University of Southern California Dissertations and Theses

Ectopic expression of a truncated isoform of hair keratin 81 in breast cancer alters biophysical characteristics to promote metastatic propensity

(USC Thesis Other)

Ectopic expression of a truncated isoform of hair keratin 81 in breast cancer alters biophysical characteristics to promote metastatic propensity

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

ECTOPIC EXPRESSION OF A TRUNCATED ISOFORM OF HAIR KERATIN 81 IN
BREAST CANCER ALTERS BIOPHYSICAL CHARACTERISTICS TO PROMOTE
METASTATIC PROPENSITY

By

Diane Seung Kang

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
MEDICAL BIOLOGY

May 2022

Copyright 2021 Diane Seung Kang

ii
DEDICATION
This thesis is dedicated to all patients suffering from cancer and their caregivers –
I hope to have contributed to the scientific body of knowledge and somehow aided the
ongoing efforts to improving treatment options and outcomes; to my husband and two
girls who support and motivate me to always strive for better; and to my family both near
and far.

iii
ACKNOWLEDGEMENTS
I would like to acknowledge my primary mentor and role model, Dr. Min Yu, for the
financial and career support, and in helping me to develop as an independent scientist. I
am forever grateful for her dedication to my success, her unceasing positive
encouragement, and creativity and commitment to science.
To Dr. Ite Offringa, who was my very first mentor and champion as a freshly
graduated college student and who I am incredibly lucky to have met. She encouraged,
educated, supported, strengthened and emboldened me these past ten years to make
me who I am today.
To Dr. Josh Neman, for the time and support towards helping me to achieve my
research goals.
To the countless collaborators, core facilities, mentors, referees, lab members, and
colleagues that I have met along the way, I am constantly in awe of the intelligence, the
kindness, and the collaborative spirit that exists in the scientific community. Thank you
for your time, education, and support.
And last but not least, I would like to acknowledge all the friends and wonderful
people I have met here at USC and in the PhD Program. Thank you so much for the
countless laughs, the good food, and the shared advice and many conversations about
our malleable and hopeful futures.

iv
TABLE OF CONTENTS

DEDICATION…………………………………………………………………………………… ii

ACKNOWLEDGEMENTS…………………………………………………………………….. iii

TABLE OF CONTENTS………………………………………………………………..……... iv

LIST OF TABLES………………………………………………………………..…………….. vi

LIST OF FIGURES………………………………………………………………………….… vii

ABBREVIATIONS…………………………………………………………………………….. viii

ABSTRACT……………………………………………………………………………………... x

CHAPTER 1: BREAST CANCER AND THE METASTATIC CASCADE………………..... 1
1.1 Breast cancer statistics…………………………………………………………..... 1
1.2 Breast anatomy………………………………………………………….…............ 1
1.3 Different molecular expression profiles of breast cancer dictate treatment
options…………………………………………………………………..………..... 3
1.4 The metastatic cascade…………………………………………………………… 7
1.5 Peripheral blood liquid biopsies…………………………………………………. 11
1.6 Circulating Tumor Cells (CTCs)…………………………………………………. 12

CHAPTER 2: ENGINEERING AND PHYSICAL APPROACHES TO INVESTIGATING
CIRCULATING TUMOR CELLS…………………………………………………………….. 14
2.1 CTCs in the bloodstream are metastatic precursors………………………….. 14
2.2 CTC detection and live cell isolation methods…………………………………. 15
2.3 Introduction to the physical environment of CTCs…………………………….. 22
2.4 CTCs experience fluid shear stress…………………………………………….. 24
2.5 CTCs experience numerous collisions…………………………………………. 28
2.6 CTCs experience traction forces………………………………………………… 29
2.7 CTCs experience compressive forces………………………………………….. 31
2.8 Conclusions……………………………………………………………………….. 34

CHAPTER 3: THE CELL CYTOSKELETON: THE CONTRIBUTION OF KERATINS IN
CANCER DIAGNOSIS, PROGNOSIS, AND PROGRESSION………………………...... 36
3.1 Cell Cytoskeleton…………………………………………………………….…… 36
3.2 Actin microfilaments: assembly and function………………………….……. …37
3.3 Microtubules: assembly and function…………………………………………… 38
3.4 Intermediate filaments……………………………………………………………. 39
3.5 Keratins: assembly and function………………………………………………… 41
3.6 Post-translational modifications (PTMs) regulate keratins…………………… 45

v
3.7 Keratins as diagnostic and prognostic markers of breast cancer………….… 47
3.8 The functional landscape of keratins in breast cancer……………………....... 48

CHAPTER 4: ECTOPIC EXPRESSION OF A TRUNCATED ISOFORM OF HAIR
KERATIN 81 (KRT81) IN BREAST CANCER ALTERS BIOPHYSICAL
CHARACTERISTICS TO PROMOTE METASTATIC PROPENSITY…………………… 50
4.1 Introduction……………………………………………………………………...... 51
4.2 Results…………………………………………………………………………….. 52
4.3 Discussion………………………………………………………………………… 73
4.4 Methods…………………………………………………………………………… 77

CHAPTER 5: DISCUSSION……………………………………………………………...…. 95
5.1 Introduction……………………………………………………………………...... 95
5.2 Future directions………………………………………………………………….. 97
5.3 Conclusions……………………………………………………………………….. 99

REFERENCES………………………………………………………………………………. 100

vi
LIST OF TABLES

Table 4.1. List of differential RNAseq comparisons with KRT81 upregulation…………. 53
Table 4.2. List of qPCR primers……………………………………………………………… 80

vii
LIST OF FIGURES

Figure 1.1. Molecular subtypes of breast cancer……………………………………………. 7

Figure 1.2. Overview of the metastatic cascade………………………………………….... 11

Figure 2.1. The physical environment of CTCs…………………………………………….. 24

Figure 3.1. Keratin assembly process and structural function……………………………. 42

Figure 4.1. KRT81 expression in BRx07 and isogenic lung-metastatic derivatives
and publicly available dataset……………………………………………………………….. 53

Figure 4.2. Identification of a truncated isoform of KRT81 in cell lines and TCGA
dataset………………………………………………………………………………………..... 56

Figure 4.3. Graphical model of the disruption of the keratin filament assembly process
by the expression of tKRT81………………………………………………………...…...….. 57

Figure 4.4. Microscopic examination of intermediate filament structures with regard
to tKRT81 expression………………………………………………………………………… 58

Figure 4.5. Transmission electron microscopy imaging of desmosomes……………….. 59

Figure 4.6. KRT18 and tKRT81 physically interact………………………………………… 61

Figure 4.7. Examination of the impact of tKRT81 on the transcription and activation
of various signaling pathways…………………………………………...………………...… 64

Figure 4.8. Biophysical and morphological impact of tKRT81 expression……..……….. 66

Figure 4.9. Functional consequences of tKRT81 expression…………………………….. 69

Figure 4.10. Expression of tKRT81 enhances in vivo lung metastasis………………….. 71

Figure 4.11. Expression of tKRT81 does not enhance in vivo lung metastasis in some
cell lines………………………………………………………………………………………... 72

Figure 4.12. Graphical model of the impact of tKRT81 expression on enhanced
metastatic propensity…………………………………………………………………………. 74

Figure 4.13. Western blot validation of tKRT81 knockdown by CRISPRi (sg10) or
shRNA system (sh1, sh2) in various cell lines…………………………………………….. 79

viii
ABBREVIATIONS

AFM Atomic force microscopy
AUC Area under curve
BRCA Breast cancer
CNV Copy number variation
CTC Circulating tumor cell
DC Ductal cancer
DCIS Ductal cancer in situ
DEP Dielectrophoresis
ECM Extracellular matrix
EMP Epithelial-mesenchymal plasticity
EMT Epithelial to mesenchymal transition
EpCAM Epithelial cell adhesion molecule
ER Estrogen receptor
F-actin Filamentous actin
FSS Fluid shear stress
G-actin Globular actin
GFAP Glial fibrillary acidic protein
HD-SCA High definition single-cell analysis
HER2 Human epidermal growth factor receptor 2
ICAM1 Intercellular adhesion molecule 1
IDC Invasive ductal cancer
IDP Inner dense plaque
IF Intermediate filament
ILC Invasive lobular cancer
IVFC In vivo flow cytometry
KF Keratin filament
Ki-67 Marker of proliferation ki-67
LC Lobular cancer
LCIS Lobular cancer in situ
LINC Linker of nucleoskeleton and cytoskeleton
LSCE live-single-cell extractor
MPA Micropipette aspiration
MTOC Microtubule organizing center
ODP Outer dense plaque
OSCC Oral squamous cell carcinoma

ix
P1a Plectin 1a
PAAD Pancreatic adenocarcinoma
PBMC Peripheral blood monocytes
PDAC Pancreatic ductal adenocarcinoma
PDL1 Programmed death-ligand 1
PDMS Polydimethylsiloxane
PR Progesterone receptor
PRD Proline-rich domain
PTM Post-translational modifications
RBC Red blood cell
SEER Surveillance, Epidemiology, and End Results
SREBP1 Sterol regulatory element-binding protein
SUMO Small ubiquitin-related modifier
TCGA The Cancer Genome Atlas
ULF Unit length filament
VCAM1 Vascular cell adhesion molecule 1
WASP Wiskott-Aldrich syndrome protein
WBC White blood cell
WH2 Wiskott-Aldrich syndrome protein homology 2
WSS Wall shear stress
YAP1 Yes-associated protein 1

x
ABSTRACT

Metastasis is the leading cause of cancer deaths and is difficult to control, predict,
and treat in the clinic. Circulating tumor cells that are detected in the peripheral blood of
cancer patients are the best models for studying the metastatic cascade as they contain
subpopulations that represent the closest biological precursors of metastatic lesions.
Recent advancements in the isolation and ex vivo expansion of this extremely rare cell
population have allowed for a more accurate investigation of the metastatic process. In
this dissertation project, we identified the upregulation of a truncated isoform of KRT81 in
in vivo models of breast cancer metastasis using circulating tumor cells that were isolated
from the peripheral blood of breast cancer patients. The KRT81 gene encodes for a type
II keratin protein that is typically expressed in hair and contributes to structural rigidity and
mechanical resilience. In breast cancer cells, the truncated KRT81 (tKRT81) protein was
found to alter the cell cytoskeleton and biophysical properties of the cell, leading to
functional changes that enhance metastatic propensity. This work highlights the
biomechanical mechanism underlying the functional contribution of keratins in cancer
beyond their traditional use as diagnostic biomarkers.

1
CHAPTER 1
BREAST CANCER AND THE METASTATIC CASCADE

1.1 Breast cancer statistics
According to the National Cancer Institute’s Surveillance, Epidemiology, and End
Results (SEER) Program, approximately 12.9% of women have a lifetime risk of
developing breast cancer. Female breast cancer is the most frequently diagnosed type of
cancer and is expected to be the second leading cause of cancer-related deaths in
women [1], and the third leading cause of cancer-related deaths amongst both men and
women in 2021 [2]. Although rare, men can also develop breast cancer, with about 14%
of male breast cancer cases attributable to mutations in the BRCA2 gene [3].
In female breast cancer, about 5-10% of cases are thought to be caused by
hereditary gene mutations [4]. The most common hereditary gene mutations occur in the
BRCA1 and BRCA2 genes, which are tumor suppressor genes normally involved in DNA
damage repair [5]. Other non-genetic risk factors for developing breast cancer include
age, reproductive history, breast density, hormone replacement therapy or oral
contraceptives, and other behavioral factors such as obesity, tobacco, and alcohol
consumption.

1.2 Breast anatomy
The female breast is composed of 15 to 20 lobules, or bulb-like glands that secrete
milk. These lobules are located deep within the breast tissue and are attached to ducts
that carry the milk generated in the lobules to the nipple, which on average contains a

2
cluster of about 9 ducts [6]. Although breast cancers can arise from abnormal cells in the
lobules, termed lobular cancer (LC), it is more common for cancers to arise from the ducts,
termed ductal cancer (DC) [7]. Less frequently observed are mixed ductal-lobular
morphology breast cancers that are thought to arise from the progression of a proportion
of ductal tumors to a lobular morphology [8].
The entire mammary gland, from lobule to connecting duct and to nipple, is
comprised of a two-cell epithelial layer. The cells lining the lumen are called luminal cells
and can be identified by the specific expression of keratins 7, 8, 18, and 19 [9]. These
luminal cells lay on top of a basal, or myoepithelial, cell layer that is distinguishable from
the luminal epithelium by expression of keratins 5, 14, and 17 [10]. The myoepithelial cell
layer forms a protective sheet to separate the glandular tissue from the stromal
compartment and actively synthesizes and deposits basement membrane proteins [11].
Depending on whether or not the cancerous cells have breached the basement
membrane underlying the myoepithelial cells and spread into the surrounding tissue, the
LC or DC is labeled as either invasive (ILC/IDC) or in situ (LCIS/DCIS).
In situ cancers have not invaded the surrounding tissue, are confined to their site
of origin, and are considered Stage 0 breast cancers. On the other hand, invasive cancers
are categorized into Stages I-IV depending on the extent of local invasion and systemic
spread to lymph nodes and distant organs. Tumor staging takes into account how the
tumor looks morphologically, called tumor grading. Breast cancers are most commonly
diagnosed in Stage I: localized and confined to the primary site. Stage I breast cancers
have good outcomes with a 99% 5-year relative survival rate [12]. Stage II breast cancers
are defined by metastatic spread to lymph nodes surrounding the breast, while Stage III

3
and Stage IV breast cancers indicate increasing amounts of metastatic spread to a large
number of lymph nodes or distant organs. The 5-year relative survival rate for patients
with Stage IV breast cancers is only 28% [13] but is estimated to be improving among
women newly diagnosed with metastatic breast cancer between the ages of 15-49 [14].

1.3 Different molecular expression profiles of breast cancer dictate treatment
options
The Prosigna® Breast Cancer Prognostic Gene Signature Assay has been
developed to categorize breast cancers into five intrinsic molecular subtypes based on
the gene expression signatures of 50 genes, called the PAM50 gene signature. The five
molecular subtypes are: 1) Luminal A, 2) Luminal B, 3) Triple-Negative/Basal-Like, 4)
HER2-enriched, and 5) Normal-like (Figure 1.1).
Luminal A cancers express estrogen receptor (ER) and/or progesterone receptor
(PR), are negative for genome amplification of human epidermal growth factor receptor 2
(HER2) and have low levels of the proliferation marker ki-67 (Ki-67). Luminal A cancers
are the most commonly diagnosed and also have the best prognosis [15,16]. Luminal B
is ER+ and/or PR+, can be HER2 +/-, and has high levels of Ki-67. The triple-
negative/basal-like subtype is ER-/PR-/HER2- and is more common in younger and Black
women [17], and women with BRCA1 gene mutations. HER2-enriched cancers are ER-
and PR- but are positive for HER2 amplification (HER2+). Normal-like breast cancers are
similar to the Luminal A subtype in terms of ER+ and/or PR+, HER2- and low Ki-67, but
have a gene expression profile similar to normal breast tissue [18]. According to data
compiled from 2012-2016 by the American Cancer Society, luminal A was the most

4
prevalent breast cancer subtype (73% of breast cancer cases), followed by triple-
negative/basal-like (12%), luminal B (11%) and HER2-enriched (4%) [1].
Molecular subtyping improves patient stratification into prognostically defined risk
groups and provides guidance on optimal course of treatment [19]. For example, luminal
cancers with expression of growth hormones like ER/PR can be treated with aromatase
inhibitors that decrease circulating levels of the estrogen hormone, thereby dampening
the estrogen receptor signaling that encourages a cancer to grow. HER2-enriched
cancers can be treated with targeted therapy agents such as trastuzumab (Herceptin®),
a monoclonal antibody that binds the HER2 receptor and prevents its tumor-promoting
function. Triple-negative cancers, on the other hand, are the most difficult subtype to treat
as they lack targetable tumor-specific receptors. Treatment options for the triple negative
subtype are typically systemic chemotherapy and/or radiation. As of July 2021 however,
pembrolizumab (Keytruda®), an immunotherapy that blocks the immunosuppressive PD-
1 pathway, received FDA approval for use in unresectable, locally advanced, or
metastatic triple-negative breast cancers that are positive for the PD-1 ligand (PDL1).
Another immune checkpoint inhibitor, dostarlimab (Jemperli®), also received FDA
approval in August 2021 for a subset of advanced breast cancer patients with tumors that
have a DNA mismatch repair deficiency.
In addition to molecular profiling by Prosigna® there are also several other
genomic tests available and currently being developed for clinical use. Oncotype DX® is
21-gene expression assay that scores early stage primary tumors from luminal A breast
cancer patients into low-, intermediate- and high-risk of recurrence groups. Using these
test results, both patients in the low risk group who did not receive unnecessary

5
chemotherapy and patients in the high risk group who did receive prophylactic
chemotherapy saw significant improvements in overall survival [20]. This assay is now
the standard of care and included in clinical guidelines to help predict the risk of distant
recurrence and responsiveness to chemotherapy.
MammaPrint® is another commercially available microarray-based diagnostic that
evaluates the expression of 70 genes to assess the risk of recurrence 5 to 10 years
following diagnosis for early stage breast cancer patients, and can also be used to
determine whether or not patients will benefit from chemotherapy. Whereas Oncotype
DX® is only suitable for ER+ luminal A patients, Mammaprint® can be used for both ER+
and ER- tumors, thus broadening applicability. Agendia, the company that makes
Mammaprint®, is also currently evaluating the Blueprint® assay, an 80-gene test that can
more accurately diagnose HER2+ breast cancers and its subtypes compared to
immunohistochemical- or in situ hybridization-based methods. For Blueprint®-identified
HER2+ patients, this test provides improved guidance on the addition of pertuzumab
(Perjeta®) to standard neoadjuvant therapy (chemotherapy with trastuzumab treatment)
[21].
Other tumor profiling tests are currently being developed to help optimize clinical
outcomes for different subsets of breast cancer patients. However, both tumor
heterogeneity and patient heterogeneity add layers of complexity that present a
substantial clinical challenge [22]. One study identified 6 distinct subtypes within the triple-
negative/basal-like molecular subtype that responded differentially to various drug
treatments and were associated with differential clinical outcomes [23]. And although
tumor genomic testing has improved precision medicine and patient outcomes, cancers

6
that are resistant to treatment and recur or metastasize suddenly become a more dire
and challenging situation. Despite recent advances in immunotherapy treatments, such
as immune checkpoint inhibitors, adoptive T cell therapy, and cancer vaccines for
controlling and preventing metastasis, metastasis remains the leading cause of cancer-
related deaths. Furthermore, patients with advanced and metastatic breast cancers
require the most resources and shoulder enormous emotional and financial burdens [24].
The crucial first step to developing rational treatment options that take patient quality of
life and financial burden into account is understanding the complex, multistep biology
underlying the process of how a tumor gains the ability to evolve, survive, and spread
throughout the body.

7

Figure 1.1. Molecular subtypes of breast cancer. The five main molecular subtypes of breast
cancer are associated with the expression of hormone receptors (HRs), including estrogen and
progesterone receptors, and HER2 gene amplification status. Luminal breast cancers have the
best prognosis whereas triple negative breast cancers have the worst. Molecular subtyping can
be used to help guide treatment options. Created with BioRender.com

1.4 The metastatic cascade
The process by which a cancer cell spreads and travels from the site of origin is
called metastasis and occurs in several distinct steps termed “the metastatic cascade”.
For a tumor cell to successfully seed and form a secondary cancerous lesion, it must
survive and progress through all the steps of the metastatic cascade, which starts with
tumor development at the site of origin.
Tumor development can be broken down into three stages: initiation, promotion,
and progression. Initiation is when a normal cell undergoes a genetic mutation that

8
increases the risk of developing into a cancer. Next is tumor promotion, whereby the
mutated cell gains the expression of tumor-promoting genes or loses the expression of
tumor-suppressing genes, ultimately leading to a proliferating, uncontrolled mass of cells.
The last stage of tumor development, progression, is defined by a gain of invasive and
aggressive characteristics that increases the metastatic potential of the tumor mass.
Tumor development is associated with neoangiogenesis (distinct from angiogenesis, the
term used in normal blood vessel development), which is the induction of new, immature,
and aberrant vascular structures in or around tumors to facilitate the diffusion of waste
and nutrients. Metastasis begins once the proliferating tumor breaches the basement
membrane and gains invasive and migratory properties (Figure 1.2). Invasion and
migration are accompanied by a remodeling of the local tumor microenvironment and
extracellular matrix (ECM) and recruitment of tumor-promoting stromal cells.
The next step in the metastatic cascade is intravasation, or entry into the
circulatory system via hematogenous or lymphatic routes (Figure 1.2). Tumor cells can
intravasate passively by simple proximity to aberrant blood vessels and by traveling down
gradients of fluid pressure. Intravasation can also be an active process, where an invasive
cancer cell exhibiting loss of cell-cell and cell-matrix adhesions uses adhesion receptors
to attach to the vascular endothelium and proteases to degrade the endothelial cell barrier
and gain entry into the bloodstream [25]. The intravasation process can also be facilitated
by stromal cells such as tumor-educated macrophages [26], cancer-associated
fibroblasts [27], and tumor-recruited neutrophils [28].
Tumor cells that have entered into circulation are referred to as circulating tumor
cells (CTCs) (Figure 1.2). CTCs must be able to overcome apoptosis induced by the lack

9
of cell-cell and cell-matrix contacts (termed anoikis), survive the physical shear stresses
imposed by turbulent interactions with other blood cell components, bypass immune
surveillance, resist oxidative stress, and cope with the lack of growth factors and
cytokines [29]. Within the vascular environment, the multitude of physical and biological
stressors leads to destruction of the majority of CTCs.
The small percentage of surviving CTCs must then exit the bloodstream and into
distant tissues in a process called extravasation, which can also be a passive or active
process (Figure 1.2). Passive extravasation can occur when a CTC, due to its size or
anatomical considerations of the distant site, gets lodged or occluded in the capillary bed
of a distant site during circulation. Active extravasation has been described where a CTC
traveling through the bloodstream slows down and attaches to the luminal side of the
vascular endothelium and comes to a stop. Adhered CTCs can then cross the
endothelium into the distant organ to become a disseminated tumor cell [30].
Disseminated tumor cells that have gained access to a distant organ must then
overcome the barriers of a new microenvironment to colonize the distant site and present
overt metastases (Figure 1.2). These metastatic lesions are proliferating masses of tumor
cells that survived the multiple treacherous stages of the metastatic cascade by
undergoing further mutations and acquiring additional genetic advantages compared to
the primary tumor, contributing to its survival. Although clinically detectable metastases
are generated by proliferating tumor cells, there can also be a population of cells that
neither die nor proliferate but remain in an undetectable, quiescent state known as
dormancy [31].

10
Each stage of the metastatic cascade is a very niche and complicated area of
study. The temporal dynamics of metastasis are also not well understood. Metastasis was
initially considered a late event in cancer progression – the result of an accumulation of
genetic aberrations leading to a progressive increase in metastatic propensity. However,
it is now understood that metastasis can be an extremely early event. In one 2003 study,
for example, copy number variation (CNV) analysis of single disseminated breast cancer
cells detected in bone marrow compared to that of the bulk primary tumor showed
significantly fewer CNV alterations [32]. This indicates that the disseminated tumor cells
occurred as an early event and arose from a less progressed mutational state. But
whether early metastasis is typical of all cancers or just a specific molecularly-primed
subset of cancers is yet unknown. Recent advances in liquid biopsy research are critical
to answering these questions, by providing a non-invasive method for repeated blood
sampling to monitor patient progress and tumor evolution. This emerging landscape of
liquid biopsies presents a unique opportunity to improve our knowledge of metastasis.

11

Figure 1.2. Overview of the metastatic cascade. The stages of metastasis include tumor cell invasion at
the primary tumor site, intravasation into the bloodstream, travel through the circulatory system to reach a
distant site, extravasation out of the vasculature, and colonization at the site of metastasis. Created with
BioRender.com.

1.5 Peripheral blood liquid biopsies
The blood contains a dynamic mixture of biological materials that can originate
from anywhere in the body and can therefore reveal information about what kinds of
biological changes are occurring in anatomical sites that may be directly inaccessible.
There are many types of analytes that can be obtained through a liquid biopsy of a
peripheral blood sample from a cancer patient, including circulating tumor DNA,
microRNA, protein fragments from dying tumors, tumor exosomes, and CTCs. These
analytes can be used to develop a more holistic understanding of the behavior of a tumor
within a patient such as treatment response, tumor evolution, or metastasis. To this end,
many different research groups are racing towards innovative and improved analyte
separation techniques as well as developing more sensitive quantification methods to
assess the potential for clinical insights and improve precision medicine.

12
1.6 Circulating Tumor Cells (CTCs)
One exciting area in the field of liquid biopsies is the study of CTCs. These are
cells that have successfully completed a portion of the metastatic cascade and are
therefore considered to be the precursors to metastasis, or metastatic intermediaries.
Historically, CTCs were extremely difficult to study because their relatively low abundance
(1-10 CTCs among millions of white blood cells (WBCs) and billions of red blood cells
(RBCs)) meant that there were only a few types of assays that could be performed, which
was to fix and stain blood samples for the detection and enumeration of CTCs in a given
amount of blood. Over time, many studies have reported significant prognostic and
predictive value of CTC enumeration. For example, in metastatic breast cancer, detection
of 5 or greater CTCs in a 7.5 mL blood sample was significantly correlated with decreased
progression-free and overall survival in a large multi-site study [33]. There have also been
similar studies and correlations discovered in prostate [34], colorectal [35], and lung
cancers[ 36] indicating the broad application and value of CTCs. Due to the rarity of CTCs
and because the process of detection and enumeration uses up the entire available
sample, there have been great efforts to increase the amount of information able to be
extracted from a small number of CTCs. One such technology is the high-definition single-
cell analysis (HD-SCA) assay that combines phenotypical and morphological data with
genome wide CNV profiles from each and all of the CTCs detected in a given blood
sample [37].
Since first discovery in 1869 until recently, CTCs were simply fixed, stained, and
observed under a microscope. In 2014, however, Yu et al. published a transformative
report detailing a method to isolate CTCs from the peripheral blood of breast cancer

13
patients for the ex vivo expansion and establishment of long term cultures of CTCs,
thereby providing a renewable source of biological material [38]. This has paved the way
for the characterization of the biology of rare CTCs in various in vitro and in vivo assays
that require large cell numbers. When xenografted into immune deficient mouse models,
these patient-derived CTC lines could recapitulate and in one case, predict, the sites of
metastatic relapse in the corresponding patient [39]. These cultured CTCs were also used
to identify drivers of breast-to-brain metastasis [39], paving the way for future research
into CTC biology with broader implications for the advancement of precision medicine and
therapeutic treatments.
It is important to note, however, that there are only a few representative CTC lines
growing in culture. There are six stable, long term CTC cultures derived from luminal
breast cancer patients [38], one from prostate [40], and one from colon cancer [41]. Many
other attempts to isolate and establish long term CTC cultures have failed, highlighting
the difficulty of this process as well as the potentially unique conditions that may be
required for different cancers and subtypes. CTCs also display a huge range of
heterogeneity in morphology and biophysical and molecular characteristics that make it
difficult to isolate the total population of true CTCs from other blood components.
Furthermore, a single blood sample merely represents one point in time and can therefore
obscure the molecular and temporal dynamics of CTCs over time, and is therefore
susceptible to sampling variation. Regardless of these caveats, the clinical value of CTCs
is abundantly clear and highlights the critical need for and future impact of understanding
CTC biology.

14
CHAPTER 2
ENGINEERING AND PHYSICAL APPROACHES TO
INVESTIGATING CIRCULATING TUMOR CELLS

2.1 CTCs in the bloodstream are metastatic precursors
CTCs are tumor cells that have entered into the bloodstream. Although the vast
majority do not survive the numerous stressors in the fluid environment [42], the small
fraction of CTCs that are able to resist destruction and elimination have the potential to
form metastatic colonies and can therefore be considered metastatic precursors, or
intermediaries. Additionally, the survival duration of CTCs in circulation was found to be
the most critical step of the metastatic cascade contributing to the development of
clinically detectable metastases, according to an in silico modeling study [43]. CTC
survival in the circulatory system is dependent on a number of factors including
interactions with the physical environment and intrinsic biophysical characteristics, such
as stiffness and deformability. CTCs have also been reported to display interesting
biological characteristics that are distinct from primary tumors. In one study, for example,
84% of CTCs analyzed from 19 HER2- breast cancer patients had acquired HER2
expression and some were observed to interconvert between HER2+ and HER2- states
[44], highlighting the unique properties and molecular plasticity of CTCs. Altogether, these
findings lend further importance and urgency to understanding and characterizing CTCs.
Research is limited by the rarity of these cells and the difficulty of live cell isolation.
Despite these limitations, understanding CTC biology and the underlying molecular and

15
biophysical mechanisms contributing to metastasis, organotropism, and other clinically
relevant parameters will generate a more accurate and comprehensive understanding of
this devastating phenomenon.

2.2 CTC detection and live cell isolation methods
Currently, the only FDA approved system for CTC detection is the CellSearch®
platform. This system uses an antibody-conjugated ferromagnetic approach to first isolate
all epithelial cell adhesion molecule (EpCAM)-expressing cells in the buffy coat (thin cell
fraction in an anticoagulated blood sample that contains a concentration of WBCs and
platelets after centrifugation) of a blood sample. These EpCAM+ cells are stained with
fluorescently labeled antibodies against cytokeratins 8, 18 and/or 19 (pan-cytokeratin,
pCK) and CD45, aligned to a single focal plane by a magnet, and then microscopically
identified as a true CTC by pCK+ and CD45- status.
The caveat to positive antibody-based selection approaches, like the CellSearch®
system, is that CTCs are reported to be an extremely heterogeneous population. For
example, a CTC population within an individual exhibits a spectrum of gene expression
along the epithelial-to-mesenchymal transition (EMT) continuum that can preclude more
mesenchymal-like CTCs from EpCAM-based detection [45]. These CTCs also show
epithelial-mesenchymal plasticity (EMP) and can possess molecular characteristics of
one or the other, or even both simultaneously [46]. In addition to the molecular
heterogeneity, there is also a large degree of phenotypic heterogeneity that is correlated
with metastatic behavior. For example, CTCs can exist as single cells, or less frequently
as homotypic clusters, or as heterotypic cell clusters that can include platelets and other

16
cell types [47]. Homotypic CTC clusters were reported to have 23-50-fold increased
metastatic potential in mouse models [48] and heterotypic CTC clusters, such as those
coated with platelets, were also reported to be more resistant to anoikis [49] and shielded
from destruction by natural killer cells [50].
In addition to molecular and phenotypic heterogeneity, CTCs are also reported to
show a wide range of diversity in physical properties not only within a single patient, but
also between CTC populations of different cancer types. For example, whereas ER+/PR+
breast cancer CTCs are reported to have diameters ranging from 9 to 19 μm, most
melanoma CTC diameters are >12 μm [51]. Many groups are therefore focusing on
developing technologies that can isolate CTCs based on various combinations of
molecular and biophysical properties. As mentioned in the previous chapter, there are
currently only a handful of patient-derived CTC lines that have been successfully
established for long-term culture and expansion: six from luminal breast cancer patients
[38], one from prostate [40], and one from colon cancer [41]. Although advanced
technologies exist for the fixation, detection, and enumeration of CTCs in set volumes of
patient blood (such as High Definition-Single Cell Analysis (HD-SCA) platform [37] and
the RareCyte® commercial isolation platform [52,53]), there is an urgent need for further
development of techniques that can isolate live, rare CTCs for ex vivo expansion and in
vitro and in vivo characterization. Because antibody-based approaches select for a subset
of heterogeneous CTC populations and can impact downstream viability, here we focus
on technologies that exploit the biophysical properties of CTCs for label-free, live cell
detection and isolation.

17
Negative selection approach to CTC isolation
Because antibody-based positive selection approaches selectively enrich for only
a subset of CTCs, many groups are working on the development of negative selection or
antibody-agnostic approaches to CTC isolation. PIC&RUN is a recently developed assay
based on the AccuCyte® density-based cell separation instrument and RareCyte®
robotic detection and retrieval system [54]. In the negative selection module of the
PIC&RUN protocol, the buffy coat of a blood sample is retrieved by AccuCyte®-based
separation and subsequently stained with a live cell dye and an antibody cocktail
detecting immune cells. The RareCyte® system is then used to detect cells that are
labeled with the live cell dye but negative for immune markers. The identified cells are
then isolated by a computer-controlled robotic needle and transferred to a cell culture
vessel. The CTCs isolated in this antibody-agnostic, negative selection approach
remained viable and were able to be expanded in culture [54].
Compared to antibody-based positive selection approaches, the label-free
negative selection approach captures a broader range and higher numbers of CTCs
[54,55,56]. However, CTC selection without the use of antibodies is difficult and requires
the integration of multiple approaches (such as the use of live cell dyes and leveraging
the physical properties of CTCs) in various combinations and permutations optimized for
different types of cancers. Too much cell manipulation can impact downstream viability
whereas too little can lead to impurities and contaminants that contribute to noise and
obscure biological data.

18
Density-gradient-based CTC isolation
The majority of CTCs have densities that are similar to that of the buffy coat
(<1.077 g/mL) [57], although some very small CTCs and CTC clusters can have much
lower or higher densities. Density-based approaches therefore involve mixing blood
samples with solutions that are formulated to precise densities in order to achieve
separation of RBCs and granulocytes that have densities >1.077 g/mL from the less-
dense CTC-containing buffy coat layer.
In general, the AccuCyte® instrument and RosetteSep™ procedure are both
density-based CTC isolation methods that work by the same principles but differ in their
approaches to the challenges of getting a clean separation of the different cell layers. The
AccuCyte® overcomes this challenge by utilizing a float with specific density in the blood
collection tube, which settles between the RBC layer and the buffy coat after
centrifugation. A CyteSealer® device then forms a tight seal around the float to accurately
and reproducibly separate the buffy coat from the RBCs without human manipulation. The
RosetteSep™ procedure, on the other hand, works by crosslinking the undesired cells
(like leukocytes) to the RBCs before density centrifugation, thereby pelleting unwanted
cells and leaving behind CTCs for downstream analysis. Of note, the RosetteSep™
procedure using CD45 depletion prior to density-based separation with Ficoll-Paque was
used to successfully isolate viable CTCs from one prostate cancer patient to establish a
patient-derived prostate cancer CTC line [40].

19
Size/deformability-based CTC isolation
CTCs from different types of cancers are reported to be different sizes. In a study
that used the CellSearch® platform to detect a total of 71,612 patient-derived CTCs from
across different cancer types, the median computed diameter was reported to be 12.4
μm in breast cancer, 10.3 μm in prostate cancer, 7.5 μm in colorectal cancer, and 8.6 μm
in bladder cancer[58]. Leukocytes, on the other hand, had a median computed diameter
of 9.4 μm in this study [58], whereas other groups that used different detection and
measurement methods have reported diameters ranging between 7-20 μm [59].
In addition to cell size, deformability is another biophysical parameter that can be
exploited to isolate CTCs. Deformability is defined as the amount of change in shape
under application of a defined amount of pressure without rupture. Related to
deformability is the concept of cellular stiffness, which is defined as the amount of
resistance to deformation under an applied force. Although higher deformability of tumor
cells is correlated with increased metastatic propensity [60,61], they are reported to be
comparatively less deformable than RBCs and WBCs [62].
The Parsortix® system takes advantage of both the size and deformability
parameters of CTCs to isolate them from other blood components. This technology
involves injecting a whole blood sample through a filtration cassette that captures CTCs
based on their large size and low deformability (high resistance to compressive forces)
while letting other blood components pass through. Although Parsortix® is able to isolate
significantly more mesenchymal CTCs compared to the antibody-based CellSearch®
system [63], CTC sizes can overlap with some WBCs, and the deformability of different
types of cancer cells compared to the surrounding blood components is yet to be further

20
explored. This leaves room for improvement on current size- and deformability-based
capture techniques.
Inertial particle focusing is another size-based separation technique. This method
takes advantage of the physics of fluid forces and particle dynamics flowing through a
microchannel [64] to separate out CTCs. The first prototype was designed with a double
spiral inertial focusing configuration [65,66], but subsequent iterations have investigated
various designs to achieve improved CTC separation [67,68,69,70]. The CTC-iChip,
which employs a sinusoidal wave like configuration coupled with hydrodynamic cell
sorting and magnetophoresis, has gained traction as a popular platform for the inertial
separation of CTCs [51]. This platform has been used to successfully establish several
long term cultures of patient-derived CTC lines from breast cancer patients [38].

Acoustofluidic-based CTC isolation
Acoustofluidics is a field of research that uses tunable ultrasonic acoustic waves
to achieve microfluidic-based cell separation. Acoustofluidic separation of CTCs takes
advantage of how acoustic waves interact with the physical properties of fluids and the
particles embedded within, and enables a contactless and more biocompatible approach
to CTC isolation. The first proof-of-concept acoustofluidic separation of tumor cells from
WBCs was developed using prostate cancer cell lines spiked into whole blood [71]. In this
first generation prototype of acoustofluidic-based tumor cell isolation from blood samples,
RBCs were depleted by lysis buffers before processing the remaining WBCs and tumor
fraction by acoustic focusing [71]. Since then, protocols have been optimized to detect
CTCs from breast cancer [72] and metastatic castration resistant prostate cancer patients

21
[73]. Of note, these protocols require a RBC lysis step, which may result in loss of rare
CTCs or be incompatible with downstream ex vivo culture. Furthermore, acoustofluidic-
based approaches are not yet capable of isolating viable CTCs from patient samples,
thereby presenting opportunities for improvement.

Dielectrophoresis (DEP)-based CTC isolation
Many molecules possess an inherent charge or polarity. Cell composition and
physical properties, such as size and shape, affect the overall electrical properties. DNA,
for example, is negatively charged, whereas the phospholipids that compose cell
membranes endow an overall negative electrical surface potential that attracts positively
charged molecules in the environment. Different cell populations therefore possess
unique electrical characteristics that can be exploited in dielectrophoretic (DEP)-based
separation of CTCs. DEP-based approaches involve application of electrical fields to
either move or retain CTCs and separate them from leukocytes. ApoStream™ is a
commercially available DEP-based technology that was first developed in 2012 [74],
which separates cancer cells from peripheral blood monocytes (PBMCs) and has been
proven to work on patient samples [75]. DEPArray™ is a similar technology, but has the
ability to generate ~30,000 DEP “cages” that can each individually levitate and capture
single CTCs avoiding adhesive interactions with other cells and surfaces [76].

Caveats and future directions
Different isolation methods have different downstream consequences – some
methods affect cell viability and are therefore not suitable for live cell isolation. Employing

22
singular physical approaches to CTC detection and isolation is also not recommended
due to overlapping biophysical parameters – for example, employment of a size-based
approach to isolating CTCs from bladder cancer would be unadvisable since they are
reported to have diameters similar in range to leukocytes [58]. Moreover, some CTC
isolation methods involve WBC depletion, but CTCs can exist in heterotypic cell clusters
with WBCs [77] and depletion can unintentionally reduce the already low numbers of
CTCs. It is also evident that the molecular and physical properties of CTCs can vary
widely between different types of cancers. Taking a singular approach to broadly capture
a heterogeneous population of CTCs is unlikely. However, different approaches can be
individually optimized and integrated together in various ways in order to improve the
broad recovery of heterogenous CTCs. For example, density-based RBC depletion with
WBC crosslinking can be combined with acoustofluidic isolation to improve CTC viability,
recovery, and purity. In line with both inter-patient and inter-cancer CTC heterogeneity,
the standard criteria for CTC identification may require refinement based on a cancer-
type or subtype basis. As such, large scale profiling of the range of physical parameters
of CTCs from different cancers will be needed in future studies to improve our knowledge
of CTC biology.

2.3 Introduction to the physical environment of CTCs
The human vascular system is a highly variable and physically treacherous
environment in which CTCs must survive to successfully metastasize. It encompasses a
large range of lumen diameters from the large aorta to tiny capillaries, possesses different
functional properties depending on whether it is an arterial or venous vessel, displays a

23
wide range of organ-specific endothelial permeability (liver versus blood-brain barrier),
and experiences different blood flow velocities depending on the amount of cardiac output
at rest or during exercise. In addition to the rate of flow, there are also different types of
blood flow patterns, such as disturbed, or turbulent flow in aberrant blood vessels near
tumor cells, and undisturbed, laminar flow in healthy vessels. Additionally, CTCs
experience collisions with other blood cells such as RBCs and WBCs, and the branch
points of furcating blood vessels and the vessel wall. In the smallest capillaries where
CTC diameter is larger than the capillary diameter, tumor cells experience compressive
forces and undergo large deformations, the effect of which on CTC biology needs further
investigation. Furthermore, differences in cardiovascular anatomy and function, as well
as the development, progression, and presentation of vascular diseases are very different
between men and women [78]. It is therefore conceivable that physiological blood flow
profiles measured in men may not be applicable for in vitro experimentation of female
breast cancer. Modeling the human vascular system is therefore incredibly complex and
requires a more thorough description of such parameters in both healthy and disease
states if we are to grasp its full impact on CTCs and the metastatic cascade (Figure 2.1).

24

Figure 2.1. The physical environment of CTCs. CTCs experience various physical stressors as they travel
through the circulatory system. These stressors include fluid shear stress, collisions with RBCs and the
vascular endothelium, traction forces, and compressive forces. Created with BioRender.com

2.4 CTCs experience fluid shear stress
The velocity of blood in the center of a blood vessel is higher than that of blood
moving closer to the vessel wall [79]. This difference in velocity generates a force vector
that is parallel to the cross section of the blood vessel and is defined as fluid shear stress
(FSS) whereas the frictional forces exerted by blood flowing against the endothelial wall
is specifically called the wall shear stress (WSS). WSS has been well studied as a factor
influencing remodeling of the vascular endothelium, atherosclerosis, and general health
of blood vessels. The effect of FSS on CTCs or adherent cell lines grown in suspension

25
conditions has been the most well studied physical parameter in vitro. Physiological FSS
in humans is reported to range from 0.5-4 dynes/cm
2
in venules and 4-30 dynes/cm
2
in
arteries [80], but can vary greatly based on type of blood vessel, location, physical activity
and underlying health conditions.
In a peristaltic pump-based microfluidic assay where cells are continuously
circulated for 4 hours, the application of high FSS of 60 dynes/cm
2
, associated with
vigorous exercise, was found to destroy more CTCs than physiological levels of FSS that
are typical of human arteries in the resting state [81]. The high FSS condition resulted in
the death of 90% of CTCs within the first 4 hours of circulation and the majority of surviving
cells underwent apoptosis within 24-hours. In clinical studies as well, physical activity was
significantly associated with reduced CTC count among stage I-III colon cancer patients
[82]. Reduced blood flow, on the other hand, was permissive for the arrest and stable
adhesion to the vascular endothelium, and subsequent successful extravasation in a
zebrafish model [83], highlighting the potential importance of blood flow velocities and
shear stress in CTC survival. Although these studies are informative, there are still many
unknown parameters that make it difficult to accurately model and test CTCs. For
example, it is unknown how much shear stress a CTC experiences and for how long in
the human body. One study applying an in vivo flow cytometry (IVFC) approach with a
highly metastatic and adherent MDA-MB-231 cell line in mouse models found that single
MDA-MB-231 cells could be detected circulating in the bloodstream with a half-life of
about 25-30 minutes whereas the half-life of circulating tumor clusters was only 6-10
minutes [48].

26
Other groups looked at the contribution of FSS to the molecular and phenotypic
characteristics of CTCs. There are various reports on the effect of FSS on the viability
and proliferation of colon cancer CTCs [84], on the induction of EMT in breast cancer
CTCs [85], on induction of RhoA/Actomyosin-dependent resistance to mechanical
destruction [86], and on CTC metabolism [87]. However, the biggest caveat of these
studies is that they were performed on adherent cell lines simply grown or assayed in
suspension conditions to mimic the circulatory environment, and not on real CTCs. The
differences in physical and molecular properties between CTCs and adherent cell lines
across different cancer types has been well characterized [58,88,89,90]. Understandably,
the rarity of CTCs in blood samples combined with the paucity of available CTC lines that
can be cultured and expanded long term for experiments necessitates the use of CTC-
like cells [91]. However, the generalization of phenomenon observed in CTC-like cells to
true CTCs must be carefully considered and the terminology used to describe CTC
research needs to be more specific, obvious, and rigorously defined. Despite these
caveats, the experimental approaches to studying tumor cells in suspension have proven
to be simple for in vitro approaches and provide useful insights. Future experimentation
on existing CTC lines or freshly isolated patient CTCs will advance our understanding of
FSS on CTC biology.

Engineering approaches
Controlled FSS can be applied to cells in various microfluidic configurations.
Syringe-pumps connected to microfluidic devices made of polydimethylsiloxane (PDMS)
channels adhered to glass coverslips can model the effect of various shear flow

27
conditions on cells. In a recently reported method, glass coverslips could be directly or
indirectly functionalized with various ligands that mimic endothelial cell surface receptors,
such as vascular cell adhesion protein 1 (VCAM1) [92]. Indirect functionalization involved
embedding freely diffusible receptors in a lipid layer, thus more accurately modeling fluid
cell membranes. Direct visualization of cell morphology and adhesion under various flow
conditions is possible through the glass coverslip, and adhered cells can later be analyzed
for transcriptional or signaling changes.
Peristaltic microfluidic devices [93,81] are similar except that they are closed-loop
systems where the rotational motion of various rollers pumps cell-containing fluid
suspensions through the microfluidic device at desired speeds to control flow rate and
shear stress in a pulsatile manner.
Cone-and-plate shear devices [94,95] were the earliest techniques to examine the
effect of shear stress on cells and involve a rotating cone inside a flat plate filled with cells
and fluid. The rotation rate of the cone can be manipulated to control the amount of shear
stress applied to the sample in the plate. This technique is based on mechanical principles
of non-Newtonian fluids, such as whole blood. A parallel-plate flow chamber, or
viscometer, is another device where a cell suspension is injected between two parallel-
plates with a defined gap size, and the rate lateral motion of one of the plates exerts a
shear stress on the sample in the gap.
In vivo flow cytometry is a type of intravital microscopy method in which cells are
injected into a live, anesthetized animal, and fluorescently labeled cells are detected in a
blood vessel that is between 20-50 μm in diameter in the mouse or rat ear [96]. This
method involves a more complicated set up but can be used to quantitate CTC dynamics

28
in the circulatory system over time, with the caveats being differences between the animal
and human circulatory systems and that the animal must be immobilized. Since initial
invention, this technique has been further optimized for increased sensitivity and sampling
rate, and for detection in larger vessels [97]. Development of a different type of intravital
microscopy, where a microscopic window is mounted over the dorsal skin fold to
illuminate the superficial blood vessels of an awake and mobile mouse [98], removes the
requirement for an animal to be immobilized and can potentially be used to examine the
effect of exercise-induced increased blood flow and shear stress on CTC biology and
circulatory dynamics.

2.5 CTCs experience numerous collisions
As previously mentioned, CTCs can exist as clusters or as single cells. Both can
also exist together with non-tumor blood components such as leukocytes and platelets.
The differing configurations of CTC composition can affect their survival in circulation as
they collide with surrounding RBCs, WBCs, and the vascular anatomy. For example,
CTCs coated in platelets were reported to be physically shielded from destruction by
natural killer cells [99] and adhere better to the endothelial wall [100], both of which
contribute to an increased potential for metastasis. In general, physical interactions
between CTCs and their physical environment have been most extensively studied by in
silico mathematical modeling. Specifically, these collisions generate wall-directed forces
that change CTC trajectory and cause margination of cells from the center of a blood
vessel to the periphery where fluid velocity is comparatively slower [101] and adhesive
interactions can occur [83]. Computational experiments predict that tumor cells frequently

29
adhere to microvascular bifurcations and that increasing numbers of collisions with RBCs
is positively correlated with tumor adhesion to the endothelium [102].
Collisions can also result in cell rupture and death. Only cells with favorable
intrinsic biophysical characteristics and cytoskeletal composition are able to survive
collision events. Various studies have examined the contributions of actin and tubulin to
CTC survival and metastatic propensity [103]. For example, the ratio of globular to
filamentous actin was reported to be higher in malignant cells compared to normal cells
[104], but exactly how that translates to mechanical resistance to rupture is unclear.
Another study used an in vitro microfluidic device to investigate the correlation between
metastatic propensity and resistance to fragmentation upon flow-induced collision at a
micropillar-based bifurcation. Results showed that highly metastatic prostate cancer cells
were more resistant to fragmentation compared to weakly metastatic cells [105].
In vitro methods for studying the effect of collisions on CTCs are limited to
microfluidic platforms that mimic obstacles encountered in the bloodstream. The obstacle
can be on a single cell level [105] or on a much larger scale to model turbulent flow of a
mixture of cells [106].

2.6 CTCs experience traction forces
Besides physical occlusion, which is a passive process, CTCs can adhere to the
endothelial cell lining in a multi-step active process that is very similar to leukocytes [107].
CTCs initially experience margination by collisions with RBCs. Closer to the endothelial
wall, slower blood flow velocities allow marginated tumor cells to form weak adhesions
and begin to roll along the endothelium. This rolling motion increases the probability and

30
strength of adhesive contacts between the cancer cell and the endothelium that
eventually results in cell arrest. The continual movement of blood flow over the adhering
tumor cell applies a FSS that not only impacts cell morphology, but also contributes to
traction forces exerted by the cell on the endothelium [108]. Traction forces exerted by
leukocytes on vascular endothelial cells were observed to facilitate the opening of
junctions before transendothelial migration [109], but whether CTCs also co-opt this
mechanism is yet unclear. FSS can also increase the adhesion strength of already
adhered cells – this was according to a study that used adhered glioma cells with or
without FSS exposure [110] in a novel live-single-cell extractor (LSCE) assay, which
measures the amount of time it takes for a detaching cell to move a specified distance
[111]. Adhesion to the endothelium is also affected by the cortical tension exhibited by
CTCs. FSS can induce higher cortical tension [110], which is correlated with rounded
morphology and decreased deformability [110] as well as faster margination and
prolonged time in circulation [112]. As adhesive strength between CTCs and the
endothelium increases, lower cortical tension, and therefore higher deformability, allows
for lower rolling velocities and a higher degree of cell spreading that allows for the
formation of firm adhesions and decreased time in circulation [113]. Once adhered,
metastasizing cells must continue to resist rupture under the forces experienced and
achieve transendothelial migration.

Engineering approaches
In addition to microfluidics-based approaches to examine CTC rolling adhesion on
selectin coated channels, traction forces can be measured using PDMS micropillars. Cells

31
are seeded on a bed of micropillars that can bend like a spring depending on the pillar
dimensions. Taking into account size and spacing of the micropillars, traction forces can
be calculated from the amount of displacement experienced by the apex of the micropillar
[114].
Traction force microscopy [115] is another method that couples the amount of
substrate deformation with the amount of traction force exerted by a cell on that substrate.
The substrate can be a wide range of ECM components or hydrogels that are embedded
with fluorescent beads. Traction forces upon the substrate causes a measurable
displacement of the fluorescent beads that can be quantified and converted into a traction
force.
Cortical tension measurements are typically achieved using micropipette
aspiration assays, in which a vacuum is applied to the surface of the cell membrane in
order to induce a membrane deformation. The amount of negative pressure applied and
the amount of membrane deformation can be used to determine the cortical tension of a
single cell. High throughput micropipette aspiration arrays have been developed to
assess the individual cortical tensions of a population of cells [116].

2.7 CTCs experience compressive forces
It takes less than a minute for blood to circulate around the entire body. When
CTCs enter small capillary vessels, they can either squeeze through or become trapped.
The number of compressive events and the amount of force experienced by the cell
depends on the circulatory half-life, the amount of time spent trapped in the blood vessel
(which is related to cell deformability [62]), the size differential between the CTC and the

32
vasculature, and whether they exist as single cells or clusters. Many of these parameters
are unknown, but some studies provide clues. CTC clusters, for example, were reported
to exhibit upregulation of plakoglobin, a desmosomal protein that links the cytoskeletons
of adjacent cells together and facilitates tight cell-cell adhesions [48]. These plakoglobin-
enriched clusters showed greater resistance to apoptosis than single CTCs and had 23-
50-fold higher metastatic potential [48]. Perhaps the supracellular cytoskeletal scaffold
generated by plakoglobin upregulation and enhanced desmosomal junctions is a potential
mechanism by which mechanical resilience is achieved by distribution of compressive
forces [117] across the CTC cluster therefore leading to the observed phenotypes. In
contrast to the idea that CTC clusters maintain their structure and organization under
compression, a different study reported that individual cells within an aggregate can
rearrange into single-file chain-like configurations in order to squeeze through capillary
sized 5-10 μm constrictions [118]. This indicates that each individual cell could experience
the same amount of compressive forces, which is highly dependent on cluster size an
shape [119]. These are merely speculations, however, and there are many unknowns
that warrant further investigation.
On a cellular level, compressive forces can lead to either deformation or rupture.
Individual CTCs were observed to undergo extreme physical elongation and deformation
in capillaries. Tumor cells observed inside mouse capillaries had 4x increased cellular
length and 1.6x greater nuclear length compared to uncompressed cells [120].
Furthermore, nuclear volume and stiffness have been inversely correlated with the ability
and speed at which a cell moves through a smaller constrictive vessel [121]. This
suggests that cells with nuclei that are more resistant to compressive forces remain

33
lodged in capillary vessels for longer periods of time and can increase the probability of
metastatic colonization. Nuclear flattening also results in increased nuclear transport of
the yes-associated protein 1 (YAP1) transcription co-factor that activates genes involved
in cancer progression [122]. Therefore, the ability of CTCs and their nuclei to deform and
resist rupture and fragmentation under compressive forces determines the rate of
metastasis [123].
Indeed, highly metastatic cell lines have been reported to be softer and more
deformable than less metastatic cell lines [61], and their parental, isogenic counterparts
[124]. Atomic force microscopy (AFM) measurements of metastatic cancer cells isolated
from the pleura of patients with lung, breast, and pancreatic cancer were also 70% softer
than benign cells [125]. In a study where cell stiffness was directly manipulated by hypo-
or hyperosmotic environmental conditions to induce swelling or compression,
respectively, increasing the tumor cell stiffness at the periphery of a breast cancer
organoid led to a significant decrease in the formation of invasive structures [126].
Furthermore, mechanical compression of brain tumor cells impairs proliferation but can
induce migration via MEK1/Erk1 pathway activation [127]. Despite the evidence for the
effect of compressive forces on tumor cell behavior and the biophysical characteristics
related to modulation of cellular response, there are very limited studies specifically
examining CTCs. CTC survival and subsequent metastasis is likely a combination of both
a selective and adaptive event and future investigations are needed.
Though there are many ways to apply compressive forces on adherent cells or
tissues, the methods described to date for suspension cells are mainly pore- or
microfluidics-based. Pore-based approaches involve driving cell suspensions through

34
membranes containing micron-sized pores. However, microfluidic devices fabricated with
constricted microchannels of specific dimensions can better model the anatomy of a blood
capillary. A high throughput microconstriction array has been developed that can drive
cells through a constrictive channel and simultaneously provide readouts for the
deformability and stiffness of each individual cell in the given population [128]. However,
the dimensions of the microchannels do need to be optimized for different cell lines due
to cell/nuclear size considerations in order to find the optimal balance between high
throughput and high sensitivity.

2.8 Conclusions
Mechanical phenotyping of CTCs is quickly being investigated as a way to predict
cancer progression. Characterization of stiffness, deformability, and fragmentability are
all emerging physical approaches to generating a better understanding of measurable
biophysical parameters associated with the metastatic propensity of CTCs. In spite of
these technological advances, the challenge of CTC rarity remains. This has prompted
the use of adherent cell lines grown briefly in suspension conditions to model CTCs;
However, the validity of using adherent cell lines as a model for CTCs has never been
tested and there are yet no reports comparing these two distinct biological resources.
Since the establishment of the first long term cultures of CTC lines generated from
breast cancer patients[38], there have been continued efforts to develop methods
compatible with the isolation of viable CTCs for cell line establishment and the generation
of renewable sources of CTCs [129]. Simultaneously, there are also parallel efforts
generating biophysical profiles of real CTCs from different cancer types using

35
combinations of the detection, isolation, and mechanical phenotyping platforms
described. This will allow for the accurate determination of parameters necessary for in
vitro and in silico modeling of CTCs in circulation as well as the ability to deduce
appropriate conclusions from in vivo models.

36
CHAPTER 3
THE CELL CYTOSKELETON: THE CONTRIBUTION OF
KERATINS IN CANCER DIAGNOSIS, PROGNOSIS, AND
PROGRESSION

3.1 Cell cytoskeleton
The cell cytoskeleton is an integral part of a cell’s biophysical properties, which is
intimately linked to morphology, behavior, and resistance to various external stressors. It
is composed of three main components: actin microfilaments (~3-6 nm), intermediate
filaments (~10 nm), and microtubules (~25 nm). Whereas microtubules form hollow tube
filaments, actin microfilaments and intermediate filaments do not. These three
cytoskeletal components coordinate together to spatially organize intracellular organelles
and execute various biological processes, but also have unique individual functions as
well. Furthermore, each are differentially regulated, assembled, and dynamically
modulated. The cytoskeleton not only endows cells with their biophysical properties, but
is also involved in mechanosensing environmental cues and transducing signals to the
nucleus. The cytoskeletons of adjacent cells can also be interconnected to form a
supracellular scaffold that provides mechanical resilience to tissues. Therefore, structure
is intimately related to function.

37
3.2 Actin microfilaments: assembly and function
Unpolymerized, globular actin (G-actin) is an intrinsically polar ATPase that serves
as the basic building block for the elongated polymers of filamentous actin (F-actin). There
are three isoforms of this gene, with alpha actin mainly expressed in skeletal muscle cells,
and beta and gamma actin typically co-expressed in the majority of all other cell types.
De novo actin filament polymerization is nucleated by a wide range of protein
complexes that usually involve at least one component possessing a proline-rich domain
(PRD) and a tandem repeat of the Wiskott-Aldrich syndrome protein (WASP)-homology
2 (WH2) domain. At the site of nucleation, ATP bound actin monomers begin to elongate
into filaments by hydrolyzing ATP into ADP and inorganic phosphate (Pi). ATP-actin has
a higher affinity for the polymerized state than ADP-actin, which generates differential
assembly and disassembly kinetics between the two ends of the actin filament. In general,
elongation occurs by the addition of ATP-actin to the fast-growing barbed end and the
pointed ends of actin filaments can be disassembled into monomeric subunits by
nucleotide exchange [130]. This leads to a phenomenon called treadmilling, where it
appears that the filament is moving directionally due to growth at one end and shrinkage
at the opposite end. Even in a steady state where it appears that microfilaments are
neither elongating nor shrinking, they are still continuously undergoing cycles of
polymerization and depolymerization that highlights the dynamic and responsive nature
of the actin cytoskeleton [131].
Actin is involved in the generation of lamellipodia, which are sheet-like ruffles of
membrane protrusions, and filopodia, which are thin finger-like probing protrusions that
are transient. Invadopodia are also actin-rich structures similar to filopodia but differ in

38
various ways. Whereas filopodia contain the motor protein, myosin X, and are involved in
cell migration, adhesion, and environmental sensing, invadopodia contain ECM
degrading proteases and are involved in invasion [132].
Actin can also be found complexed with myosin II, which generates a cytoskeletal
structure that is inherently capable of contraction. When these actomyosin contractile
structures are aberrantly regulated, they have pathological consequences such as
nuclear fragmentation and genome instability [133], dysregulation of focal adhesions
maturation [134], and changes in cortical tension that affect the metastasis-promoting
architecture and behavior of cancer cells [135].

3.3 Microtubules: assembly and function
Unlike actin filaments, polymerization of tubulin generates rigid and hollow rod-like
structures. Tubulin, the basic subunit of microtubules, is a heterodimer of alpha-tubulin
and beta-tubulin. A third type of tubulin, gamma-tubulin, is involved in initiating
microtubule assembly at the centrosome of dividing cells. Tubulin subunits are nucleated
by various types of microtubule organizing centers (MTOCs) that can determine the
structure and orientation of growing microtubules. At these MTOCs, GTP-bound tubulin
molecules bind in a linear, head-to-tail orientation to form an elongated protofilament.
Thirteen protofilaments then bind in a lateral fashion to form the hollow microtubule.
Similar to actin microfilaments, microtubules are intrinsically polar structures and have
dynamic polymerization and depolymerization activities at both ends that results in
microtubule treadmilling – GTP-tubulin is continuously added to the growing, or plus, end,
whereas GDP-tubulin is removed from the minus end [136].

39
Microtubules are normally involved in cellular processes such as chromosome
segregation during metaphase and anaphase, kinesin- and dynein-mediated intracellular
transport, and organelle and vesicle positioning and localization [137]. However,
dysregulation of tubulin and changes in expression patterns of different tubulin isotypes
have been frequently observed and reported in cancer [138]. Microtubules can also affect
cellular stress responses by regulating p53 levels, a tumor suppressor protein involved in
cell cycle arrest and apoptosis, and respond to various stress conditions such as hypoxia
and oxidative stress [138]. Furthermore, microtubules have been implicated in a recently
identified tubulin- and vimentin-rich cell membrane protrusion called microtentacles,
which are structures involved in CTC extravasation and penetration through endothelial
cell barriers [103,139].

3.4 Intermediate filaments
Intermediate filaments (IFs) differ greatly from their microfilament and microtubule
counterparts. For one, they are intrinsically apolar proteins and do not require binding of
nucleotide cofactors for polymerization. IFs are also a comparatively large and diverse
group of proteins encompassing six families that are all related by a conserved central
rod domain. Type I and Type II IFs encompass a total of 54 different acidic and basic
keratin proteins, respectively. Type III IFs include desmin, peripherin, glial fibrillary acidic
protein (GFAP), and vimentin, which is the most extensively studied and widely expressed
protein in this group. Type IV IFs include alpha-internexin, neurofilaments, synemin, and
syncoilin. Type V IFs are comprised of nuclear lamins, and lastly, Type VI IFs include the
beaded filaments, filensin and phakinin, nestin, and other yet uncharacterized IF proteins.

40
In general, IFs span the entire cell cytoplasm from junctional structures like
desmosomes and hemidesmosomes along the plasma membrane, to mesh-like IF
networks throughout the cytoplasm, to anchor points on the outer nuclear membrane. To
date, vimentin, lamins, and keratins have been the most extensively studied in the context
of cancer.
Vimentin has been characterized in many different types of cancers. It is often
upregulated in epithelial cancers and used as a marker for cancer cells undergoing EMT
[140]. In general, vimentin expression is positively correlated with tumor growth, invasion,
metastatic potential, and poor prognosis, but there have been some conflicting studies
reported as well. Conflicting data may very well be due to context dependent differences
since vimentin has been shown to interact with at least 18 different proteins that modulate
various signaling pathways and cellular functions [141].
Lamins form dense mesh-like nucleoskeletal structures that line the inner nuclear
membrane and are responsible for the integrity and morphology of the nucleus. The
nucleoskeleton is connected to the cytoskeleton by Linker of Nucleoskeleton and
Cytoskeleton (LINC) complexes embedded in the nuclear membrane. This linkage is
involved in transducing external mechanical signals to the nucleus and bidirectional
signaling [142]. Lamins are additionally involved in chromosomal organization and play a
role in gene regulation. In the context of cancer, there is no clear direct relationship
between lamin expression and cancer progression. The only clear observation is that
lamin expression, function, and localization is aberrant in cancer cells and that these
irregularities can contribute to and/or be a product of cancer phenotypes [143].

41
Keratins, the largest family of IFs, were initially used as diagnostic biomarkers in
the cancer field due to their tissue- and cell-specific expression patterns. More recently,
however, there have been increasing reports of the functional contributions of keratins to
cancer progression and metastasis. To understand how keratins influence cell behavior,
they must each be examined on an individual and context-dependent basis.

3.5 Keratins: assembly and function
Of the ~70 types of IF proteins, keratins are the most numerous with the human
genome containing 54 functional keratin genes. These keratins are further categorized
into two families: type I acidic keratins, and type II basic keratins. Keratins form obligate
heterodimers between one molecule of a type I keratin and one molecule of a type II
keratin. Two heterodimers bind laterally to form the most basic soluble tetrameric subunit
of keratin IFs. Eight tetramers then form lateral associations to form a unit length filament
(ULF). ULFs come together longitudinally to form a short, elongated filament that then
experiences a radial compaction down to a final diameter of ~10 nm (Figure 3.1). These
short filaments can then integrate anywhere into the existing keratin filament (KF)
network, be exchanged with existing subunits, or be broken down into its smaller
counterparts to be recycled [144]. In normal tissues, keratins form IFs throughout the
cytoplasm and surround the nucleus, providing mechanical support and integrity to a cell.
In the more macroscopic cellular context, keratins are also known to form desmosomal
and hemidesmosomal junctions, providing anchorage support between adjacent cells and
from a cell to the underlying ECM (Figure 3.1).

42

Figure 3.1. Keratin assembly process and structural function. Keratin monomers self-assemble into
intermediate filament structures. Intermediate filaments span the entire cell from plasma membrane to the
nucleus and endow cells/tissues with mechanical resilience. Created with BioRender.com

Keratins in focal adhesions
Soluble keratins form a pool of readily available subunits near the cell periphery
where they begin to assemble near actin-rich focal adhesions and migrate along the actin
fibers until integration into the peripheral KF network. Disrupting the actin polymerization
disrupts the inward movement of keratin but has no effect on the formation, elongation
and fusion with the KF network [145]. Although the exact mechanism for nucleation is not
yet fully elucidated, it is evident that there is a co-dependence between different
cytoskeletal components.

43
Keratins in desmosomal junctions
Desmosomes are adhesive plaque-like junctions that form tight connections
between adjacent cells. Desmosomes contain three distinct regions – the proximal inner
dense plaque (IDP), the more distal outer dense plaque (ODP), and the extracellular
region (desmoglea) – and occur with mirror symmetry between two adjacent cells.
The IDP consists of the C-terminal region of desmoplakin, a protein that directly
binds keratin intermediate filaments and anchors them to the desmosomal plaques. It has
also been reported that desmoplakin recruited to nascent desmosomes potentially serve
as nucleation sites for KF elongation and eventually keratin filament bundling that
generates the basis for a desmosomal maturation and stable adhesion [146]. Located
between the IDP and the cell membrane, the ODP consists of the N-terminal domain of
desmoplakin, plakoglobin, plakophilin, and the intracellular domains of two types of
desmosomal cadherins called desmoglein and desmocollin. These calcium-dependent
proteins have a single pass transmembrane region with an intracellular binding domain
and five extracellular domains, which mediate heterophilic binding interactions in the
extracellular space and compose the desmoglea [147].

Keratins in hemidesmosomal junctions
Hemidesmosomes are also adhesive plaque-like junctions that form between cells
and the basement membrane, and they are located in close proximity to focal adhesions.
There are two types of hemidesmosomes: Type I, found in stratified and pseudostratified
epithelia, and Type II, found in simple epithelia. The core components of
hemidesmosomes include integrin α6β4, which binds to basement membrane proteins

44
like laminin, and plectin 1a (P1a), a cytoskeletal linker protein. Whereas type I
hemidesmosomal structures also contain tetraspanin CD151, bullous pemphigoid antigen
(BPAG) 1 isoform e, or BPAG1e (also known as BP230), and BPAG2 (also known as
BP180), type II HDs lack BPAG1e and BPAG2.
Hemidesmosomes also consist of two regions: the outer and inner dense plaques.
In the outer dense plaque, closest to the cell periphery, transmembrane protein
complexes (integrin α6β4, BPAG2, and CD151) connect the basement membrane to the
inner dense plaque via p1a and BPAG1e. In the inner dense plaque, these two proteins
help tether KFs to the hemidesmosomal structures [148]. It has also been reported that
recruitment of p1a and BPAG1e to nascent hemidesmosomes serve as keratin nucleation
and elongation sites [149].

Keratins attach to nuclear anchor points and a subset can have a nuclear role
Keratins span the entire cell from a soluble pool localized near the cell membrane
to a more organized filamentous network forming a perinuclear cage. Perinuclear KFs
impact nuclear shape and in some cases, can enter the nucleus to impact chromatin
organization and gene regulation [150,151]. KFs can anchor to the nucleus through
Nesprin-3, an outer nuclear membrane protein that is a component of the LINC complex.
Nesprin-3 binds plectin, which is a cytoskeletal crosslinking protein that can form direct
connections to keratin 6 and keratin 14 [152,153].
Interestingly, although keratins were previously thought to reside solely in the
cytoplasmic compartment whereas lamins were the nuclear counterparts, keratins 7, 8,
17 and 18 were identified in an antibody screen designed to assess the molecular

45
composition of the nuclear scaffold [154]. Among these four keratins, keratin 17 has been
the most well characterized and was found to affect nuclear morphology, chromatin
organization, gene regulation [150] and localization of the transcription factor,
autoimmune regulator (Aire) [155] in skin tumor keratinocytes.

3.6 Post-translational modifications (PTMs) regulate keratins
Keratins are expressed in a tissue specific manner and expression of different
types of keratins confers unique functions and properties to tissues. “Hard” keratins – for
example keratins 31-38 and keratins 81-68 –, are typically expressed in hair, skin, and
nails and their higher sulfur content and extent of disulfide bond formation contribute to
the rigidity and toughness of these tissues [156]. In addition to providing mechanical
resilience and strength to cells and tissues, they are also involved in: organelle positioning
within a cell, growth, migration, modulating signaling cascades, cellular viscoelasticity,
and physical properties of a cell. The large number of keratin proteins and the diversity of
PTMs lead to an immense number of possibilities for fine-tuned regulation in a cell type
specific and context dependent manner.
Keratins are regulated by various PTMs. Phosphorylation of serine and threonine
residues, which typically occur in the keratin head and tail domains, have large effects on
KF solubility, assembly, and organization, whereas tyrosine phosphorylation is reported
to affect keratin insolubility. Keratins containing the phosphoepitope LLS/TPL have also
been reported to function as a phosphate “sponge” or “sink” to help preserve the cell by
absorbing phosphorylation events away from pro-apoptotic proteins that lead to cell death
[157]. Phosphorylated keratin residues also provide binding sites for the adaptor protein

46
14-3-3, which is a family of proteins that bind phosphorylated serine- and threonine-rich
motifs to affect a range of cellular functions such as mitosis [158], cell growth via activation
of the Akt/mTOR pathway [159], and maturation and stability of cell adhesions [160].
However, the specific effects of phosphorylation on particular residues can have opposing
effects in different contexts. For example, phosphorylation of the keratin 8 Ser431 residue
triggered perinuclear keratin reorganization and increased migration of pancreatic and
gastric cancer cells [161] but dephosphorylation was linked with increased adhesion and
tumorigenicity in oral squamous cell carcinoma (OSCC) cell lines and correlated with size,
lymph node metastasis and stage in OSCC patients [162]. This highlights the variable
and complex context dependent effects of keratin regulation.
Phosphorylation is also linked with the sumoylation of keratins. Stress-induced
phosphorylation of keratins can regulate the attachment of small ubiquitin-related modifier
(SUMO) proteins at lysine residues located in the rod domain. Whereas moderate
sumoylation improves solubility, hypersumoylation causes keratins to precipitate and
aggregate in insoluble cellular compartments, removing them from functional contribution
to cellular processes [163].
Lysine acetylation is another type of modification that can impact how neighboring
residues are modified contributing to protein level regulation. For example, the highly
conserved Lys207 residue in the rod domain of keratin 8 was identified as a major
acetylation site that can also affect the phosphorylation state of other residues. In diabetic
mouse and human livers, high glucose levels triggered rapid acetylation that resulted in
decreased solubility [164]. Acetylation of Lys108 and the linked phosphorylated state of

47
Ser74 in keratin 8 has also been identified, but no further functional studies have been
reported [165].
Other less-well studied PTMs include methylation, glycosylation, and
transamidation. Methylation of arginine residues has also been reported to prevent rapid
degradation and turnover of keratins [165]. Glycosylation of serine residues in keratin 18
has been reported to promote the activation of Akt1- and PKC-mediated cell survival and
protection from epithelial injury [166]. It has also been reported that transamidation of
keratin lysine residues covalently links the KF network to the cornified cell envelope in the
upper layer of skin, generating mechanical resilience and contributing to barrier function
[167]. Transamidation between keratins also leads to covalent aggregation into inclusion
bodies [168].
Overall, there are many types of PTMs on a wide range of target residues in a
large panel of keratin proteins to explore. Furthermore, it seems that the same residues
can undergo different PTMs and even influence the modified state of neighboring
residues. The discovery of this overlap and crosstalk is a small but growing field that can
have a significant impact on our understanding of the function of keratin intermediate
filaments and their specific contributions to health and disease.

3.7 Keratins as diagnostic and prognostic markers of breast cancer
As aforementioned, type I and II keratins form obligate heterodimers and different
combinations of the 54 functional keratin genes are uniquely expressed in different tissue
types making them useful tools as diagnostic biomarkers for different types of cancers. In

48
breast cancer, keratins 8, 18, and 19 are used as markers for the luminal subtype [9]
whereas keratins 5, 6, 14, and/or 17 are used to identify the basal subtype [10].
Keratins have also been explored for use as prognostic indicators. In HER2+
breast carcinomas, keratin 5/6 positivity was significantly associated with decreased
overall and disease-free survival. Furthermore, keratin 5/6+ was also significantly
inversely correlated with time to treatment failure in HER2+ patients receiving
trastuzumab therapy [169]. In a separate study, keratin 5/6 and/or keratin 17 expression
was associated with poor clinical outcome in lymph node-negative breast cancers [170].
Detection of keratin 19 mRNA in the CTCs of breast cancer patients before adjuvant
chemotherapy was prognostic for reduced disease free survival and overall survival in
ER-, triple-negative, and HER2+ populations of breast cancer patients [171] and
associated with chemotherapy-resistant residual disease when detected in CTCs from
breast cancer patients after adjuvant chemotherapy regardless of subtype [172]. Keratin
10 expression in breast cancer was also reported to be correlated with worse outcome
[173]. Loss of keratin expression can be used as a prognostic marker as well. For
example, loss of keratin 18 expression was significantly correlated with advanced tumor
stage, high grade, and reduced overall survival [174].

3.8 The functional landscape of keratins in breast cancer
Beyond the diagnostic and prognostic value, keratins also play a functional role in
the behavior of cancer cells. Depletion of all keratins in mouse keratinocytes led to a more
invasive phenotype and enhanced 3D colony forming ability [175]. Keratin 14 is enriched
in the leader cells of collectively invading cluster of cells and knockdown abrogates the

49
invasive phenotype at the border between tumor and stroma but does not affect tumor
growth [176]. Keratin 18 physically localizes the ER coreceptor, LRP16, away from the
cell membrane and in the cytoplasm to decrease ER signaling [177]. The keratin head
domain of keratin 19 was found to physically interact with the C-terminal domain of HER2
to activate downstream signaling [178]. Keratin 19 also physically interacts with cyclin D3
to control cell cycle progression and knockdown increases the sensitivity of breast cancer
cells to cyclin-dependent kinase inhibitors [179]. Studies also show that while keratin 19
expression in breast cancer cells led to a decrease in cell migration and invasion, it was
associated with cell surface localization and E-cadherin and required for maintenance of
tight cell-cell adhesion [180]. In endocrine-resistant ER+ breast cancer cells, enhancer
activation by sterol regulatory element-binding protein 1 (SREBP1) leads to upregulation
of keratin 80, which promotes invasion and is associated with poor survival [181].
Upregulation of keratin 8 in a highly invasive breast cancer cell line decreased
proliferation, migration, and invasion, whereas downregulation of keratin 8 in a poorly
invasive breast cancer cell line increased these parameters [182].
In conclusion, keratins are a large and diverse class of intermediate filament
proteins that are involved in not only structural but also functional properties in both health
and disease. The range of different combinations of various post-translational
modifications introduces further complexity and requires further followup studies. In light
of all the evidence gathered to date, it is becoming increasingly clear that keratins
contribute to the biophysical properties favorable for metastasis and are involved in
molecular mechanisms underlying multiple stages of the metastatic cascade.

50
CHAPTER 4
Ectopic expression of a truncated isoform of hair keratin 81
(KRT81) in breast cancer alters biophysical characteristics to
promote metastatic propensity

Diane S. Kang
1,2
, Amal Thomas
3
, Jia Hao
4
, Aidan Moriarty
1,2
, Bret A. Unger
5
, Shamim Ahmmed
6
,
Yonatan Amzaleg
1,2
, Remi Klotz
1,2
, Siva Vanapalli
6
, Ke Xu
5
, Keyue Shen
4
, Andrew Smith
3
, and
Min Yu
1,2*

1
Department of Stem Cell Biology and Regenerative Medicine, Keck School of Medicine of the
University of Southern California, Los Angeles, CA, USA.
2
USC Norris Comprehensive Cancer Center, Keck School of Medicine of the University of
Southern California, Los Angeles, California, USA.
3
Department of Molecular and Computational Biology, USC David and Dana Dornsife College of
Letters, Arts and Sciences, University of Southern California, Los Angeles, California, USA.
4
Department of Biomedical Engineering, Viterbi School of Engineering, University of Southern
California, Los Angeles, CA 90089
5
Department of Chemistry, University of California at Berkeley, Berkeley, CA 94720
6
Department of Chemical Engineering, Texas Tech University, Lubbock, TX 79409

51
4.1 Introduction
Keratins were originally developed as diagnostic biomarkers for various different
cancer types. In breast cancer, luminal subtypes are identified by keratins 8, 18, and/or
19 positivity [9], whereas keratins 5, 6, 14 and/or 17 are used for basal subtypes [10].
Due to associations between various keratins and different clinicopathological features of
tumors as well as evidence for the prognostic value in patients, there has been a growing
interest in elucidating the mechanisms linking keratin expression to cancer progression
and metastasis.
Indeed there have been reports of the involvement of keratins in the invasive
phenotype and metastatic potential of CTCs and modulation of cancer-related signaling
pathways. Some keratins have also been newly attributed with a nuclear role as outlined
in Chapter 3. In our studies, we examined the metastatic potential of patient-derived CTCs
in an immune deficient mouse model and identified an upregulation of a truncated isoform
of KRT81 (tKRT81) in clinical samples, various breast cancer cell lines, and in vivo
selected lung-metastatic CTC-derivatives. We further defined the biophysical and
functional consequences of tKRT81 upregulation and how it contributes to metastatic
progression.

52
4.2 Results

KRT81 is upregulated in metastatic derivatives of CTCs compared to the isogenic,
parental CTC lines and is associated with lung-metastasis free survival

A previous study characterizing the metastatic tropism of ex vivo cultured CTC
lines (referred to as BRx07, BRx68, BRx50, and BRx42)[39] generated RNA sequencing
data for lung, brain, bone, and ovary metastases arising from the isogenic, patient-
matched parental CTC lines. In collaboration with Amal Thomas from Dr. Andrew Smith’s
lab at the University of Southern California, various bioinformatic and differential RNA
sequencing analyses were performed. KRT81, a type II hard keratin gene located on
chromosome 12, was identified as significantly upregulated in metastatic sites compared
to matched parental CTC lines (Table 4.1).
KRT81 upregulation in lung-metastatic CTC derivatives of BRx07 (referred to as
LuM1 and LuM2) compared to the parental, isogenic CTC line (BRx07) was confirmed by
qPCR using primers targeted towards the 3’-end of the gene (Figure 4.1A). Interestingly,
in a small cohort of 204 early stage breast cancer patients who were followed over time
for metastatic relapse (GSE12276), higher KRT81 expression levels in the primary tumor
were significantly associated with a shorter lung- and brain- (Figures 4.1B and C
respectively) but not bone-metastasis free survival (Figure 4.1D).

53
Comparison p-adjusted Fold change
Lung metastases vs. isogenic BRx07 CTCs 2.68 x 10
-4
25.8
Ovary metastases vs. isogenic BRx07 CTCs 5.09 x 10
-5
35.5
All metastatic sites vs. isogenic BRx07 CTCs 1.36 x 10
-5
30.7
All metastases vs. all CTC lines 4.68 x 10
-5
10.1
All lung metastases vs. all other metastatic sites 6.87 x 10
-3
16.0
Bone metastases vs. isogenic BRx68 CTCs 1.70 x 10
-4
27.9
Table 4.1. List of differential RNAseq comparisons with KRT81 upregulation. Several analyses identify
significant upregulation of KRT81 expression in different metastatic scenarios (FDR cutoff 0.1).

Figure 4.1. KRT81 expression in BRx07 and isogenic lung-metastatic derivatives and publicly available
dataset. (A) Quantitative PCR of KRT81 in two lung metastatic CTC derivatives (LuM1 and LuM2)
compared to the isogenic parental CTC line (BRx07). Kaplan-Meier survival curves of (B) lung- (C) brain-
and (D) bone-metastasis free survival in publicly available dataset, GSE12276. Significance was calculated
by log-rank test.

54
Identification of truncated KRT81 in breast cancer cell lines and patient samples
Alignment of the RNA sequencing reads to the genome showed that only the last
5 out of 9 exons were ectopically expressed in our lung-metastatic CTC derivatives
(LuM1, LuM2) and some commonly used breast cancer cell lines (MDA-MB-361, MCF7)
whereas full length KRT81 was normally expressed in skin cells (Figure 4.2A). A previous
study identified a methylation-responsive cryptic promoter containing two proximal Sp1
binding sites within the fourth intron of the KRT81 gene that was involved in generating
the truncated isoform [183]. While the full length mRNA encodes for an intact keratin
including the N-terminal keratin head domain, the conserved central rod domain, and the
C-terminal tail domain, the truncated isoform expressed by exons 5-9 encodes for only
the C-terminal half of the rod and tail domains (Figure 4.2B). Expression of tKRT81 was
confirmed by western blot showing a ~27 kDa band corresponding to the smaller protein
product translated from the truncated transcript (Figure 4.2C).
To determine the relevance of tKRT81 expression to a larger patient dataset, RNA
sequencing data available from breast cancer patients in TCGA (TCGA-BRCA) were
analyzed and 531 patient samples that express tKRT81 were identified. These libraries
were stratified into high, medium, and low quantiles based on the gene coverage plots
(Figure 4.2D). Interestingly, although nearly half of all the libraries were of luminal
subtype, the basal subtype was significantly over-represented in the high tKRT81
expression group (p<2.2 x 10
-16
based on chi-square test) (Figure 4.2E) and high tKRT81
expression was significantly correlated with decreased survival probability (Figure 4.2F).
In addition to breast cancer, KRT81 has also been described in various other
cancer types including non-Hodgkin’s lymphoma[184], gastric cancer[185],

55
melanoma[186] and pancreatic ductal adenocarcinoma (PDAC) [187]. In PDAC, KRT81
is currently being investigated as a marker for an aggressive subtype [187]. However,
many of the primers, probes, and antibodies used in these studies are unable to
differentiate between the full length and truncated isoforms. We therefore checked the
RNA sequencing read alignments to the KRT81 gene in the TCGA pancreatic
adenocarcinoma (TCGA-PAAD) dataset and found evidence of tKRT81 expression in at
least 8 samples, indicating the need for further clarification (Figure 4.2G). In our studies
in breast cancer, we found that a truncated isoform of KRT81 expressing only the last five
3’-exons is specifically expressed.

56

Figure 4.2. Identification of a truncated isoform of KRT81 in cell lines and TCGA dataset. (A) RNA
sequencing reads from BRx07 CTC line (in black), isogenic lung-metastatic BRx07 CTC derivatives (in
red), and publicly available data from commonly used breast cancer cell lines (in blue) aligned to the KRT81
gene showing that only the last five out of nine exons are expressed. In a context where KRT81 is normally
expressed, such as skin (in green) all exons are expressed. (B) A schematic depicting the protein domains
encoded by a full length KRT81 mRNA transcript compared to that of truncated KRT81. (C) Western blot
shows a band corresponding to KRT81 at ~27 kDa, corresponding to the size of a protein product translated
from a truncated KRT81 RNA transcript. HSP90 was used as loading control. (D) Sample gene coverage
plot from a high tKRT81 expressing TCGA-BRCA sample (red) compared to a low tKRT81 expressing
patient library (blue). (E) Binning of stratified TCGA-BRCA patient samples based on PAM50 subtype

57
shows that although nearly half of the samples were from the luminal A subtype, high tKRT81 expressing
patient libraries were significantly overrepresented by the basal subtype (significance calculated by chi-
square test). (F) High tKRT81 expression in TCGA-BRCA dataset showed a statistically significant inverse
correlation with overall survival (significance calculated by log-rank test). (G) Examination of KRT81 gene
coverage in TCGA-PAAD datasets shows evidence of truncated KRT81 expression in a small number of
samples.

Truncated KRT81 is associated with keratin filament disorganization and physically
interacts with KRT18
Whereas keratin assembly normally occurs in the presence of full length keratins,
expression of tKRT81, a type II keratin, was hypothesized to interact with type I keratins
at some point along the assembly process to disrupt the structure and function of keratin
intermediate filaments (Figure 4.3).

Figure 4.3. Graphical model of the disruption of the normal keratin filament assembly process by the
expression of tKRT81.

58
To examine the structural features of the keratin filament network, we conducted
super resolution imaging in collaboration with Bret Ungar from Dr. Ke Xu’s lab at
University of California Berkeley. MDA-MB-361 cells, which express a relatively high level
of tKRT81, were used to generate tKRT81 knockdown cell lines via CRISPR-mediated
silencing with sgRNA targeting the cryptic promoter of tKRT81 (sg10) (Figure 4.13A). In
control cells that express tKRT81, super resolution imaging of cells immunostained for
pan-cytokeratin shows more diffuse, thinner, and disorganized cytoplasmic keratin
intermediate filament (IF) structure compared to tKRT81 knockdown cells (Figure 4.4A).
In addition to a more diffuse cytoplasmic keratin filament network, tKRT81-expressing
control cells showed decreased perinuclear keratin organization compared to the tKRT81
knockdown cells (Figure 4.4B).

Figure 4.4. Microscopic examination of intermediate filament structures with regard to tKRT81 expression.
(A) Super resolution microscopy images of the keratin filament network (green) in tKRT81-expressing MDA-
MB-361 cells (control) compared to tKRT81 knockdown (knockdown). (B) Perinuclear KRT18 network in
control versus knockdown cells. Yellow lines mark the outer edges of the nuclei, and white arrows indicate
where the nuclei are located. Images are full focus z-stack images taken with 100X oil immersion lens on
Keyence BZ-X810 microscope.

59
There were also observable differences in desmosomal structures between
tKRT81 expressing control and knockdown cells when imaged by transmission electron
microscopy. In the tKRT81-expressing control cells, the desmosomes appeared to be
more frequently electron-dense and formed thicker mirror image plaques across the
extracellular space of two adjacent cells. When tKRT81 was knocked down, we observed
that desmosomes formed thinner electron-dense plaques close to the cell membrane
(Figure 4.5A). These observations were quantified in a blinded study where randomized
images were quantified by measuring the width and depth (Figure 4.5B). Although there
was no significant difference between desmosome widths (Figure 4.5C), the depths of
desmosomes were significantly thinner in tKRT81 knockdown cells compared to the
controls (Figure 4.5D).

Figure 4.5. Transmission electron microscopy imaging of desmosomes. (A) Transmission electron
microscopy images of desmosomes from control and knockdown cells. (B) Desmosome sizes were
quantified in a blinded analysis where plaque width and depth were measured. (C) Desmosomal plaque
widths were similar between the control and knockdown conditions. (D) Desmosomal plaque depths were
significantly shorter in tKRT81 knockdown cells compared to the controls. Statistical significance was
calculated by unpaired t-test.

60
To determine tKRT81 binding factors, tKRT81-DDK was ectopically expressed in
MDA-MB-361 knockdown cells via lentiviral induction and used to immunoprecipitate
interacting proteins for mass spectrometry analysis, which identified KRT18 as an
interacting factor (Figure 4.6A). The physical interaction between tKRT81 and KRT18 was
further verified by colocalization in super resolution imaging. Microscopy showed tKRT81-
DDK signal localizing to filamentous KRT18 and forming brush-like hubs surrounding the
KRT18 filaments (Figure 4.6B). Interestingly, areas of the cell that displayed high tKRT81
expression showed diminished KRT18 filaments (white circle, Figure 4.6C), whereas
areas that showed low tKRT81 expression showed thicker, more intact KRT18 filaments
(yellow circle, Figure 4.6C).

61

Figure 4.6. KRT18 and tKRT81 physically interact. (A) Mass spectrometry identifies KRT18 protein in co-
immunoprecipitation experiment with tKRT81-DDK. (B) Super resolution imaging showing co-localization
of KRT18 (green) filaments and tKRT81-DDK (magenta) signal. (C) Areas of tKRT81 expression (magenta
signal in white circle) have diminished signal for KRT18 filaments whereas areas without tKRT81
expression (yellow circle) show thick KRT18 filaments.

62
Truncated KRT81 expression leads to subtle changes in global transcription and
is not linked to YAP/TAZ, cGas-STING, or ERα signaling pathways in breast cancer
cells
To try to understand what signaling pathways may be affected by tKRT81
expression, RNA sequencing of tKRT81 expressing control cell line, MDA-MB-361, and
the corresponding isogenic knockdown and rescue cell lines was performed. Although
differential genes were identified, the changes were subtle and there were no large effects
on any single gene or pathway. Despite these subtle effects, however, comparisons
between the differential pathways in the control versus knockdown comparison (Figure
4.7A, Comparison 1) and the rescue versus knockdown comparison (Figure 4.7A,
Comparison 2) identified some pathways in common. In conditions where tKRT81 is
expressed, there is a suppression of kinetochore metaphase signaling, and suppression
of the mitotic roles of polo-like kinases which control the G2/M transition, and an activation
of the G2/M DNA damage checkpoint compared to the knockdown condition (Figure
4.7A).
Based on observed phenotypes and previously published literature, the effect of
tKRT81 expression on the activation of various signaling pathways was evaluated by
qPCR analysis of downstream genes in knockdown and overexpression cell lines
generated for MDA-MB-361, MCF7, and BRx07 (Figure 4.13A, B, and F respectively). To
test if the disruption in KF structure leads to differences in the YAP/TAZ pathway, which
is involved in sensing the cell’s physical nature, transcript levels of YAP/TAZ downstream
targets CTGF and CYR61 were measured by qPCR but there were no significant
correlations (Figure 4.7B). To determine if the decreased perinuclear support translates

63
to increased nuclear instability and changes in the cGas-STING pathway, transcript levels
of IFNB and IL6 were measured by qPCR but there were no significant correlations
(Figure 4.7C). Prior studies have reported that KRT18 can modulate estrogen receptor α
(ERα) signaling by sequestering an estrogen receptor coactivator, LRP16, in the
cytoplasm[177]. To determine if tKRT81 colocalization of KRT18 affects ERα signaling,
transcript levels of GREB1, PGR, and PS2 were measured by qPCR but there were no
significant correlations (Figure 4.7D).

64

Figure 4.7. Examination of the impact of tKRT81 on the transcription and activation of various signaling
pathways. (A) Qiagen’s Ingenuity Pathways Analysis (IPA) software was used to perform a comparison
analysis to identify changes in biological states related to tKRT81 expression in MDA-MB-361 cells.
Comparison 1 represents the core analysis performed to identify differentially regulated genes between the
tKRT81-expressing control and knockdown whereas comparison 2 represents the core analysis between
the tKRT81 rescue and knockdown (n=2 per group). Comparing the two core analyses to each other, eight
pathways related to tKRT81 expression and cell cycle and metabolic changes were identified. Signaling
pathway activation with regard to tKRT81 expression was assessed by qPCR for the (B) YAP/TAZ pathway
(measured by CTGF and CYR61), (C) cGas-STING pathway (measured by IFNB and IL6), and (D) ERα
signaling pathway (measured by GREB1, PGR, and PS2) with graphs representing two biological
replicates.

65
Truncated KRT81 influences cell morphological and biophysical changes
We next evaluated the potential biophysical changes induced by tKRT81
expression. To examine cell stiffness and cortical tension, a high throughput microfluidic
pipette aspiration (MPA) assay was performed in collaboration with Shamim Ahmmed
from Dr. Siva Vanapalli’s lab at Texas Tech University. In this experiment, a PDMS chip
device was fabricated to hold 1440 channels distributed over 16 rows. Each of the
channels is designed to trap a single cell in order to produce stiffness measurement. Cells
are flowed onto the chip and a vacuum is applied in order to trap the cell in the channel.
The applied negative pressure of the vacuum draws a portion of the cell membrane into
a fixed size channel, where a fluorescently labeled cell (Calcein-AM) can be visualized in
order to measure the radius of the cell (R) and the length of the cell deformation into the
channel (L) (Figure 4.8A). These measurements are used to calculate the Young’s
Modulus of each cell within a channel. Lung-metastatic CTC derivatives (LuM1) that
express tKRT81 were used to generate isogenic knockdown cell lines using shRNA
against tKRT81 (Figure 4.8D). Knocking down tKRT81 (shtKRT81) led to statistically
significant increase in intrinsic cell stiffness (Figure 4.8B) compared to the tKRT81-
expressing controls (shSCR), which corroborates the microscopy data that showed that
tKRT81 knockdown in MDA-MB-361 cells resulted in restoration of a more structurally
intact IF cytoskeleton and perinuclear cage (Figure 4.4A and B).
In addition to increases in cell stiffness, tKRT81 knockdown in MDA-MB-361 cells
led to significant decreases in median cell size, as measured by area of F-actin signal
(Figure 4.8C) and median nuclear size, measured by area of DAPI signal (Figure 4.8D).
MCF7 breast cancer cells overexpressing tKRT81 also showed a more strongly adherent

66
morphology compared to empty vector (EV) controls (Figure 4.8E). This difference in
morphology trended towards statistical significance in a blinded study where three
individuals were asked to quantify the number of “strongly adherent” and “weakly
adherent” cells in randomized images (Figure 4.8F).

Figure 4.8. Biophysical and morphological impact of tKRT81 expression. (A) Schematic of the high-
throughput microfluidic aspiration (MPA) device developed by Shamim Ahmmed in Dr. Siva Vanapalli’s lab
at Texas Tech University. Cells are labeled with live cell dye Calcein AM and then applied to the device,
which can assay 1440 cells simultaneously. For each cell inside a channel, a computer script is used to
take unbiased measurements of the aspiration length (L) and cell radius (R) to calculate the different cell
stiffness parameters. Knocking down tKRT81 expression (shtKRT81) in LuM1 cells significantly increased
intrinsic (B) cell stiffness compared to tKRT81-expressing control cells (shSCR). Significance was
calculated by Mann-Whitney test and ****: p-value<0.0001. Using the Keyence BZ-X800 Analyzer Software,
MDA-MB-361 tKRT81-expressing control (NT) and knockdown (sg10) cells were examined for (C) cell size
by F-actin staining, and (D) nuclear size by DAPI staining. Significance was calculated by unpaired t-test
and ****: p-value<0.0001. (E) MCF7 cell lines were adhered to collagen coated plates and stained by crystal
violet to examine morphology. Cells that do not express tKRT81 (MCF7 EV) showed more instances of a
“weakly adherent” phenotype (red arrows) compared to cells overexpressing tKRT81 (MCF7 tKRT81). (F)
The qualitative morphological differences were quantified in a blinded study where three people were each
given five randomly assigned images from each condition and asked to quantify the number of cells that
looked strongly versus weakly adhered. Significance was calculated by unpaired t-test.

67
Truncated KRT81 enhances the ability of a single cell to form a mammosphere and
adhesion to collagen, but does not affect migration and invasion
Mammospheres are aggregates of mammary epithelial stem cells and the
mammosphere formation assay is typically used to quantify the potential stemness
property of cancer cells[188]. To examine the ability for a single cell to give rise to a
mammosphere, we modified this assay by FACs sorting single cells into 96-well ultra-low
adhesion (ULA) plates and counting the number of cells in each well over 4 weeks.
Doxycycline inducible tKRT81 shRNA knockdown (sh1, sh2) in lung-metastatic CTC
derivatives (LuM1) compared to controls (shSCR) showed statistically significant
decrease in mammosphere forming ability when cultured in doxycycline containing media
(Figure 4.9A).
To examine the effect of tKRT81 in cell adhesion, we performed cell adhesion
assays on collagen coated plates. Knockdown of tKRT81 in MDA-MB-361 cells (sg10)
significantly reduced adhesion compared to the control (NT) (Figure 4.9B), whereas in
MCF7 cells, overexpressing tKRT81 increased adhesion compared to the control (EV)
(Figure 4.9C). Similarly, we tested adhesion to various substrates under shear stress
conditions using a microfluidic chip attached to a dual-channel syringe pump in
collaboration with Jia Hao from Dr. Keyue Shen’s lab at the University of Southern
California (Figure 4.9D). The microfluidic chip is coated on the surface with various
substrates and cells are added to the channels and allowed to adhere. The number of
cells that have adhered is quantified as “before shear flow”. Increasing amounts of shear
flow is then applied (ramping up from 0 to 30 mL/min, with 10s holding of each flow rates
in a stepwise fashion) [92] and the number of cells left attached to the surface is quantified

68
at each rate of flow and quantified as “after shear flow”. Adhesion to collagen under shear
stress conditions was statistically significantly greater in tKRT81 expressing LuM1 control
cells (shSCR) compared to isogenic tKRT81 knockdown cells (sh2), and this phenotype
could be rescued by tKRT81 overexpression (tKRT81) in the same cell lines (Figure
4.9E). In addition to adhesion phenotypes, keratins were also implicated in migratory and
invasive cell behaviors[189]. However, there was no difference in in vitro transwell
migration (Figure 4.9F), in vitro transwell invasion (Figure 4.9G), or in vivo lung
transendothelial migration (Figure 4.9H) with respect to tKRT81 expression.
Since cytoplasmic keratins have been reported to localize proteins to either the
cell surface or cytoplasm [177,190,180], we tested whether integrin β1 localization was
affected by tKRT81 expression in order to determine if this was the mechanism behind
the observed adhesion phenotype. However, flow cytometry analysis in LuM1 control,
tKRT81 knockdown, and rescue cells showed no difference of cell surface expression of
integrin β1 (Figure 4.9I).

69

Figure 4.9. Functional consequences of tKRT81 expression. (A) LuM1 cells expressing two different short
hairpins (sh1 and sh2) targeting tKRT81 and controls (shSCR) were single cell sorted into 96-well ultra low
adherence plates for single cell mammosphere assay. Plots are showing the area under the curve (AUC)
of each single cell growth over time. Red bar shows the median value. Statistical significance is calculated
among the three samples by Kruskal-Wallis test (p-value=0.0007). Significance is also calculated by Mann-
Whitney test between pairwise comparisons with ***: p=0.005 and ****:p<0.0001. (B) In vitro cell adhesion
assay with MDA-MB-361 tKRT81-expressing control (NT) and knockdown (sg10) cells. Significant by
unpaired t-test (**:p=0.0039) (C) In vitro cell adhesion assay with MCF7 control (EV) and tKRT81
overexpression (tKRT81) cells. Significant by unpaired t-test (**:p=0.0089) (D) Schematic diagram of dual-
channel microfluidic device to assay cell adhesion under shear stress conditions. (E) Microscopic images
of adhesion channels in microfluidic device depicting the number of cells adhered to a collagen coated
channel before and after application of shear force at a flow rate of 20 mL/min. LuM1 control (shSCR) and
tKRT81 overexpression cell lines showed significantly increased adhesion over a range of increasing flow
rates compared to tKRT81 knockdown cells (sh2). (F) In vitro transwell migration assays in MCF7 and
MDA-MB-361 cell lines after manipulation of tKRT81 expression levels. Not significant by unpaired t-test.
(G) In vitro transwell invasion assays using matrigel-coated inserts in MCF7 and MDA-MB-361 cell lines
after manipulation of tKRT81 expression levels. Not significant by unpaired t-test. (H) Plot showing
luciferase activity in in vivo lung transendothelial migration assay using MDA-MB-231 control (EV) and
tKRT81 overexpression cell line. Not significant by unpaired t-test. (I) There was no detectable difference
in cell surface localization and expression of APC-integrin β1 when analyzed by flow cytometry (both
median and mean APC signal) in both LuM1 and MDA-MB-361 tKRT81-expressing control (shSCR),
knockdown (shtKRT81), and overexpression (tKRT81) cells.

70
Expression of truncated KRT81 enhances in vivo lung metastasis
We then evaluated the effect of tKRT81 expression on tumorigenesis and
metastasis in immunodeficient NSG mice. Primary tumors generated by mammary fat
pad orthotopic injection of 2.5x10
5
LuM1 cells with tKRT81 expression control (shSCR)
or knocked down (shtKRT81) showed no significant difference in size and growth over
the 26-week period (Figure 4.10A). To test for lung recolonization ability, the same cells
were injected by lateral tail vein and mice were monitored by bioluminescent imaging for
a period of 32 weeks. At experimental endpoint, although there was no statistical
significance in whole body bioluminescent signal (Figure 4.10B), there was a significant
difference in lung tumor burden when imaged ex vivo (Figure 4.10C). Similar lateral tail
vein injections with the MDA-MB-361 tKRT81-expressing control (NT), tKRT81
knockdown (sg10), and rescue (tKRT81) cell lines showed a significant difference
between the rescue and knockdown groups both by whole body imaging (Figure 4.10D)
and in ex vivo measured lung tumor burden (Figure 4.10E). Although we did not detect
any differences between the tKRT81-expressing control and tKRT81 knockdown groups,
this may be due to differences in total protein expression levels. The rescue cell line has
a 2-fold higher abundance of tKRT81 than the endogenous levels present in the control
cell line (Figure 4.10F) and therefore an in vivo phenotype may become more evident if
the experiment was carried out for a longer period of time.

71

Figure 4.10. Expression of tKRT81 enhances in vivo lung metastasis. (A) LuM1 lung-metastatic CTC
derivatives were injected into the mammary fat pad and tumor growth was observed by bioluminescent
imaging over a period of 26-weeks. There was no difference in tumor growth at the orthotopic site between
the control (shSCR) and knockdown (shtKRT81) groups. LuM1 cells were injected into mice by lateral tail
vein and followed for 32 weeks to examine lung metastatic ability. (B) Whole body bioluminescent imaging
shows no significant difference at experimental endpoint, but (C) ex vivo lung imaging showed a significant
difference in lung tumor burden between the shtKRT81 knockdown and shSCR control groups. MDA-MB-
361 cells were injected into mice by lateral tail vein and followed for 11-weeks to examine lung metastatic
ability. (D) There was a significant difference in whole body bioluminescent signal between the tKRT81
knockdown (sg10) and rescue (tKRT81) groups at experimental endpoint, and (E) ex vivo lung imaging also
showed a significant difference in lung tumor burden between tKRT81 knockdown (sg10) and rescue
(tKRT81) group. Statistical significance was calculated by one-tailed Mann-Whitney test with *:p-value<0.05
and **:p-value<0.005. (F) Western blot of MDA-MB-361 tKRT81 expressing control (NT) versus tKRT81
rescue shows 2-fold higher tKRT81 protein levels in the rescue cell line compared to the control.

72
Although we detected an in vivo lung phenotype in the aforementioned cell lines,
there was no correlation between tKRT81 expression and lung tumor burden in mice
cohorts injected with MCF7 and T47D control and overexpression cells (Figures 4.11A
and B, respectively). Combined with the in vitro data, in which tKRT81 affected cell
adhesion but not migration and invasion, we hypothesized that the observed in vivo lung
phenotype may be due to cell adhesion and survival after transendothelial migration
rather than a difference in migration or invasion properties. LuM1 and MDA-MB-361 cell
lines were injected by lateral tail vein into NSG mice and lungs were harvested 2-days
later for fixation, sectioning, and immunofluorescent analysis of the proportion of apoptotic
tumor cells measured by the ratio of cleaved caspase-3 to GFP signal. There was no
significant correlation between tKRT81-expression and apoptosis after transendothelial
migration in both cell lines (Figures 4.11C and D, respectively).

Figure 4.11. Expression of tKRT81 does not enhance in vivo lung metastasis in some cell lines. (A) Ex vivo
lung tumor burden is not significantly difference between MCF7 EV and tKRT81 overexpression groups 8-
weeks post tail vein injection. (B) Ex vivo lung tumor burden is not significantly different between T47D EV
and tKRT81 overexpression groups 8 weeks post tail vein injection. There was no difference in apoptosis
after transendothelial migration between (C) LuM1 control (shSCR) and tKRT81 knockdown (shtKRT81)
groups, and (D) MDA-MB-361 rescue (tKRT81) and knockdown (shtKRT81) groups.

73
4.3 Discussion
In this study using immune deficient mouse models, we found that tKRT81 was
upregulated in the in vivo-selected lung metastases generated from patient-derived CTC
lines. Upregulation in the primary tumor was significantly correlated with lung-metastasis
free survival in a small cohort of breast cancer patients and overall survival in TCGA-
BRCA datasets. Expression of tKRT81 was associated with the disruption of the keratin
filament network in the cytoplasm and around the nucleus and with more electron-dense
desmosomes. Furthermore, tKRT81 was discovered to physically interact with and disrupt
the filamentous organization of KRT18 in breast cancer cells. The cytoskeletal changes
associated with the expression of tKRT81 resulted in larger cells with larger nuclei that
were also softer and more deformable. Knocking down tKRT81 diminished the ability of
single cell lung-metastatic CTC derivatives to form mammospheres in vitro and
decreased cell adhesion while having no effect on migration or invasion. In vivo
experiments also show increased lung metastasis due to tKRT81 expression, which can
be abrogated by knocking down tKRT81 levels. These data point to a working model in
which upregulation of tKRT81 leads to alterations in the cell cytoskeleton that increase
cell deformability, cluster formation, and adhesion, leading to enhanced lung metastasis
(Figure 4.12).

74

Figure 4.12. Graphical model of the impact of tKRT81 expression on enhanced metastatic propensity.

In the literature, KRT81 appears in numerous different cancer studies. A single
nucleotide polymorphism in the 3’UTR of the KRT81 gene is associated with risk in gastric
cancer [185], prognosis in non-Hodgkin’s lymphoma [184], recurrence in non-small-cell
lung cancer [191], and survival in multiple myeloma [192]. KRT81 is also currently being
developed as a marker for quasi-mesenchymal pancreatic ductal adenocarcinoma, the
most aggressive subtype [187], but whether the full length or truncated isoform is
expressed should be further investigated. KRT81 was identified in several different
studies on breast cancer as well. It was identified, but not further explored, in a study that
generated mouse models of lung metastasis using the MDA-MB-231 breast cancer cell
line to develop a 54-gene panel comprising a breast cancer lung-metastasis gene

75
signature [193]. Another group looking at genes involved in GATA3-mediated lung
metastasis found that GATA3 can repress KRT81 among several other genes involved in
breast-to-lung metastasis [194]. Gene expression profiling between pure breast ductal
carcinoma in situ (DCIS) cells with DCIS associated with synchronous invasive breast
cancer identified KRT81 upregulation as one of nine genes associated with those patients
with synchronous invasive breast cancer [195]. In the most recent study, one group found
that both full length and truncated KRT81 could be detected in both normal and breast
cancer cell lines, and that knockdown resulted in diminished in vitro migration and
invasion of the MDA-MB-231 cell line [196]. Another study performed single cell RNA
sequencing on the “normal” adjacent tissue of a patient diagnosed with DCIS using the
10X Genomics platform. Sequencing analysis identified KRT81 expression in one of three
clusters of epithelial cells. The identity of this KRT81-expressing cluster was ambiguous,
however, due to the expression of both luminal and basal epithelial keratins as well as
having an overall signature most similar to triple negative breast tumors. Moreover, it
cannot be distinguished whether expression was of the full length or truncated KRT81
due to the inherent 3’bias of the 10X Genomics platform [197].
Our data both aligns with and conflicts with the existing literature. In our studies,
only the truncated isoform of KRT81 was detected in breast cancer cell lines. Additionally,
tKRT81 only affected the biophysical characteristics and adhesion phenotype of our
breast cancer CTCs and cell lines. Although tKRT81 was not correlated with a migratory
or invasive phenotype in our study, tKRT81 upregulation does affect the keratin filament
structure and organization and increases cell deformability, which have all been reported
to be biophysical characteristics that are permissive for CTC adhesion, extravasation, and

76
metastasis [112,113,198]. We also uncovered a direct, physical interaction of tKRT81
with KRT18, which has never been reported before. Furthermore, although the CTC lines
initially used to identify tKRT81 upregulation were from luminal breast cancer patients,
there exists evidence pointing to a stronger association with the basal or triple negative
phenotype. A caveat of our study, however, is that the exact protein residues expressed
endogenously have not yet been mapped. One prior study that initially identified
upregulation of the truncated transcript reported that they were able to in vitro translate
the mRNA [199], but no one has yet identified the specific residues expressed. There are
two methionine residues close to the 5’-end of the mRNA that are both capable of being
the starting methionine and are separated by the following amino acid sequence:
DCIIAEIKAQYDDIVTRSRAEAESWYRSKCEE. This sequence amounts to a predicted
3.78 kDa difference and contains six residues capable of being phosphorylated that could
potentially have dramatic effects on the function and turnover of the resultant protein. In
our hands, expression of tKRT81 starting from the first methionine resulted in an
upregulation at the transcript level, but no expression at the protein level, whereas
expression of tKRT81 starting from the second methionine was able to generate both
mRNA and protein.
All these published and our own data suggest an important role for tKRT81 in
breast cancer metastasis. Of note, there have been many correlational studies implicating
KRT81 in breast cancer and other types of aggressive cancers. However, the exact
function of KRT81 and its truncated form, tKRT81, has not yet been investigated in cancer
metastasis. In fact, many of the probes and primers used in the literature are unable to
distinguish between the full length and truncated forms, contributing to the confusion. This

77
study contributes to the clarification of the role of truncated KRT81 in breast cancer
metastasis.

4.4 Methods

Cell culture
BRx07 CTCs and the corresponding isogenic, in vivo-selected lung-metastatic
derivatives, LuM1 and LuM2, were isolated and cultured as previously described [38,39].
Briefly, these cells were cultured in suspension on ultra low adhesion tissue culture plates
in 4% O2 and 5% CO2 in CTC media (RPMI 1640 medium, 20 ng/mL EGF, 20 ng/mL
bFGF, 1X B27, and 1X antibiotic/antimycotic). The adherent cell lines used in this study
were purchased from American Type Culture Collection (ATCC) and cultured in DMEM,
10% FBS, 1% Penicillin-Streptomycin (MDA-MB-231, MDA-MB-361, MCF7) or RPMI,
10% FBS, 1% Penicillin-Streptomycin (T47D). All cell lines were routinely tested for
mycoplasma contamination using a commercially available kit (MycoAlert, Lonza Group).

Construct generation
DDK-tagged truncated KRT81 construct was cloned from Lenti ORF clone of
Human keratin 81, myc-ddk-tagged, from Origene (Catalog# RC220726L3). This
construct was modified by PCR cloning to move the puromycin resistance gene upstream
of the P2A sequence followed by the truncated KRT81 sequence that is C-terminally
tagged with Myc and DDK before the stop codon.

78
The shRNA sequence targeting human KRT81 used was: (sh1) 5’-CCG GCA CTC
CTG GCC TCA CAT TTC TCT CGA GAG AAA TGT GAG GCC AGG AGT GTT TTT TG-
3’ and (sh2) 5’-CCG GAG CAA GTG CTC AGC TAC TTC TCT CGA GAG AAG TAG CTG
AGC ACT TGC TTT TTT TG-3’ (both targeting the 3’UTR) and was cloned into pLKO.1
Tet-on neo vector. Similarly, the shRNA for the non-targeting scramble (SCR) control
sequence used was: 5’-CCG GCC TAA GGT TAA GTC GCC CTC GCT CGA GCG AGG
GCG ACT TAA CCT TAG GTT TTT G-3’. Doxycycline was used at a final concentration
of 100 ng/mL for 48 hours to induce hairpin constructs.
The two part CRISPR inactivation lentiviral system was purchased from Addgene:
Lenti-dCas9-KRAB-Blast (Addgene, 89567) and pLKO5.sgRNA.EFS.tRFP657
(Addgene, 57824). The online Gene Perturbation Platform (GPP) sgRNA Designer from
the Broad Institute was used to design sgRNAs targeting intron 4 of the KRT81 gene (5’-
CAC CGA AAC CTG GCA GCC AGC AGA G) and cloned into the
pLKO5.sgRNA.EFS.tRFP657 vector as directed in the protocol associated with the
product. Cells were first transduced with the dCas9-Krab lentivirus, selected by blasticidin
concentrations optimized for the different cell lines, then subsequently transduced with
the sgRNA containing lentivirus and purified for tRFP657 positive populations by FACS.

79

Figure 4.13. Western blot validation of tKRT81 knockdown by CRISPRi (sg10) or shRNA system (sh1, sh2)
in various cell lines. Truncated KRT81 protein levels normalized to β-actin in (A) MDA-MB-361 tKRT81-
expressing control (NT), tKRT81 inactivation (sg10), and tKRT81 rescue (tKRT81), (B) MCF7 control (EV)
and tKRT81 overexpression (tKRT81) (C) T47D control (EV), and tKRT81 overexpression (tKRT81), (D)
LuM1 tKRT81-expressing control (shSCR) and shRNA knockdown (sh1, sh2), and (E) LuM1 tKRT81-
expressing control (shSCR) and tKRT81 overexpression (tKRT81), and (F) BRx07 control (EV) and tKRT81
overexpression (tKRT81).

Lentiviral production and generation of stable cell lines
Lentivirus was generated using the published protocol from The RNAi Consortium
(TRC) Broad Institute. In brief, low passage 293T cells were co-transfected with second-
generation lentiviral packaging vectors and the aforementioned expression constructs
using TransIT-LT1 transfection reagent (Mirus). Next day, cells were cultured in high
serum growth medium and collected at both 48 and 72 hours post-transfection. Viral
media was either used directly to transduce adherent cell lines, or concentrated with
Lenti-X concentrator (Clontech) and resuspended in PBS to remove serum for
transduction of CTCs and lung-metastatic CTC derivatives. All cells were transduced with
virus in the presence of 8 ug/mL polybrene and selected by 1-week antibiotic selection
based on experimentally derived kill curve assays, or by FACS.

80
Gene expression profiling by qRT-PCR
RNA was extracted using Quick RNA Microprep kit from Zymo. Between 200-1000
ng RNA was reverse transcribed using 5X iScript supermix (Bio-Rad). Ten ng of cDNA
was then used per real-time quantitative PCR reaction using iQ SYBR Green Supermix
(Bio-Rad) and run on a CFX iCycler real-time PCR machine (Bio-Rad). The primer
sequences used are listed in Table 4.2.

Gene Forward sequence (5’-3’) Reverse sequence (5’-3’)
36B4 GTGTTCGACAATGGCAGCAT AGACACTGGCAACATTGCGGA
TBP CCACTCACAGACTCTCACAAC CTGCGGTACAATCCCAGAACT
KRT81 CCAATTGAACACCACCTGCG CACACCCAGGGAGCTGATAC
CTGF CAGCATGGACGTTCGTCTG AACCACGGTTTGGTCCTTGG
CYR61 ACCGCTCTGAAGGGGATCT ACTGATGTTTACAGTTGGGCTG
IFNB ATGACCAACAAGTGTCTCCTCC GGAATCCAAGCAAGTTGTAGCTC
IL6 CCTCGACGGCATCTCAGCCCT TCTGCCAGTGCCTCTTTGCTGC
GREB1 ATGGGAAATTCTTACGCTGGAC CACTCGGCTACCACCTTCT
PGR CATTTGCACAAACCTGATGG CATGGTGTACAAGGCCACTG
PS2 TCGGGGTCGCCTTTGGAGCAG GAGGGCGTGACACCAGGAAAAC
Table 4.2. List of qPCR primers

Western blot
Cells were washed in PBS and lysed in Laemmli Buffer (50 mM Tris pH=6.8, 1.25%
SDS, 15% glycerol) and heated at 95C for 15 minutes. After protein quantification by

81
Lowry protein assay (Bio-Rad) samples were reduced with 5% (v/v) beta-
mercaptoethanol. Lysates were then mixed with bromophenol blue (0.01% v/v) and run
on a denaturing 4-15% Mini-PROTEAN TGX Precast gels (Bio-Rad). Gels were
transferred by semi-dry method to low fluorescence PVDF membranes in a Trans-Blot
Turbo Transfer System (Bio-Rad) using the standard Bio-Rad listed protocol. Membranes
were then blocked in 5% NFDM in TBS followed by primary antibody incubation in
blocking buffer with 0.2% Tween 20 at 4
o
C overnight. Membranes were then washed (3
x 10 minutes each, 1XTBST) and incubated with LICOR secondary antibodies diluted in
blocking buffer with 0.2% Tween 20 and 0.01% SDS for 1 hour at RT. Membranes were
then washed (3 x 10 minutes each, 1XTBST), rinsed in TBS, and imaged on a LI-COR
imaging instrument. The following antibodies were used: Pan-Keratin (C11) Mouse mAb
Cell Signaling 4545S; DYKDDDK Tag (D6W5B) Rabbit mAb Cell Signaling 14793S;
Keratin 18 (DC10) Mouse mAb Cell Signaling 4548; HSP90 antibody Abcam ab13492;
Anti-basic Hair Keratin K81 guinea pig polyclonal, serum Progen GP-hHb1; KRT81
polyclonal antibody ProteinTech 11342-1-AP; APC Anti-integrin beta 1 antibody [P5D2]
Abcam ab221241; b-Actin (mouse) Sigma A5441; IRDye® 800CW goat anti-rabbit IgG
secondary antibody LI-COR 925-32211; IRDye® 680RD goat anti-mouse IgG secondary
antibody LI-COR 926-68070; IRDye® 800CW donkey anti-guinea pig IgG secondary
antibody LI-COR 926-32411.

RNA-seq and Differentially Expressed Genes (DEG) analysis
RNA was isolated using Zymo RNA isolation kit. RNA quantity and quality were
measured by NanoDrop and TapeStation before the KAPA Stranded RNA-Seq Kit with

82
RiboErase (HMR) was used to generate the sequencing libraries. Libraries were
sequenced at either USC Translational Genomics Core or CHLA. Some libraries were
generated and sequenced by Novogene Corporation. CTC sequencing reads were
mapped to hg19 (GRCh37) reference using STAR v2.5.2b [200]. Genes annotated in the
ENSEMBL GRCh37.p13 GTF (release 75) were quantified using HTSeq-count [201].
Differential analysis was performed using DESeq2 [202]. Differential genes (DE) across
the target metastatic sites (FDR ≤ 0.05) were identified after controlling for the cell line,
dissociation, and culture effects. Coverage tracks were created using deepTools [203] for
visualization in the UCSC genome browser [204]. RNA-seq data of different breast cancer
cell lines were obtained from Daemen et al. 2013.

Clinical data analysis
GSE12276 dataset was analyzed in Partek Genomics Suite 6.6. Kaplan-Meier
survival curves were calculated on KRT81 expression split into two quantiles based on
the median value. Comparisons were considered statistically significant by log-rank p
value < 0.05.
TCGA-BRCA data were analyzed. All TCGA-BRCA data were downloaded and
RNA sequencing reads aligned to exons 1-9 of KRT81. Libraries expressed tKRT81 were
selected for survival analysis. Patients across different subtypes with high tKRT81
expression levels (67
th
percentile) were significantly associated with decreased overall
survival compared to those patients with low expression. The RNA-seq BAM files of
TCGA-BRCA patients were downloaded using GDC data transfer tool. For each tumor
library, reads falling into exons 1-9 were quantified and counts were normalized for the

83
library size. To identify expression levels of full-length KRT81 and tKRT81 transcripts, a
goodness of fit test is performed based on read counts in exons 1-4 and exons 5-9. The
coverage at base pair resolution across the gene body is calculated and the coverage
plots were also compared to identify tKRT81 expression. Only libraries with TPM
expression ≥5 were considered for the analysis. An FDR level of 0.1 is used for multiple
testing corrections. Gene body coverage for house-keeping genes was calculated using
RSeQC [205] and libraries that showed 3’ coverage bias were filtered out. Libraries were
stratified into high or low tKRT81 expression and clinical data of the patients were
obtained using TCGAbiolinks R package [206]. Survival analyses were performed using
the survival R package [207].

Immunocytochemistry
Cells were seeded on 18mm Coverglass No. 1 coverslips and cultured overnight.
After washing with PBS, cells were fixed with 4% paraformaldehyde in PBS for 10 minutes
at RT, washed in PBS, then permeabilized with 0.1% Triton X-100 in PBS for 10 minutes
at RT. Cells were then blocked for 1 hour at RT in 5% goat serum in PBS with 0.1%
Tween 20. Primary antibodies were diluted in blocking buffer and incubated with
coverslips overnight at 4
o
C. After 3 x 5 minute washes with PBST, secondary antibodies
were diluted in blocking buffer and incubated with coverslips for 1 hour at RT in the dark.
The coverslips were then washed 3 x 5 minutes in PBST, stained with a nuclear dye 4,6-
dianidino-2-phenylindole (DAPI) for 5 minutes, and mounted using ProLong Gold Antifade
Mounting media (Thermo Fisher) overnight in the dark. The next day, coverslips were

84
imaged using the Keyence BZ-X810 microscope. Cell parameters were quantified using
the Hybrid Cell Count and Macro Cell Count features of the BZ-X800 Analyzer Software.

Super resolution microscopy
Super resolution microscopy was performed in collaboration with Bret Ungar from
Dr. Ke Xu lab at University of California Berkeley. In brief, cells were seeded on 18mm
Coverglass No. 1 coverslips and cultured overnight. After washing with PBS, cells were
fixed with 4% paraformaldehyde in PBS for 10 minutes at RT, washed in PBS, and
permeabilized and blocked in blocking buffer (3% w/v BSA, 0.1% v/v Triton X-100 in PBS)
for 1 h. Afterward, the cells were incubated with primary antibodies (below) in the blocking
buffer for 12 h at 4 ̊C. After washing in a washing buffer (0.3% w/v BSA and 0.01% v/v
Triton X-100 in PBS) for three times, the cells were incubated with dye-labeled secondary
antibodies (below) for 1 h at room temperature. Then, the samples were washed 3 times
with the washing buffer and 3 times with PBS. Primary antibodies used: pan-cytokeratin
(Sigma C2562), DYKDDDDK Tag (D6W5B) (Cell Signaling Technologies 14793S), and
KRT18 (DC10) (Cell Signaling Technologies 4548S). Secondary antibodies used: Alexa
Fluor 647-labeled goat anti-mouse (Invitrogen A21240) and CF-568 conjugated to
AffiniPure donkey anti-rabbit IgG (H+L) (Jackson ImmunoResearch 711-005-152). 3D-
STORM super-resolution microscopy [208, 209] was carried out on a homebuilt setup
using a Nikon CFI Plan Apo λ 100x oil immersion objective (NA 1.45), as described
previously[210]. Briefly, the sample was mounted with an imaging buffer consisting of 5%
(w/v) glucose, 100 mM cysteamine, 0.8 mg/mL glucose oxidase, and 40 µg/mL catalase
in a Tris-HCl buffer (pH 7.5). For two-color imaging of DELE1 and TOM20, the two targets

85
were labeled by Alexa Fluor 647 and CF568, respectively, and were imaged sequentially
using 647- and 560-nm excitation lasers. These lasers were passed through an acousto-
optic tunable filter and illuminated a few micrometers into the sample at ~2 kW cm
-2
, thus
photoswitching most of the labeled dye molecules in the sample into the dark state while
allowing a small, random fraction of molecules to emit across the wield-field over different
camera frames. Single-molecule emission was passed through a cylindrical lens of focal
length 1 m to introduce astigmatism [209], and recorded with an Andor iXon Ultra 897
EM-CCD camera at a framerate of 110 Hz, for a total of ~50,000 frames per image. The
raw STORM data were analyzed according to previously described methods [208, 209].

Transmission Electron Microscopy
Cells were grown to confluence on collagen coated tissue culture plates then fixed
in TEM Buffer (2.5% glutaraldehyde, 2% paraformaldehyde, 7% w/v sucrose in 0.1M
HEPES) and submitted to USC’s Core Center of Excellence in Nano Imaging run by
senior scientist Dr. Carolyn Marks.

Immunoprecipitation and Mass Spectrometry
All steps were done on ice and all buffers were supplemented with 1X Roche
cOmplete
TM
EDTA-free Protease Inhibitor Cocktail prior to use. Roughly 10
6
-10
7
cells
were seeded in large 15 cm tissue culture plates. The next day, cells were washed in
PBS, scraped, and pelleted. Cell pellets were frozen at -80
o
C for at least 1 hour to help
in the lysis process. Thawed pellets were then lysed in 200 cold lysis buffer (10 mM Tris-
HCl pH=7.5, 150 mM NaCl, 0.5 mM EDTA, 0.5% Nonidet P40 Substitute) for 30 minutes

86
on ice with frequent pipetting. Insoluble debris were pelleted by centrifugation at 17,000g
for 10 minutes at 4
o
C and the resulting supernatant was then diluted to 500 uL in dilution
buffer (10 mM Tris-HCl pH=7.5, 150 mM NaCl, 0.5 mM EDTA). The diluted lysate was
precleared using 25 uL of mNeonGreen-Trap Magnetic Agarose beads (Chromotek). The
precleared lysate was then incubated with 25 uL of pre-equilibrated Myc-Trap Magnetic
Agarose beads (Chromotek) for 1 hour at 4
o
C. The beads were then washed 5 x 5 minutes
each in cold wash buffer (10 mM Tris-HCl pH=7.5, 150 mM NaCl, 0.5 mM EDTA, 0.05%
Nonidet P40 Substitute) and binding protein complexes were eluted in 30 uL of 2X SDS-
Sample buffer (120 mM Tris-HCl pH=6.8, 20% glycerol, 4% SDS, 0.04% bromophenol
blue, 10% beta-mercaptoethanol). Proteins were then run on 7.5% and 18% SDS-PAGE
gels to resolve high- and low-molecular weight proteins, respectively. Gels were then
silver stained by sequential incubation in the following buffers made with deionized and
distilled water: 50% methanol for 10 minutes, 5% methanol for 10 minutes, 32 micromolar
DTT solution, 0.1% AgNO3 solution for 10 minutes, two quick rinses with deionized water
followed by incubation with developing solution (0.02% paraformaldehyde, 3% Na2CO3)
until bands appear. The developing solution was then neutralized by empirically adding
citric acid powder. Differential bands were identified between the control cell line and cell
line expressing recombinant tKRT81-DDK-Myc and extracted from the gel and submitted
to the mass spectrometry core at USC’s School of Pharmacy run by director Dr. Alireza
Abdolvahabi. In brief, silver-stained IP gel pieces were cut and de-stained using a solution
containing 100 mM sodium thiosulfate and 30 mM potassium ferricyanide (1:1 ratio) for
30 min with gentle shaking. Gel pieces were then washed with 25 mM ammonium
bicarbonate in 50% acetonitrile (ACN), reduced with 5 mM dithiothreitol (DTT) for 30 min

87
at 60 °C, alkylated with 20 mM iodoacetamide (IAA) at dark for 30 min, and digested
overnight with trypsin Gold (Promega) at a final concentration of 6 ng/µl. Digestion was
then quenched by addition of 2% formic acid (FA). Peptides were extracted twice with
50% ACN/2% FA each time for 30 min with vigorous shaking. Extracted peptides were
evaporated to complete dryness and reconstituted in 5 µl of MALDI matrix solution (10
mg/ml dihydroxybenzoic acid in 70% ACN/0.1% FA). Half microliter of this solution was
spotted on a 384 Big Anchor MALDI target, let dry under ambient conditions, and
analyzed using a Rapiflex MALDI-TOF-TOF mass spectrometer working under Linear
Mode. Prior to running sample, the mass spectrometer was calibrated using a peptide
calibration solution containing bradykinin, angiotensin, substance P, bombesin, ACTH,
and somatostatin. The resulting peptides were searched against SwissProt library using
BioTools software (Bruker Daltonics) for protein identification. The parent mass tolerance
was set to 50 ppm. Protein identification was performed using peptide mass fingerprinting
(PMF) and validated with MS/MS.

Microfluidic pipette aspiration (MPA) assay
MPA devices were generated by Shamim Ahmmed at Texas Tech University in
Dr. Siva Vanapalli’s lab. Microfluidic devices for MPA were made using standard soft
lithography [211]. The design of MPA contains 1440 aspirator channels with each channel
having a cross-section of 5 mm ´ 5mm. 1500 cells in a 15 uL volume were aspirated using
a vacuum pump (Fluigent Inc.) at a negative pressure of DP = -600 Pa. Prior to loading
of cells in the MPA devices, cells were tagged with calcein AM to enable easy visualization
of the aspiration length. Images of trapped cells in the aspirator channels were obtained

88
using a Keyence BZ-9000 microscope. A custom written MATLAB (Mathworks Inc.)
routine was developed for processing images and quantifying the equilibrium aspiration
length L. The Young’s modulus E was calculated using the expression E = 3RDPf/2pL,
where R is the hydraulic radius of the aspirator channel (= 5 mm) and f is the wall function
with a typical value of 2.1 [212].

Single cell mammosphere assay
Cells were incubated with viability dye 7AAD and single cell sorted on a BD
FACSAria instrument into 96-well V-bottom ultra low adherence plates containing
appropriate media. Excluding the wells lining the edge of the plate, each well was counted
immediately after sorting to ensure that it received exactly one cell. The number of cells
in each wells were counted once a week for up to four weeks. Growth was charted for
each well and the sum of the area under the curve (AUC) were compared to determine
differences in the ability for a single cell to generate a mammosphere. Statistical
significance was calculated by Kruskal-Wallis test.

Cell adhesion assay and morphology assessment
Rat tail collagen I was diluted in 0.2N acetic acid to a final concentration of 100
ug/mL and used to coat wells for 1 hour at RT. The collagen was then gently removed
and washed 2x with PBS. The coated ECM was blocked in DMEM with 10% FBS and
incubated for 30 minutes at 37
o
C. During this time, cells were trypsinized for 3 minutes,
neutralized, and quickly counted and adjusted to an optimized final seeding density
(MDA-MB-361: 20,000 cells/well, MCF7: 30,000 cells/well) and adhered to the collagen

89
coated wells for 30 minutes at 37
o
C. Unadhered cells were then removed by plate
inversion and the wells were washed 2x with cold PBS containing 1 mM CaCl2 and 1 mM
MgCl2. Cells were then fixed with cold methanol for 10 minutes at RT, followed by three
washes with PBS. Crystal violet (0.5% wv crystal violet in 20% ethanol) was incubated on
the cells for 10 minutes at RT with gently shaking. Excess crystal violet was removed by
immersing the plates sequentially in 3 x 2L beakers of distilled water for 1 minute each.
Crystal violet was then recovered by adding 200 uL of 100% methanol to each well for 15
minutes at RT with gentle shaking. 100 uL of the recovered crystal violet solution was
transferred to flat-bottom 96-well plates and absorbance measured at 590 nm. Adhered
cells stained with crystal violet were also imaged at 40X with 5 random images taken per
condition. These images were then randomized and given to three blinded individuals
who qualitatively binned cells into two nondescript categories: strongly adhered, and
weakly adhered. The cell counts were quantified between the three individuals to obtain
a quantitative measurement of cell morphology.

Cell adhesion under shear stress conditions
This experiment was performed by Jia Hao from Dr. Keyue Shen’s lab at USC. To
prepare supported lipid bilayers and protein tethered surfaces, lipid components, 18:1
(Δ9-Cis) 1,2-Dioleoyl-sn-glycero-3-phosphocholine (DOPC) and 5% 18:1 1,2-dioleoyl-sn-
glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid)succinyl] (nickel salt) (DGS-
NTA(Ni)), dissolved in chloroform were purchased from Avanti Polar Lipids and mixed.
The lipids were air-dried in round-bottom flasks and desiccated for 2 hours with house
vacuum pump in a chemical fume hood. The lipid mixture was resuspended by bath

90
sonication in 1X PBS at a final concentration of 2.5 mg/ml and extruded 10 times through
a membrane with 50 nm pore size (Avanti Polar Lipids) into small unilamellar vesicles
(SUVs). The SUV solutions were then diluted 1:1 in 1X PBS (pH 7.4) before being loaded
onto the detergent cleaned and dried glass coverslip through the loading chamber, and
incubated for 2 min to spontaneously form the lipid bilayers. The chambers were then
washed with a 10X excess volume of 1X PBS.
For protein capturing on lipid bilayer, the substrate was blocked with 1% BSA for
1h and a solution of 10 µg/mL recombinant mouse ICAM-1 with poly-histidine tag(Cat.
50440-M08H, SinoBiological) was injected to supported lipid bilayer, incubated at RT for
40 min and tethered to 18:1 DGS-NTA(Ni) through chelation. Tethered SLB was washed
excessively with 1X PBS before use. For the immobilization of ICAM-1, 10 µg/mL
recombinant protein A (Cat.101100, Thermo Fisher) in 1X PBS was injected to detergent
cleaned and dried glass coverslip, incubated for 30 min, washed with 1X PBS, blocked
with 1% BSA for 1 h, before 10 ug/mL recombinant mouse ICAM-1 with Fc-tag (Cat. 796-
IC, R&D systems) was injected and incubated at RT for 40 min. The resulted substrate
was then washed with 1X PBS before use. For ECM protein coating on substrates,
cleaned coverslips were incubated with 1mg/mL collagen or 10 µg/mL fibronectin for 1 h
at RT, and rinsed with PBS before use.
We created the microfluidic device in-house using a micromilling platform, design
and fabrication protocols, and soft-lithography techniques for shear flow and adhesion
analysis [92]. Within each device, SLBs were formed in two geometrically identical
(mirrored), parallel microfluidic channels separated by a 250 μm barrier. The design and
toolpaths for the double channel microdevice (channel height 1 mm, channel width 2 mm,

91
length 16 mm) were created in Autodesk Fusion 360 (San Rafael, CA) and custom-milled
(Shapeoko, Carbide 3D, Torrance, CA) out of polycarbonate. The final device was
manufactured by pouring polydimethylsiloxane (PDMS) mixed at 10:1 base to curing
agent ratio (Sylgard 184 elastomer kit; Dow Corning). PDMS was cured at 80°C for 3 h,
peeled off, and cut into individual devices. Channel inlets and outlets with 0.75 mm
diameter were punched at both ends of microfluidic channels. The PDMS devices were
permanently bound to the detergent-cleaned glass coverslips after plasma treatment for
50 seconds (Harrick Plasma, Model PDC-001-HP) for the subsequent lipid bilayer
formation and substrate modification.
A dual-channel syringe pump (New Era Pump Systems, NY) was used to apply
controlled shear flow to the two channels through 10 mL glass syringes (inner diameter
14.57 mm) and tubing connections. Cells were labeled with Calcein AM (Cat. C1430,
Thermo Fisher) following vendor’s instructions, incubated with substrates for 1 hr under
hypoxia conditions (37°C, 4% CO2 and 5% O2), before infusing serum-free RPMI 1640
media at controlled flow rates (ramping up from 0 to 30 mL/min, with 10s holding of each
flow rates in a stepwise fashion) under a 37°C environment. The design enables real-time
imaging and direct comparison of two cell types on the same substrate under the same
flow rates. BF images were taken once every second using a 2x objective (CFI60 Plan
Apochromat Lambda Lens, NA 0.1, WD 8.5mm). The remaining cells under each flow
rate were normalized as a percentage by the starting cell numbers in the same regions
of interest (ROIs). Each ROI is a 500x500 μm square containing 20-80 cells randomly
selected along the center of the channel. Shear stress at the SLB surface (bottom of
channel) was calculated at https://www.elveflow.com/microfluidic-calculator/, where the

92
fluidic properties were assumed the same as water at 37°C considering the serum-free
nature of the RPMI 1640 media.

In vitro migration and invasion assay
Boyden chamber experiments were conducted using 8 micron transwell inserts. In
brief, cells were serum-starved for 24-hours prior to staining with a live-cell dye
(CellTracker™ Green) and seeding at a density of 5 x 10
4
cells in the upper chamber in
serum-free media. Chemoattractant (10% FBS) was added to the bottom chamber and
cells were allowed to migrate across the membrane for 18 hours. After brief fixation, cells
attached to the top of the insert were gently removed after fixation and the insert with cells
migrated to the bottom of the insert were mounted on slides, imaged, and counted using
the Keyence BZ-X810 microscope quantified in the BZ-X800 Analyzer Software. Invasion
assays were performed in the same way, except using inserts that were purchased pre-
coated in Matrigel (Corning).

In vivo experiments
Orthotopic tumors were established by mammary fat pad injections into 6-8 week
old female NSG mice. Mice were given analgesic (Ketoprofen 5 mg/kg) and general
anesthesia (2% isoflurane) and placed in supine position on a heating pad with limbs
immobilized. The fur around the fourth mammary gland on the mouse’s right abdomen
was shaved and disinfected with three alternating scrubs of chlorohexidine/iodine and
sterile alcohol prep pads. A surgical incision was made medial to the fourth nipple in the
abdominal skin and 100 uL of tumor cells suspended in a 1:1 mixture of PBS:Matrigel

93
was slowly injected into the fat pad. The wound was closed and animals were monitored
for at least 3-days post-surgery to ensure recovery.
Lateral tail vein injections were performed on 6-8 week old female NSG mice. Mice
were placed in heated chambers before restraint in order to dilate the tail veins for
injection. After visualization of the lateral tail veins, tails were sterilized with an alcohol
swab and 100 uL suspension of cells in PBS were injected using a 26 and 5/8th gauge
needle. After injection, pressure was firmly applied to the injection site for at least 1 minute
to stop bleeding.
For all mouse experiments were luminal, ER+ cell lines were used, an estrogen
pellet was subcutaneously implanted. For doxycycline induction of hairpins in mice,
doxycycline was administered in the drinking water at 1 mg/mL with 1% sucrose in sterile
water and changed every 2-3 days. In vivo bioluminescent imaging was performed by
intraperitoneal injection of 100 uL of D-luciferin substrate and subsequent imaging on the
IVIS Lumina III instrument (PerkinElmer). Luciferase assay for measuring
transendothelial cell migration was performed by harvesting and snap-freezing lungs in
liquid nitrogen. Frozen lungs were pulverized into a powder using a pre-chilled mortar and
pestle and weighed. Using the luciferase assay system (Promega), cells were
mechanically and chemically lysed and luciferase activity was quantified per mg of tissue
in a Lucetta Luminometer (Lonza) using a 2 second delay and 10 second integration of
luciferase signal.

94
Immunofluorescence on tissue sections
At experimental endpoint, mice were euthanized by CO2 asphyxiation and cervical
dislocation before harvesting lungs. Lungs were briefly rinsed in ice cold PBS and placed
in cold 4% paraformaldehyde for 4.5 hours with rocking, followed by 3 x 10 minute washes
with cold PBS and cryoprotected in 30% sucrose solution in PBS at 4
o
C overnight. The
next day, tissues were embedded in OCT and stored at -80
o
C prior to sectioning.
Cryosections were incubated at 4
o
C overnight with antibodies against chicken anti-
GFP (1:2000 dilution, Abcam ab13970) and rabbit anti-cleaved caspase 3 (1:400 dilution,
Cell Signaling Technologies 9661). The next day, secondary antibodies goat anti-chicken
IgY Alexa Fluor 488 (1:500 dilution, Life Technologies A11039) and goat anti-rabbit IgG
Alexa Fluor 647 (1:500 dilution, Life Technologies A32733) were incubated for one hour
at room temperature. Images were taken on a Keyence BZ-X810 microscope and signal
was quantified in the BZ-X800 Analyzer Software.

95
CHAPTER 5
DISCUSSION

5.1 Introduction
By most recent estimations, 1 in 2 men and 1 in 3 women in the United States will
develop cancer at some point in their lives [213]. For women with breast cancer, earlier
detection [214], the development of some targeted therapies, such as CDK4/6 inhibitors
[215], and the recent surge in immunotherapeutic treatments [216] have contributed to a
1% annual decrease in overall death rate from 2013-2018 despite the average rise of
incidence rates by 0.3% per year measured from 2009-2018 [2]. Still, breast cancer is the
second leading cause of cancer-related deaths in women in the United States [1] and is
in large part due to metastatic progression.
CTCs can be considered metastatic precursors and represent a unique biological
resource to investigate the mechanisms that contribute to cancer metastasis. However,
the problem remains that CTCs are a rare biological resource and it has proven to be
difficult to purify viable CTCs to establish long term proliferating cell lines. While these
efforts are ongoing, there have been a handful of studies developing biophysical profiles
of CTC populations detected in patient peripheral blood samples [217,125,51,62]. These
studies suggest mechanical phenotyping of CTCs as a method to predict cancer
progression.
In the research presented in this dissertation, I discovered the ectopic expression
of a truncated isoform of type II hair keratin 81 (tKRT81) and showed that it increased
deformability, enhanced adhesion, and promoted in vivo lung tumor burden. Despite the

96
inability to pinpoint the exact signaling mechanisms behind this phenotype, it is clear that
it is not a result of the impact of tKRT81 on migration, invasion, apoptosis, or any
identifiable signaling pathways examined in this study. In light of the negative data
produced in this study, perhaps tKRT81 expression and the fine-tuned modulation of the
cell cytoskeleton and the resulting changes in biophysical properties represents a wholly
mechanical mechanism. By the physical selection of softer, more deformable cells that
can resist the various physical stressors of the circulatory environment that encourage
cell rupture and destruction, it is conceivable that tumor cells that are already molecularly
primed for uninhibited proliferation and invasion in the primary tumor are physically
selected for completion of the metastatic cascade. The metastatic process cannot be
purely physically or purely molecularly driven series of events. It is much more likely that
an optimal combination of both factors contributes to metastatic propensity. Therefore,
the combination of both physical and molecular markers of metastasis may provide a
more accurate and complete measurement for metastasis. However, the metastasis
promoting physical aspect is not well studied and our finding shed new insights into its
importance.
These results also provide a compelling argument for mapping the keratin
landscape in various cancers. Keratins comprise the largest family of intermediate
filament proteins by far, yet are relatively understudied and not as well characterized as
other cytoskeletal components beyond their frequent use in tumor diagnostics. A more
detailed understanding of the regulation, structure, and function as a result of the
combinatorial expression of various keratins, particularly potential other truncated

97
variants like tKRT81, in different types of cancers may be useful for the quantification of
metastatic propensity.

5.2 Future directions
CTCs can be used as a way to monitor treatment response (for example, by
enumeration) but can also be used as a way to predict patient response to therapy or
sites of metastatic relapse [39] although the amount of time required to investigate these
parameters in the lab is not yet amenable to informing clinical decisions. The field of CTC
biology is newly emerging, however, and technologies continue to improve over time.
Furthermore, clinical trials regarding the clinical utility of CTCs in guiding treatment
decisions are still ongoing [218].

Caveats of CTC research
Due to a dearth of available long term cultures of CTC lines, some research groups
have used adherent cell lines grown in suspension in order to model CTCs. However, it
is important to note the distinction between these different biological materials. There are
efforts underway to develop long term cultures of CTCs, but whether the variable that
controls short term versus long term culture survival is inherently biological in nature or a
technical parameter that must be optimized is yet to be understood. Indeed, in search of
such efforts, different groups have done drug-screens to identify drugs that can perhaps
aid in long-term CTC culture [129].
Additionally, the CTCs detected in a single blood collection are only representative
of the time of collection and do not provide a dynamic picture. It is unknown if CTCs are

98
continuously shed or shed in bursts correlated with tumor proliferation. It is also unknown
for how long a CTC circulates in the bloodstream and how much cumulative shear is
experienced by successfully metastasizing CTC in the human body. Furthermore, CTCs
are postulated to arise from both the primary tumor and metastatic lesions, and
metastasis can be a very early event. Disseminated tumor cells found in breast cancer
patient bone marrow showed significantly less chromosomal aberrations than the primary
tumor when isolated at the same time, showing that there is potential for metastatic
seeding even before the primary tumor has developed overt metastases [32].
Development of CTC assays that can differentiate between different anatomical origin
sites and infer tumor evolutionary timescales can potentially be very useful for metastasis
management.

Targeting the physical components of metastasizing cells
Thus far, the large majority of cancer research is focused on the genetic and
molecular features of cancer cells. However, there is an entire field of research studying
the biophysical aspects of tumors and CTCs that is yet to be utilized for developing
treatment options. This is a difficult task because, whereas molecular features can be
chosen to distinguish cancerous cells from normal ones, substantial overlap can exist
between physical parameters. Classes of cytoskeletal drugs that target the cytoskeleton
of proliferating cells, such as microtubule and actin inhibitors, preferentially target highly
proliferating cells, which can be cancer cells or the normal epithelium lining the respiratory
or digestive tract, but can miss dormant cells. One could imagine in the distant future that
once there is a better understanding of CTC dynamics in the bloodstream, there could be

99
treatment options for selectively destroying CTCs based on physical and/or molecular
properties and preventing them from seeding metastatic lesions. In this vein, one group
developed a novel shearing device that capitalizes on the lower shear stress-resistance
of cancer cells compared to other blood cells for the selective destruction of CTCs, but
has yet to be tested on clinical samples [219].
Exercise is also being investigated as a way to apply high shear forces to CTCs in
the circulatory system and control cancer in various clinical trials. Clinical trial
NCT03988595 investigates the combination of exercise with standard therapy in
metastatic breast cancer and how it relates to the development of resistance for hormone
therapy and tumor growth. Whether exercise causes a physical destruction or induces a
molecular response is unclear, but regardless of the mechanism there is still some
evidence for a link between exercise and improved overall survival [220]. Interestingly, a
small pilot clinical trial showed that exercise reduced the number of detectable CTCs
among patients with resected Stage I-III colon cancer [82], but whether this was caused
by a direct or indirect effect is yet unknown.

5.3 Conclusions
CTCs are not a newly discovered phenomenon but have recently gained traction
and sparked new research interest with the development of improved technologies and
the ability to ex vivo culture and expand this rare resource. There is still quite a way to go
to fully understand and exploit the biology of CTCs for clinical guidance and personalized
medicine. The discoveries of tKRT81 and its function from this thesis contributed to new
knowledge of biophysical properties that promote metastatic propensity of CTCs.

100
REFERENCES
1. American Cancer Society. Breast Cancer Facts & Figures 2019-2020. Atlanta:
American Cancer Society; 2019. Atlanta: American Cancer Society, Inc. 2019.
2. Cancer Stat Facts: Female Breast Cancer. [cited 2021 November 13]; Available
from: https://seer.cancer.gov/statfacts/html/breast.html.
3. Couch, F.J., et al., BRCA2 germline mutations in male breast cancer cases and
breast cancer families. Nat Genet, 1996. 13(1): p. 123-5.
4. Apostolou, P. and F. Fostira, Hereditary breast cancer: the era of new
susceptibility genes. Biomed Res Int, 2013. 2013: p. 747318.
5. Mahdavi, M., et al., Hereditary breast cancer; Genetic penetrance and current
status with BRCA. J Cell Physiol, 2019. 234(5): p. 5741-5750.
6. Pandya, S. and R.G. Moore, Breast development and anatomy. Clin Obstet
Gynecol, 2011. 54(1): p. 91-5.
7. Du, T., et al., Invasive lobular and ductal breast carcinoma differ in immune
response, protein translation efficiency and metabolism. Sci Rep, 2018. 8(1): p.
7205.
8. McCart Reed, A.E., et al., Mixed ductal-lobular carcinomas: evidence for
progression from ductal to lobular morphology. J Pathol, 2018. 244(4): p. 460-
468.
9. Taylor-Papadimitriou, J., et al., Keratin expression in human mammary epithelial
cells cultured from normal and malignant tissue: relation to in vivo phenotypes
and influence of medium. J Cell Sci, 1989. 94 ( Pt 3): p. 403-13.
10. Dairkee, S. and H.W. Heid, Cytokeratin profile of immunomagnetically separated
epithelial subsets of the human mammary gland. In Vitro Cell Dev Biol Anim,
1993. 29A(5): p. 427-32.
11. Pandey, P.R., J. Saidou, and K. Watabe, Role of myoepithelial cells in breast
tumor progression. Front Biosci (Landmark Ed), 2010. 15: p. 226-36.
12. Henson, D.E., et al., Relationship among outcome, stage of disease, and
histologic grade for 22,616 cases of breast cancer. The basis for a prognostic
index. Cancer, 1991. 68(10): p. 2142-9.
13. American Cancer Society. Survival Rates for Breast Cancer. Atlanta: American
Cancer Society; 2021. https://www.cancer.org/cancer/breast-
cancer/understanding-a-breast-cancer-diagnosis/breast-cancer-survival-
rates.html.

101
14. Mariotto, A.B., et al., Estimation of the Number of Women Living with Metastatic
Breast Cancer in the United States. Cancer Epidemiol Biomarkers Prev, 2017.
26(6): p. 809-815.
15. Pandit, P., et al., Prevalence of Molecular Subtypes of Breast Cancer: A Single
Institutional Experience of 2062 Patients. Eur J Breast Health, 2020. 16(1): p. 39-
43.
16. Voduc, K.D., et al., Breast cancer subtypes and the risk of local and regional
relapse. J Clin Oncol, 2010. 28(10): p. 1684-91.
17. McCarthy, A.M., et al., Relationship of established risk factors with breast cancer
subtypes. Cancer Med, 2021. 10(18): p. 6456-6467.
18. Liu, Z., X.S. Zhang, and S. Zhang, Breast tumor subgroups reveal diverse clinical
prognostic power. Sci Rep, 2014. 4: p. 4002.
19. Yao, K., et al., Molecular subtyping improves diagnostic stratification of patients
with primary breast cancer into prognostically defined risk groups. Breast Cancer
Res Treat, 2015. 154(1): p. 81-8.
20. Schaafsma, E., et al., Impact of Oncotype DX testing on ER+ breast cancer
treatment and survival in the first decade of use. Breast Cancer Res, 2021. 23(1):
p. 74.
21. Beitsch, P., et al., Pertuzumab/Trastuzumab/CT Versus Trastuzumab/CT
Therapy for HER2+ Breast Cancer: Results from the Prospective Neoadjuvant
Breast Registry Symphony Trial (NBRST). Ann Surg Oncol, 2017. 24(9): p. 2539-
2546.
22. Ciriello, G., et al., Emerging landscape of oncogenic signatures across human
cancers. Nat Genet, 2013. 45(10): p. 1127-33.
23. Lehmann, B.D., et al., Identification of human triple-negative breast cancer
subtypes and preclinical models for selection of targeted therapies. J Clin Invest,
2011. 121(7): p. 2750-67.
24. Montero, A.J., et al., The economic burden of metastatic breast cancer: a U.S.
managed care perspective. Breast Cancer Res Treat, 2012. 134(2): p. 815-22.
25. Bockhorn, M., R.K. Jain, and L.L. Munn, Active versus passive mechanisms in
metastasis: do cancer cells crawl into vessels, or are they pushed? Lancet
Oncol, 2007. 8(5): p. 444-8.
26. Wyckoff, J.B., et al., Direct visualization of macrophage-assisted tumor cell
intravasation in mammary tumors. Cancer Res, 2007. 67(6): p. 2649-56.

102
27. Peña, C., et al., STC1 expression by cancer-associated fibroblasts drives
metastasis of colorectal cancer. Cancer Res, 2013. 73(4): p. 1287-97.
28. Bekes, E.M., et al., Tumor-recruited neutrophils and neutrophil TIMP-free MMP-9
regulate coordinately the levels of tumor angiogenesis and efficiency of
malignant cell intravasation. Am J Pathol, 2011. 179(3): p. 1455-70.
29. Micalizzi, D.S., S. Maheswaran, and D.A. Haber, A conduit to metastasis:
circulating tumor cell biology. Genes Dev, 2017. 31(18): p. 1827-1840.
30. Wirtz, D., K. Konstantopoulos, and P.C. Searson, The physics of cancer: the role
of physical interactions and mechanical forces in metastasis. Nat Rev Cancer,
2011. 11(7): p. 512-22.
31. Endo, H. and M. Inoue, Dormancy in cancer. Cancer Sci, 2019. 110(2): p. 474-
480.
32. Schmidt-Kittler, O., et al., From latent disseminated cells to overt metastasis:
genetic analysis of systemic breast cancer progression. Proc Natl Acad Sci U S
A, 2003. 100(13): p. 7737-42.
33. Bidard, F.C., et al., Clinical validity of circulating tumour cells in patients with
metastatic breast cancer: a pooled analysis of individual patient data. Lancet
Oncol, 2014. 15(4): p. 406-14.
34. de Bono, J.S., et al., Circulating tumor cells predict survival benefit from
treatment in metastatic castration-resistant prostate cancer. Clin Cancer Res,
2008. 14(19): p. 6302-9.
35. Tsai, W.S., et al., Circulating Tumor Cell Count Correlates with Colorectal
Neoplasm Progression and Is a Prognostic Marker for Distant Metastasis in Non-
Metastatic Patients. Sci Rep, 2016. 6: p. 24517.
36. Hiltermann, T.J.N., et al., Circulating tumor cells in small-cell lung cancer: a
predictive and prognostic factor. Ann Oncol, 2012. 23(11): p. 2937-2942.
37. Thiele, J.A., et al., Single-Cell Analysis of Circulating Tumor Cells. Methods Mol
Biol, 2019. 1908: p. 243-264.
38. Yu, M., et al., Cancer therapy. Ex vivo culture of circulating breast tumor cells for
individualized testing of drug susceptibility. Science, 2014. 345(6193): p. 216-20.
39. Klotz, R., et al., Circulating Tumor Cells Exhibit Metastatic Tropism and Reveal
Brain Metastasis Drivers. Cancer Discov, 2020. 10(1): p. 86-103.
40. Gao, D., et al., Organoid cultures derived from patients with advanced prostate
cancer. Cell, 2014. 159(1): p. 176-187.

103
41. Cayrefourcq, L., et al., Establishment and characterization of a cell line from
human circulating colon cancer cells. Cancer Res, 2015. 75(5): p. 892-901.
42. Fidler, I.J., Metastasis: quantitative analysis of distribution and fate of tumor
emboli labeled with 125 I-5-iodo-2'-deoxyuridine. J Natl Cancer Inst, 1970. 45(4):
p. 773-82.
43. Dujon, A.M., et al., Is There One Key Step in the Metastatic Cascade? Cancers
(Basel), 2021. 13(15).
44. Jordan, N.V., et al., HER2 expression identifies dynamic functional states within
circulating breast cancer cells. Nature, 2016. 537(7618): p. 102-106.
45. Yu, M., et al., Circulating breast tumor cells exhibit dynamic changes in epithelial
and mesenchymal composition. Science, 2013. 339(6119): p. 580-4.
46. Hassan, S., et al., Diversity of Epithelial-Mesenchymal Phenotypes in Circulating
Tumour Cells from Prostate Cancer Patient-Derived Xenograft Models. Cancers
(Basel), 2021. 13(11).
47. Aceto, N., Bring along your friends: Homotypic and heterotypic circulating tumor
cell clustering to accelerate metastasis. Biomed J, 2020. 43(1): p. 18-23.
48. Aceto, N., et al., Circulating tumor cell clusters are oligoclonal precursors of
breast cancer metastasis. Cell, 2014. 158(5): p. 1110-22.
49. Velez, J., et al., Platelets promote mitochondrial uncoupling and resistance to
apoptosis in leukemia cells: a novel paradigm for the bone marrow
microenvironment. Cancer Microenviron, 2014. 7(1-2): p. 79-90.
50. Nieswandt, B., et al., Lysis of tumor cells by natural killer cells in mice is impeded
by platelets. Cancer Res, 1999. 59(6): p. 1295-300.
51. Ozkumur, E., et al., Inertial focusing for tumor antigen-dependent and -
independent sorting of rare circulating tumor cells. Sci Transl Med, 2013. 5(179):
p. 179ra47.
52. Ramirez, A.B., et al., Circulating tumor cell investigation in breast cancer patient-
derived xenograft models by automated immunofluorescence staining, image
acquisition, and single cell retrieval and analysis. BMC Cancer, 2019. 19(1): p.
220.
53. Pauken, C.M., et al., Heterogeneity of Circulating Tumor Cell Neoplastic
Subpopulations Outlined by Single-Cell Transcriptomics. Cancers (Basel), 2021.
13(19).
54. Kamal, M., et al., PIC&RUN: An integrated assay for the detection and retrieval
of single viable circulating tumor cells. Sci Rep, 2019. 9(1): p. 17470.

104
55. Drucker, A., et al., Comparative performance of different methods for circulating
tumor cell enrichment in metastatic breast cancer patients. PLoS One, 2020.
15(8): p. e0237308.
56. Kang, H., et al., Evaluation of Positive and Negative Methods for Isolation of
Circulating Tumor Cells by Lateral Magnetophoresis. Micromachines (Basel),
2019. 10(6).
57. Morgan, T.M., P.H. Lange, and R.L. Vessella, Detection and characterization of
circulating and disseminated prostate cancer cells. Front Biosci, 2007. 12: p.
3000-9.
58. Mendelaar, P.A.J., et al., Defining the dimensions of circulating tumor cells in a
large series of breast, prostate, colon, and bladder cancer patients. Mol Oncol,
2021. 15(1): p. 116-125.
59. Prinyakupt, J. and C. Pluempitiwiriyawej, Segmentation of white blood cells and
comparison of cell morphology by linear and naïve Bayes classifiers. Biomed
Eng Online, 2015. 14: p. 63.
60. Hakim, M., et al., A new insight to deformability correlation of circulating tumor
cells with metastatic behavior by application of a new deformability-based
microfluidic chip. Anal Chim Acta, 2021. 1186: p. 339115.
61. Ahmmed, S.M., et al., Multi-sample deformability cytometry of cancer cells. APL
Bioeng, 2018. 2(3): p. 032002.
62. Shaw Bagnall, J., et al., Deformability of Tumor Cells versus Blood Cells. Sci
Rep, 2015. 5: p. 18542.
63. Kitz, J., et al., EMT-independent detection of circulating tumor cells in human
blood samples and pre-clinical mouse models of metastasis. Clin Exp
Metastasis, 2021. 38(1): p. 97-108.
64. Ying, Y. and Y. Lin, Inertial Focusing and Separation of Particles in Similar
Curved Channels. Sci Rep, 2019. 9(1): p. 16575.
65. Sun, J., et al., Double spiral microchannel for label-free tumor cell separation and
enrichment. Lab Chip, 2012. 12(20): p. 3952-60.
66. Sollier, E., et al., Size-selective collection of circulating tumor cells using Vortex
technology. Lab Chip, 2014. 14(1): p. 63-77.
67. Xu, X., et al., Recent progress of inertial microfluidic-based cell separation.
Analyst, 2021. 146(23): p. 7070-7086.
68. Lin, E., et al., High-Throughput Microfluidic Labyrinth for the Label-free Isolation
of Circulating Tumor Cells. Cell Syst, 2017. 5(3): p. 295-304.e4.

105
69. Abdulla, A., et al., High-Throughput Isolation of Circulating Tumor Cells Using
Cascaded Inertial Focusing Microfluidic Channel. Anal Chem, 2018. 90(7): p.
4397-4405.
70. Smith, K.J., et al., Inertial focusing of circulating tumor cells in whole blood at
high flow rates using the microfluidic CTCKey™ device for CTC enrichment. Lab
Chip, 2021. 21(18): p. 3559-3572.
71. Augustsson, P., et al., Microfluidic, label-free enrichment of prostate cancer cells
in blood based on acoustophoresis. Anal Chem, 2012. 84(18): p. 7954-62.
72. Li, P., et al., Acoustic separation of circulating tumor cells. Proc Natl Acad Sci U
S A, 2015. 112(16): p. 4970-5.
73. Wu, M., et al., Circulating Tumor Cell Phenotyping via High-Throughput Acoustic
Separation. Small, 2018. 14(32): p. e1801131.
74. Gupta, V., et al., ApoStream(™), a new dielectrophoretic device for antibody
independent isolation and recovery of viable cancer cells from blood.
Biomicrofluidics, 2012. 6(2): p. 24133.
75. Shim, S., et al., Antibody-independent isolation of circulating tumor cells by
continuous-flow dielectrophoresis. Biomicrofluidics, 2013. 7(1): p. 11807.
76. Di Trapani, M., N. Manaresi, and G. Medoro, DEPArray™ system: An automatic
image-based sorter for isolation of pure circulating tumor cells. Cytometry A,
2018. 93(12): p. 1260-1266.
77. Cho, E.H., et al., Characterization of circulating tumor cell aggregates identified
in patients with epithelial tumors. Phys Biol, 2012. 9(1): p. 016001.
78. Huxley, V.H., Sex and the cardiovascular system: the intriguing tale of how
women and men regulate cardiovascular function differently. Adv Physiol Educ,
2007. 31(1): p. 17-22.
79. Tangelder, G.J., et al., Wall shear rate in arterioles in vivo: least estimates from
platelet velocity profiles. Am J Physiol, 1988. 254(6 Pt 2): p. H1059-64.
80. Mitchell, M.J. and M.R. King, Fluid Shear Stress Sensitizes Cancer Cells to
Receptor-Mediated Apoptosis via Trimeric Death Receptors. New J Phys, 2013.
15: p. 015008.
81. Regmi, S., A. Fu, and K.Q. Luo, High Shear Stresses under Exercise Condition
Destroy Circulating Tumor Cells in a Microfluidic System. Sci Rep, 2017. 7: p.
39975.
82. Brown, J.C., et al., Effects of exercise on circulating tumor cells among patients
with resected stage I-III colon cancer. PLoS One, 2018. 13(10): p. e0204875.

106
83. Follain, G., et al., Hemodynamic Forces Tune the Arrest, Adhesion, and
Extravasation of Circulating Tumor Cells. Dev Cell, 2018. 45(1): p. 33-52.e12.
84. Fan, R., et al., Circulatory shear flow alters the viability and proliferation of
circulating colon cancer cells. Sci Rep, 2016. 6: p. 27073.
85. Xin, Y., et al., Fluid Shear Stress Induces EMT of Circulating Tumor Cells via
JNK Signaling in Favor of Their Survival during Hematogenous Dissemination.
Int J Mol Sci, 2020. 21(21).
86. Moose, D.L., et al., Cancer Cells Resist Mechanical Destruction in Circulation via
RhoA/Actomyosin-Dependent Mechano-Adaptation. Cell Rep, 2020. 30(11): p.
3864-3874.e6.
87. Park, H.A., S.R. Brown, and Y. Kim, Cellular Mechanisms of Circulating Tumor
Cells During Breast Cancer Metastasis. Int J Mol Sci, 2020. 21(14).
88. Rao, C.G., et al., Expression of epithelial cell adhesion molecule in carcinoma
cells present in blood and primary and metastatic tumors. Int J Oncol, 2005.
27(1): p. 49-57.
89. Park, S., et al., Morphological differences between circulating tumor cells from
prostate cancer patients and cultured prostate cancer cells. PLoS One, 2014.
9(1): p. e85264.
90. Lazar, D.C., et al., Cytometric comparisons between circulating tumor cells from
prostate cancer patients and the prostate-tumor-derived LNCaP cell line. Phys
Biol, 2012. 9(1): p. 016002.
91. Park, J.Y., et al., Development of suspension cell culture model to mimic
circulating tumor cells. Oncotarget, 2018. 9(1): p. 622-640.
92. Hao, J., et al., Membrane-bound SCF and VCAM-1 synergistically regulate the
morphology of hematopoietic stem cells. J Cell Biol, 2021. 220(10).
93. Voyvodic, P.L., D. Min, and A.B. Baker, A multichannel dampened flow system
for studies on shear stress-mediated mechanotransduction. Lab Chip, 2012.
12(18): p. 3322-30.
94. Franzoni, M., et al., Design of a cone-and-plate device for controlled realistic
shear stress stimulation on endothelial cell monolayers. Cytotechnology, 2016.
68(5): p. 1885-96.
95. Ye, C., et al., Novel cone-and-plate flow chamber with controlled distribution of
wall fluid shear stress. Comput Biol Med, 2019. 106: p. 140-148.
96. Novak, J., et al., In vivo flow cytometer for real-time detection and quantification
of circulating cells. Opt Lett, 2004. 29(1): p. 77-9.

107
97. Tan, X., et al., In Vivo Flow Cytometry of Extremely Rare Circulating Cells. Sci
Rep, 2019. 9(1): p. 3366.
98. Sasportas, L.S. and S.S. Gambhir, Imaging circulating tumor cells in freely
moving awake small animals using a miniaturized intravital microscope. PLoS
One, 2014. 9(1): p. e86759.
99. Palumbo, J.S., et al., Platelets and fibrin(ogen) increase metastatic potential by
impeding natural killer cell-mediated elimination of tumor cells. Blood, 2005.
105(1): p. 178-85.
100. Burdick, M.M. and K. Konstantopoulos, Platelet-induced enhancement of LS174T
colon carcinoma and THP-1 monocytoid cell adhesion to vascular endothelium
under flow. Am J Physiol Cell Physiol, 2004. 287(2): p. C539-47.
101. Secomb, T.W., Hemodynamics. Compr Physiol, 2016. 6(2): p. 975-1003.
102. Wang, S., et al., Margination and adhesion dynamics of tumor cells in a real
microvascular network. PLoS Comput Biol, 2021. 17(2): p. e1008746.
103. Matrone, M.A., et al., Microtentacles tip the balance of cytoskeletal forces in
circulating tumor cells. Cancer Res, 2010. 70(20): p. 7737-41.
104. Katsantonis, J., et al., Differences in the G/total actin ratio and microfilament
stability between normal and malignant human keratinocytes. Cell Biochem
Funct, 1994. 12(4): p. 267-74.
105. Kamyabi, N. and S.A. Vanapalli, Microfluidic cell fragmentation for mechanical
phenotyping of cancer cells. Biomicrofluidics, 2016. 10(2): p. 021102.
106. Tan, J., et al., Simulation of circulating tumor cell transport and adhesion in cell
suspensions in microfluidic devices. Biomicrofluidics, 2019. 13(6): p. 064105.
107. Rejniak, K.A., Circulating Tumor Cells: When a Solid Tumor Meets a Fluid
Microenvironment. Adv Exp Med Biol, 2016. 936: p. 93-106.
108. Liu, Z., N.J. Sniadecki, and C.S. Chen, Mechanical Forces in Endothelial Cells
during Firm Adhesion and Early Transmigration of Human Monocytes. Cell Mol
Bioeng, 2010. 3(1): p. 50-59.
109. Yeh, Y.T., et al., Three-dimensional forces exerted by leukocytes and vascular
endothelial cells dynamically facilitate diapedesis. Proc Natl Acad Sci U S A,
2018. 115(1): p. 133-138.
110. Li, W., et al., Responses of Cellular Adhesion Strength and Stiffness to Fluid
Shear Stress during Tumor Cell Rolling Motion. ACS Sens, 2019. 4(6): p. 1710-
1715.

108
111. Mao, S., et al., Adhesion analysis of single circulating tumor cells on a base layer
of endothelial cells using open microfluidics. Chem Sci, 2018. 9(39): p. 7694-
7699.
112. King, M.R., et al., A physical sciences network characterization of circulating
tumor cell aggregate transport. Am J Physiol Cell Physiol, 2015. 308(10): p.
C792-802.
113. Bose, S., et al., A semianalytical model to study the effect of cortical tension on
cell rolling. Biophys J, 2010. 99(12): p. 3870-9.
114. Coppola, S., et al., Quantifying cellular forces and biomechanical properties by
correlative micropillar traction force and Brillouin microscopy. Biomed Opt
Express, 2019. 10(5): p. 2202-2212.
115. Mulligan, J.A., et al., Traction Force Microscopy for Noninvasive Imaging of Cell
Forces. Adv Exp Med Biol, 2018. 1092: p. 319-349.
116. Davidson, P.M., et al., High-throughput microfluidic micropipette aspiration
device to probe time-scale dependent nuclear mechanics in intact cells. Lab
Chip, 2019. 19(21): p. 3652-3663.
117. Green, K.J. and C.L. Simpson, Desmosomes: new perspectives on a classic. J
Invest Dermatol, 2007. 127(11): p. 2499-515.
118. Au, S.H., et al., Clusters of circulating tumor cells traverse capillary-sized
vessels. Proc Natl Acad Sci U S A, 2016. 113(18): p. 4947-52.
119. Khavari, A. and A.J. Ehrlicher, Nuclei deformation reveals pressure distributions
in 3D cell clusters. PLoS One, 2019. 14(9): p. e0221753.
120. Yamauchi, K., et al., Real-time in vivo dual-color imaging of intracapillary cancer
cell and nucleus deformation and migration. Cancer Res, 2005. 65(10): p. 4246-
52.
121. Lautscham, L.A., et al., Migration in Confined 3D Environments Is Determined by
a Combination of Adhesiveness, Nuclear Volume, Contractility, and Cell
Stiffness. Biophys J, 2015. 109(5): p. 900-13.
122. Elosegui-Artola, A., et al., Force Triggers YAP Nuclear Entry by Regulating
Transport across Nuclear Pores. Cell, 2017. 171(6): p. 1397-1410.e14.
123. Weiss, L., et al., Lethal deformation of cancer cells in the microcirculation: a
potential rate regulator of hematogenous metastasis. Int J Cancer, 1992. 50(1):
p. 103-7.
124. Liu, Z., et al., Cancer cells display increased migration and deformability in pace
with metastatic progression. FASEB J, 2020.

109
125. Cross, S.E., et al., Nanomechanical analysis of cells from cancer patients. Nat
Nanotechnol, 2007. 2(12): p. 780-3.
126. Han, Y.L., et al., Cell swelling, softening and invasion in a three-dimensional
breast cancer model. Nat Phys, 2020. 16(1): p. 101-108.
127. Kalli, M., et al., Mechanical Compression Regulates Brain Cancer Cell Migration
Through MEK1/Erk1 Pathway Activation and GDF15 Expression. Front Oncol,
2019. 9: p. 992.
128. Lange, J.R., et al., Microconstriction arrays for high-throughput quantitative
measurements of cell mechanical properties. Biophys J, 2015. 109(1): p. 26-34.
129. Teng, T., et al., N-Acetyl-L-cysteine Promotes Ex Vivo Growth and Expansion of
Single Circulating Tumor Cells by Mitigating Cellular Stress Responses. Mol
Cancer Res, 2021. 19(3): p. 441-450.
130. Rottner, K., et al., Actin assembly mechanisms at a glance. J Cell Sci, 2017.
130(20): p. 3427-3435.
131. Neuhaus, J.M., et al., Treadmilling of actin. J Muscle Res Cell Motil, 1983. 4(5):
p. 507-27.
132. Gross, S.R., Actin binding proteins: their ups and downs in metastatic life. Cell
Adh Migr, 2013. 7(2): p. 199-213.
133. Takaki, T., et al., Actomyosin drives cancer cell nuclear dysmorphia and
threatens genome stability. Nat Commun, 2017. 8: p. 16013.
134. Chrzanowska-Wodnicka, M. and K. Burridge, Rho-stimulated contractility drives
the formation of stress fibers and focal adhesions. J Cell Biol, 1996. 133(6): p.
1403-15.
135. Bryan, D.S., et al., 4-Hydroxyacetophenone modulates the actomyosin
cytoskeleton to reduce metastasis. Proc Natl Acad Sci U S A, 2020. 117(36): p.
22423-22429.
136. Downing, K.H. and E. Nogales, Tubulin structure: insights into microtubule
properties and functions. Curr Opin Struct Biol, 1998. 8(6): p. 785-91.
137. Goodson, H.V. and E.M. Jonasson, Microtubules and Microtubule-Associated
Proteins. Cold Spring Harb Perspect Biol, 2018. 10(6).
138. Parker, A.L., M. Kavallaris, and J.A. McCarroll, Microtubules and their role in
cellular stress in cancer. Front Oncol, 2014. 4: p. 153.

110
139. Whipple, R.A., et al., Epithelial-to-mesenchymal transition promotes tubulin
detyrosination and microtentacles that enhance endothelial engagement. Cancer
Res, 2010. 70(20): p. 8127-37.
140. Thiery, J.P., Epithelial-mesenchymal transitions in tumour progression. Nat Rev
Cancer, 2002. 2(6): p. 442-54.
141. Satelli, A. and S. Li, Vimentin in cancer and its potential as a molecular target for
cancer therapy. Cell Mol Life Sci, 2011. 68(18): p. 3033-46.
142. Hieda, M., Signal Transduction across the Nuclear Envelope: Role of the LINC
Complex in Bidirectional Signaling. Cells, 2019. 8(2).
143. Dubik, N. and S. Mai, Lamin A/C: Function in Normal and Tumor Cells. Cancers
(Basel), 2020. 12(12).
144. Windoffer, R., et al., Cytoskeleton in motion: the dynamics of keratin intermediate
filaments in epithelia. J Cell Biol, 2011. 194(5): p. 669-78.
145. Windoffer, R., et al., Focal adhesions are hotspots for keratin filament precursor
formation. J Cell Biol, 2006. 173(3): p. 341-8.
146. Moch, M., et al., The keratin-desmosome scaffold: pivotal role of desmosomes
for keratin network morphogenesis. Cell Mol Life Sci, 2020. 77(3): p. 543-558.
147. Stokes, D.L., Desmosomes from a structural perspective. Curr Opin Cell Biol,
2007. 19(5): p. 565-71.
148. Walko, G., M.J. Castañón, and G. Wiche, Molecular architecture and function of
the hemidesmosome. Cell Tissue Res, 2015. 360(2): p. 363-78.
149. Moch, M. and R.E. Leube, Hemidesmosome-Related Keratin Filament Bundling
and Nucleation. Int J Mol Sci, 2021. 22(4).
150. Jacob, J.T., et al., Keratin 17 regulates nuclear morphology and chromatin
organization. J Cell Sci, 2020. 133(20).
151. Havaki, S., et al., Nuclear localization of cytokeratin 8 and the O-linked N-
acetylglucosamine-containing epitope H in epithelial cells of infiltrating ductal
breast carcinomas: a combination of immunogold and EDTA regressive staining
methods. Ultrastruct Pathol, 2006. 30(3): p. 177-86.
152. Wilhelmsen, K., et al., Nesprin-3, a novel outer nuclear membrane protein,
associates with the cytoskeletal linker protein plectin. J Cell Biol, 2005. 171(5): p.
799-810.
153. Almeida, F.V., et al., The cytolinker plectin regulates nuclear
mechanotransduction in keratinocytes. J Cell Sci, 2015. 128(24): p. 4475-86.

111
154. Kumeta, M., et al., Antibody-based analysis reveals "filamentous vs. non-
filamentous" and "cytoplasmic vs. nuclear" crosstalk of cytoskeletal proteins. Exp
Cell Res, 2013. 319(20): p. 3226-37.
155. Hobbs, R.P., et al., Keratin-dependent regulation of Aire and gene expression in
skin tumor keratinocytes. Nat Genet, 2015. 47(8): p. 933-8.
156. Bawden, C.S., et al., A unique type I keratin intermediate filament gene family is
abundantly expressed in the inner root sheaths of sheep and human hair follicles.
J Invest Dermatol, 2001. 116(1): p. 157-66.
157. Ku, N.O. and M.B. Omary, A disease- and phosphorylation-related
nonmechanical function for keratin 8. J Cell Biol, 2006. 174(1): p. 115-25.
158. Margolis, S.S., et al., Role for the PP2A/B56delta phosphatase in regulating 14-
3-3 release from Cdc25 to control mitosis. Cell, 2006. 127(4): p. 759-73.
159. Kim, S., P. Wong, and P.A. Coulombe, A keratin cytoskeletal protein regulates
protein synthesis and epithelial cell growth. Nature, 2006. 441(7091): p. 362-5.
160. Mariani, R.A., et al., 14-3-3 targets keratin intermediate filaments to mechanically
sensitive cell-cell contacts. Mol Biol Cell, 2020. 31(9): p. 930-943.
161. Busch, T., et al., Keratin 8 phosphorylation regulates keratin reorganization and
migration of epithelial tumor cells. J Cell Sci, 2012. 125(Pt 9): p. 2148-59.
162. Alam, H., et al., Loss of keratin 8 phosphorylation leads to increased tumor
progression and correlates with clinico-pathological parameters of OSCC
patients. PLoS One, 2011. 6(11): p. e27767.
163. Snider, N.T., et al., Keratin hypersumoylation alters filament dynamics and is a
marker for human liver disease and keratin mutation. J Biol Chem, 2011. 286(3):
p. 2273-84.
164. Snider, N.T., et al., Glucose and SIRT2 reciprocally mediate the regulation of
keratin 8 by lysine acetylation. J Cell Biol, 2013. 200(3): p. 241-7.
165. Jang, K.H., et al., Liver disease-associated keratin 8 and 18 mutations modulate
keratin acetylation and methylation. FASEB J, 2019. 33(8): p. 9030-9043.
166. Ku, N.O., et al., Cytoskeletal keratin glycosylation protects epithelial tissue from
injury. Nat Cell Biol, 2010. 12(9): p. 876-85.
167. Candi, E., et al., A highly conserved lysine residue on the head domain of type II
keratins is essential for the attachment of keratin intermediate filaments to the
cornified cell envelope through isopeptide crosslinking by transglutaminases.
Proc Natl Acad Sci U S A, 1998. 95(5): p. 2067-72.

112
168. Kwan, R., et al., Keratin 8 phosphorylation regulates its transamidation and
hepatocyte Mallory-Denk body formation. FASEB J, 2012. 26(6): p. 2318-26.
169. Martin-Castillo, B., et al., Cytokeratin 5/6 fingerprinting in HER2-positive tumors
identifies a poor prognosis and trastuzumab-resistant basal-HER2 subtype of
breast cancer. Oncotarget, 2015. 6(9): p. 7104-22.
170. van de Rijn, M., et al., Expression of cytokeratins 17 and 5 identifies a group of
breast carcinomas with poor clinical outcome. Am J Pathol, 2002. 161(6): p.
1991-6.
171. Ignatiadis, M., et al., Different prognostic value of cytokeratin-19 mRNA positive
circulating tumor cells according to estrogen receptor and HER2 status in early-
stage breast cancer. J Clin Oncol, 2007. 25(33): p. 5194-202.
172. Xenidis, N., et al., Cytokeratin-19 mRNA-positive circulating tumor cells after
adjuvant chemotherapy in patients with early breast cancer. J Clin Oncol, 2009.
27(13): p. 2177-84.
173. Kim, J. and R. Villadsen, The Expression Pattern of Epidermal Differentiation
Marker Keratin 10 in the Normal Human Breast and Breast Cancer Cells. J
Histochem Cytochem, 2020. 68(8): p. 561-570.
174. Woelfle, U., et al., Down-regulated expression of cytokeratin 18 promotes
progression of human breast cancer. Clin Cancer Res, 2004. 10(8): p. 2670-4.
175. Seltmann, K., et al., Keratins significantly contribute to cell stiffness and impact
invasive behavior. Proc Natl Acad Sci U S A, 2013. 110(46): p. 18507-12.
176. Cheung, K.J., et al., Collective invasion in breast cancer requires a conserved
basal epithelial program. Cell, 2013. 155(7): p. 1639-51.
177. Meng, Y., et al., Keratin 18 attenuates estrogen receptor alpha-mediated
signaling by sequestering LRP16 in cytoplasm. BMC Cell Biol, 2009. 10: p. 96.
178. Ohtsuka, T., et al., Interaction of cytokeratin 19 head domain and HER2 in the
cytoplasm leads to activation of HER2-Erk pathway. Sci Rep, 2016. 6: p. 39557.
179. Sharma, P., et al., Keratin 19 regulates cell cycle pathway and sensitivity of
breast cancer cells to CDK inhibitors. Sci Rep, 2019. 9(1): p. 14650.
180. Alsharif, S., et al., Keratin 19 maintains E-cadherin localization at the cell surface
and stabilizes cell-cell adhesion of MCF7 cells. Cell Adh Migr, 2021. 15(1): p. 1-
17.
181. Perone, Y., et al., SREBP1 drives Keratin-80-dependent cytoskeletal changes
and invasive behavior in endocrine-resistant ERα breast cancer. Nat Commun,
2019. 10(1): p. 2115.

113
182. Iyer, S.V., et al., Understanding the role of keratins 8 and 18 in neoplastic
potential of breast cancer derived cell lines. PLoS One, 2013. 8(1): p. e53532.
183. Boulay, A., et al., Transcription regulation and protein subcellular localization of
the truncated basic hair keratin hHb1-DeltaN in human breast cancer cells. J Biol
Chem, 2001. 276(25): p. 22954-64.
184. Xie, Y., et al., A miR-SNP of the KRT81 gene is associated with the prognosis of
non-Hodgkin's lymphoma. Gene, 2014. 539(2): p. 198-202.
185. Wang, C., et al., A polymorphism at the microRNA binding site in the 3'-
untranslated region of C14orf101 is associated with the risk of gastric cancer
development. Exp Ther Med, 2016. 12(3): p. 1867-1872.
186. Zhang, K., et al., Knockdown Inhibits Malignant Progression of Melanoma
Through Regulating Interleukin-8. DNA Cell Biol, 2021. 40(10): p. 1290-1297.
187. Muckenhuber, A., et al., Pancreatic Ductal Adenocarcinoma Subtyping Using the
Biomarkers Hepatocyte Nuclear Factor-1A and Cytokeratin-81 Correlates with
Outcome and Treatment Response. Clin Cancer Res, 2017.
188. Lombardo, Y., et al., Mammosphere formation assay from human breast cancer
tissues and cell lines. J Vis Exp, 2015(97).
189. Seltmann, K., et al., Keratins mediate localization of hemidesmosomes and
repress cell motility. J Invest Dermatol, 2013. 133(1): p. 181-90.
190. Ju, J.H., et al., Cytokeratin19 induced by HER2/ERK binds and stabilizes HER2
on cell membranes. Cell Death Differ, 2015. 22(4): p. 665-76.
191. Campayo, M., et al., A dual role for KRT81: a miR-SNP associated with
recurrence in non-small-cell lung cancer and a novel marker of squamous cell
lung carcinoma. PLoS One, 2011. 6(7): p. e22509.
192. de Larrea, C.F., et al., Impact of MiRSNPs on survival and progression in
patients with multiple myeloma undergoing autologous stem cell transplantation.
Clin Cancer Res, 2012. 18(13): p. 3697-704.
193. Minn, A.J., et al., Genes that mediate breast cancer metastasis to lung. Nature,
2005. 436(7050): p. 518-24.
194. Dydensborg, A.B., et al., GATA3 inhibits breast cancer growth and pulmonary
breast cancer metastasis. Oncogene, 2009. 28(29): p. 2634-42.
195. Doebar, S.C., et al., Gene Expression Differences between Ductal Carcinoma in
Situ with and without Progression to Invasive Breast Cancer. Am J Pathol, 2017.
187(7): p. 1648-1655.

114
196. Nanashima, N., et al., Hair keratin KRT81 is expressed in normal and breast
cancer cells and contributes to their invasiveness. Oncol Rep, 2017. 37(5): p.
2964-2970.
197. Peng, S., et al., Single-Cell RNA Sequencing of a Postmenopausal Normal
Breast Tissue Identifies Multiple Cell Types That Contribute to Breast Cancer.
Cancers (Basel), 2020. 12(12).
198. Liu, Z., et al., Cancer cells display increased migration and deformability in pace
with metastatic progression. FASEB J, 2020. 34(7): p. 9307-9315.
199. Régnier, C.H., et al., Expression of a truncated form of hHb1 hair keratin in
human breast carcinomas. Br J Cancer, 1998. 78(12): p. 1640-4.
200. Dobin, A., et al., STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 2013.
29(1): p. 15-21.
201. Anders, S., P.T. Pyl, and W. Huber, HTSeq--a Python framework to work with
high-throughput sequencing data. Bioinformatics, 2015. 31(2): p. 166-9.
202. Love, M.I., W. Huber, and S. Anders, Moderated estimation of fold change and
dispersion for RNA-seq data with DESeq2. Genome Biol, 2014. 15(12): p. 550.
203. Ramírez, F., et al., deepTools2: a next generation web server for deep-
sequencing data analysis. Nucleic Acids Res, 2016. 44(W1): p. W160-5.
204. Kent, W.J., et al., The human genome browser at UCSC. Genome Res, 2002.
12(6): p. 996-1006.
205. Wang, L., S. Wang, and W. Li, RSeQC: quality control of RNA-seq experiments.
Bioinformatics, 2012. 28(16): p. 2184-5.
206. Colaprico, A., et al., TCGAbiolinks: an R/Bioconductor package for integrative
analysis of TCGA data. Nucleic Acids Res, 2016. 44(8): p. e71.
207. Therneau, T.M. and P.M. Grambsch, Modeling Survival Data: Extending the Cox
Model. 2000, Springer, New York, NY.
208. Rust, M.J., M. Bates, and X. Zhuang, Sub-diffraction-limit imaging by stochastic
optical reconstruction microscopy (STORM). Nature Methods, 2006. 3(10): p.
793-795.
209. Huang, B., et al., Three-dimensional super-resolution imaging by stochastic
optical reconstruction microscopy. Science, 2008. 319(5864): p. 810-813.
210. Wojcik, M., et al., Graphene-enabled electron microscopy and correlated super-
resolution microscopy of wet cells. Nature Communications, 2015. 6: p. 7384.

115
211. Xia, Y. and G.M. Whitesides, Soft Lithography. Angew Chem Int Ed Engl, 1998.
37(5): p. 550-575.
212. Theret, D.P., et al., The application of a homogeneous half-space model in the
analysis of endothelial cell micropipette measurements. J Biomech Eng, 1988.
110(3): p. 190-9.
213. Lifetime Risk of Developing or Dying From Cancer. 2020 [cited 2021 24 Nov];
Available from: https://www.cancer.org/cancer/cancer-basics/lifetime-probability-
of-developing-or-dying-from-cancer.html.
214. Niraula, S., et al., Incidence, Characteristics, and Outcomes of Interval Breast
Cancers Compared With Screening-Detected Breast Cancers. JAMA Netw
Open, 2020. 3(9): p. e2018179.
215. Johnston, S.R.D., et al., Abemaciclib Combined With Endocrine Therapy for the
Adjuvant Treatment of HR+, HER2-, Node-Positive, High-Risk, Early Breast
Cancer (monarchE). J Clin Oncol, 2020. 38(34): p. 3987-3998.
216. Nanda, R., et al., Effect of Pembrolizumab Plus Neoadjuvant Chemotherapy on
Pathologic Complete Response in Women With Early-Stage Breast Cancer: An
Analysis of the Ongoing Phase 2 Adaptively Randomized I-SPY2 Trial. JAMA
Oncol, 2020. 6(5): p. 676-684.
217. Osmulski, P., et al., Nanomechanical biomarkers of single circulating tumor cells
for detection of castration resistant prostate cancer. Prostate, 2014. 74(13): p.
1297-307.
218. Schochter, F., et al., Are Circulating Tumor Cells (CTCs) Ready for Clinical Use
in Breast Cancer? An Overview of Completed and Ongoing Trials Using CTCs
for Clinical Treatment Decisions. Cells, 2019. 8(11).
219. Weth, A., et al., A novel device for elimination of cancer cells from blood
specimens. Sci Rep, 2020. 10(1): p. 10181.
220. Delrieu, L., et al., Analysis of the StoRM cohort reveals physical activity to be
associated with survival in metastatic breast cancer. Sci Rep, 2020. 10(1): p.
10757.

Abstract (if available)

Abstract Metastasis is the leading cause of cancer deaths and is difficult to control, predict, and treat in the clinic. Circulating tumor cells that are detected in the peripheral blood of cancer patients are the best models for studying the metastatic cascade as they contain subpopulations that represent the closest biological precursors of metastatic lesions. Recent advancements in the isolation and ex vivo expansion of this extremely rare cell population have allowed for a more accurate investigation of the metastatic process. In this dissertation project, we identified the upregulation of a truncated isoform of KRT81 in in vivo models of breast cancer metastasis using circulating tumor cells that were isolated from the peripheral blood of breast cancer patients. The KRT81 gene encodes for a type II keratin protein that is typically expressed in hair and contributes to structural rigidity and mechanical resilience. In breast cancer cells, the truncated KRT81 (tKRT81) protein was found to alter the cell cytoskeleton and biophysical properties of the cell, leading to functional changes that enhance metastatic propensity. This work highlights the biomechanical mechanism underlying the functional contribution of keratins in cancer beyond their traditional use as diagnostic biomarkers.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Developing a robust single cell whole genome bisulfite sequencing protocol to analyse circulating tumor cells

PDF

Corisol's role in breast-to-brain metastasis

PDF

Heterogeneity and plasticity of malignant and non-malignant circulating analytes in breast carcinomas

PDF

Exploring the effects of CXCR4 inhibition on circulating tumor cell populations in metastatic prostate cancer

PDF

An image-based identification of aggressive circulating tumor cell subtypes

PDF

Development of a colorectal cancer-on-chip to investigate the tumor microenvironment's role in cancer progression

PDF

Decreased levels of expression of transmembrane protein 56 (TMEM56) in breast cancer tissues

PDF

Neuronal master regulator SRRM4 in breast cancer cells facilitates CNS-acclimation and colonization leading to brain metastasis

PDF

Molecular signature of aggressive disease and clonal diversity revealed by single-cell copy number analysis of prostate cancer cells across multiple disease states

PDF

Identification of CBP/FOXM1 as a molecular target in triple negative breast cancer

PDF

Molecular mechanisms of chemoresistance in breast cancer

PDF

The role of endoplasmic reticulum chaperone protein GRP78 in breast cancer

PDF

Context-dependent role of androgen receptor (AR) in estrogen receptor-positive (ER+) breast cancer

PDF

RNA methylation in cancer plasticity and drug resistance

PDF

Mechanism of a new CK2 inhibitor triggering senescence in breast cancer cells

PDF

Characterization of a new HER2 positive mouse model of metastatic breast cancer

PDF

Breast epithelial cell type specific enhancers and functional annotation of breast cancer risk loci

PDF

Developing a bioluminescence tracking system targeting on tumor cells and T cells for cancer immunotherapy

PDF

The opening of the blood brain barrier by homogenized perillyl alcohol

PDF

Quantitative evaluation of the effectiveness of an integrin antagonist, vicrostatin, in cell migration

Asset Metadata

Creator Kang, Diane Seung (author)

Core Title Ectopic expression of a truncated isoform of hair keratin 81 in breast cancer alters biophysical characteristics to promote metastatic propensity

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Medical Biology

Degree Conferral Date 2022-05

Publication Date 01/14/2022

Defense Date 12/09/2021

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag breast cancer,cancer,circulating tumor cells,CTCs,cytoskeleton,intermediate filament,keratin,KRT81,Metastasis,OAI-PMH Harvest,tKRT81

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Offringa, Ite (committee chair), Neman-Ebrahim, Josh (committee member), Yu, Min (committee member)

Creator Email 89kangd@gmail.com,dianesle@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC110520234

Unique identifier UC110520234

Legacy Identifier etd-KangDianeS-10347

Document Type Dissertation

Format application/pdf (imt)

Rights Kang, Diane Seung

Type texts

Source 20220124-usctheses-batch-908 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu