically active cells. To determine whether the
drug treatments might induce apoptosis, a western blot of PARP cleavage is often
conducted [37,46].
An additional functional assay to consider when determining whether drug
treatments may cause cytotoxicity is a washout test. For a wash-out test, media
with drug treatment is removed and several subsequent exchanges of clean media
and removal are administered to remove drug exposure from the cells. Next, the
cells would be reimaged and the same analyses executed as control and treatment.
Finally, wash-out results are compared to control results to ensure that the drug
treatments had no additional adverse effects.
Drug reagents were obtained from Sigma Aldrich and concentrations were
based on previous studies: 5µM Latrunculin-A [37, 44], 25µM Blebbistatin [41, 52],
10µM Y-27632 [39,46]. Cell viability for the above concentrations were determined
by previous work in the Martin Lab employing a variety of different assays at 24-
hour time points. To determine whether drug treatment kills the cells by necrosis,
an XTT or CellTiter assay is frequently employed [37, 39, 44, 46, 53]. The XTT
assay works by staining for metabolically active cells. To determine whether the
62
drug treatments might induce apoptosis, a western blot of PARP cleavage is often
conducted [37,46].
2.5.2 Indirect Immunofluorescence
MCF-7 cells grown on glass coverslips were treated with vehicle (DMSO 0.1%)
or the indicated drug for 30 minutes, Blebbistatin (25µM), Latrunculin-A (5µM) or
Y-27632 (10µM). Cells were then fixed (3.7% formaldehyde/PBS, 15 minutes), per-
meabilized (0.25% Triton X-100/PBS, 10 min), and blocked for 1h [PBS/5% bovine
serum albumin (BSA)/0.5% NP40]. Immunostaining was performed for one hour at
4 ◦C with monoclonal α-tubulin DM1A (1:1000, Sigma) and Alexa488-conjugated
secondary antibody (1:1000, Invitrogen), along with Hoechst 33342 (1:5000, Sigma)
to label DNA and Alexa594-Phalloidin (1:200, Molecular Probes) to visualize poly-
merized actin. Images were acquired on an Olympus FV1000 laser scanning confocal
microscope (Olympus, Center Valley, PA).
2.5.3 Transfection and confocal microscopy
To observe the growth of microtubules, cells were transfected using Exgen-500
(Fermentas) with full-length APC-associated end-binding protein-1 (EB1) that was
fused at its C-terminus to green fluorescent protein (EB1-GFP) [108]. To compare
cells with relatively similar levels of EB1 expression, cells were imaged at a short
time point (16 hours post-transfection) and cells with comparably low levels of
particulate EB1-GFP fluorescence were used for analysis. Cells with bright and
63
diffuse cytoplasmic EB1-GFP were excluded from the analysis, since this level of
expression presumably saturates the microtubule binding sites for EB1. Cells were
imaged on an Olympus FV-1000 point scanning confocal microscope, and live images
were taken every 2 seconds in 50 sets of 256x256 image frames, using a 60x oil
immersion objective. GFP fluorescent signal was collected using excitation from a
488nm laser line (10% power) and FITC emission filter set (Olympus, 41001). For
boundary analysis datasets only, cells were co-transfected with mCherry to serve as a
non-targeted cytoplasmic marker. Time-lapse videos were collected from at least 19
cells for each of the 4 different treatment conditions in 2 independent experiments:
Vehicle control (0.1% DMSO), Latrunculin-A (5µM), Blebbistatin (25µM), and Y-
27632 (10µM).
2.5.4 Image analysis, particle tracking and computation
EB1 localizes into clusters bound to the tips of growing microtubules. Trajec-
tories of the EB1 clusters were computed using established particle tracking meth-
ods [124]. To remove noise, a bandpass filter was applied to the images with a lower
threshold at 1 pixel and higher bandpass threshold at 4 pixels. The position of each
cluster was determined with subpixel accuracy via peak fitting of the bright region.
To measure the dynamics of the microtubules in an image sequence, the locations of
clusters in successive images are linked as belonging to the same microtubule using
established maximum likelihood principles for particle tracking [124]. To reduce
noise from centroid computation, a moving average over three frames was applied
64
to the trajectory of each EB1 cluster, truncating the first and last location of each
cluster. Furthermore, only clusters that moved a minimum displacement of 1 pixel
were included in velocity vector computations. The average speed was calculated
by taking the mean of all speeds between two frames.
In order to find the cell body boundary, we modified image analysis methods
developed previously [125]. Since the cells do not move significantly, aside from lo-
calized protrusions, the boundary was measured from the mean image of 50 frames
in the cytoplasmic (mCherry) channel. This yields a cell body boundary that ignores
short lived cellular extensions (microtentacles, filopodia, etc.), i.e. such extensions
are considered ”outside” the cell body boundary for our further analysis. In our
analysis we did not include image sequences that were very noisy, where less than
85% of identified particles were successfully tracked; or less than 85% of the trajec-
tories showed notable displacements of ≥1 pixel. For further analysis, we defined
the distance from the boundary to be positive inside the cell body (cytoplasmic)
boundary, and negative outside the boundary of the main cell body, e.g. in McTNs.
Microtubule tip locations were characterized as outside the cell body, in the cell
interior, or near the boundary. The cell body boundary region is defined as the 10%
of the cell area closest to the boundary from the inside.
2.5.5 Statistics
For all statistical analysis, code was written in MATLAB. In order to determine
whether a the data had a normal distributions, measurements of skew and kurtosis
65
were conducted. Data was accepted as normal if the skew had a value ±2 and
the kurtosis was between 0 and 6. For comparing multiple conditions, anova tests
measure p-values. In the event that the data did not meet the criteria for a normal
distribution, an additional Kolmogorov-Smirnov (ks-test) was conducted in parallel.
2.6 Acknowledgements
This research supported by R01-CA154624 from the National Cancer Institute
and an Era of Hope Scholar award from the Department of Defense (BC100675).
WL was supported by NIH grant R01GM085574. We thank Lynne Cassimeris and
Michelle Piehl for generously providing the EB1-GFP expression vector.
66
Chapter 3: Extracting Microtentacle Dynamics in Non-
adherent Cells
This chapter was adapted from Ory, E.C-h., et al, "Extracting Microtentacle Dynam-
ics in Non-adherent Cells", In preparation for Oncotarget, 2016. Experiments were
proposed and conducted by Ory. Tethering slides were prepared by Chakrabarti.
First iteration of anisotropic filtering code was written by Chen. All other code and
analysis was conducted by Ory.
3.1 Abstract
The study of circulating tumor cells (CTCs) is one of the fastest growing fields
of cancer research and holds exciting potential for less invasive clinical research
and diagnostics. Using recently developed tethering techniques, we present a novel
image analysis framework for extracting the outlines of microtentacles (McTNs).
From full cell outlines, we were able to measure total number of microtentacles,
distance of McTN tips from cell body, as well as develop 2 total McTN phenotype
metrics. Having a robust, automated analysis of microtentacle phenotype allowed us
to measure dynamic morphological fluctuations by computing the autocorrelations
coefficients. We demonstrate that while microtubule stabilizing drug treatment
taxol increases total microtentacle phenotype, it reduces microtentacle dynamics.
This work has potential implications for developing techniques to quickly measure
drug response to small patient samples of CTCs.
67
3.2 Introduction
The study of circulating tumor cells is a rapidly growing field of research and
diagnostics [126]. Considering that 90% of cancer fatalities are the result of metas-
tasis [1], tumors surviving in the circulation is a rate-limiting step in the metastatic
cascade and presents a valuable opportunity for understanding patient prognosis
and preventing dissemination. Already, research has demonstrated that CTCs can
be detected early during cancer disease progression [126,127]; furthermore, a higher
CTC count is correlated with a poorer prognosis [126]. Better understanding of CTC
characteristics and reattachment mechanisms represent an underutilized approach
for improving patient diagnostics and drug therapies.
Historically, the incredibly low concentration of CTCs, which are as rare as
1 CTC in 100 million to 1 billion blood cells, has posed a technological hurdle to
further research and better understanding of the role of CTCs in metastasis [128],
Recently, an abundance of emerging technologies has improved the efficacy and
efficiency of capturing and segregating CTCs [126]. For example, Vortex Bioscience′s
technology now has a capture efficiency of 83% and processes up to 800 µL of blood
per minute [129]. The Celsee Diagnostics′s capture method has a sensitivity of 94%
and specificity of close to 100%. In these studies, between one and 2,457 CTCs
were captured in 2 mL of blood. [128]. Cellsieve has a capture efficiency of 98%
for chemically-fixed cells and 85% for live, unfixed cells [130]. It is now feasible to
capture 10 CTCs or more from a typical patient blood sample size.
Though CTCs can now be extracted from the bloodstream, further charac-
68
terization of the cells is very limited, in particular, characterization of the cells in
their native environment of suspension. Currently, the most frequent downstream
analysis only enumerates total number of CTCs or presence of particular biomark-
ers using immunostaining [126]. Most image analysis techniques for suspended cells
have focused on detecting and measuring immunofluorescence levels for a particular
biomarker. Due to the dynamic and motile nature of non-attached cells, few studies
have focused on time-lapsed single cell imaging.
Little is known about which circulating tumor cells succeed in surviving the
blood stream and ultimately forming metastases [131]. However, one likely mor-
phological phenotype of cell reattachment that was found in numerous metastatic
breast tumor cell lines is the presence of microtentacles, McTNs [42, 59]. McTNs
are tubulin-based protrusions found in detached cells; McTN positive cells reattach
to endothelial cells, get trapped in the capillaries of the lungs, and are a poten-
tial indicator for evaluating reattachment potential [38,42]. A higher microtentacle
number is found in more invasive breast cancer cell lines [42]. Furthermore, molec-
ular mechanisms that support McTNs are associated with increased metastasis and
poor patient prognosis [38,132].
Standard image analysis techniques cannot reliably capture the faint, thin
McTN structures since most standard fluidic systems for CTCs allow cells to float
out the field of view quickly, thus limiting data to snapshots of CTCs as they passed
by the imaging area. Thus, most prior work on McTNs relied on a blinded observer
to manually score the presence or absence of McTNs..
Here we use a novel cell tethering technique, recently developed and validated
69
by the Martin group that allows us to hold a cell in place within the field of view of
the microscope over long time periods, and thus enable extended time-lapse imaging
of McTN behavior [133]. To analyze these much larger datasets we adapted image
analysis approaches to quantify microtentacle number, microtentacle tip distance,
and their dynamics.
3.3 Results
As the study of circulating tumor cells progresses, there is a growing need
for techniques to analyze such unattached cells. We have previously shown that
tethering suspended cells is an effective technique for studying non-adherent tumor
cells [133]. The tethering technique attaches a small part of the cell’s membrane to
a surface while allowing the cell to retain its non-adherent characteristics. Here, we
added metrics to previous techniques further validating the benefits of cell tether-
ing quantitatively as well as demonstrating the geometric dynamics in response to
tubulin targeting drugs for tethered cells.
3.3.1 Anisotropic Filter allows us to capture outline of microtentacles
To test the ability of our image analysis to detect microtentacles, we used
MDA-436 cells, a mesenchymal triple-negative cell line with a high metastatic po-
tential known to form microtentacles. In order to compare free-floating cells with
tethered cells, microfluidics chambers were prepared with 2 different surface treat-
70
ments. For free-floating cells, microfluidic chambers were coated with pluronic F-
127, a generally cytophobic coating to prevent cell attachment. For tethered cells,
microfluidic chambers were coated with a cytophobic polyelectrolyte multilayers
(PEMs) followed by a lipophilic coating of DOTAP to engage the lipid membrane
while maintaining free-floating cell behaviors (Figure 3.1 A) [133].
Combining several existing image analysis techniques, we devised a framework
optimized for finding both the McTNs and cell body. There currently exist im-
age analysis techniques that are optimized for attached globular shapes as well as
techniques for stress fibers [134, 135]. Thus, we combined the most apposite pieces
of previously developed image analysis techniques, tailored, and adapted them to
finding full cell outlines that included McTNs. The most critical step for better ex-
traction of the faint, hairlike structures of McTNs was convolving the images with
a rotating anisotropic filter, taking the output of the rotating anisotropic filter, and
repeating the rotating anisotropic filtering for several iterations prior to thresholding
(Figure 3.1 B). Using our image analysis protocol, we were able for the first time
extract pertinent attributes of suspended cells including full cell outline, cell body
outline, and McTN tips (Figure 3.1C).
71
B	
  
C	
  
i	
   ii	
   iii	
  
i	
   ii	
   iii	
  
A B
Tumor	
  
cells	
  A BA BA	
  
i	
   ii	
   iii	
  
Figure 3.1: A. Lipid tethering and Image Analysis Techniques allows us to obtain morpho-
logical attributes quantitatively. A. Lipid Tethering Schematic picturing Ibidi microfluidics
chamber wells (i) coated with alternating coats of lipophilic and cytophobic polyelectrolyte
multilayer (PEMs) (ii) allows circulating tumor cells to attach to lipid tethers (iii). B. Image
analysis methods for full cell outline enhance microtentacle visualization by taking the max-
imum intensity profile of 5 stack z-projection for a particular time point (i) and undergoing
several iterations of anisotropic filtering (ii) before thresholding the results into a binary image
(iii). C. Attributes derived from image analysis consists of outline of the full cell including mi-
crotentacles (i), outline of the cell body′s boundaries excluding microtentacles (ii) and tips of
microtentacles (iii) derived from maximum local curvature of skeletonization of full cell outline
(scalebar = 10µm).
72
3.3.2 Tethering prevents cells from drifting and improves visualization
of microtentacles
Previous research demonstrated that tethered cells stay attached to the surface
after several washes better than free-floating cells [133]. In this study, image analysis
techniques allowed us to visualize drift of individual suspended and free-floating
cells qualitatively by looking at the maximum intensity over time, overlays of cell
body boundary as a function of time, and overlays of the centroid of the cell body
boundary as a function of time (Figure 3.2A). Computing total distance traveled
by cell body centroid, we demonstrated quantitatively that free-floating cells have
significantly more lateral drifting than tethered cells (Figure 3.2B) where t-test p-
value was 9.02e-15 and ks-test p-value was 5.5276e-09.
73
05
10
15
Free Floating Tethered
Total Distance traveled of Cell Body Centroid
 for 436 cells
Di
st
an
ce
 (µ
m
) 
A	
  
B	
  
iii	
  ii	
  i	
  
Te
th
er
ed
	
  
Fr
ee
-­‐fl
oa
0n
g	
   Cell	
  Body	
  C ntroid	
  Dri7	
  
v	
  iv	
   vi	
  
Original	
   Cellbody	
  Outline	
   Centroid	
  
Figure 3.2: Measurements of lateral cell drifting compare free floating cells to tethered cells. A.
Time projection profile of maximum intensity z-stacks for Free-floating cells (top) and Tethered
Cells (bottom). Overlays of cell body outline (ii) from initial (blue) to final (red) time points
for Free-floating cells (top) and Tethered Cells (bottom). Centroid of cell body (iii) from initial
(blue) to final (red) time points for Free-floating cells (top) and Tethered Cells (bottom). B.
Average total distance traveled by centroid of cell body for free-floating cells is greater than
tethered cells. Horizontal bar represents average across cells; shaded area, SEM; and individual
dots, mean per cell (scalebar = 10µm).
74
Using binary cell body image results from image analysis, the average cell
body area over time across all cells and cell body area variance per cell across time
was computed. Results showed that cell body area of free-floating cells was slightly
smaller than tethered cells but not significantly with a t-test p-value of .14 (Figure
3.3B). Because all z-stacks were centered at the largest part of the cell, the slightly
larger cell body area for tethered cells may indicate that free-floating cells were also
drifting vertically to smaller cross-sectional cell areas in the z-plane (Figure 3.3A).
Furthermore, free-floating cells had a significantly higher variance (ks-test p=.0496)
in the cell body area which further substantiated that the cells were moving slightly
out of plane along the z-axis (Figure 3.3C).
75
050
100
150
200
250
300
350
400
Cell Body Area for 436 cells
Ar
ea
(µ
m
2 )
Free Floating Tethered
215
220
225
Mean Variance for Cell Body Area for 436 cells
0
10
20
30
40
50
60
Free Floating Tethered
Va
ria
nc
e 
of
 C
el
l B
od
y 
Ar
ea
(µ
m
2 )
Cell	
  Body	
  Area	
   Cell	
  Body	
  Area	
  Variance	
  
C	
  
A	
  
B	
  
Z	
  0	
  (t=0)	
  	
  
Z	
  n	
  (t=n)	
  
A	
  0	
  
A	
  n	
  
A	
  0	
  	
  	
  >	
  	
  A	
  n	
  
	
  
Figure 3.3: Measurements of cell body attributes for Free-floating and tethered cells. A.
Schematic of cell′s cross-sectional Area at different z-planes. Cross-sectional area is largest,
when the slice crosses the center. B. Average cell body cross-sectional area of free-floating and
tethered cells has no significant difference p=.14. C. Variance of cell body area over time for
free-floating and tethered cells ks-test p=.0496. Horizontal bar represents average across cells;
shaded area, SEM; and individual dots, mean per cell
76
We extended our image analysis technique to capture quantitative McTN met-
rics for both tethered and free-floating cells. From our image analysis code, we es-
timated McTN length by measuring the distance of the McTN tip from the nearest
cell body boundary point. We found that tethered cells have a larger distance of
McTN tip from cell body boundary than free-floating cells with a t-test p-value .02
(Figure 3.4A). Another way we measured total McTN phenotype was by taking the
ratio between the full cell perimeter and the cell body perimeter; this allowed us
to compare McTN perimeter, while normalizing by cell size. Tethered cells exhib-
ited a higher ratio of full cell perimeter to cell body perimeter than free-floating
cells (t-test p= 8.9944e-06) suggesting that tethering allows one to better capture
McTNs than the free-floating technique (Figure 3.4B). For the interpretation of this
analysis we assume that the average McTN length and number should be the same
for tethered as for free-floating cells. While this is consistent with prior published
work, prior studies did not have the accuracy of our quantitative analysis. Thus we
plan to carry out manual analysis on the images to support this assumption more
carefully.
77
11.5
2
2.5
3
3.5
4
Free Floating Tethered
Ratio of Full Cell Perimeter to Cell Body Perimeter for 436 cells
Ra
tio
0.5
1
1.5
2
2.5
3
3.5
Free Floating Tethered
Average Distance of Protrusion tip from cell body boundary
Di
st
an
ce
 (µ
m
) 
Perimeter	
  Ra*os:	
  Full	
  Cell	
  to	
  Cell	
  Body	
  	
  
B	
  
Distance	
  of	
  Microtentacles	
  
A	
  
Figure 3.4: Statistics of Free floating verses Tethered Cells metrics suggest that tethered cells allow better
visualization of microtentacles. A. Average distance of microtentacle tips from cell body boundary for
free-floating and suspended cells (p=0.02). B. Average ratio of Perimeter of the full cell outline to cell
body outline for free-floating and suspended cells (t-test p= 8.9944e-06). Horizontal bar represents average
across cells; shaded area, SEM; and individual dots, mean per cell.
78
3.3.3 Image Analysis captures microtentacles qualitatively and quanti-
tatively of drug treatments
Once we determined that we were able to visualize and effectively quantify
more McTNs with the combination of cell tethering and image analysis, we used
tethered cells for systematic studies of McTN length and count. For our first analy-
sis, we selected drugs taxol and colchicine which have previously demonstrated the
ability to enhance or diminish microtentacles by respectively stabilizing or destabi-
lizing the microtubules that support McTNs. On tethered surfaces, we calculated
the attributes of cell body boundary, full cell outline, and microtentacle tips for
cells treated with the vehicle or 0.1% DMSO, 1 µg/mL taxol, and 125 µM colchicine
(Figure 3.5).
Deriving metrics from our image analysis attributes, we were able to quantify
statistically-significant differences for cells treated with vehicle, 1 µg/mL taxol, and
125 µM colchicine. Colchicine treated cells had significantly fewer McTNs than the
vehicle with an anova p-value of 0.0026 (Figure 3.6A). However, there was no signifi-
cant difference in the number of McTNs between vehicle and taxol treated cells (Fig-
ure 3.6A). Looking beyond McTN number per cell, the distance of the McTN tips
from the cell body boundary was significantly higher in taxol-treated cells compared
to vehicle-treated cells with an anova p-value of 0.0026 (Figure 3.6B). In addition to
the significant decrease in McTN number with colchicine, McTN tip distance was
79
also significantly lower (anova p =0.0001)in colchicine treated cells than in vehicle
treated cells (Figure 3.6B). A cell may be perceived as having a stronger McTN
phenotype either by increasing the number of McTNs or by increasing the length of
McTNs. One way that we measured the aggregate McTN phenotype, was to multi-
ply the number of McTNs by the average distance of McTN tips from the cell body
boundary per frame per cell; in essence, the cumulative McTN tip distance within a
frame. Taxol-treated cells had a significantly higher (anova p= 0.0151) cumulative
tip distance compared to vehicle (Figure 3.6C). Colchicine treated cells, on the other
hand, had a significantly lower (anova p=0.0018) cumulative tip distance compared
to control (Figure 3.6C). An additional metric we utilized to measure overall McTN
phenotype was to take the ratio between the full cell outline and cell body outline;
this method has the advantage of including the entire length and curve of McTNs
unlike the cumulative tip distance metric, but was still normalized to the size of the
cell body. For the ratio of outlines metric (Figure 3.6D), we found that taxol treated
cells had a possibly higher ratio than vehicle (t-test p=0.0325 anova p=0.1318) while
colchicine had a lower ratio than vehicle (t-test p= 5.4288e-04 anova p<.0001). Cu-
mulative tip distance and ratio, were more robust than number or distance metrics
independently as demonstrated by the statistical separation (Figure 3.6C and D).
80
original	
   Cellbody	
  Outline	
   McTN	
  outline	
   5p	
  
Co
lc
hi
ci
ne
	
  
Ta
xo
l	
  
Ve
hi
cl
e	
  
A	
  
B	
  
i	
   ii	
   iii	
   iv	
  
i	
   ii	
   iii	
   iv	
  
i	
   ii	
   iii	
   iv	
  
C	
  
Figure 3.5: Drug Panel shows all the Image analysis attributes for all 3 drug treatments. A. Max projection
of z-stack for MDA-436 cells treated with vehicle (i) is analyzed to find cell body boundary (ii), outline of
full cell, (iii) and tips of microtentacles (iv). B. Max projection of z-stack for MDA-436 cells treated with
1 µg/mL taxol (i) analyzed for cell body boundary (ii), outline of full cell, (iii) and tips of microtentacles
(iv) shows increase in microtentacles. C. Max projection of z-stack for MDA-436 cells treated with 125
µM colchicine (i) analyzed for cell body boundary (ii), outline of full cell, (iii) and tips of microtentacles
(iv) shows a decrease in microtentacles (scalebar = 10µm).
81
020
40
60
80
100
120
140
160
180
control taxol colchicine
Cummulative Distance of Protrusion tips from cell body
per frame for 436 cells
Di
st
an
ce
 x
 N
um
be
r (
µ
m
) 
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
5.5
Average Distance of Protrusion tip from cell body boundary
Di
st
an
ce
 (µ
m
) 
control taxol colchicine0
5
10
15
20
25
30
35
40
Number of Protrusions for suspended 436 cells
Nu
m
be
r
control taxol colchicine
0
0.5
1
1.5
2
2.5
3
3.5
4
Ratio of Full Cell Perimeter to Cell Body Perimeter for 436 cells
Ra
tio
control taxol colchicine
Number	
  of	
  Tips	
   Distance	
  of	
  3ps	
  
Cu ula3v 	
  Tip	
  Distance	
   Ra3o	
  Full	
  Cell	
  to	
  Cell	
  body	
  Perimeters	
  
A	
   B	
  
D	
  C	
  
Figure 3.6: Measurements of microtentacle attributes for different drug treatments. A. Average number
of microtentacle tips for cells treated with vehicle, 1 µg/mL taxol, and 125 µM colchicine. B. Average
Distance of microtentacle tips from cell body boundary for cells treated with vehicle, 1 µg/mL taxol,
and 125 µM colchicine. C. Average cumulative tip distance, calculated by multiplying total number of
microtentacle tips by the average distance of microtentacle tip from cell body per frame, is shown for cells
treated with vehicle, 1 µg/mL taxol, and 125 µM colchicine. D. Ratio of perimeters for full cell outline
to cell body boundary is shown for cells treated with vehicle, 1 µg/mL taxol, and 125 µM colchicine.
Horizontal bar represents average across cells; shaded area, SEM; and individual dots, mean per cell.
82
3.3.4 Dynamic analysis of morphology measures stability of drug treat-
ments
Having a technique to identify our attributes, allowed us to measure dynamic
fluctuations in response to drug treatment. Plotting cumulative distance traces
as a function of time for each individual cell, we observed that some cells have
more fluctuations in the cumulative tip distance than other cells (Figure 3.7A). The
mean cumulative distance for each drug treatment as a function of time, however,
was relatively stable suggesting a normal distribution (Figure 3.7A). In order to
measure the fluctuations, the autocorrelation coefficient of cumulative tip distance
was computed for different time intervals apart: 10 seconds, 20 seconds and 30
seconds. The data showed that up to 30 seconds apart, cells treated with taxol and
colchicine had less fluctuations than the vehicle for cumulatie tip distance (Figure
3.7B). Consistent with the autocorrelation coefficient of cumulative tip distance, the
autocorrelation coefficient of the ratio of full cell outline to cell body outline showed
less fluctuations in cells treated with Colchicine compared to control as far as 30
seconds apart. For Taxol treated cells, the autocorrelation coefficient showed less
fluctuations than control as far as 10 seconds apart.
83
10 20 30ï0.1
0
0.1
0.2
0.3
0.4
lag (seconds)
Co
rre
la
tio
n 
Co
ef
fic
ie
nt
Temporal Correlation Coefficient for 
of Cummulative Distance of Protrusion tips from cell body
Veh  Tax  Col Veh  Tax  Col Veh  Tax  Col 
10 20 30ï0.1
0
0.1
0.2
0.3
0.4
lag (seconds)
Co
rre
la
tio
n 
Co
ef
fic
ie
nt
Temporal Correlation Coefficient for 
of Ratio of Full Cell Perimeter to Cell Body Perimeter for 436 cells
Veh  Tax  Col Veh  Tax  Col 
Veh  Tax  Col 
0 50 100 150 2000
20
40
60
80
100
120
140
160
180
200
Cummulative Distance of Protrusion tips from cell body
for colchicine
time (s)
Di
st
an
ce
 x
 N
um
be
r (
µ
m
s)
 
0 50 100 150 2000
20
40
60
80
100
120
140
160
180
200
Cummulative Distance of Protrusion tips from cell body
for taxol
time (s)
Di
st
an
ce
 x
 N
um
be
r (
µ
m
s)
 
0 50 100 150 2000
20
40
60
80
100
120
140
160
180
200
Cummulative Distance of Protrusion tips from cell body
for control
time (s)
Di
st
an
ce
 x
 N
um
be
r (
µ
m
s)
 
Vehicle	
   Taxol	
   Colchicine	
  
A	
  
B	
  
i	
   iii	
  ii	
  
Cumula1ve	
  Tip	
  Distance	
  
Cu ula1ve	
  Tip	
  Distance	
  
C	
  
Ra1o	
  of	
  Full	
  Cell	
  Perimeter	
  to	
  Cell	
  Body	
  
	
  	
  
)	
  	
  
Figure 3.7: Dynamic behavior is assessed by analyzing cumulative tip distance and the ratio
of full cell perimeter to cell body perimeter. A. Time traces or cumulative tip distance for
individual cells cells treated with vehicle (i), 1 µg/mL taxol (ii), and 125 µM colchicine(iii).
Bold time trace is average cumulative tip distance over all individual cells. B. Fluctuations
of cumulative distance is shown by computing the autocorrelation coefficient at time lags
0 to 30 seconds for cells treated with vehicle, 1 µg/mL taxol, and 125 µM colchicine. C.
Fluctuations of Ratio between full cell outline and cell body boundary is shown by computing
the autocorrelation coefficient at time lags 0 to 30 seconds for cells treated with vehicle, 1
µg/mL taxol, and 125 µM colchicine.
84
3.4 Discussion
Even with the recent advances made in CTC isolation, the small quantity of
cells isolated may be as little as 1-10 CTC/mL of blood [129]; this presents major
limitations on which downstream analysis techniques are feasible. Desirable analysis
techniques like flow cytometry can require as many as 10,000 cells for statistically
relevant results. Furthermore, CTC-derived patient xenografts (CDX) have had a
very low success rate [136]. Currently, the only FDA approved technique for ana-
lyzing CTCs is CELLSEARCH. Unfortunately, CELLSEARCH can only enumerate
cells and requires fixation; thus, limiting parallel and downstream analyses. Our
technique demonstrates the ability to yield statistically significant results with as
few as 19 cells without killing the cells, potentially enabling downstream or parallel
cancer assays.
Current techniques for McTN analysis require double-blinded studies where
the researcher tediously counts the McTNs by hand. Such techniques are time-
consuming and potentially lack uniformity. For the first time, we are able to evaluate
McTN number automatically and systematically. We introduce multiple quantita-
tive measurements of cell phenotype such as distance of McTN tip from cell body
and number of McTNs. Previous research based on qualitative assessment of positive
or negative McTN phenotype concluded that taxol appeared to increase mostly the
length of McTNs and a combination of latrunculin and taxol appeared to increase
85
the number of McTNs [59,96]. In this study we were able verify quantitatively that
taxol increases the McTN behavior by specifically increasing the length of existing
McTNs rather than number of McTNs in general. Such distinctions may pave the
way towards determining whether an increase in the number of McTNs or the length
in McTNs is more likely to increase CTC reattachment.
The most promising techniques for determining targeted, personalized thera-
pies include using patient derived xenografts (PDX) and circulating tumor DNA or
ctDNA [136]. However, in patients where the primary tumor is too small, PDX is
not always an option. PDX has a highly variable success rate ranging from 23-75%
depending on tumor types; for breast cancer specifically, 23% [137, 138]. Thus, the
PDX model has a high risk of losing precious patient samples and not gaining any
information. Additionally, the PDX model can take up to 6 months to establish and
2 years for a complete drug study. ctDNA, on the other hand gathers tumor DNA
from the bloodstream and has the advantage of being noninvasive. While ctDNA
analysis can predict drug resistance in some patients, the prediction is purely correl-
ative. The tethering and image analysis technique allows us to quickly and directly
test drug response.
For the first time ever, we were able to estimate morphological stability by
measuring the autocorrelation coefficients of metrics for total McTN phenotype.
The lower fluctuations in drug treatments of our autocorrelation results suggest that
morphological stability is consistent with biochemical stability. Previous research
shows that cells from EMT-induced cell lines have more McTNs, higher reattachment
rates and embed themselves into endothelial cell layers. [37]. Cell lines rich in
86
McTNs due to tau-induced microtubule stabilization trap more efficiently in the
lung capillaries of living mice [59]. Neither McTN dynamics of metastatic cells lines
has been analyzed yet, nor whether cells treated with McTN stabilizing drugs are
more likely to get stuck in the capillaries. While cells with more McTNs have higher
reattachment, less dynamic McTNs may suggest cells are less able to extravasate
through endothelial cells and out of the bloodstream. Future work should explore
whether drug induced McTN cells that form less dynamic McTNs are more likely
to get trapped in the capillaries of distant tissues in vivo, as well as whether the
McTNs are more or less dynamic in different breast cancer subtypes or different
stages of metastatic progression.
The ability to automatically and quantitatively compare circulating tumor
phenotypes has potential applications in improving basic understanding of the role of
McTNs mechanism and reattachment, studying patient CTCs, and perhaps selecting
appropriate drug therapies for patients. Measuring McTNs is a fast assay that may
give insight to cancer progression and drug response without the complications of
culturing cells like PDX′s which can take up to 2 years for a drug study.
3.5 Materials and Methods
3.5.1 Cell Culture
Human MDA-MB-436 cells derived from a metastatic pleural adenocarcinoma
were obtained from the American Type Culture Collection were used for all experi-
87
ments. Human MDA-MB-436 cells were selected as a cell model for metastatic po-
tential and presence of microtentacles [96]. MDA-436 cells were cultured in DMEM
media containing 10% Fetal Bovine Serum and 1% penicillin/streptomycin. Cells
were detached from cell culture plates at a minimum of 80% confluency using trypsin.
For drug treatments, all reagents were obtained from Sigma Aldrich and con-
centrations were based on previous studies. For microtubule stabilization, 1.2 µM
Taxol was administered, and previous studies determined cell viability using XTT,
CellTiter 96, and PARP assays [40, 46, 52, 133]. For microtubule destabilization,
125 µM Colchicine was administered, and previous studies determined cell viability
using CellTiter 96 and PARP assays [40,52].
3.5.2 Free Floating Cells
For the experiments involved in suspended free floating cells, ibidi microfluidics
chambers were coated with 1% pluronic F-127 solution for 30 minutes. Cells were
treated with a 1:10,000 dilution of CellMask-Orange (Life Technologies) membrane
stain in order to visualize microtentacles. Next, cells were treated with the vehicle
or a drug treatment of 1 µg/mL taxol or 125 µM colchicine. A 150 µL sample of
treated cells was added to each ibidi channel at a concentration of 30,000 cells per
channel. Cells were incubated at 37C to allow absorption of Orange mask and drug
treatment for 30 minutes prior to imaging.
88
3.5.3 Tethered Cells
For tethered cell experiments, cytophobic polyelectrolyte multilayers (PEMs)
deposited on microfluidic substrates were used to prevent tumor cell adhesion, and
the addition of lipid moieties to tether tumor cells to these surfaces through in-
teractions with the tumor cell membranes (Figure 3.1A). Cells received the same
treatment of Cell-Mask Orange and drugs as free floating cells.
3.5.4 Confocal Microscopy
All imaging was conducted on an Olympus FV-1000 confocal at a 60x magnifi-
cation. For videos of suspended cells, a set of five 0.5 µm/slice z-stacks were imaged
every 6.5 seconds for a total time series of 20 z-stacks. For tethered cells, z-stack
slices were 1 µm thick and stacks were imaged every 10 seconds. In all cases, the
middle z-slice was calibrated along the z-axis to where the cell appeared largest.
3.5.5 Image Analysis
For each time point, a max intensity image of each z-stack was computed, all
further processing was derived from max intensity images (Figure 3.8). Outlines
for the cell-body and full-cell were computed separately. Cell-body outlines were
identified by using image analysis methods published previously [134].
In order to get clear outlines of the microtentacle features, we modified and
89
combined previously published image analysis techniques optimized for cell shape
along with techniques optimized for stress fibers by using a rotating anisotropic
filter [134,135]. Consequently, analysis for full cell outlines processed and optimized
parameters for 3 distinct cellular regions separately: microtentacles, bright cell-
body border and globular base of protrusion region, and the cell center. The full
cell outline was derived from a binary image comprised of the 3 distinct analyses.
(Figure 3.9).
Due to the fact that tentacle features were significantly dimmer than the cell
body and filamentous rather than globular, the first analysis was optimized specif-
ically for the McTNs. First, a 2 x 2 median filter was applied to the maximum
z-projection per time-point in order to give a very fine-featured, localized smoothing
optimized for approximately half the width of microtentacles (Figure 3.10A). Follow-
ing, the output underwent and initial rough convolution with a rotating anisotropic
filter that will be described in more detail below (Figure3.10B). After the initial it-
eration of rotating anisotropic filtering was applied, contrast adjustment algorithms
in matlab were optimized to make the protrusions rather than cytoplasm uniformly
white (Figure 3.10C). Additionally, to specifically extract the microtentacle features,
the output underwent another iteration of rotating anisotropic filtering; multiple
iterations of the rotating anisotropic filtering were repeated using the combined
output of the previous series of filtering.(Figure 3.10D) [135]. The 60 anisotropic
filters consisted of convolving a Laplacian with a 60 Gaussian kernels at different
angles (Figure 3.11). The contrast adjusted image (Figure 3.10C) is convolved with
anisotropic filters at 60 different angles (Figure 3.12). Each individual anisotropic
90
filter emphasizes alignment along a different angle (Figure 3.13). Next, the max
projection of all 60 anisotropic filter results was computed and another pass of con-
trast adjustments (Figure 3.13). The image underwent several iterations of rotating
anisotropic filtering followed by compiling them. Finally, the resulting anisotropic
results was linearly multiplied with the initial rotating anisotropic results before the
contrast was adjusted again and then thresholded (Figure 3.10E and F) Because
the rotating anisotropic filter selects preferentially for line-like features occasionally
truncating, rather than intersecting with the cell body or potentially incorrectly
biasing cell-body curvature near the base of the protrusion, a second analysis opti-
mized for features including and near the cell body boundary was conducted inde-
pendently. For this second region, the initial image underwent a matched filtering
technique originally designed for retinal segmentation (Figure 3.9B) [139]; this tech-
nique showed preference for the base of the protrusions. The retinal segmentation
technique was image multiplied with the initial anisotropic filtering results prior to
being inputed into the previously establish local, global curvature technique [134].
Lastly, to prevent the analysis optimized for tentacles from creating an annular out-
line, a rough estimate of the cell center was computed by using the matlab built
in function of imfilter to blur the image, thresholding, and then using matlabÕs
bmorph to remove any spuring and erode shape to prevent it to contributing to cell
outline information (Figure 3.9C). Cell center analysis did not require any contrast
optimization. Once all 3 analyses were complete, results were added together for a
composite binary image (Figure 3.9D) and cleaned of debris (Figure 3.9E). Finally,
the composite binary image was inputted into an active contour algorithm as an
91
initial estimate (Figure 3.9S2F).
Once the outline of the full cells perimeter was computed, the tips were com-
puted by skeletonizing the binary microtentacle image. From the images′ skeletons,
tips were selected by locating the coordinates of maximum local curvature.
Attributes derived from image analysis consist of microtentacle inclusive out-
line of the full cell, outlines of the cell body exclusively, centroid of cell body, and
the tips of the microtentacles. From these attributes, measurements of microtenta-
cle behavior were derived including, area of the cell body, variance in the cell body
area, total distance traveled by the centroid of the cell body boundary, ratio of the
perimeter of the full cell perimeter to the cell body perimeter, distance of micro-
tentacle tips from cell body perimeter, number of microtentacle tips. Additionally,
cumulative tip distance was measured by multiplying the number of microtentacles
by the average distance of microtentacle tips from the cell body boundary per frame
per cell. All attributes and metrics were computed in matlab.
92
Max	
  projec+on	
  over	
  Z	
  
5	
  z-­‐slices	
  	
  1	
  μm/slice	
  	
  	
  
Figure 3.8: Max Intensity of z-projections. For the 5 z-slices at each time point (top), a max
intensity image of each z-stack was computed (bottom).
93
Composite	
  of	
  3	
  
Outline	
  of	
  
microtentalces	
  
Fine	
  features	
   Globular	
  features	
   Cell	
  center	
  
Composite	
  Clean	
  Composite	
  
B	
  A	
  
D	
  
C	
  
A	
  
A	
  
A	
  
F	
  E	
  
Figure 3.9: Full cell outline is the composite of analyses optimized for 3 distinct cellular regions.
A. Binary image results for algorithms optimizing for fine featured tentacles. B. Binary image
results for algorithms optimizing for globular features. C. Binary image results for computing
a rough cell center. D. Binary image composite of 3 binary results from different cell regions.
E. Binary image composite cleaned of debris. F. Outline overlay of final results on initial image
includes microtentacles.
94
McTN	
  preprocessing	
  
Ini/al	
  median	
  
	
  filtering	
  
Ini/al	
  
Anisotropic	
  filtering	
  
Contrast	
  adjustment	
  
Mctn	
  filtering	
  
(mul/ple	
  anisotropic	
  filters)	
  
Contrast	
  adjustment	
  Linear	
  mul/ply	
  with	
  
ini/al	
  anisotropic	
  filtering	
  
A	
   C	
  B	
  
F	
  D	
   E	
  
Figure 3.10: Results of processing steps optimizing for fine featured tentacles. A. Median
filter was applied to the maximum z-projection results B. Results from first convolution with a
rotating anisotropic filter routine called ‘pixelalign′ C. Contrast adjustment results optimizing
contrast for fine featured tentacles D. Results from multiple iterations of a rotating anisotropic
filter. E. Results from linear multiplication between intial and final rotating anisotropic filter
results. F. Final contrast adjustment results prior to thresholding.
95
2 4 6 8 10 12 14 16
2
4
6
8
10
12
14
16
0.5 1 1.5 2 2.5 3 3.5
0.5
1
1.5
2
2.5
3
3.5
Laplacian	
   60x	
  line	
  kernel	
  with	
  Gaussian	
  
2 4 6 8 10 12 14 16
2
4
6
8
10
12
14
16
60x	
  Anisotropic	
  filter	
  
Figure 3.11: Computing rotating Anisotropic filter. A Laplacian kernel (top left) is convolved
with 60 Gaussian line kernels (top right) at 60 different angles to get 60 different Anisotropic
filters.
96
2 4 6 8 10 12 14 16
2
4
6
8
10
12
14
16
Adjusted	
  
Contrast	
  input	
  
Adjusted	
  
Contrast	
  input	
  
Convolved	
  Image	
  
Result	
  
60x	
  
Figure 3.12: Rotating Anisotropic filter. The contrast adjusted image (top left) is convolved
with the 60 different anisotropic filters (top right) resulting in 60 filtered results biased at
different angles (bottom right).
97
N=15/60	
   N=30/60	
   N=45/60	
   N=60/60	
  
Max	
  Composite	
   Composite	
  imadjust	
  
A	
  
B	
   C	
  
Figure 3.13: Composite rotating Anisotropic filter results. A. Results of Anisotropic at different
angles B. Maximum of composite of anisotropic results C. Contrast adjustment results of
anisotropic filtering.
98
3.5.6 Metrics and Dynamics
To better understand the dynamics of the cells, the fluctuations of the cumu-
lative tip distance and ratio of full cell perimeter to the cell body perimeter as a
function of time, was measured by computing the autocorrelation coefficient in 10
second increment lags 0 to 30 seconds. Autocorrelation code was written in matlab.
All dynamic measurements were only conducted on tethered cells.
3.5.7 Statistics
All statistical analysis was implemented in MATLAB. Determination of normal
distribution was calculated by measuring skewness and kurtosis where any distribu-
tion of normal with a skewness of ±2 or kurtosis 0-6 was considered close enough
to normal to apply standard statistical analysis. To measure p-values between two
conditions (ie free-floating vs tethered), the built in MATLAB ttest2 was used; for
comparing across multiple conditions, ANOVA analyses were implemented. In the
event of a non-normal distribution, an additional Kolmogorov-Smirnov (ks-test) was
conducted to compare 2 conditions.
99
Chapter 4: Summary and Future directions
4.1 Summary and Discussion of Results
Prior to the work outlined in my thesis, the only existing techniques for mea-
suring McTNs consisted of making use of double-blinded studies where the evaluator
scored samples for the absence or presence of McTNs. Additionally, no McTN stud-
ies had quantified the microtubule protrusions in attached cells under actomyosin
cortex perturbations. Lastly, no studies had measured morphological fluctuations
of McTNs under tubulin drug perturbations.
4.1.1 EB1 Results
In chapter 2, I showed that manipulating the actin cortex changes the growth
dynamics of microtubules in attached cells. I demonstrated that particle tracking
algorithms could successfully be adapted to biological EB1 tips. Results support
the hypothesis that the actomyosin cortex acts as a barrier. When the actomyosin
barrier is removed by latrunculin or severed via Y-27632, EB1 tips are allowed
to move further beyond the cell body boundary compared to control as measured
both by average distance from the cell body boundary and percent of EB1 tips
outside the cell body boundary. EB1 trajectories move slower outside the cell body
100
boundary compared to inside the cell body boundary across control and all drug
treatments suggesting that the membrane is the larger inhibitor of speed. Speed
of EB1 trajectories in Blebbistatin treated cells move the slowest across all regions.
Due to the fact that in attached cells there is a protrusive force outward as the result
of actin polymerization [116], this may outweigh the inward contractile force of actin
for attached cells. Lastly, EB1 tips in cells treated with Blebbistatin move the least
straight followed by those treated with Latrunculin. EB1 tips in cells treated with
Blebbistatin and Latrunculin also move more slowly.
As demonstrated by the speed of the EB1 tips beyond the cell body bound-
ary, membrane tension most likely places a significant force on microtubule growth.
It would be interesting to repeat the above experiments and decrease the mem-
brane tension using sucrose. Decreasing the membrane tension should theoretically
increase the ability of EB1 tips to move beyond the cell body boundary for con-
trol and Blebbistatin treated cells. Additionally, a reduction of membrane tension
should increase the overall speed and straightness of EB1 trajectories in Latrunculin-
A. Similar experiments have been conducted to study the effect of membrane tension
on clathrin polymerization, where decreasing the tension either by micropipette as-
piration or sucrose solution increased clathrin polymerization [140]. Another study
shows that decreasing membrane tension in migrating cells results in loss of polar-
ity and more spread out cells; additionally increasing membrane tension caused a
decrease in actin polymerization [141].
Finally, some additional considerations should be discussed regarding the lim-
itations of the drug treatments. First, imaging the GFP tagged EB1 tips required
101
exposure to blue light and research has shown that blue light exposure can inacti-
vate Blebbistatin in in vitro cells [142]. Thus, the efficacy of Blebbistatin treatments
may have had a limited effect on the cells. On the other hand, because these studies
were conducted on a confocal microscope, blue light exposure was confined to a
small area of the cell at a time and for a maximum total exposure of only 2 minutes,
so diffusion may have maintained continued drug exposure. Also, Y-27632 has some
off-target effects like inhibiting PRK2, a kinase involved in cell to cell adhesion [143].
It is unclear how PRK2 inhibition would impact the above results.
4.1.2 Suspended Cells
In chapter 3, I show that manipulating tubulin stability manipulates McTNs
geometry and dynamics. Successful analysis techniques were developed for analyz-
ing the full outline of cells including McTNs. Results show that tethering cells does
not significantly alter McTN length; however, tethering allows visualization of more
McTNs than the free-floating method. It is worth noting, however, that the com-
plete absence of attachment in free-floating cells may increase the contractility of
the actomyosin cortex which could also conceivably repress McTNs [82,83]. Stabiliz-
ing tubulin by using Taxol treatment significantly increases the distance of McTNs
tips from the cell body, and destabilizing tubulin using Colchicine treatment signifi-
cantly decreases distance of McTNs tips from cell body. In order to measure overall
McTN morphology, I proposed using 2 composite McTN metrics: cumulative tip
distance of McTN tip from cell body boundary and the ratio of the full cell outline
102
to the cell body outline. Both metrics measure an increase in McTN phenotype
for Taxol treated cells compared to control and a decrease in McTN phenotype for
Colchicine treated cells when compared to control. All results are consistent with
the model where an increase in tubulin stabilization enhances the ability for micro-
tubule protrusions to move beyond the cell body boundary and the converse results
for microtubule destabilization. Based on EB1 study above, the average distance
of microtubule tip from the cell body boundary in attached cells is 1.6158 µm±.08
compared to 2.2907 µm±.13 in tethered cells. This is consistent with the hypoth-
esis where a lack of traction force in suspended cells decreases the ability of actin
polymerization to push the plasma membrane outward; thus, allowing the plasma
membrane to deform more easily around the microtubule protrusion.
In my experiments looking at suspended cells, I also measure morphological
dynamics and find that both drug treatments Taxol and Colchicine cause less mor-
phological fluctuation compared to untreated cells. These results may have impli-
cations for how CTCs reattach; previous research studies show that Taxol increases
reattachment in impedance assays [40]. However, CTCs in vivo have the additional
hurdle of getting through an endothelial layer of cells. It is plausible that McTNs
might require a requisite amount of flexibility and motion in order to push McTN
protrusions between endothelial cells.
A conceivable next step, might be to conduct the same dynamics measure-
ments on tethered cells with the same actomyosin targeting drugs discussed in the
EB1 experiments. I would propose that for all drug types (Latrunculin-A, Bleb-
bistatin, and Y-27632), there would most likely be an increase in the length of
103
McTNs, and that Latrunculin-A with its complete lack of actomyosin cortex may
additionally increase the total number of McTNs. Results may have interesting
implications if the morphological fluctuations in drug treated cells increase or are
comparable to control conditions. I expect that while stabilization of microtubules
may decrease morphological fluctuations, destabilizing the actomyosin cortex would
preserve fluctuations if not increase them.
4.2 Future Directions
4.2.1 Effects of Cancer Relevant Mutations
Work from my thesis demonstrates changes in microtubule dynamics as a
result of drug treatment. A natural next step would be to make the same mea-
surements but expand from drugs to genetic modifications. Here, I propose us-
ing 4 different genetic modifications spanning a spectrum of different cytoskeletal
states and metastatic capabilities. The parental line used as control for these ex-
periments is MCF10A; MCF10A is a non-tumorgenic mammary epithelial cell line
well-established in the literature as a model for normal breast cell behavior [144].
The genetic modifications would include MCF10A with a patient mutation in the
phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit alpha or PI3CA;
MCF10A with tumor suppressor gene pten homogenously knocked out (PTEN-/-);
and MCF10A with the transcription factor twist overexpressed. These choices of
cell lines will serve as an in vitro cell line model that separates, non-tumorgenic
104
(control), invasiveness (PIK3CA-H1047R), McTN formation (pten -/-), and cancer
stemness (twist).
In terms of McTNs, MCF10As form few McTNs. It has been shown that
MCF10A pten(-/-) exhibits McTNs, while MCF10A PIK3CA produce very few
McTNs [41]. MCF10As overexpressing twist have not been tested for McTN pres-
ence, but McTNs are expected. Previous research demonstrates that ectopic ex-
pression of twist in MCF10As induces cancer stemness [145]; also, overexpression of
twist in another non-tumorgenic breast cancer cell line HMLE induces McTNs [37].
Within the mechanical framework of McTNs, stronger more stabilized micro-
tubules contribute positively to McTNs, while weaker tubulin contributes negatively.
Glu-tubulin has a long persistence time, slow turnover rate, and is potentially an
important indicator of McTN formation [51, 59, 146]. Of the 4 cell lines, MCF10A,
pten(-/-), and PIK3CA-H1047R are known not to have high levels of glu-tubulin;
however, MCF10A cells transfected with twist are expected to have higher levels of
glu-tubulin based on experiments that use other cell lines transfected with twist [37].
Expression of vimentin is another potential indicator of microtubule strength and
is known to colocalize and align with glu-tubulin [42] which may help stabilize the
McTNs. Furthermore, in the clinical scope vimentin has been identified as a marker
for breast cancer cells more likely to metastasize [42]. Based on previous research
in the Martin lab, only the cell line MCF10A with PIK3CA-H1047R of the 4 cell
line models is known to express significant levels of vimentin; we hypothesize that
MCF10A with overexpression of twist will also express significant levels of vimentin
based on research of other cell lines overexpressing twist [37]. The vimentin contri-
105
bution to McTNs formation is not completely clear, however. From rheological in
vitro studies, a direct interaction between actin and vimentin has demonstrated a
stiffer actin network [147]; this would contribute negatively to McTN formation and
may explain why there are less McTNs in the case of PIK3CA mutation.
The second half of the mechanical framework of McTNs, is the actomyosin cor-
tex where a more contractile actomyosin cortex contributes negatively to McTNs and
a weak actomyosin cortex contributes positively to McTNs. There are 4 main ways
to weaken the actomyosin network: either by cutting actin filaments, depolymer-
izing actin, weakening actin crosslinking, or weakening myosin-based contractions.
As demonstrated in my EB1 research, removing the actomyosin cortex network as
opposed to weakening it has distinctly different growth dynamics. Cofilin, for ex-
ample is an actin severing protein that would be expected to weaken the actomyosin
cortex [148]; The only cell line known to contain high levels of activated cofilin is
MCF10A pten(-/-) cells (See Figure 4.1). Another important consideration of the
actomyosin cortex is the myosin light chain; the myosin head binds to the actin
filaments and allows the actomyosin cortex to contract. Currently, we don′t have
direct information about the myosin light chain′s activation state, but based on the
presence or absence of McTNs described above, we have formed hypotheses of the
myosin light chain′s activation state. Myosin light chain is predicted to contribute
to a tighter more contractile actin cortex in MCF10A cells with PIK3CA mutation
and twist. For MCF10A control cell line, myosin light chain is predicted to be
contractile but significantly less contractile than cells with PIK3CA mutation and
twist. Lastly, myosin light chain is expected to cause a loose and non-contractile
106
actin network in MCF10A pten(-/-) (See Figure 4.1).
Based on all the above information, we hypothesize a positive or negative
contribution of the actin cortex state and microtubules towards McTN formation
(See Figure 4.1). Using this qualitative model, we hypothesize whether McTN
formation is driven more by stable tubulin or weaker actin. For MCF10A, the
actin cortical tension and microtubules counteract each other, so very few McTNs
form. For MCF10A pten(-/-), the microtubule behavior is normal, but the absence
of strong actin cortical tension results in McTNs. For MCF10A PIK3CA-H1047R
cells, the vimentin aligns with the microtubules potentially making things stiffer,
but vimentin interactions with actin can also make for a stiffer network; thus, actin
and microtubule interactions still cancel each other out. For MCF10A with twist
overexpression, the glu-tubulin in the microtubules drives the cell to form McTNs.
It is expected that since pten (-/-) cells have a looser actin network compared
to MCF10A cells, EB1 tips will move slower on average than MCF10A cells in a
manner similar to Latrunculin-A treated cells. EB1 trajectories in Pten (-/-) cells
might move slower near the cell body periphery due to the cell membrane acting
as a barrier. Different from pten(-/-), PIK3CA-H1047R mutation has an increased
expression of vimentin, but is also thought to have a stronger actomyosin cortex.
It is expected that EB1 trajectories in PIK3CA-H1047R activated cells move at a
slower speed and move less straight due to the actomyosin cortex acting as a barrier.
However, because of the potentially vimentin-stabilized microtubules, overall length
of trajectories may be longer. As twist-transfected cells are hypothesized to have a
weaker actin cortex but stronger tubulin growth due to the increase in glu-tubulin,
107
EB1 trajectories are expected to move faster in both the bulk and cell-body periph-
ery of the cell compared to control cells. Cells transfected with twist are expected to
have slower moving EB1 tips in the cell body periphery, due to a more active actin
cortex. EB1 tips in twist-transfected cells are also expected to move well beyond
the cell body boundary.
I have conducted preliminary experiments for MCF10As, pten (-/-), and PIK3CA-
H1047R. Preliminary results demonstrate successful EB1 transfection and clearly
defined EB1 tips. By doing a time average of the videos, we were additionally able
to see the 3 phenotypes qualitatively for EB1 trajectories (See Figure 4.2). Overall,
results look promising for EB1 tracking.
108
mcf10a	
   Pten	
  (-­‐/-­‐)	
   pi3kca	
   Twist	
  
	
  
Schema9c	
  of	
  cytoskeletal	
  phenotypes	
  
*	
  Hypothesized	
  cytoskeletal	
  state	
  
m
f1
0a
	
  
Pt
en
	
  (-­‐
/-­‐
)	
  
Pi
3K
CA
	
  
Tw
is
t	
  
Biomolecular	
  
Glu	
   -­‐	
   -­‐	
   -­‐	
   +++	
  
Vimen@n	
   -­‐	
   -­‐	
   ++	
   +++	
  
cofilin	
   -­‐	
   +++	
   -­‐	
   ?	
  
mLC	
   +*	
   -­‐*	
   ++*	
   ++*	
  
Mechanical	
   Ac@n	
  
+	
   -­‐	
   ++	
   ++*	
  
MT	
   +	
   +	
   ++	
   +++	
  
Overall	
   mcTn	
   -­‐	
   ++	
   -­‐	
   +++	
  
Figure 4.1: Different balances of forces between microtubules and actin cortex for 4 different
cyotoskeletal phenotypes where thicker lines indicate stronger force and dashed lines indicate
weaker force; red is actin, green is tubulin. Chart indicates state of each cytoskeletal component
based and overall mechanical contribution, where a ’-’ indicates not present; ’+’, present; ’+++’
strongly present.
109
mcf10a	
   Pten	
  -­‐/-­‐	
   pi3kCA	
  
si
ng
le
	
  
T-­‐
av
e	
  
Figure 4.2: Live-cell confocal microscopy of EB1-GFP. (Top) Single image frame of EB1-GFP
transfected cells for all 3 conditions: mcf10a, pten (-/-), and PIK3CA-H1047R. (Bottom) Mean
image across all 50 frames at time steps 2 seconds apart.
110
4.2.2 Direct Mechanical Measurements and Preliminary Optical Stretcher
Results
While the results presented in my thesis show changes in the ability of mi-
crotubules to move beyond the cell body boundary both by manipulating either
the actomyosin cortex or the tubulin stability, it does not measure direct material
properties nor integrity of the actomyosin cortex.
One way to measure actomyosin integrity more directly is to use an optical
stretcher to measure properties such as stiffness, deformability, and viscoelasticity.
The stretcher consists of two lasers with divergent optics facing each other; due
to the higher index of refraction on the inside of the cell, light loses momentum
when it exits the cell. The result is a uniaxial tensile force on the order of a few
hundred pN exerted on the cell [149]. Using rapidly acquired videos of the ‘stretch’
in combination with very robust edge-detection image analysis techniques, one can
deduce material properties of the cell. In 2012, I was able to collect preliminary
results from Josef Kas′s lab in Leipzig Germany.
Preliminary results used Human Mammary Epithelial Cells (HMEC): one
GFP-expressing control (HMEC-GFP) and one which over-expresses twist (HMEC-
twist). Overexpression of twist has been shown to induce epithelial-to-mesenchymal-
transition (EMT) morphology and McTN formation [37]. Measurements consisted
of extracting outlines from 182 HMEC-GFP cells and 155 HMEC twist cells. Using
111
code written by members Josef Kas’s lab in Leipzig Germany, the outline of the edge
of the cells at 152 time points over the course of 5 seconds were extracted. From the
outlines, the long axis along of the cell, or direction of stretch controlled for rotation
was computed. To compare across cell types and drug treatments, the rotationally
adjusted long axis was normalized or divided by the initial, relaxed state and set
to zero (by subtracting normalized relaxed state); the result is the relative, rotation
adjusted long axis measurement plotted as a function of time (Figure 4.3). Prelim-
inary results show that twist is more contractile than the control. This suggests
that McTNs are driven more by the stability of the MTs than the weakness of actin
cortex. One of the disadvantages of the optical stretcher is cell heating; thus, the
preliminary results shown above could also indicate that twist cells respond more
quickly to cytoplasmic heating.
112
Figure 4.3: Optical Stretcher Results GFP vehicle vs Twist show the normalized, rotation
adjusted, relative long-axis results as a function of time. Twist appears to be stiffer or have a
more contractile actomyosin cortex.
113
4.3 Outlook and Final Thoughts
The work outlined in my thesis is just the beginning of understanding the cy-
toskeletal behavior that gives rise to McTNs. With several emerging technologies on
the horizon, comes the promising possibility of further measuring and understanding
many of these mechanical questions. For example, there is now new microfluidics
stretcher that is even higher-throughput than the optical stretcher without several
of the technical challenges like heating the cells up [150]. Nanofabrication of PDMS
fabricated ridges may help better understand how CTCs sense the environment once
they attach [125]. Considering that currently many cancer drug therapies target the
cytoskeleton, better understanding of these mechanisms is of critical importance to
clinical therapies.
One interesting area of future work involves better relating cytoskeletal inter-
actions between attached and suspended cells mechanically. One valuable technique
that holds promise for assaying both suspended and attached cells is the aspirator
micropipette assays where a portion of the membrane is deformed under negative
pressure. Already micropipette assays have shown that that bone marrow-derived
mesenchymal stem cells (hMSCs) become stiffer on stiffer substrates where E0 on
stiffer substrate was 155±76 Pa and on softer substrates was 35±6 [151]. An-
other study assayed suspended hMSCs and found viscoelastic solid behavior with a
Young′s moduli for E0 and E∞ to be 886 ± 289 Pa and 372 ± 125 Pa [152]. The
114
trend of increased Young′s moduli values for attached cells compared to suspended
cells is consistent with AFM and optical stretcher results [82,83].
The study of McTNs is a new area of cancer research and still has many
interesting questions remaining. Preliminary unpublished work in the Martin Lab
has found McTNs in fresh patient cells, but currently no other in vivo discoveries
of McTNS in other human systems have been observed. Previously, I discussed the
potential role of leukocytes in mediating CTC reattachment, it is currently unknown
whether immune cells have any interactions with McTNs, but it may be worth
exploring. Likewise, no studies have looked for receptors on the surface of McTNs
that may help mediate reattachment. Also, interesting unanswered questions that
may help us better answer the cytoskeletal interactions between microtubules and
the actomyosin cortex is the amount of force exerted by McTNs as well as McTN
stiffness.
Research on McTNs is at the forefront for improving understanding the role
of the cytoskeleton on reattachment in metastasis. Already, it is widely accepted
that changes in cytoskeleton play a critical role in cancer invasiveness. The vast
majority of the cancer invasiveness body of literature is limited to the migration of
attached cells though and not the reattachment of CTCs. Clinically, many of the
most common chemotherapies already target the cytoskeleton such as Vinblastine,
Paclitaxel, and Demecolcine. Given the potential for cytoskeleton alterations to
influence CTC reattachment, the implications of improving our understanding in
this underappreciated and poorly understood area of research is critical. The work
outlined in my thesis lays the foundation of building better analytical techniques
115
to studying the balance of forces between microtubules and the actomyosin cortex
especially in suspended cells. Additionally, I present an organizing framework in
order to better understand cytoskeletal interactions and show preliminary results
supporting this model.
116
Bibliography
[1] Patrick Mehlen and Alain Puisieux. Metastasis: a question of life or death.
Nature reviews: Cancer, 6:449–458, June 2006.
[2] Yibin Kang. New tricks against an old foe: molecular dissection of metastasis
tissue tropism in breast cancer. Breast Disease, 26:129–38, 2007.
[3] Gaorav P Gupta and Joan Massague. Cancer metastasis: Building a frame-
work. Cell, 127:679–695, November 17 2006.
[4] Steven I Hajdu. A note from history: Landmarks in history of cancer, part 1.
Cancer, 117(5):1097–1102, March 1 2011.
[5] Rosalie David and Michael R. Zimmerman. Cancer: an old disease, a new
disease or something in between? Science and Society, 10(10), October 2010.
[6] Steven I Hajdu and M Vadmal. A note from history: Landmarks in history
of cancer, part 6. Cancer, 119(23):14058–82, September 16 2013.
[7] Stephen F Keevil. Physics and medicine: a historical perspective. The Lancet,
379(9825), April 21 2012.
[8] Karl Pearson. The Grammar of Science, volume 17. Walter Scott, London,
UK, 1892.
[9] Wei Wang. Radiotherapy in the management of early breast cancer. Journal
of Medical Radiation Sciences, 60(1):40–46, March 2013.
[10] Steven I Hajdu. A note from history: Landmarks in history of cancer, part 4.
Cancer, 118(20):4914–28, October 15 2012.
[11] John S. Laughlin. History of medical physics. Physics Today, 36(7), July 1983.
[12] Sridhar Ramaswamy, Ken N Ross, Eric S Lander, and Todd Golub. A molecu-
lar signature of metastasis in primary solid tumors. Nature Genetics, 33:49–54,
December 9 2002.
117
[13] Andy J Minn, Gaorav P Gupta, Peter M Siegel, Paula D Bos, Weiping Shu,
Dilip D Giri, Agnes Viale, Adam B Olshen, William L Gerald, and Joan
Massague. Genes that mediate breast cancer metastasis to lung. Nature,
436:518–524, July 28 2005.
[14] Victoria Seewaldt. Ecm stiffness paves the way for tumor cells. Nature
Medicine, 20:332–333, April 2014.
[15] Sanjay Kumar and Valerie M. Weaver. Mechanics, malignancy, and metastasis:
the force journey of a tumor cell. Cancer Metastasis Review, 28(1-2):113–27,
January 21 2009.
[16] Vinay Swaminathan, Karthikeyan Mythreye, E Tim O’brien, Andrew
Berchuck, Gerard C Blobe, and Richard Superfine. Mechanical stiffness grades
metastatic potential in patient tumor cells and in cancer cell lines. Cancer Re-
search, 71(15), 2011.
[17] Ali Rafiq. A glimpse into the historical journey of how we came to understand
cancer. Journal of Pioneering Medical Sciences Blogs, December 12 2014.
[18] Jiaquan Xu, Sherry L. Murphy, Kenneth D. Kochanek, and Brigham A. Bas-
tian. Deaths: Final data for 2013. National Vital Statistics Reports, 64(2),
February 16 2016.
[19] Debra L. Blackwell, Jacqueline W Lucas, and Tainya C Clarke. Summary
health statistics for u.s. adults: National health interview survey. Vital Health
Statistics, 10(260), 2014.
[20] Jonathon Sleeman and Patricia S Steeg. Stopping cancer in its tracks: Cancer
metastasis as a therapeutic target. European Journal of Cancer, 46(7):1177–
1180, 2010.
[21] Andrzej B Poplawski, Michal Jankowski, Stephen W Erickson, Teresita Dõaz
de Stahl, E Christopher Partridge, Chiquito Crasto, Jingyu Guo, John Gibson,
Uwe Menzel, Carl EG Bruder, Aneta Kaczmarczyk, Magdalena Benetkiewicz,
Robin Andersson, Johanna Sandgren, Barbara Zegarska, Dariusz Bala, Ewa
Srutek, David B Allison, Arkadiusz Piotrowski, Wojciech Zegarski, and
Jan P Dumanski. Frequent genetic differences between matched primary and
metastatic breast cancer provide an approach to identification of biomarkers
for disease progression. European Journal of Human Genetics, 18:560–568,
2010.
[22] Mika Suzuki and David Tarin. Gene expression profiling of human lymph node
metastases and matched primary breast carcinomas: Clinical implications.
Molecular Oncology, 1:172–180, April 7 2007.
[23] Myriam Labelle and Richard O Hynes. The initial hours of metastasis: the
importance of cooperative host-tumor cell interactions during hematogenous
dissemination. Cancer Discovery, 2(12):1091–1099, December 2012.
118
[24] Patricia S Steeg. Tumor metastasis: mechanistic insights and clinical chal-
lenges. Nature Medicine, 12(8):895–904, August 2006.
[25] David Owen Morgan. The Cell Cycle: Principles of Control. New Science
Press, Sunderland, Massachusetts, 2007.
[26] Peter Gassman and Joerg Haier. The tumor cell-host organ interface in the
early onset of metastatic organ colonisation. Clinical Experimental Metastasis,
25:171–181, December 5 2008.
[27] Neta Erez and Lisa M. Coussens. Leukocytes as paracrine regulators of metas-
tasis and determinants of organ-specific colonization. International Journal of
Cancer, 128(11):2536–2544, June 1 2011.
[28] Carina Strell and Frank Entschladen. Extravasation of leukocytes in compar-
ison to tumor cells. Cell Communication and Signaling, 6(10), December 4
2008.
[29] Simone Bersini, Jessie S. Jeon, Matteo Moretti, and Roger D. Kamm. In vitro
models of the metastatic cascade: from local invasion to extravasation. Drug
Discovery Today, 19(6):735Ð742, June 2014.
[30] Carina Strell, Kerstin Lang, Bernd Niggemann, Kurt Zaenker, and Frank
Entschladen. Surface molecules regulating rolling and adhesion to endothelium
of neutrophil granulocytes and mda-mb-468 breast carcinoma cells and their
interaction. Cellular and Molecular Life Sciences, December 2007.
[31] Ja Hye Myung, Khyati A. Gajjar, Ryan M. Pearson, Cari A. Launiere,
David T. Eddington, and Seungpyo Hong. Direct measurements on cd24-
mediated rolling of human breast cancer mcf-7 cells on e-selectin. Analytical
Chemistry, 83(3):1078–83, February 1 2011.
[32] Joseph S. Palumbo, Kathryn E. Talmage, Jessica V. Massari, Christine M. La
Jeunesse, Matthew J. Flick, Keith W. Kombrinck, Markeńta Jirouskova, and
Jay L. Degen. Platelets and fibrin(ogen) increase metastatic potential by
impeding natural killer cellÐmediated elimination of tumor cells. Blood,
104(1):178–185, 2005.
[33] Nicolas Reymond, Barbara Borda d’Agua, and Anne J.ÊRidley. Crossing the
endothelial barrier during metastasis. Nature Reviews Cancer, 13:858–870,
December 2013.
[34] Jonathan GarcŠa-RomĞn and Alejandro Zentella-Dehesa. Vascular perme-
ability changes involved in tumor metastasis. Cancer Letters, 335(2):259–269,
July 28 2013.
[35] Jorg Haier and Garth L. Nicolson. Tumor cell adhesion under hydrodynamic
conditions of fluid flow. Acta Pathologica, Microbiologica et Immunologica
Scandinavica, 109(4):241–262, April 2001.
119
[36] Matthew Barnes, Jones T Nauseef, and Michael D Henry. Resistance to fluid
shear stress is a conserved biophysical property of malignant cells. PloS ONE,
December 2012.
[37] Rebecca A Whipple, Michael A Matrone, Edward H Cho, Eric M Balzer,
Michele I Vitolo, Jennifer R Yoon, Olga B Ioffe, Kimberly C Tuttle, Jing
Yang, and Stuart S Martin. Epithelial-to-mesenchymal transition promotes
tubulin detyrosinatiion and microtentacles that enhance endothelial engage-
ment. Cancer Research, 70:8127–8137, 2010.
[38] Michael A Matrone, Rebecca A Whipple, Eric M Balzer, and Stuart S Martin.
Microtentacles tip the balance of cytoskeletal forces in circulating tumor cells.
Cancer Research, 70, October 2010.
[39] Lekhana Bhandary, Rebecca A. Whipple, Michele I. Vitolo, Monica S. Char-
pentier, Amanda E. Boggs, Kristi R. Chakrabarti, Keyata N. Thompson, and
Stuart S. Martin. Rock inhibition promotes microtentacles that enhance reat-
tachment of breast cancer cells. Oncotarget, 6(8):6251–66, January 31 2015.
[40] Rebecca A Whipple, Michele I Vitolo, Amanda E Boggs, Monica S Charpen-
tier, Keyata Thompson, and Stuart S Martin. Parthenolide and costunolide
reduce microtentacles and tumor cell attachment by selectively targeting dety-
rosinated tubulin independent from nf-κb inhibition. Breast Cancer Research,
15(5), 2013.
[41] Michele I. Vitolo, Amanda E. Boggs, Rebecca A. Whipple, Jennifer R. Yoon,
Keyata Thompson, Michael A. Matrone, Edward H. Cho, Eric M. Balzer,
and Stuart S. Martin. Loss of pten induces microtentacles through pi3k-
independent activation of cofilin. Oncogene, June 11 2013.
[42] Rebecca A Whipple, Eric M Balzer, Edward H Cho, Michael A Matrone, Jen-
nifer R Yoon, and Stuart S Martin. Vimentin filaments support extension of
tubulin-based microtentacles in detached breast tumor cells. Cancer Research,
2008.
[43] Kristi R. Chakrabarti, Rebecca A. Whipple, Amanda E. Boggs, Lindsay K.
Hessler, Lekhana Bhandary, Michele I. Vitolo, Keyata Thompson, and Stu-
art S. Martin1. Pharmacologic regulation of ampk in breast cancer affects cy-
toskeletal properties involved with microtentacle formation and re-attachment.
Oncotarget, 6(34):36292–307, November 3 2015.
[44] Jennifer R. Yoon, Rebecca A. Whipple, Eric M. Balzer, Edward H. Cho,
Michael A. Matrone, Michelle Peckham, and Stuart S. Martin. Local anesthet-
ics inhibit kinesin motility and microtentacle protrusions in human epithelial
and breast tumor cells. Breast Cancer Research and Treatment, 129(3):691–
701, 2011.
120
[45] Preeti Shah, Yael Gau, and Gauri Sabnis. Histone deacetylase inhibitor enti-
nostat reverses epithelial to mesenchymal transition of breast cancer cells by
reversing the repression of e-cadherin. Breast Cancer Research and Treatment,
143(1):99–111, 2014.
[46] Nicole A. Perry, Michele I. Vitolo, Stuart S. Martin, and Aikaterini Kon-
trogianni Konstantopoulos. Loss of the obscurin-rhogef downregulates rhoa
signaling and increases microtentacle formation and attachment of breast ep-
ithelial cells. Oncotarget, 5(18):8558–8568, August 10 2014.
[47] Eric M. Balzer, Rebecca A. Whipple, Keyata Thompson, Amanda E. Boggs,
Jana Slovic, Edward H. Cho, Michael A. Matrone, Toshiyuki Yoneda,
Susette C. Mueller, and Stuart S. Martin. c-src differentially regulates the
functions of microtentacles and invadopodia. Oncogene, 29(48):6402–6408,
2010.
[48] RA Whipple, AM Cheung, and SS Martin. Detyrosinated microtubule pro-
trusions in suspended mammary epithelial cells promote reattachment. Ex-
perimental Cell Research, 313(7):1326–1336, April 2007.
[49] Alison NKillilea, Roseann Csencsits, Sam Kenny, Ke Xu, and Kenneth H.
Downing. Structure of microtubule-based microtentacles. Microscopy and
Microanalysis, 21:235–236, August 2015.
[50] A Mialhe, L Lafanechere, I Treilleux, N Peloux, C Dumontet, A Bremond,
MH Panh, R Payan, J Wehland, RL Margolis, and D Job. Tubulin detyrosi-
nation is a frequent occurrence in breast cancers of poor prognosis. Cancer
Research, 61:5024–5027, 2001.
[51] Sadiqa Khawaja, Gregg Gundersen, and Jeannette Bulinski. Enhanced stabil-
ity of microtubules enriched in detyrosnated tubulin is not a direct function
of detyrosination level. Journal of Cell Biology, 106:141–149, 1988.
[52] Eric M. Balzer, Ziqiu Tong, Colin D. Paul, Wei-Chien Hung, Kimberly M.
Stroka, Amanda E. Boggs, Stuart S. Martin, and Konstantinos Konstantopou-
los. Physical confinement alters tumor cell adhesion and migration phenotypes.
Federation of American Societies for Experimental Biology, 26, June 15 2012.
[53] Monica S. Charpentier, Rebecca A. Whipple, Michele I. Vitolo, Amanda E.
Boggs, Jana Slovic, Keyata N. Thompson, Lekhana Bhandary, and Stuart S.
Martin. Curcumin targets breast cancer stemÐlike cells with microtentacles
that persist in mammospheres and promote reattachment. Cancer Research,
74(4):1250–1260, February 15 2014.
[54] Sendurai A. Mani, Wenjun Guo, Mai-Jing Liao, Elinor Ng. Eaton, Ayyakkannu
Ayyanan, Alicia Y. Zhou, Mary Brooks, Ferenc Reinhard, Cheng Cheng
Zhang, Michail Shipitsin, Lauren L. Campbell, Kornelia Polyak, Cathrin
Brisken, Jing Yang, and Robert A. Weinberg. The epithelial-mesenchymal
121
transition generates cells with properties of stem cells. Cell, 133(4):704 –715,
May 16 2008.
[55] Lao H. Saal, Karolina Holm, Matthew Maurer, Lorenzo Memeo, Xi-
aomei Wang Tao Su, Jennifer S. Yu, Per-Olof Malmstrom, Mahesh Man-
sukhani, Jens Enoksson, Hanina Hibshoosh, Ake Borg, and Ramon Parson.
Pik3ca mutations correlate with hormone receptors, node metastasis, and
erbb2, and are mutually exclusive with pten loss in human breast carcinoma.
Cancer Research, 65(7), April 1 2005.
[56] Jing Li, Clifford Yen, Danny Liaw, Katrina Podsypanina, Shikha Bose,
Steven I. Wang, Janusz Puc, Christa Miliaresis, Linda Rodgers, Richard Mc-
Combie, Sandra H. Bigner, Beppino C. Giovanella, Michael Ittmann, Ben
Tycko, Hanina Hibshoosh, Michael H. Wigler, and Ramon Parsons. Pten, a
putative protein tyrosine phosphatase gene mutated in human brain, breast,
and prostate cancer. Science, 275(5308):1943–1947, 1997.
[57] Shunyou Wang, Jing Gao, Qunying Lei, Nora Rozengurt, Colin Pritchard,
Jing Jiao, George V. Thomas, Gang Li, Pradip Roy-Burman, Peter S. Nelson,
Xin Liu, and Hong Wu. Prostate-specific deletion of the murine pten tumor
suppressor gene leads to metastatic prostate cancer. Cancer Cell, 4(3):209–
221, September 2003.
[58] Kanika A. Bowen, Hung Q. Doan, Binhua P. Zhou, Qingding Wang, Yuning
Zhou, Piotr G. Rychahou, and B. Mark Evers. Pten loss induces epithelial-
mesenchymal transition in human colon cancer cells. Anticancer Research,
29(11):4439–4449, 2009.
[59] MA Matrone, RA Whipple, K Thompson, EH Cho, MI Vitolo, JR Yoon
EM Balzer, OB Ioffe, KC Tuttle, M Tan, and SS Martin. Metastatic breast
tumors express increased tau, which promotes microtentacle formation and
the reattachment of detached breast tumor cells. Oncogene, 29:3217–3227,
June 2010.
[60] Donald E. Ingber. Tensegrity i. cell structure and hierarchical systems biology.
Journal of Cell Science, 116:1157–1173, April 1 2003.
[61] Donald E Ingber, Ning Wang, and Dimitrije Stamenovi? Tensegrity, cellu-
lar biophysics, and the mechanics of living systems. Reports on Progress in
Physics, 77(4), April 2 2014.
[62] Mechanobiology Institute, National University of Singapore. https://www.
mechanobio.info/what-is-mechanobiology/, 2016. [Online; accessed 19-
July-2016].
[63] N Kumar and M Flavin. Modulation of some parameters of assembly of micro-
tubules in vitro by tyrosination of tubulin. European Journal of Biochemistry,
127:215–222, 1982.
122
[64] Leticia Peris, Michael Wagenbach, Laurence LafanechŔre, Jacques Brocard,
Ayana T. Moore, Frank Kozielski, Didier Job, Linda Wordeman, and Annie
Andrieux. Motor-dependent microtubule disassembly driven by tubulin ty-
rosination. Journal of Cell Biology, 185(7):1159–1166, 2009.
[65] Daniel Webster, Juergen Wehland, Klaus Weber, and Gary Borisy. Detyrosi-
nation of alpha tubulin does not stabilize microtubules in vivo. Journal of
Cell Biology, 111, July 1990.
[66] Daniel Webster, Gregg Gundersen, Jeannette Bulinski, and Gary Borisy. As-
sembly and turnover of detyrosinated tubulin in vivo. Journal of Cell Biology,
105:265–276, 1987.
[67] Geri Gurland and Gregg G. Gundersen. Stable, detyrosinated microtubules
function to localize vimentin intermediate filaments in fibroblasts. Journal of
Cell Biology, 131(5):1275–1290, 1995.
[68] Geri Kreitzer, Liao Guojuan, and Gregg Gundersen. Detyrosination of tubulin
regulates the interaction of intermediate filaments with microtubules in vivo
via a kinesin-dependent mechanism. Molecular Biology of the Cell, 10, January
28 1999.
[69] P. Wagner, B. Wang, E. Clark, H. Lee, R. Rouzier, and L. Pusztai. Microtubule
associated protein (map)-tau: a novel mediator of paclitaxel sensitivity in vitro
and in vivo. Cell Cycle, 4(9):1149Ð1152, July 19 2005.
[70] Roman Rouzier, Radhika Rajan, Peter Wagner, Kenneth R. Hess, David L.
Gold, James Stec, Mark Ayers, Jeffrey S. Ross, Peter Zhang andThomas
A. Buchholz, Henry Kuerer, Marjorie Green, Banu Arun, Gabriel N. Hor-
tobagyi, W. Fraser Symmans, and Lajos Pusztaia. Microtubule-associated
protein tau: a marker of paclitaxel sensitivity in breast cancer. Proceed-
ings of the National Academy of Sciences of the United States of America,
102(23):8315–20, June 7 2005.
[71] Mohammad Soheilypour, Mohaddeseh Peyro, Stephen J. Peter, and Moham-
mad R. K. Mofrad. Buckling behavior of individual and bundled microtubules.
Biophysical Journal, 106:1718Ð1726, April 2015.
[72] Guillaume Salbreux, Guillaume Charras, and Ewa Paluchemail. Actin cortex
mechanics and cellular morphogenesis. Trends in Cell Biology, 22(10), 2012.
[73] Andrew G. Clark, Ortrud Wartlick, Guillaume Salbreux, and Ewa K. Paluch.
Stresses at the cell surface during animal cell morphogenesis. Current Biology,
24(10):484–494, May 19 2014.
[74] Guillaume T. Charras andChi Kuo Hu, Margaret Coughlin, and Timothy J.
Mitchison. Reassembly of contractile actin cortex in cell blebs. Journal of Cell
Biology, 175(3):477–490, November 6 2006.
123
[75] Hydrostatic pressure and the actomyosin cortex drive mitotic cell rounding.
Nature Letter, 469:226Ð230, January 2 2011.
[76] Dimitrije Stamenovic and Donald E. Ingber. Tensegrity-guided self assembly:
from molecules to living cells. Soft Matter, 5(6):1093, March 21 2009.
[77] Andrei I. Ivanov, Ann M. Hopkins, G. Thomas Brown, Kirsten Gerner-Smidt,
Brian A. Babbin, Charles A. Parkos, and Asma Nusrat. Myosin ii regulates the
shape of three-dimensional intestinal epithelial cysts. Journal of Cell Science,
121(11):1803–1814, 2008.
[78] Meng-Horng Lee, Pei-Hsun Wu, Daniele Gilkes Ivie Aifuwa, and Denis Wirtz.
Normal mammary epithelial cells promote carcinoma basement membrane in-
vasion by inducing microtubule-rich protrusions. Oncotarget, 6(32):32634–45,
August 3 2015.
[79] Carsten Schwan, Barbel Stecher, Tina Tzivelekidis, Marco van Ham, Manfred
Rohde, Wolf-Dietrich Hardt, Jurgen Wehland, and Klaus Aktories. Clostrid-
ium difficile toxin cdt induces formation of microtubule-based protrusions and
increases adherence of bacteria. PLOS Pathogens, 5(10), October 16 2009.
[80] Ewa Paluch, Matthieu Piel, Jacques Prost, Michel Bornens, and Cecile Sykes.
Cortical actomyosin breakage triggers shape oscillations in cells and cell frag-
ments. Biophysical Journal, 89(1):724–733, July 2005.
[81] Ning Wang and Donald E. Ingber. Control of cytoskeletal mechanics by ex-
tracellular matrix, cell shape, and mechanical tension. Biophysical Journal,
66(6):2181–2189, June 1994.
[82] Seyed Mohammad Ali Haghparast, Takanori Kihara, and Jun Miyake. Distinct
mechanical behavior of hek293 cells in adherent and suspended states. PeerJ,
3(e1131), July 30 2015.
[83] John M. Maloney, Dessy Nikova, Franziska Lautenschlager, Emer Clarke,
Robert Langer, Jochen Guck, and Krystyn J. Van Vliet. Mesenchymal stem
cell mechanics from the attached to the suspended state. Biophysical Journal,
99(8):2479–2487, October 10 2010.
[84] Nir S. Gov and Ajay Gopinathany. Dynamics of membranes driven by actin
polymerization. Biophysical Journal, 90(2):454–469, January 15 2006.
[85] Jay D. Humphrey, Eric R. Dufresne, and Martin A. Schwartz. Mechanotrans-
duction and extracellular matrix homeostasis. Nature Reviews Molecular Cell
Biology, 15(12):802–812, December 2014.
[86] John D.Hood and David A. Cheresh. Role of integrins in cell invasion and
migration. Nature Reviews Cancer, 2:91–100, February 2002.
124
[87] Anna Huttenlocher and Alan Rick Horwitz. Integrins in cell migration. Cold
Spring Harbor Perspectives in Biology, 3(9), September 2011.
[88] Thomas W. Marshall, Isaac E. Lloyda, Jean Marie Delalandea, Inke NŁthkeb,
and Jody Rosenblatt. The tumor suppressor adenomatous polyposis coli con-
trols the direction in which a cell extrudes from an epithelium. Molecular
Biology of the Cell, 22(21):3962–70, September 7 2011.
[89] Carmen Birchmeier, Walter Birchmeier, and Beate Brand-Saberi. Epithelial-
mesenchymal transitions in cancer progression. Acta Anatomica, 1996.
[90] Sharona Even-Ram, Andrew D. Doyle, Mary Anne Conti, Kazue Matsumoto,
Robert S. Adelstein, and Kenneth M. Yamada. Myosin iia regulates cell motil-
ity and actomyosinÐmicrotubule crosstalk. Nature Cell Biology, 9(3):299–309,
February 18 2007.
[91] Ying Wen, Christina H. Eng, Jan Schmoranzer, Noemi Cabrera-Poch, Edward
J. S. Morris, Michael Chen, Bradley J. Wallar andArthur S. Alberts, and
Gregg G. Gundersen. Eb1 and apc bind to mdia to stabilize microtubules
downstream of rho and promote cell migration. Nature Cell Biology, 6(9):820–
30, 2004.
[92] James B. Moseley, Francesca Bartolini, Kyoko Okada, Ying Wen, Gregg G.
Gundersen, and Bruce L. Goode. Regulated binding of adenomatous polyposis
coli protein to actin. Journal of Biological Chemistry, 282(17):12661–8, 2007.
[93] Clifford P. Brangwynne, Frederick C. MacKintosh, Sanjay Kumar, Nicholas A.
Geisse, Jennifer Talbot, L. Mahadevan, Kevin K. Parker, Donald E. Ingber,
and David A. Weitz. Microtubules can bear enhanced compressive loads in
living cells because of lateral reinforcement. Journal of Cell Biology, 173, 2006.
[94] Donald E. Ingber. Cancer as a disease of epithelial-mesenchymal interactions
and extracellular matrix regulation. Differentiation, 70(9-10):547–60, 2002.
[95] Hideki Yamaguchia and John Condeelis. Regulation of the actin cytoskeleton
in cancer cell migration and invasion. Biochimica et Biophysica Acta (BBA)
- Molecular Cell Research, 1773(5):642–52, 2007.
[96] Eric M Balzer, Rebecca A Whipple, Edward H Cho, Michael A Matrone, and
Stuart S Martin. Antimitotic chemotherapeutics promote adhesive responses
in detached and circulating tumor cells. Breast Cancer Research Treatment,
121:65–78, 2010.
[97] Jian Yu Rao and Ning Li. Microfilament actin remodeling as a potential target
for cancer drug development. Current Cancer Drug Targets, 4(4):345–54, 2004.
[98] Gerhard Fritz and Bernd Kaina. Rho gtpases promising cellular targets for
novel anticancer drugs. Current Cancer Drug Targets, 6(1):1–14, 2006.
125
[99] Kazuyuki Itoh, Kiyoko Yoshioka, Hitoshi Akedo, Masayoshi Uehata, Toshi-
masa Ishizaki, and Shuh Narumiya. An essential part for rho-associated kinase
in the transcellular invasion of tumor cells. Nature Medicine, 5(2):221–225,
1999.
[100] Dominico Vigil, Tai Young Kim, Ana Plachco andAndrew J. Garton, Linda
Castaldo, Jonathan A. Pachter, Hanqing Dong, Brianna Tokar Xin Chen,
Sharon L. Campbell, and Channing J. Der. Rock1 and rock2 are required
for non-small cell lung cancer anchorage-independent growth and invasion.
Cancer Research, 72(20):5338–47, 2012.
[101] Lili Wang, Lexun Xue, Hongxia Yan, Jie Li, and Yucheng Lu. Effects of
rock inhibitor, y-27632, on adhesion and mobility in esophageal squamous cell
cancer cells. Molecular Biology Reports, 37(4):1971–7, 2010.
[102] Francis J. Alenghat, Surya M. Naulib, Robert Kolbb, Jing Zhoub, and Don-
ald E. Ingber. Global cytoskeletal control of mechanotransduction in kidney
epithelial cells. Experimental Cell Research, 301(1):23–30, November 15 2004.
[103] Hirokazu Ohata, Tatsuya Ishiguro, Yuki Aihara, Ai Sato, Hiroaki Sakai,
Shigeki Sekine, Hirokazu Taniguchi, Takayuki Akasu, Shin Fujita, Hitoshi
Nakagama, and Koji Okamoto. Induction of the stem-like cell regulator
cd44 by rho kinase inhibition contributes to the maintenance of colon cancer-
initiating cells. Cancer Research, 72(19):5101–10, October 2 2012.
[104] Xuefeng Liu, Virginie Ory, Sandra Chapman, Hang Yuan, Chris Albanese,
Bhaskar Kallakury, Olga A. Timofeeva, Caitlin Nealon, Aleksandra Dakic,
Vera Simic, Bassem R. Haddad, Johng S. Rhim, Anatoly Dritschilo, Anna
Riegel, Alison McBride, and Richard Schlegel. Rock inhibitor and feeder cells
induce the conditional reprogramming of epithelial cells. American Journal of
Pathology, 180(2):599–607, February 2012.
[105] Frank A. Suprynowicza, Geeta Upadhyaya, Ewa Krawczyka, Sarah C. Kram-
era, Jess D. Heberta, Xuefeng Liua, Hang Yuana, Chaitra Cheluvarajub,
Phillip W. Clappb, Richard C. Boucher, Christopher M. Kamonjoha, Scott H.
Randellb, and Richard Schlegela. Conditionally reprogrammed cells represent
a stem-like state of adult epithelial cells. Proceedings of the National Academy
of Sciences of the United States, December 4 2012.
[106] Mihaly Kovacs, Judit Toth, Csaba Hetenyi, Andras Malnasi-Csizmadia, and
James R. Sellers. Mechanism of blebbistatin inhibition of myosin ii. Journal
of Biological Chemistry, 279(34):35557–63, June 16 2004.
[107] Michelle Piehl and Lynne Cassimeris. Organization and dynamics of growing
microtubule plus ends during early mitosis. Molecular Biology of the Cell,
14(3):916–25, December 7 2003.
126
[108] Michelle Piehl, U. Serdar Tulu, Pat Wadsworth, and Lynne Cassimeris. Cen-
trosome maturation: measurement of microtubule nucleation throughout the
cell cycle by using gfp-tagged eb1. Proceedings of the National Academy of
Sciences of the United States, 101, September 4 2004.
[109] Guillermo Velasco, Chris Armstrong, Nick Morrice, Sheelagh Frame, and
Philip Cohen. Phosphorylation of the regulatory subunit of smooth muscle
protein phosphatase 1m at thr850 induces its dissociation from myosin. Fed-
eration of European Biochemical Societies Letters, 527(1-3):101–4, September
11 2002.
[110] Takahiro Tsuji, Toshimasa Ishizaki, Muneo Okamoto, Chiharu Higashida,
Kazuhiro Kimura, Tomoyuki Furuyashiki, Yoshiki Arakawa, Raymond B.
Birge, Tetsuya Nakamoto, Hisamaru Hirai, and Shuh Narumiya. Rock and
mdia1 antagonize in rho-dependent rac activation in swiss 3t3 fibroblasts.
Journal of Cell Biology, May 20 2002.
[111] Ilan Spector, Nava R. Shochet, Dina Blasberger, and Yoel Kashman.
Latrunculins–novel marine macrolides that disrupt microfilament organiza-
tion and affect cell growth: I. comparison with cytochalasin d. Cell Motility
and the Cytoskeleton, 13(3):127–44, 1989.
[112] Walter M. Morton, Kathryn R. Ayscough, and Paul J. McLaughlin. Latrun-
culin alters the actin-monomer subunit interface to prevent polymerization.
Nature Cell Biology, 2(6):376–8, June 1 2000.
[113] Adam G. Hendricks, Jacob E. Lazarus, Eran Perlson, Melissa K. Gardner,
David J. Odde, Yale E. Goldman, and Erika L.F. Holzbaur. Dynein tethers
and stabilizes dynamic microtubule plus ends. Current Biology, 22(7):632–7,
April 10 2012.
[114] Aaron Christian Ponti, Matthias Machacek, Stephanie L. Gupton, Clare M
Waterman-Storer, and Gaudenz Danuser. Two distinct actin networks drive
the protrusion of migrating cells. Science, 305(5691):1782–1786, September
17 2004.
[115] Pieta K. Mattila and Pekka Lappalainen. Filopodia: molecular architecture
and cellular functions. Nature Reviews Molecular Cell Biology, 9(6):446–454,
June 2008.
[116] Wasim A. Sayyad, Ladan Amin, Paolo Fabris, Erika Ercolini, and Vincent
Torre. The role of myosin-ii in force generation of drg filopodia and lamellipo-
dia. Scientific Reports, 5(7842), January 19 2015.
[117] Margaret L. Gardel, Benedikt Sabass, Lin Ji, Gaudenz Danuser, Ulrich S.
Schwarz, and Clare M. Waterman. Traction stress in focal adhesions corre-
lates biphasically with actin retrograde flow speed. Journal of Cell Biology,
183(6):999–1005, December 15 2008.
127
[118] Clifford P. Brangwynne, F. C. MacKintosh, and David A. Weitz. Force fluctu-
ations and polymerization dynamics of intracellular microtubules. Proceedings
of the National Academy of Sciences of the United States, October 9 2007.
[119] Dominique Seetapun and David J. Odde. Cell-length-dependent microtubule
accumulation during polarization. Current Biology, 20(11):979–88, June 8
2010.
[120] Gianluca De Santis, Alex B. Lennon, Federica Boschetti, Benedict Verhegghe,
Pascal Verdonck, and Patrick J. Prendergast. How can cells sense the elasticity
of a substrate? an analysis using a cell tensegrity model. Eur Cell Mater,
22:202–13, October 11 2011.
[121] Piyush B. Gupta, Tamer T. Onder, Guozhi Jiang, Kai Tao, Charlotte Kuper-
wasser, Robert A. Weinberg, and Eric S. Lander. Identification of selective
inhibitors of cancer stem cells by high-throughput screening. Cell, 138(4):645–
59, August 21 2009.
[122] Birthe Willipinski-Stapelfeldt, Sabine Riethdorf, Volker Assmann, Ute
Woelfle, Thomas Rau, Guido Sauter, Jochen Heukeshoven, and Klaus Pan-
tel. Changes in cytoskeletal protein composition indicative of an epithelial-
mesenchymal transition in human micrometastatic and primary breast carci-
noma cells. Clinical Cancer Research, 11(22):8006–14, November 15 2005.
[123] Linda A. Gordon, Kellie T. Mulligan, Helen Maxwell-Jones, Matthew Adams,
Rosemary A. Walker, and J. Louise Jones. Breast cell invasive potential relates
to the myoepithelial phenotype. International Journal of Cancer, 106(1):8–16,
April 16 2003.
[124] John Crocker and David Grier. Methods of digital video microscopy for col-
loidal studies. Journal of Colloid and Interface Science, 179(1):298–310, April
15 1996.
[125] Meghan K Driscoll, Xiaoyu Sun, Can Guven, John T Fourkas, and Wolfgang
Losert. Cellular contact guidance through dynamic sensing of nanotopography.
The Journal of Physical Chemistry Nano, 8(4):3546–3555, 2014.
[126] Meghaan M. Ferreira, Vishnu C. Ramani, and Stefanie S. Jeffrey. Circulating
tumor cell technologies. Molecular Oncology, 10(3), March 2016.
[127] Emilian Racila, David Euhus, Arthur J. Weiss, Chandra Rao, John McConnell,
Leon W. M. M. Terstappen, and Jonathan W. Uhr. Detection and character-
ization of carcinoma cells in the blood. Proceedings of the National Academy
of Sciences of the United States of America, 95(8):4589–4594, April 14 1998.
[128] Priya Gogoi, Saedeh Sepehri, Yi Zhou, Michael A. Gorin, Carmela Paolillo,
Ettore Capoluongo, Kyle Gleason, Austin Payne, Brian Boniface, Massimo
Cristofanilli, Todd M. Morgan, Paolo Fortina, Kenneth J. Pienta, Kalyan
128
Handique, and Yixin Wang. Development of an automated and sensitive mi-
crofluidic device for capturing and characterizing circulating tumor cells (ctcs)
from clinical blood samples. Plos One, 11(1), January 25 2016.
[129] James Che, Victor Yu, Manjima Dhar, Corinne Renier, Melissa Matsumoto,
Kyra Heirich, Edward B. Garon, Jonathan Goldman, Jianyu Rao, George W.
Sledge, Mark D. Pegram, Shruti Sheth, Stefanie S. Jeffrey, Rajan P. Kulkarni,
Elodie Sollier, and Dino Di Carlo. Classification of large circulating tumor cells
isolated with ultrahigh throughput microfluidic vortex technology. Oncotarget,
7(11):12748–12760, February 6 2016.
[130] Daniel L. Adams, R. Katherine Alpaugh, Stuart S. Martin, Monica Charp-
entier, Saranya Chumsri, Massimo Cristofanilli, Diane K. Adams, Olga V.
Makarova, Peixuan Zhu, Shuhong Li, Cha-Mei Tangh, and Steingrimur Ste-
fanssoni. Precision microfilters as an all in one system for multiplex analysis
of circulating tumor cells. Royal Society of Chemistry, 6:6405–6414, January
11 2016.
[131] Joan Massague and Anna C. Obenauf. Metastatic colonization by circulating
tumour cells. Nature, 529(7586):298–306, january 21 2016.
[132] Amanda E. Boggs, Michele I. Vitolo, Rebecca A. Whipple, Monica S. Char-
pentier, Olga G. Goloubeva, Olga B. Ioffe, Kimberly C. Tuttle, Jana Slovic,
Yiling Lu, Gordon B. Mills, and Stuart S. Martin. ?-tubulin acetylation ele-
vated in metastatic and basal-like breast cancer cells promotes microtentacle
formation, adhesion and invasive migration. Cancer Research, 75(1):203–215,
January 1 2015.
[133] Kristi R. Chakrabarti, James I. Andorko, Rebecca A. Whipple, Peipei Zhang,
Elisabeth L. Sooklal, Stuart S. Martin, and Christopher M. Jewell. Lipid
tethering of breast tumor cells enables real-time imaging of free-floating cell
dynamics and drug response. Oncotarget, 7(9):10486–10497, February 8 2016.
[134] Meghan K Driscoll, John T Fourkas, and Wolfgang Losert1. Local and global
measures of shape dynamics. Physical Biology, August 10 2011.
[135] Assaf Zemel, Florian Rehfeldt, Andre E X Brown, Dennis E. Discher, and
Samuel A. Safran. Optimal matrix rigidity for stress-fibre polarization in
stem cells. Nature Physics, 6:468–473, March 21 2010.
[136] Maria Romina Girotti, Gabriela Gremel, Rebecca Lee, Elena Galvani, Do-
minic Rothwell, Amaya Viros, Amit Kumar Mandal, Kok Haw Jonathan Lim,
Grazia Saturno, Simon J. Furney, Franziska Baenke, Malin Pedersen, Jane Ro-
gan, Jacqueline Swan, Matthew Smith, Alberto Fusi, Deemesh Oudit, Nathalie
Dhomen, Ged Brady, Paul Lorigan, Caroline Dive, and Richard Marais. Ap-
plication of sequencing, liquid biopsies, and patient-derived xenografts for per-
sonalized medicine in melanoma. Cancer Discovery, 6(3):286–99, December
29 2015.
129
[137] Despina Siolas and Gregory J. Hannon. Patient derived tumor xenografts:
transforming clinical samples into mouse models. Cancer Research,
73(17):5315–5319, September 1 2013.
[138] Hyeong-Gon Moon, Keunhee Oh, Jiwoo Lee, Minju Lee, Ju-Yeon Kim, Tae-
Kyung Yoo, Myung Won Seo, Ae Kyung Park, Han Suk Ryu, Eun-Jung
Jung, Namshin Kim, Seongmun Jeong, Wonshik Han, Dong-Sup Lee, and
Dong-Young Noh. Prognostic and functional importance of the engraftment-
associated genes in the patient-derived xenograft models of triple-negative
breast cancers. Breast Cancer Research and Treatment, 154(1), 2015.
[139] Thitiporn Chanwimaluang and Guoliang Fan. An efficient blood vessel de-
tection algorithm for retinal images using local entropy thresholding. Circuits
and Systems, 5, May 25-28 2003.
[140] Mohammed Saleem, Sandrine Morlot andAnnika Hohendahl, John Manzi,
Martin Lenz, and Aurelien Roux. A balance between membrane elasticity
and polymerization energy sets the shape of spherical clathrin coats. Nature
Communications, 6(6249), February 2015 2015.
[141] Kazuya Tsujita, Tadaomi Takenawa, and Toshiki Itoh. Feedback regulation
between plasma membrane tension and membrane-bending proteins organizes
cell polarity during leading edge formation. Nature Cell Biology, 17:749–758,
May 4 2015.
[142] Takeshi Sakamoto, John Limouze, Christian A. Combs, Aaron F. Straight,
and James R. Sellers. Blebbistatin, a myosin ii inhibitor, is photoinactivated
by blue light. Biochemistry, 44(2):584–588, December 15 2005.
[143] Stephen P. Davies, Helen Reddy, Matilde Caivano, and Philip Cohen. Speci-
ficity and mechanism of action of some commonly used protein kinase in-
hibitors. Biochemical Journal, 351(1):95–105, October 1 2001.
[144] Physical Sciences le Oncology Centers Network. A physical sciences network
characterization of non-tumorgenic and metastatic cells. Scientific Report,
2013.
[145] Farhad Vesuna, Ala Lisok, Brian Kimble, and Venu Raman. Twist modu-
lates breast cancer stem cells by transcriptional regulation of cd24 expression.
Neoplasia, 11(12):1318–1328, December 2009.
[146] Daniel R Webster, Gregg G Gundersenand Jeannette C Bulinski, and Gary G
Borisy. Differential turnover of tyrosinated and detyrosinated microtubules.
Proceedings of the National Academy of Sciences, 84(24):9040–9044, December
1987.
[147] Osigwe Esue, Ashley A Carson, Yiider Tseng, and Denis Wirtz. A direct
interaction between actin and vimentin filaments mediated by the tail domain
of vimentin. Journal of Biological Chemistry, 281(41):30393–30399, 2006.
130
[148] Dmitry Pavlov, Andras Muhlrad, John Cooper, Martin Wear, and Emil
Reisler. Actin filament severing by cofilin. Journal of Molecular Biology,
365(5):1350–1358, February 2007.
[149] J Guck, R Ananthakrishnan, T. J. Moon, C Cunningham, and J Kas. Optical
deformability of soft biological dielectrics. PRL, 84:5451–5454, 2000.
[150] Alexander Mietke, Oliver Otto, Salvatore Girardo, Philipp Rosendahl, Anna
Taubenberger, Stefan Golfier, Elke Ulbricht, Sebastian Aland, Guck, and Elis-
abeth Fischer-Friedrich. Extracting cell stiffness from real-time deformabil-
ity cytometry: Theory and experiment. Biophysical Journal, 109:2023–2036,
November 15 2015.
[151] Qinwei Xu, Cheng Li, Yuejun Kang, and Yilei Zhang. Long term effects of
substrate stiffness on the development of hmsc mechanical properties. Royal
Society of Chemistry, 5:105651–105660, December 8 2015.
[152] Samuel CW Tan, Wen X Pan, Gang Ma, Ning Cai, Kam W Leong, and Kin
Liao. Viscoelastic behaviour of human mesenchymal stem cells. BMC Cell
Biology, 9(1):1–7, 2008.
131